Skip to main content

Full text of "Transactions of the Cambridge Philosophical Society"

See other formats



















^^f "'■ ■■ ll 

! '^H| 




^^%&^ kjf 

^ : ■^•v> 

V '^Ti^^ IT. 

"^^^^^^fs^W;..- ■ / 

^^K^;i:^^ '^■. 




























-r\rf ^^'^^^>f%^r^^^'^^ r' 

i . t 3. JT 




ESTABLISHED November 15, 1819. 








Part I. 


N°. I. Mathematical Investigations concerning the Laws of the Equilibrium of Fluids 
analogous to the Electric Fluid, with other similar Researches: by George 
Green, Esq. Communicated by Sir Edward Ffrench Bromhead, Bart, M.A. 
F.R.S.L. & E I 

II. On Elimination between an Indefinite Number of Unknown quantities : by the 

Rev. R. Murphy 65 

III. On the General Equation of Surfaces of the Second Degree : by Augustus 

De Morgan, Esq 77 

IV. On a Monstrosity of the Common Mignionette : by the Rev. Professor Henslow. . . 95 

Part II. 

V. On the Calculation of Newton's Experiments on Diffraction: by Professor Airv... 101 

VI. Second Memoir on the Inverse Method of Definite Integrals: by the Rev. 

R. MUBPHY 113 

VII. On the Nature of the Truth of the Laws of Motion : by the Rev. W. Whewkll 149 

VIII. Researches in the Theory of the Motion of Fluids : by the Rev. James Challis US 

IX. Theory of Residuo- Capillary Attraction; being an Explanation of the Pheno- 

mena of Endosmose and Exosmose on Mechanical Principles: by the Rev. 

J. PovfEH 205 

X. On Aerial Vibrations in Cylindrical Tubes: by William Hopkins, M.A 231 

XI. On the Latitude of Cambridge Observatory: by Professor Airy 271 


Paet III. 


N° XII. On the Diffraction of an Object-glass with Circular Aperture: bt Professor 

Airy 283 

XIII. On the Equilibrium of the Arch: by the Rev. Hbnry Moseley 293 

XIV. Third Memoir on the Inverse Method of Definite Integrals: by the Rev. 

R. MuHPHY 315 

XV. On the Determination of the Exterior and Interior Attractions of Ellipsoids 

of Variable Densities : by George Green, Esq 395 

XVI. On the Position of the Axes of Optical Elasticity in Crystals belonging to 

the Oblique- Prismatic System: by W. H. Miller, Esq 431 


The Society as a hody is not to he considered responsible for any 
Jucts and opinions advanced in the several Papers, which must rest 
entirely on the credit of their respective Authors. 

The Society takes this opportunity of expressing its grateful 
acknowledgments to the Syndics of the University Press, for their 
liberality in taking upon themselves the expense of printing this 
Part of its Transactions. 




Vol. V. Part I. 








^AL rtV 

I. Mathematical Investigations concerning the Laws of the Equilibrium 
of Fluids analagous to the Electric Fluid, with other similar Researches. 
By George Green, Esq. Communicated hy Sir Edward Ffrench 
Bromhead, Bart. M.A. F.K.S.L. and E. 

[Read Nov. 12, 1832.] 

Amongst the various subjects which have at different times occupied 
the attention of Mathematicians, there are probably few more interesting 
in themselves, or which offer greater difficulties in their investigation, 
than those in which it is required to determine mathematically the 
laws of the equilibrium or motion of a system composed of an infinite 
number of free particles all acting upon each other mutually, and ac- 
cording to some given law. When we conceive, moreover, the law of 
the mutual action of the particles to be such that the forces which 
emanate from them may become insensible at sensible distances, the 
researches to which the consideration of these forces lead will be greatly 
simplified by the limitation thus introduced, and may be regarded as 
forming a class distinct from the rest. Indeed they then for the most 
part terminate in the resolution of equations between the values of 
certain functions at any point taken at will in the interior of the sys- 
tem, and the values of the partial differentials of these functions at the 
same point. When on the contrary the forces in question continue 
sensible at every finite distance, the researches dependent upon them 
become far more complicated, and often require all the resources of 
the modern analysis for their successful prosecution. It would be easy 
so to exhibit the theories of the equilibrium and motion of ordinary 
fluids, as to offer instances of researches appertaining to the former 
class, whilst the mathematical investigations to which the theories of 
Electricity and Magnetism have given rise may be considered as in- 
teresting examples of such as belong to the latter class. 

Vol. V. Pakt I. A 


It is not my chief design in this paper to determine mathematically 
the density of the electric fluid in bodies under given circumstances, 
having elsewhere* given some general methods by which this may be 
effected, and applied these methods to a variety of cases not before 
submitted to calculation. My present object will be to determine the 
laws of the equilibrium of an hypothetical fluid analagous to the electric 
fluid, but of which the law of the repulsion of the particles, instead of 
being inversely as the square of the distance, shall be inversely as any 
power n of the distance ; and I shall have more particularly in view 
the determination of the density of this fluid in the interior of con- 
ducting spheres when in equilibrium, and acted upon by any exterior 
bodies whatever, though since the general method by which this is 
effected will be equally applicable to circular plates and ellipsoids. 

1 shall present a sketch of these applications also. 

It is well known that in enquiries of a nature similar to the one 
about to engage our attention, it is always advantageous to avoid the 
direct consideration of the various forces acting upon any particle p of 
the fluid in the system, by introducing a particular function V of the 
co-ordinates of this particle, from the differentials of which the values 
of all these forces may be immediately deduced f. We have, therefore, 
in the present paper endeavoured, in the first place, to find the value 
of V, where the density of the fluid in the interior of a sphere is given 
by means of a very simple consideration, which in a great measure 
obviates the difficulties usually attendant on researches of this kind, 
have been able to determine the value F^, where p, the density of the 
fluid in any element dv of the sphere's volume, is equal to the product 
of two factors, one of which is a very simple function containing an 
arbitrary exponent fi, and the remaining one J" is equal to any rational 

* Essay on the Application of Mathematical Analysis to the Theories of Electricity and 

t This function in the present case will be obtained by taking the sum of all the molecules 
of a fluid acting upon p, divided by the (n — 1)* power of their respective distances from^; 
and indeed the function which Laplace has represented by F in the third book of the 
Mecanique Celeste, is only a particular value of our more general one produced by writing 

2 in the place of the general exponent n. 


and entire function whatever of the rectangular co-ordinates of the element 
dv, and afterwards by a proper determination of the exponent /3, have 
reduced the resulting quantity ^ to a rational and entire function of 
the rectangular co-ordinates of the particle p, of the same degree as 
the function f. This being done, it is easy to perceive that the reso- 
lution of the inverse problem may readily be effected, because the 
coefficients of the required factor f will then be determined from the 
given coefficients of the rational and entire function V, by means of 
linear algebraic equations. 

The method alluded to in what precedes, and which is exposed in 
the two first articles of the following paper, will enable us to assign 
generally the value of the induced density p for any ellipsoid, what- 
ever its axes may be, provided the inducing forces are given explicitly 
in functions of the co-ordinates of p ; but when by supposing these axes 
equal we reduce the ellipsoid to a sphere, it is natural to expect that 
as the form of the solid has become more simple, a corresponding degree 
of simplicity will be introduced into the results ; and accordingly, as 
will be seen in the fourth and fifth articles, the complete solutions both 
of the direct and inverse problems, considered under their most general 
point of view, are such that the required quantities are there always 
expressed by simple and explicit functions of the known ones, inde- 
pendent of the resolution of any equations whatever. 

The first five articles of the present paper being entirely analytical, 
serve to exhibit the relations which exist between the density p of our 
hypothetical fluid, and its dependent function V; but in the following 
ones our principal object has been to point out some particular appli- 
cations of these general relations. 

In the seventh article, for example, the law of the density of our 
fluid when in equilibrium in the interior of a conductory sphere, has 
been investigated, and the analytical value of p there found admits of 
the following simple enunciation. 

The density p of free fluid at any point p within a conducting sphere 
A, of which O is the centre, is always proportional to the {n - 4)"' power 
of the radius of the circle formed by the intersection of a plane per- 
pendicular to the ray Op with the surface of the sphere itself, provided 

A 2 


n is greater than 2. When on the contrary n is less than 2, this law 
requires a certain modification ; the nature of which has been fully 
investigated in the article just named, and the one immediately fol- 

It has before been remarked, that the generality of our analysis will 
enable us to assign the density of the free fluid which would be induced 
in a sphere by the action of exterior forces, supposing these forces are 
given explicitly in functions of the rectangular co-ordinates of the point 
of space to which they belong. But, as in the particular case in which 
our formulae admit of an application to natural phenomena, the forces in 
question arise from electric fluid diffused in the inducing bodies, we 
have in the ninth article considered more especially the case of a con- 
ducting sphere acted upon by the fluid contained in any exterior bodies 
whatever, and have ultimately been able to exhibit the value of the 
induced density under a very simple form, whatever the given density 
of the fluid in these bodies may be. 

The tenth and last article contains an application of the general 
method to circular planes, from which results, analagous to those formed 
for spheres in some of the preceding ones are deduced; and towards 
the latter part, a very simple formula is given, which serves to express 
the value of the density of the free fluid in an infinitely thin plate, 
supposing it acted upon by other fluid, distributed according to any 
given law in its own plane. Now it is clear, that if to the general ex- 
ponent 11 we assign the particular value 2, all our results will become 
applicable to electrical phenomena. In this way the density of the 
electric fluid on an infinitely thin circular plate, when under the in- 
fluence of any electrified bodies whatever, situated in its own plane, 
will become known. The analytical expression which serves to repre- 
sent the value of this density, is remai-kable for its simplicity ; and by 
suppressing the term due to the exterior bodies, immediately gives the 
density of the electric fluid on a circular conducting plate, when quite 
free from all extraneous action. Fortunately, the manner in which 
the electric fluid distributes itself in the latter case, has long since 
been determined experimentally by Coulomb. We have thus had the 
advantage of comparing our theoretical results with those of a very 


accurate observer, and the differences between them are not greater 
than may be supposed due to the unavoidable errors of experiment, 
and to that which would necessarily be produced by employing plates 
of a finite thickness, whilst the theory supposes this thickness infinitely 
small. ]\Ioreover, the errors are all of the same kind with regard to 
sign, as would arise from the latter cause. 

1. If we conceive a fluid analogous to the electric fluid, but of 
which the law of the repulsion of the particles instead of being in- 
versely as the square of the distance is inversely as some power n of 
the distance, and suppose p to represent the density of this fluid, so 
that dv being an element of the volume of a body A through which 
it is diffiised, pdv may represent the quantity contained in this element, 
and if afterwards we write g for the distance between dv and any 
particle /> under consideration, and these form the quantity 

the integral extending over the whole volume of A, it is well known 
that the force with which a particle p of this fluid situate in any 
point of space is impelled in the direction of any line q and tending 
to increase this line will always be represented by 



1-n \dq) ' 

?^, being regarded as a function of three rectangular co-ordinates of 

p, one of which co-ordinates coincides with the line q, and (—7-) 

being the partial differential of V, relative to this last co-ordinate. 

In order now to make known the principal artifices on which the 
success of our general method for determining the function V mainly 
depends, it will be convenient to begin with a very simple example. 

Let us therefore suppose that the body ^ is a sphere, whose centre, 
is at the origin O of the co-ordinates, the radius being 1 ; and p is 
such a function of x', y', %, that where we substitute for x', y', »' their 
values in polar co-ordinates 


X = r' cos 0', y' = / sin 9' cos tst', %' = r' sin Q' sin tr', 
it shall reduce itself to the form 

P = (l-/y./(0; 

f being the characteristic of any rational and entire function what- 
ever: which is in fact equivalent to supposing 

p = (1 - /' - y" - %'f.f{x" + y" + z'^). 

Now, when as in the present case, p can be expanded in a series 
of the entire powers of the quantities x, y', %', and of the various 
products of these powers, the function V will always admit of a similar 
expansion in the entire powers and products of the quantities x, y, %, 
provided the point p continues within the body A*, and as moreover 
V evidently depends on the distance Op — r and is independent of 6 
and -sr, the two other polar co-ordinates of p, it is easy to see that the 
quantity V when we substitute for x, y, z these values 

x = r cos 9, y = r sin 9 cos w, z = r sin 9 sin tst 

will become a function of r, only containing none but the even 
powers of this variable. 

But since we have 

dv = r"dr d9' d-ur sin 0', and /> = (1 - ry.f{r'% 

the value of V becomes 

V= f-^, = jr'^dr'd9'd-w' sin 9' (1 - r''ff{r") .g'"", 
J g" 

the integrals being taken from tst' = to tr' = 2 tt, from 9' = to 9' = w, 
and from r' = to r' = l. 

* The truth of this assertion will become tolerably clear, if we recollect that V may be 
regarded as the sum of every element pdv of the body's mass divided by the (n—l)"" power 
of the distance of each element from the point p, supposing the density of the body A to be 
expressed by p, a continuous function of x, y, z. For then the quantity V is represented 
by a continuous function, so long as p remains within A ; but there is in general a violation 
of the law of continuity whenever the point p passes from the interior to the exterior space. 
This truth, however, as enunciated in the text, is demonstrable, but since the present paper 
is a long one, I have suppressed the demonstrations to save room. 


Now V may be considered as composed of two parts, one V due 
to the sphere B whose centre is at the origin O, and surface passes 
through the point p, and another V" due to the shell S exterior to B. 
In order to obtain the first part, we must expand the quantity g^~" 



in an ascending series of the powers of — . In this way we get 

^1 -« _ ^1 ^2rr {cos 9 cosff -\- sin 9 sin 9' cos (^' - -sr)] + r'^] 


= r' " " , 

If then we substitute this series for g^'" in the value of F", and 
after having expanded the quantity (1 — r'^f , we effect the integrations 
relative to r, 0', and w', we shall have a result of the form 

r' = r*-'' [A-i-Br+Cf^ + Sic.] 

seeing that in obtaining the part of V before represented by V, the 
integral relative to r' ought to be taken from r =0 to r' = r only. 

To obtain the value of F", we must expand the quantity g^-" in 
an ascending series of the powers of — , and we shall thus have 


g^-''={r^ — 2rr' [cos 6 cos 0' + sin 9 sin 6' cos (tst - -nr')] + r"') 

the coefficients Qo, Qi, Q2, &c. being the same as before. 

The expansion here given being substituted in P", there will arise 
a series of the form 

of which the general term T, is 

T,= fd9'd^' sin ff QJr-'dr ^^^(l-ry.f{ry, 

the integrals being taken from r' = r to r' = l, from 0' = O to & — it, and 
from •z«r' = to 'ar' = 27r. This will be evident by recollecting that the 


triple integral by which the value of V" is expressed, is the same as 
the one before given for V, except that the integration relative to r, 
instead of extending from /=0 to r'=l, ought only to extend from 
r =r to r= 1. 

But the general term in the function J'{r'^) being represented by 
Atr'^\ the part of T^ dependent on this term will evidently be 

(2) Atr'fde'dw' sin 9'.QJr'''+^-^-''dr' {l-r'y-, 

the limits of the integrals being the same as before. 

We thus see that the value of T, and consequently of F'" would 
immediately be obtained, provided we had the value of the general 



which being expanded and integrated becomes 

^ 1 +Mzl). 1 _&c. 

b + l l'b + 3 1.2 'b + 5 

^+1 Q fj,+3 /3(/3-l) r'+» 

+ T • i — ;r — ~ 7 r. • ~i = + <^c- 

6 + 1 16 + 3 1.2 '6 + 5 

but since the first line of this expression is the well known expansion of 


r lf\ r li 




when n = 2.p — h + \ and 5' = 2(/3 + l) we have ultimately, 

By means of the result here obtained, we shall readily find the 
value of the expression (2) which will evidently contain one term multi- 
plied by r' and an infinite number of others, in all of which the quantity 
r is affected with the exponent n. But as in the case under considera- 
tion, n may represent any number whatever, fractionary or irrational, 


it is clear that none of the terms last mentioned can enter into V, 
seeing that it ought to contain the even powers of r only, thence the 
terms of this kind entering into V", must necessarily be destroyed by 
corresponding ones in V. By rejecting them, therefore, the formula (2) 
will become 

m —^ ^-^ Air'.fde'd-Br' smd'Qs. 

But as V ought to contain the even powers of r only, those terms 
in which the exponent s is an odd number, will vanish of themselves 
after all the integrations have been effected, and consequently the only 
terms which can appear in V, are of the form 

r(#+2-y-|) r(/3 + i) 

(4) ^ Atr'^fde'dTir' sin ff Q,r, 

2r(^ + /3 + 3-*'-|) 

where, since s is an even number, we have written 2 s' in the place of 
s, and as Qu- is always a rational and entire function of cos 9', sin 0' 
cos w', and sin 9' sin -sr', the remaining integrations may immediately be 

Having thus the part of T'a,- due to any term Atr'''* of the function 
y(r'*) we have immediately the value of T'.^' and consequently of F'", 

r"= u'+ t:+ t:+ t:+ T:+kc. -, 

U' representing the sum of all the terms in F" which have been rejected 
on account of their form, and T,' T,' T,' the value of T, Ty T„ &c. 
obtained by employing the truncated formula (2) in the place of the 
complete one (2). 

But -v=v'+ V" = r'+ u^Ti^ r;+ t: + 7v+ &c. 

or by transposition, 


and as in this equation, the function on the left side contains none 
but the even powers of the indeterminate quantity r, whilst that on 
Vol. V. Part I. B 


the right does not contain any of the even powers of r, it is clear that 
each of its sides ought to be equated separately to zero. In thi& way 
the left side gives 

(5) r=T:+T,'+T:+T:+kc. 

Hitherto the value of the exponent /3 has remained quite arbitrary, 
but the known properties of the function r will enable us so to 
determine /3, that the series just given shall contain a finite number 
of terms only. We shall thus greatly simplify the value of F) and 
reduce it in fact to a rational and entire function of r*. 

For this purpose, we may remark that 

r(0)=«, r(-l)=oo, r { — 2) = CO, in infinitum. 

If therefore we make — - + /3 = any whole number positive or 

negative, the denominator of the function (4) will become infinite, and 
consequently the function itself will vanish when s is so great that 

1- /3 + i + 3 - *' is equal to zero or any negative number, and as 


the value of t never exceeds a certain number, seeing that f{i^^) is 
a rational and entire function, it is clear that the series (4) will termi- 
fMrte of itself, and V become a rational and entire function of r*. 

(2) The method that has been employed in the preceding article 
where the function by which the density is expressed is of the particular 

may by means of a very slight modification, be applied to the far more 
general value 

P = (1 - ryf{^, i, a') = (1 - x" - y" - --'ff{x, y', z) 

tvhere f is the characteristic of any rational and entire function what- 
ever : and the same value of /3 which reduces V to a, rational and entire 
function of r^ in the first case, reduces it in the second to a similar 
function of x, y, % and the rectangular co-ordinates of p. 


To prove this, we may remark that the con-esponding value V will 


F = fr"dr'de'd^' sin 6' (1 - ryf{x', y', «')^'-"; 

tJie integral being conceived to comprehend the whole volume of the 

Let now the function y be divided into two parts, so that 

fi^, y, %') =/ ix', y', z') +f, ix', y', ^') ; 

/i containing all the terms of the function J] in which the sum of the 
exponents of af, y, %' is an odd number ; and ^ the remaining terms, or 
those where the same sum is an even number. In this way we get 

the functions F'l and V^a corresponding to^ andj^, being 

V, = fr"dr'de'dvr' sin'0' {l~ryf, {x', y\z')g'-', 

V^ = lr"dr'd&d-^ sin & (1 - ryf, {x', y% a') g^-\ 

"We will in the first place endeavour to determine the value J^j; and 
for this purpose, by writing for x, y, %' their values before given in 
r', ff, w', we get 


the coefficients of the various powers of r'^ in ^{r'^) being evidently 

rational and entire functions of cos 0„ sin & cos w', and sin sin w. 


V, = jr^dr'dffdTs' sin 6' (1 - ry />/.(/') ^'-"; 

this integral, like the foregoing, comprehending the whole volume of 
the sphere. 

Now as the density corresponding to the function Fi is - 


it is clear that it may be expanded in an ascending series of the entire 
powers of x', y, »', and the various products of these powers consequently, 
as was before remarked (Art. 1.), Fl admits of an analagous expansion 
in entire powers and products of x, y, ■%. Moreover, as the density /i, 

B 2 


retains the same numerical value, and merely changes its sign when 
we pass from the element dv to a point diametrically opposite, where 
the co-ordinates x, y , % are replaced by - x , - y' , —% \ it is easy to 
see that the function V-^, depending vxpon /s,, possesses a similar property, 
and merely changes its sign when x, y, %, the co-ordinates of p, are 
changed into - x, —y, — as. Hence the nature of the function Vi is 
such that it can contain none but the odd powers of r, when we sub- 
stitute for the rectangular co-ordinates x, y, %, their values in the polar 
co-ordinates r, 6, ■zs. 

Having premised these remarks, let us now suppose Vx is divided 
into two parts, one V^ due to the sphere B which passes through the 
particle p, and the other V" due to the exterior shell aS*. Then it is 
evident by proceeding, as in the case where p = (1 - r"^Yf{i% that Vi 
will be of the form 

the coefficients A, B, C, &;c. being quantities independent of the variable r. 

In like manner we have also 

F/' = fr'^dr'ae'dsr' sin ff {\-ry .r'>\,{r'')g^-''; 

the integrals being taken from r' = r to r = l, from 6' = to 0' = 7r, and 
from Gr' = to 'z<r' = 2 7r. 

By substituting now the second expansion of g^" before used (Art. 1.), 
the last expression will become 

r," = t; + Ti + r. + ^3 + &c. 

of which series the general term is 

T, = fd9'dw' sin ff Q, fr"-dr' (1 - ry ^ x/. {r"). 

Moreover, the general term of the function \l^ {r'-) being represented by 
Air'^\ the portion of 1\ due to this term, will be 

(a) r fdffdw' smO' Q,Atjr''-''^''-Ulr' {l-ry-, 

•the limits of the integrals being the same as before. 


If now we effect the integrations relative to r' by means of the for- 
mula (3), Art. 1, and reject as before those powers of the variable /•, 
in which it is affected, with the exponent w, since these ought not enter 
into the function Fi, the last formula will become 


(a) ;r-7rrj ^-rr: r'fde'diir' sin 6' Q,A„ 

2 r C ^ + ^/J-w + g^-jy x •' ^' 

and as F, ought to contain none but the odd powers of r, we may make 
* = 2*' + l, and disregard all those terms in which s is an even number, 
since they will necessarily vanish after all the operations have been 
effected. Thus the only remaining terms will be of the form 

t^''^' fde'dsr' sin 9' Q,,.^, A,; 

2.T ' 


where, as At and 02/+ 1 are both rational and entire functions of cos 0', 
sin ff cos •ht', sin ff sin -sr', the remaining integrations from 6' = to 9' = tt, 
and Tsr' = to tjt' = 2 tt, may easily be effected in the ordinary way. 

If now we follow the process employed in the preceding article, and 
suppose To', Ti, T2, &c. are what T^, Ti, 71, &c. become when we use 
the truncated formula («') instead of the complete one (a), we shall 
readily get 

F, = t: + t: + t: + r/ + &c. 

In like manner, from the value of V^ before given, we get 

r," = fr'dr'd&dsr' sin &{1- ry(p{r")g' -" ; 

the integrals being taken from r' = r to r = l, from 9' = to 9' = ^, and 
from -ar = to tsr = 2 tt. 

Expanding now g^'" as before, we have 

r;'= t;-„+t7,+ z7.+ j7, + &c. 


U. = fdffd-sr' sin ■sr'Qjy'-'^dr'il-ry ^ (/*), 


acnd the part of U, due to the general term i?(/-''^' in (/*), will be 

(J) r'fd&diir' sm 9' Q^Bt/lr'^-"^^'-' dr' {l-ry-, 

which, by employing the formula (3') Art. 1., and rejecting the inad- 
missible terms, gives for truncated formula 

[ 2 j 

By continuing to follow exactly the same process as was before 
employed in finding the value of Fl, we shall see that * must always 
be an even number, say 2 s'; and thu« the expression immediately {Br- 
eeding will become 

,, l4!-n + 2t-2s 

^ (6-n + 23 + 2t-2s' 

2r I 

; — r'^' fdO'dw sin d'^2.- B,. 

2 J 

Moreover, the value erf V^ will be 

r, = u: + u: + u: + u: + &c. ; 

U^, Ui, Ui, U3, &c. being what Uo, Ui, Ui, &c. become when we use 
the formula {b') instead of the complete one (h). 

The value of V answering to the density 

p = p, + p, = (l-ry. /{:>/, y',z'), 

by adding together the two parts into which it was originally divided, 
therefore, becomes 

r = r,+r,= t: + t: + t: + t/ + &c. 

+ £/■„' +t4'+C7;'+t7e' + &c. 

When /3 is taken arbitrarily, the two series -entering into V extend 
in infinitum, but by supposing as before. Art. 1., 

— n n 


w representing any whole number, positive or negative, it is clear from 
the form of the quantities entering into JLs'+i and U2/, and from the 
known properties of the function F, that both these series wiU terminate 
of themselves, and the value of F' be expressed in a finite form ; which, 
by what has preceded, must necessarily reduce itself to a rational and 
entire function of the rectangular co-ordinates x, y, ss. It seems needless, 
after what has before been advanced, (Art. 1.) to offer any proof of this: 
we will, therefore, only remark that if 7 represents the degree of the 
function f{x', y\ &'), the highest degree to which V can ascend will be 

7 + 2 a> + 4. 

In what immediately precedes, w may represent any whole number 
whatever, positive or negative ; but if we make w= —2, and consequently, 

^ = ^ the degree of the function J^ is the same as that of the factor 

A^\ y', ^), 

comprised in p. This factor then being supposed the most general of 
its kind, contains as many arbitrary constant quantities as there are 
terms in the resulting function V. If, therefore, the form of the rational 
and entire function V be taken at will, the arbitrary quantities contained 
in fkpd, y, %') will in case w = — 2 always enable us to assign the corres- 
ponding value of p, and the resulting value of J'{a;', y, %') will be a rational 
and entire fimction of the same degree as T-^. Therefore, in the case 
now under consideration, we shall not only be able to determine the 
value of F' when p is given, but shall also have the means of solving 
the inverse problem, or of determining p when V is given ; and this 
determination will depend upon the resolution of a certain number of 
algebraical equations, all of the first degree. 

3. The object of the preceding sketch has not been to point out 
the most convenient way of finding the value of the function ^, but 
merely to make known the spirit of the method ; and to show on what 
its success depends. Moreover, when presented in this simple form, 
it has the advantage of being, with a very slight modification, as ap- 
plicable to any ellipsoid whatever as to the sphere itself. But when 


spheres only are to be considered, the resulting formula?, as we shall 
afterwards show, will be much more simple if we expand the density p 
in a series of functions similar to those used by Laplace {Mec. Cel. 
Liv. iii.) : it will however be advantageous previously to demonstrate 
a general property of functions of this kind, which will not only serve 
to simplify the determination of F, but also admit of various other 
applications of dcr. 

Suppose, therefore, J^''' is a function of 9 and trr, of the form con- 
sidered by Laplace {Mec. Cel. Liv. iii.), r, 9, -zs- being the polar co-ordi- 
nates referred to the axes JT, Y, Z, fixed in space, so that 

ar = r cos 0, y = r %\w9 cos Tsr, x = /• sin sin vr ; 

then, if we conceive three other fixed axes Xi, Y^, Z,, having the same 
origin but different directions, P'^'^ will become a function of 0, and •zjti, 
and may therefore be expanded in a series of the form 

(6) r <^> = r/"' + F.*'> + F/^' + F/^' + &c. . 

Suppose now we take any other point p and mark its various co-ordinates 
with an accent, in order to distinguish them from those of p ; then, if 
we designate the distance pp by {p, p), we shall have 

^ - = f r' - 2rr' [cos 9 cos ff + sin 9 sin & cos {tn- - •sr')] + r'^\ "* 

as has been shewn by Laplace in the third book of the Mec. Cel., where 
the nature of the different functions here employed is completely ex- 

In like manner, if the same quantity is expressed in the polar co- 
ordinates belonging to the new system of axes X-,, F„ Z,, we have, 
5ince the quantities r and r' are evidently the same for both systems, 

{^p, p) r \^ r r IT I 

^nd it is also evident from the form of the radical quantity of which 


the series just given are expansions, that whatever number i may re- 
present, Qi*** will be immediately deduced from Q*'> by changing 9, sr, 

9', -sr' into 0„ "sr,, 9/, ^r,'. But since the quantity - is indeterminate, 

and may be taken at will, we get, by equating the two values of . , 

. f 
and comparing the like powers of the indeterminate quantity -, 

If now we multiply the equation (6) by the element of a spherical 
surface whose radius is unity, and then by Q<*' = Q/*>, we shall have, 
by integrating and extending the integration over the whole of this 
spherical surface, 

fdf.dwQ"^ r® = fdfx, d-ar, Q/** { F/"' + Y^ + F*^' + &c. } . 

Which equation, by the known properties of the functions Q**' and Y^^\ 
reduces itself to 

when h and i represent different whole numbers. But by means of a 
formula given by Laplace {Mec. Cel. Liv. iii. No. 17.) we may imme- 
diately effect the integration here indicated, and there wiU thus result 

"-2^ + 1-^^ ' 

F;'<*> being what Fi''" becomes by changing 9^, tsti into 0,', •ar/, and as 
the values of these last co-ordinates, which belong to p, may be taken 
arbitrarily like the first, we shall have generally F,**', except when 
h = i. Hence, the expansion (6) reduces itself to a single term, and 

F® = F®. 

We thus see that the function F<'' continues of the same form even 
when referred to any other system of axes X„ F„ Z„ having the same 
origin O with the first. 

This being established, let us conceive a spherical surface whose center 
is at the origin O of the co-ordinates and radius r', covered with fluid. 
Vol. V. Part I. C 


of which the density p = P''*'' ; then, if d<r' represent any element of 
this surface, and we afterwards form the quantity 

the integral extending over the whole spherical surface, g being the 
distance p, da and y\f the characteristic of any function whatever. I 
say, the resulting value of V will be of the form 

V= Y^B; 

R being a function of r, the distance Op only and K<'' what Y'^^ becomes 
by changing 9', w, the polar co-ordinates, into 9, tit, the like co-ordinates 
of the point p. 

To justify this assertion, let there be taken three new axes JT,, I^„ Z„ 
so that the point p may be upon the axis Xx ; then, the new polar 
co-ordinates of da' may be written r', ff, tjt', those of p being r, 0, •sr. 
and consequently, the distance will become 

g = ^{r" - 2 rr' cos 9,' + r^) ; 

and as da^' = r'^d9i'd'sri sin 9,', we immediately obtain 

r = fY'^'Vde.d-sr, sin 9, f (/•-- 2rr' cos d,' + O 

= r'^SZd9; sin 0/ ^{r'-^rr' cos 0/ -f r'^)f^Zd-ur( Y' <". 

Let us here consider more particularly the nature of the integral 

In the preceding part of the present article, it has been shown that 
the value of Y'^'^, when expressed in the new co-ordinates, will be of 
the form P'/*'' ; but aU functions of this form (Vide Mec. Cel. Liv. iii.) 
may be expanded in a finite series containing 2 « + 1 terms, of which 
the first is independent of the angle "sr,', and each of the others has 
for a factor a sine or cosine of some entire multiple of this same angle. 
Hence, the integration relative to ro-/ will cause all the last mentioned 
terms to vanish, and we shall only have to attend to the first here. 
But this term is known to be of the form 

, / ,. i.i — \ ,. „ i.i—l.i-2.i — S ,,• , « N 

mh green, on the laws of the equilibrium of fluids. 19 
and consequently, there will result 

where ni = cos 9^ and ^ is a quantity independent of 6/ and tr/, but 
which may contain the co-ordinates 9, -ar, that serve to define the 
position of the axis JCi passing through the point p. 

It now only remains to find the value of the quantity k, which may 
be done by making 0i' = O, for then the line r coincides with the axis 
JTi, and K*'' during the integration remains constantly equal to Y^\ 
the value of the density at this axis. Thus we have 

^ ^rin ^ 7 [-. ii—l i.i — l.i—2.i — 3 „ \ 
V 2.2?— 1 2.4.2«— 1.2^ — 3 I 

or, by summing the series within the parenthesis, and supplying the 
common factor 2 7r, - 

•jr(i) _ ^-^-^ ^ J, 

1.3.5 2«-l ' 

and, by substituting the value of k, draAvn from this equation in the 
value of the required, integral given above, we ultimately obtain 

If now, for abridgement, we make 

^^> = ^' - 2:27:11^' + 2.4.2i-1.2i-3 ^' -^^- 

we shall obtain, by substituting the value of the integral just found in 
that of V before given, 

r= r(^27rr'%i44^^^-^^^^^/_}r?^/(^H(^-2rr'M/ + r'^); 

which proves the truth of our assertion. 

From what has been advanced in the preceding article, it is likewise 
very easy to see that if the density of the fluid within a sphere of 
any radius be every where represented by 

c 2 


<p being the characteristic of any function whatever; and we afterwards 
form the quantity 

where dv represents an element of the sphere's volume, and g the dis- 
tance between dv and any particle p under consideration, the resulting 
value of V wiU always be of the form 

V^^ being what I^'*" becomes by changing 9^, nr , the polar co-ordinates 
of the element dv into Q, w, the co-ordinates of the point p; and R 
being a function of r, the remaining co-ordinate of p, only. 

4. Having thus demonstrated a very general property of functions 
of the form P"*'', let us now endeavour to determine the value of F" 
for a sphere whose radius is unity, and containing fluid of which the 
density is every where represented by 

p = {l-x''-y"-zyf{x',y',z'); 

on', y , z' being the rectangular co-ordinates of dv, an element of the 
sphere's volume, and Jl the characteristic of any rational and entire 
function whatever. 

For this purpose we will substitute in the place of the co-ordinates 
x', y , z' their values 

x = r cos &\ y = r sin & cos w'. z = r' sin ff sin -bt' ; 

and afterwards expand the function/(a;', y', s) by Laplace's simple method, 
{Mec. Cel. Liv. iii. No. 16.). Thus, 

(7) /{x, y, z) =/<«>+/'" +/'<^> + &c +/'«; 

s being the degree of the function /{x, y', z'). 

It is likewise easy to perceive that any term /'■''' of this expansion 
may be again developed thus, 

/'(•■) =/;(•■>/* +/'«/-= +^'<V^+^ + &c.; 

and as every coefficient of the last developement is of the form [/'", 
(Mec. Cel. Liv. iii.), it is easy to see that the general term y'''V'+^' may 
always be reduced to a rational and entire function of the original 
co-ordinates x, y', »'. 

If now we can obtain the part of ^ due to the term 

we shall immediately have the value of V by summing all the parts 
corresponding to the various values of which i and t are susceptible. 
But from what has before been proved (Art. 3.), the part of V now 
under consideration must necessarily be of the form F"*'^; representing, 
therefore, this part by F"/'', we shall readily get 

r/" =/J/'+^'+^</r' (1 - ryjtl-nr'de' sin ^/'<'^^'-". 

Moreover from what has been shown in the same article, it is easy 
to see that we have generally 

fV'^'^clu'de' sin e'^ig"-) = Stt F« ^f'f •:'^'~^ /-i'c?Mi' (i) yl^{r'-2rr',x,' + r") ; 

\(/ being the characteristic of any function whatever, and P'^'' what P''"* 
becomes by substituting 9, w the polar co-ordinates of p in the place 
of 6', TB-', the analogous co-ordinates of the element dv. If therefore 
in the expression immediately preceding, we make 

F'«=/'« and fig^) =^'.- = (^^)^, 

and substitute the value of the integral thus obtained for its equal in 
Vf-'^ there will arise 

where yj® is deduced from Jl'^^ by changing 9", %r' into 6, w, and (i), for 
abridgement, is written in the place of the function 

,; i.i-1 ,;_; , i.i-l.i-2.i-3 ,^_^ . 
'"~2.2e-l'" ^2.4.2i-1.2«-3'" *'''• 

As the integral relative to n\ which enters into the expression on 
the right side of the equation (8) is a definite one, and depends therefore 
on the two extreme valvies of fj.\ only, it is evident that in the deter- 
mination of this integral, it is altogether useless to retain the accents 


by which n\ is affected. But by omitting these superfluous accents, 
we shall have to calculate the value of the quantity 

2 , 

fj.dfi. (^) . (r' - 2 rr'/m + r") 

,. . i.i — 1 . i.i-i,i-2.i-3 ; , - 

^'^ = ''-2:2^:1'' + 2.4.2i-1.2i-3 -^ -^^- 

The method which first presents itself for determining the value of 


the integral in question, is to expand the quantity {r^ — 2rr'/u. + r'^) ^ by 
means of the Binomial Theorem, to replace the various powers of m by 
their values in functions similar to (i) and afterwards to effect the in- 
tegrations by the formulee contained in the third Book of the Mec. Cel. 
For this purpose we have the general equation 

.-s i ... , i.i — 1 ,. „. , i.i—l.i — 2.i — 3,. ,, 

^^^ '^ =^^)+ 2:271:1 (^-^^-^2X2I33:27::5(*-*) 

i.i-l.i-2.i-3.i-4!.i-5 . 
2.i.6.2i-5.2i-7.2i-9 ^' '") + ^^' 

To remove all doubt of the correctness of this equation, we may 
multiply each of its sides by (i, and reduce the products on the right 
by means of the relation 

which it is very easy to prove exists between functions of the form (?). 
In this way it will be seen that if the equation (9) holds good for any 
power fx' it will do so likewise for the following power ^'+^ and as it 
is evidently correct when i='l, it is therefore necessarily so, whatever 
whole number i may represent. 

Now by means of the Binomial Theorem, we obtain when r^r' 

= 2, 

r"'-K(r'-2rr'^ + r"y = (i-2m J + ^,) 
y>n—l.n + l.n-\-3 n + 2s — 3 



If now we conceive the quantity (2ju- rj to be expanded by 


the same theorem, it is easy to perceive that the term having f— | 
for factor is 

i + 2t' 

7i-l.n-\-l.n + 3 « + 2^• + 4^'-3 ,^,„ ,^„, /r\ 

2.4.6 2e + 4r '^ W) 

i + iV 

n-l.n + 1 n + 2i + 4>t'-5 ,^,,_, ,+,„_, fn'^"''!! i+^t-1 

2 . 4 2? + 4#'-2 '" """ ' \r') 

2.4 2« + 4ir-4 ^ '^^ VJ /•''■ 1.2 

— &c &c &c 

/^\ i + 2(' 

and therefore the coefficient of I — I in the expansion of the function 

will be expressed by 

v "-^-^ + l ^ + 2^'+4^'-2^- 3 ,,,,.,, .. 

2.4 2« + 4^'-2* ^""^^ '^ ^ 

« + 2#'-*.^ + 2^'-«-l ^■ + 2^-2* + l 

Hence the portion of this coefficient containing the function (i), when 
the various powers of /i have been replaced by their values in functions 
of this kind agreeably to the preceding observation will be found, by 
means of the equation (9), to be 

. .X „ n — l.n + 1 n + 2i+4<t'~2s — 3 

^^^ 2 . 4 2? + 4^'-2* 

^^ + 2/-2^.^• + 2^'-2,y-l i + 1 

"^ 2.4 2^-2*x2^■ + 2if'-2* + 1.2^ + 2f-2#-1...2^■ + 3 

i + 2t'-s.i+2t'-s-l i + 2^-2#+l 



n-l.n + 1. n + S n + 2i+it' -2s-3 ,,,,.,.,.. 

■^^2.4.6 2« + 4^-2« ^ ^ 


^^ + l.^^ + 2.^^ + 3.^' + 4 i + ^t' -s 

^ 1.2.3 *x2.4.6 2^-2* X 2? + 2^'- 2* + 1 2e + 3 

_ . (-1V.W-1.W + 1.W + 3 w + 2? + 4^' -2.9 -3 

-2'-W-2g^ 2?x2.4 2*x2.4 2#'-2* x 2^■ + 2^'-2*+1...2^■ + 3 

3.5.7 2« + l ... n-\.n + \.n + S ?« + 2m2^'-3 



1.2.3 i ^^ 3.5.7 2« + 2r + l 

r - 1 V ^ + ^^' + ^^-1 w + 2? + 4/-2.y-3 

^ "^ ^ 2.4.6 2^-2* 

2t + 2^'-2.y + 3 2? + 2/' + 1 

^ 2.4.6 2* 

where all the finite integrals may evidently be extended from * = 
to * = 00 , and it is clear that the last of these integrals is equal to the 
coefficient of a^ in the product 

(, w + 2i + 2#'-l » + 2i + 2/-l.w + 2i + 2/' + l „ „ . . ., 
{1+ x + ^-^ oi? + hc.tninf.\ 

,, 2? + 2^ + l , 2i + 2r + 1.2? + 2#'-l „ . . . ^, 
X {1 ~ x+ ^-^ af-kc. intnf.] 

If now we write in the place of the series their known values, the 
preceding product will become 

n + 2i+2f-l 2i+2t' + l i-n 

(l-or) " x(l-;r) ' ={\-x)\ 

and consequently the value of the required coefficient of af^ is 

« — 2.W.W + 2 w + 2/' — 4 

2 .4. 6 2^' 

This quantity being substituted in the place of the last of the finite 
integrals gives for the value of that portion of the coefficient of 

which contains the function (i) the expression 

3.5.7 2? + l n-l.n + 1 w+2?'+2^'-3 n-2.n m+2^ -4 , ., 

1.2.8 i ^3.5 2e + 2^ + l ^ 2.4 2t' ^*^' 


By multiplying the last expression by ( — ) , and taking the sum 
of all the resulting values which arise when we make successively 

^ = 0, 1, 2, 3, 4, 5, 6, &c. in infinitum, 
we shall obtain the value of the term I^<'> contained in the expression 

, l-n 

(l - 2m ^ + ^,) ~ = Y^'^ + rc> + F(=> + F^^' + &c. 

1.2 i ^^^ 3.5 2i + 2f+l 

n-2.n n + ^t'-i /rV*^*' 

"^ 2.4 2t' [?) ' 

the finite integral extending from t' = to t' = oc. 

But by the known properties of functions bf this kind, we have 
by substituting for F'"' its value 

/_\d^ (i) (l - 2m p + ^^~=/-\d^ (i) . F« 

3.5.7 2« + l ., ,.,„ ^n-l.n + l n + 2i + 2t'-3 

= 1.2.3 i /^^(O^x^ 3 . 5 2e + 2^' + l 

n-2.n « + 2/-4/rV*"' 

2.4 2t' 


^ 1.2.3 i n-l.n + 1 n + 2i + 2t'-3 

~ 1.3.5 2?-l 3 . 5 2i + 2t' + 2 

71 — 2. n n + 2t' — ^! fr' 

t' ~ \r'} ' 

2.4 2t 

since by what Laplace has shown {Mec. Cel. Liv. iii. No. 17.) 

■^^^ W^= 2m I 1.3.5 2e-l j • 

Vol. V. Paet I. D 


If now we restore to n. the accents with which it was originally 
affected, and multiply the resulting quantity by r'""\ we shall have when 

(10) /Ad^\(i) if-^rr't^, + ■' =/»-y_;</M'i ii) (l - UJ-. + ^-.) 

_ ,j_„ 1.2.3 i w-l.w + 1 ?^ + 2^^ + 2^-3 

~ '1.3.5 2«-l 3 . 5 2e + 2#'+l 

;« — 2.W « + 2#'-4 /r 

2.4 2jf' 


j + 21' 

and in order to deduce the value of the same integral when /•' /. r, we 
shall only have to change r into /, and reciprocally, in the formula 
just given. 

We may now readily obtain the value of Vi^ by means of the 
formula (8). For the density corresponding thereto being 

:/;«/•'+=' (l-r"7, ■ 

it follows from what has been observed in the former part of the 
present article, that ^'®r'^^' may always be reduced to a rational and 
entire function of icf, y, %' the rectangular co-ordinates of the element 
dv, and therefore the density in question will admit of being expanded 
in a series of the entire powers of x, y', %' and the various products of 
these powers. Hence (Art. 1.) F/'' admits of a similar expansion in 
entire powers, he. of x, y, z the rectangular co-ordinates of the point p, 
and by following the methods before exposed Art. 1 and 2, we readily 

^t ^-^J J r' ui y>. , ) .^ 3.5 2t + 2t' + l 

n-2.n.n + 2 » + 2#' — 4 /r\'+"' 


2 .4. 6 2t' 

-4 /r^y^-"" 

" V'j ' 

and thence we have ultimately, 

(,ii; rt Airj, ^33 2i + 2t' + 1 2.4 2/' 


r r-%^"-'' )r(3+i) r{/3 + i)r(i^) 


=2Trf,^\ ^^ - — -r' 

/ 2t-2t' + 2fi + 6-n \ ■^' / 6 + 2/3-w\ 

4-n.6-n 2t~2t'+2-n n-2.n n + 2t'-4! 

6 + 2fi-n 2t-2t'+2l3 + 4>-n ^ 2.4 W 

n—1 .n + 1 n + 2i + 2t'-3 ^ 

"^ 3 . 5 2i + 2t' + l ' 

the finite integrals being taken from t' =0 to t'=cD and r being the 
characteristic of the well known function Gamma, which is introduced 
when we effect the integrations relative to r' by means of the formula 
(3), Art. 1. 

Having thus F"/" or the part of F corresponding to the term j^''*', 
in J'(x', y, as') we immediately deduce the complete value of V by giving 
to i and t the various values of which these numbers are susceptible, 
and taking the sum of all the parts corresponding to the different terms 
hi the expansion of the function fix', y', &'). 

Athough in the present Article we have hitherto supposed J" to be 
the characteristic of a rational and entire function, the same process will 
evidently be applicable, provided y"(a;', y, z') can be expanded in an 
infinite series of the entire powers of x', y, z' and the various products 
of these powers. In the latter case we have as before, the development 

fix', y, z') =/'<»> +/'<•> + /'® +/'<^) + &c. 

of which any term, as for example f'^''> may be farther expanded as 

/'« =/;«r" + /'«r"+^ +/'«/•"+*+ &c. 

and as we have already determined F"/*' or the part of V corresponding 
toyt'''V'+^'', we immediately deduce as before the required value of V, 
the only difference is, that the numbers i and t, instead of being as 
in the former case confined within certain limits, may here become in- 
definitely great. 

D 2 


In the foregoing expression (11) /3 may be taken at will, but if we 

qq ^ 

assign to it such a value that -~ — may be a whole number, the 

series contained therein will terminate of itself, and consequently the 
value of Vt^^ will be exhibited in a finite form, capable by what has 
been shown at the beginning of the present Article of being converted 
into a rational and entire function of x, y, %, the rectangular co-ordinates 
of p. It is moreover evident, that the complete value of V being com- 
posed of a finite number of terms of the form Vt-'^ will possess the same 
property, provided the function fix, y , %) is rational and entire, which 
agrees with what has been already proved in the second Article, by a 
very different method. 

(5) We have before remarked, (Art. 2.) that in the particular case 

where /3 = — — — , the arbitrary constants contained in y(a;', y' , %) are just 

sufficient in number to enable us to determine this function, so as to 
make the resulting value of V equal to any given rational and entire 
function of x, y, z, the rectangular co-ordinates of p, and have proved 
that the corresponding functions V and J" will be of the same degree. 
But when this degree is considerable, the method there proposed becomes 
impracticable, seeing that it requires the resolution of a system of 

^ + 1 .^ + 2,.s + 3 

linear equations containing as many unknown quantities ; s being the 
degree of the functions in question. But by the aid of what has been 
shown in the preceding Article, it will be very easy to determine for 
this particular value of /3 the function J'{x', y, %') and consequently the 
density p when F' is given, and we shall thus be able to exhibit the 
complete solution of the inverse problem by means of very simple 

For this purpose, let us suppose agreeably to the preceding remarks, 
that p the density of the fluid in the element dv is of the form 

p = {l-r^)-^/{x',y',z); 


f being the characteristic of a rational and entire function of the same 
degree as V, and which we will here endeavour so to determine, that 
the value of V thence resulting, may be equal to any given rational 
and entire function of x, y, % of the degree s. 

Then by Laplace's simple method {Mec. Cel. Liv. iii. No. 16.) we 
may always expand F" in a series of the form 

r= r<«> + r(» + r® + &c + r«. 

In like manner as has before been remarked, we shall have the 
analogous expansion 

f{x',y', ,')=/''«' +/'<'>+/'^=>+/'<'>+ &c +/'«, 

of which any termy*'' for example, may be farther developed as follows, 

/'« =^'('V' +y;'''V"+'^ +//»/'+* + &c. = r" (/'« +y;'('V'^ +/'»;.'^ + &c.) 

y", yj'<'>, j^''*^, &c. being quantities independent of / and all of the form 
K'"' {Mec. Cel. Liv. iii.) Moreover F/" the part of F' due to the general 

term Jl'^'^r''+^* of the last series, will be obtained by writing for (i 

in the equation (11), and afterwards substituting for 

(n — 2\ _ f4i-n 

r(!t^)r(l^) us value- 


In this way we get 

27r;/;'V ±-„,e-n 2t-2f + 2-n 

' . fn-2 \ 2.4 2t~2t' 

sm (-^.j 

n-2.n w + 2#'-4 w — 1 . w + 1 n + 2i + 2f — 3 ^ 

^ 2.4 2? ^ 3 . 5 ...... 2i + 2t' + l ' 

yj<'^ being what J]''-^ becomes by changing 6', -ar' into 6, sr, and the finite 
integral being taken from t' = to t'=<x . 

Let us now for a moment assume 


n-2.n w + 2#' — 4 «-l.« + l n + 2i + 2t'-3 


2.4 2t' 3.5 2i + 2t' + l ' 


then the expression immediately preceding may be written 

■^J7\r 4<-n.6-n Q,t-2t' + 2 

(n-2 \ 2.4 2t-2t' 

dn [-^ .) 


and by giving to t the various values 0, 1, 2, 3, &c. of which it is sus- 
ceptible, and taking the sum of all the resulting values of F/'' the quantity 
thus obtained will be equal to V^^ or that part of V which is of the 
form Y^\ Thus we get 

27r^ r' 




+ &c &c &c. 

since aU the terms in the preceding value of Vi-^ in which t'>t vanish 
of themselves in consequence of the factor 

/ 2^-2^ +4-w \ 

= (when t > t). 

2 . 4 2i?-2^ „,. ., . _ „ /4-w 

(-'■«) rC-?) 

But F"^'^ as deduced from the given value of V may be expanded in 
a series of the form 

r«=?''. {r„®-i- r;(v^+ r,»./^+ v^'>t^+kc.\ 

and if in order to simplify the remaining operations, we make generally 

__„ 27r^ n-2.n M+2/-4 n-\.n + l n + 2i+2t-3 ^„., 

' . [n-2 \ 2.4 2t 3.5 2« + 2^+l ' 

sm [—.] 




. /n-2 \ 
sm (— .) 

the equation immediately preceding will become 

^(0= ^I-:£ X {(j) (0) . f7o"' + 0(l) . t7;(W(^ (2) C7,». r' + &cc.\ 

[n — 2 \ 

which compared with the foregoing value of F^'\ will give by suppressing 

the factor '— , common to both, and equating separately the 


coefficients of the different powers of the indeterminate quantity r the 
following system of equations 

&c=...&c &c &c. 

for determining the unknown functions fo'-^, /<*', f/\ &c. by means of the 
known ones f7"o*'\ Ui'-'\ ZJg*", &c. In fact the last equation of the system 
gives U^^=fP, and then by ascending successively from the bottom to 
the top equation, we shall get the values of fs^\ /,%, f}%, &c. with 
very little trouble. It will however be simpler still to remark, that the 
general type of all our equations is 

where the symbols of operation have been separated from those of 
quantity and e employed in its usual acceptation, so that 

But it is evident we may satisfy the last equation by making 


Expanding now and replacing e?7„'''; e^UJ^^, &c. by these values UJIi, 
U^%, &c. we get 

from which we may immediately deduce ^'® and thence successively 

/'« = r" (/„'« +/'« r'^ +/;« r" +//« r'» + &c.) 
fW, !/, 85') =/'<"' +/'«+/'<^' + &c +/'« 

and > = (1 - X'"- - y" - z'"-y^.f(x', y , %), 

Application of the general Methods exposed in the preceding Articles 
to Spherical conducting Sodies. 

(6) In order to explain the phenomena which electrified bodies 
present. Philosophers have found it advantageous either to adopt the 
hypothesis of two fluids, the vitreous and resinous of Dufay for 
example, or to suppose with jEpinus and others, that the particles of 
matter when deprived of their natural quantity of electric fluid, possess 
a mutual repulsive force. It is easy to perceive that the mathematical 
laws of equilibrium deducible from these two hypotheses, ought not to 
differ when the quantity of fluid or fluids (according to the hypothesis 
we choose to adopt) which bodies in their natural state are supposed 
to contain, is so great, that a complete decomposition shall never be 
effected by any forces to which they may be exposed, but that in 
every part of them a farther decomposition shall always be possible by 
the application of still greater forces. In fact the mathematical theory 
of electricity merely consists in determining p* the analytical value of 

* It may not be Improper to remark that p is always supposed to represent the density 
of the free fluid, or that which manifests itself by its repulsive force; and therefore, when 
the hypothesis of two fluids is employed, the measure of the excess of the quantity of either 



the fluid's density, so that the whole of the electrical actions exerted 
upon any point p, situated at will in the interior of the conducting 
bodies may exactly destroy each other, and consequently p have no 
tendency to move in any direction. For the electric fluid itself, the 
exponent n is equal to 2, and the resulting value of p is always such 
as not to require that a complete decomposition should take place in 
the body under consideration, but there are certain values of n for which 
the resulting values of p will render fpdv greater than any assignable 
quantity ; for some portions of the body it is therefore evident that 
how great soever the quantity of the fluid or fluids may be, which 
in a natural state this body is supposed to possess, it will then become 
impossible strictly to realize the analytical value of p, and therefore some 
modification at least will be rendered necessary, by the limit fixed to 
the quantity of fluid or fluids originally contained in the body, and 
as Dufay's hypothesis appears the more natural of the two, we shall 
keep this principally in view, when in what follows it may become 
requisite to introduce either. 

7. The foregoing general observations being premised, we will proceed 
in the present article to determine mathematically the law of the density 
p, when the equilibrium has established itself in the interior of a con- 
ducting sphere A, supposing it free from the actions of exterior bodies, 
and that the particles of fluid contained therein repel each other with 
forces which vary inversely as the w"" power of the distance. For this 
purpose it may be remarked, that the formula (1), Art. 1, immediately 
gives the values of the forces acting on any particle p, in virtue of 
the repulsion exerted by the whole of the fluid contained in A. In 
this way we get 

1 dV 
- _ .-jr = the force directed parallel to the axis X, 

1 dV 

- ■ _ . y— = the force directed parallel to the axis Y, 

fluid which we choose to consider as positive over that of the fluid of opposite name in any 
element dv of the volume of the body is expressed by pdv, whereas on the other hypothesis 
pdv serves to measure the excess of the quantity of fluid in the element dv over what it 
would possess in a natural state. 

Vol. V. Paet I. E 


1 dV 

.-^- = the force directed parallel to the axis Z. 

1 — » d% 

But since, in consequence of the equilibrium, each of these forces is 
equal to zero, we shall have 

„ dV J . dV , , dV . .-, 
= -5— dx + -7— dy + -J- d% = dV\ 
dx dy d% 

and therefore, by integration, 

F = const. 

Having thus the value of V at the point p, whose co-ordinates are 
X, y, %, we immediately deduce, by the method explained in the fifth 


/w-2 \ 


P = 



seeing that in the present case the general expansion of K there given 
reduces itself to 

If moreover Q serve to designate the total quantity of free fluid in 
the sphere, we shall have, by substituting for 

sin f TT j its value 


\ 2 / rrz-ij >i.Ji/i '2\""S~ ^"^ 


Q = /pe/«; = ^5 i F/liW^dril-r") 

See Legendre. Exer. de Cal. Int. Quatrieme Partie. 

In the preceding values, as in the article cited, the radius of the 
sphere is taken for the unit of space ; but the same formula may 

readily be adapted to any other unit by writing — in the place of r', 

and recollecting that the quantities p, V, and Q, are of the dimensions 
0, 4 — «, and 3 respectively, with regard to space; a being the number 


which represents the radius of the sphere when we employ the new 
unit. In this way we obtain 

P = — V4 — - r{a'-r")~, and Q = lii — ; . V. 

Hence, when Q, the quantity of redundant flviid originally introduced 
into the sphere is given, the values of V and of the density p are like- 
wise given. In fact, by writing in the preceding equation for 

ry, and sin(^7r), 

their values, we thence immediately deduce 

and F= ^ ' ■ ' a'-.Q. 


The foregoing formulae present no difficulties where « > 2, but when 
H < 2, the value of p, if extended to the surface of the sphere Jl, would 
require an infinite quantity of fluid of one name to have been origi- 
nally introduced into its interior, and therefore, agreeably to a preceding 
observation, could not be strictly realized. In order then to determine 
the modification which in this case ought to be introduced, let us in 
the first place make n>2, and conceive an inner sphere S, whose 
radius is a — Sa, in which the density of the fluid is still defined by 
the first of the equations (12); then, supposing afterwards the rest of 
the fluid in the exterior shell to be considered on ^'s surface, the portion 
so condensed, if we neglect quantities of the order Sa, compared with those 
retained, will be 

-- r f^±i) 

2* V 2 / ^, 

' V2/ 

E 2 





and since, in the transfer of the fluid to ^'s surface, its particles move 
over spaces of the order ^a only, the alteration which will thence be 
produced in V will evidently be of the order 

n — 2 n 

and consequently the value of V will become 

k being a quantity which remains finite when ^a vanishes. 

In establishing the preceding results, ti has been supposed greater 
than 2, but p the density of the fluid within S and the quantity of it 
condensed on ^'s surface being still determined by the same formulae, 
the foregoing value of V ought to hold good in virtue of the generality 
of analysis whatever n may be, and therefore when w is a positive quantity 
and hi is exceedingly small, we shall have without sensible errors 

v;^m^m«'--* ' 

Conceiving now P' to represent the density of the fluid condensed 
on A's surface, 47ra^P' will be the total quantity thereon contained, which 
being equated to the value before given, there results 



and hence we immediately deduce 

fn + V 

n — 4 -p 



Moreover as Q represents the total quantity of redundant fluid in the 
entire sphere A, the quantity contained in B is 

Mh green, on the laws of the equilibrium of fluids. 37 


If now when w is supposed less than 2, we adopt an hypothesis 
similar to Dufay's, and conceive that the quantities of fluid of opposite 
denominations in the interior of A are exceedingly great when this 
body is in a natural state, then after having introduced the quantity Q 
of redundant fluid, we may always by means of the expression just 
given, determine the value of Sa, so that the whole of the fluid of 
contrary name to Q, may be contained in the inner sphere S, the 
density in every part of it being determined by the first of the equa- 
tions (12). If afterwards the whole of the fluid of the same name as 
Q is condensed upon A's svirface, the value of V in the interior of S 
as before determined will evidently be constant, provided we neglect 


indefinitely small quantities of the order ht'\ Hence all the fluid con- 
tained in J3 will be in equilibrium, and as the shell included between 
the two concentric spheres, A and S is entirely void of fluid, it follows 
that the whole system must be in equilibrium. 

From what has preceded, we see that the first of the formulae (12) 
which served to give the density p within the sphere A when n is 
greater than 2, is still sensibly correct when n represents any positive 
quantity less than 2, provided we do not extend it to the immediate 
vicinity of A's surface. But as the foregoing solution is only approxi- 
mative, and supposes the quantities of the two fluids which originally 
neutralized each other to be exceedingly great, we shall in the follow- 
ing article endeavour to exhibit a rigorous solution of the problem, 
in case w < 2, which will be totally independent of this supposition. 

8. Let us here in the first place conceive a spherical surface whose 
radius is a, covered with fluid of the uniform density P', and suppose 
it is required to determine the value of the density p in the interior 
of a concentric conducting sphere, the radius of which is taken for 
the unit of space, so that the fluid therein contained, may be in equi- 
librium in virtue of the joint action of that contained in the sphere 
itself, and on the exterior spherical surface. "• 


If now V represents the value of V due to the exterior surface, 
it is clear from what Laplace has shown, {Mec. Cel. Liv. ii. No. 12.) that 

^ = !y^ = (3^:^ K«+^r"-(«-'-M; 

rfo- being an element of this surface, and g being the distance of this 
element from the point p to which V is supposed to belong. 

If afterwards we conceive that the function V is due to the fluid 
within the sphere itself, it is easy to prove as in the last article, that 
in consequence of the equilibrium we must have 

V +V= const. 

But V and consequently V is of the form F^"', therefore by employ- 
ing the method before explained, (Art. 4.) we get 

/(ar', y', %) =/'(»' =/„("> +/(">. r'' +//>. r'' + &c. = B, + B,r'' + B, r" + &c. ; 

where, as in the present case, ^''°>, yi'<°', ^''% &c. are all constant 
quantities, they have for the sake of simplicity been replaced by 

J?o, jBi, B.^, &c. 

Hitherto the exponent /3 has remained quite arbitrary, but by making 

/3= — -— ^ the formula (11) Art. 4. will become when « = 0, 

ir(o)_o 7? < 2y V 2 ; ^ ,,, 4-».6-w 2t-9.t' + ^-n 

^' -2'^^'- YW) ' 4 . 6 2^-2^+2 

n-2.n-l « + 2^- 3 

"^ 2.3.4 2^ + 1 

{/i-2)Tr'Bt ^ ,„ 4-W.6-M 2^-2/' + 2-« w-2.«-l « + 2/-3 

2.?^ . ^ -r^ ^w . n ^ 

. (71-2 \ • 4.6 2^-2^ + 2 2 . 3 2^' + l ' 

sm(— .) 

Giving now to / the successive values 0, 1, 2, 3, &c. and taking 
the sum of the functions thence resulting, there arises 

r= F<°' = Fo<"> + rr + T^-P + ^3<°* + &c. = s. r/"> 

(«-2)7r'^ ^P^ ,,, 4-W.6-W 2t-2f + 2-n n-2.n-l «+2^'-3 

"""T^ri"^ ' 4.6.8 2^-2^' + 2 "" 2 . 3 2/' + l ' 

sm (^.) 

where the sign S is referred to the variable t and 2 to ^. 


Again, by substituting for V and V their values in the equation 
V^ + V= const, and expanding the function V we obtain 

,/^*' ti-2.n-l.n n + 2f-3 

const. =47rP'«'-".2 

«^"" 2 . 3 .4 2^ + 1 

(w-2)7r' „ „ ,, /4-».6-?? 2^-2^'+2-w n-2.n-l ti+2t'+3 

"^ (n-2 X '^^^''' 4 . 6 2f-2f + 2 ^ 2.3.4 2^ + 1 




which by equating separately the coefficients of the various powers of 
the indeterminate quantity r, and reducing, gives generally 


2P'«'-"-"'. sin /• 


2 I „2 — «.4 — « 2s — n 

-^ 2 . 4> 2s ^''^'-' 

Then by assigning to t' its successive values 1, 2, 3, &c. there results for 
the determhiation of the quantities B^,Bi, JB^, &c. the following system 
of equations, 

2P' „ . (n-2 \ p ^ 2-n „ 2-n.4<-n „ , » 

2P' . fn-2 \ „ „ 2 — « „ 2 — ;«.4 — ra ^ . 

=^ «'-".sin — r-TT .«-^ = P,-f — r- P. + — ^— r— P3 + &C. 

TT V 2 / 2 2.4 

2P' , . (n-2 \ ... „ 2-M„,2-ra.4-/< 

a'-", sm ( -^— t] • '^ = P.' + —^ -^^ + ~~;> 4 — P4 + &C. 

&c &c &c &c 

But it is evident from the form of these equations, that we. may satisfy 
the whole system by making 

B, = B^.a\ B, = B.a-\ B, = B,,(r\ B, = B^.a-M>ic. 
provided we determine Po by 

2P' . /n-2 \ „ ,, 2-n , 2 — n.4>-n ... . , 

= P„(l-a-)' 
„ 2P . (n-2 \ . , ,,^ 

,-2\ 2 

n — 2 
Hence as in the present case, ^ = —^ , we immediately deduce the 

successive values 

2P . fn-2 


/(or', y\ «')=/'(") = ^„ + P,r'^ + ^.r'^ + &c. = if„ (l - ^']", 
and p = (1 - O ^ ./(x', y', %') = ^ sin {—^ ^) . («= - 1)"... 

(«^ - /'')-Ml - 0~- 

In the value of p just exhibited, the radius of the sphere is taken 
as the unit of space, but the same formula may easily be adapted to 

any other unit by writing j and y- in the place of a and / respectively, 

and recollecting at the same time that in consequence of the equation 

•dv.p . rdaP' 


= r^r^j!^ + ji 

s -^ g 

before given, ^ , is a quantity of the dimension — 1 with regard to 

space: h being the number which represents the radius of the sphere 
when we employ the new unit. Hence we obtain for a sphere whose 
radius is bg, acted upon by an exterior concentric spherical surface 
of which the radius is a, 

2P'a.sin {—-■"] 2-n ^ 

(/3) p = -^ ff-b') ' {a'-r")-' {b'-r") ^ ; 


P' being the density of the fluid on the exterior surface. 

If now we conceive a conducting sphere A whose radius is a, and 
determine P' so that all the fluid of one kind, viz. that which is re- 
dundant in this sphere, may be condensed on its surface, and afterwards 
find b the radius of the interior sphere S from the condition that it 
shall just contain all the fluid of the opposite kind, it is evident that 
each of the fluids will be in equilibrium within A, and therefore the 
problem originally proposed is thus accurately solved. The reason for 
supposing all the fluid of one name to be completely abstracted from 
S, is that our formulas may represent the state of permanent equilibrium, 
for the tendency of the forces acting within the void shell included 
between the surfaces A and B, is to abstract continually the fluid of 
the same name as that on ^'s surface from the sphere S. 


To prove the truth of what has just been asserted, we will begin 
with determining the repulsion exerted by the inner sphere itself, on 
any point p exterior to it, and situate at the distance r from its centre 
O. But by what Laplace has shown {Mec. Cel. Liv. ii. No. 12.) the 
repulsion on an exterior point p, arising from a spherical shell of which 
the radius is r', thickness dr and center is at O will be measured by 

I'Kr'dr'p d_ (r + r)^-" - (r-/)^-" 

1 — ti.S — n'dr' r ' 

the general term of which when expanded in an ascending series of 

the powers of — is, 

+ ^"- 27Tl ^-.r-prfr, 

and the part of the required repulsion due thereto will, by substituting 
for p its value before found, become 
%F . (n-2 \ ,o ,,,^ -2 + wx«.» + l M+2*-3xw + l*-l 

Si' . fn-2 \ , , ,„,: 

2.3 . 4 2« + l 

,'2\ -1 »-2 


It now remains to find the value of the definite integral herein con- 

tained. But when 11 -\ is expanded, and the integrations are 

effected, by known formulae, we obtain 

(14) M 1 - ^ I (*' - O ' r''^^'dr'=/o' 2^ -jj {b'-r") ' .r"'+'dr' 
^^ "a'' r^, ,3 X "^ 'f n 3\ 

, 2^ + 3 y 2^ + 3.2^ + 5 ¥ 

* 25 + 3 + w a' 2* + 3 + ».2« + 5 + wa' "*■ ^'^ 

^,^, ^ (2) ^ r "^ 2) {^s+l-^n){l-xy'' r^ 

_, , ^, ,.„ . _ „ , , „ 'dx 

, 1J2S + 1 

2° • / « 3> 

r (.+ - + -) (1-.^)- 

Vol. V. Pakt I. 


r f^^l r (-\ — ' 

\2J \2) 1.3.5 2s + lb''+'+" (l-af)' fx'^^'dx 

1+w ^ ^•'^+'+" ^o' 


J {\-x) 

Avhere after the integrations have been effected, x ought to be made 
equal to - . 

The value of the integral last found being substituted in the expres- 
sion immediately preceding, and the finite integral taken relative to s 
from * = to * = X gives for the repulsion of the inner sphere. 

a ^ ll+n\ f2 — n 

¥~J V'~2 


-2.7i.n + 2 n + 2s-4> (by"(l-x')'' raf'^'dx 

''^" 2.4.6 2s [r) ic^'+'-^" J„ 



-47rv/,rP^a^/-" » n-2.n.n + 2 w + 2^-4 (a^ rj..^„ /, _^.^\ 

„fl+n\ l2-n<:^' 2.4.6 2s (rj ■'"'^^'^ ' ^* ^' ' 

i ) [ 2 J 

since F (^) =\/ir, sin ( tt] = 


and as was before observed, a; = - . 


But we have evidently by means of the binomial theorem, 

/ _ ff-j;-\i^ _ . n-2.n. w4-2 w + 2^-4 /«^y, 

I rW ' "" ° 2.4.6 2* \r] ' 

and therefore the preceding quantity becomes 

(15) 7^^ L dxaf 1 ' (l-x^)" . 


T X* 

If HOW we make x — — , the same quantity may be written 
(16) , . , -; tx'dxiX-x^)' 1 —\'- 

Having thus the vahie of the repulsion due to the inner sphere B 
on an exterior point p, it remains to determine that due to the fluid 
on ^'s surface. But this last is represented by 

, 2 7raP^ (L_ {a-\-rf-''-{a-rf- 

' l — 7i.S — ndr r 

{Mec. Cel. Liv. ii. No. 12.) Now by expanding this function there re- 

1~ ■ r "*" 4.5 ^ "*" 

^TrP'a'-'r. ^^ ■ U + "T l^ 2'- + """,T^":" ' " .g^+^cl 

. r./ i» 2-w .^, ra.w + l.?« + 2 /i + 2*-l, ^.r" 

= 4.PV-V.-g-.2„ ,.5.6 ,,^3 (^-^1)^- 

The last of these expressions may readily be exhibited vinder a finite 
form, by remarking that 

flx"dx{l-a^) ^ (l - -^) ^ =/lx''dx(l-x') " S ^^ g ^^^ .-^ 

/ 2^ + w + l \ / 4-/A 

«.?f + 2.;? + 4 w + 2jf-2 y^' V 2 / V 2 / 

~ " 2. 4 .6 2s 'a^'' ^^l2s + 5\ 

2 r 

r {lzl\ r fl+^^ 

V 2 / \ 2 I 2-n ^^ n.n + l.n + 2 n + 2s-l , ,.r^' 


3 ' " 4. 5 . 6 2* + 3 ' 'a 

Hence, since r (^) = v^tt, the value of the repulsion arising from ^'s 
surface becomes 

F 2 


Now by adding the repulsion due to the inner sphere which is given 
by the formula (16), we obtain, (since it is evidently indifferent what 
variable enters into a definite integral, provided each of its limits re- 
main unchanged) 

f'afdxCl -x") '' .{1 , , 

1 + w\ / 2-w \ •'^ ^ ' V (f ) 

\ 2 j ' 

for the value of the total repulsion upon a particle p of positive fluid 
situate within the sphere A and exterior to S. We thus see that 
when P' is positive the particle p is always impelled by a force whicli 
is equal to zero at JS's surface, and which continually increases as p 
recedes farther from it. Hence, if any particle of positive fluid is 
separated ever so little from 2?'s surface, it has no tendency to return 
there, but on the contrary, it is continually impelled therefrom by a 
regularly increasing force ; and consequently, as was before observed, 
the equilibrium can not be permanent until all the positive fluid has 
been gradually abstracted from B and carried to the surface of A, 
Avhere it is retained by the non-conducting medium with which the 
sphere A is conceived to be surrounded. 

Let now q represent the total quantity of fluid in the inner sphere, 
then the repulsion exerted on p by this will evidently be 


when r is supposed infinite. Making therefore r infinite in the expression 
(15), and equating the value thus obtained to the one just given, there 

q= — :: tclx-afil-x')'. 

2 J \ 2 

When the equilibrium has become permanent, q is equal to the total 
quantity of that kind of fluid, which we choose to consider negative, 
originally introduced into the sphere A ; and if now qi represent the 


total quantity of fluid of opposite name contained within A, we shall 
have, for the determination of the two unknown quantities P' and b, 
the equations 

5', = 4nra'.P', 

and ^ = — , "^""^ X" dxx" (1 - af)^, 

and hence we are enabled to assign accurately the manner in which the 
two fluids will distribute themselves in the interior of A; q and «/, , the 
quantities of the fluids of opposite names originally introduced into A 
being supposed given. 

9. In the two foregoing articles we have determined the manner 
in which our hypothetical fluids wiU distribute themselves in the interior 
of a conducting sphere A when in equilibrium and free from all exterior 
actions, but the method employed in the former is equally applicable 
when the sphere is under the influence of any exterior forces. In fact, 
if we conceive them all resolved into three JT, Y, Z, in the direction 
of the co-ordinates x, y, « of a point j9, and then make, as in Art. 1, 

r pdv 

we shall have, in consequence of the equilibrium, 

1 dr „ ^ \ dV ^ ^ \ dV „ 

0= -J— + X, = 5- + ^' = 7- + Z, 

1 — ndx \ — ndy 1 — ndz 

which, multiplied by dx, dy and d% respectively, and integrated, give 
const. = =-^ V + f{Xdx + Ydy + Zdz) ; 

X ^~ ft/ 

where Xdx + Ydy + Zd% is always an exact differential. 

We thus see that when X, Y, Z are given rational and entire functions 
V will be so likewise, and we may thence deduce (Art. 5.) 

p = (1 _ ;j;'^ _ y'2 _ «'^)f^ ./{a;', y', x'), 

where / is the characteristic of a rational and entire function of the Same 
degree as V. 


The preceding method is directly applicable when the forces X, Y, Z 
are given explicitly in functions of x, y, x. But instead of these forces, 
we may conceive the density of the fluid in the exterior bodies as given, 
and thence determine the state which its action will induce in the con- 
ducting sphere A. For example, we may in the first place suppose 
the radius of A to be taken as the unit of space, and an exterior con- 
centric spherical surface, of which the radius is a, to be covered with 
fluid of the density U"^'^: ZJ"'"* being a function of the two polar co- 
ordinates 6" and ■zsr" of any element of the spherical surface of the same 
kind as those considered by Laplace {Mec. Cel. Liv. iii.). Then it is 
easy to perceive by what has been proved in the article last cited, that 
the value of the induced density wiU be of the form 

p = [/-'Wr'' (1 - r"'y' .f{r") ; 

r', &, -ar' being the polar co-ordinates of the element dv, and £/'<'* what 
Z7"<'> becomes by changing Q", -sr" into 9', tst'. 

Still continuing to follow the methods before explained, (Art. 4. and 5.) 
we get in the present case 

f{af, y', «') = t7'<Vy(r'^) =/«, 
and by expanding y(r'^), we have 

/(r'^) = i?o + B,r" + B,i'' + B,r" + &c. 
Hence, /'" = B,U'^\ and 

' . ln-% \-"'^'''^ 2.4.6 2^-2^ ^ 2.4 2^' 


n-l.n + 1 w + 2? + 2if-3 

^3.5 2«-l-2«r + l ■ 

Then, by giving to t all the values 1, 2, 3, &c. of which it is sus- 
ceptible, and taking the sum of all the resulting quantities, we shall 
have, since in the present case V reduces itself to the single term V^\ 

sm (-^.) 

n-l.n + 1 n + 2i + 2t'-3 ^ 

^ S . 5 2i + 2f + l ' 

the sign S belonging to the unaccented letter t. 


If now V represents the function analagous to V and due to the 
fluid on the spherical surface, we shall obtain by what has been proved 
(Art. 3.) 

V = WK 27r«^ W'^ ^'7^ /-i(/M {i) (f-2artx-V a'i^', 

X • ^ ,Ot« • • • t 

{i) representing the same function as in the article just cited. 

Moreover, it is evident from the equation (10) Art. 4, that 

,. ,j ,., ,^ ^ , ,.^ ^,1.2.3 i ^n-l.n + 1 n + 2i + 2t'-S 

/ild^{t){r^-2ar^ + a^)^ = 2«'-" ^ 3 ^■_^ 2 ^ ^ ^i + ^ TTT 

n-2.n n + 2t'-4> IrV^^^' 

"" 2.4 2^ \a) ' 

and consequently, 

(.9) r'=cr<o.w-.. '';';''^';;;;;'-^f;^;:^7^ 

«-2.w ?« + 2if'-4 /r\' + '' 

2.4 2^' \a) 

the finite integrals extending from t' = Q to t'=<x). 

Substituting now for F' and ?^' their values in the equation of equi- 

(20) const. = r'+ f; 

we immediately obtain 

const. = i7".47rflr' ".2 

3 . 5 2i + 2t' + l 

n-2.?i w + 2^'- 4 //•\'+''' 

2 .4 2f 


^ 2^" rm 97? v,.<+2,' ^-1-^^ + 1 « + 2i + 2^'-3 

^ . fn-2 \ ' "3.5 2» + 2r-l 

sm(-^.) Ti-{-2t'-4> 4-W.6-W 2^-2^' + 2-« 

"^ 2 .4 2? ^ 2 . 4 2^-2^ 

the constant on the left side of this equation being equal to zero, except 
when i = 0. 


By equating separately the coefficients of the various powers of the 
indeterminate quantity r, we get the following system of equations : 

^ . (n-2 \ 

2 sm — — — TT . AC 

2 . 4 

^ • fn-2 \ 

, I o -n Tk 4 — w „ 4i — n.6 — n . 

TT 2 2.4 

^ . fn-2 \ 

TT 2 2.4 

&c &c &c. 

But it is evident from the form of these equations, that if we make 
generally ^,+i = a'^Bt, they will all be satisfied provided the first is, and 
as by this means the first equation becomes 

2 sm — -— TT , A AC 

Tr»/, 4 — « „ 4 — M.b — « . . \ 
«-»- = J?„ (l + -g-«- + 3 ^ a-^ + &c.) 

= J5o(l-«-'^) ^ = .Bo«*-"(a'-l) ^ . 
there arises 

„ . /M-2 


(n-2 \ 

^^ La-'-'{a'-l)' , B, = B,.a-\ 

Bo = -— -a-'-'ia'-l)' , B, = B,.a-\ B, = B,.a-\ !>ic. 



f{r") = B, + By' + B,r'' + &c. = ^„ (^1 + ^ + ^ + &c.) 

»8 2 sm I — — - TTJ ^_„ 

= ^„fl_ !l)-i = ^„a=(a^_/^)-'= L* ia'+'(a^-l)~(a^-r'^)-\ 

\ a I If 

and the required value of p becomes 

(21 ) (D = C7'< V; (1 - r'^)^/(r'0 


(a' - l)'^"a i7'» f^') '(«= - /=)-' (1 - r"y\ 


But whatever the density P on the inducing spherical surface may 
be, we can always expand it in a series of the form 

P = C7"<°>+ Z7"<'>+ C7"<^>+ Z7"® + &c. in inf. 

and the corresponding value of p by what precedes will be 

„ . /n-Q, 

P = - 

<a{a'-l) ' .{a'-r")-H^-r") 

X { t7'W+ [/'<')- + t7'<='^ + f7'<^>^ + &c. in inf.] ; 

Ijm^ U'M^ jjm^ &c. being what U"^'\ U"^'\ U"^% &c. become by changing 
d", w" into ff, Ts-', the polar co-ordinates of the element dv. But, since 
we have generally 

^d&'d-uy" sin 6)"PQ« = fdff'd^" sin 6" C7"<"Q» = ^^ C7<", 

{Mec. Cel. Liv. iii.) the preceding expression becomes 


p = _> a{a'-\) ' K-r'^)-'(l-0 ' jd&'d-sr" sin &'. 

2:(2e + l)PQ«^; 

the integrals being taken from 0" = O to 0" = 7r, and from ■bt" to ■sr" = 27r. 

In order to find the value of the finite integral entering into the 
preceding formula, let R represent the distance between the two ele- 
ments dff, dv ; then by expanding -^ in an ascending series of the powers 

of — we shall obtain 

— = ^ _ 2°°Q<*>.— -, 

B Va^ - 2ar' [cos 0' cos 0" + sine' sine" cos (-ar' -•23-")+/* ° *«" 

Mec. Cel. Liv. iii.). Hence we immediately deduce 

^ = .r«»e^, and .^4,^^ = K(^.>1)«?»^. 
Vol. V. Part I. G 

If now we substitute this in the value of p before given, and after- 

// o ft^ __ >«'2 

wards write — and „3 in the place of their equivalents, 

dd"dnr" smO", andvV'^'^, 

clr R 

we shall obtain 

. (n-2 

p- i7^ — («^-i)^ (i-O^ /-^; 

the integral relative to da being extended over the whole spherical sur- 

Lastly, if p^ represents the density of the reducing fluid disseminated 
over the space exterior to A, it is clear that we shall get the corres- 
ponding value of p by changing P into pida in the preceding expression, 
and then integrating the whole relative to a. Thus, 

, = - !iy4 (i-..)=i^/a-«.)*-?/**£i. 

But dada = dvx\ dvi being an element of the volume of the exterior 
space, and therefore we ultimately get 

fn — 2_ 

4— n 

. /n — 2\ 

(22) p= y5 -i^-r")'^ .fp^dv, ^ , 

where the last integral is supposed to extend over all the space exterior 
to the sphere and R, to represent the distance between the two elements 
dv and dv^. 

It is easy to perceive from what has before been shown (Art. 7.), that 
Ave may add to any of the preceding values of p, a term of the form 

h being an arbitrary constant quantity : for it is clear from the article 
just cited, that the only alteration which such an addition could produce 
would be to change the value of the constant on the left side of the 


general equation of equilibrium ; and as this constant is arbitrary, it is 
evident that the equilibrium will not be at all affected by the change 
in question. Moreover, it may be observed, that in general the additive 
term is necessary to enable us to assign the proper value of p, when 
Q, the quantity of redundant fluid originally introduced into the sphere, 
is given. 

In the foregoing expressions the radius of the sphere has been taken 
as the unit of space, but it is very easy thence to deduce formula^ 

adapted to any other unit, by recollecting that —, -p, j^ and y^^, 

are quantities of the dimensions 0, — 1, — 1 and S — n respectively with 
regard to space: for if h represents the sphere's radius, when we employ 

any other unit we shall only have to write, t> j, -j- > -jr- and j- in the 

place of r, r, R, dvi and a, and afterwards to multiply the resulting 
expressions by such powers of h, as will reduce each of them to their 
proper dimensions. 

If we here take the formula (22) of the present article as an example, 
there will result, 

• / W-Q ^ 4-n 

(23).... p= 1-|_-I(i"-/^)^ fp,dv^-^-^, 

for the value of the density which would be induced in a sphere A, 
whose radius is b, by the action of any exterior bodies whatever. 

When w > 2, the value of p or of the density of the free fluid here 
given offers no difficulties, but if » < 2, we shaU not be able strictly to 
realize it, for reasons before assigned (Art. 6. and 7.) If however n 
is positive, and we adopt the hypothesis of two fluids, supposing that 
the quantities of each contained by bodies in a natural state are ex- 
ceedingly great, we shall easily perceive by proceeding as in the last 
of the articles here cited, that the density given by the formula (23) 
will be sensibly correct except in the immediate vicinity of A's surface, 
provided we extend it to the surface of a sphere whose radius is 
h—^b only, and afterwards conceive the exterior shell entirely deprived 
of fluid: the surface of the conducting sphere itself having such a 

G 2 


quantity condensed upon it, that its density may every where be repre- 
sented by 

ftl 2 \ „-4 „_2 4-n 

Application of the general Methods to circular conducting Planes, &f:. 

10. Methods in every way similar to those which have been used 
for a sphere, are equally applicable to a circular plane as we shall im- 
mediately proceed to show, by endeavouring in the first place to determine 
the value of V when the density of the fluid on such a plane is of 

the form 

p = {\-ry.f{x',y'): 

f being the characteristic of a rational and entire function of the degree * ; 
x\ y' the rectangular co-ordinates of any element dcr of the plane's 
surface, and r', & the corresponding polar co-ordinates. 

Then we shall readily obtain the formula 

r= ff^ = r rrdr'd9'{l-ry.f{x',y') ^ . 
'' g"'' ''■^ {f^-Zrr' cos {9-9') + r"f^' 

where r, 9 are the polar co-ordinates of p, and the integrals are to be 
taken from 9' = to 0' = 27r, and from r' = to /•' = !; the radius of 
the circular plane being for greater simplicity considered as the unit 
of distance. 

Since the function /{x', y') is rational and entire of the degree j, 
we may always reduce it to the form 

(24) f{x', y') = A^°^ + A^'^ cos 9' + A^'^ cos 20' + ^*'' cos 39' + 

+ ^« sin 9' + B'-'^ sin 29' + B^'^ sin 30' + 
the coefficients A'-''\ A^'\ A^'\ &c. B^'\ B^% B^\ &c. being functions 
of r' only of a degree not exceeding *, and such that 

^('•'=«'o°' + «<V^ + 4"V'* + &c.; ^« = («?> + alV + 4'V'V)/; 
^(» = ( jw -I- J(/)r'^ + i(»r'^ + &c.) r' ; B'^ = {bf> + hfr" + &c.) r'\ 


We will now consider more particularly the part of V due to any 
of the terms in f as -<4^'^ cos i& for example. The value of this part 
will evidently be 

r r /dr'dff{\ - r"fA^^ cos iff 

{r'-^rr' cos (Q - ff) + r'')~^ ' 

the limits of the integrals being the same as before. But if we make 
6' = 9 + (p, there will result dff = d<p, and cos i9' = cos id cos e0 — sin iO sin «0, 
and hence the double integral here given by observing that the term 
multiplied sin i<p vanishes when the integration relative to (p is effected, 

cos ie/lA^'^r'dr' (1 - ry f^ ^"^ ^"^ ''I' —^ ; 

° {r'' — 2rr' cos (p + r"^)~^ 

If now we write F"/*^ for that portion of V which is due to the term 
«/*^r"+^* in the coefficient A^'^ we shall have 

r,» = «/'> . cos ieflr^^'^^'dt" (1 -ry /" "^"^ ^"^ "^ ^ . 

" {r^ — 2rr' cos (p + r"^)~^ 

But by well known methods we readily get 

•^'^ d(p cos i(}) 


{r' — ^rr' cos (f> + r'^) ^ 

i ^i-.-iv» ^"' n-l.n + 1 n+2t'-S n-l.n + 1 n+2i+2t'-3 

-2irr.r i„^„,. 2 _ ^ 2^ "" 2 . 4 2i + 2t ' 

when r'>r, and when /<r, the same expression will still be correct, 
provided we change r into r' and reciprocally. 

This value being substituted in that of Fj*'' we shaU readily have by 
following the processes before explained, (Art. 1. and 2.) 

F,w = 27ra/'V* cos 10 2o r ' — r 7 -—p 

P^^ . ,.„ ,S + 2t-2f~n\ 

»-l.» + l « + 2« + 2#'-3 "^f^ ' I \ 2 

X — ^:: -. 7;r- — :;^-r, X 

(/3 + i)r[^ 

2 . 4 2i + 2/' ( 2li + 5 + 2t-2 f) 

^^[ 2 J 


= TTtt/'V COS i6 . —5 — . 

, „P n-1 .n + 1 n + 2f — 3 7i-l . ji + 1 n + 2i + 2t'-3 

" 2.4 2t' ^ 2 . i 2i + 2f 

3-n.5-n 1 + 2^-2^'-?? ^ 

^ 2(i + 5-n 2fi + 3 + 2t+2t'-n' 

the sign of integration 2 belonging to the variable f. 

Having thus the part of V due to the term a,''' cos i9' in the expansion 

of J'iaf, if) it is clear that we may thence deduce the part due to the 

analogous term J/'^ sin i& by simply changing «/" cos iQ into J/'' sin iO, and 

consequently we shall have the total value of V itself, by taking the 

sum of the various parts due to all the different terms which enter 

into the complete expansion of y(a;', y'). 

11 3 

If now we make iS = — - — and recollect that 



the foregoing expression will undergo simplifications analogous to those 
before noticed (Art. 5.) Thus we shall obtain 

TT^a/" , .^^„„M-1.« + 1 n-k2t-3 

r/" = "-:^ r' cos iQ . 2 r''' 

' . (n 
sin I - 

n-1 \ 2.4 2t' 

sin — - — IT 


n-1 .n + 1 n + 2i + 2f-3 3-n. 5-n 1 + 2^-2^- w 

'^ 2 . 4 2i + 2t' ^2.4 2if-2^ 

or by writing for abridgment 

,. ^ n-1. n + 1 n + 2t'-3 n-l.n + 1 n + 2i + 2f-3 

"P^^'*^'- 2 . 4 2f "" 2 . 4 2i + 2t' ' 

there will result this particular value of /8 

,^(. ^«/'> . .^ „ ,„ 3-n. 5-n l + 2t-'2f-n ^,. „^ 

^' = . (n-1 . ^^osze.^r-^. ^^^ ^^_^^, .<p{t;t'), 

and afterwards by making 

ro = ro» + r/'> + r.» + r « + r« + &c. 

we shall have 


V^'^= , r* cos i6 into x 

/« — 1 

sin (-^vr) 


+ a<'>.?^.<^(e;0) + «?>.!. 0(«; 1) . r= 

+^'- 2.4.6 •<^(^;o) + «^"-2r-^— 0(^;i)-^ 

+ «^^^.0(«; 2).r' + af.l.cj>{i;3).f^ 

+ &C +&C +&C +&C 

Conceiving in the next place that F is a given rational and entire 
function of x, y, the rectangular co-ordinates of p, we shall have since 
X = r cos 0, y = r sin 0. 

{25) r= C<") + C('> cos 6 + C-'^ cos 20 + C('> cos 3 + &c. 

+ ^« sin + ^('' sin 29 + E^'^ sin 30 + &c. 

of which expansion any coefficient as C^'> for example, may be still 
farther developed in the form 

C« = ""'-^ {di\(p{i', 0) + c>{K(p{i; l).r'+4^.(p{i; 2).r' + ke.}. 

sm (-^ .J 

Now it is clear that the term C> cos iO in the developement (25) 
corresponds to that part of F which we have designated by F''', and 
hence by equating these two forms of the same quantity, we get 

F» = Cw cos ie, 


which by substituting for F"*'' and C^ their values before exhibited, and 
comparing like powers of the indeterminate quantity r gives 

/> , 1-^ 3 — n,. 3 — n.5 — n,i. 3 — n.5 — n.7 — n,., , 

&iC.— &c &c 

of which system the general type is 

C<'> = (1 - e)~ . «« ; 

the symbols of operation being here separated from those of quantity, 
and e being used in its ordinary acceptation with reference to the lower 
index u, so that we shall have generally 

f.m „(i) _ ^ (0 

The general equation between «!'' and cll^ being resolved, evidently gives 
by expanding the binomial and writing in the place of eci'', e''&i\ ^c'i\ &iC. 
their values c„*j\ , cj-i\, Cu%, &;c. 

(26) ««=(l-e)^c« = c<;>+^c„« + ''~^-''~^ 

2 — ' ■ 2 

... « — 3 . ra — 1 . w + 1 (i) , s 
%+ 2.4.6 "-- + ^"- 

Having thus the value of af we thence immediately deduce the value 
of ^<'' and this quantity being known, the first line of the expansion 
(25) evidently becomes known. 

In like manner when we suppose that the quantity J5^'> is expanded 
in a series of the form 

j5:« = — ^TTT ^^»"- *^ (^' ' 0) + ^'* "^ (^' ; 1) • ^' + ^^'* <^ («■ ; 2) . f^ + &c. ^ 



we shall readily deduce 

A«= (1 - ef^e^^^ + ^e.%+ ""'^'""l^ e.% + &c., 

and ii,^ being thus given, B'-'^ and consequently the second line of the 
expansion (25) are also given. 

From what has preceded, it is clear that when V is given equal to 
any rational and entire function whatever of x and y, the value of 
f{x', y') entering into the expression 

will immediately be determined by means of the most simple formulas. 

The preceding results being quite independent of the degree s of 
the function f(x', y) will be equally applicable when s is infinite, or 
wherever this function can be expanded in a series of the entire powers 
of x, y', and the various products of these powers. 

We will now endeavour to determine the manner in which one fluid 
will distribute itself on the circular conducting plane A when acted 
upon by fluid distributed in any way in its own plane. 

For this purpose, let us in the first place conceive a quantity q of 
fluid concentrated in a point P, where /• = « and 6 = 0, to act upon a 
conducting plate whose radius is unity. Then the value of V due to this 
fluid will evidently be 

g V' 

((^ — 9,ar cos Q + r^)~^ 

and consequently the equation of equilibrium analogous to the one marked 
(20) Art. 10., will be 

(27) const. = ^ ^+ F; 

(«'-2«rcos e + r^)~ 

V being due to the fluid on the conducting plate only. 

If now we expand the value of V deduced from this equation, and 
Vol. V. Part I. H 


then compare it with the forrnulag (25) of the present article, we 
shall have generally E^^ = 0, and 

C"'=-2ga-^.l^(r;O)+0(e;l)^+«^(/; 2) ^ +,^ («; 3) J + &c.^ 

except when i = 0, in which case we must take only half the quantity 
furnished by this expression in order to have the correct value of C*"'. 
Hence whatever u may be, 

2 sin I ^-Q— T I 
^ - 0, and cf = ^^ qa} -"-*-^" ; 


the particular value f=0 being excepted, for in this case we have agreeably 
to the preceding remark 






and then the only remaining exception is that due to the constant 
quantity on the left side of the equation (27). But it will be more 
simple to avoid considering this last exception here, and to afterwards add 
to the final result the term which arises from the constant quantity thus 

The equation (26) of the present article gives by substituting for 
d" its value just found. 

^ . (n-l \ 
2 sm I—- — tt) 

«»= l^f L qa'"-'-'". {1 + '^.a-' 

n — 3.n-l ^, n — 3.n-l.n-l ., . , 
2.4 2.4.6 ' 

^ • /«-! 
2 sm 

(n-1 \ 

l^i i qa'-'-'-"- (1 - a-')— 

2 sin I -—^ 


and consequently, 

sin (^.) 

. (n-\ 

2 sin I —7: /r I 3_^ ,2 ,4 


•* sin 1 ^ "I B-B „M 


the particular value -4'°^ being one half only of what would result from 
making i = in this general formulje. 

But 4'' = evidently gives £^''^=0, and therefore the expansion of 
f{a!, y') before given becomes 

fix', y') = J^'^ + A^'^ cos ff + A-'^ cos 20' + ^''^ cos 30' + &c. 

= 1-^ -g(«^-l) ^ («=-r'0-'.{| + - cos0'+ -cos20' + &c.| 

or by summing the series included between the braces, 

. (n — 1 \ 3-» 

JKx,y)- -^ ^a^-2«r'cos0' + r"' 

sin (^ .) 


iJ being the distance between P, the point in which the quantity of 
fluid q is concentrated, and that to which the density p is supposed to 

Having thus the value of /(a;', y') we thence deduce 

(n — 1 'N 3-, 

p = (1 - /»)-/(x', y') = - — i-| i (1 - /') ^ ? — ^. 

sin 1 — ::: — TT I „_3 (a^ — Yy^ 

H 2 


The value of p here given being expressed in quantities perfectly 
independent of the situation of the axis from which the angle 6' is 
measured, is evidently applicable when the point P is not situated upon 
this axis, and in order to have the complete value oi p, it will now 
only be requisite to add the term due to the arbitrary constant quantity 
on the left side of the equation (26), and as it is clear from what has pre- 
ceded, that the term in question is of the form 


const. X (1 - /') 2 , 
we shall therefore have generally, wherever P may be placed. 

P = (l-r-) 


- [n-l \ 3.„ 

const. - 

^- 7?-' 1 

The transition from this particular case to the more general one, 
originally proposed is almost immediate : for if p represents the density 
of the inducing fluid on any element dai of the plane coinciding with 
that of the plate, p^da-i will be the quantity of fluid contained in this 
element, and the density induced thereby will be had from the last 
formula, by changing q into pidai. If then we integrate the expression 
thus obtained, and extend the integral over all the fluid acting on the 
plate, we shall have for the required value of p 

p=(l-0^ .jconst. \f ^ fp^da ^" J^ }; 

B being the distance of the element dai from the point to which p belongs, 
and a the distance between da^ and the center of the conducting plate. 

Hitherto the radius of the circular plate has been taken as the unit 

of distance, but if we employ any other unit, and suppose that b is 

the measure of the same radius, in this case we shall only have to 

.^ a r' d(Tx , R . ,, , „ , , 

write ^ > ^ ' -^ and -g- m the place of a, r, da, and R respectively, 

recollecting that -^ is a quantity of the dimension with regard to space, 
by so doing the resulting value of jo is 



sin I -— — TT 


.p = {]r-r')' .|const. Jp.dcr,- — ^^] 

By supposing w = 2, the preceding investigation will be applicable 
to the electric fluid, and the value of the density induced upon an 
infinitely thin conducting plate by the action of a quantity of this 
fluid, distributed in any way at will in the plane of the plate itself 
will be immediately given. In fact, when n = 2, the foregoing value of 
p becomes 

1 7 , y/a'-b'] 

^ = 7ltptHst--^/^'^<^' 


If we suppose the plate free from all extraneous action, we shall 
simply have to make pi = in the preceding formula; and thus 

,„^, const. 

(29) p = 


Biot (Traite cle Physique, Tom. ii. p. 277.)> has related the results of 
some experiments made by Coulomb on the distribution of the electric fluid 
when in equilibrium upon a plate of copper 10 inches in diameter, but 
of which the thickness is not specified. If we conceive this thickness 
to be very small compared with the diameter of the plate, which was 
imdoubtedly the case, the formula just found ought to be applicable 
to it, provided we except those parts of the plate which are in the 
immediate vicinity of its exterior edge. As the comparison of any 
results mathematically deduced from the received theory of electricity 
with those of the experiments of so accurate an observer as Coulomb 
must always be interesting, we will here give a table of the values of 
the density at different points on the surface of the plate, calculated 
by means of the formula (29), together with the corresponding values 
found from experiment. 


Distances from the 
Plate's edge. 



5 in 















We thus see that the differences between the calculated and observed 
densities are trifling; and moreover, that the observed are all something 
smaller than the calculated ones, which it is evident ought to be the 
case, since the latter have been determined by considering the thickness 
of the plate as infinitely small, and consequently they will be somewhat 
greater than when this thickness is a finite quantity, as it necessarily 
was in Coulomb's experiments. 

It has already been remarked that the method given in the second 
article is applicable to any ellipsoid whatever, whose axes are a, h, c. 
In fact, if we suppose that x, y, % are the co-ordinates of a point p 
within it, and x', y', z' those of any element dv of its volume, and 
afterwards make 

X = a. cos 9, y — i.sin 6 cos w, a = c.sin 6 sin ■sr, 
x'= a. cos ff, y' = J. sin 9' cos w', z'= c.sin 9' sin w', 
we shall readily obtain by substitution. 

2 . 

■ V=abcf p. r'^dr'd&d-ar' si-n 9'. {Xr"- 2 nrr'-¥vr'^) 
the limits of the integrals being the same as before (Art. 2.), and 
\ = «^ cos 9^ + If sin 9^ cos ts^ -\- & sin 0- sin Tsr^, 

ft. = a^ cos 9 cos 9' + U sin 9 sin 9' cos tb- cos in-' ■\-<? sin 9 sin 9' sin •ar sin w', 
V = «' COS 9" + ¥ sin 9'^ cos sr'^ + e sin 9'^ sin ■ar'^ 


Under the present form it is clear the determination of V can offer 
no difficulties after what has been shown (Art. 2.). I shall not there- 
fore insist upon it here more particularly, as it is my intention in a 
future paper to give a general and purely analytical method of finding 
the value of V, whether p is situated within the ellipsoid or not. I 
shall therefore only observe, that for the particular value 

(30) , = ^\^-ii^-%-i]' = ^^(^ -'") ' ' 

the series Uo + U2' + U/ + &c. (Art. 2.) will reduce itself to the single 
term Uo, and we shall ultimately get 

2sin("— .) 

which is evidently a constant quantity. Hence it follows that the ex- 
pression (30) gives the value of p when the fluid is in equilibrium 
within the ellipsoid, and free from all extraneous action. Moreover, 
this value is subject, when n < 2, to modifications similar to those of 
the analagous value for the sphere (Art. 7.)- 


11. On Elimination between an Indefinite Number of Unknown Quantities. 
By the Rev. R. Muephy, M. A. Fellow of Cuius College, a?id of 
the Cambridge Philosophical Society. 

[Read Nov. 26, 1832.] 



Fourier, in his treatise, * Theorie de la Chaleur,' * has given an 
example of the determination of an indefinite number of unknown 
quantities, subject to the same immber of conditions. If n be the 
number of those quantities, in order to discover their law by this 
method, it will be necessary to eliminate successively the first (ni- 1) 
and the last (« — »^) unknown quantities, thus determining the »^'^ by 
a final equation containing that quantity only. 

This process is obviously too laborious, and the results too compli- 
cated, to be practically useful, in most cases. 

The same objection applies to the elegant method of Laplace, which 
makes the determination of one of the unknown quantities, depend 
on the discovery of all the (w — 1) arbitrary multipliers introduced in 
the process. It has besides the disadvantage of not seizing, in many 
cases, the facilities offered by the peculiar forms of the proposed equa- 

• Vid. Fourier, p. 1 69 to 174- 
Vol. V. Paet I. I 

G6 mk murphy, on elimination between an indefinite 

In the physical investigations, which conduct to an indefinite num- 
ber of equations, it is of great importance to discover the law of those 
quantities, corresponding to the law by which the given equations are 
connected. The method which I here propose for this object is founded 
on the two following principles. 

First, if we make the right-hand member of the a;*'' equation dis- 
appear by transposition, the left-hand member is then a function of x, 

which vanishes when x is any number of the series 1, 2, 3, w; and 

therefore it must be of the form 

P.(.r- 1) {x-2) (x-S) (x-n). 

Secondly, if an identity exist between two formulas which are 
partly integer, partly proper algebraic fractions (of which the numerators 
are of lower dimensions than the denominators) the integer and fractional 
parts are separately equal. 

To demonstrate this principle, let 

represent such an identity, where each symbol denotes an entire function 
of X, and the dimensions of P, P' are respectively lower than those of 
Q, Q'; then we have 

(N-N)QQ = PQ- PQ'. 

If therefore N—jV' be not identically nothing, we shall have the 
entire function, represented by the left-hand member, identical with one 
of lower dimensions ; but this is impossible, because in integer formulae 
we may equate like powers of x, hence we must have iV=iV' and, 
therefore also, 

Z! - ^ 

Q- Q' 

By means of this principle, we shall be able to expand a given 
entire function P, in terms of other given functions, whenever such an 
expansion is possible. 



Application of the First Principle. 

The first principle alone is sufficient, in a great number of instances, 
to resolve the proposed equations ; we shall illustrate its application by 
selecting three distinct classes of equations to be resolved. 

First, when the terms which compose the general or a;**" equation are 
proper fractions. 

Example : 

To find the values of the n unknown quantities ssi, %i, sss, s,, sub- 
ject to the n equations following, 

»1 «2 ^ »„ ^ _ 1 

3 4 "*■ 5 "^ » + 2 2' 

«i , ^ ^ , g« =_ 1 

4 "*" 5 6 "^ « + 3 3* 

» + l « + 2 ra + 3 2w »■ 

The general, or a;**" equation, when its right-hand member is trans- 
posed, becomes 

- + — Y — -^ \- H — =0. 

a; x+1 x + ^ x + n 

Suppose these fractions are actually added, and let -^ represent the 

sum; where D = x{x-\-\)(x-\-9l) {x + n) and A'' is some function of x 

of n dimensions. . ,- • .• 



Hence we have 7^ =0, and therefore iV=0, provided a; is any num- 
ber of the series 1, 2, 3,....ti and consequently iV (which is of ?i dimensions) 

has a factor (^ — 1) U — 2) (x — ti); and can therefore admit of no 

other factor, but a constant c. 

Hence we have in general, 

^^ X x + 1 x + 2 x+n x(x+l){x + 2) {x + 3)...{x+n)' 

Multiply this equation by x, and then put x = 0, hence c = ( — 1)\ 

Multiply the same by x + 1, and then put x = —I; 

, n n + 1 

hence ssi = - - . — - — . 

Similarly, multiply by x + 2, and put x= —2, 

_ n.{tt—l) {n + !)(« + 2) 
• • '^ ~ 1 . 2 • 1 . 2 ' 

and generally, if we multiply equation (a) by x + m, and then put 
x= -m, we get 

'"'^ ' 'I. 2.3 m ' 1.2 m ' 

It is clear from this example, that if the general or x^^ equation were 

a+bx^ a' + h'x ^ d'\¥x ^ «<"' + i<"'x ~ "' 

we should find the sum of the fractions composing the left-hand member 
to be 

c .{x—\){x — 2) {x — n) 

(a + hx) [a + h'x) («" + h"x) (a" + i^a;) ' 

then multiplying by n + bx and putting x= — j, we should find e, 

multiplying by a'+b'x and putting .r=-,,, we should find s,, 

&c &c &c 


In the example above taken, we have supposed that the number 
of equations and unknown quantities were the same, but if we supposed 
that following the same law as in that example, the number of equa- 
tions were n + m, then the numerator N which was shown to be of 
n dimensions, ought to vanish when x is any number of the series 

1, 2, 3 n + m; that is, the equation A^=0 has more roots than it has 

dimensions, which is impossible ; it is therefore equally impossible to 
satisfy all the given equations. 

On the other hand, if the number of the given equations was 
only n — m, then n would by the preceding reasoning have a factor 

{x — l){x — 2) {x-7i-{tn), 

and since it is of n dimensions, it must have another factor of m dimen- 
sions, as C {x - a^) {x — a-i) (x — a„). 

Hence - -\ — ^ H "* ^ ^- 

X x + 1 x + 2 x + n 

_ C(x—'l){x—2) .{x~ tl + m){x — ai) jx — a-^ { x~a,„) ^ 

■~ ''"'xT{x + l){x + 2) [x + n) ' 

following now the same steps as before, we find 

^^ g. . g^( ly "■('»-^) jn-m + i) 

«.(« — 1) {n-m + 1)' '' ' ' ai.a-i a,„ 

c(-l)".(l+ai)(l+a.> (Ifg.) ^ (l+aiXl+aa) (!+«„,) w n-m + l 

'~ (w-1)(m — 2) {n — m + 2) o, . a^ a,„ '1' 1 

^. ., , (2 + a,)(2+a,) {2+a^) n.{n + l) (ti-m+l) {n-m+ 2) 

Similarly, %■^= - ~ — • ., ^ • i ^ • 

The quantities a„ a.^ a„ are evidently arbitrary, and each of the 

required quantities », z-., he. x„_„, are here determined in such a manner, 
as to contain the m arbitrary constants. This is therefore the most, 
complete solution of the problem. 


Another useful observation may be made in this place ; if the function 
which represents the a;"' eqvxation were discontiiuious, i. e. if any of the 
equations, for instance the second, were 

3 + 5 ' 7 ~ ' 

2 2 2 

and consequently an exception to the general law expressed by the x^^ 

equation, we should have then N—0 when x=\,S, 4 w, also when 

A' = ^, but not when a; = 2, hence in this case, 

iV=c. (ar-i)(x-l) {x-S) (x-4) {x-n); 

after this the remainder of the process would be the same as before. 

We have been thus particular about the preceding example, as being 
well calculated to shew the spirit and advantages of the present method. 

The next class of equations, which may be solved by the first principle 
alone, consists of those in which the terms composing the a;*'' equation 
contain common factors ; for if we then assign to x such values as may 
successively cause such factors to vanish, the unknown quantities will 
be determined. 

Example : 

To find the values of asj, %2, %^ a, subject to the n equations 

following; viz. 

a:, + 1 ,2.S!2 + 1.2.3.S53+ + 1.2.3 «s!„= -1, 

2s!, + 2.3.a!2 + 2.3.4.X3+ + 2 . 3 . 4...(w + l)x„= -1, 

3a, + 3.4.&, + 3.4.5.«3 + + 3 . 4. 5...(w+2)a;„= -1, 

n8Sl + ?i(w + l)8:2+M(w + l)(w + 2)S83+ + W (w + 1) (« + 2)...2W2!„= -1. 

If we transpose the right-hand member of the above equations, the 
.r"" or general equation becomes 

\ + x%,+ X {x + l)%.,^ X {x-\-\){x + ^) .%; + +a;(d;+l)(a:+2)...2x.!£„ = 0. 


This equation is evidently of w dimensions with respect to x, and its 

roots by the first principle are 1, 2, 3 n; the left-hand member 

must therefore be identical with the product 

c{x-l\{x -2){x-S) (x-ti), 

whatever value may be assigned to x. 

i — lY 

Put therefore x = 0. Hence c = - — ^ ' , 

1.2.3 « 

X ^^ X.»..>........V[ ^^ -^ fif 

^-~" 2a— ,g ^ , 

n . (w — 1) 


_ _ «(w-l)(w-2) 

* ^ a:,-- 1222.32 

&c &c 

and generally, «,„ = ^^ ^ . 3 ^)^~ •(-!)• 

We may verify this result by observing, that if we substitute this 
quantity for 25,„ in the general or x^^ equation, then its left-hand member 

n . ( w-1) X . ( x + l) w.(w-l)(w-2) ^(£+_l)_(a; + 2) , ' 
^"■"'^'*" 1.2 •~T:2 1.2 . 3 • 1.2 7 3 ^*'''- 

This quantity is evidently the part which does not contain h in the 

f, , x{x + l) ,„ x{x + l)ix + 2) ,3,o„l f, n , n{n- l) 1 1 

or in (l-//)-'.(l-|)". 

it is therefore the coefficient of //" in the expansion of 

But this coefficient is manifestly when x is any positive integer, which 
evidently agrees with the proposed conditions. 

72 Mr murphy, on elimination BETWEEN AN INDEFINITE 

Another class of equations which may easily be resolved by the 
first principle, occurs when the x^^ equation is of n dimensions, and 
arranged according to the powers of some function of .r ; it is then 
merely necessary to expand 

c.a:{x — l)(x-Q) (x — n) 

according to the powers of that function ; and equate the coefficients 
of like powers in both cases. 


Ki + SSs + «3 + + SS, = - 1, 

2a!, + 2-S2 + S'xs + + 2»ss„ = - 1, 

Sz, + 3-S5, + 3^X3 + + 3'%„ = - 1, 

w^i + w'asa + ?r%3 + + n''z^ = — 1, 

to find «„ %2 

The general or «"" equation in this case, is 

1 + x»i + x"%i + + afz^ = 0, 

the roots of which equation are x = l, 2, 3 ;/. 

Hence, the left-hand member is identical with the product 

c.{x-l){x-2){x-3) (x-tt), 

or c(-iy{S„-xS„., + x'S,_,- (-l)".x"|, 

where S„ denotes the sum of the quantities 1, 2, 3 n when taken 

in products m and m together. 

Hence, by equating, we get 

c(-lY S =1- • r = \^y .. 

- c(-1)".aS'„_,= «,; .-. x,= -aS'_,; 
c( — 1)" .«>„_2= »2; .•. SS2=— 0,2; 
and generally £., = S,,, 

where S.„ denotes the sum of the reciprocals of the quantities of which 
a9„ represents the sum. 



Amplication of the Second Principle. 

To expand a given function of x as P, in terms of other given 

Qo, Q., Q. Qn, 

all being supposed of n dimensions in x. 

Let P=aoQo + «iQi + «2Q2+ +«nQ», 

where a^, ffi, Oa «» are constants to be determined. 

Divide all the functions by Qo, and let the corresponding quotients 
be respectively 

P', Qo, Q'l, Q.....Qn, 
and the remainders 

p', g^o, q\, q'i q\- 

Then by attending to the second principle, we have 

P' = «oQ'o + «lQ'l + (kQ2+ +«„Q'n, 

p' =aoq'o + aig-'i + «25''2 + +a„q'n, 

when we obviously have Q'o=l and §''0=0. 

Dividing the last equation by q'l and using a similar notation, we 
get in like manner 

P'=«.Q". + «2Q"2 + ««Q"„, 

p"= aiq"i + (hq"2 + anq\, 

where Q"i = l and q'\ = 0. 

Divide the equation last obtained by q"i, and we obtain 

P"' = a,Q"',+ +a„Q\, 

p"'==a»q"', + +«„^"„, 

in the latter of which equations the first term = and in the former 
it equals unity. 

Vol. V. Pakt I. K 


The systems of the first equations thus obtained may be written in 
an inverse order thus, 

&c. = &;c 

whence «„, «„_i, a„_2, &c. are successively known. 

We have supposed all the functions to be of n dimensions, for 
these necessarily comprise all of lower degrees. 

Example : 
To expand unity in terms of the functions 

af, {x-^hy, (a; + 2A)", (a; + «A)°. 

Put l=«o*" + «i(^ + ^0° + '''^(^ + 2^)" + + «„(a;+wA)"; dividing by 

a;", we get 

= ao + ffi +«2 + + «„, 

1= ai?'i + Oag-'s + + «»<?'», 

where we have o'„ = A. {waf"*.»i-l — , ^ ' hm^af-^+ \. 

^ 1.2 

Divide now by g'l and we obtain 

= ai + 2a2 + + wa», 

1= «2g'"2+ + a„g^'„, 

where in general g'"„ = A^ |— ^ — ^— ^ a;""* {m^ — m) + &c. > . 

This process is easily continued, and we obtain successively the 


= 1.2a2 + 2.3a3 + (»— 1). w«„, 

0= 1.2. 3 03+ {n- 2) (71-1) nan. 

and lastly, ^= 1.2.3 «a„ 


From these equations taken in the inverse order, we get 

^ 1 

"' ~ 1.2 .3 nh"' 

«„-!= - na, 


„ _ n(n-\) 

1 . 2 
&C.= &C 

Hence the required expansion is 

To apply this principle to equations, we may observe that when 
the general or a;"' equation is cleared of fractions and its right-hand 
member transposed, it is of the form 

-P+ XiXi +^2X2 + +i8„X„ = 0, 

where ssi, sss a, are the unknown quantities, and P, Xx, X^.... 

known functions of x. 

The left-hand member must, by the reasoning of the preceding 
Section, be divisible by (« — l)(x— 2) {x—n). 

Let Xi, Xi, &c. when divided by this quantity leave the re- 
mainders Q'l, Q'2, &c. and P, the remainder P', hence 

where all the functions are necessarily of less than n dimensions, the 
application of the process above described, would then determine the 
quantities ssi, asj, »„. 


Caics College, 
March 5, 1833. 


III. On the General Equation of Surfaces of the Second Degree. 
By Augustus De Morgan, of Trinity College. 

[Read Nov. 12, 1832.] 

The present investigations are a continuation of those upon lines 
of the second degree, published in Vol. IV. Part I. of these Transactions. 
I have omitted various algebraical developments, as unnecessary, and 
tending to swell this communication to a length more than proportional 
to its importance. 

As the theory of the reduction of oblique to rectangular co-ordinates 
is a very necessary part of what follows, I proceed first to give the 
equations which will be required under this head. Let x, y, %, be 
oblique, and x', if, a' rectangular co-ordinates to the same point, with 
a common origin. Let the angles made by the first system be 

A A A ^ 

y% = ?, %x = t), xy = ^, 

and let the rectangular and oblique co-ordinates be so related that 

COS xsd = a, COS yx' = /3, cos xyf = a', &c. ; 

whence the following equations: 

a/ = ax + fiy + yx, 

y' = a'x +, /3'y + y'z (1), 

S8' =a"x + fi"y + 7"i8; 

l = a'+a" + a'", COS ? = /37 + (i'y' + fi"y'\ 

l=l3f>+ fi'^+ fi'% COS t, =ya + y'a! + y"a" (2), 

1 = y + y^ + y% cos ^ = a/3 + a'/3' + a")8". 



Make the following abbreviations, to which, for facility of reference, 
are annexed those which will afterwards appear in treating the general 
equation of the surface, 

aaf + bif-ir cz" + 2ai/z + 2bzx + 2cxy + 2aa? + 2% + 2cs! +/ = (3), 

the co-ordinates of the center of which call X, Y, and Z. Throughout 
this paper, all subscript indices indicate the dimension of the quantity 
signified, in terms of the coefficients of (3) : 

p =/3'7"-/3"7', 
/ = /3"7-/37", 
p"-=^y' -d'y. 


^1 ff If 

q = y a—ya , 
q"= 7«' -7'«' 

r =a'/3"-a"/3', 
t" = a"(i-a(i".. 
r"=«/3' -a' 13. 


a^^= be — a', 
b^, = ca- b% 
c„ — ab — (?. 

tto = sin' I, 
b^= sin'*;., 
Co = sin^ ^. 


a^= 6 + c — 2acos^, 
b,= c + a — 2b cos rj, 
Cf = a + b — 2c cos ^, 

l^^ —bc-aa, I, = &cos^+ccos»7-« — acos^, \ = cos v cos ^— cos ^, 
m, = ca — bb, mj=: ccos^ + acos^—b -bcosrj, 7»o= cos^cosf — cos ^...(6), 
91^1 =■ ab— cc, n,=a cos j? + 6 cos ^—c~c cos ^, «„ = cos ? cos tj — cos ^. 

=»?„«o-ao 4-^ cos A 

= «o lo-boMo-T- COSri\...(7), 

.= /o««o-Co«o-T-cosg 




V, = aai + bb^-\-cCa + 2ala + 2bma + 2cnf, 

Vi = a„ + b^, + c„ + 2/,, cos '^■\-2m„ cos r\-\-<iLn„ cos ^ 

= */;C//-/,; ^«1 


Vz~abc^2(ibc — a<i — b¥ — c& \ =.c„a„—m,J-^b\ 

r; = a,,a' + J,,6^ + c,,c^ + 2/,,6c + 2»?,,ca + 2w,,«F (11), 

= m„n„-a„l„ -=ra 

■.nj„ -b„m„^b 

= l„m„-c„n„ ^c 

W=-^ +/= aX+ 6 F+cZ+/ 



From (4), we find by inspection that the following six quantities 
are severally equal: 

pa + qft + ry, pa + p'a' + p" a'\ 

p'a' + g^/3' + r'y', qfi + ^/3' + q"fi" (13), 

p"a" + q"(i" + r"7", ry + r'y' + r"y", 

and moreover, that any symmetrical interchanges of accents in the first 
three, or of letters in the second, give results severally equal to nothing. 
Such are joa' + g'iS' + ry, p li + p (i' + p" fi" , &c. Let the common value 
of the first six be T. We have then 

pa -{■ q& + ry = T, 

pa +ql3'+ry'=0 (14), 

pa" + qli" + ry" = 0. 

From which, by obvious multiplications and additions, looking at equa- 
tions (2), we have 

p +3' cos (^+r cos ri=Ta, 

pcos ^+q +r COS ^=T(i (15), 

p cos t] +q cos^+r = Ty. 

From either of which sets we deduce 

1^ ■¥<f -Vi^-^^qr cos + 2rp cos n-\-^pq cos^= T^ (16), 

and similar relations may be deduced between jo', g', r', and jt>", ^", r' ; 
T being the same throughout. 

Again, form the several quantities 

flo, /o, &c. or 1 - cos^ f, cos n cos ^- cos f, &c. 

from the second set of equations in (2), and make the results homo- 
geneous and symmetrical from the first set; for example, write for 
Oa and /o 

(7« + V«' + 7"«") («/3 + a')3' + a"/3") - \c?^oi^^cl'''\ (fiy+fi'y'+(i"y"), 


in which the factors equal to unity, and introduced for symmetry, 
have the brackets []. Develope these expressions, from which we 
obtain the following equations: 

fl„=p^+y^+jo"% l,= qr-^qr' + q"r", 

h^=q' + q'' + q"\ m,= rp^-t'p' + r"p" (17), 

Co = r^ + /" + r"*, n„=pq +p'q' + jo'Y'. 

These, added together, the three last having been respectively multi- 
plied by 2 cos I, 2 cos rj, 2 cos ^, give from (16) 

«o + *o+Co + 24 cos f + 2»?o cos »? + 2«„cos ^=3T\ 
The first side of which, developed from (5) and (6) gives 3 V^* whence 

T=y/Vo (18). 

If the process by which (17) w^as obtained from (2) be repeated 
upon (17), that is, if at,ha-lo, Wana—a^la, &c. be formed, we shall have 
equations of a similar form, substituting instead of p, p' &;c. such functions 
of them, as they themselves are of a, y3, &c., the first sides of the equations 
being from (7), ^o ^^ *^^ ^^^^ three, and V^ cos f, F^ cos ri, Vg cos ^, in 
the last three. These equations are such as would arise from sub- 
stituting in (2), 

^ ^ ,^ — instead of a y~ — and y-^ ^ for a and a", &c...(19), 

which are therefore the values of a, a', &c. in terms of p, q, &c. 

From (1), by means of (14) and (18), can be deduced the following : 


VT,y = qaf + q'y'+q"fi (20), 

-v/Fo as = r x' + / 2^' + r"%', 

and the equations of the axis of x', referred to the oblique axes 
X, y, and k, are any two of the three, 

qx-py — 0, ry — q% = 0, p%-rx=Q (21), 


The equations of the center, central line, or central plane, as the 
case may be, of the surface expressed by (3) are 

aX+'cT+bZ+a = 0, 

cX+bY+aZ+% = (22), 

bX+aY+cZ+c = 0, 

and in the two following sets of quantities, it will be found that the 
sum of the products made by taking a term from each in the same 
horizontal line is = F^ ; while if the terms be taken from different horizontal 
lines, it will be = 0. 




















aa„-\-cn„-\-bm„—Vz, an,, + cb^^+bl,, = 0, &c. 

Hence, if the three equations in (22) be independent of one another, 
the co-ordinates of the center are 

j^^ _ a,,a + n,J + m,,c ^ y^ _ n„a^bj)^l„c ^^ _ m^^a^-l^-\-c,fi 


The equation of the surface, referred to this center, and to axes 
parallel to the primitive axes, becomes, calling ^ {x, y, %) the first side 
of equation (3), 

aa? ^-bf +cz^ + 2ay% + 2bzx ■{■^'cxy+^{X, Y, Z) = ...;:.. (24), 

and by multiplying the three equations in (22) by X, Y, and Z respec- 
tively, and adding, we get 

0(X, F, Z) = aX+bY+7z+/=JV (25). 

When only two of the equations (22) are independent, there is a 
central line. The conditions of this case are, that the numerators and 
Vol. V. Paut I. L 


denominators in (23) must be severally equal to nothing; but if 
f^3 = 0, the equations in (10) shew that it is sufficient that one of 
the numerators should be equal to nothing; or that the conditions 
may be stated thus, 

r, = 0, a/o^, « + \/T,* + 'v/cIc = (26). 

When F'i = 0, F't is a perfect square, (10) and (11), its root being 

the second expression in (26). Hence W appears in the form - . From 

two of equations (22), substitute in (25) values of any two co-ordinates 
of the center in terms of the third; it will be found that the co- 
efficient of the third disappears under the conditions in (26), and that 
the resulting value of W, which we denote by W, may be expressed 
in either of the following ways: 

„^, b(^ — 2cab + a¥ , „ cb^ — 2acb + b<f ^ 
ab — & bc — tt 

^ _ ad'-^bac + ca' .^^. 

ac — V 

When no two of the equations (22) are independent, there is a 
central plane. The conditions of this case are, as appears from the 
equations, that a„, 6,,, c^,, /„, «»,,, «,,, must be severally = ; of which how- 
ever it is sufficient that any three should exist. We have moreover 


a : c '. b (28). 

From all which it appears that W is now in the form -. From 

one of the equations (22) substitute in (25) the value of one of the 
co-ordinates in terms of the other two; the coefficients of the last two 
will disappear, as before, and the different forms of the value of W, 
which we call W", will be 

W"^ - I +/= - J +/= - 7 +/• (29). 

By substituting W or W", when necessary, for W or <p {X, Y, Z) 
in (24) the equation of the surface will be obtained, referred to any 
point in its central line or plane. 


Let the equation of the surface, referred to the principal axes, be 
Aa;" + A'y" + A"z''+W=0 (30), 

which must be identical with (24) when the values of x', y', «', found in 
(1) are substituted. We must then have 

a==Aa' +A'a" +A"a"\ 

h = Ali' +A'(i"' +A"(i"\ 

c = Ay^ +A'y" +A"y'\ 


a = Al3y +A'fi'y' + A"li"y", 

h=-Aya +A'y'a -{■A"y"a', 

'^ = Aafi+A'a'^ +A"a"li", 

which equations are reduced to those in (2) by substituting unity for 
A, A', A", a, h, and c; and cos f, cos n, and cos X, for a, h, and c. Thus, 
whatever equation is deduced from these, we immediately find another, 
containing a, /3, &c. in the same way, by the last mentioned substitu- 
tion. Multiplying the first of these by p, the last by q, and the last 
but one by r; and adding, we obtain by the use of (14), 

pa + qc +rb =Aa\/Vo 

p +qcoS(^ + rcosr]= ay/V^ 

from which, and similar processes, we obtain 


p{A — a) + q{A cos ^—c) + r {A cos t} — b) = 0, 

p{A cos^-c) + q(A-h) + r(^cosf-a) = (33), 

p{Acc^ri — b) + q{Acosl^—a) + r{A — c) =0; 

which agree in form with (22), if a, J, and c be struck out, and A — a 
substituted for a, ^cos^ — a for a, &c. But 1^3 = is the result of (22), 
with the last terms erased ; that is, if in V^ the substitutions just men- 
tioned be made for a, a, &;c. the result developed and equated to zero 
wiU give the equation for determining A, A', and A". That equation is 

r,A'- r^A'+r,A-v,=o (34). 

L 8 


We also find from (33), for substitution in (21), 
-:-:-:: l^ — l^A + loA' : m^^ — m^A + m^A^ : n^, — n,A + n^A^ (35). 

The equation (34) must have all its roots possible. For from (31) 
it appears that A' and A" cannot be of the forms X + m V-l and X— m \/- 1, 
unless a' and a", /3' and /3", 7' and 7" are of the same form ; from which, 

{K + xV'^){o-(p\/'^) - («-x\/^)(0 + 0\/^T) 

is of the form k\/ — 1, it will follow that p, q, and r (4) must be of 
this form : which is inconsistent with (32), if we suppose V^ positive ; 
since it may be seen from (31), and will presently appear otherwise, 
that a is possible when A is possible. 

We might find equations of the third degree to determine jh q, &c. 
but it will be more convenient to express them in terms of A, &c., 
supposed to be found from (34). To do this, let a,,, a^, I,,, I,, &c. (5) 
and (6), be found in terms of A, a, he. by substituting the values of 
a, h, a, h, he. from (31). The results, after reduction, are 

a,,=A'A"jf +A"Ap" +AA'p"', a,=U'+'^")f +{A"+A)p" +{A+A')p"% 
h,^A'A"(f +A"Aq" +AA'q"\ b={A'+A")q' +{A"+A)q" +{A+A')q"% 
c„=A'A"f^ +A"Ar" +AA'r"', c=U'+^"V +U"+A)r" +{A+A')r"\ 


l„=A'A"qr+A"Aqr'+AA'q"r", 1,={A'+A")qr+{A"+A)q'r'+{A+A')q"r", 

m„=A'A"rp+A"Ar'p+AAy'p", m={A'+A")rp+{A"+A)rp'+{A+A')r"p", 

n,=A'A"pq+A"Ap'q'+AA'p"q", n=U'+^")Pq+i^"+-^)p'q'+i^+^')P'Y' 

which equations, with those marked (17), give the following values 
of p'', qr, he. 

a,-a^A + a,A' _ l„ - l,A + kA' 

^~ {A-A'){A-A")' ^ {A-A'){A-A"y 

" — ^i i~^t ^ +hoA ^ _ m,,-m,A + irigA^ . 

^'~ {A-A%A-A"y ''P~ {A-A'){A-A") ^^^' 

•i _ C// — g, ^ +CoA ^ _ n,, - n,A + n^A' 

^ ~ XA^^t^ - ^"') ' ^^~ {A- A) {A - A") ■ 


In which equations the letter p, q. A, &c. may be accented throughout 
singly or doubly, striking off three accents from any A which thus 
obtains three or more. 

By squaring the equations (15), writing V^ for 7", substituting 
the values just obtained for p^, qr, &c. and then multiplying the same 
equations together two and two, and making 

Li = *„Co + b^c,, - 2 IX, Z/2 = in,,n, + m^n,, - aj,, - a„k, 

Ni = «„*„ + «o*// - 2 n„n^, Ni = l,,m, + lotn^^ - c^n„ — c„«o, 

we get 

^_ F,-L,-{F,-aK)A + F,A' r,-M,-{F-br,)A+ F.A^ 

**" V,{A-A'){A-A") ' ^~ r,{A-A'){A-A") 

2 F-N,-{r,-cr:)A+V,A' ,„. 
^- K{A-A'){A-A") ^^^^' 

„ _ FiCos^ — Lz — jVi cos^—aF'o)A+ FgCosBA^ 
^'y~ F{A-A'){A-A") 

_ Fcos tj — Mi- (Fj cos tj -bFp) A + FqCos t/A^ 
'y"- F,{A-A'){A-A") 

^ F, cos t-'^2-{F, cos ^-c F„) A + F„ cos ^A' 
"^~ F^{A-A'){A-A") 

in which the letters may be singly or doubly accented as before, and 
from which the determination of the position of the principal diameters 
is made to depend directly upon the solution of (34). 

Let the surface whose equation is (3) be referred to another origin 
and other axes, and let the quantities corresponding . to those already 
given or deduced, which belong to the new origin or axes, be denoted 
by the same letters and accents enclosed in brackets [ ]. Thus the 
angles made by the new axes are [^1 [>;], and [^] ; the coefficients of 


the new equation are [a], [«], &c.; the functions of these coefficients 
already noticed are [«J, [/„], [F,], &c. Since the principal diameters 
of the surface are the same, from whatever equation they are derived, 

w r w'l 

that is, since — 'T ~ ~ rlT ' ^^' *^^ roots of (34) bear to those of [34] 

the proportion of W^ to [ W^ ; whence, \ being an indeterminate quan- 
tity, since one coefficient in (3) is indeterminate. 


LjO'^'k' M-^' 

These equations* correspond to the general relations (6), (7), and (9), 
given in my former paper, and from them may be deduced the pro- 
perties of systems of conjugate diameters, and the remarkable property 
of the reciprocal squares of three semi-diameters at right angles to one 

Let wT', V, and Z', be the co-ordinates of the second origin referred 
to the first, so that if the co-ordinates be changed, [y] and (p{JC', Y', Z') 
will be corresponding terms of two equations, the terms of which should 
be respectively proportional. Assume X, the indeterminate quantity 
above-mentioned, so that 

[/] = \4>{X', Y', Z') (40). 

and multiply together the first and last of (39), recollecting that 


= -r.-^f^ [^i = -[Fj^t/]. 

* These relations have been given by M. Cacchv, for the case of rectangular co- 
ordinates, in his " Lcfons sur les applications du Calcnl Infinitesimal d la Geometrie," Vol. i. 
p. 2441. The equation (34) of this paper, in as general a form, has also been given, since 
this was written, by Mr Lubbock, in the Philosophical Magazine. 



and we obtain 


Substitute from the last of (39) for [jfl, and develope <f>{X', V, Z'), 

removing the term which contains it to the left hand side; which 


r, + F, jaX" + br" + cZ'' + &c. &c.) 


answering to (8) in my former paper. 

We shall afterwards proceed to some applications of these general 
formulas, and now enquire into the several varieties of the equation (3), 
and the criteria for distinguishing between them. The following table, 
immediately to be explained, gives a synoptical view of the various 
caseSi inhcrfHt i->9.R :i;f<>1 '"\'r^hr' 5r .-^.v: 

When the Equations of the 
Center denote 


W changes its sign. 

^ negative, 

A point 

A Right Line. W 
substituted for W. 

A Plane. W" sub- 
stituted for W. 


Single Hyper- 




(W=o) Point. 

(W= oc) Elliptic Paraboloid. 

(W=0) Cone. 
(W=oc) Hyperbolic Para- 

(W'=0) Right line. 
(W"=oc) Parabolic Cylinder. 

(W'=0) Intersecting Planes. 
(W'=oc) Parabolic Cylinder. 

fW"=0 \ ^. 
^, Plane. 


Double Hyperboloid. 

Elliptic Cylinder. 
Hyperbolic Cylinder. 
Parallel Planes. 


Taking the first line of this table, and the signs of W, V^, V^, and 
V^, (on which, as will presently be shewn, the variety of the equation 
depends,) being such as to denote that the equation is impossible, a 
change of sign in W only will indicate the ellipsoid, the elliptic cylinder, 
or parallel planes, according as the centre is a point, a line, or a plane. 
When the sign changes, if W be then = 0, the variety of the equation 
belongs to a point, a right line, or a plane ; while if W be infinite, 
we have an elliptic paraboloid, a parabolic cylinder, or a plane. In 
using W, we mean its real value, W or W", when the primitive form 

of W becomes - . 

The following table, from which the preceding may be deduced, and 
which I proceed to establish, gives the signs of W, &c., and also of V^, 
&c., for the different cases. When p alone, or p and n occur on the 
same line, p may signify either sign, provided n stands for the other. 
Also when a sign is enclosed in brackets, it is a necessary consequence 
of what precedes it, and not an independent assumption. The num- 
bers over the headings are references to the equations. 

The last part of the table, including all the varieties under W= - , 

forms a similar synoptical table for the curves of the second degree. 
The following are the values of W, W", V^ and Fi, expressed in the 
notation of my former paper, the equation of the curve being 

ay* + hxy + ca^ + dy->rex +f= ; 

and the angle made by the axes being Q, 

.^, _ cd^ + ae^ — hde „ 

™,„ _ _ dr-^a£_ &-^^cf 
id ~~ 4c ' 

V, =- (b'-iac), 

Vi = a + c — h cos Q. 




•, C 




§ 1 










.O" 4) 
CO i^C 














^ go 


- E 



3, and the next 
y be = 0, 



5, and all which 
bstituted for 7, 





-.' S 



■s s 

-^ S 








^ u- 












•■6 'o 














c8 o 

*^ 'o 


} Straight lin 


1 = 


1 S 

2 CL( 








•1 .J 







a, a 





a, a 

a, a 



o 8 

^— , ' 

/— \ 

+ + 

+ 1 

+ 1 

+ I 


+ 1 




+ 1 


O ^ 

■ • — 

^ • 

- ■ y • 








" — 

O '^ 

Rh a 









o ■ 


Vol. V. Paet I. 



First, with regard to the coefficients K^, Vi, V^, V3 in equation 
(34) it appears from spherical trigonometry, that V^ is always positive 
when '(;, J/, and ^ are the sides of a spherical triangle; while from the 
possibility of the roots, as well as from the quantities themselves, we 
infer that if V3 is finite, Fj and Vi can never vanish at the same 
time, while if ^i = 0, and ^ = 0, Fj, must be negative. 

If we suppose TV finite, and the order of signs in (34) to be 

H (-- or + + + +, in which case all its roots are of one sign ; 

that is, if K2 be positive, and Vi and V3 of the same sign, the equa- 
tion (30) shews that the surface is impossible or an ellipsoid, according 
as W and F'a have the same or different signs. From (36) it appears 
that in this case, a^^, b,^, and c„ must be positive, whence a, h, and c have 
the same sign ; which conditions, together with that of V^ having the 
same sign as a, are equivalent to those given in the Table for the 
impossible case or the ellipsoid. If we examine independently into 
the conditions under which the aggregate of the first six terms of 
(24) always has the same sign, we shall find them to be that a^, b„, 
and c„ must be positive, and V3 must have the common sign of a, h, 
and c. And it is evident that the first three terms of (30) are the first 
six terms of (24) in a different form. It may be worth noticing, that 

these conditions are equivalent to supposing ,-- , -- — , —7=5= to be 

's/ he \/ca y/ab 

the cosines of the sides of a spherical triangle. When any other order 

of signs except the two already noticed, is found in (34), we shall have 

one positive root only, or one negative root only, according as V3 is 

positive or negative ; that is to say, one possible axis, or a double 

hyperboloid, when V^ and W have contrary signs ; and one impossible 

axis or a single hyperboloid, when they have the same signs. 

When W—0, V^ being finite, equation (30) represents a point, or 
a cone; the first when all the roots of (34) have the same sign, the 
second in any other case. When V3 = 0, Vi being finite, or W infinite, 
the center is at an infinite distance, and the equation belongs to an 
elliptic or hyperbolic paraboloid, according as V^ is positive or negative. 
Since when V3 = 0, «,,, 5„, and c,, have the same sign, (10), which is 


also the sign of V^, a,^ may be substituted for Vi. In this case, (10) 
and (9), V2 has the form 

P+ Q + 2? + 2\/QK" cos ?+2\/;BP cos .7+2 ^/PQ cos ^, 

which, when P, Q, and R have the same sign, is always of that sign; 
and therefore can only be = when P, Q, and B are severally = 0. 

When ^"3=0, and F'i = 0, in which case W appears in the form -, 

and its real value is W (27), the simplest criteria of which are ex- 
pressed in (26) the equations (30) and (34) assume the forms 

Aaf' + A'y"+Jr'=0 (42), 

KA'- r,A + r,=o (43), 

the first of which, if V^ be positive, and F", and W of the same sign, 
is impossible, and belongs to an elliptic cylinder if V^ be positive, 
and Fi and W of different signs. As before, we may substitute a„ 
for Vi. If V2 or a^i be negative, (42) belongs to an hyperbolic cylinder : 
and if V2 — O, in which case a^^ = 0, h,i = 0, and c^^ = and W is infinite, 
we have a parabolic cylinder. It appears therefore, that any surface of 
the second order, which has three parabolic sections, not having a 
common line of intersection, is a parabolic cylinder. The central line 
of this surface is at an infinite distance. When W' = and V is 
positive, equation (42), considered as of two dimensions, represents 
only the origin, and therefore belongs to a straight line, the axis of 
iB'. When Fa is negative, W being =0, (42) is the equation of two 
planes intersecting at an angle whose tangent is 

2^/-AA' 2V-F,r, 

A + A' ' °^ r, 

When the equations of the center belong to a plane, and W as 
well as W appears in the form -, the real value of W is W", given 
in (29) and the simplest conditions are, as in (28), 

«// = *// = C// = 0, 

a : b : c '.: a : c : b. 

M 2 


The equations (42) and (43) take the forms 

Ax"+jr" = (44), 

F,A - F, =0 (45). 

The first of which is impossible if W and Fl have the same sign, 
that is, if W" and a have the same sign ; for when a„ = b,^ = c„ = 0, 
T^i takes the same form with respect to a, b, and c which Vs took with 
respect to a„, b^,, and c„ in the last case. When a and W" have 
different signs (44) belongs to two parallel planes, which coincide in 
one where W" = 0. That is (29) the surface is impossible, two parallel 
planes, or one plane, according as af—c^ is positive, negative, or nothing. 
When W becomes infinite, or a = 0, in which case b, c, a, b, and c are 
severally = 0, the proposed equation (3) is in fact of the first degree. 

Though oblique co-ordinates have hitherto been used, yet they 
might have been dispensed with so far as the criteria of distinction 
between the different classes of surfaces are concerned. It would take 
some space, and complicated algebraical operations, to prove this in 
all the individual cases, but the following general consideration is equally 
conclusive. So long as we only consider those distinctions which are 
implied in calling the surface bounded or unbounded, of one sheet or 
of two sheets, &c. in which no numerical relations of lines, &c. appear, 
it is evident that any equation will preserve the same character, how- 
ever the axes on which its results are measured are inclined to one 
another. That is, when the sign of a quantity is alleged to be a cri- 
terion of distinction, it cannot stand as such, if by any alteration of 
I, >/, or ^, consistent with V^ remaining positive, the sign of that quantity 
can be changed. Again, if the signs of two out of the three, a, b, and c 
be changed, as well as that of the third letter in a, b, and c, (those of 
a, b, and c, for example) it is evident that the surface remains the same 
in form and magnitude, those parts which were below one of the co- 
ordinate planes being transferred above it, and vice versa. That is, 
it is impossible that any aggregate of terms of an odd degree, with respect 
to a, b, and c, b, c, and a, or c, a, and b, can affect the sign of any 


of the criteria. If we look at F'l, V^, Fs, and F^, we find that those 
terms, and those terms only, which are multiplied by cosines of f, &c., 
are of the first or third degree, with respect to any of the three sets 
just mentioned. 

The case is very much altered when we consider any numerical 
relation, however simple. For example, I give the condition which 
expresses a surface of revolution, or a surface two of whose axes are 
equal. If A and A' belong to the equal axes, a, a, &c. become in- 
determinate; hence the numerators of the six equations (38), will, when 
equated to zero, have a common root. Eliminate F - FA + FoA" from 
the values of a* and (3y, &c. in (38), which gives 

.-_ XaCOsf-Z^ MiCO^ri-D iVgCOS^-iV^ ,^ 

AFa = r"=- = -7 T~ - y - V4b), 

a cos ^— a cos t] — o ccos^— c 

which does not admit of any material simplification. There are evidently 
other ways of obtaining corresponding conditions from (38). I have 
chosen this because the corresponding formulae have been given in the 
case of rectangular co-ordinates. In this case, 

cos ^ = cos >; = cos ^ = 0, and Li = - l„, &c. 


^ _ ^/ _ ^/ 
a b c 

(See Mr Hamilton's Analytical Geometry, p. 323.) 

To apply the formulae (39) and (41), let there be two planes whose 
equations, separately considered, are 

\'x+ fi'y+ i/'a + l =0 J 

but which together must be one of the varieties of equation (3). Let 
new and rectangular axes be taken, the intersection of the planes 
being that of x. Their equation will then be 

[c] z" + 2[a]yss = 0, 













i.ei w 



-« lCk° 


C8 O 
S M 

« -a 




aj to 

fc: ^-i 

bi3 +^ 


g - ^ 

•^ 5 S 

£ ^ c 
g " § 

'S -^ ■=« 

o .in 


o Si 
e9 .iH H-i 








































§ I 

Sh '^ 

0; o 

^ s 

«4-i 5s 



'-J II w 

O F=f 

I — I 

I I 


I — 1 
I I 


I — I 

I I 

I — I 

I I 


r— 1 




















IV. On a Monstrosity of the Common Mignionette. By Rev, J. S. 
Henslow, M.A. Professor of JBotany in the University of Cambridge, 
and Secretary to the Cambridge Philosophical Society. 

[Read May 21, 1832.] 

Having met with a very interesting monstrosity of the common 
Mignionette {Reseda odorata,) in the course of last summer (1831), 
I made several drawings of the peculiarities which it exhibited. I beg 
to present the Society with a selection from these, as I think they may 
both serve to throw considerable light upon the true structure of the 
flowers of this genus, which is at present a matter of dispute among 
our most eminent Botanists, and also tend to illustrate the manner 
in which the reproductive organs of plants generally, may be con- 
sidered as resulting from a modification of the leaf. 

It is well known to every Botanist, that Professor Lindley has 
proposed a new and highly ingenious theory, in which he considers 
the flowers of a Reseda to be compounded of an aggregate of florets, 
very analogous to the inflorescence of a Euphorbia. Mr Brown, on 
the other hand, maintains the ordinary opinion of each flower being 
simple, and possessed of calyx, corolla, stamens, and pistil. I shall 
not here enter upon any examination of the arguments by which 
these gentlemen have supported their respective views, but will refer 
those who are desirous of seeing them to the " Introduction to the 
Natural System of Botany, by Prof. Lindley," and to the "Appendix 
to Major Denham's Narrative, by Mr Brown," My present object will 
be little more than to describe the several appearances figured in plates 
1 and 2. 


Fig. 1. is one of the slightest deviations that was noticed from the 
ordinary state of the flower. It consists in an elongation of the pistil (a), 
and a general enlargement of its parts, indicating a tendency in them to 
pass into leaves. This is accompanied by a slight diminution in the size 
of the central disk. The number of the sepals was either six or seven. 

Fig. 2. is a portion of the ovarium of the same flower opened, in 
which three of the ovules are somewhat distorted. 

Fig. 3. Here the three valves of the ovarium have assumed a dis- 
tinctly foliaceous character (a); the same has happened to some of the 
stamens {b), and to the petals (c) ; but the sepals are unaltered. The 
central disk has entirely disappeared. 

Fig. 4. This is a still closer approximation to the ordinary state of 
a proliferous flower bud, when developed. Those parts which would 
have formed the pistil, if the flower had been completed, are no longer 
distinguishable, and only a few of the stamens are to be seen, disguised 
in the form of foliaceous filaments crowned by distorted anthers (h). 

Fig. 5. A slight deviation in one of the petals from the usual 
character. The fleshy unguis is somewhat diminished, and the fimbriae 
are becoming green and leaf-like. These are aggregated into three 
distinct bundles, the middle one being composed of a single strap, 
and the two outer ones of five straps each, blended together at the 

Fig. 6. The line of demarcation between the unguis and the fimbriae 
has completely disappeared, and the number of the latter is considerably 
reduced. The whole is more green and leaf-like than fig. 5. 

Fig. 7. The fimbrige reduced to a single strap ; the position of the 
lateral bundles being indicated by slight projections only. Other in- 
stances occurred in which the petal appeared as a single undivided 
uniform green strap. 

Fig. 8. The two exterior whorls of a flower, consisting of seven 
regularly formed sepals, and eight petals. The latter deviate more or 
less from the forms represented in fig. 6 and 7. The whole of a green 
tint, and leaf-like. 

7>^nsactM>ns afthe Ceimi. I'Ail.Sor. VoLTTTf^J 

JSa-n,l,.u- ,1,1 f 


Transaetians of the C'affiiJ'ful.SrK.Til.KIt.Z. 

XlkKStmnfy X'"^ 


Figs. 9, 10. These are parts of one and the same flower dissected 
to shew the several whorls more distinctly. The whole has assumed 
a regular appearance, and is composed of seven sepals, alternating with 
seven green strap-shaped petals, which are succeeded by about twenty 
stamens without any fleshy disk ; the pistil is somewhat metamorphosed. 
This is perhaps the most remarkable deviation that was noticed from 
the ordinary state of the flower, and as several examples of it occurred, 
it is not likely that there is any error in this account of it. It appears 
to lead us in a very decided manner to the plan on which the flowers 
of the genus may be considered to be constructed, and to shew us 
that they are really simple and not compound. 

Fig. 11 to 15, represent the appearances assumed by some of the 
stamens, indicating various degrees of deviation from the perfect state 
towards a foliaceous structure. 

There were other circumstances, besides the appearances in figs. 9. 
and 10, which may lead us to conclude the structvire of the flowers 
of the genus to be simple and not compound. A compound flower 
arises from the development of several buds in the axillee of certain 
foliaceous appendages more or less degenerated from the character of 
leaves, and consequently these buds and the florets which they develop 
are always seated nearer to the axis than the foliaceous appendages 
themselves. If we suppose a raceme of the mignionette to degene- 
rate into the condition of a compound flower, we must allow for the 
abortion of the stem on which the several flowers are seated, so that 
these may become condensed into a capitulum, each floret of which 
will be accompanied by a bractea, more or less developed, at its base. 
Let us compare this supposition with the diagrams represented in 
figs. 16, 17, 18. 

Fig. 16. is an imaginary section of the flower in its ordinary state, 
(a) the pistil, (b) the stamens on the fleshy disk, (c) the petals, {d) the 
sepals alternating with them. 

Fig. 17. represents the position of the several buds (e) which com- 
pose the florets of the flower on the supposition of its being com- 
pound. Here it will be noticed that these buds alternate with the 

Vol. V. Part I. N 


sepals instead of being placed in their axils where we might rather 
expect to find them. 

Fig. 18. represents a fact which was observed in the present case, 
where some of the latent buds in the axils of the altered petals were 
partially developed. This development might perhaps be considered as 
indicating the construction of a compound flower, and those buds which 
in ordinary cases compose the outer and abortive florets, it might be 
said, are here manifesting themselves. But the axes of these buds lie 
nearer to the axis of the whole flower than the petals in whose axils 
they are developed; whereas it appears by fig. 17, that they ought to 
be further from it, since the centres of the five outer circles marked (e) 
would represent the axes of the several buds, whose partial develop- 
ment must be supposed to be on the side next the axis, if we allow 
any weight to the analogy between the position of the abortive 
stamens on the supposed calyx, and the fertile stamens on the central 

These figures are all that I have thought it necessary to give for 
the purpose of illustrating the structure of the flower; but as there 
were several interesting appearances noticed upon dissecting the pistil, 
I have selected some of them for the second plate, as they may 
possibly serve to throw some light upon the relationship which the 
several parts of the ovarium bear to the leaf, and to support the 
theory of their being all of them merely modifications of that im- 
portant organ. 

Fig. 19. is a pistil in which the three ovules have become foliaceous, 
and the central, or terminal bud of the flower-stalk is developing in 
the proliferous form represented in fig. 4. 

Fig. 20. The central bud is not developing ; but the three axillary 
buds in the bases of the transformed valves of the pistil are here 
assuming the form of branches on which one or two pair of leaves are 

Fig. 21. 22. unite the appearances in fig. 19 and 20, with the 
addition of a glandular body seated between the leaves at their 


junction. This apparently originates in the union of the two glandular 
stipules seated at the base of the leaves of this genus, and which 
may also be seen to accompany the scale-like leaves on the central 
bud within. 

Figs. 23. to 25. Interior views of metamorphosed pistils, in which 
the ovules are seen transformed to leaves, and the glandular stipules 
are all that remain of the leaves which should compose the central 
bud, their limbs having entirely disappeared. 

Fig. 26. The appearance of these stipules on a leaf-bud, develop- 
ing under ordinary circumstances. 

Fig. 27. One of them more highly magnified. 

Figs. 28. 29. Their appearance on the small scale-like leaves of the 
central buds in fig. 21, 22. 

Fig. 30. Similar to fig. 23, but without any appearance of the 
transformed ovules; the glandular stipules are seen in the bottom of 
the ovarium. 

These glandular bodies assume a very prominent character in the 
anatomy of the metamorphosed pistils, and I was for some time 
puzzled to account for them, thinking that they might represent an 
altered condition of the ovules. I believe however that I have rightly 
considered them as the only representatives of the various leaves which 
would have made their appearance on the branch if the bud had 
developed in the ordinary way. They do not appear to diminish in 
size though the limb of the leaf has disappeared. 

Fig. 31. Four pedicillated semitransformed ovules, seated on a pla- 
centa of a pistil metamorphosed similarly to that in fig. 9- 

Figs. 32. to 35. Other appearances of a similar kind, all representing 
various approaches of the ovules to a foliaceous character. The little 
theca-shaped appendages are hollow, with a perforation at their apex, 
representing the foramen. 


Fig. 36. One of these dissected, exhibiting a free clavate cellular 
body within, resembling the columella in the theca of a moss, and 
"here probably representing the nucleus of the ovule. 

Fig. 37. In this case the theca-shaped body was partially open 
exposing the included nucleus. 

Fig. 38. This nucleus more highly magnified. 

These appearances surely indicate a development of the investing 
coats of the nucleus into leaves ; but how far these developments 
might be extended, and whether the nucleus itself is capable of being 
further separated into a series of investing coats does not appear from 
these specimens. 





Vol. V. Part II. 








On the Calculation of Newton's Experiments on Diffraction. By 
George Biddell Airy, M.A. late Fellow of Trinity College, 
and Plumian Professor of Astronomy and Experimental Philosophy 
in the University of Cambridge. 

[Read May 7, 1833.] 

Since the publication of Fresnel's experiments on Diffraction, it has 
been usual to employ as the source of light, in all experiments of this 
class, the image of the Sun formed by a lens of short focal length. On 
the undulatory theory, the effect of light thus produced is precisely 
the same as if the minute image of the Sun were the real origin of 
the light diverging with equal intensity through a solid angle whose 
diameter is many degrees. The spherical or chromatic aberration of 
the lens produces no sensible effect in any of the common experiments, 
in all which the angle, made by rays which afterwards interfere, is small. 
In calculating experiments thus conducted we proceed therefore with 
full confidence that no consideration is left out of sight, the omission 
of which could cause sensible error. 

Newton's experiments however were conducted in a different way. 
His origin of light was a hole, from Jg^ to ^ of an inch in diameter, 
through which the Sun's light was made to pass. The effect of this 
light, on the undulatory theory, is not the same as if the bright hole 
were the origin of light. It becomes then a matter of some interest 
to examine mathematically what is the effect produced by transmitting 
the sun-beams directly through a hole of sensible size ; and whether this 
effect, in practice, will differ much from the effect produced by forming 
an image of the Sun with a lens of short focal length. 

The integrals which occur in this investigation are of such a kind 
that their values cannot be exhibited even in tables of numbers (except 
Vol. V. Part II. O 


of course in any particular case, when by very tedious summation nume- 
rical results might be obtained). The only thing which can be attempted 
is, to shew that the integrals are precisely the same as those that occur 
in a very different instance where Fresnel's method of experimenting 
is adopted. Even thus far however I have not succeeded except in one 
case, namely, where the hole is a rectangular parallelogram of any length, 
and where the diffracting aperture is also a rectangular parallelogram 
in a similar position ; including in this general case the particular instance 
in which one or both parallelograms have no boundary on one side. 

To consider, in the first place, a case similar to Newton's. A plane 
wave is supposed to enter an external parallelogram and then to pass 
through a slit with sides parallel to those of the parallelogram ; and the 
intensity of the light which falls upon a screen at a given distance is to 
be found. First, it is to be observed, that in estimating the comparative 
intensity of light in a direction parallel to one side of the parallelograms 
(suppose for instance the shorter) there is no necessity to take into ac- 
count the length of the parallelograms in the other direction ; as it will 
easily be seen, upon attempting an integration, that the intensity of light 
is expressed by the product of two quantities, of which one depends only 
on the lengths of the parallelograms and the position of the point of 
the screen in one dimension, and the other depends only on the breadth 
of the parallelograms and the position of the point of the screen in the 
other dimension. The intensity of light along a given line parallel 
to one side of the parallelogram will therefore, so far as it depends on 
the other side, be affected only with a constant multiplier. Neglecting 
therefore the lengths (by which term I designate that dimension of the 
parallelograms which is perpendicular to the line on which the comparative 
brightness is to be ascertained), suppose a normal to the front of the 
wave to be di-awn, and suppose the limits of the breadth of the external 
aperture measured from this line to be a, fi, (the distance of any point 
of the aperture being v), and suppose the limits of the breadth of the 
slit to be 7, 5, (the distance of any point of the slit being w)'. and 
suppose the distance of the point on the screen, whose illumination we 
wish to ascertain, to be x. Let the distance of the external aperture 
from the slit be a, and the distance of the slit from the screen h. Suppose 


the front of the wave where it enters the external aperture to be divided 
into a great number of small parts ^v ; and suppose each of these to 
be the origin of a small wave which diverges from it as a center. The 
distance from the point v of the aperture to the point w of the slit is 

^{a' + {v-wY]=a+ —(v-wy; 

and the disturbance produced at w by the small wave spreading from the 
space Sv at v will therefore be proportional to 

^tj.sin. — {vt- A — a— —-(v — wY]. 

Integrating this with respect to v, the coefficient of sin — {vt — A- a) 
will be 

L cos ~{v-wy, 

and the coefficient of cos —- (\t— A — a) will be 


The first of these integrals = X cos ^ iv \/ —r — w V -r-J '' 

and putting ^(s) for f. cos f- »M, this integral between the limits v = a, 
v = l3, will be proportional to 

<h\^\/ -- —W\/ -—] — d>\a \/ -— - W\/ ^\. 

(TV \ TT 

- xM , the integral - /„ sin -— (v — ivy 

between the same limits will be proportional to 

- ^ l*^ ^ - ■" ^^) + H° ^Fx - =" ^i)' 

o 2 



The whole displacement at the point w will therefore be 

sm — 

{yi-A-a)x{ (p(iB\/^-w\/^]-<p(a\/%-w\/-^]] 
' I ^V aX aXJ ^\ aX aXl j 

2ir f / 

+ cos —— (yt — A — a) X < - \l/ [j3 


— w 

aXJ ^ \ aX aXl ] 

Suppose now this displacement to be the origin of a small wave 
which diverges from it as a center. The distance of the point w of 
the slit from the point x of the screen is 

^{¥ + {w-xY]=h^^{w-x)\ 

and this distance must be added to ^ + a in the expressions 

sin -^{yt-A-a) and cos ~ {vt— A-a), 
X X 

in order to find an expression proportional to the displacement produced 
by it on the screen at the point x. The expression must also be mul- 
tiplied by Sw, the breadth of the small space from which the wave 
proceeds. Thus we find for the whole displacement at the point x of 
the screen. 


-.)..{ ^(/5V^-«,v'|;)-?.(«\/|;-».V|;)}' 

COS J— {w 
■{vt-A-a-b)y. { — — 

+COS -:^{vt- A -a-b)x j ■■ 

si„i(„-.)..{-*(^ v^-. V„4).«(<.v/|;-. V|)}j 

.co.i (.-.)'x{-V.(/3 V|;-» Vi) .+(« V|;-«- X/|) Y 


where the integrals are to be taken between the limits w = y, w = S. The 
brightness at the point x of the screen will then be proportional to the 

sum of the squares of the coefficients of sin---(yi—A — a — h) an(f 
cos — (yt— A — a — b). 


To consider in the second place a case in which the illumination is 
produced in Fresnel's method. Let the distance from the origin of light 
to the aperture be a', and from the aperture to the screen V. Let a 
line be drawn from the origin of light perpendicular to the screen, and 
let the limits of the aperture measured from this line, in the same 
direction as the breadths of the parallelograms in Newton's case, be e and ^ 
(the general letter for the distance of any point in this direction being p), 
and let the limits in the direction perpendicular to this be rj + np, 6 + nj), 
where m is constant. (It is readily seen that this implies the figure to 
be rhomboidal, with two sides parallel to the length of the parallelograms 
in Newton's case.) Let q be the general letter for distance in this second 
direction : also let of and y' be the distances, in the directions of p and q, 
of a point on the screen from the same line. The distance from the 
origin of light to the point p, q, in the aperture is 

and the displacement there will therefore be proportional to 

The distance from the point p, q, in the aperture to the point or', y', on 
the screen, is 

and this must be added to 

A + a' + ^ + ^,, 

in the expression for the displacement, in order to find the displacement 
produced at the point x',y', of the screen by the wave diverging from 


the point p, q, of the aperture. For the effect of the wave spreading 
from the small rectangle whose sides are ip, ^q, we must multiply by 
^p, Sq. Hence we find that the quantity to be integrated is 

where, after integrating with respect to q, the limits of q must be ex- 
pressed in terms of p before the next integration. 

Puttmg A' + a' + b'+ ^ ^ = B', this expression becomes 

The first integral is 



-COS - {..-2? - ^— (p _ _^) J /^ sm {2 .-^r^ (? - ^) } 
/tt 2(«'+ft')/ «y^'l , TT / ./2{a'+b') ,^/ 2m:' y 

which between the limits q = r}-irnp, q — Q-^np, is proportional to 

^1 ^ «7yx ^ ^ i' («' + *') x^ «'*'x J 

The quantity proportional to .4 sin j^.i^lt*} L _ 4^) | will be ex- 
pressed in the same manner, putting >// in the place of 0. 

The whole displacement of ether at the point x', y', will therefore 
be found to be 




\ a'h' V a' + b'J I ^\ ^ a'b'X ^ ^ b'la' + b')\ 

sm~(yt-B')x f ^ 

+ sin; 

27r, . 
f-COS— (v^ 

-B')x\ < 


. TT a'+b' f a'x'Y r f . . /2 (a' + b') ,./ 2a' i 

+ COS; 

TT a'+b' 1^ a'x' Y r , L./2(«' + 6') ,./ 2^^ 

where the integrals are to be taken between the limits p = €, p = ^. 
The brightness at the point x', i/, of the screen will then be propor- 
tional to the sum of the squares of the coefficients of sin — (v#— ^') 


and cos -— {v t - B'). 


We have now to shew that, for a constant value of y', and a vari- 
able value of x', these expressions may be made similar to those ob- 
tained in the first case. For this purpose it will be necessary, first, to 
make the coefficients of the expressions under the integral sign equal: 
secondly, to make the limits of integration the same. 


rr,! /. , •! i- • •^ IT a' + b' a'x' 

1 he first consideration gives vis j— = - . — rrr ' x = -; — rt ; 

^ h\ \ a'b' af + b" 

"^aX ''^ «'6'X y ^ b'{a' + b')\' ^ a\ ^ a'b'X ' 

and the second consideration gives 7 = e; S = ^; whence 5 — 7 = ^-c. The 
first set of equations, reduced, are 

6 = ^' + 

1 1 *,OaA ^./l 'aA 

a\/ - = tj'S/ J- — y' \/ jn\ whence (/3 — a) v - = — >?; and w= — V- 

The purport of these equations, in common language, may be stated 
thus : 

If in Newton's method light pass through a rectangular hole whose 
horizontal breadth is /3 — a, and through a slit whose horizontal breadth 
is 5-7, at the distance a from the former, and fall finally on a screen 
at the distance b from the slit: 

And if in Fresnel's method light pass through a rhomboidal hole, 
with two vertical sides, at the distance a' from the Sun's image; and 
fall on a screen or eyepiece at the distance V from the hole, so that 
1 1__ 1 
a'^ b'~ b' 

And if the length of the vertical sides of the rhomboid be \/- x 


the horizontal breadth of the external hole in the first case (or /3 — a); 
and the horizontal breadth of the rhomboid be equal to the horizontal 
breadth of the slit in the first case (or 5 -7); and the tangent of the 

angle made by the sides of the rhomboid be \/ j, (the acute angle of 

the rhomboid being on the side where x is negative and y positive). 


Then the proportion of the intensities of light along the horizontal 
line in the first case will be the same as the proportion of the inten- 
sities of light along a horizontal line in the second case: the distance 

x' = x y. -r in the second case corresponding to the distance x in the first 

If in the first case the center of the hole is opposite to the center 
of the slit, the horizontal line in the second case must be drawn over 
the middle of the illumination on the screen. But if in the first case 
the center of the hole is not opposite to the center of the slit, but 
deviates in the direction which makes x positive, then the horizontal 
line in the second case must not be drawn over the middle of the 
illumination, but on that side on which y' is negative. In general, 
or when one side of either aperture in the first case is wanting, the 

may be used. 

When the inequality of the sides of the rhomboid is considerable, 
the form of the illumination is not very different from the illumination 
when the hole is parallelogrammic. The coloured bars will be a little 
inclined, so that those which for a parallelogram would be perpendi- 
cular to its longest sides, will approach towards the direction perpendi- 
cular to the longer diagonal of the rhomboid. Besides these, there is 
a faint brush of light projecting from each part which corresponds to 
an obtuse angle, and nearly in the direction of a line bisecting that 
angle produced. These general notions will assist the reader in judging 
what ought, theoretically, to be expected in the different circumstances 
of Newton's experiments. 

In Newton's experiments the external hole was in fact circular. 
What would be the effect of this form it is impossible (theoretically) 
to say: but judging from the insignificance of the effect produced by a 

Vol. V. Tart II. P 


rectangular hole, I am inclined to think that, when the apertures are 
centrally opposite, the same investigation will apply well to it. 

I may now without impropriety mention the circumstances which 
induced me to make this investigation. 

In Newton's Optics, Book iii. Observation 6, Newton describes in 
very striking language the effect of narrowing a slit on which the 
sun-light fell after having passed through a hole a quarter of an inch 
in diameter. He states that when the breadth of the slit was about 
—t\\ of an inch, the illumination on the screen was interrupted by 
a black shadow in the middle. It is certain, theoretically and prac- 
tically, that if the experiment had been made in Fresnel's method the 
center would be the brightest part. It seemed therefore worth while 
to ascertain, by the best kind of investigation that svich an un- 
manageable case admits of, whether the size of the external hole 
could account for the dark shadow. From consideration of the form 
of the illumination in the second case above, it appears certain that 
it could not. The only resource (which the dullness of the weather 
at that time denied me) was to repeat the experiment. This I have 
now done three separate times in the presence of as many different 
persons : I have used both parallelogrammic and circular holes of dif- 
ferent sizes (the largest circular hole being ^inch in diameter) and 
have sometimes diminished the aperture to as little as j^ inch (by 
estimation). The distances have been 30 inches each, which appear 
to have been the distances in Newton's experiments. In every in- 
stance the center has been bright. I can account for this inaccuracy 
in Newton's observation only by supposing that his eye was in such 
a state as not to recover from the sudden impression which is pro- 
duced by rapidly diminishing the central light on the screen (which 
makes it for an instant appear black), and by referring to his candid 
avowal in the Advertisement, that " the third book and the last pro- 
" position of the second were put together out of scattered papers," 
and that " The subject of the third book I have also left imperfect, 
" not having tried all the experiments which I intended when I was 


"about these matters, nor repeated some of those which I did try 
"until I had satisfied myself about all their circumstances." I may 
add that Newton's measures of the distances at which the first dark 
bar was formed are so irreconcileable with those of his admirer Biot 
that, referring to the avowal above-cited, I think no reliance ought to 
be placed on the accuracy of his observations of diffraction. 

Since writing the above, I find that Biot has repeated the experi- 
ment with the same result which I have obtained {Traite cle Physique, 
Tom. IV. p. 749). He has not commented on or even mentioned 
Newton's observation. 

G. B. AIRY. 


May 6, 1833. 

p 2 

VI. Second Memoir on the Inverse Method of Definite Integrals. 
By the Rev. R. Muuphy, M.A. Fellow of Cuius College, and of 
the Cambridge Philosophical Society. 

[Read Nov. 11, 1833-3 


The object of my former Memoir on the present subject, pub- 
lished in the Fourth Volume of the Society's Transactions, was to 
investigate the principles by which we might revert from a function 
outside the sign of definite integration, to the function under that 
sign, whenever the latter belonged to any of those classes usually 
received in analysis. In that case the function outside the sign of 
integration possessed the characteristic property of converging to zero 
when a variable quantity x was made to increase indefinitely ; in the 
present Memoir I have endeavoured to complete this theory, by the 
research of the forms and properties of the functions under the sign 
of integration, when the characteristic above mentioned is not pos- 
sessed by the function resulting from integration : and as the subject 
increased in difficulty, those methods of analysis which possessed greater 
simplicity and uniformity have been most adhered to, in the follow- 
ing investigations. 

The fourth Section is devoted to the research of the nature and 
properties of the function under the sign of integration, when the 
integral always vanishes between the limits (0 and 1) of the indepen- 
dent variable which have been uniformly adopted in this as in the 
first Memoir. The class of functions thus investigated possess the re- 
markable property of vanishing an indefinitely great number of times 


in a finite extent; such functions correspond to an extended and 
curious class of pheenomena in nature, when any principles of action 
which have been observed, under peculiar circumstances cease to produce 
the observed effects, as when equal charges of opposite electricities 
are communicated to a body, or when a body electrised by influence 
is removed from the vicinity of the influencing system ; or lastly, as 
when heat in its thermometric effects disappears in the chemical 
changes which bodies undergo. 

The properties of this class of functions are of great use and 
importance in analysis, as they conduct directly to the theory of 
reciprocal functions. This term I have here employed to denote such 
functions, two of which being multiplied together the integral of the 
product vanishes, except in one particular case. That function which 
is in this sense reciprocal to another, is also in general different in its 
nature. There are however many functions which are reciprocal to 
functions of their own nature, and to this class belong the only two 
species of reciprocal functions hitherto introduced into analysis ; namely, 
the sines or cosines of the multiples of an angle, the integral of the 
product of which always vanishes (when taken between proper limits) 
except in the particular case of equimultiples; and secondly, such 
functions as satisfy the well-known partial differential equation in the 
third book of the Mecanique Celeste; where the integral of the product 
also vanishes except in the particular case where the functions are of 
the same order. It is this exception which renders reciprocal func- 
tions particularly useful, as is evident from the application of the 
trigonometrical functions in the theory of heat, and of Laplace's functions 
in investigations relative to the distribution of electricity. In the same 
Section I have shewn generally the means of discovering all species 
of reciprocal functions, and given several examples : as an instance of 
one of the most simple species possessing properties very analogous to 
those of Laplace's functions, but giving a simpler integral in the case 
where that integral does not vanish, it is proved in the succeeding 


Section that if T„ be the coefficient of h" in - — j , then when n and in 
are vmequal ftT„T„ = 0, but when n = vu ftT„T„ = l. 


The theory of reciprocal functions is applied in the fifth Section 
to the complete solution of the question, which was the object of this 
and the preceding JMemoir, namely, to revert from any function what- 
ever to that under the sign of definite integration, those reciprocal 
functions being employed which are most convenient in each particular 

The last application in this Memoir of the theory of reciprocal 
functions, is to the development of given functions of x in descending 
powers or other forms which vanish when x is infinitely great; the 
results of which may be applied to the valuation of functions of 
very great numbers, and to a great variety of physical problems. 
These series have also the peculiarity, generally, to terminate for the 
functions of integer numbers. 



Inverse Method for Definite Integrals which vanish; and Theory of 

Reciprocal Functions. 

1. When the equation fif{t).t' = (p{x) is supposed to be restricted 
to particular values of x, then whatever may be the form of (p {x), 

J'{t) may always be determined ; the values to which x is restricted 

we shall suppose to be the natural numbers 0, 1, 2, 3 (w — 1), and 

the method here pursued will also apply if the values of n should be 
different from those mentioned. 

2. * First, let f,f{t).t' = 0, the limits of t being always and 1, 
and let us seek for f{t) a rational function of t of the lowest possible 
dimensions, which shall satisfy this equation when x is any integer from 
to n — 1 inclusive. 

Any value of f{t) which answers the proposed conditions may be 
divided by the absolute term, and the quotient, it is evident, will 
equally fulfil those conditions; we may therefore take the first or 
absolute term in f{f) to be unity, and as the conditions to be satisfied 
are w in number, we must have n coefficients in f{t), which will hence 
be a rational function of the form 

1 + Alt + A,f + + Ant"; 

and therefore (j>{x) = + — -^ + — ^ + + "—-^, 


or = T^r by actual addition, 

putting Q for (a; + !)(« + 2), (x + w + 1), and P representing a function 

oi X oi n dimensions. 

Hence P=0, provided x be any number of the series 0, 1, 2....(/i — 1); 
these are therefore all the roots of that equation, P being of n dimensions ; 
hence we must have 

P = c.x.{x-l){x-2) {x-n + \)\ 

c representing a constant quantity. 

* I have resolved this question in a different manner in the " Treatise on Electricity." 


We have thus 

1 Ai A2 An c,x.{x-\) ....(ar-w + l) 

« + l "^ x + ^ "^ x + ^6 "^ x + n + 1 ~ {x ■\-\) .{x + ^)....{x -{-n + 1)' 

Multiply by x + 1, and then put x= —\\ hence c = ( — 1)", 

by ar + a, a:= — 2; -(4i= — - . — — , 

by . + 3, .= -3; 4=-"-^-^.^^^±ii^>; 

&c &c. 

1- j^/js -. « w + 1 ^ «.(«-l) (?i + l).(w + 2) .„ J 
hence /(0 = 1- j •-]-• ^+ ^T^-^-^ H^ ^./^-&c. 

dt" 1.2.3....W 

3. Denoting by P„ the value of f{t) which has been investigated 
in the preceding article, it possesses the remarkable property ; that 
ftP„P„ = 0, except when m — n, and then 

r p jj __ ^ . 

•'' '" "~2w + l' 

the limits being always and 1. 

For when m and n are unequal, one of them as n is the greater, 
P„ contains then only powers of t inferior to n, the integral of each 
of which vanishes by the natvire of P„. 

When m = n, the last term of P„, namely 

(w + l)(w + 2)....2w , 
1 . 2 ....n ^ ^'' 

is the only term of which, when multiplied by P„, the integral does 

• This value of _/(<) lias been shewn in the " Treatise on Electricity " to be the coefficient 
of /j' in {1-2//. (1-2/)+/*^}-^. 

Vol. V. Part II. Q 


not vanish ; and since in general 

/■p.. _/_ix. ^-(^-1) ....{x-n + 1) 
■'' " " ^ '^^ •{x+l){x + 2)....{x + n + l)' 

it is evident that in this case ftPj 

2n + l ' 

4. To illustrate the observation in Art. 1, with respect to the 
generality of this method, let it now be required, to find a rational function 
of t, as f{t), of the lowest possible dimensions, to satisfy the equation 
fif{t).t' = 0, when x is any number of the series 

p, p + 1, p + 2, .p + n-l. 

Putting as before /(^) = \ + A,t + A-J'- + + Aj\ we have 

/• f(f\ ft _ 1 , ^' , -^2 , -^- 

■"•^^^' x + 1 x-^2 x + S x + n + 1' 

the sum of all which fractions must by the reasoning of Art. 2, be 

c . jx—p) {x—p—l)....{x — p — ti + 1) _ 

{x + l){x + 2){x + 3)....{x+p + x) ' 

and determining c, Ai, Ai, &c. in the same manner as in the Article 
referred to, we have 


c = (-l)». 

(j9 + l).(jO + 2)....(jO + W)' 

. _ n n+p+1 
"^'~~1- p^\ ' 

. _ n.{n-\) (»+/? + l).(«+jP + 2) 
'~ 1.2 • (^ + l).(;j + 2) ' 

&c &c. 

and therefore 

'^ ' 1 ^ + 1 1.2 (jo + l).(jo + 2) 

t^ d^ j /t _ ^ , n.{n-l) ^ „ 1 

~ip + l).{p + 2)....(p + n)-dt''-y -V 1-^+ 1.2 .^-&c.|; 


or, putting l — t=^t', we obtain 

J^' (jo + l).(jo + 2)....(jp + «)" r/^" 

5. From this result it follows that if we put 

then shall ftj'{t).t'' = 0, provided x is any number of the series 

0, 1, 2 (n-l); 

Op representing any constant quantity. 

Now OpfP may be taken for the general term of an arbitrary function J^; 
hence the most general function which satisfies the equation ftf{t)-t'' — 0, 
is expressed by 

.,,. _ d^jtH'T) 

In fact we have (supposing the integrals to commence from ^ = 0,) 

f,f{t) . r = t^f, it) - xt^-'f, {f)\X.{x- 1) ./s it), &C. 

representing by fn {t) the ri^ successive integral oi fit), and putting for x 
0, 1, 2....(w — 1) successively, it follows that 

Mt) = 0, f,{t) = .f„{t) = 0, when ^=1; 

that is,Jn{t) and its n differential coefficients vanish when t = and when 
t=l; therefore y^ (/) contains a factor of the form ^".(1 — ^)", and con- 
sequently the most general form of f{t) is 

dt" ■ 

6. Hence we deduce the following general property: '' If /{t) he 
any function which satisfies the equation [tf{t) . t* = 0, a; being any integer 

from to in — V) inclusive, then the equation f{f) = will always have n real 
roots lying between and 1." 

For the equation r.^'°F=0 has n roots t = and n roots ^=1; and 
therefore f{t) which is the n^^ derived equation must have n roots be- 
tween and 1. 



Hence, if we suppose the equation J,f{t) .^ = to hold true for an 
indefinite number of entire values of x, the equation f{t) = will also 
have an indefinitely great number of roots all lying between and 1, 
and the curve, of which the ordinate is f{t), and the abscissa t, would 
intersect that portion of the axis of x, of which the length is unity 
measured from the origin in an indefinitely great number of points; 
thus we have a property characteristic of this class of functions.* 

7. We have supposed J'{t) to consist of terms involving the 
powers of t, but as we may proceed in like manner for any other 
assumed form, we take the following as an example, because it leads 
to some remarkable results. 

To find a rational function of h. 1. (f) as y(h. 1. t) of the lowest 
possible dimensions, which may satisfy the equation ftf(h.\.t).t' = 0, 
X being any integer from to n—1 inclusive. 

Put /(h. \.t) = \ + A, h.\.t+ A^ (h. 1. ff + + A„ (h. 1. t)', 

and observing that J,{h.\.{t)]"'.f = {-\f. •^^■^•:;\ , 

we get f,f{\,.l.t).t' = ^^-j^^^,+~^^^^- ± -(^^nyr.T-. 

and actually adding the fractions in the right-hand member of this 
equation, the numerator which is a function of n dimensions, ought 
to vanish when x is any number of the series 0, 1, 2...(w-l); that is, 

{x + 1)» -A,{x + !)"-■ + 1 . 2^2 (a; + 1)""' - 1 . 2 . 3 ^3 (a; + 1)""' 

= C.x.{x-\){x-^) {x-n->r\). 

Let Si represent the sum of the natural numbers 1, 2, 3.,..(«-l), n, 
Si the sum of their products two by two, 
^^3 the sum of their products three by three, &c. 

* Vide Art. (4) in my first Memoir on the Inverse Method of Definite Integrals. 



Then by the theory of equations, the right-hand member of this 
equation is equivalent to 

c {(x + 1)' - s,{x+iy-' + SA^+iy-' - SA^+i)"-', &c.| 

whence c = l, ^i = aS',, ^2=, „, ^3= J -^, &c. hence the required 
function is 

8. It has been proved, that the function thus obtained (which we 
shall denote by L„) in common with all others which possess the 
property that ftj'(t) .f = 0, when x is any integer from to n — 1 in- 
clusive, is of the form 

d\ {ft'" V) 
dt" ' 

to verify this in the present case, we must sum the preceding series 
which is represented by Z/„. 

First, by the nature of multiplication, we have 

hr + SJi^-' + S.h"-"- + +S„ = {h + \){h + ^) {h + n), 

and the development of an exponential gives 

i+7.h.l.(^+-A^ + + i.a.3..,:, +&c.=/-, 

the coefficient of h" in the product of both the latter series is iden- 
tical with that by which Z/„ is expressed. 

But since that product =^(A + 1) (/« + 2) (A + w) 


= ^{r(l+Ah.l.^4-^^l^^&C.)|, 

it follows that the coefficient of h" is also expressed by 

d" [f {h.\. ty\ 
1.2.3 ndf' 


this quantity is therefore the sum of the series which we proposed to 

Now the equation h. 1. {t) = is satisfied by ^ = 1 ; hence h. 1. t is 

i' t'^ 
of the form t'. Q, {where Q= — (1 + - + — + &;c.)}, and therefore if we 


put — ^ = J^, we get the value of L„ to be 


which was the formula we had required to verify. 

We may also observe that since in the equation L„ = 0, / must have 
n values lying between and 1, therefore h.l. {t), according to the powers 
of which L„ is arranged, must have n real negative roots, which we 
see confirmed by the positive signs of all the terms which compose L,,. 

9. If we form the equation 

u (1 — h h. 1. u) =t, 
we have by Lagrange's theorem 

. J..UW.N, ^' d{t\ ^ ¥ d'{t\\.\.tf , 
« = . + ;i.h.l.(0+^.-^^^— ^ -f-^-^3.-A^^+&c. 

from whence it appears that Li„ is the coefficient of h" in the value 
of -^. Similarly if in Article (12) we form the equation 

u \\ -h. (1 - u)] =/, 


we have P„ = the coefficient of h" in -rj . 

10. If Q„ i<? the coefficient of h" in -j-, supposing u to he deter- 
mined by the equation u{l — hU) = t, U bei)ig a function of u which 
vanishes when u = l, and T the same function oft, then shall 

j,Q„f ^ x.{x-\){x-2) {x-n + 1) , 

j/F'T 1.2.3 n '^ '' 


For if we put ti = in the equation u{l — hU) = t we get 1 = 0, 
and putting u = 1 we have by supposition U = and therefore t = I, 
hence the limits of u are the same as the limits of f. 

But j;Q„f = the coefficient of h" in f^^.f, 

^i U/t 

and l^^.if = JJ^ = !„u''{l-hUr 

expanding the part under the sign of integration, and taking the co- 
efficient of h" we obtain 

hHnt - 1.2.3 n -(-l)^^ ■^- 

11. If U he a rational and entire Junction of u which vanishes 
when u = \, and if Q„ be the term independent of u in the product 

U"- \\— —\ , then shall Q„ be itself a rational and entire function of 

t possessing the property of ftQj'^ = 0, x being any integer from to 
n—\ inclusive. 

For it has been proved in my former Memoir on the Resolution 
of Equations*, that the root of the rational equation <f){x) = is the 

coefficient of - in — h. 1. ^— , hence the value of u in the equation 

u(\ -hU)=t, is the coefficient of - in -h.l. j(l--] -hul, and 

differentiating, it follows that the value of -tt is the term independant 
of u in 


* Camb. Trans. Vol. iv. p. 131, 


because the u under the logarithmic sign is the same as if we had 

placed there, a or any arbitrary symbol, and is therefore treated as a 

constant in the differentiation; hence the coefficient of h" in ~t- is the 

term independant of m in 

(> - -:) 

n + i ' 

that is, its value is Q„, and therefore by the preceding Article /Q.r 
vanishes between the limits of x, and n — \, its general value being 

T being the same function of / that U is of u. 

By this theorem, every possible variety of rational and entire func- 
tions which possess the above-mentioned property may be found, as in 
the following 


To find a rational function of t, in which the powers of the variable 
are in arithmetical progression, such that jiQ,nt'=0 when x is any number 
of the series 0, 1, 2 {n — 1). 

In this instance put U = 1 — u"", m being any positive integer. 
Hence Q„ = term independent of u in 

/ t\ "*""*"'* 


^ « (w + l)(w + 2)...(w+m) w. (w-l) (M+l)(w+2)...(?i+2OT) ^,„_. 

1' 1.2...m ■ 1.2 ■ 1.2, ..2m 

in which if we take in particular m =1, we get the value of P„ before 
found in Art. (2). 

This formula for Q„ may be written in another form by which it 
will comprise the case where /w is a fraction, thus 

n (m+l)(m+2)...( m+n) ,„, w(«+l) (2»^+l)(2w^+2)...(2w^-^w) „ „ 

^-=^-i- r^:::^ -^ ^"ttt- t:2::ji -^ "*'''• 


and it is, moreover, evident that either of those values are identical with 

1.2...nde' ' 

which is included in the general form given in Art. 5. viz. 

d\ {ft'" V) 
dj" ■ 

12. 2'o find a rational and entire function of f of h dimensions, 
which if multiplied hy a rational and entire function of t' of less than n 
dimensions, the integral of the product may vanish between the limits t = 
and t=l. 

Let the required function be represented by (p, q),„ so that 

{p,q\^l + A,t^ + A,f-f + A,,f^, 

and by the proposed conditions we must have 

lAp, qXt-" = 0, 

ni being any integer from to « — 1 inclusive, put t^ = T, the limits 
of 7' are the same as those of t. 

Hence J^ip, ?)» T~^~' ■ T''= 0. 

Now ij), q)„ T" , is a function of T of which the indices are in 

arithmetical progression, - being the common difference, and T' the 

first term ; and as the nature of the question affords m independant 
equations for the determination of the n coefficients Au A-,...A„, it 
follows that there is only one function of the kind, which will satisfy 
the proposed conditions, and by Art. 5, it is evident that the function 

5 2'"''' (1 


1 1 ,A/ 1 
n+ 1 w+- - 

\ q l\ q 

' 1 

answers those conditions, and is manifestly of the required form, it 
Vol. V. Paet II. R 


follows that if we divide this function by T"^ , and then substitute f 

for T, we shall obtain the value of {p, §-)„ ; we have thus, 

_ ip + \){p + l+q){p^-\ + 2q)....\p + 1+{n-l).q] n 

ia+g){l + 2q)....{l + (n-l).g\ l' 

, (2p + l)(2p + l+q)....{2p-i-l + (n-l).q} n.( ?i-l) 

l.{l+q)....{l+{n-l).q} ' 1.2 ^ .'^^• 

13. The functions {p, q\ and (5-, jo)„ may be termed reciprocal func- 
tions, and possess the remarkable property, that if n and «' are any 
different integers, then shall 

ft(p,q)n.{q,p)n' = 0. 

For if n>n' then {q, p)„' is a rational and entire function of t^ of 
less than n dimensions, and therefore by the preceding Article the 
integral of the product must vanish ; again if n' > n, then {p, q)„ is a 
function of f^ of less than n' dimensions, and therefore when multiplied 
by (q,p)„' the integral ought to vanish. 

To determine the value of the same integral when 71 = n', it is 
evident by the nature of the function {p, q)„ that we need only attend 
to the last term in the expansion of {q,p)n, namely 

. {nq^l){nq + l+p)....{nq^-l + {n-\).p} 

^ "->•'' 1.0.+p)....{l^{n-l).p} 


Now if we put for {p, q)^ the series assumed in Art. (12) aiid multi- 
plying then by f, integrate from ^=0 to ^ = 1, we have 

ar + 1 a;+j9 + l a; + 2/> + l "*" x-\rnp + l 
and actually adding these fractions, the denominator of the sum is 

{x + '\){x +p + l)(;r + 2jo + 1) {x + np + 1); 

and since the numerator is of n dimensions in x, and vanishes when 

x = 0, q, 2q....{n- I) . q, 
it follows that the sum is of the form 

c .X . (x — q) (x—2q)....{x — {n — l).q] 

{x + l).{x+p + l)....{x + np + l) 

Multiply by ^ + 1 and then put x= —1; hence 

^^ c.{-l)\l.(q + l)(2q + l)....{{n-l).q + l} ^ 
p . 2p . 

whence deducing the value of c, and substituting in the above integral, 
we obtain 

^'^^'^^''•^^^~P^'''l.{q + l){2q + l)....{{n-l).q+l\ 

^^ x.(x-q)(x-2q)....{x-{n-l).q} 
{x+ l){x +p + l)....{x + np + l) ' 

hence y;(^, g)„ .^"^ = (-^y . ^ ^^^ ^| ; ^^^^^ ^^ ^y 

nq (nq - q) [nq — 2y) .... \nq - {n — 1) . q] 
{nq + l){tiq+p + l)....(nq + np + l) 

from whence we obtain finally 

n" 1 . 2 . 3 . . . .?{ 

f^ ip, q)n {q, p)., = „(^ + ^) + i • i(^q^l)„..{{n-.l).q+l\ 

nq(n q — q) {nq — 2q)....{nq—{n — l). q} 

''~'Up + l){2p + l)....{{n-l).p + l} ■ 

R 2 


that is, it 

_ {pqY V .^' . 3' 

n{p+q) + l'l.{p + l)(q+l){2p + l){2q + l)....{{n-l).p+l\{{n-l).q + l\ 

, 2 _ y f 1.2.3....W 1^ 

COR. j,(p,p}n- 2n + l-\l.{p + l).{2p + l)....{{n-J).p-i.l}f- 

14. To find the reciprocal function to that denoted by L„ in Art. 8, 
, d" {f (h.\. ty\ 
^' 1.2....W df ' 


L„ consists of the powers of h. 1. /, and possesses the property of 
ftL„t' = when x<n; suppose now that we investigate a rational function 
X„ which shall possess the property JtK {h.l. t)' = when x<n; then it 
is evident that j^X„i„/ = when n and n are unequal; and therefore they 
are reciprocal functions. 

Put K=l + AJ + A,f +....AJ'', 

Put Ar = 2" + 'B„ A, = 3'' + 'B, A„ = {n + lY^\B,r, 

hence we must have when x<n, 

1'-' + a"-'^, + 3"-"^2 + (« + i)"-^jB, = 0. 

Now the left-hand member of the equation is the same as 

putting t = after the differentiations. 

Hence the differential coefficients from the 1" to the w* inclusive 
of the function between the brackets vanishes when ^=0; that function 
of e' ought therefore to contain no power of t inferior to the (w + 1)"', 
and conversely, a function of e* which does not contain such a power 
of t, will fulfil the required conditions. 


Now this is the case with (1 — e')''''"^ which is also when expanded 
of the same form as the part between the brackets; hence equating like 
terms, we have 

Hence A,= -\.T, A,= '^^^^ .S" ^„ + , = (- 1)". (« + !)"; 

and therefore 

X„ = l-p2"^+'^^.3'7^- (-1)". (« + !)«. r. 

Cor. 1. When ?^ and n' are unequal, then ftL„'\,„ = 0. 

But when Ti'=fi, we need only take the last term of L,„ namely, (h. 1. /)"; 

j;x„z>„ = j;(h.i.^)"{i-^.2"^+'i^j^^.3"^^-&c.| 

= (-l)..,...S....„{.-f.l.^^).l-.e.} 

_ (-l)''.1.2.3....w 
~ ft + l 

Cor. 2. j;x„(h.l.^)^' 

= i - ly .1 . 2 . 3....X ll'-^-' - n Q.""-' + ^~^ .3"-"-' - kc.\ 

= ( - 1)"- M .2.3 ...x A" . (A"-*-'), 

h being put = 1 after the operation of taking the «* finite diiFerenc<» 
on the supposition that the increment of k is unity ; from whence it 
is easy to deduce 

^^■^' = <-')-'^--^- 

Cor. 3. All the roots of the equation X„ = are real, and lie between 
and 1. 

For if we put h. 1. (/) = it, and X„e"= U, 

then ;x„ (h. 1. ty = f. Uu^ = ti^f^ U- xw-^f.: U+ ^4^^ // U, &c. 


and putting x = 0, 1, 2, &c. successively, it follows that fu"U and its (« — 1) 
successive differential coefficient vanish when u = and w = - oo , Hence 
U=0 has n real negative roots; and therefore X„ = has n real positive 
and fractional roots. 

15. In general let U,„ V„ be any functions of the variable t and the 
integer n, and let A-^.-.A,,, ai...a„ represent constant quantities; or de- 
pending on n only. 

Put T„ = C/„ + A,U, + A,U, + .... + A^U„, 
and T:= K + «i^> + «-.F, + .... + a,r„. 
Then the n equations 

j;r„r„=o, f,T„r,=o, j,t„v,=q ;r„r;_,=o, 

Avill serve to determine the constants A^, A.,....A„. 

In like manner let the corresponding constants «i, a2....a„ be de- 
termined from the n equations 

the functions T„ and T„' which are thus determined, are reciprocal func- 
tions, and possess the general property ft T„ TJ = 0, except when n - n', 
and then 

ft 2\ T: = aJtT^K = A,, ft T: C7„ ; 

this is the general principle of reciprocal functions. 

Cor. Let f{t) be any function of t represented by the series 
f{t) = c,T, + c. 2\ + c, T, .... &c. 

where Co, c,, Cg, &c. are constant coefficients to be determined, then 
multiply by T^, T(, T~U &c. and integrate the successive products, 
and we get 

c,ftT,Tl = ff{t)T^, 

c^ftT.TI^ ftf(f).Tl, 

c.fT,T^ = f,f{t).T.I, 

&c &c. 

by means of which equations the required coefficients are given. 


16. Let «„, h„, c„, &c. be any functions of t, the reciprocal functions 
to which for simple integration are «„', J„', c'„', &c. 

Let a„, &c. be any function of another variable T, and let a/, &c. 
represent the corresponding reciprocal function. 

Put S„ = a„a^ + Kai + C^a^ + 

and S,! = an'uo + i/a/ + c„'a.2 + 

then S„, Sn are general forms for reciprocal functions with respect to 
the double integration relative both to t and T. 

For if we put m for n in the latter series, and multiply the series 
for S„ and S,„' together, the integral of the products of any two terms 
which do not hold the same place in either series when taken relative 
to T must vanish, since a„, a„' are reciprocal functions. 

Hence frSnSJ = a„a„' fj.aoaa + b„b,„' fraiai + c^cj frO^a./ + 

Integrate now with respect to t, observing that when m and n are un- 
equal, then 

_^ «„«,„' = 0, ftKbJ = 0, ftC„c„' = 0, &c. 

Hence /_4*S'„«S',„' = 0, when m is not equal to n, 

and ftfrSuSn = ftfr {a„a^aoa^ + Kb„'aiai' + CnC^'a-^a^ +...]. 

Cor. 1. In the same manner reciprocal functions of any number 
of independent variables may be formed. 

Cor. 2. The equation S„ = has n real roots or values of t lying 
between and 1, whatever value be assigned to 7', when a„, b„, c„, kc. 
are functions possessing the property ftaj' = 0, &c, x being any integer 

from to w - 1 inclusive ; for then «„ must be of the form — —jj-„ — - > 
by Art. 5, and similarly 

, d\{t'-t"'V') _ d\{t^t"'F") 
"~ dt" ' ^"~ df ' 
and therefore 

Hence »S',=0 must have n real roots between and 1. (Art. 6.) 


17. If it is necessary that the terms which compose the reciprocal 
functions S,„ S,! should follow a simple law, it will be most convenient 
to get first two reciprocal functions of t, as R,,, R,', which may contain 
an arbitrary constant r, and to put for «„, J„, c„ &c. the values acquired 
by R„ when r = 0, 1, 2, &c. ; and similarly for «„', i„', c,,', he. the cor- 
responding values of R\. 

Example : 

Thus, put R^^iiff'"^-^, and RJ = {ttf'" '^, P„ being the 

d" (tt'Y 
function so denominated in Art. 3, namely, — ^,^ ; then, integrating 

by parts, we have 

the part outside the sign of integration vanishes between the limits of 
/, and repeating the same operation any number of times, the part out- 
side the sign of integration is evidently of the form 

dt'-" ' dt"-' \ dt' 

the latter differential coefficient will vanish between limits when k is 
any number from to r inclusive, because it will always contain the 
factor {tt'Y~'"^^ ; also when n and m are unequal we may suppose w to 


be the greater, and since ft'' — j-^ is of in + ;• dimensions, it follows 

that if k> n + r, then k -1> in + 7-; and consequently the latter dif- 
ferential coefficient will be identically zero. 

^*-i / d' P 
The only instance in which the factor , .._, iff' , .'" j does not 

Aanish between limits is, therefore, where k lies between r + 1 and r-\-n 
inclusive, but then the first factor is changed to ft'''P„; and since k — r 
is now some immber from 1 to « inclusive, this factor vanishes between 
limits (vid. Art. 5.), and therefore the part outside the sign of integration 
vanishes in all cases, and we thus obtain , 

f,R.R.„ -(-1; j^-^^,-^.-^[tt -^jr)' 


Put now h = r, the first factor under the sign of integration becomes 
simply P„, and the second factor is then of m dimensions; and there- 
fore, by the nature of P„, the integral vanishes; and therefore, when 
n>m, ftIl„Bm' = 0: and the same reasoning applies when m>n, only sub- 
stituting RJ instead of R^ throughout the process, hence R„ and R,„' are ' 
reciprocal functions. 

When m = n, then in the general expression 

j;R«.' = (-irj;p.^(«-^); 

we need only take the term involving the highest power of t in 

dr K^ dr )' 

/ ,.... (« + !)•(« + 2)...(2w) d^ (..rd-.n 
^ ' 1.2... n dr \ dt' 1 

. ,, , (« + l) . (« + 2)...2« , . , ,^ , ,x , 

and observing that /JP„#" = ( — 1)" . -, .,, , — ''" /_ — -^. ; 

it follows that ftR„R,!=- . {n + r) {n + r-1) {n + r-2)...{n-r). 

The reciprocal functions a„, a„' may be obtained by putting r = 
in R„ and RJ ', similarly, if we put r = l, we get b„, b„', &c., and thence 
we obtain the reciprocal functions relative to double integration, namely, 

dP d^P d^P 

S,'=:ao'{tt'Y ^n + «.'(«')^^'^ + «^'(«T^"^-f" + «3'(«')^-"^", &c. 

In the same manner if we vary the constant a while r remains constant, 
we obtain the reciprocal functions 

Vol. V. Part II. S 


Cor. 1. The simplest form for a„ is the sine or cosine of the w'" 
multiple of an arc of which the limits are and 2w7r, as 

A„ sin (2 nicT) + B„ cos (2 wtt T), 

where A„, B„ are arbitrary constants, then we have (putting for sim- 
plicity a = 0), 

S„ = A,P„ + {A, sin ^TTT + B, cos ^-n-r) -^ 

+ {Ai sin 4 TTT + Bi cos 4 ttt) 


■i > 

this is the most general form for all the reciprocal functions which occur 
in the Mecanique Celeste. (Vid. Prop, xi. Treatise on Electricity.) 

CoK. 2. If T„, T,' are arbitrary functions of t, which do not become 
infinite when ^=0 or 1, then, putting 

Rn = {tt'f Tr*^, and R,: = {tt'f T; .^ , 

the same reasoning as that used in the preceding example will show 
that R^, R„' are reciprocal functions, and thus we get for a^^, aS",,' the 
very general forms 

S„ = «„ T,P„ + «. y. ^ («')* + «^ T^ -^ m + «3 T, ^ {tt'f + &c. 

S: = a„' 2;'P„ + a/ T; "^ {tt'f + a.: T^ ^ {tt') + ai Ti ^{tt'f + &C. 

Cor. 3. If f{t, t) is any function of the variables /, t, which is ex- 
panded under the form 

f{t, t) = a,S^ + a,Si + a^S; + 

then, to determine the coefficients a^, Ui, a-i, &c., multiply successively 
by So, Si, SJ.... and integrate from t=0 to t=l, and from t = to 
T = 1 : we thus get 

do ft fr So So = ftfTSo'J'{t, t), 

aiftfrSiSi' = ftf,Si/{t, t), 
aJJ^S.,S.; = f,f^S./f{t,T); 
from whence the required coefficients are known. 



Inverse Method for Definite Integrals which are expressed in positive 
powers of x, or under any form. 

18. Let <^{x) represent any function of x, such that Stf(Jt) .f = (p{x) 
when X is any integer from to n — 1 inclusive, then excluding the 
case of (p {x) = 0, which has been considered in the preceding Section, 
it is evident that by putting 

f{t) = A, + A,t + A,f + +An-,tf-\ 

the conditions of the question give n equations, which suffice to de- 
termine the coefficients A^, Ai, A^, A„.^\ if we represent the 

particular value of f{t) thus deduced by T„^i, and seek its most 
general value, we have 

;/(0 .t^ = <p {x), 

.-. f,{f(t)-T„.,}.t^ = 0. 
Hence by the preceding Section, the most general value of f{t)— Tn-i is 

dt- ' 

and therefore the most general value oi f(t) is found by adding this 
appendage to its prime value T„_i. 

19. When <p{x) is a rational and entire function of x, of m di- 
mensions, we have by the proposed conditions 

'P^'^'' x+l^ x + 2^ x + 3^ x + n' 

and actually adding the terms which compose the right-hand member 
of this equation, the common denominator is 

(x + l){x + 2) (x + n), 



and tlie numerator will be a function of ft — \ dimensions, represented 
by v„, so that 



{x+ l)(ar + 2) (a; + M)' 

when X is any integer from to (^^ - 1) inclusive; and if we multiply 
by a;+ 1 and put x= — 1, and again by a; + 2 and put a-= — 2, &c. as in 
the preceding Section, we get 


V - 







= - • 

1 * 











'~ 1.2 ■ 1.2.3....(« - J)' 

&c.= &c. 

Now the equation 

^{x) . {x + 1) (a; + 2) {x + n)- v„ = 0, 

is of m + n dimensions, and is by hypothesis satisfied, when 

^• = 0, 1, 2, («-l); 

therefore if u^ represent some function oi x of m dimensions, we must 
have the identity 

(p{x) .{x + V) (ar + 2) (ar + «)-», = M,.ar . (^-l)(ar- 2) (x-n + \), 

hence if we divide 

<f>{x){x + \){x^2) (x + w) by x{x-'\){x-^) {x-n^l\ 

and retain only the part of the quotient which is an entire function of x, 
u, will be completely determined. 

Put now —1, —%,...— n successively for x in the preceding identity, 
and we get 

t;., = (-l)»+M .2.3....».«_,, 


«_, = (- 1)"+'. 1.2.3....n/-~-.u.2, 

&c. = &c. 

from whence the values of A^, Ai, A-z, &c. are known, and being sub- 
stituted, give 

J .M_2^ 

n.(«.H)(.-f2) («-l)(.-2) I 

^ 1.2 1.2 ' J 

Example : 
Let 0(a;) = 1, then «, = 1, and therefore 

«.(« + l)(w + 2) (w-l)(/?-2) 
"*" 1.2 ■ 1.2 


20. The function Tn-\ possesses a property analogous to the charac- 
teristic property of those in the former Section, that is, the equation 
2\_^ = admits of n — m-l roofs between and 1, and consequently 
vanishes an indefinitely great number of times between the limits / = 
and t=\ when n is taken indefinitely great. 

For since r„., = (- 1)-' |«M., - ^^^^jtil . ^V «., ^ 
n{n+l){n+2) {n-l)(n-2) ] 

■^ TTa • 1.2 •«-3^&c.j 

_ (-1)- ^\t^(u -VlzI u 1 1 (»-i)-(^^-a) „ t. .,„^l 
= i.2.3....(.*-i)-rfrr \ ' 1 ^ ^'^^ i:% -"-a^-^c.JI 

= -r 2.3..U-i) -£^^'^"'"-^^"'^- 


tlie operation A being performed on the supposition that the finite 
increment of x is unity, and x being put =1 after the operation A""' 
has been performed. 

Put i=l — f, and therefore, 
A"-'(M_,r-') = A"-'M_,-^A'-'M_.(a;-l)+— — A"-'M_,(a;- 1) (a--2)-&c. 

and since m_^ is of m dimensions, the first term of this series which does 
not vanish is 


- 1 ■ 2.-(«-/»- 1) •^""'"-^^'^~ ^^ (^-^)--^^ -n + m + l), 
and therefore the whole expansion is of the form 

t'"-"-' r, 1.2.3 (w-l), 

which being substituted gives 

_ d'{t''t"'-'"-T} 

and since the equation t''t''""'-'^F'=0 has at least 2n-m—l real roots, 
viz. ti of them =0, and n — m—l of them = 1, it follows that the w"" 
derived equation T„ = has n — m—l real roots lying between and ] . 

COK. Since r„_. = ,.,.3.1(,_^) • ^ {r^-^u.J-^}, 
if we actually differentiate we get 

^-^= 1.2.3.!..(«-l) -^""'^^-^-^+^>—^^ + "~^^"-^"'^- 

21. Let now <l>{x) be any function whatever, and let it be required, 
in general, to find J'(i), so that ftj'it) ■ f = ^{x), provided x be any 
integer from to w — 1 inclusive. 

It has been shewn in Art. 18, that a function T„.i of w— 1 di- 
mensions may always be found to satisfy the imposed conditions, and 
for the most general value oi f{t) we shall then have 



Now 7'„-i contains only n constants, being of » — 1 dimensions, and 
therefore if we denote by P„ the same quantity as in the preceding 
Section, namely the coefficient of h" in 

{1- 2h{l-2t) + h:'}-i, 

we may put 

T„.i = ttoPo + a^P, + (hP2 + + a„_,P„_i, 

the right-hand member being of the same dimensions with the left, and 
containing the same number of constants. 

Now by the properties of P„ we have j;P„P„ = 0, when m and « are 
unequal, and 

2« + l 
Hence we have fiP^T„_.,= «„ 


Jl'* 2 -* n-1 — "^ • 

But by the conditions of the question, 

jc being any integer less than n. 

j;P„7;-i = ^r„_i = 0(O) = (f>{h) when h is put =0, 

iP 7'„_,=j;2;_, (1 - 2o=0(o)-20 (1)= - A ^^y^ .<^ (A), 


and generally 


= (-l)'"A"'.^^ '\ ^ ' ^^ '-.d){h). When h is put =0. 

' 1 . 2 m ^ ' ^ 

and by comparing the former integrals with the latter, the values of 
ffo, «i, a-i, &c. are known, and being substituted give 

T._, = P,0(/^)-3P,A^.0(A) + 5P.A^^^±^^^±^.0(A) 

J. • <« • t7 

// being put =0, after the operations are performed. 

It should be observed here that the terms of this expansion are 
perfectly independant of «, which only fixes the number of the terms; 
hence this series may be continued to any number of terms, and we 
shall always have ftT„.it^ = (p{x) provided x is any integer less than that 
number, and consequently if the series be continued ad infinitum, the 
equation will be true for all integer and positive values of x. 

Cor. Multiply both sides by if and integrate from t = to /f=l, 
hence «^ (^) = ^ • <^ ('') + ^ . ^^^j;;^^^^^ A ^ . A 

+ ^-(x + l)(ar + 2)(x + 3)^ 1.2 '?>^ + *'C- 

when /* is put =0. 

This series may be used, not only for the integer and positive 
values of x, but for any values which will not render it divergent. 
(Vid. First Memoir, 'On the Inverse method of Definite Integrals,' 
Art. 2.) 


22. When 0(a;) is given we may obtain f{t) in an infinite variety 
of forms by means of the theory of reciprocal functions given in the 
preceding Section. For instance, if we denote by S^ the sum of the 

products of the natural numbers 1, 2, 3. n when taken m and m 

together, and put 

i.=i+«.h.i,^.j«^.(h.M.+ ^.(h.M-+....+ r^.ch.ur 

,5!lM!!iM. (Art. 8. Section IV.) 

and \„ = l-?.2"/+^4^^.3"f- ±{n+l)''t 

= (-l)"A"{(A + l)''^*}, when h is put =0, 
then L„ and \„ are reciprocal functions. (Sect. iv. Art. 14.) 

Put therefore y*(^)=aoZ/o + «iZ/i + a2Z/2+a3i3 + &c. 
and observing that 


ftKL„ = {-lY.- 

w + 1 

we have «„ = ( - 1)" . ^ ^"^ ^ . ftf{t) . X„. 
But jl/{t) . \ = ftf{f) . ( - 1)'. A" . (A + 1)" . t. When h is put = 0, 

=(-i)"A"(a + i)»j;/(o.^ 

= (-l)».A".(A + l)«.0(A), since ft/{t).f = (p{x). 
Hence «„= ^ ^^ — - . A» . (A + 1)" . (k), 

and therefore 

f(f) = Lo(p{h) + 2Li — ^ ^' r ^.gjr,^ — i 2 +^-^^- — i 2 3 ' 
Vol. V. Part II. T 



which series when convergent will satisfy the equation jtf{t) . f' = ^ («) 
for all values of x\ but even if not convergent, it will satisfy that 
equation for all the integer values of x from to n — \ inclusive, 
provided it be continued for at least n terms. 

If we multiply by f and integrate as before, we get 

which series when convergent may be used for any value of x, but 
only positive and integer values when divergent. 

23. In Art. 21. when ftf(t).t'^(p(x) a given function of x, we have 
found y(0 in a series expressed by functions of t of the same nature 
as P„, now P„ is only a particular value of the general function (jo, q)„ 
investigated in the former Section, Art. 12., namely, when p = q = l; we 
shall now express /{t) according to this more general class of functions, 
that is, under the form 

fit) = «o ip, q)o + «i (p, q)i + «2 {p, q)2 + &c. 

Now in Art. 12. above referred to, we have found 

, . _ {p + l){p + l+q)....{p^l + (m-l).q] m 
Kp,qh-i l(l^q)....{i^{m-\).q} l"^ 

(2p + l)(2p + l+g)....{2jo + l + (?»-l).g} 
■^ i.(\+q)....{\ + {m.-l).q} ' 1.2 '^ " *''• 

To simplify this expression, put 

77 = (/>^ + l)(M + l+9)--{p^ + l+(^-l)-g} 

'■'' l{l+q)....{l+{m-l).q} 

Let yj^ express the operation of changing h into h + 1 (Vid. former 
Memoir, Note B. 2.), >//^ the repetition of this operation a second 
time, &c. ; the preceding series will then become 


{p, 9),„ = H,r - f . ^H,,f" + ^^^ >\^^H,.r 

on the supposition that we put h = after the operations above indicated, 
are performed. 

Separate in this expression the symbols of operation and of quantity, 
and we shall obtain the equation 

(p,q),„ = (l-fr.H,J'': 

But \U — 1 or \^ T- x//° indicates that we must subtract the original 
value of Hp,q, from the value it receives when h + 1 is put for h, 
that is, it is the same as performing the operation A of finite differences ; 
this consideration transforms the preceding equation, to this 

(p, q)m = (-iy" A" . Hf.qt"", when h is put =0. 

In like manner if we put 

„ ^ i,qh-\-\) {qh + l+p) {qh + 1 -\-{m-l) .p} 

"■' 1(1+^) {\ + {m-\).p} 

we have (g-, jo)„ = (-l)"' A'" Jf^.pi?"', when A = 0. 

Now observing that by the nature of reciprocal functions we have 

S* ip, q)m (q, p)n = 0, except when m = n, 

and by Art. 13., fi{p, q\{q,p)^ 

_ ip, q)'" 3 .m . m 

~ 1 + mip+q) '1.1. (l+ju)(l + 9)(l + 2^)(l+29)...{l + (»w-l) .p} {l + {m-l).q} ' 

then since f{t) = «„ (p, q)o + «i (p, q)i + a, {p, q)^ + &c. 

we have ftf{t) . (q, p)„ 

_ (pq)'" 3 m . m 

"**"'• l + m(p+q)'l.l{l+p){l+q) {l + im-l).p} {l+{m-l).q} ' 



But if we put for (q, p)„ the value above found, and observe that the 
operations A and fi are with respect to different variables h and t, and 
therefore their order is transmutable, we have also, 

= {-iy A"^ H,,p<p{qh), by hypothesis. 
Comparing this value of the integral with that already found, we get 

'" ^ ' {pqT I'l" 2 ■ 2 • 3 • 3 ■■■ 

l + {m-l).p \ +{m-\) .q 

X A" JZ", p {qh), when h = 0, 
from whence we have finally 

At) = {p, q). (qh) - ip, q), . ^+f/^ • T ' T " ^ ^V. <l> W 
, , 1 + 2(0 + 0) 1 1 1+p 1+q .,„„ , , 

_(« «N l+^Ci>+g) 1 1 l+£ l+i 1±2£ 1+22 A3 W^'" ri.r«M 
^^'^'°- (pqf ri- 2 •^^- 3 •—^■^■^'>f't>'^W 

+ &c &c. 

h being put =0, after the operation, and H', H", H', &c. being the 
values of Hp,, when m = \, 2, 3, &;c. successively. 

Cor. 1. Multiply by t\ and then integrate from ^ = to ^=1; for 
Itf{t).t' put its value <p{x), and for ft{p,q)mt'' its value 

/ ,v„„,„ 1-2. 3. .-^^ xix-q)...{x-{m-l).q\ 

^ 'P ■ 1 (1 + g)(l + 2y)...{l -!-(»«- 1| .^)*(a;+l)(a;+jt) + l)...(a; + »w^+l)' 


by Art. 12 ; and lastly, put for H,,^ its value 

(gA + 1) {qh + 1. +i)). . ..\ g^ + l+ { m-\).p\ . 

1.(1 +;?).. .|i+(»w-iy:jo"i 

we thus obtain 

+ »(^- g> L+a(i>±2) ^. (^ ^ (^ (^ 

(a: + l)(a;+jo + l)(a?+2^ + l) Sg'' '^ '^^ ^ /rv"/ / 

x(a; — 5')(a; — 2^) 

■*" {x + l)(a:+jt> + l)(ar + 2/> + l)(2 + 3ja + l) 

^ ^ 1^.2^^^ ^'^^^' + ^^^^^ +-^ + ^^^^^ + aja + 1) {qh) 
+ &c. when A is put = 0, 
and where /> and q are perfectly arbitrary. 

Cor. 2. Put ^ = ^ = 0, and make 

where ^(0), 0'(O), 0"(O), and the values of ^{x) and its successive 
differential coefficients when a; = 0, and the above expansion will become 

</>(x) = ^„.^ + ^,.^^-., + ^..^3 + &c. 

If, moreover, we put 

rr, , « , , , W.(»-l) (h. 1. O'' „ 

which is the same as A„ when we put f for ^ (a;), then it is easily 
seen by the principles of the first Memoir, that jj Tj' = -, r-r—r , and 

/ r r , ji n (a; + l)"+'' 

since we have also fij'it) . /^ = {x), it follows that 


24. The functions which have been all along designated by {p, q)„ and 
{q, p)„, have been already shewn to be reciprocal one to the other; putting 
p = q, the resulting function {p, p)„ must be reciprocal to itself; that is, 
ft{p, p)„{p, p),„ = when m and n are unequal positive integers; when 
p = l the function {p, j)),, is then identical with that denoted by P„ , which 
has been before shewn to be reciprocal to itself; again, the function T„ or 

n n.(n~\) {hA.ty ^ n.{n-l) .{n-^) (h. 1. If 

is reciprocal to itself, for if we mviltiply by (h. 1. ty, and integrate, we get 

j;r„(h.l.0"' = 1.2.3...«.(-ir{l-f^-^!^ (^±i)^_&e.}. 

The expression between the brackets is the term independent of h in the 

product (1+/^)"(1 + t) , or the coefficient of //-('"+'> in (1+A)"-'"-'; 

it is therefore zero when n>m, but when n = m its value is ( — I)'", 
and when n<m, its value is 

, _ (?w + l -n){m + 2-n)...m 
^~ ' ' 1 .2 ...n '■ 

Hence fi T„ 7'„ = 0, when m and n are unequal, and 

1.2. ..n 

25. Put h. 1. (^) = T, and substituting in T„, we have 

1.2...W J'„e'^ = e"|l.2...M + w.2.3...Wx+^^^^^\3.4...WT^ + &C.| 
(dw c?"-'t" n.(n-l) d"-W „ 1 

_ d"{e'^r'') 
„ _, 6-^d" (e-T") 

Hence 7; = -^—p^ t-^ . 

1.2. ..war" 


From , this formulae it appears that the equation T„ = has n real 
values of t all negative; and therefore n corresponding values of t, 
which are all included between and 1. 

Moreover, if we form the equation 

u = T + hu, or u 


it follows by the theorem of Lagrange, that T„ is the coefficient of h" 

de" e'~* 

in ^'^•-j-> that is, in - — y, and putting t for e% T„ is clearly the co- 


efficient of h" in the expansion of the function y . 

Conversely, we may now prove that the coefficient of h" in the ex- 


pansion of - — - is a reciprocal function; for when h = 0, this function 

A ""■ ft 

is reduced to unity, we may therefore put generally 

= ro+T,^+T,A^ + &c. where T, = \. 



Let h' represent any other arbitrary quantity, and we have 


j= T,+ T,h'+T,h" + &ic. 

Multiply both series term by term and integrate, the result in the 
left-hand members is 

{i-h){i-h')^' ~ i-hh" 

/which expanded becomes 1 + hh' + h^h'^ + kc.; which being identical with 
the integral of the product of the right-hand members, will necessarily 
require that the integrals of those terms which are not in corresponding 
places in both series must vanish, and the integrals of the products of 
the corresponding coefficients to be unity, which are the same properties 
that have been demonstrated in Art. 24. 


Cor. Put .; — r = ^» and the series 

t~^ = (\~.h){ T„ +T,h+ TJi^ + &c. \ becomes 

^^=^+ ^■•r4T^+ ^-7;:ttv3 + *'^- 

The principles which have been used in this Section to obtain ex- 
pansions such as the preceding by means of reciprocal functions relative 
to simple integration, will apply with equal simplicity to reciprocal 
functions relative to any number of integrations. 


Caius College, 
Bee. 18, 1833. 

VII. On the Nature of the Truth of the Laws of Motion. By the 
Rev. W. Whewell, M.A. Fellow and Tutor of Trinity College. 

[Bead Feb. 17, 1834.] 

1. The long continuance of the disputes and oppositions of opinion 
which have occurred among theoretical writers concerning the elementary 
principles of Mechanics, may have made such discussions appear to some 
persons wearisome and unprofitable. I might, however, not unreasonably 
plead this very circumstance as an apology for offering a new view of 
the subject; since the extent to which these discussions have already 
gone shews that some men at least take a great interest in them ; 
and it may be stated, I think, without fear of contradiction, that 
these controversies have not terminated in the general and undisputed 
establishment of any one of the antagonist opinions. 

The question to which my remarks at present refer is this: "What 
is the kind and degree of cogency of the best proofs of the laws of 
motion, or of the fundamental principles of mechanics, exprest in any 
other way?" Are these laws, philosophically considered, necessary, and 
capable of demonstration by means of self-evident axioms, like the 
truths of geometry ; or are they empirical, and only known to be true 
by trial and observation, like such general rules as we obtain in natural 
history ? 

It certainly appears, at first sight, very difficult to answer the argu- 
ments for either side of this alternative. On the one hand it is said, 
the laws of motion cannot be necessarily true, for if they were so, the 
denial of them would involve a contradiction. But this it does not, 
for we can readily conceive them to be other than they are. We can 
conceive that a body in motion should have a natural tendency to 
move slower and slower. And we know that, historically speaking, 
Vol. V. Paet II. U 


men did at first suppose the laws of motion to be different from 
what they are now proved to be. This would have been impossible 
if the negation of these laws had involved a contradiction of self-evi- 
dent principles, and consequently had been not only false but incon- 
ceivable. These laws, therefore, cannot be necessary ; and can be duly 
established in no other way than by a reference to experience. 

On the other hand, those who deduce their mechanical principles 
without any express reference to experiment, may urge, on their side, 
that, by the confession even of their adversaries, the laws of motion 
are proved to be true beyond the limits of experience ; — that they are 
assumed to be true of any new kind of motion when first detected, as 
well as of those already examined; — and that it is inexplicable how 
such truths should be established empirically. They may add that the 
consequences of these laws are allowed to hold with the most complete 
and absolute universality; for instance, the proposition that "the quan- 
tity of motion in the world in a given direction cannot be either 
increased or diminished," is conceived to be rigorously exact; and to 
have a degree and kind of certainty beyond and above all mere facts 
of experience ; what other kind of truth than necessary truth this 
can be, it is difficult to say. And if the conclusions be necessarily 
true, the principles must be so too. 

This apparent contradiction therefore, that a law should be neces- 
sarily true and yet the contrary of it conceivable, is what I have now 
to endeavour to explain ; and this I must do by pointing out what 
appear to me the true grounds of the laws of motion. 

2. The science of Mechanics is concerned about motions as deter- 
mined by their causes, namely, forces ; the nature and extent of the 
truth of the first principles of this science must therefore depend upon 
the way in which we can and do reason concerning causes. In what 
manner we obtain the conception of cause, is a question for the meta- 
physician, and has been the subject of much discussion. But the general 
principle which governs our mode of viewing occurrences with reference 
to this conception, so far as our present subject is concerned, does not 
appear to be disturbed by any of the arguments which have been 


adduced in this controversy. This principle I shall state in the form 
of an axiom, as follows. 

Axiom I. Every change is produced by a cause. 

It will probably be allowed that this axiom expresses a universal 
and constant conviction of the human mind ; and that in looking at 
a series of occurrences, whether for theoretical or practical purposes, 
we inevitably and unconsciously assume the truth of this axiom. If a 
body at rest moves, or a body in motion stops, or turns to the right 
or the left, we cannot conceive otherwise than that there is some 
cause for this change. And so far as we can found our mechanical 
principles on this axiom, they will rest upon as broad and deep a 
basis as any truths which can come within the circle of our know- 

I shall not attempt to analyse this axiom further. Different per- 
sons may, according to their different views of such subjects, call it a 
law of our nature that we should think thus, or a part of the con- 
stitution of the human mind, or a result of our power of seeing the 
true relations of things. Such variety of opinion or expression would 
not affect the fundamental and universal character of the conviction 
which the axiom expresses; and would therefore not interfere with our 
future reasonings. 

3. There is another axiom connected with this, which is also a 
governing and universal principle in all our reasoning concerning 
causes. It may be thus stated. 

Axiom II. Causes are measured by their effects. 

Every effect, that is, every change in external objects, implies a 
cause, as we have already said : and the existence of the cause is known 
only by the effects it produces. Hence the intensity or magnitude of 
the cause cannot be known in any other manner than by these effects: 
and, therefore, when we have to assign a measure of the cause, we 
must take it from the effects produced. 

In what manner the effects are to be taken into account, so as 
to measure the cause for any particular purpose, will have to be 


153 mk whewell, on the nature of the truth 

further considered ; but the axiom, as now stated, is absolutely and 
universally true, and is acted upon in all parts of our knowledge in 
which causes are measured. 

4. But something further is requisite. We not only consider that 
all changes of motion in a body have a cause, but that this cause may 
reside in other bodies. Bodies are conceived to act upon one another, 
and thus to influence each other's motions, as when one billiard ball 
strikes another. But when this happens, it is also supposed that the 
body struck influences the motion of the striking body. This is inclu- 
ded in our notion of body or matter. If one ball could strike and 
affect the motions of any number of others without having its own 
motion in any degree affected, the struck balls would be considered, 
not as bodies, but as mere shapes or appearances. Some reciprocal in- 
fluence, some resistance, in short some reaction, is necessarily involved 
in our conception of action among bodies. All mechanical action upon 
matter implies a corresponding reaction; and we might describe matter 
as that which resists or reacts when acted on by force. Not only 
must there be a reaction in such cases, but this reaction is defined 
and determined by the action which produces it, and is of the same 
kind as the action itself The action which one body exerts upon 
another is a blow, or a pressure; but it cannot press or strike with- 
out receiving a pressure or a blow in return. And the reciprocal 
pressure or blow depends upon the direct, and is determined altogether 
and solely by that. But this action being mutual, and of the same 
kind on each body, the effect on each body will be determined by the 
effect on the other, according to the same rule ; each effect in turn 
being considered as action and the other as reaction. But this cannot be 
otherwise than by the equality and opposite direction of the action and 
reaction. And since this reasoning applies in all cases in which bodies 
influence each others motions, we have the following axiom which is 
universally true, and is a fundamental principle with regard to all me- 
chanical relations. 

Axiom III. Action is always accompanied by an equal and opposite 



5. I now proceed to shew in what manner the Laws of Motion 
depend upon these three axioms. 

Bodies move in lines straight or curved, they move more or less 
rapidly, and their motions are variously affected by other bodies. This 
succession of occurrences suggests the conceptions of certain properties 
or attributes of the motions of bodies, as their direction and velocity, 
by means of which the laws of such occurrences may be exprest. 
And these properties or attributes are conceived as belonging to the 
body at each j)^^^^ of its motion, and as changing from one point to 
another. Thus the body, at each point of its path, moves in a 
certain direction, and with a certain velocity. 

These properties, direction and velocity for instance, are subject 
to the rule stated in the first axiom : they cannot change without 
some cause ; and when any changes in the motions of a body are 
seen to depend on its position relative to another body or to any part 
of space, such other body, or such other part of space, is said to 
exert a Jbrce upon the moving body. Also the force exerted upon 
the moving body is considered to be of a certain value at each 
point of the body's motion ; and though it may change from one point 
to another, its changes must depend upon the position of the points 
only, and not upon the velocity and direction of the moving body. 
For the force which acts upon the body is conceived as a property of 
the bodies, or points, or lines, or surfaces among which the moving body 
is placed; the force at all points therefore depends upon the position 
with regard to the bodies and spaces of which the force is a property ; 
but remains the same, whatever be the circumstances of the body 
moved. The circumstances of the body moved cannot be a cause 
which shall change the force acting at any point of space, although 
they may alter the effect which that force produces upon the body. 
Thus, gravity is the same force at the same point of space, whether it 
have to act upon a body at rest or in motion ; although it still remains 
to be seen whether it will produce the same effect in the two cases. 

6. This being established, we can now see of what nature the 
laws of motion must be, and can state in a few words the proofs 


of them. We shall have a law of motion corresponding to each 
of the above three axioms ; the first law will assert that when no force 
acts, the properties of the motion will be constant; the second law 
will assert that when a force acts, its quantity is measured by the 
effect produced ; the third law will assert that, when one body acts 
upon another, there will be a reaction, equal and opposite to the 
action. And so far as the laws are announced in this form, they will 
be of absolute and universal truth, and independent of any particular 
experiment or observation whatever. 

But though these laws of motion are necessarily and infallibly 
true, they are, in the form in which we have stated them, entirely 
useless and inapplicable. It is impossible to deduce from them any 
definite and positive conclusions, without some additional knowledge or 
assumption. This will be clear by stating, as we can now do in a 
very small compass, the proofs of the laws of motion in the form 
in which they are employed in mechanical reasonings. 

7. First, of the first Law ; — that a body not acted upon by any force 
will go on in a straight line with an invariable velocity. 

The body will go on in a straight line : for, at any point of its 
motion, it has a certain direction, which direction will, by Axiom I, 
continue unchanged, except some cause make it deviate to one side or 
other of its former position. But any cause which should make the 
direction deviate towards any part of space would be a force, and the 
body is not acted upon by any force. Therefore, the direction cannot 
change, and the body will go on in the same straight line from the 

The body will move with an invariable velocity. For the velocity 
at any point will, by Axiom I, continue unchanged, except some 
cause make it increase or decrease. And since, by supposition, the 
body is not acted upon by any force, there can be no such cause 
depending upon position, that is, upon relations of space; for any 
cause of change of motion which has a reference to space is force. 

Therefore there can be no cause of change of motion, except 
there be one depending upon time, such, for instance, as would exist 


if bodies had a natural tendency to move slower and slower, according 
to a rate depending on the time elapsed. 

But if such cause existed, its effects ought to be considered sepa- 
rately ; and it would still be requisite to assume the permanence of 
the same velocity, as the first law of motion ; and to obtain, in addi- 
tion to this, the laws of the retardation depending on the time. 

Whether there is any. such cause of retardation in the actual 
motions of bodies, can be known only by a reference to experience; 
and by such reference it appears that there is no such cause of the 
diminution of velocity depending on time alone; and therefore that 
the first law of motion may, in all cases in which bodies are exempt 
from the action of external forces, be applied without any addition or 
correction depending upon the time elapsed. 

It is not here necessary to explain at any length in what manner 
we obtain from experience the knowledge of the truth just stated, that 
there is not in the mere lapse of time any cause of the retardation of 
moving bodies. The proposition is established by shewing that in all 
the cases in which such a cause appears to exist, the cause of retar- 
dation resides in surrounding bodies and not in time alone, and is 
therefore an external force. And as this can be shewn in every in- 
stance, there remains only the negation of all grovind for the assump- 
tion of such a cause of retardation. We therefore reject it altogether. 

Thus it appears that in proving the first law of motion, we obtain 
from our conception of cause the conviction that velocity will be 
uniform except some cause produce a change in it ; but that we are 
compelled to have recourse to experience in order to learn that time 
alone is not a cause of change of velocity. 

8. I now proceed to the second Law : — that when a force acts 
upon a body in motion, the effect is the same as that which the same 
force produces upon a body at rest. 

This law requires some explanation. How is the effect produced 
upon a moving body to be measured, so that we may compare it with 


the effect upon a body at rest? The answer to this is, that we here 
take for the measure of the effect of the force, that motion which 
must be compounded with the motion existing before the change, in or- 
der to produce the motion which exists after the change: the rules for 
the composition of motion being established on independent grounds 
by the aid of definition alone. Thus if gravity act upon a body 
which is falling vertically, the effect of gravity upon the body is 
measured by the velocity added to that which the body already has : 
if gravity act upon a body which is moving horizontally, its effect 
is measured by the distance to which the body falls below the hori- 
zontal line. 

The effect of the force which we consider in the second Law of 
motion, is its effect upon velocity only : and it is proper to mark 
this restriction by an appropriate term : we shall call this the accele- 
rative effect of force; and the cause, as measured by this effect, may 
be termed the accelerathe quantity of the force.* 

A law of motion which necessarily results from our second Axiom 
is, that the accelerative quantity of a force is measured by the acce- 
lerative effect. But whether the accelerative effect depends upon the 
velocity and direction of the moving body, cannot be known indepen- 
dently of experience. It is very conceivable, for instance, that the 
force of gravity being every where the same, shall yet produce, upon 
falling bodies, a smaller accelerative effect in proportion to the velocity 
which they already have in a downward direction. Indeed if gravity 
resembled in its operation the effect of any other mode of mechanical 
agency, the result would be so. If a body moved downwards in 

* The accelerative quantity of a force (the quantitas acceleratrix vis cujusvis of Newton) 
is often called the accelerating forces and we may thus have to speak of the accelerating 
force of a certain force, which is at any rate an awkward phraseology. It would perhaps 
have been fortunate if Newton, or some other writer of authority, at the time when the 
principles of mechanics were first clearly developed, had invented an abstract term for 
this quantity : it might for instance have been called acceleralivity. And the second law 
of motion would then have been, that the acceleralivity of the same force is the same, 
whatever be the motion of the body acted on. 


consequence of the action of a hand pushing it with a constant effort, 
or of a spring, or of a stream of fluid rushing in the same direction, 
the accelerative effect of such agents would be smaller and smaller 
as the velocity of the body propelled was larger and larger. We can 
learn from experience alone that the effects of the action of gravity 
do not follow the same rule. 

We assert that the accelerative quantity of the same force of gra- 
vity is the same whatever be the motion of the body acted on. It 
may be asked how we know that the force of gravity is the same 
in cases so compared ; for instance, when it acts on a body at rest 
and in motion ? The answer to this question we have given already. 
By the very process of considering gravity as a force, we consider 
it as an attribute of something independent of the body acted on. 
The amount of the force may depend upon place, and even time, for 
any thing we know a priori ; but we do not find that the weight of 
bodies depends on these circumstances, and therefore, having no evi- 
dence of a difference in the force of gravity, we suppose it the same 
at different times and places. And as to the rest, since the force is a 
force which acts on the body, it is considered as the same force, 
whatever be the circumstances of the passive body, although the ejects 
may vary with these circumstances. If the effects are liable to such 
change, this change must be considered separately, and its laws investi- 
gated ; but it cannot be allowed to unsettle our assumption of the 
permanence of the force itself. It is precisely this assumption of a 
constant cause, which gives us a fixed term, as a means of estimating 
and expressing by what conditions the effects are regulated. 

It appears by observation and experiment, that the accelerative 
quantity of the same force is not affected by the velocity or direction 
of the body acted on : for instance, a body falling vertically receives, 
in any second of time, an accession of velocity as great as that which 
it received in the first second, notwithstanding the velocity with which 
it is already moving. The proof of this and similar assertions from 
experiment produced, historically speaking, the establishment of the 
second law of motion in the sense in which we now assert it. And 
here, as in the case of the first law, we may observe that an important 
Vol. V. Part II. X 


portion of the process of proof consisted in shewing that in those cases 
in which the accelerative effect of a force appeared to be changed by 
the circumstances of the motion of the body acted on, the change was, 
in fact, due to other external forces ; so that all evidence of a cause 
of change residing in those circumstances was entirely negatived; and 
thus the law, that the accelerative effect of the same force is the 
same, appeared to be absolutely and rigorously true. 

9. When the motions of bodies are not affected merely by forces 
like gravity, which are only perceived by their effects, but are acted 
upon by other bodies, the case requires other considerations. 

It is in such cases that we originally form the conception of force; 
we ourselves pull and push, thrust and throw bodies, with a view, it 
may be, either to put them in motion, or to prevent their moving, 
or to alter their figure. Such operations, and the terms by which 
they are described, are all included in the term force, and in other 
terms of cognate import. And in using this term, we necessarily 
assume and imply the co-existence of these various effects of force 
which we have observed universally to accompany each other. Thus 
the same kind of force which is the cause of motion, may also be 
the cause of a body having a form different from its natural form ; 
when we draw a bow, the same kind of pull is needed to move the 
string, and to hold it steady when the bow is bent. And a weight 
might be hung to the string, so as to produce either the one or 
the other of these effects. By an infinite multiplicity of experiments 
of this kind, we become imbued with the conviction that the same 
pressure may be the cause of tension and of motion. Also as the 
cause can be known by its effects only, each of these effects may be 
taken as its measure ; and therefore, so long as one of them is the 
same, since the cause is the same, the other must be the same also. 
That is, so long as the pressure or force which shews itself in 
tension is the same, the motion which it would produce must, under 
the same circumstances, be the same also. This general fact is not 
a result of any particular observations, but of the general observation 
or suggestion arising unavoidably from universal experience, that both 


tension and motion may be referred to force as their cause, and have 
no other cause. 

We come therefore to this principle with regard to the actions of 
bodies upon each other, that so long as the tension or pressure is the 
same, the force, as shewn by its effect in producing motion, must 
also be the same. 

10. This force or action of bodies upon one another, is that which 
is meant in the Third Axiom, and we now proceed to consider the 
application of this axiom in mechanics. 

Pressures or forces such as I have spoken of, may be employed in 
producing tension only, and not motion ; in this case, each force prevents 
the motion which would be produced by the others, and the forces 
are said to balance each other, or to be in equilibrium. The science 
which treats of such cases is called Statics, and it depends entirely 
upon the above third axiom, applied to pressures producing rest. It 
follows from that axiom, that pressures, which acting in opposite di- 
rections thus destroy each other's effects, must be equal, each measuring 
the other. Thus if a man supports a stone in his hand, the force or 
effort exerted by the man upwards is equal to the weight or force 
of the stone downwards. And if a second stone, just equal to the 'first, 
were supported at the same time in the same hand, the force or effort 
must be twice as great ; for the two stones may be considered as 
one body of twice the magnitude, and of twice the weight; and 
therefore the effort which supports it must also be twice as great. 
And thus we see in what manner statical forces are to be measured 
in virtue of this third axiom ; and no further principle is requisite to 
enable us to establish the whole doctrine of statics. 

11. The third axiom, when applied to the actions of bodies in 
motion, gives rise to the third law of motion, which Ave must now con- 
sider. Here, as in the cases of the other axioms, we must inquire 
how we are to measure the quantities to which the axiom applies. What 
is the measure of the action which takes place when a body is put 
in motion by pressure or force? In order to answer this question, we 

X 2 


must consider what circumstances make it requisite that the force 
should be greater or less. If we have to lift a stone, the force which 
we exert must be greater when the stone is greater : again, we must 
exert a greater force to lift it quickly than slowly. It is clear, there- 
fore, that that property of a force with which we are here concerned, 
and which we may call the motive quantity of the force,* increases both 
when the velocity communicated, and when the mass moved, increase, and 
depends upon both these quantities, though we have not yet shewn 
what is the law of this dependence. 

The condition that a quantity P shall increase when each of two 
others V and M does so, may be satisfied in many ways : for instance, 
by supposing P proportional to the sum M+ V (all the quantities being 
expressed in numbers), or to the product, MV, or to MF'-, or in many 
other ways. 

When, however, the quantities ^ and M are altogether hetero- 
geneous, as when one is velocity, and the other weight, the first 
of the above suppositions, that P varies as M + V, is inadmissible. 
For the law of variation of the formula M+ V depends upon the 
relation of the units by which M and V respectively are measured; 
and as these units are arbitrary in each case, the result is, in like 
manner, arbitrary, and therefore cannot express a law of nature. 


12. The supposition that the motive quantity of a force varies as 
M^-V, where M is the mass moved and V the velocity, being thus 
inadmissible, we have to select upon due grounds, among the other 
formulae MV, MV\ M'V, &c. 

And in the first place I observe that the formula must be propor- 
tional to M simply (excluding M.^ &;c.) for both the forces which 

* The motive quantity of a force {vis cujusvis quantitas matrix of Newton) is sometimes 
called moving force; we are thus led to speak of the moving force of a force, as we 
have already observed concerning accelerating force. Hence, as in that case, we might 
employ a single term, as motivity, to denote this property of force; and might thus speak 
of it and of its measures without the awkwardness which arises from the usual phrase. 


produce motion and the masses in which motion is produced are capa- 
ble of addition by juxtaposition, and it is easily seen by observation 
that such addition does not modify the motion of each mass. If a 
certain pressure upon one brick (as its own weight) cause it to fall 
with a certain velocity, an equal pressure on another equal brick wiU 
cause it also to fall with the same velocity ; and these two bricks 
being placed in contact, may be considered as one mass, which a dou- 
ble force will cause to fall with still the same velocity. And thus 
all bodies, whatever be their magnitude, will fall with the same velo- 
city by the action of gravity. Those who deny this (as the Aristo- 
telians did) must maintain, that by establishing between two bodies 
such a contact as makes them one body, we modify the motion which 
a certain pressure will produce in them. And when we find experi- 
mentally (as we do find) that large bodies and small ones fall with the 
same velocity, excluding the effects of extraneous forces, this result 
shews that there is not, in the union of small bodies into a larger one, 
any cause which affects the motion produced in the bodies. 

It appears, therefore, that the motive quantity of force which puts 
a body in motion is, cceteris paribus, proportional to the mass of the 
body ; so that for a double mass a double force is requisite, in order 
that the velocity produced may be the same. Mass considered with 
reference to this rule, is called Inertia. 

13. The measure of mass which is used in expressing a law of 
motion, must be obtained in some way independent of motion, other- 
wise the law will have no meaning. Therefore, mass measured in 
order to be considered as Inertia must be measured by the statical 
effects of bodies, for instance, by comparison of weights. Thus two 
masses are equal which each balance the same weight in the same 
manner; and a mass is double of one of them which produces the 
same effect as the two. And we find, by universal observations, that 
the weight of a mass is not affected by the figure or the arrange- 
ment of parts, so long as the matter continues the same. Hence it 
appears that the mass of bodies must be compared by comparing their 
weights, and Inertia is proportional to weight at the same place. 


Since all bodies, small or large, light or heavy, fall downwards with 
equal velocities, when we remove or abstract the effect of extraneous 
circumstances, the motive quantity of the force of gravity on equal 
bodies is as their masses ; or as their weight, by what has just been said. 

14. For the measure of the motive quantity of force, or of the action 
and reaction of bodies in motion, we have, therefore, now to chuse 
among such expressions as MV, and MV^. And our choice must be 
regulated by finding what is the measure which will enable us to 
assert, in all cases of action between bodies in motion, that action and 
reaction are equal and opposite. 

Now the fact is, that either of the above measures may be taken, 
and each has been taken by a large body of mathematicians. The former 
however {MV) has obtained the designation which naturally falls to the 
lot of such a measure ; and is called momentum, or sometimes simply 
quantity of motion : the latter quantity {MV^) is called vis viva or liv- 
ing force. 

I have said that either of these measures may be taken : the former 
must be the measure of action, if we are to measure it by the effect pro- 
duced in a given time; the latter is the measure if we take the whole 
effect produced. In either way the third law of motion would be true. 

Thvis if a ball B, lying on a smooth table, be drawn along by a 
weight A hanging by a thread over the edge of the table, the motion 
of B is produced by the action of A, and on the other hand the 
motion of A is diminished by the reaction of B; and the equality 
of action and reaction here consists in this, that the momentum {MV) 
which B acquires in any time is equal to that which A loses : that is, 
so much is taken from the momentum which A would have had, if 
it had fallen freely in the same time; so that A falls more slowly by 
just so much. 

But if the weight A fall through a given space from rest, as 1 foot, 
and then cease to act, the eqviality of action and reaction consists in 
this, that the vis viva which B acquires on the whole, is equal to the 
vis viva which A loses ; that is, the vis viva of A thus acting on B is 


smaller by so much than it would have been, if A had fallen freely 
through the same space. 

15. In fact, these two propositions are necessarily connected, and 
one of them may be deduced from the other. The former way of 
stating the third law of motion appears, however, to be the simplest mode 
of treating the subject, and we may put the third law of motion in 
this form. 

In the direct mutual action of bodies, the momentum gained and lost 
in any time are equal. 

This law depends upon experiment, and is perhaps best proved by 
some of its consequences. It follows from the law so stated, that the 
motive quantity of a force is proportional to the momentum generated in 
a given time; since the motive quantity of force is to be equivalent 
to that action and reaction which is understood in the third law of 
motion. Now, if the pressure arising from the weight of a body P 
produce motion in a mass Q, since the momentum gained by Q and 
that lost by P in any time are equal, the momentum of the whole 
at any time will be the same as if P's weight had been employed 
in moving P alone. Therefore, the velocity of the mass Q will be 
less, in the same proportion in which the mass or inertia is greater: 
and thus the accelerating quantity of the force is inversely propor- 
tioned to the mass moved. This rule enables us to find the accele- 
rative quantity of the force in various cases, as for instance, when bodies 
oscillate, or when a smaller weight moves a large mass; and we 
can hence calculate the circumstances of the motion, which are found 
to agree with the consequences of the above law. 

16. But the argument may be reduced to a simpler form. Our 
object is to shew that, for an equal mass, the velocity produced by a 
force acting for a given time is as the pressure which produces the 
motion; for instance, that a double pressure will produce a double 
velocity. Now a double pressure may be considered as the union of 
two equal pressures, and if these two act successively, the first will 
communicate to the body a certain velocity, and the second will com- 


municate an additional velocity, equal to the first, by the second law 
of motion ; so that the whole velocity thus commvinicated will be the 
double of the first. Therefore, if the velocity communicated be not 
also the double of the first when the two pressures act together, the 
difference must arise from this, that the effect of one force is modified 
by the simultaneous action of the other. And when we find by expe- 
rience (as we do find) that there is no such difference, but that the 
velocity communicated in a given time is as the pressure which com- 
mimicates it, this result shews that there is nothing in the circumstance 
of a body being already acted on by one pressure, which modifies the 
effect of an additional pressure acting along with the first. 

17- I have above asserted the law, of the direct action of bodies 
only. But it is also true when the action is indirect, as when by 
turning a winch we move a wheel, the main mass of which is farther 
from the axis than the handle of the winch. In this case the pres- 
sure we exert acts at a mechanical disadvantage on the main mass of 
the wheel, and we may ask whether this circumstance introduces any 
new law of motion. And to this we may reply, that we can conceive 
pressure to produce different effects in moving bodies, according as it 
is exerted directly or by the intervention of machines; but that we 
find no reason to believe that such a difference exists. The relations 
of the pressures in different parts of a machine are determined by con- 
sidering the machine at rest. But if we suppose it to be put in 
motion by such pressures, we see no reason to expect that these pres- 
sures should have a different relation to the motions produced from 
what they would have done if they were direct pressures. And as 
we find in experiment a negation of all evidence of such a differ- 
ence, we reject the supposition altogether. We assert, therefore, the 
third law of motion to be true, whatever be the mechanism by 
the intervention of which action and reaction are opposed to each 

From this consideration it is easy to deduce the following rule, 
which is known by the designation of D'Alembert's principle, and 
may be considered as a fourth law of motion. 


WJien any forces produce motion in any connected system of matter, 
the motive quantities of force gained and lost by the different parts 
must balance each other according to the connexion of the system. ■ 

By the motive quantity of force gained by any body, is here 
meant the quantity by which that motive force which the body's mo- 
tion implies (according to the measures already established) exceeds 
the quantity of motive force which acts immediately upon the body. 
It is the excess of the effective above the impressed force, and of course 
arises from the force transmitted from the other bodies of the system 
in consequence of the connexion of the parts. The motive quantity 
of force lost is in like manner the excess of the impressed above the 
effective force. And these two excesses, in different parts of the sys- 
tem, must balance each other according to the mechanical advantage 
or disadvantage at which they act for each part. 

This completes our system of mechanical principles, and authorizes 
us to extend to bodies of any size and form the rules which the 
second law of motion gives for the motion of bodies considered as 
points. And by thus enabling us to trace what the motions of bodies 
will be according to the rule asserted in the third law of motion, 
(namely, that the motive quantity of forces is as the momentum pro- 
duced in a given time,) it leads us to verify that supposition by experi- 
ments in which bodies oscillate or revolve or move in any regular 
and measurable manner, as has been done by Atwood, Smeaton, and 
many others. 

18. We have thus a complete view of the nature and extent of 
the fundamental principles of mechanics; and we now see the reason 
why the laws of motion are so many and no more, in what way they 
are independent of experience, and in what way they depend upon 
experiment. The form, and even the language of these laws is of 
necessity what it is; but the interpretation and application of them is 
not possible without reference to fact. We may imagine many rules 
according to which bodies might move (for many sets of rules, dif- 
ferent from the existing ones, are, so far as we can see, possible) and 
we should still have to assert — that velocity could not change without 
Vol. V. Pakt II. Y 


a cause, — that change of action is proportional to the force which pro- 
duces it, — and that action and reaction are equal and opposite. The 
truth of these assertions is involved in those notions of causation and 
matter, which the very attempt to know any thing concerning the rela- 
tions of matter and motion presupposes. But, according to the facts 
which we might find, in such imaginary cases as I have spoken of, 
we should settle in a different way — what is a cause of change of ve- 
locity, — what is the measure of the force which changes motion, — and 
what is the measure of action between bodies. The law is necessary, 
if there is to be a law ; the meaning of its terms is decided by what 
we find, and is therefore regulated by our special experience. 

19. It may further illustrate this matter to point out that this 
view is confirmed by the history of mathematics. The laws of motion 
were assented to as soon as propounded; but were yet each in its turn 
the subject of strenuous controversy. The terms of the law, the form, 
which is necessarily true, were recognised and undisputed ; but the 
meaning of the terms, the substance of the law, was loudly contested; 
and though men often tried to decide the disputed points by pure 
reasoning, it was easily seen that this could not suffice ; and that since 
it was a case where experience could decide, experience must be the 
proper test: since the matter came within her jurisdiction, her authority 
was single and supreme. 

Thus with regard to the first law of motion, Aristotle allowed that 
natural motions continue unchanged, though he asserted the motions 
of terrestrial bodies to be constrained motions, and therefore, liable to 
diminution. Whether this was the cause of their diminution was a 
question of fact, which was, by examination of facts, decided against 
Aristotle. In like manner, in the first case of the second law of 
motion which came under consideration, both Galileo and his oppo- 
nent agree that falling bodies are uniformly accelerated ; that is, that 
the force of gravity accelerates a body uniformly whatever be the 
velocity it has already ; but the question arises, what is uniform acce- 
leration ? It so happened in this case, that the first conjecture of Ga- 
lileo, afterwards defended by Casraeus, (that the velocity was propor- 


tional to the space from the beginning of the motion) was not only 
contradictory to fact, but involved a self-contradiction; and was, 
therefore, easily disposed of. But this accident did not supersede the 
necessity of Galileo and his pupils verifying their assertion by refer- 
ence to experiment, since there were many suppositions which were 
different from theirs, and still possible, though that of Casrasus was 

The mistake of Aristotle and his followers, in maintaining that 
large bodies fall more quickly than small ones, in exact proportion 
to their weight, arose from perceiving half of the third law of motion, 
that the velocity increases with the force which produces it ; and from 
overlooking the remaining half, that a greater force is required for the 
same velocity, according as the mass is larger. The ancients never 
attained to any conception of the force which moves and the body 
which is moved, as distinct elements to be considered when we en- 
quire into the subject of motion, and therefore could not even propose 
to themselves in a clear manner the questions which the third law of 
motion answered. 

But, when, in more modern times, this distinction was brought into 
view, the progress of opinion in this case was nearly the same as with 
regard to the other laws. 

It was allowed at once, and by all, that action and reaction are 
equal ; but the controversy concerning the sense in which this law is to 
be interpreted, was one of the longest and fiercest in the history of ma- 
thematics, and the din of the war has hardly yet died away. The 
disputes concerning the measure of the force of bodies in motion, 
or the vis viva, were in fact a dispute which of two measures of action 
that I have mentioned above should be taken ; the effect in a given 
time, or the whole effect : in the one case the momentum {MV) in the 
other the vis viva, {MV'^) was the proper measure. 

20. It may be observed that the word momentum, which one party 
appropriated to their views, was employed to designate the motive 
quantity of force, or the action of bodies in motion, before it was 



determined what the true measure of such action was. Thus Galileo, 
in his "Discorso intorno alle cose che stanno in su I'Acqua,'" says, that 
momentum "is the force, efficacy, or virtue with which the motion 
moves and the body moved resists; depending not on weight only, 
but on the velocity, inclination, and any other cause of such virtue." 

The adoption of the phrase vis viva is another instance of the extent 
to which men are tenacious of those terms which carry along with their 
use a reference to the fundamental laws of our thought on such matters. 
The party which used this phrase maintained that the mass multiplied 
into the square of the velocity was the proper measure of the force 
of bodies in motion; but finding the term moving force appropriated 
by their opponents, they still took the same term force, with the 
peculiar distinction of its being living force, in opposition to dead 
force or pressure, which they allowed to be rightly measured by the 
momentum generated in a given time. The same tendency to adopt, 
in a limited and technical sense, the words of most general and fun- 
damental vise in the subject, has led some writers (Newton for instance,) 
to employ the term motion or quantity of motion as synonymous with 
momentum, or the product of the numbers which express the mass 
and the velocity. And this use being established, the quantities of 
motion gained and lost are always equal and opposite; and, therefore 
the quantity which exists in any given direction cannot be increased 
or diminished by any mutual action of bodies. Thus we are led to the 
assertion which has already been noticed, that the quantity of motion 
in the world is always the same. And we now see how far the 
necessary truth of this proposition can be asserted. The proposition is 
necessarily true according to our notions of material causation ; but the 
measure of "quantity of motion," which is a condition of its truth, is 
inevitably obtained from experience. 

21. It is not surprising that there should have been a good deal 
of confusion and difference of opinion on these matters : for it appears 
that there is, in the intellectual constitution and facvdties of man, a 
source of self-delusion in svich reasonings. The actual rules of the 
motion and mutual action of bodies are, and must be, obtained from 



observation of the external world : but there is a constant wish and 
propensity to express these rules in such terms as shall make them 
appear self-evident, because identical with the universal and necessary 
rules of causation. And this propensity is essential to the progress of 
our knowledge ; and in the success of this effort consists, in a great 
measure, the advance of the science to its highest point of simplicity 
and generality. 

22. The nature of the truth which belongs to the laws of motion 
will perhaps appear still more clearly, if we state, in the following- 
tabular form, the analysis of each law into the part which is necessary, 
and the part which is empirical. 





Velocity does not change 
without a cause. 

The accelerating quantity 
of a force is measured by the 
acceleration produced. 

Reaction is equal and op- 
posite to action. 


The time for which a body has al- 
ready been in motion is not a cause of 
change of velocity. 

The velocity and direction of the mo- 
tion which a body already possesses are 
not, either of them, causes which 
change the acceleration produced. 

The connexion of the parts of a body, 
or of a system of bodies, and the action 
to which the body or system is already 
subject, are not, either of them, causes 
which change the effects of any ad- 
ditional action. 

Of course, it will be understood that, when we assert that the con- 
nexion of the parts of a system does not change the effect of any 
action upon it, we mean that this connexion does not introduce any 
new cause of change, but leaves the effect to be determined by the 
previously established rules of equilibrium and motion. The connexion 
will modify the application of such rules ; but it introduces no ad- 
ditional rule: and the same observation applies to all the above stated 
empirical propositions. 


This being understood, it will be observed that the part of each law 
which is here stated as empirical, consists, in each case, of a negation 
of the supposition that the condition of the moving body with respect 
to motion and action, is a cause of any change in the circumstances of 
its motion; and from this it follows that these circumstances are de- 
termined entirely by the forces extraneous to the body itself. 

23. This mode of considering the question shews us in what 
manner the laws of motion may be said to be proved by their sim- 
plicity, which is sometimes urged as a proof. They undoubtedly have 
this distinction of the greatest possible simplicity, for they consist in 
the negation of all causes of change, except those which are essential 
to our conception of such causation. We may conceive the motions 
of bodies, and the effect of forces upon them, to be regulated by the 
lapse of time, by the motion which the bodies have, by the forces 
previously acting ; but though we may imagine this as possible, we do 
not find that it is so in reality. If it were, we should have to con- 
sider the effect of these conditions of the body acted on, and to com- 
bine this effect with that of the acting forces ; and thus the motion 
would be determined by more numerous conditions and more complex 
rules than those which are found to be the laws of nature. The laws 
which, in reality, govern motion are the fewest and simplest possible, 
because all are excluded, except those which the very nature of laws 
of motion necessarily implies. The prerogative of simplicity is possessed 
by the actual laws of the universe, in the highest perfection which is 
imaginable or possible. Instead of having to take into account all the 
circumstances of the moving bodies, we find that we have only to 
reject all these circumstances. Instead of having to combine empirical 
with necessary laws, we learn empirically that the necessary laws are 
entirely sufficient. 

24. Since all that we learn from experience is, that she has no- 
thing to teach us concerning the laws of motion, it is very natural 
that some persons shovdd imagine that experience is not necessary to 
their proof. And accordingly many writers have undertaken to esta- 
blish all the fundamental principles of mechanics by reasoning alone. 


This has been done in two ways: — sometimes by attending only to the 
necessary part of each law (as the parts are stated in the last para- 
graph but one) and by overlooking the necessity of the empirical 
supplement and limitation to it; — at other times by asserting the part 
which I have stated as empirical to be self-evident, no less than the 
other part. The former way of proceeding may be found in many 
English writers on the subject; the latter appears to direct the reason- 
ings of many eminent French mathematicians. Some (as Laplace) have 
allowed the empirical nature of two out of the three laws ; others, as 
M. Poisson, have considered the first as alone empirical ; and others, as 
D'Alembert, have assumed the self-evidence of all the three indepen- 
dently of any reference whatever to observation. 

25. The parts of the laws which I have stated as empirical, 
appear to me to be clearly of a different nature, as to the cogency 
of their truth, from the parts which are necessary ; and this difference 
is, I think, established by the fact that these propositions were de- 
nied, contested, and modified, before they were finally established. If 
these truths could not be denied without a self-contradiction, it is 
difficult to understand how they could be (as they were) long and 
obstinately controverted by mathematicians and others fully sensible to 
the cogency of necessary truth. 

I will not however go so far as to assert that there may not be 
some point of view in which that which I have called the empirical 
part of these laws, (which, as we have seen, contains negatives only,) 
may be properly said to be self-evident. But however this may be, 
I think it can hardly be denied that there is a difference of a fun- 
damental kind in the nature of these truths, — which we can, in our 
imagination at least, contradict and replace by others, and which, his- 
torically speaking, have been established by experiment; — and those 
other truths, which have been assented to from the first, and by all, 
and which we cannot deny without a contradiction in terms, or reject 
without putting an end to all use of our reason on this subject. 

26. On the other hand, if any one should be disposed to maintain 
that, inasmuch as the laws are interpreted by the aid of experience 

172 Mh WHEWELL, on the NATURE OF THE TRUTH, &c. 

only, they must be considered as entirely empirical laws, I should not 
assert this to be placing the science of mechanics on a wrong basis. 
But at the same time I would observe, that the form of these laws is 
not empirical, and would be the same if the results of experience 
should differ from the actual results. The laws may be considered as 
a formula derived from a priori reasonings, where experience assigns 
the value of the terms which enter into the formula. 

Finally, it may be observed, that if any one can convince himself 
that matter is either necessarily and by its own nature determined to 
move slower and slower, or necessarily and by its own nature deter- 
mined to move uniformly, he must adopt the latter opinion, not only 
of the truth, but of the necessity of the truth of the first law of 
motion, since the former branch of the alternative is certainly false : and 
similar assertions may be made with regard to the other laws of motion. 

27. This enquiry into the nature of the laws of motion, will, I 
hope, possess some interest for those who attach any importance to the 
logic and philosophy of science. The discussion may be said to be 
rather metaphysical than mechanical ; but the views which I have en- 
deavoured to present, appear to explain the occurrence and result of 
the principal controversies which the history of this science exhibits ; 
and, if they are well founded,' ought to govern the way in which the 
principles of the science are treated of, whether the treatise be intended 
for the mathematical student or the philosopher. 


VIII. Researches in the Theory of the Motion of Fluids. By the Rev. 
James Challis, late Fellow of Trinity College, Cambridge, 
and Fellow of the Cambridge Philosophical Society. 

[VieaA March 3, 1834-3 

1. The subjects treated of in this communication are of a miscel- 
laneous character, referring to several points of the theory of fluid 
motion, respecting which the author conceived he had something new 
to advance. In illustration of the principles he has attempted to establish, 
solutions are given of two problems of considerable interest: — the 
resistance to the motion of a ball-pendulum ; and, the resistance to the 
motion of a body partly immersed in water and drawn along at the 
surface in the horizontal direction. The principal object in the solution 
of the latter problem is to account for the rising of the body in the 
vertical direction on increasing the velocity of draught, which in some 
recent experiments on canal navigation has been observed to take place. 
In the course of the paper I have had occasion to refer several times 
to a previous communication* to this Society respecting fluid motion, 
for the purpose of giving to the views there advanced some corrections 
and confirmations which have been suggested by more mature considera- 
tion. For the sake of distinctness the subjects of the present essay 
are divided into sections. 

• Camb. Phil. Trans. Vol. HI. Part in. 
Vol. V. Part II. > Z 



On the Integral of the Equation -j-^ + -~ = Q. 

2. This equation is applicable to all problems respecting the motion 
of incompressible fluids, which require for their solutions the consideration 
of motion in one plane only. Mathematicians have obtained integrals 
of it suited to the particular questions they were discussing ; for instance, 
in solving the problem of waves propagated in a canal of uniform 
width, M. Poisson has given a value of (p, which, while it satisfies the 
equation in question, is exclusively applicable to that problem. But 
it is well known that by the common method of finding the integrals 
of linear partial differential equations of the second order between 
three variables, a value of cp may be found prior to any consideration 
of the circumstances under which the fluid was put in motion. There- 
fore any inferences respecting the nature of the motion, which may be 
drawn from this integral, must be equally applicable to all problems of 
this class. To obtain such inferences is the object of the following 

3. The integral I speak of is. 

To ascertain its general signification, I propose to determine the forms 
of the functions F and jf, independently of any hypothesis respecting 
the mode in which the fluid was put in motion. The quantity (j) is 
subject to the condition {d(p) = udx-^vdy, where u and v are the 
velocities at the poiiit xy in the directions of the axes of x and y 

respectively. Hence ^^=«, -j^=v, and 

u = F' {x + y^/'^\)+f {x-y^/'^), 
v=V^lF'ix + y\/^^)-\/'^fix-yV^l). 


First, it may be observed that u and v are not both possible for any 
values of x and y, unless the functions F' andy be the same. Again, 
as the form of F' we are seeking for is to be independent of all that 
is arbitrary, it will remain the same whatever direction we arbitrarily 
assign to the axes of co-ordinates. Let therefore the axis of y pass 
through the point to which the velocities u, v, belong. Then 

y = 0, u = 2F'{x), v = 0. 

If now the axes be supposed to take any other position, the origin 

remaining the same, u will be equal to / ^ ^ F' {^/x^ + y^). 


F'{x + y^^) + F'(x-yV-l)=-^^y^^=^.F'(./^FTr), 

a functional equation for determining the form of F'. Let 

x + yy^ - l=m, and x — y^/— 1 =n; 

2x = m + n, and "s/ x^ -k-y^^s/ mn. 


It is easily seen that if F\y/mn) =— f=, the equation is satisfied. 
Hence ^ 

^ = ^-7=+ ^== = A^ and^--^^. ^-^ 

dx x + yV^l^ a;-yV-l x'^ + y^' dy~x^ + y'-' 

2C -^ - 

and consequently the velocity at xy, or \/ u'^ + v^ = —-r=^ ^- ,., ,, 

These results shew that the velocity is directed to or from the origin 
of co-ordinates, and varies inversely as the distance from it. But we 
must observe that this limitation as to the point to which the velocity 
is directed, is owing to the particular forms, x+y'\/~^, x-ys/ — \, 



of the quantities which the function F' involves. For the equation 
^j% + -T-? = 0, is also satisfied by the following, 

((> = F {{a^- X cos d - y m\ Q) + {^ + xsm9 +y cosO) ^/^^} 

+y {(" + a; cos 0-y sin 0) - (/3 + a; sin f y cos 9) V~^\ '■> 

and this analytical circumstance has its interpretation in reference to 
the motion of the fluid. By supposing the function f to be the same 
as F, and giving to F' the same form as before, we shall find, 

' d^ _ 2 C(.r + g cos g + /3 sin 9) 

dx ~ (a + xcos9 — i/ sin 9y + (13 + x sin 9 + y cos 9)- 

d^ 2C(y + /3cosg-as in 9) 

dy {a -\- X cos 9 — y sin 9'f + (/3 + a; sin + y cos 9Y 

/d^y /d^Y 4C^ 

\dx ) \dy) ~ {a + x cos9-y sin9y + (^ + x sm9 + y cos9y 
Or, if a cos 9-\-fisin9= —a, and /3 cos 9 — a sin 9= —b, 

d(f> 2C(x-a) 

dx (x — ay + {y — bf ' 

^ or V- ^^(y-*) 
dy ^x-ay + iy-by 

Vu' + v^ = 

y/ix-af + iy-by' 

This shews that the velocity is directed to the point whose co-ordinates 
are a, b, and varies inversely as the distance from it. And as we have 
arrived at this result without considering any circumstances under which 
the fluid was caused to move, the inference to be drawn is, that such 
is the general character of the motion. Nothing forbids our considering 
C, a, and b, functions depending on the time and the given conditions 
of motion in any proposed problem. Also if at a given instant, a line 
commencing at any point, be drawn continually in the direction of the 
motions of the particles through which it passes, C, a, and b, may be 


supposed to vary in any manner along this line. The foregoing- 
reasoning only proves that in passing at a given instant from one 
point to another indefinitely near along the line, these quantities may 
be considered constant. 

4. The nature of the integral we have been discussing will perhaps 
be understood by comparing it to the general integral of a common 
differential equation, which has a particular solution. The latter, we 
know, is that which gives the answer to a proposed problem, and the 
general integral is used (though not necessarily) to obtain this solution. 
So, I conceive, the integral above is useful for arriving at the particular 
functions of x, y, and t, which give the velocity and direction of the 
velocity at any point and instant in any proposed question. The 

integral of -—^ + -r^ = , which M.M. Poisson and Cauchy have 

obtained for the solution of the problem of waves, may be called the 
particular solution of the equation, for that particular problem ; and I 
think it probable that the same might have been obtained by employing 
what I would call the general integral, though I am not prepared to 
state exactly the process. 

5. The following considerations are added in confirmation of the 
foregoing reasoning. In whatever manner the fluid is put in motion, 
we may conceive a line, commencing at any point, to be continually 
drawn in a direction perpendicular to the directions of the motions at 
a given instant of the particles through which it passes. This line 
may be of any arbitrary and irregular shape, not defined by a single 
equation between x and y. But it must be composed of parts either 
finite or indefinitely small, which obey the law of continuity. Con- 
sequently the motion, being at all the points of the line in the directions 
of the normals, must tend to or from the centres of curvature, and 
vary, in at least elementary portions of the fluid, inversely as the 
distances from those centres. An unlimited number of such lines 
may be drawn through the whole extent of the fluid mass in 


6. If we put (f} = (pi + (p2 + (f>3 + Sac. we shall have 

d^ d^_ (d^ d^\ Id^ dy,\ id^ d-^<pA , . _ „ 
daf "*" df ~ \dx' ^ dfj '^ \dx' "^ df) ■*■ \dx' ^ df)'^^-~^- 

Hence if there be any number of functions which severally satisfy the 
given equation, the sum of these will satisfy it. But from what has 
been proved above, if 

d^i Ci(x — a,) 

d(p, _ C:(y-/3,) 

dx ~ (ar-aO' + (y-/3,)" 

dy (x-a,y + (y-l3,r' 

a 02 Cs {x — aj) 

d(j>, Q{y-fi,) 

dx ~ ix-a,r + oj-(i,y' 

dy (^x-a,f + iy-fi,Y' 

&c. = &c. 

&c. = &c. 

<p\, 02. 03, &c. will severally satisfy it; therefore 0i + 02 1- 0, + &c. 
will also. And we have, 

dx dx dx 

v=^ +^ +^ +&,c 
dy dy dy 

These equations prove that the velocity at any point may be the re- 
sultant of several velocities produced by different causes; and that any 
given cause will have the same effect in producing velocity at a given 
point, whether or not other causes may be operating to produce 
velocities at the same point. 

7. We may here also determine the manner in which the motion 
of the fluid is affected, when the rectilinear transmission of an impulse 
tending from any centre is interrupted by a plane surface. For suppose 
two impulses tending from two centres to be of equal magnitude and in 
every respect alike. Then if the straight line joining these centres be 
bisected at right angles by a plane, there will be no motion of the par- 
ticles contiguous to the plane in a direction perpendicular to it, because 
the resultant of the velocities from the two causes must lie wholly in 


the plane. Hence as the division of fluids* may be effected without the 
application of force, nothing will be altered if we suppose the plane to 
become rigid and to intercept the communication of the fluid on one side 
with that on the other. The motion on each side will then be reflected, 
and the angle of incidence will be equal to the angle of reflection. - 

8. I propose now to adduce an application of the proposition 
above demonstrated (Art. 3.) respecting the general law of fluid motion, 
which may serve to shew its utility. Suppose water in a cylindrical 
vessel (for instance, a glass tumbler,) to be caused to revolve with con- 
siderable rapidity about the axis of the cylinder. There is no practical 
difficulty in making the fluid revolve so that every particle shall de- 
scribe approximately a horizontal circle about the axis. Then, the fluid 
being left to itself after the disturbance, each particle may be considered 
to move as it does, by reason of a centripetal force tending to the 
axis in a horizontal plane. This force must be owing to the action 
of the cylindrical surface on the fluid particles in contact with it, 
deflecting them continually from a rectilinear course. If V be the 
velocity of the particles in contact with the surface, and a the radius 

of the cylinder, the force tending to the axis is — , the effect of 

friction being neglected. The deflections which this force is continually 

producing in the directions of radii, are transmitted through the fluid, 

and as they tend to a centre, will vary, according to the proposition 

above proved, inversely as the distance from the centre.f Hence the 

V^ a V^ 

centripetal force at the distance r is — x -, or — . This shews 
^ a r r 

that at any distance r the velocity is still V. Experience seems to 

confirm this result. For if light substances be strewed on the surface 

of the water, those nearer the centre always perform their revolutions 

* The introduction of this consideration here is merely reverting to a principle -which 
Professor Airy (very properly, I think,) has proposed to make the basis of the mathematical 
treatment of fluids. Without referring to a principle of this nature, I do not see that 
problems of reflection can be satisfactorily solved. 

+ The total motion is compounded of these deflections and rectilinear motions along 
tangents to the circles, which by Art. 6. may be considered separately. ''' ' •"'-'■'■ 


in less time than those more remote. This is particularly observable 
in two of the floating particles which are near each other, and at nearly 
equal distances from the centre. That which is less distant overtakes 
the other, as it ought to do, supposing it to describe a less circle with 
equal velocity. At the centre a kind of eddy is formed, the more 
observable as the motion at every point of the surface is more nearly 
in concentric circles. When the revolving motion takes place in a 
conical tunnel from which the water is issuing, the appearance at the 
axis is very remarkable, a hollow space like a sack, being formed a 
considerable way down the axis. What has been here said may serve 
to explain in some measure the manner in which eddies in any case 
are produced. 


On the Integration of the Equation -^ + -r^ + -r^ = 0. 

9. M. Poisson has expressed the general integral of this equation 
by means of definite integrals ; {Memoires de rAcademie des Sciences, 
Ann. 1818), and this, I believe, admits of a discussion similar to that 

applied above (Art. 3.) to the integral of -^ + -~ = 0. But perhaps 

the following reasoning, analogous to what was indicated in Art. 5., 
may be considered sufficient. In whatever manner the fluid is put in 
motion, we may conceive a surface to be described, which shall be 
every where perpendicular to the directions of the motions at a given 
instant of the particles through which it passes. This surface may be 
of an arbitrary and irregular shape, not necessarily defined by a single 
equation between x, y, and %. But it must be composed of parts either 
finite or indefinitely small, which are continuous, and consequently have 
radii of curvature subject to the same conditions as those of regular 
curve surfaces. Hence the normals to all the points of any element 
of the surface will pass through two focal lines, situated at the centres 
and in the planes of greatest and least curvature, and cutting the 


directions of the normals at right angles. The motion, being in the 
normals, will be directed to the focal lines. If we describe another 
surface indefinitely near the first, and cutting in like manner the direc- 
tions of the motion at right angles, all the points of any fluid element 
intercepted between two opposite elements of the surfaces, will at a 
given instant ultimately have their motion directed to the same focal 
lines : but this cannot be said in general of more than an elementary 
portion. If we suppose the form of the superficial element to be a 
rectangle, the normals through all the points of its sides, will inclose 
a wedge-shaped mass, the transverse section of which at any point, it 
is easy to shew, will vary as the product of the distances of that point 
from the focal lines. Hence the velocity in passing at a given instant 
from the first to the second of the surfaces above-mentioned wiU vary 
inversely as this product. Let therefore r and r + l he the distances 
of the point whose velocity is V, from the focal lines to which the 


motion is directed. Then V= . j-, in which expression C, /, and 

the positions of the focal lines are constant at a given instant, when 
r varies through a space which may either be finite or indefinitely small. 
Let a, /3, 7, be the co-ordinates of the middle of that focal line which 
is distant by r from the point in question. The velocity (m) in x will 

then be V. ; the velocity {v) in y, V. ; and the velocity 

{w\ in », V. ^. Hence 

' r 

udx + vdy + wd%= Vi — ~dx + ^ , dy H -d%\ . 

Now since r- = {x — af + {y — fif-\-{%~yY, if we make r vary with 
X, y, and %, while a, )3, 7, remain constant according to what has just 
been said, we shall have rdr — {x — a)dx + {y-fi)dy + {% — y)dti. Hence 
tfdx + vdy + wdz=F^dr; and as F" is a function of r and /, the right 
side of the equation is a complete differential of a function of 
X, y, %, and t, with respect to the three first variables, t being con- 
stant. Therefore also the left side is the same. Let the function be <p. 
Vol. V. Part II. A a 



dr ' dx ' dy ' d% 

We proceed to shew next that the equation 

d^d) d^d) d^d> „ du dv dw 

zr^ + j^ + -j^ = 0, or J- + T- + ^- = 0, 
daf df d%^ dx dy dx 

is satisfied by the kind of motion we have been describing. 

10. Let P (Fig. 1.) be the point whose motion we are considering; 
Or, Nq, the focal lines to which the motion of the element at P is 
directed. Let PNO be the straight line which passes through P and 
the focal lines, cutting them in N and O. Suppose O to be the 
origin of a system of axes, of which ONP is the axis of x, Oy coinciding 
with the focal line Or the axis of y, and 0% perpendicular to the plane 
yOx, the axis of %. The co-ordinates of P referred to another system 
of rectangular axes AX, AY, AZ, are X, Y, Z: p is a point 
indefinitely near to P, Pp is parallel to AZ, and the co-ordinates of 
p are X, Y, Z+SZ: pqr is the straight line which passes through p 
and the focal lines cutting them in q and r. Now let the equations 
of Pp referred to the system Ox, Oy, 0%, be x = a% + a, y — b% + fi, 
and the equations of pqr, x = dz-\-a, y = b'z + li'. Then 

„ l+aa' + bb' 

cos ^ Ppq = „ , — . 

Let ON=l, NP=r. Hence because Pp passes through P whose 
co-ordinates referred to the axes Ox, Oy, Oz, are I + r, 0, 0, it follows 
that l + r = a, and /3 = 0. Thus the equations of Pp become x = az + l + r, 
y = bz. Again, because the line ^^gr passes through r, whose co-ordinates 
are x — 0, z = 0, we have a' = ; and because it passes through q, whose 
co-ordinates are y = 0, x = l, we have l=a'z, and = i's: + /3'. Hence 

a: = - = - -n, and consequently ft' = r. Thus the equations of 

pqr become x — a'%^ y^V% y . Also because Pp and pqr pass 


through the same point p, a; = «'x = a« + /+r, and therefore ^ = -7 . 

And y = hz = h'% ^i therefore z = -rm — tx • Hence -; = —rrr, — 7t> 

'' a' a{b'-h) a -a a{b-b) 

which gives h' = — ; j^. From p draw ps perpendicular on Ox, 

and let P.? = 5. Then ^ = x-{r + l). Bnt x = a'z = t!^±Il, Therefore 

' a —a 

^ = — 7 '- . Hence it will be found that a' = — — » , and 

a —a 6 

V = — J — -. This latter quantity, if we neglect powers of S above 

the first, is equal to (l H — rj A. Therefore by substitution 

„ d r \ r{l + r)J 

cos / Jr«o = — -. . , , ,. 

a'(^r + l)+(l+d' + b'—]s 
= (neglecting ^, &c.) V ^ / 

Here / „ === is the cosine of the angle pPs. Hence if ^ = the 
V 1 + a^ + o^ 

velocity at P in Ox, and w the part resolved in the direction parallel 


to AZ, w — — / 2~ ^i • ^^^ ^ ~ ^^ resolved portion of the velocity 

at p in the same direction. Now the velocity at p is ultimately the 

same as that at s, and is therefore equal to V . -, A^ \ — r- , 

according to the law of variation from P to s determined . by the 




considerations with which we commenced this investigation. Neglecting 

powers of I above the first, this quantity becomes V \\ ^ J . 



But S = SZcospPs = SZ X , 'g.J p' Hence 

w'-w _ „ / \-d 1 ¥-a^ 1 

~IZ U +«' + *'■ / + r "^ l+«2 + fr^'r 

If now a, ft, 7, be the angles which the axis AZ makes with 
Ox, Oy, 0%, respectively, we have 

Hence passing from differences to differentials, 

-7- = (COS^'V — COS'o)^ + (cOS^/3-COS*a) - (1). 

d% ' ' l + r ' r ' 

So if d, /3', 7', be the corresponding angles for the axis of Y, and 
a", /3", 7", for that of ^, v the velocity in F", and u that in ^, we shall 
have by a like process, 

^ = (COSV - COs'a') ,— + (coS^/3' - COS^a') ....... (2) , 

^ = (cos^y- COs'a") y^ + (cos'/3"-COS^a") ^ (3). 


But as a, a, a", are the angles which Ox makes with three rectangular 


cos" a + cos" a + COS" a" = 1, 

so cos-/3 + cos^/3' + cos^/3" = l, 

and cos'^7 + cos'^7' + cos^7" = l. 

Therefore by adding the equations (1), (2), (3), 

du dv dw _ 
7lX^ dY^dZ~ 

11. The general conclusion from all that precedes is, that the law 
of the variation of the velocity from any point to another indefinitely 
near in the direction of the motion, at a given instant, may be expressed 


by -^ — J-, the quantities C, r, and I, being such as we have stated 


in Art. 9- If 1=0, we have- as a particular case, V=-^. In my 

former paper on the motion of fluids, I assumed, as it now appears, 


incorrectly, that — represents the general law of the variation of the 

velocity. None, however, of the results in that paper are affected by 
the assumption. For instance, the expression for 

as it only requires that (p should be a function of r and /, will remain 
the same. This expression may also be briefly obtained thus. We 

have seen that -~- = V. Now as r is ultimately in tlie direction in 

which the velocity V takes place, if a line commencing at a given 
point be drawn constantly in the direction of the motion at a given 
instant of the points through which it passes, dr may be considered the 
increment of this line. Hence if we call its length s reckoned from 

the fixed point, -j^ = -^ = F. Then integrating, c}> = jVds -^/{t); 


and differentiating under the sign /, ^ = f -r-^^ +J''(^)- Hence 
substituting for -^ in the known expression for the pressure {p), 

p = f(Xdx + Ydy + Zdz) - f^ds - ^ -fit). 

If f^ be always the same in quantity and direction at the same point, 

dr V^ 

-^ = : so that, p = j{Xdx + Ydy + Zd%) - -— -f{t). 

This equation may thus be considered to be strictly deduced from the 
general equations of fluid motion. 

Considerations analogous to those applied (Arts. 6 and 7) to motion 
in a plane, might here be introduced to shew that the motion at any 
point, when due to several causes, is the resultant of the motions which 
would be produced by the causes acting separately ; and also to determine 
the same law of reflection at a plane surface. 

12. The following simple instance of fluid motion may serve to 
illustrate some points of the preceding theory. BCD (Fig. 2.) is a 
conical vessel with its axis vertical. A mass of fluid DBhd is made 
to descend so that its lower surface hd is bounded by a horizontal 
plane to which any arbitrary velocity is given. The upper surface is 
also supposed to be plane and to be kept horizontal by the force of 
gravity. It is required to find the consequent velocity and pressure 
at every point of the fluid. 

It is evident that the motion will be in vertical planes passing through 
the axis, and will be, the same in all such planes. Take therefore two 
planes making an indefinitely small angle with each other, and let 
AB, AE, be their intersections with the upper surface, ab, ae, with 
the lower. Let PQSB be an element of the upper surface, P and B 
being equidistant from A, as also Q and S. If now at any instant 
lines commencing at the four points P, Q, B, S, be continually drawn 


in the direction of the motion at the points through which they pass, 
these lines must be rectiUnear, because there is no curvilinear motion 
at the boundaries of the fluid, and therefore no cause to impress a 
curvilinear motion on the parts interior. The straight lines commencing 
at P and R will intersect ah and ae at p and r, points equidistant 
from a, and those commencing at Q and S will intersect the same 
lines at q and s also equidistant from a. Now from the law of 
the variation of the velocity above found, at every point of the cunei- 
form element Ps, the velocity will be inversely proportional to its 
transverse section. Let therefore V =^ the vertical velocity with which 
(lb is made to descend, and v the vertical velocity with which the 
surface DB descends. Let AB = a, AQ = x, PQ = X, ah = h, aq = x', 
pq = 'S.', and the angle BAE = e. Then the element PQSR^^xeX, 
and pqsr = x'e\'. These elements are proportional to the transverse 
sections at P and p ; and the vertical velocities V, v, are to each 

other as the velocities at p and P in Pp. Hence — = -; — , = -V-, • 

-' ■* V X e\ xX 

F • Wence ^, - ^ 

because the motion is along the slant surface. Therefore in this case 

X a 

r-, =" T. Suppose X to be given, and let Xi be the consequent value 

X o 

of x'. Then — = -r, and -. = y . If now x be taken = « — X, from 

X, o b-Xi o 

what has been just shewn, x' will = 6 — X, . Hence 4 — = j^, and 

\0 — Xi) X2 o 

consequently — = t- Therefore X2 = Xi; and so on. From this it 

X2 o 

follows that if AB and ab be divided into the same number of 
indefinitely small equal parts, the straight lines joining the corresponding 
points of division will give the directions of the motion, which is 
consequently every where directed to the vertex of the cone. Hence the 

velocity af^ any point W whose distance Cp W from C is p, varies as — . 

Let CA =h, Ca = k, z AC W= 9 ; then the velocity at p=V sec 9, and 

the velocity at W= V sec 9 . ^ — ; this resolved in the vertical 

But — also = jj, . Hence ^j^, = 71 • If we take x = a, x' must = h. 


f^k'sec'6 VTf 
direction gives j— — , which = — — = velocity at Z. Hence the 

vertical velocity is the same at all points of any horizontal plane, and 
the fluid will consequently descend in parallel slices.* Let us now 
determine the pressure at any point on the particular supposition that 
V is uniform. Then if 

Vk"&eee ^, , .. , „r dw Vsec'd ^,dk 2F"Asec'0 

w = :: the velocity at W, -7- = ; — x 2«-7-- = y, . 

p' ■'at p^ at p- 


Idt^' ^-Jdt'^P-J 7 = -p + ^ 



„ 2r'kse&9 r'k'sec'e 
p = C-g. + . __. 

And as when z = h, p — 0, and p cos Q = h, it follows that 

The above solution I do not consider to be of any value, except as 
illustrating the process to be followed in determining mathematically 
the way in which the interior of a mass of fluid is affected as to 
velocity and pressure, in consequence of given conditions at the 
boundaries. This part of the theory of fluid motion is very 

* I obtained this result in the number of the Phil. Mag. and Annals of Philosophy 
for .Jan. 1831, but omitted to shew that it is entirely dependent on the arbitrary condition 
that the inferior -surface of the fluid is bounded by a horizontal plane. Qji any other 
supposition the problem would be one of much greater difficulty. This omission has not 
without reason caused a misapprehension as to the application of the solution, on the part of 
Berzelius in a notice taken of it in his Annual Review. {Jahres-Bericht, 1833.) 



Application of the Principles of the foregoing Section to an instance of 
the Resistance of an Incompressible Fluid to a Body hounded by a 
Spherical Surface moving in it. 

13. Let a solid sphere, partially immersed in water, being of less 
specific gravity than the fluid, be drawn along in a horizontal direction 
with a given uniform velocity ; it is required to find the height of 
its centre above the horizontal surface of the water. 

We shall suppose for the sake of simplicity, that the fluid is 
unlimited in extent both in the vertical and horizontal directions, and 
that the surface of the sphere is so smooth that it impresses no velocity 
on the water in contact with it in the direction of a tangent plane. 
Let CDJBJE (Fig. 3.) be the sphere, O its centre, ADE the intersection 
of the horizontal surface of the fluid by a vertical plane through the 
centre of the ball; OQ a line through the centre parallel to ADE. 
This will be the direction of the motion of O, since the velocity is 
supposed to have become uniform, and ON to be constant. Let A, 
a fixed point in ADE, be the origin of co-ordinates, AN=a, NO = 'y, 

at any instant. Then the velocity {V) of O = -r-. Draw OB vertical; 

let P be any point of the surface immersed; through P draw the 
spherical arcs PQ, PB, and let the angle QOP=6, and the angle 
PQB = to. The velocity impressed by the sphere on the fluid at P 
is F'cos 9, as none is impressed in the direction of a tangent plane. 
This velocity is directed to the point O, because in the case of a 
spherical surface / = 0. Hence if « = the radius of the sphere, 


FcosO = — . (Art. 11.) The velocity at every point of the line OP 

produced, wiU at a given instant be in the direction of this line, 
Vol. V. Part II. B b 


because when the fluid is of unlimited extent, there is no cause* to 
produce motion at any point of the line, but the impression made at P, 
which is transmitted instantaneously, varying at different distances 
from O according to the law of the inverse square. Hence if ^ be a 

point in OP produced, and OR = r, the velocity at R — —, = ^ — . 

Let ADE be the axis of x, a vertical through A the axis of z reckoned 
positive downwards, and a line through A perpendicular to the plane 
of these two the axis of y. Then if the co-ordinates of R be x, y, %, 

we shall have r- = (:r — a)" -I- y^ + (s + 7)' ; and cos0= . Therefore 

the velocity {v) at R, 

•A-A I'? 


Ka;-ay^ + y' + (» + 7)-}5" 



dv dv da re rr :i 4. ..k 

~j- = 7- • 77 > (lor ^ and 7 are constant), 

_ F«^(3cos-e-l) (la 
r" ' dt 



Therefore, gravity being the only force acting on the fluid, the pressure 
ip) at R, 

* This cannot be said of the parts of the fluid adjacent to the radii produced which pass 
through the circle in which the surface of the water meets the surface of the sphere, because 
the water outside of the conical surface formed by these radii must be put in motion by that 
within by reason of the difference of pressure occasioned by the motion. On account of the 
difficulty of estimating this effect, it is left out of consideration in our solution, which can 
therefore be only considered approximate. 


= ^« + -27^(3008-'^- 1) - -g^-cos'e -f{t). 

When r is indefinitely great this equation becomes p=g^—f{t)', and 

as for this value of r the velocity = 0, p must = g% ; therefore /{t) = 0. 

If now we put r = a, and i8 = ss,, the co-ordinate of P, we obtain the 

?^* cos 20 
pressure (/>,) at P, = gz, -\ . The portion of this resolved in 

the vertical direction is jo, x cos i FOB. But from the spherical 
triangle PQB, cos /. FOB = cos w sin 9. Therefore the vertical pressure 
is p, cos w sin 6. The element of the surface at F = ad9 x a sin ddw. 
Hence the whole vertical pressure = //jOia'sin''^ cos wt/^c^w 

=ga^ff%i sin^OeoswdOdw + ——— ff sin^ 9 cos29 cos wd$dw. 


The first term is plainly the weight of water displaced, and is there- 
fore equal to — -(2«' — 3«'7 + 7*), the specific gravity of the water 
being 1. The integrations with respect to u> must be taken from 
a,= — cos"^ — jT—r to -f cos"^ — ^-;r , and the integrations with respect 

to 6 from sin"'— to the supplement of that arc. Between these limits 

of w, fcoswda) = 2\/i T ; and between the limits of 9, 


2fsm'9cos29d9 x/i _ , '^^ =_ J fi _:>:!') . 
•^ ^ «*sin''0 2 V «V 

Therefore if JV = the weight of the sphere, which is the same as the 
whole vertical pressure, and w = the weight of fluid displaced, 

IV =w- 


B B 2 



This result shews that the weight of fluid displaced is greater than 
the weight of the sphere, and consequently that the centre O is lower 
than it would be in a state of rest. 

Suppose a portion of the sphere to be cut off" by a horizontal 
section at the distance of b from the centre ; and let 7 become 7', the 
centre being still above the surface of the water. Then if we suppose 
the motion to be always in the direction of the radii*, and the horizontal 
bottom to have no effect in impressing motion, the equation for this 
case will be. 



= w : — 

The difference between W and w is here less than before on account 

of both the factors —; and 1 — -yr ; for -?- is greater than - . This 

a* b* b ^ a 

seems to shew that curved bottoins tend to depress the vessel when it 

begins to move, and consequently to increase the resistance. 

As the term —-— ff sm^6eos26 cos uidOdu) is positive from 0=:sin"' — 
to = 45°, and from = 135° to 6 = the supplement of sin"' — , let us 


integrate for the portion of the surface corresponding to these limits, 
or what amounts to the same, take the double of the integral between 
the first limits, those of w remaining the same as before. In order to 
abstract from the consideration of the portion of the surface not taken 
into account in this integration, we may suppose the portions for 
which we integrate to be connected by a cylindrical surface, the radius 
of which = a sin 45°. The length of this cylindrical part may be any 
we please : the vertical pressure against it will be only equal to the 

* This again cannot be true in the direction of the radii which pass through the lower 
circular boundary of the surface. 


weight of fluid displaced. Also the shape of the floating body above 
the part immersed is of no importance to the problem. The form of 
the whole body may be such as is described in Fig. 4, ABCDEF 
being a half cylinder of which the axis is GH, and ALC, FKD, the 
extreme portions of the body, bounded by spherical surfaces which have 
their centres at M and N. Now in general ^ jjsin^d co^^O co& wdQdw, 

commencing at = sin~'— , and ending at any other value of 9, will 

be found to be 

cosefssin^e + l-^') V sin^e-^ -\(\-—i 

And if we put cos 9 = — ?= , we shall have 




W = w + 

As the second term is necessarily positive, the floating body will be 
higher than it would be in a state of rest, and consequently the 
surface against which the resistance acts becomes less by an increase 
of velocity. 

To obtain a numerical result respecting the rise of the body 
corresponding to a given velocity, we will suppose for the sake of 
simplicity of calculation that when the vessel is at rest, the centres 
of the spherical ends and consequently the axis of the cylindrical part, 
are in the plane of the horizontal surface of the water. This circum- 
stance may be produced by loading the upper part of the body 
. without altering its specific gravity. Let / = the length of the axis 
of the cylindrical portion. Then the area of the horizontal section of 

the vessel at the level of the water surface is ID H ■— — , its 

4 2 

breadth being D. Now W—w must be equal to the difference of the 


quantities of fluid displaced in the states of rest and motion, and is 
therefore equal to yg \ID+'^ — j , 7 being small. Therefore 

neglecting powers of — above the first, 

Let ^ = 3. It will then be found that F' = 696** x 7. And if 7 = one 

inch, or ^, this equation gives ^=519 miles per hour; consequently 
if ^=10*4 miles per hour, 7 = 4 inches. 


In general, neglecting "—, &c. 



sin e cos e (2 sin'0 + ^ ) ~ |) ' 

also W — w = yg llD + ^---iAO - sin 9 cos6)\ nearly ; 

therefore, as I> = Zasm9, it will be found that 

F- sin2 0(2sin'0 + l)-0 , . ^ / 

y = -r- •'-4 '--TT, — • r.n r,n y w? bemg put for -y\. 

' 4!g 4!msm'9-sm26 + 29 " * D 

If 9 be assumed equal to 15°, and 711 = 3, this equation gives ^"=7-35 
miles per hour when 7 = 4 inches. 

These results, which probably are but very rough approximations 
to matters of fact, may yet suffice to shew that when vessels and boats 
of the usual forms sail in the open sea, they may be expected to rise 
in some degree upon an increase of their velocity, and so much the 
more as they are less adapted to cleave the water. Our theory shews 
that the rise is the same for bodies of the same shape and proportions, 
moving with the same velocity, whatever be their absolute magnitudes; 
also that this effect is equally due to the pressures on the front and 


stern of the vessel. The theory, in fact, determines these pressures to 
be in every respect alike, so that if we proceeded to investigate the 
total pressure in the horizontal direction, we should find it to be 
nothing, when the motion is uniform. This may serve to shew that, 
if friction be left out of consideration, a front ill-adapted to cleave the 
water, is not unfavorable to speedy motion, if the stern be of the same 
shape; and that the resistance to the motion of vessels in the open 
sea is principally owing to the friction of the water against their 
surface. This cause operates to produce unequal actions on the front 
and stern, making the directions of the motions of the particles in 
contact with the surface of the former, less inclined to the horizon 
than they would be in the case of no friction, and of those in contact 
with the surface of the latter more inclined. To counteract this inequality 
probably the stern should be less curved than the front. 


General Propositions respecting the Motion of Compressible Fluids. 

14. The considerations applied at the beginning of Section II. to 
incompressible fluids, are equally applicable to compressible. I shall 
therefore assume that in a mass of fluid in which the density varies 
as the pressure, the directions of the motion at all the points of any 
element pass at a given instant through two focal lines. Let p be 
the density at a point distant by r and r -vl from the focal lines, and 
V the velocity : p and V the same for a point indefinitely near the 
former. Also let the transverse section of a cuneiform element aclk 
(Fig. 5.) which is bounded by four pli.nes passing through the focal 
lines kl, mn, be at the first point efgh, and at the other, abed. The 
pressure and consequently the density will be the same at all points 
of the section eg; as also the velocity; at least our reasoning does not 
apply to cases in which this condition is not fulfilled. The same may 
be said of the section ac and of all sections intermediate to ac and eg. 


Let now the area of eg = m, and that of ac = m'. Then if the motion 
which exists at a given instant, be supposed to be continued uniform 
for the small time t, the quantity of fluid which passes the section eg 
in that time, is mpF^T, and that which passes ac is m'p'Vr. Hence 
the increment of matter between the two sections is — {m' p'V'T — mpVT), 
whether the velocity tend from or to the focal lines, being considered 
negative in the latter case. The increment of density {Ip) of the element 

in the time t, is consequently — ^^ — - — r—, — —. — — ■ But — = — ^^ =^ . 

^ •' m{r'-r) m r{r + l) 


pT'r'ir' + D-prnr + l) _^^^^^^Sp_^ 

And passing from differences to differentials, 
^^^^^'dt ~ dr 


dp dV ,.dp ,^ /i 1 \ 

As before udx + vdy + wd% = V dx + V^^ — — dy + V d% = Vdr, 

" f* T T 

if a, /3, 7, be the co-ordinates of the middle point of the focal line hi. 
Now as we have supposed that in passing from one point to another 
of tlie element acge, the change of velocity at a given instant depends 
only on the change of r, we may consider V a function of r and t, 
and Vdr a differential of a fimction of r and t. Then udx ^ vdy 
+ wdfi = d(l), a complete differential of a function of x, y, and as; and 

-~ = V. But in this case we have the known equation, 

a' Nap. log. p^fiXdx + Vdy + Zd%) - ^ - ~ (B.) 

Therefore considering X, V, Z, to be independent of the time, 

d'dp _ d^(p dV d'(ji d(p d'<p 

pdt ~ df dt ~d¥ ~ 'dr ' drdt ' 


But from (A), 

pdt pdr ' dr dr' dr \r r + l) ' 

And differentiating (B) with respect to space only, 

^1^ = Xdx+ Vdy + Zdx-d.^ - VdV. 
p at 

If the variation be from one point to another in the direction of the 

motion, dx = dr, dy = - — — dr, dz = dr. Hence, 

r ^ r r 

a\dp ^ X ^-° , Y y~^ + Z ^^^ _-^ d(p d'(f> 
pdr ' r ' r ' r drat dr ' di^ ' 

Substituting this value of — ^ in the foregoing equation, and then 

equating the two values of ,'] , we shall obtain, 

/ d£\d^_Q^ d^ d^t . ,.d^(l , J_\ 

\ ~ dt^j dr' dr ' drat df "*" dr\r "^ r + l) 

+ ^ (x^^ + ry^ + Z'-^) =0 (C.) 

dr \ r r r I 

This is an equation of general application. If, as a particular case, 
I, a, /3, 7, each = 0, we shall have the equation I obtained in my 
former paper (Art. 4.) by assuming ^ to be a function of v^a^ + y^ + s!^ 
and t in the equation {n) of the Mecanique Analytique (Part II. 
Sect. XII. Art. 8.) 

It may be proved as in Art. 11, that -^ = /~77 ^*' ^^ ^'^^ incom- 
pressible fluids, and that the equation applicable to steady motion is, 

a' h. 1. p = fiXdx + Ydy + Zd%) - ^ + fit) . 
Vol. V. Part II. Cc 


15. If r be indefinitely great in equation (C), the motion is in 

parallel lines, and putting r = c + s, j=j- Let -^ = w, and 

suppose no force to act ; the equation for this case becomes 

d''(p 2w (Pep 1 d'(p _ 

~d? ~ '^^' ■ dsdt "^ o^^T^ ' dF~^' 

This equation combined with a* N. 1. p = — -^ — — , gives as a particular 

integral, u] = al:iA. p =/"{«- {a + w)t\. By varying a little the mode of 

_ ^( as. 

integrating, I found also w — a^A. p =/( ■ atj, {Camb. Phil. 

Trans. Vol. III. Part III. p. 399), and endeavoured to shew the way 
in which each integral ought to be applied. But this enquiry was 
unnecessary ; for the integral may present itself under an unlimited 
number of different forms. The equations 

w = a^.\.p=f{.^-{a + io)t + ^{w)], or «, = «N.]. jo=/(^^^^%i^l , 

will equally satisfy the differential equations, being, in fact, only 

different forms of the first-mentioned integral. The principle according 

to which it now appears to me, an integral of this nature should be 

employed, is to apply it immediately only to the parts of the fluid 

immediately acted upon by the arbitrary disturbance, in order to 

determine the law according to which the initial velocity is transmitted 

to the contiguous parts ; then to determine the law of transmission 

from these to the next; and so on in succession. In the present 

instance by making * and t vary so that w and p remain the same, 

we shall find a + w for -j~ the velocity of transmission, under whatever 

form the integral may appear. The second term m of this quantity 
is due to the transmission of velocity through space by the motion of 
the particles themselves ; the other a is the velocity of propagation 
along the particles. In this example, as the velocity and density are 
propagated uniformly and undiminished, it is easy to determine at any 


instant the velocity and density at any given point, which result from 
a given disturbance. In other cases in which the velocity of propaga- 
tion is variable, the determination would be more difficult, but must 
probably be arrived at by the same principle of reasoning. Variable 
propagation is analogous to variable motion, as uniform propagation to 
uniform motion, and would seem to require integration to determine 
the time at which the effect of a given disturbance is felt at a given 

16. If in the equation (C), a be an indefinitely great quantity, 
the terms which do not contain a^ as a factor may be neglected in com- 
'parison of those which do, and the equation will become 

dr^ ^ dr\r ^ r + l) ' 

which by integration gives -^ = — j-, the same as for incompressible 

fluids. This result was to be expected, because a, as is well known, 
is the velocity of propagation in the compressible fluid, and when this 
becomes infinite, the propagation is instantaneous, and the fluid there- 
fore incompressible. 

If / be indefinitely great, it will be found in the same way that 

-r~ — — , and the motion is such as was considered Art. 3. 
dr r 

Let now -^ be very small compared to «, and X, V, Z, and / 
each = 0. The equation (C) reduces itself to 

"-11?^^"^ dr df-^' '''''• dr' -~dF~' 

a particular integral of which is r^=^'P{r — at). This gives 

d^ ^ F\r-at) _ F{r-at) 

dr ~ r r^ ^ "' 



At the same time, because a^'N.l.p= — -^ nearly, we shall have 

, T , F'{r-at) ,^, 

«.N.l.p = ^ -(2.) 

The equations (1) and (2), involving but one arbitrary function, can 
apply only to a single disturbance, which takes place in a direction 
tending from a centre, as I have elsewhere shewn*. It is important 
to observe that when r is very small, the term of equation (1) which 
involves r"- in the denominator may be much greater than that in- 
volving r. In fact, if we expand the fxmctions, supposing r to be 
very small. 


_ F{-at) _ F'{-at) _ F"{-at) 

When therefore the disturbance is made by a sphere of very small 
radius r, the motion is transmitted from its surface to other parts of 
the fluid, nearly as if the fluid were incompressible. 


Application of the Principles of the foregoing Section to determine the 
Resistance of the Air to the Motion of a Sail-Pendulum. 

17. For the sake of simplicity, I will suppose gravity not to act. 
The ball being spherical and perfectly smooth, the direction of the 
motion of the aerial particles in contact with its surface tends at every 

* Camh. Phil. Trans. Vol. HI. p. 402. 


instant from its centre. Therefore / = 0. Also if the radius of the 

ball be supposed very small, the equation -f- = ^-t^> obtained at the 

end of the preceding Article, will be approximately applicable to the 
motion of the fluid in contact with the ball. Hence the velocity which 
is impressed at any point of the spherical surface may be considered 
to be transmitted instantaneously in the direction of the radius through 
that point, and to decrease according to the law of the inverse square 
of the distance. The problem, with the limitations above made is 
solved in the same manner for air as for water. 

Let now the origin of co-ordinates be A, (Fig. 6.), the position 
of the centre of the ball when it hangs at rest. I^et its centre perform 
oscillations of very small extent in nAN, which we will consider to 
be rectilinear. Suppose N to be the position of the centre at the 
time t reckoned from a given epoch, and call AN, a. Take P any 
point of the surface, join NP and produce it to R, and let NPR make 
an angle Q with ANQ, and the plane RNQ an angle /3 with the 

plane SAQ. The velocity of the centre = ^; and the velocity of 


the air at P — -rrCosO. Hence if NP=h, and NR = r, the velocity 

at ^ = -„ — . -^ . Now if AN be the axis of x, AS of a, and a 

r- at 

line through A perpendicular to the plane SAN, the axis of y, and 

the co-ordinates of R be x, y, %, then r^ = {x — aY + y^ + %^. Consequently 

the velocity (^) at R=, r^ ^ — 2 • ;77- Hence differentiating V^ 

With respect to the time only, 

dr _ d'a b^cos9 2b^cose{x-a) d^ h^ da d.cosO 
dt ~ W-' r' "^ 7 • dt^ '^ r"' df dt ' 

^ ^ x — a d.cosO 1 da cos^6 da sin^O da 

But as cos9 = , r: — = --77 + 


dt . r ' dt r ' dt r ' dt 

dV_d^ ¥cos9 ^b'cos'e da' b'sin'0 d^ 
dt ~ df r^ ^ r' ' dt' r' ' dt^ 



J dt df • r 2? • df ' 

Substituting in equation (B), 

j\-. - d^a FC0S9 b'^ ir, 2/, -am da b* COS^ 9 da „.^^ 

« ^-^-P = df ■ —r- + ap (2cos=0-sm=0) ^ - -^^ . ^ -M. 
When /• = infinity, /o = 1 : therefore f{t^ = 0. Hence when /• = A, 

„,T. , «?^a , . COS 20 <:?a^ 

Where p = 1, let j9 = n = a^ Hence when (O = 1 + o-, p = e' (1 4- a-) = n + aV. 
But because a- is very small, «^N. = «V very nearly. Therefore, 

„ d^a , - cos 20 rfa^ 

^ = n+^.*cos0 + -^.^. 

The total pressure resolved in the direction NA is ffp¥ eos6sm9d9dfi, 
from /3 = to /3 = 27r, and from = to = 7r. It will consequently 

be found to be equal to — — . -^ : and if A = the ratio of the specific 

gravity of the ball to that of air, the accelerative force produced by 

1 d'a 

this pressure is — . -7-7 . But the accelerative force of gravity in the 

same direction, if SA = A, is ^ ( 1 ~ t" ) » taking account of the weight 
of air displaced. Hence 

_ cP_a _g^(-._}\ j_ d^ 

d'a ^ 


«^__^ ±^_§^(l_l] nearly 

1 + K 


Therefore if L be the length of the seconds pendulum in vacuum, 

2s * 

I in air, / = Z« ( 1 — — j 

The correction of the length of the pendulum is thus determined 
to be double of what it would be if the motion of the air were not 
considered. It is to be observed that these calculations apply strictly 
only to the case of a very small ball. The experiments of M. Bessel 

give 1"956 for the coefficient of — . Those of Mr Baily, which were 

made most nearly under the circumstances which the theory supposes, 
give 1"864. The effects of friction and of the suspending wire, would 
tend to make the coefficient rather greater than less than 2. I am 
therefore unable to account for the difference between the experimental 
and theoretical determinations, which it appears by Mr Baily's experi- 
ments, is greater as the radius of the ball is greater, excepting perhaps 
the confined space of the apparatus may have had some effect on the 
experimental results. 

It would not be difficult to shew from the nature of the analytical 
expressions, that if the confined space in which the balls vibrate were 
taken into account in the theory, the same results would be obtained 
for two balls of different diameters, vibrating in different spaces, if the 
linear dimensions of the spaces were in the proportion of the diameters, 
their forms being alike. If this could be verified experimentally, it 
would shew that the difference of the values of the numerical coefficient 
which Mr Baily calls n, for balls of different diameters, as well as its 
deviation from the theoretical value 2, is very probably owing to the 
confined space of the vacuum apparatus. It would at any rate be de- 
sirable to ascertain by experiment whether the same ball gives the same 
value of n, when it oscillates in apparatus of different dimensions. 

Papworth St Everard, 
March S, 1834. 

* This result I obtained in the London and Edinburgh Philosophical Magazine (September, 
1833), by assuming the principle of the conservation of vis viva, without employing equa- 
tion (B). 


IX. Theory of Residuo-Capillarij Attraction; being an Explanation of 
the Phenomena of Endosmose and Exosmose on Mechanical 
Principles. By the Rev. J. Power, M.A. Fellow and Tutor 
of Trinity Hall, and late Fellow of Clare Hall, Cambridge. 

[Read March 17, 1834-3 

1. The curious and elegant law, according to which an interchange 
takes place between two fluids separated from each other by a thin 
membrane, one of the fluids generally (but not universally) the lighter 
of the two, being transmitted in greater abundance, was discovered a 
few years ago by Dutrochet.* 

His experiments tended to show that the unknown force which 
operated this effect, whether measured by the fluid transmitted in a 
given time, or by the pressure required to stop the process, was, for 
the same substances, proportional to the difference of densities of the 
mixtures on each side of the membrane. 

The vast importance of this law in animal and vegetable physiology, 
renders it highly desirable that its theory should be investigated on 
mechanical principles, and such is the object of the present enquiry. 

2. The opinion which would attribute this phenomenon to the 
existence of electrical currents, is now pretty nearly abandoned, even by 
Dutrochet himself, with whom it originated, and who maintained it with 
great zeal, until the publication of his later researches, in which he 

* L'Agent immedial du Mouvement Vital, (Paris, 1826), and Nouvelles Recherches sur 
I'Endosmose et VExosmose (Paris, 1828). 

Vol. V. Part II. Dd 


confesses himself compelled to resign it, though he does so with 
manifest reluctance. That electricity, artificially excited, is capable of 
accelerating the process, is indeed sufficiently established by the experi- 
ments of Dutrochet; but it is equally certain that this agent is by no 
means essential to the operation, since, in the natural process, the most 
delicate galvanometer gives no indication of its existence. 

3. To me it appears unquestionable, that the phenomenon results 
from the corpuscular attractions, which the particles constituting the 
membrane and the fluids, exert upon each other : that electricity, 
by heightening or modifying these attractions, should produce a sensible 
effect upon the operation, is nothing more than its ordinary chemical 
agency would lead us to expect. 

4. By corpuscular attractions are meant the forces which the 
ultimate atoms of different materials, whether simple or compound, 
exert upon each other. These forces are enormously great (though not 
infinite) when the particles are in immediate contact, but diminish with 
extreme rapidity, as the particles separate, becoming insensible at a 
sensible distance. The effects of corpuscular attraction are different, 
according as it is exerted between particle and particle, or between 
mass and mass. In the former case it gives rise to the phenomena of 
chemical affinity ; and in the latter, to those of cohesion, adhesion, and 
capillary attraction, which may be regarded in general, as the mutual 
attraction of contiguous masses, being the combined effect of the 
corpuscular attractions of their integrant particles. It is under this 
point of view that La Place has considered the subject of capillary 
attraction, and his theory will be of the greatest use in the present 

5. Although no pores can be detected in the membranous partition 
by the help of the most powerful microscope, yet the fact that the 
fluids are transmitted, is a certain proof that such pores exist. They 
must indeed be extremely minute, and it will be seen that it is on 
this very minuteness that the energy of the sustaining force depends. 
These pores must be regarded as communicating with the opposite fluids 


at their two extremities, while the fluids meet and mix in the 

6. Dutrochet argues that capillary attraction cannot be the cause 
of endosmose, because it can only raise a fluid to a small height in a 
capillary tube, and is utterly incapable of drawing it beyond the limits 
of the tube. 

In stating these objections, he perhaps does not consider that the 
height at which a fluid may be sustained in a capillary tube is inversely 
as its diameter, and consequently in a tube of so extremely small a 
diameter as those of which it is necessary to suppose the membrane to 
consist, that height might be almost indefinitely great. It is true that 
in the case of a single fluid, this effect would require for its production 
that the tubes themselves should be coextensive with the fluid raised ; 
but this is no longer necessary when the two ends of the tube are 
immersed in different fluids. The reason why a homogeneous fluid 
cannot be drawn beyond the limits of the tube, is, that, were it to 
be so, the tube, acting equally at its two ends, would produce no 
effect whatever upon the fluid. But the circumstances are very different 
when the extremities communicate with different fluids. In that case the 
full residual effect, consisting of the difference of effects, which the same 
tube indefinitely extended, is capable of impressing separately upon the 
two fluids, might be produced by an extremely small length of tube, 
not exceeding a small multiple of the sphere of attraction of the par- 
ticles of the tube, and there is no doubt that the thickness of the 
finest membrane is a considerable multiple of this magnitude. In fact, 
if we cut off" from the ends of the tube a distance greater than the 
tube's sphere of sensible attraction, it is plain that the fluids which 
occupy the intermediate part, in whatever way they may communicate 
there, will suffer no effective attraction from the tube, since every 
elementary portion will be drawn by it equally in both directions. The 
only effective attractions will therefore be those exerted by an insensible 
portion at each extremity ; we may therefore imagine these two por- 
tions to be brought together as near as we please without any diminution 
of effect. 

D D 2 


7. In order to form some sort of estimate of the forces which may 
be expected to result from residual attractions of this kind, let us 
suppose the fluids to be water and alcohol, and the tube to be of glass. 
Now Gay Lussac found by experiments of great accuracy, that in a 
tvibe of glass whose diameter was 1.29441 millimetres, water would 
stand at the height of 23"'.3791, and alcohol of specific gravity O.8I96 
(that of water being 1) at the height of g^^'.SgSOS. This column of 
alcohol would be equivalent to 7™.7176 of water; the difference of 
effects would therefore be measured by a column of water of I5"'\66l5. 
Suppose now the diameter of the tube to be diminished a thousand 
times, or to become 0'"'.001294, the column of water which measures 
the difference of effects would be 1566l™'.5: or, since the French 
millimetre = .0393708 of an English inch, a glass tube of diameter 
0'".0000507, or about the twenty-thousandth of an inch, would produce 
a residual effect, with water and alcohol, measured by 616.6 inches or 
51" 4'" of water, which is equivalent to the pressure of nearly two 
atmospheres. When it is considered that a platina wire of one three- 
thousandth of an inch in diameter may be seen by the naked eye, it is 
probable that the magnitude we have assigned to the capillary tube 
is considerably greater than the diameter of the membranous pores, 
which evade the powers of the strongest microscope. From this ex- 
ample I think the conclusion may be fairly drawn, that, so far at least 
as the magnitude of the force is concerned, we need be under no 
apprehension but that the residual capillary forces are sufficient to 
account for the sustaining force of endosmose. How far they will 
account for the law of its variation will be seen hereafter. 

8. An attempt to explain the phenomenon by the principles of 
capillary attraction has been already made by a distinguished mathema- 
tician, Mons. Poisson. He first abstracts from the pressure of the 
adjacent fluids, by supposing their altitudes above the membrane to be 
inversely as their densities. The fluid in the tube being now equally 
pressed on both sides, he supposes that that liquid, for which the tube 
has the stronger attraction is drawn by this attraction to the opposite 
end, thus filling the whole tube. The fluid within the tube, he now 
argues, will be urged by two forces : 1st, the attraction of the liquid 


to which it belongs ; 2dly, the attraction of the opposite liquid. If then 
the latter attraction be superior to the former, the fluid which fills the 
tube, he says, will be drawn in an uninterrupted stream into the 
opposite vessel. 

Dutrochet justly objects to this theory, that it will only account for 
a motion in one direction, whereas the phenomenon of exosmose requires 
a corresponding motion in the opposite direction. 

Professor Henslow, in a number of the Foreign Quarterly, suggests 
as a modification of Poisson's theory, that whilst the fluid within the 
tube is carried in the direction of the stronger attraction, the natural 
tendency of the fluids to mix, may carry the other fluid (or, perhaps, 
a slight infusion of it) in the opposite direction, and thus produce the 

I perfectly agree with Professor Henslow that the natural process 
of mixture is the cause of the exosmose, it being only necessary to 
suppose that the rapidity with which this process extends itself witliin 
the tube is somewhat greater than the velocity with which the whole 
mass of fluid which fills the tube is drawn in the opposite direction. 

But the theory of Poisson is further objectionable on this account, 
that it makes the continuation of the process solely dependent on the 
action of the fluids, whereas the experiments of Dutrochet incontestably 
demonstrate that it depends mainly on the action of the membrane. 
No doubt, the effect both of the fluids upon themselves, and of the 
membrane upon the fluids, ought to be taken into consideration, and 
this will be done in the following theory. 

9. If a capillary tube be divided into two parts by a plane perpen- 
dicular to its axis ; the attraction of one of these parts upon a fluid 
which exactly fills the other part is \cH, c being the contour of the 
inner surface of the tube, and H a certain definite integral or constant, 
depending solely on the materials of which the tube and the fluid 
consist. The contour of the tube may be of any shape whatever, curved 
or polygonal. (See Mec. Cel. Sup. au X* Liv. pp. 14 — 21.) 


It is convenient to give a name to the quantity H \ we will call 
it the capillary affinity between the two materials of which the tube 
and fluid are composed. 

It is easy to see that the quantity H will remain unchanged if we 
conceive the tube and the fluid to exchange their materials; for, by 
the equality of action and reaction, the elementary attractions, of which 


—— is the sum, will be equal in the two hypotheses. The tube may 

be regarded either as solid or fluid, and this fluid may be either the 
same as that which fills its interior or a different one. 

If we conceive the density of the inner fluid to be diminished in 
any ratio, all the elementary attractions, and therefore H, will be 
diminished in the same ratio ; and if, further, the density of the tube 
be diminished in any ratio, H will be diminished in the compound 

10. Next, let u and v be the original quantities by volume of two 
vmmixed fluids. Then, if no penetration of dimensions takes place, 
u + v will be their volume after mixture. If we regard the fluids after 
mixture as coexisting, each with a diminished density, within the same 
volume u + v, calling r, and pi these diminished or partial densities, 
(r) and {p) the densities of the unmixed fluids, we shall have 

,^ J and 7— , 

{r) u + v \p) u + v 


^ +-^ =1 

{r) ^ (p) 


ri + pi = r, 

r being the total or ordinary density of the mixture. The two last 
equations serve to express ri and pi in terms of /•, (;•) and {p). 


If then we have a second mixture of the same two original fluids, 
we shall have 

— + -^ = 1 

and r-i + p-i = p , 

where rj and p-i are the two partial densities, and p the total density of 
this second mixture. These equations serve in like manner to express 
r-i and p-i in terms of p, {r) and {p). 

11. Let us now endeavour to express the mutual capillary affinities 
which exist between the two mixtures just mentioned, and a third 
material (as that of a membrane or tube), in terms of the densities 
of these mixtures and the mutual capillary affinities between this same 
material and the unmixed fluids. 

Let the former affinities be denoted by H, K, L, M, N, namely, 
H between the tube and the first mixture, 
K between the tube and the second mixture, 
L between the first mixture and the second, 
M between the first mixture and its like, 
N between the second mixture and its like; 

and let the latter affinities be denoted by {H), {K), (L), {M), (A^), 


{H) between the tube and the fluid of density (r), 
(K) between the tube and the fluid of density (p), 
{L) between the fluids of densities (r) and {p), 
{M) between the fluid of density (/•) and its like, 
(iV) between the fluid of density (p) and its like. 

The attraction ^cH of No. (9) will be the sum of two partial 
attractions, namely, that of the tube upon two coexistent cylinders of 
the opposite fluids, whose densities are those of the original unmixed 
fluids diminished in the ratios r^ : (r) and p^ : (p). Hence by the latter 
part of that No., 

ic^=ic(^)^ + ic(^).^; 



By similar reasoning, superposing all the different attractions, each 
diminished in the ratio of the densities of the attracting and attracted 
materials, we shall have 


By combining each of the last five equations with the four equations 
of No. 10, and eliminating r,, ^2, /a,, p^, we shall obtain H, K, L, M, 
N, in terms of the actual densities r, p, the original affinities (H), 
(K), (L), (M), (N), and the original densities (r) and (p). 

12. Let us now proceed to apply the principles of the three last 
numbers to explain the experiments of Dutrochet. And first let us 
consider those which relate to the statical force of endosmose. In 
these experiments the process was allowed to continue until the fluid 
raised, or rather the mercurial column which was hydrostatically sub- 
stituted for it, attained its maximum altitude ; at this moment the 
densities of the two liquids were experimentally determined ; and 
instituting different experiments with different mixtures of the same 
substances, Dutrochet found that the maximum altitudes were propor- 
tional to the corresponding differences of densities. 

The substances employed in his experiments were saccharine or 
gummy solutions on the one hand, and water on the other, and the 



water was found to be transmitted in greater abundance. Common 
treacle is a very convenient substance for experiments. 

Let us suppose then that the lower part of the endosmometer is 
filled with treacle, and having a thin membrane tied over its mouth, 
is immersed in water ; and let us suppose that the fluid is allowed 
to ascend until the operation ceases. 

At this moment we may regard the capillary pore which traverses 
the membrane, as communicating at its two extremities with fluid in 
the same state of mixture as the fluid in the contiguous vessels, 
there being a gradual transition from one end to the other. 

Let C1C1C2C2, be a portion of 
the membrane, AiAiAsA^ one of 
its capillary pores, with its axis at 
right angles to the plane of the 
membrane, communicating originally 
with the water at A^A^, and with 
the treacle at A^A^, but when the 
fluid has reached its maximum al- 
titude, communicating with the ^ 
first mixture of No. (10) at AiA^, 
and with the second mixture of 
that No. at A2A2. 



















Imagine the geometrical figure of the tube, (not its material) to 
be produced both ways to D, and D^, and cut off" from each end 
of the tube a distance AiBi, AiB^, equal to the tube's sphere of 
sensible attraction. 

Since A^Bi, and A^Bi, are insensible, we may regard the fluids in 
AiA^BiBi, and A^A-^BiB^, as in the same state of mixture as the 
fluids in the contiguous vessels. 

Vol. V. Part II. E e 


Let us now estimate all the forces which tend to move the central 
column DyD^D-iDi in direction of its axis. 

It is plain that, in whatever manner the fluids may communicate 
in the interior of the tube, the tube can produce no effect upon 
ByBiBiBi, since every elementary portion of this part of the fluid 
will be drawn in both directions as by an infinitely extended tube. 

We may also neglect, as producing equal and opposite forces in 
both directions, the attraction between the tube A^B^ and the fluid 
AiAiBiBi; between the tube A^Bi, and the fluid A^AzBiB,; be- 
tween the fluid tube dA^Di, and AiA^D^D^ ; between C2A2D,, and 
AzAzDiDs, between the membrane and C^A^D^-, between the mem- 
brane and CiA-iD-i. 

Lastly, we may neglect all the mutual actions of the particles 
composing the central column DyD^DiDi, their tendency being only 
to mix the opposite fluids, and not to move the column as a 

Of the remaining attractions we shall have at one end the 

attraction of the tube B^Bt, upon B-^B^A^A^, ( = \ cH) + the 

attraction of the tube A^B^, upon D^D^A^A^, {= ^ cH) — the 

attraction of the fluid tube C^A^D^, upon A^A^B^B^, {=\cM); 

constituting the capillary force - {2H — M). This will be opposed 

by a similar force — {^K—N) exerted at the other end of the tube. 
The residual sustaining force is therefore 

I .{2H-2K-M+N'). 

It now only remains to express this force in terms of the actual 
densities r and p, and the initial constants 

{r), ip), {H), (K), (L), {M), {N). 


13. For this purpose let 

^-=... and ^=..; 

therefore by No. (10). 

making these substitutions in the equations of No. 11., we have 

K = s,{H) + il-s,){K). 

L =s,.{l-s,){L) + s,.{l-s,){L) + s,s,{M) + {l-s,){l-s,){N'). 

M= 2s, (1 -*0 {L) + s,' {M) + (1 -*,)' (^)- 

N = 2s, (1 - s,) {L) + si (M) + (1 - s,y (N). 

Hence 2H-2K-M+ N=A (H) + BiK) + C {L) + D{M) + E{N-), 

where A = 2. {s, — S2). 

B = 2.{l-s,)-2.{l-s,) 

= -2{si-s,). 
C=-2s,.{l-s,) + 2s, (1 - s,) 

= -2{s,-s,) + 2{8{'-si). 
E=-{l-s,Y + {l-s,Y 

= 2{s,-s,)-{s,'-'Si). 



The residual force is therefore 

I {s,-s,){2{H)-^{K)-2{L) + 2{N)} 


+ l-{s,'-s.'){2{L)-{M)-{N)}. 

Again, r = r^+p, = {r).^^+{p).■^ 

= ('•) *! + (/») (1-*.); 

••. *, 

=={r)s, + {p).{l-s,); 

■• '~ip)-irr 

_ p — r 
ip) - {r) 

„ . „_?ie)zik±rl. 


{ip)-ir)r {{p)-ir)r' 
The expression for the residual force is, therefore, 

p^ -r 



which may be put under the form 


^-(^)-(r)-r^) ^^^)+(^)-(r)L^^> {p) + {r) J}' 


{ip)-ir)rv' 2 r 

The agreement of theory with experiment, then, requires that 

jM) + jN) 
(^) 2 

should be either nothing, or very small compared with 

14. When I first began to investigate this subject, certain con- 
siderations, which it would be tedious to detail, led me to imagine 
that the fluids might communicate in the interior of the tube, 
forming a series of interlacing cylinders one within another, and I 
found the forces which tended to protrude the cylinders into the 

opposite fluids, all multiplied by (L) — - — '-—^ — - . I therefore looked 

upon this expression as a measure of the tendency of the fluids to 
mix, and this tendency being, as experience shows, very small in the 
case of treacle and water, as well as in the case of the gummy 
solutions and water, afforded an explanation why the force should 
be so nearly proportional to the difference of densities, as Dutrochet's 

* I have elsewhere erroneously stated, that the residual force is c A(p—r) + c B(p—ry, 
a mistake which I am glad to have this opportunity of correcting. 


experiments seemed to indicate. But the preceding theory being 
perfectly independent of the mode in which the fluids communicate, 
it is better not to have recourse to a supposition, which is in the 
slightest degree precarious, especially as I am now prepared to show, 
that, in whatever way the fluids may arrange themselves within the 
tube, the rapidity of the mixing process will depend upon the mag- 

nitude of (X)- (^);W . 

15. In fact, in whatever manner the mixing process may be 
effected,- we may at any moment imagine the fluid to be divided 
into an indefinite number of contiguous strata, of any arbitrary or 
convoluted form, the density being the same for the whole extent of 
any one stratum, but varying from one to another. 

If the surface which separates two contiguous strata be a perfect 
plane, it is evident, by the equality of action and re-action, that this 
would be a position of momentary equilibrium, (abstracting from 
gravity, which I am not here considering.) 

Suppose, now, that this surface becomes 
undulated in an arbitrary way, and take any 
point A upon it, and draw a tangent plane 
BAD, including with the surface EAC, a kind 
of lens BDEC, which, with La Place, we 
may call a meniscus. Draw the normal FAG ; 
and let Ri, and R^ be the radii of greatest and 
least curvature at the point A. 

Now La Place has shown that the attraction of such a meniscus 

upon the column of fluid AF is ("»"+ p")--^> where H is the 

capillary affinity between the material of the meniscus, and that of 
the fluid in the sense already defined. (See Supp. au X* Liv. 
page 14 — 17.) 


He has also shown that the attraction of the meniscus is the 
same whichever way it be turned. 

If the meniscus instead of consisting of the left hand fluid, (as 
in the figure), consisted of the right hand fluid, the common boundary 
being the plane BAD, there would be equilibrium, the column 
AF being attracted by the right hand fluid, just as much as the 
column AG is by the left. 

Since then the meniscus consists of the left hand fluid instead of 
the right, the effect of the disturbance upon the column AF, tending 
to draw it in the direction FA, is the attraction of the meniscus 
upon AF, regarding it as consisting of the left hand fluid, minus 
the attraction of the same meniscus regarding it as consisting of the 
right, that is 

\R, ^ EJ \2 2 

supposing the left hand fluid to be the first mixture of No. (10), 
or the lower fluid of No. (12). 

If then we estimate the effect in the direction AF, it is 

1 1\ /L M\ 

(1_ J_\ (± _M 

In the same way, the effect of the disturbance upon AG, in the 
direction GA, is the attraction of the meniscus, regarded as consisting 
of the left hand fluid, minus the attraction of the same meniscus, 
regarded as consisting of the right, that is 

Ui "^ BJ • \2 ^ 

2 j- 

Hence the whole attraction in the direction GF, is 



If we substitute for L, M, N, the expressions at the commence- 
ment of No. (13), we shall find 

^L-M-N = A{L) + B{M) + C{N), where 

^ = 2*i.(l-«2) + 2*2.(l-*,)-2*i.(l -*i)-2 52.(l-«2) 
= 2 *, - 2 *,«2 + 2 *2 - 2 «i*8 - 2 *, + 2 «i' - 2 *2 + 2 «/ 
= 2*,*- 4*1*2 + 2*/ 
= 2(*i-*,f. 

= -(*!- s^f. 

= -{(i-*0-(i-*.)}' 
= -(«i-*2)'; 

.-. ^L-M-N={s,-s,)\{2 (L) - {M) - {N)} 
The effect of the disturbance in the direction GF, is therefore 

consequently if (i) be greater than ^ — '-^ — '- , or, if the capillary 

affinity of the opposite fluids exceed an arithmetic mean between 
the capillary affinities of the two fluids for fluids of their own kind, 
the tendency will be to depart still farther from the position of 


equilibrium, and the tendency is the greatest where the curvature is 
the greatest. 

16. Hence it is easy to see that the protruding segments of each 
fluid will become more and more pointed at their summits of greatest 
curvature as they advance into the opposite fluids, thus forming 
interlacing spiculse, shooting into the opposite fluids, and at the same 
time inosculating with each other by their lateral protrusion, and 
that this process cannot cease until the fluids have divided each 
other into segments of a magnitude comparable with that of the 
sphere of sensible attraction. 

Beyond this limit the theory does not hold. It is very possible then, 
that in some cases a limit may be attained where the mixing fluids 
have arrived at such a state of subdivision, that the conditions for 
continuing the subdivision are no longer satisfied ; in other cases it 
is possible that the subdivision may proceed until the ultimate atoms 
of the opposite fluids act upon each other by ones, twos, and threes, 
thus effecting a chemical decomposition : nature presents numerous 
instances of both kinds. 

17. But though the mathematical theory is not strictly applicable 
when the subdivided segments are of less magnitude than the sphere 
of sensible attraction, it may be considered as an approximation to the 
truth considerably beyond this limit. For, the most effective part of 
the attraction of each segment being that exerted by the particles 
in immediate contact with the normal column, the diminution of 
the segments will only have the effect of removing the more feeble 
part of the attractions which the theory takes into the account. It is 
therefore probable that, even in cases where no chemical decomposition 
takes place, the subdivision of the fluids may be carried to a limit far 
beyond that to which the theory is strictly applicable. Besides, the 
processes of nature are not interrupted of a sudden; the tendency 
therefore to farther subdivision cannot be suddenly arrested, but in 
cases where it is ultimately reduced to nothing, it must be so by 
passing through all degrees of magnitude. This reasoning is further 

Vol. V. Paet II. F f 


confirmed by those experiments which demonstrate the almost infinite 
subdivision of matter by repeated dilution, experiments which are 
familiar to every one. This infinite subdivision is, in fact, involved in 
the mathematical conception upon which this theory is founded, namely, 
that in the state of mixture the two fluids may be regarded as 
coexisting within the same volume, each with a diminished density. 
This conception cannot of course be a rigorous representation of nature ; 
but is sufficiently so for the application of La Place's theory, or, which 
comes to the same thing, for the summation of the attractions by the 
principles of the Integral Calculus. 

18. In cases of simple mixture, unattended with a chemical change, 
the ultimate segments of the opposite fluids, though in an extreme 
state of subdivision, have a separate and independent existence, which 
renders it highly probable, that the volume of the mixed fluids should 
equal the sum of the volumes of the unmixed fluids. This supposi- 
tion has been made in the preceding theory, and I find by experiment 
that in mixtures of treacle and water it is accurately true. The same, 
I believe, is true in all cases of simple mixture, where no chemical 
result takes place, such as the precipitation of solids, or the disengage- 
ment of heat or other volatile constituents. To liquids whose union 
is accompanied by such phenomena the present theory is inapplicable, 
not only on account of the penetration of dimensions, with which 
such phenomena are generally attended, but on account of the change 
of affinities, which the escape of some of the constituents must 
necessarily produce, including heat, which, regarded as a chemical 
constituent, is as important as any. 

19. The addition of a third fluid to one of the liquids, by altering 
the chemical affinities-, must likewise alter the capillary aflfinities, which 
are only a different modification of the same corpuscular attractions 
which produce the former. It is not surprising then, that Dutrochet 
should have discovered some substances which accelerated the process 
in his experiments, and others which retarded it or stopped it 


Water impregnated with sulphuretted hydrogen was found not only 
to stop the process, but to destroy the energy of the membrane for 
subsequent experiments with pure water and pure saccharine solutions. 
No doubt the sulphuretted hydrogen had decomposed the surface of 
the capillary pore, leaving a coating of putrid matter, which was not 
possessed of such capillary properties as to supply the place of the 
material of the membrane. That this is the true explanation is shown 
by the fact, that when the membrane was for a long time steeped in 
water and well washed, its energy was restored : in fact, the putrid 
matter being washed away, the membrane presented an unvitiated 
surface to the fluids. Heat and electricity may be classed amongst 
these chemical agents, as they operate their effect precisely in the same 
way, namely, by changing the chemical and consequently the capillary 

20. If we wish to compute the height to which the fluid will 
rise in the endosmometer, let ^ be the height of the supported column 
above the surface of the membrane, and z the height of the lower fluid 
above the same, w the transverse section of the tube; the difference 
of the pressures of the cylindrical columns w^ and wg, having the 
common section w, is gpco^—grioz: this must be counterbalanced by the 
sustaining force cA(p — r) + cS{p^ — r"), which denotes a pressure on the 
same scale ; 

" o ff \ p J oi ff \ p / p 

If a column of mercury be hydrostatically substituted for the 
ascending fluid, as in the experiments of Dutrochet, calling Z the 
altitude of the mercury, and R its density, we must have 

„ c A (p-r\ cB (p^-r^\ r 

^ = -.- g • K-w) ■" -.^ ■K-R-) "■ R^' 

this of course being subject to a correction when the cistern of the 
mercury is not on a level with the membrane. 

F F2 


21. If the pore be circular, let ^ be its diameter, then 

c = 2'7r.-, and to = tt . — : 
2 4t 

€ 4 

•'. - = -J ;. 


the sustaining force is therefore inversely proportional to the diameter 
of the pore, as in ordinary capillary attraction. 

Hence we see how the membrane^s delicacy of texture contributes 
to the intensity of the sustaining force. 

22. It is now easy enough to see in what manner the process is 
effected. The residual force cA{p — r) + cB{p'^ — r"), which would result 
if the ends of the tube communicated with fluid of the densities 
r and p, being greater than the altitudinal pressure upon the section w, 
would cause the fluid within the tube to move as a mass into the 
endosmometer, thus bringing fluid more and more diluted to the 
issuing orifice; this will continue until the residual force is weakened 
to such a degree as exactly to counterbalance the altitudinal pressure. 
Contemporaneously with the former motion, the mixing process will 
transfer the two fluids in opposite directions, the current from the 
endosmometer towards the water producing the exosmose, and the 
opposite current supplying the deficiency caused by the exosmose, and 
therefore not contributing to the endosmose. The diluted fluid which 
was carried into the endosmometer by the residual force, will gradually 
mix with the treacle within, whether that mixture be carried on near 
the orifice of the tube, or whether the diluted fluid be raised by its 
specific levity higher up in the endosmometer. The extremely small 
portion of diluted fluid which has thus been transmitted, and the 
viscosity of the treacle, render it most* probable that it would not be 

* This probability amounts nearly to certainty when we consider that the denser fluid 
has no access to the lower part of the transmitted fluid. It is only when a lighter body 
is insulated, or partially insulated^ in a denser that it rises by its specific levity. 


carried up by its specific levity, but rather adhere to the membrane 
in the way that bubbles of air adhere to the sides of vessels containing 
water or mercury. But, be this as it may, the end of the tube which 
communicates with the endosmometer, will soon be surrounded by a 
stronger infusion of the treacle, which will again bring the residual 
force into action ; thus a fresh portion of the fluid will be introduced 
into the endosmometer, and the same process will be repeated as before. 
For the sake of explanation, I have supposed the residual force to 
produce its eflPect discontinuously, but it is easy to see that the process will 
really be continuous, the united actions of the endosmose and exosmose 
always keeping the orifices of the tube surrounded by fluid in such a 
state of dilution that the magnitude of the residual force will be exactly 
sufficient to create a supply proportioned to the demand arising from 
the mixing process which is continually proceeding within the endosmo- 
meter. The residual force cannot be less than this, for if it were, the 
encroachment of the treacle upon the issuing orifice would immediately 
increase it ; nor can it be greater, for then the accumulation of the more 
diluted fluid at that same orifice would immediately diminish it, 

23. The quantity transmitted in a given time must depend more 
upon the rapidity with which the mixing process is carried on within 
the endosmometer than on the magnitude of the residual force. This 
force is certainly essential to the transmission, but its effect is no other 
than that of a pump which supplies the fluid from below as fast as it is 
wanted, and no faster, and that of a catch or valve to sustain it when it 
is once elevated. The moving force at the summit of any protruding 

spicula is by No. (14) represented by [^ + ~p) ^(p — ^)^ and is, 

therefore, for spicule of given shape, as the square of the difference 
of densities. It might appear then, at first sight, more probable that 
the quantity of the lower fluid absorbed by the fluid in the endosmo- 
meter in a given time, would be more nearly as the square of the 
difference of densities, than as the simple power of this difference, which 
is the law the experiments of Dutrochet tend to establish. But such 
a conclusion would be very precarious, as will appear by the following 


24. Let us imagine two different experiments, all circumstances, 
as regards the materials, form and disposition of the apparatus, being 
exactly similar, but the proportions in which the substances are mixed 
on each side the membrane, being different in the two experiments. 
Let us suppose also that the mixing process takes place in both experi^ 
ments after exactly the same type, only with different velocities, that 
is to say, that at certain times, t and t', t + T and t' + r, # + 2t and 
#' + 2t', &c., the protruding spiculae from the lighter fluid exist in 
exactly the same state in both experiments, as regards their number, 
shape, size and situation. 

This supposition being made, the volume of the lighter fluid absorbed 
by the fluid in the endosmometer in the two experiments, will be equal 
in the intervals t and t' : also the summits of the spiculae will have 
described the same paths in the two experiments during these same 
corresponding intervals. Let t and t be indefinitely small, and let us 
equate the spaces described by the summits of any two corresponding 
spiculae between the epochs t and # + r, t' and t' + t', and also between 
the epochs t and ^ + 2t, t' and #' + 2t'. 

Let a be the sphere of sensible attraction, and imagine a small 
normal column 2 a at the vertex of each spicula, being half in one 
fluid and half in the other. 

The two spiculse having by the hypothesis the same shape, the 
moving forces upon these columns are as {p — rf and [p' — r'f, and the 
masses moved are as ap + ar and ap' + ar', that is, as p + r and p' + r; 

the accelerating; forces will therefore be as — and '^ , / ; let 

^ p+r p +r 

(p _ rY (p — r'f 
them be k ^ '- and ^ k . ^ f- . Then if v and v' be the velocities 

p+r p +r 

of the two summits at times t and t', equating the corresponding spaces, 
we shall have 


P + r " ■ ^' p+r' 

■^ p + r ^ p +r 


These equations are equivalent to the following: 

VT = VT, and — = -^-—, — —. — ; 

p-irr p ^■r' 


v' ~ T ~ p' — r' ' p + r ' 

Let q be the volume of fluid absorbed in the times t and t', which 
we have seen to be the same in each experiment; and let Q and Q' 
be the quantities absorbed during a given time T, T not being so 
great but that r, p, r' and p may be considered the same during this 

If then there be a law connecting the quantities absorbed in a given 
time with the densities, we must regard this absorption in each experi- 
ment as uniform during the time T\ 

.: Q : q y. T : T, 

and Q '. q V. T : r'\ 

... Qr = qT=Q'T'; 

+ / 

^ T p' — r' p + r 

■ The supposition we have made, as to the exactitude of type in the 
two mixing processes, is particular ; but if there be a general law whicli 
is applicable to all cases, that, must include the case supposed, and 
therefore the result of the particular case must coincide with that of 
the general law. If then there be such a law, it is expressed by the 

Q.Q .'. 

p-r . p-r 

"s/p + r ' \/p' + r' ' 

This being true in different experiments, must be true in different 
stages of the same experiment. 


Now in the same experiment p diminishes and r increases as the 
experiment proceeds, and therefore the variation of p + r is small com- 
pared with that of p — r; the quantities absorbed will therefore be 
pretty nearly in the ratio of the difference of densities, as Dutrochet 
found them to be. Whether the proportion 

Q:Q' :: -E^ : -IzL 
y/p + r Vp + r' 

may be a more accurate representation of nature than the law of 
Dutrochet, is left to the test of experiment. 

25. It may perhaps be objected to the theory of No. (12), that 
the ordinary theory of capillary attraction supposes the dimensions of 
the tube to be incomparably greater than the sphere of sensible attrac- 
tion, whereas the fact, that these pores are so small as to elude 
microscopic observation, might lead us to apprehend that their dimensions 
were of a size comparable with that sphere. The example which has 
been calculated in No. (7), does not seem to leave any cause for such 
an apprehension. But supposing this were the case, the only difference 
it would make in the theory is this : that, whereas, on the former 
supposition, the quantities \cH, \c K, &c., denoted the results of 
integrations extending from nothing to infinity, and not otherwise 
depending on the form of the tubes than by involving the contour c 
as a multiplier; on the second supposition, the limits of the integra- 
tion will depend on the form of the tubes and the texture of the 
membrane : but these limits being the same in the cases compared, it is 
easy to see that the theory will be still true on the latter hypothesis, 
provided we look upon ^c{H), ^c{K), &c., as denoting certain 
unknown limited integrals depending not only upon the nature of the 
materials, but also upon the form and size of the capiUary pores. The 
residual force will, therefore, on this hypothesis also, be of the form 
a{p-r) + h{p'-r'). 

26. By the application of similar reasoning to the theory of No. (15), 
it is not difficult to conclude that the moving forces upon the normal 


columns at the summits of spiculae of given shape and sixe will be as 
(p — r)-, even when the dimensions of the spiculse are indefinitely less 
than the sphere of sensible attraction. For, the attraction of a meniscus 
bounded on one side by a plane surface, upon the conterminous normal 
column, will in all cases be a definite integral depending on the shape 
and size of the meniscus, and the demonstration of La Place, by which 
he shows that the attraction of such a meniscus is the same whichever 
way it be turned, is perfectly independent of its size and the shape of 
its curved surface. 

Let then I be the attraction of any meniscus upon the conterminous 
normal, the meniscus consisting of one mixture, and the normal of the 
other; m, the attraction of the same meniscus when the meniscus and 
column consist both of the first mixture; and n, the same thing when 
they consist of the second mixture. Then reasoning exactly as in 
No. (15), the moving force upon the column GF will be 2l—m — n; 
and if (/), (m), (»), be the initial values of I, m, n, it may be shown 

exactly as before, that 2l—m — n = c/C_^ >.^g • {2(/) - (m) - (n)}, the 

theory of No. (11) being equally applicable in this case. Hence, how- 
ever minute the spiculae may be, the moving force upon the central 
column will, for spiculse of given shape, be as the square of the difference 
of densities. 

This consideration applied to the theory of No. (25), gives it a 
generality which renders it as satisfactory as can well be desired. 


Trinity Hall, 

Marck 29, 1834s 

Vol. V. Part II. Gg 


X. On Aerial Vibrations in Cylindrical Tubes. By William 
Hopkins, M.A. Mathematical Lecturer of St Peter's CoUege, 
and FeUow of the Cambridge Philosophical Society. 

[Read May 20, 1833.] 

The problem which has for its object the determination of the 
motion of a small vibration propagated in an elastic medium along a 
prismatic tube of indefinite length (the motion of every particle in 
each section of the tube perpendicular to its axis being the same) was 
long since solved by Euler and Lagrange. The problem, so nearly 
allied to this — to determine the motion of an aerial pulsation in a tube 
of definite length — has not been so satisfactorily solved, the tube being 
either open at the extremity or stopped with a substance possessing 
some degree of elasticity. In addition to the difficulties of the former 
problem, we have in this latter one those still more formidable difficulties 
which exist in the determination of the circumstances of the motion 
at the confines of two elastic media in the closed tube, or at the 
extremity of the open one, where the air in the tube communicates 
with the circumambient air. These motions must no doubt be deter- 
minable from the nature of the media, and the causes producing and 
maintaining the vibrations, having nothing arbitrary in them, except 
what may be so in the original disturbance ; but I am not aware 
of any progress having been made in the direct solution of these 
questions, which now forms one of the greatest desiderata in the appli- 
cation of mathematics to physical science; and in our inability to 
determine these motions at the extremity of the tube, either by theory 
or direct observation, we are driven to the necessity of assumptions. 
It is from a difference in these assumed conditions that we have the 



different solutions which mathematicians have given of the problem in 
question. The principle on which we ought to proceed in making such 
assumptions is obvious ; they should be subjected to no restrictions, 
(not imposed on them by our theory), which are not necessary to draw 
those deductions and inferences from our mathematical results^ which 
admit of verification by experiment, to the test of which an assumption, 
in any degree arbitrary, must necessarily be subjected before it can claim 
our confidence. The physical conditions however on which the solutions 
of this problem depend, (as far as it is distinct from that of the motion 
of a wave along a uniform tube of indefinite length), have neither 
been assumed on this principle, nor subjected, as far as I am aware, 
to this experimental test. It has been principally with the view of 
remedying these defects that I have prosecuted the researches, an account 
of which I have now the honour of laying before the Society. 

1. The physical conditions assumed by Euler, and by most of those 
who have since written on the subject, are, that the particles of air at 
the extremity of a closed tube are always at rest; and that no con- 
densation of the air takes place at the extremity of an open one. The 
first condition involves the supposition of the perfect rigidity of the 
material with which the tube is stopped. This cannot be accurately 
true, but probably leads to no error very appreciable to observation. 
The second condition assumes an eqviality in the densities of the external 
air, and of that within the tube immediately at its open extremity, 
during the whole time of the vibrating motion, in the same manner as 
if the air were at rest. This supposition carries with it but little 
appearance of being even very approximately true; for it is difficult 
to conceive how a sonorous wave could thus be produced and maintained 
in the surrounding air from the open extremity of the tube, and it 
appears perfectly irreconcileable with the fact of the sudden cessation 
of sound after the cause producing it has ceased, M. Poisson, struck 
with these objections, has assumed another physical condition as appli- 
cable to any tube, whether open or stopped, viz. that there exists at the 
extremity of the tube, during the whole motion, q constant relation 
between the velocity of the particles of the fluid at any instant, and 
its condensation, this relation depending on the nature of the substance 


with which the fluid at the extremity of the tube is in immediate 
contact. This condition is manifestly less restrictive than those of 
Euler, since it involves no supposition of the perfect rigidity of bodies, 
and leaves room for a certain degree of condensation and rarefaction 
of the fluid at the extremity of the open tube, thus removing the 
difficulty above-mentioned respecting the maintaining of aerial pulsations 
from the open end, in the circumambient air ; while it enables us also 
to account in some measure for the rapid cessation of sound with the 
cessation of the cause producing the vibratory motion of the air in 
the tube. 

2. The two authors above-mentioned have written elaborately on 
this subject of the vibrations of elastic fluids in tubes. Mr Challis 
also in his paper published in the Transactions of this Society, (Vol. III.), 
has been led to the consideration of the conditions which hold at the 
closed or open extremity of the tube in which the air is in a state 
of sonorous vibration, though the determination of this point forms 
with him a collateral rather than a principal object. He assumes that 
a pulse proceeding along a cylindrical tube will be reflected from the 
further extremity if the tube be stopped, the intensity of the reflected 
pulse being equal to that of the incident one; and that if the extremity 
of the tube be open, it will pass into the circumambient air, sending 
back no reflected wave within the tube. If this were the case, it 
would immediately account for the apparently instantaneous cessation 
of sound above-mentioned ; but there are other equally obvious 
phenomena, for which this hypothesis appears to offer no adequate 

3. It will be observed, that Euler has supposed either the velocity 
of the particles or their condensation to have, at the extremity of the 
tube, a constant value, independently of the time ; while M. Poisson 
has supposed this constancy of value to belong to the quantity ex- 
pressing the relation between the velocity and condensation. It does not 
however appear to me probable that any such conditions, independently 
of the time, should hold. All the above assumptions are equally 
arbitrary, and equally require to be put to the test of experiment. In 


applying this test, I find that the deductions from the results, derived 
from any of the three hypotheses above-mentioned, do not sufficiently 
accord with the observed phenomena to be perfectly satisfactory. This 
discrepancy is more particularly observable in the position of the nodes 
or points of minimum vibration in the open tube. According to Euler's 
hypothesis, these nodes would be places of perfect rest ; and they would 

be distant from the open end by an exact odd multiple of -, where 

\ = length of a whole undulation. From the hypothesis of M. Poisson, 
their positions would be the same as in the above case, but they would 
become points of minimum vibration, and not of perfect rest. Mr 
Challis's supposition would lead to the conclusion that no nodes existed 
in this case, except they should be produced by some vibration of the 
tube itself, a cause the total inadequacy of which to produce any appre- 
ciable effect, must be immediately recognized by every one who has 
made experiments on this subject. The facts, as determined by experi- 
ment, are very obvious ; and it appears that there are nodes, which 
are points of minimum vibration and not of perfect rest ; that they are 

equidistant, but that denoting this distance by -, the distance between 

the open extremity and the nearest node is considerably less than -. 

I shall not in this place proceed further with the detail of experimental 
facts ; but shall first shew how the theory of this subject may be 
generalized by the assumption of conditions less restrictive than those 
which have been made by the writers I have mentioned. In the second 
section, I shall describe the experiments which have suggested these 
assumptions ; and shall conclude with some observations on the resonance 
of tubes, so far, more particularly, as it is allied to the investigations 
contained in this paper. 

The form under which I shall consider the problem, is that under 
which it presents itself, as nearly as possible, in the experiments I have 
to describe. 



4. Suppose the tube AB, (fig. I.), open at A, and stopped at B, 
with some substance possessing any degree of elasticity ; and suppose 
the vibrations first produced and kept up by a rigid diaphragm, vibrating 
according to a given law at A, and perfectly excluding the air within 
the tube from any communication with the external air. We have 
the usual equations 

v=f{at-x) + F{at + x)] 


as=f{at-x)-F{at + x)] 

V denoting the velocity of a particle at distance a; from the origin, 
and s the condensation at the same point at the time t, and a being 
the velocity of propagation of an aerial pulse along the tube. 

One of our conditions must necessarily be, that the velocity of the 
air within the tube and immediately in contact with the diaphragm, 
must constantly have the same velocity as the diaphragm itself, con- 
strained to move according to a given law. Let this velocity = <p{at). 
Then shall we have 

(j){af)=/{ai) + F{at) (1). 

5. To ascertain the nature of the second condition, which must 
hold at B, where the motion of the wave propagated along the tube 
is interrupted, we must consider the effect which will be produced on 
the stop by the action of the air within the tube. The vibratory motion 
wUl produce alternations of condensation and rarefaction at the ex- 
tremity B, which will tend to put the substance forming the stop in 
vibration; and if it will admit of vibrations having the same period 
as those of the air in the tube, this effect will be produced by the 
constant reiteration of the cause above-mentioned. If the substance is 
not susceptible of vibrations of this kind, no appreciable effect will be 
produced upon it. 


The determination of the nature of these vibrations, or of the 
function expressing the velocity at any instant of the extreme section 
of the stop, will necessarily depend on the material of which it is made; 
and any solution of the problem in question, independently of this 
consideration, cannot be regarded as complete. Still, whatever may be 
the nature of the stop, we know that the period of its vibrations must 
be the same as for those in the tube; and it is also manifest, that each 
vibration of the stop must begin at a time later by an interval at least 

nearly = -, (/= the length of the tube), than the corresponding vibration 

in the diaphragm at A, whence the original disturbance is supposed to 

proceed. I say that this interval is' nearly equal -, because certain 

phenomena, of which I shall speak hereafter, seem inconsistent with 

its being in particular cases exactly = -. I shall therefore, to give the 


assumption all the generality possible, consider it as generally = — f- arbi- 

trary quantity, to be determined in each particular case by experiment. 
Hence then, if ^ denote the form of the function of the time expressing 
the velocity of the extreme section of the stop, we shall have the 
velocity = v/'l «/ — (/ + c)}, c being arbitrary. This must also be the 
velocity of the extreme section of the air at B, consequently we have 
as a second condition 

•^{at-{l-^c)}=f{at-l) + F{at^-l) (2). 

We have from (1) 

(t>{at + l)=f{at + l) + F{at + l); 

and eliminating F(at + l), 

f{at + l)-/{at-l) = <p{ai + l)-f{at-{l + c)\ ; 

or, writing at + 1 (or at, 

f(flt-ir^l)=f{at)-y\f{at-c) + <t>{at + ^l) (B). 


The substance forming the stop being known, so that we might 
regard the vibrations produced in it under given circumstances de- 
terminable, the relation between the functions xj^ and J" would be 
known, and the function y would be the only unknown one in the 
above functional equation, from which, any particular form being 
assigned to (p, that of y must be determined. The arbitrary quantity 
which will be involved in the solution of this equation, must be 
determined by the original value of the function jf. 

6. We have here supposed the tube to be stopped, but the 
equation (B) will still be true for the open tube, \|/ {«/-(/ + c)}, de- 
noting always the velocity of the 'extreme section at the time f. 

Equation (2) gives us 

F{at + l)=-f{at-l) + y\,{at-{l+c)}, 
and writing at + x, for at + l, 

F{at + x)= -f{at-{2l-x)} + >// {at -{2l + c -x)}'. 


v = f{at-x)-f{at-{2l-x)} +^ {at-{2l + c-x)\-\ 

as = f{at-x)+f{at-{2l-x)}-yl^{at-{2l + c-x)}] 

The form of J" being determined by equation J?, these last equations 
will give the complete solution of the problem. 

7. Before we proceed to consider particular cases, we will exhibit 
these equations (C) under another form, which will be useful in 
deducing some general inferences as to the nature of the motion in 
the tube. 

Let T denote a period of time, from the commencement of the 
motion at A, less than that which is necessary for the pulse to 
travel twice the length of the tube ; consequently at will be less 
than 21. 

Equation (B) gives us 

/(ar + 9.1)=/ {ar) - v// («T - c) + ^ {aT + 2l), 
Vol. V. Part II. Hh 




and for ar, writing ar — x, 

/{(aT + 2/)-^}=/(aT-ar)-x//{«T-(a; + c)}+0(«T + 2/-;r) (3). 

Also putting ar + ^sl—x, for ar, 
/■{(aT + 4/)-ar}=/(aT + 2/-ar)-v|/{«T + 2/-(a; + c)} +0(«t + 4/-^) 

-x//{aT + 2/-(a; + c)} 
+ 0(aT+2/-a;) 

+ (i>{ar + 4!l-x). 
And similarly, we have 


/{»(.+ ^v^* 

^{«('^ + — )-(*' + c)}, 


f {"(-r +—)-(« + c)}, 


, ( r 2{n-l).l-\ . 


.^{«('^ + -|-)-^} 



In the same manner, 



+ ( 


(2 A 


./.{a(x+?i^^^)-(2/ + c-^)}. 



^{«(- + ^)-(2/-^)}. 

Hence we have at the time (t+ j, 

V =f(aT-x)-f{aT-(2l-x)} 

— &c. 

f, , r 2(n-l)J-\ , x> . , r 2(w-l)./n -^, ,J 

+ ^|/ {« [t + — j -(21 + C-X)} 

+ &c. 


+ ^{a[r + ^)-a^}-ct>{a[r+^)-i2l-a^)}, 





v=-f{aT-x)-f{aT-{^l-x)} , ^ 

r^n ^ it / 

. . . ^nl 

Similarly, we find 
as = f(fiT-x)+/{aT-{2l-x)}, 

+ 2,., {<t> [« (r + ^) -(.r + c)] + [« (^. 4-^^) -(2/ + C- .r)]}. ^ 


\.. .(D)(2). 

8. The function /(ar — x), in the expression for v, represents the 
velocity of any particle produced by the first wave, propagated 
along the tube from the original disturbance at A, so long as t 

is less than - ; and if this wave were reflected entirely from B, 


the first line of the above expression for v, would give us the velocity 
of any particle within the sphere of the reflected wave, the time t 

not exceeding — . 

With our supposition as to the original disturbance, the form of f 
T less than — I will be immediately known from that of (p. The 


other terms in the general vahie of v, shew how the general waves 
in which we have 


v, = /,(af-x), and v, = Jl{at-(2l-x)}, 

are formed by the superposition of successive waves, as the time 
increases. If the velocity becomes by this superposition so large, that 
it can no longer be considered extremely small as compared with 
the velocity of propagation (a), our analysis will be no longer ap- 
plicable ; but if V never exceed a certain value, the motion will 
become regular, and follow the law which our investigations indicate. 
Let us consider in what cases we may expect these effects to be 

9. We have at present imposed no restrictions on the forms of 
the functions denoted by cp, f and \//, except that their greatest 
values shall be small compared with a. In order however that the 
undulations may be sonorous, <p, and consequently y and \f/, must 
denote periodical functions, so that the values of (p {z), f (2), and ^ (ss), 
will recur as often as % is increased by a certain quantity. We will 
also iinpose an additional limitation upon them, to which, in all 
practical cases they will probably be subject very nearly, as will 
certainly be the case in the experiments to which I shall hereafter 
more immediately refer. Supposing then their values to recur, when 
s becomes %-Vm\, {m any whole number), we will also suppose them 

to recur with different signs when z becomes x±m' -; {m! being 

any odd number). 

10. First suppose the greatest value of \//, small as compared with 
that of y or 0, as must be the case in a closed tube. In the above 
expression for v, it will be observed that the quantity represented by 
% increases as we proceed from one term to the next, in a vertical 
line by 2/. 

Suppose then 

%l = m' . -, or l = m' - 

2 4 


In this case it is manifest that the consecutive terms taken in the 
order just mentioned will destroy each other ; and there will con- 
sequently be no accumulation of motion in the tube, and the 
vibrations will go on uniformly. Again, let 

2l = m\, or / = 2m. -. 


In this case the values of the successive terms taken as before in 
the expression for v will be equal, and with the same sign. Hence, 
if we take x of any value, except such as would render 

<(>{at-x) = <p {at-{9.l-x)], 

f which value of x is I — m -\ , it is manifest (since the value of (p 

is greater than that of \|/), that the motion will constantly increase 

for such points, and will soon become greater than is consistent 

with our original suppositions. Such a vibration then cannot be 
maintained. . 

11. Again suppose the functions (p, f, and ^, to be continuous, 
and suppose 

2/=m'^+2\', or / = m'^+\', 
2 4 

X' being any quantity less than -; the consecutive teims of 1.(f>(%), 


will not then destroy each other, but as the number of pairs of terms 
increases, the sum will increase till ^(s; + 2r/) becomes negative, it will 
then decrease, after having thus attained a maximum value. Maxima 
and minima values will thus occur alternately, and the same will hold 
for 2. >//(»). If these maxima values do not render v greater than our 
original suppositions allow, the vibrations may be maintained. 

Since these maxima values are 0, when l = m'.-, and greatest 

when l=m' .-, we conclude that they will be intermediate for inter- 
mediate values of I, following some continuous law. Hence we infer 


the possibility of maintaining sonorous vibrations of which the period 
is - , in stopped tubes of which the length differs considerably from 

?«' . - , particularly if the greatest value of V/ should not be very 

small. If the supposition we have made respecting the continuity of 
the function (p more particularly, should not be quite true, it is not 
likely in those practical cases to which we can best refer, to be so 
far wrong as to render the above reasoning otherwise than at least 
approximately true. 

12. Our supposition has been that the intensity of the distvu-bance 
denoted by v//, is considerably less than that indicated by (p, the tube 
being stopped with some substance having a certain degree of elas- 
ticity ; if the tube be open, it seems probable from certain pheno- 
mena, that the reverse of this supposition is true. 

Assuming this to be the case, the expansion of the expression 
for V may be put under a more convenient form. 


y{r {at-{2l+ c-x)} =2f{at-(2l- x)] - f, {at- (21+ c' -x)], 


v=f(at-x)+f{at-{2l-x)}-xj.,{at-(2l + c'-x)} (a), 

and equation (3) becomes 

/(aT + 2l-x)= -/(aT-x) + x|/, {aT-(x + c')} +<p(aT + 2l-x) (4). 

By proceeding exactly as in the former case, we obtain 

v = {-irif(aT-x)+flar-(2l-x)}} 

-fAci{'r+^)-(2l + c'-x)} 
+ 2,^,(-l)-{<^[a(T + ^) -X] + 0[« (t + ^y(2l-x)]} 





Similarly, we find 
«*=(-l)"{/(ar-^)-/[«T-(2/-a;)]} >^ 

+ ^l,,{a[T +—j-{2l■^■c-x)} 

+ 2,^,(-l)»-{<^[« (t + ?^) -.V] - 0[« {'r+~) - (2/ - x)\\. ^ 

Reasoning on the expression for v, exactly similar to that used 
above, will in this case show that sonorous vibrations cannot be 

maintained if / be too nearly equal to an odd multiple of - ; but 

that they can be continued, if / do not differ too much from an 

even multiple of - .* 

13. If we examine the expressions for as in the last article, and 
in Art. 7, it will appear that the condensations and rarefactions at 
the surface of the vibrating plate within the tube, are such as to 
produce forces opposing more strongly the motion of the plate as 
the lengths of the tubes approximate respectively to those particular 
lengths for which it will be impossible to maintain the vibrations in 

the tube ; and when the lengths differ from the above by - , these 

condensations and rarefactions are such as to promote the motion of 
the plate, instead of opposing it. 

14. The expanded expression for v may be put also under another 
form, which it may be useful to point out for the case in which 
the intensity of the disturbance denoted by \//, is considerably greater 
than that denoted by <^. 

* The quantity c' in these general inferences is not taken into account. Its value 
however is considerable, as will be seen hereafter. 


This is deduced, by assuming 

i,,{at-(x + c')}= (if (at -a;)+yl.'{ai-(x + c")}, 

x/. {«/- (x + c)} = (2 - /3) /(«^ -x)+ir' {at -(x + c")}. 

Then the equation (a) (Art. 12) becomes 

v=f{at-x) + {l-l3)f{at-{2l-x)}-i,'{at-{2l+c"-x)} (/3). 

We may observe, that since the vibration denoted by \j/, is pro- 
duced by that denoted by Jl it seems a necessary consequence that 
their periods must be the same. Their phases also are nearly so ; 
and if in addition we assume that the Jbrm of the function ex- 
pressing the one motion, does not differ very widely from that ex- 
pressing the other, (however the intensity of the vibrations may differ) 
it is manifest that /3 may be so taken that the intensity of the 
vibrations denoted by the unknown function \j^' shall be small com- 
pared with that indicated by <p. 

Equation (4) becomes 

f{ar + 2l-x)=-Cl-ft)f(aT-x) + i.'{aT~(x + c")}+(j>(aT + 2l-x) (5), 

= -hf{ar -x)^-^' {a-r - (^ + c")} + («t + 2/- x), 
if 1-/3 = *. 

This gives us 

And the equation (/3) becomes, (when t=T-\- j, 

v^{-hY {f{aT-x)+hflar-{2l-x)]} 

+ S,,,(-&)-|>/.'{«[t+ ^^^^^)-^ ]-(x+0}+&^^1«[t+ ^^''^^^-/ ]-(2/+c"-;»^)}I 

-^'{a(r + ^)-{2l + c"-x)]. 

r-n ^ W/ \ (t ) 

Vol. V. Part II. 1 1 


Since b is less than unity, and n soon becomes a very high 

number, after an extremely short time the first line in this expression 

may be neglected, as may also all the terms in the other lines in- 
volving high powers oi h. 

Whence it follows that the original disturbance (on which the 
form of the function f will depend), will cease in an extremely short 
space of time to have any effect on the form of the existing vi- 
bration, supposing the vibrations maintained by some cause distinct 
from that producing the original disturbance. 

Also, if the cause maintaining the vibrations cease, the vibrations 
themselves may cease in an extremely small space of time. 

The inferences we have drawn from the former developement (E) 
of the expression for v, may be drawn from this and perhaps with 
still greater facility. 

15. If we suppose >|/' (ss) always = 0, the expression for v will 
reduce itself to the same as that given by M. Poisson. But in this 
case it will be observed that all the functions involving the quantity c" 
disappear, which renders it impossible to account on this theory for the 
position of the modes or points of minimum vibration as determined 
by experiment*. For the purpose of determining the positions' of 
these points theoretically we will recur to the equations (C), the first 
of which is 

~ v = f{at-x)-f{at-{^l- X)} +^ {at -{2l + c - X)} (6). 

If we neglect ^{at-{2l+c — x)}, (or suppose the substance with 
which the tube is stopped perfectly rigid) we shall have » = 0, when- 

{at — x) - {at — {^l- x)}=Q, or mX, 

{m being any whole number), or when 

{l-x) = m.-. 

* See Art. 36, Sec. II. 


This condition is independent of t, and consequently at all points 

distant from the stopped end, any multiple of -, the motion will be 

the same as at that extremity, i.e. it will always equal 0, and there 
will be perfect nodes at those points. 

16. We may take the general case, and let 

f\at-{il-x)\-^ {a/-(2/ + c-ar)} =j(; {at-{<il->rc, -x)}, 
and :.v=f{flt — x) — x\a't—{^l^-Cx — x)\, 

^ being still small. The forms of J" and x// being known, that of ^ 
will be determined ; its period will also be the same as that of J" 
and ■^. It expresses the velocity of each particle produced by the 
whole wave actually reflected from B. The nodes will in this case 
be points of minimum vibration, and not of perfect rest. 

For the sake of clearness we will assume that y(x), and >//(x), are 
such that 

and therefore 


that y(»), and ^(z), {and therefore x(*)} admit of only one maximum 

value between x = 0, and 8;=-; and that the ratio which y(s!) bears 

to ylr (%) is always considerable, as by hypothesis it is when those 
functions have their maximum values. There can be little doubt but 
that these assumptions are at least approximately true in all practical 
cases ; and appear as simple as any we can make (and some must 
be made), in order to give distinctness to our inferences as to the 
positions of these points of minimum vibration. 

17. For the determination of c, in terms of c, let the origin of 
t and X be so taken that y(0) = 0, then making at- {2l — x) — 0, 
we have 

-^(-c) = x(-c,); 

or =\l/{ — c). 



By our hypotheses, x (*) must be always greater than \// (%) ; and 
if we suppose c and c^ less than the least value of z, which gives 
to ^ (%), or X (^) its maximum value, it is manifest that from this 
last equation, c, must be considerably smaller than c, and must be 

affected with a different sign. Suppose c^ = j^, where k is consider- 
ably greater than unity. It follows then that if the phase of the 
vibration of the extreme section of a stopped tube be retarded by a 
certain quantity c, the phase of the actually reflected wave will be 

accelerated by a quantity t. 

18. Giving then the proper sign to c„ we have 

v=f(at-x)-x{at-(2l-^-a;)} (7), 

and to determine the points of minimum vibration, we may observe 
that this expression is exactly the same, as if the wave for which 

v, = x{at-{2l-^-x)}, 
were reflected immediately from a section B' whose distance from A = l — —x. 

Suppose a rigid diaphragm at this section constrained to move 
exactly as the fluid does there ; we may then suppose the actual 
stop B removed, and the points of minimum vibration will remain 
the same. 

Now to determine them in this case, we observe that whenever 
at — x = at—{2l — T — x) + m\. 

the value of v will be the same as when 


at—x = at—{2l— T - x). 

In the latter case 


and in the former 

or l-^ = m\ + ^^', 

consequently, at any point in the tube whose distance from B" = m .-^, 

the velocity will be the same as at B'. These then will be points of 
minimum vibration in this hypothetical case, and therefore also, from 
what precedes, in the actual case. 

Making c = 0, we have l—x = m.-, which will give the positions 
of the nodes when there is no retardation. 

Hence we have this general conclusion with respect to the stopped 
tube — that if there be a retardation in the phase of the vibration of 
the extreme section, the positions of the points of minimum vibration 

will all be further from the stopped end by —j, than if there were 

no such retardation, the distances between these points respectively 
remaining unaltered. 

19. We will now consider the case of the open tube, in which 
we suppose >|/(a!) to be always considerably larger than J'{%). Assume, 
as in Art. (12), 

yl,{at-{2l + c-x)}-y{at-{^l-x)}~^, [at - {21 + c' - x)} (8), 

v=f{at-x)-¥f{at-{2l-x)}-^,{at-{2l + c'-x)}. 

First neglecting the function >|/, , v will = whenever 

f{at-x)^-/\at-{2l-x)}; ^ 
i. e. whenever 

at—x = at—{2l—x) + m'.- {m' an odd number), 
or I— x = m .-, 


a condition independent of /. Consequently, at every point whose 

distance from the open end is an odd multiple of -, there would be 
a perfect node. 

20. Put 

f{at-{2l-x)\-y},,{at-{2l + c'-a;)] =x {at- (2l+c,-x)\ (9). 


v=/{at-x) + x{at-{2l + c,-a;)} (10). 

To find the relation between d and c, we have from equation (8), 
(proceeding as in Art. 7, and with the same assumptions), 


or >|,i(c')= -x|/(c); 

and since >//(») is much larger than >/'i(i8), we shall have c'. considerably 
larger than c, and affected with a different sign. We may therefore put 

ki being greater than unity. 
Again from equation (9), 

-^.(-0=x'(-c.), • 

or x'(<=-^)=-Uc'). 

If we suppose x'(«) nearly equal to v//^,(i8), (which probably is not 
far from the truth), we shall have 

C2= —c' nearly, 

Hence in this case if the phase of the vibration of the extreme section 
be retarded by a quantity c, that of the actually reflected wave will 
be retarded by kic; and it will appear by the same reasoning as in the 
case of the closed tube, that the distance of the points of minimum 

vibration from the open end will be m' -r 1-, {m' being any odd 



21. If e and e' be the distances through which the nodes are moved 
by a supposed given retardation of phase, the same for each, at the 
extremities of the open and closed tubes respectively, 

e = — kki e ; 

6 will consequently be much larger than e'. 

The quantities m' i- in the open tube, and m- + -^ in the 

4 2 ^ 4 2« 

closed one, must be determined by experiment. 

22. I will recapitulate the principal inferences from this theory. 

I. In the tube AB, open at the extremity B opposite to that at 
which the vibrations are produced, there will be a series of nodes 

equidistant from each other by -, or half a whole undulation, the 

distance of the nearest node from the open extremity being considerably 

less than -, the whole system of nodes being thus brought nearer to 

the open end than the position assigned to it by the investigations of 
Euler or of M. Poisson. The distance of each node from the open 
end will be independent of the length of the tube. (Art. 20.) 

II. If the tube be closed at B, the nodes will still be equidistant as 

before by - . The distance from B of the node nearest that extremity 

will be - , or a quantity rather greater than that, if we suppose a cause 

of displacement of the whole system of nodes to exist in this case of 
the closed tube, similar to that which exists in the open one ; the dis- 
placement however being necessarily much smaller in the former than 
in the latter case, and in the opposite direction. (Art. 18.) 

III. These nodes are not places in which the air is perfectly at 
rest, but points of minimum vibration. (See Art. 16.) 


IV. Sonorous vibrations, whatever be their period, may be main- 
tained in a tube of any length, except that of which the length does 
not approximate too nearly to something less than an even multiple 

of J in the closed tube, or to an odd multiple of - in the open one. 

(Arts. 11, 12.) 

V. The intensity of the general vibrations in the tube varies with 
the length of the tube, being greatest for the lengths just mentioned, 
and least in the closed tube when its length is rather greater than an 

odd multiple of -; and in the open one, when it is something less than an 
even multiple of -r . (Art. 10.) 

VI. In these latter cases also of both tubes, the opposition afforded 
by the vibratory motion of the air within the tube, to the vibrating 
of the plate, is least; and greatest for the lengths which approximate 
to those mentioned in (IV.), as those with which the vibrations cannot 
be maintained. (Art 13.) 

VII. When the cause producing the vibrations in a tube ceases, 
the vibrations themselves may cease, not instantaneously, but in a period 
of time not exceeding the small fraction of a second, supposing the 
tube not to exceed a few feet in length. (Art. 14.) 

VIII. If we suppose the original disturbance to produce an un- 
dulation different in any respect to those produced by the cause which 
afterwards maintains the vibratory motion of the aerial column, this 
original disturbance will cease to affect the form of subsequent undula- 
tions in a period of time not exceeding the small fraction of a second, 
depending on the length of the tube*. (Art. 14.) 

* Similar inferences to the above may be drawn equally from M. Poisson's investigations, 
except that the phenomena according to his solution would take place for lengths of the open 
tube materially different from those above-mentioned. 



23. I WILL now proceed to describe the experiments which have 
been made with a view of putting the different theories on this subject 
to an experimental test. Sonorous vibrations are usually excited in a 
tube, either by directing a stream of air across the open end, as in 
blowing across the embouchure of the flute; by means of a vibrating 
tongue, as in all reed instruments ; or by placing an open end of the 
tube close to the surface of a vibrating body. In the two first cases it 
seems impossible to conceive that the same disturbance can be com- 
municated to each part of the extreme section of the air in the tube 
where the original motion is produced, a condition which is always 
assumed to hold at least approximately in all our mathematical investi- 
gations of the subject. This irregularity of the motion will no doubt 
extend to some distance within the tube, and it is impossible to say 
how it will affect the phenomena even in those parts of the tube in 
which the motion may become more uniform. In the second case too 
in particular, a stream of air must constantly be passing through the 
tube, a circumstance not contemplated in our analysis of the problem. 
This may or may not influence materially the observed phenomena, 
but at all events the danger of derangement from any such cause 
must be avoided, if we would render our experiments decisive tests 
of the truth of any theory professing to account for phenomena of so 
delicate a nature as those which are now the objects of our investigation. 
The third method, however, above-mentioned, is entirely free from the 
latter objection, and may be made almost entirely so from the former, 
and is, therefore, that which I have adopted. 

24. The apparatus is very simple. Figure I. represents it. A 
plate of common window glass is held firmly in a horizontal position 
by a pair of pincers at its middle point. AB is a gltiss tube, having 
a short brass tube closely sliding within it at the upper end B, so 
that the whole tube AB can be lengthened or shortened at pleasure. 
Within the tube a small* brass frame M, having a delicate membrane 

* Fig. (2) represents this frame with the membrane ab, which may be tuned, or rendered 
sensitive in different degrees, to the vibrations produced by any proposed note, either by 
Vol. V. Paet II. K k 


stretched across it, is suspended by a fine wire or thread from the upper 
extremity of the tube, in such a manner that it can be heightened or 
lowered at pleasure. The other parts of the apparatus are merely such 
as are adapted for facility and -accuracy of arrangement of the tube 
and plate. 

25. The air in the tube is put in a state of sonorous vibration 
by means of the plate, which is made to vibrate by drawing the bow 
of a violin equably across its edge in a direction perpendicular to its 
plane ; the vibratory motion of the air is communicated to the membrane 
suspended in the tube, and the degree of motion is indicated by the 
agitation of a small quantity of light dry sand sprinkled upon it*. 
Suppose the tube to be open at the upper end B, and let the membrane 
be drawn up near that extremity. Tf the sand indicate a considerable 
motion when the plate is vibrating, let the membrane be gradually 
lowered ; a position will thus be found in which the sand has little 
or no apparent motion, thus indicating the existence of a node. On 
lowering the membrane still further, the sand will again become strongly 
agitated, and will then come to another place of rest, (or at least of 
minimum vibration), and so on till it reach the lower end of the tube. 
These alternations of points of rest and motion can of course only take 
place when the tube is sufficiently long in comparison with the length 
of an undulation produced by the vibrating plate, to admit of them. 
These nodal points are thus found to be at equal distances from each 
other, the distance of the upper one from the top of the tube being less 
than half that between the nodes. This is independent of the length of 
the tube. These results are accordant with our theory, (Art. 22, I.), from 
which it appears that this constant distance between two consecutive 

nodes must be -. 

If we call the distance of the upper node from B, -— C, C denotes 
what I have termed the displacement of the nodes. 

altering the tension by means of the small cylinder round which the end b of the membrane 
passes, or by moving the small bridge cd, and thus altering the length of the vibrating part. 

* This was the method adopted by Savart in such a variety of caseSj in which he wished to 
ascertain the intensity of sonorous vibrations in air. 


26. If the membrane be rendered very sensitive by being exactly 
tuned to the note produced by the vibrating plate, it will not indicate 
perfect rest at the nodal points, shewing them in fact to be points of 
minimum vibration, which agrees with our theory, (Art. 22, III.). 
With such a membrane it will be difficult to determine the position of 
these points with accuracy, and its sensibility should be diminished, 
till the sand appears perfectly at rest when it is placed exactly at the 
node. If the membrane be rendered still less sensitive, it will appear 
at rest for a space on each side of the node, the position of which will 
in such case, be determined by observing those points immediately 
above, and below the node at which the motion of the sand is just 
sensible. The middle point between them will of course be the 

27. Now suppose the length of the tube to be any odd multiple 

of -, and the membrane to have such a degree of sensibility, as just 

to remain at rest only when placed in a node or within a very small 
distance of it. After it has been placed in this position, let the brass 
tube sliding within the upper part of the glass one be raised through a 

space less than - . While the whole tube is thus lengthened, let the 

distance of the membrane from the upper end B remain the same; 
the membrane will consequently be still in a node. The plate being 
now put in vibration, the membrane will remain perfectly at rest, not 
only in this position, but also when moved to one considerably above 
or below the node, the new length of the tube remaining the same. 
This indicates a less degree of motion in the tube than in the former 
case, and we find that the intensity of the vibration in the open tube 
is least when its length is equal to something less than an even 

multiple of -T, or 2m.j — C; and becomes greater as the length 

approximates to rather less than an odd multiple of -, or {2m' + 1)-—C, 

m and m' being any whole numbers. (Art. 22. V.). This diminution of 
motion is also very obvious when the membrane is placed in those 



parts of the tube where the motion is most sensible. In all cases, 
however, the distances of the nodes from B is independent of the 
length of the tube. 

28. If we take a tube closed at B instead of the open one, we 

observe the same alternations of points of greatest and least vibration, 

and (the plate being made to vibrate in the same manner as before) 

at exactly the same distances from each other as in the closed tube; 

but the distance of the upper node from the closed extremity of the 

X I 

tube is now observed to be -, the same as the distance between the 


nodes. Proceeding as in the former case, it is found also that the 

strongest vibrations are excited when the length of the tube is about equal 

to a multiple of - ; and the least vibrations when the length = an odd 

multiple of - . I find also that in the open tube stronger vibrations exist 

in the nodal points than for corresponding cases of the closed tube. 

29. In performing the above experiments with reference to the 
intensity of the vibrations in the tube, care must of course be taken 
to prevent the influence of any other cause than that of which I have 
spoken, viz. the length of the tube with respect to X. It has been as- 
sumed that the vibration of the part of the plate immediately in contact 
with the mouth of the tube is in all cases the same, which requires 
that the tvibe should always be placed over exactly the same portion 
of the plate. This portion also should be included in one and the same 
ventral segment; for if a nodal line on the plate pass across the mouth 
of the tube, the vibrations transmitted from opposite sides of this line 
will be in exactly opposite phases, and will consequently neutralize each 
other in a degree depending on the ratio which the intensity of one 
of these undulations bears to the other. If the nodal line divides the 
part of the plate in contact with the mouth of the tube into two 
equal portions, parts of similar ventral segments, the interference 
will be so complete as to destroy all sensible motion in the 


tube*. It is only however as regards the intensity of the vibrations 
that this precaution respecting the relative position of the nodal 
lines and mouth of the tube is important ; it does not affect the 
positions of the nodes. The reason is obvious — it does not affect the 
value of X. 

30. Again, taking the tube open at B, let the extreme section 
at A be made to coincide nearly with the surface of the vibrating 
plate. If the plate (the bow being applied to it) vibrate freely, let 
the length of the tube be gradually increased or diminished. It will 
thus be found, that as the tube approximates to certain lengths, the 
plate vibrates with less facility, requiring a greater pressure of the 
bow, and continuing to vibrate audibly for a shorter time after its 
removal; and in many cases, between certain limits in the length of 
the tube, it becomes almost impossible to make the plate assume that 
state of vibration which it assumes freely for other lengths ; and the 
vibration, if it be produced, appears to cease almost instantaneously 
on the removal of the bow, instead of being audible for several 
seconds, as it would be if the tube were removed, or were of a 

different length. These phenomena recur for every increase of — in 

the length of the tube ; and if I be any length with which it becomes 
almost impossible to make the plate vibrate in the manner proposed, 

then will / + - be that length with which it vibrates with the same 

facility as if the tube were removed. 

* It is easy by a very simple experiment to give ocular demonstration of the fact that the 
union of two intense sounds may produce perfect silence. Take a branch tube ABA' (Fig. 3.),"and 
stretch over the open end B a fine membrane or a piece of common writing paper. Place the 
open extremities A, A' of the equal and similar branches CA, CA' over portions of two ventral 
segments of a vibratory plate in the same phase of vibration. A small quantity of sand strewed 
over the membrane at B, will immediately shew it to be in a state of strong vibration. Let A, A 
be then carefully placed over suiiilar portions of similar ventral segments of the plate, in opposite 
phases of vibration ; the sand on the membrane will remain perfectly at rest, shewing that the 
waves propagated along AC and A'C in opposite phases so completely interfere at Cas to produce 
no undulation along CB. In other words, no sound would in this case be transmitted along the 
tube to its mouth B. 



So far these phenomena are in accordance with the results of 
theory, (Art. 22, VI.) ; but when we examine the length I just men- 
tioned, we find it entirely at variance with them. In fact on 
investigating the circumstances more narrowly, we find that the value 
of / depends in a considerable degree on the small distance between 
the vibrating plate, and the extreme section A of the tube, a cir- 
cumstance which nothing in our theoretical deductions has led us 
to anticipate. This will be seen in the results of the following ex- 
periment made with an open tube. 

Diameter of the tube = 1 . 35 inches. 

Value of- ,.=4.82 for temperature 63°. 

Position of the mouth (^A) of the 
tube (Fig. I.) 

Value of the length / 
above mentioned. 

Theoretical value of /. 

As close to the plate as 
possible without interfering 
with its vibrations 

About T^ inch from thel 

lo I 

vibrating plate. 

.12.25 inches. 

* 1 1 . 46 inches. 

12. 6 

31. This discrepancy however between the theoretical and ex- 
perimental results is only apparent. It arises from the circumstance 
of one of the conditions assumed in our mathematical investigation, 
not being accurately satisfied, namely, the perfect prevention of all 
communication between the external air and that within the tube at 
the extremity next the plate. And this is easily proved experi- 
mentally, by placing the extremity of the tube as near as possible 

* In this value of Z I have taken account of the displacement of the nodes, which is .59 
inches, as determined by experiment. (See Table, Art. S6.) 


to the surface of the plate, without interfering with its vibrating 
motion, and then putting round the edge of the tube, a small 
quantity of fluid which by its adherence to the tube and the plate 
fills up the interstice between them, and prevents communication with 
the external air. When this precaution is taken, the lengths of the 
tube which correspond to the above mentioned phenomena exactly agree 
with theory; that is — 

The .vibration of the plate is unaffected by the presence of the open 
tube, Avhen its length is equal to something less than an even multiple 

of — , or 2 m. J— C, and of the closed one when its length is equal to 

4 4 

an odd multiple of -; but as the lengths of the tubes approximate 

respectively to quantities differing by - , from the above lengths it 

becomes almost impossible to make the plate assume the same vi- 
bratory motion. (Art. 22, VI.) 

32. It might at first appear probable that the neglect of this 
precaution might have some effect on the position of the nodes, as 
well as on the phenomena above mentioned. This however is not 
the case; and the reason will be obvious if we recollect that the 
position of the nodes depends on the periodicity of the vibrations, or 
the value of X, which is unaffected by the communication with the 
external air at A ; whereas the force opposing the vibration of the 
plate depends on the condensations and rarefactions of the air, at the 
surface of the plate within the tube, which will necessarily be much 
affected by the communication just mentioned.* 

33. If we take a closed tube, a similar discrepancy or accordance 
in the results of theory and experiment will be found under the 
same circumstances as above described. 

* It does not appear so easy to account for the phenomena as above described, when the 
influence of external air is not prevented. This, however, does not immediately belong to the 
object I have proposed to myself in this paper, which is, to establish as accurately as possible the 
identity of the results of theory and of experiment in those cases in which the conditions assumed 
in our mathematical investigations are experimentally satisfied. 


The phenomena above mentioned, agree with those observed by 
Mr Willis, and described in his paper on the Vowel sounds, pub- 
lished in the Transactions of this Society, Vol. III. The manner 
however in which his experiments (having a different object from 
mine) were conducted, render them unfit for the verification of any 
of our mathematical results in this subject. 

34. From what I have above stated, respecting the difficulty of 
making the plate vibrate with certain lengths of the tube, it is manifest 
how we may avail ourselves of this phenomenon, in the determination 
of the value of X, corresponding to any particular mode of vibration 
of the plate, supposing those particular lengths of the tube can be 
ascertained with sufficient accuracy. Now this can be done almost 
as accurately as the position of a node can be determined by the 
vibrating membrane, and consequently the value of X may thus be 
found. For if A and 4 denote two observed values of /, we shall have 

— = , n being a whole number easily ascertained. (See Arts. 30, 31.) 

35. Though I have had frequent occasion to speak of this displace- 
ment of the nodes in the open tube, from the positions assigned to them 
by the common theory, I have hitherto said nothing as to the ex- 
perimental determination of its magnitude. The most direct way of 
accomplishing this, is to determine the actual positions of the nodal 
points by means of the vibrating membrane ; but this method becomes 
inconvenient when the diameter of the tube is small, as, for instance, 
less than an inch. Those which I have used most commonly are 
from 1.3 in. to 1.5 in. diameter. If the tube be larger than this, it 
will generally be too large to admit of the extreme section of it 
being placed entirely upon the same ventral segment of the plate, 
as is always desirable, (see Art. 29) ; and if much smaller it becomes 
necessary to make the surface of the membrane so small as to be 
inconvenient, in order that it may not bear too great a ratio to the 
area of the section of the tube, in which case the presence of the 
membrane might be supposed to render the vibrations in the tube 
materially different from what they would otherwise be. 


The best method therefore of determining the positions of the 
nodes in tubes considerably smaller than those I have mentioned, is 
that by which the value of \ is determined, as described in the 
last Article. 

Thus, suppose / to be the length of tube, with which it is found 
most difficult to make the plate vibrate ; then (the tube being open) 
we shall have 

l={2m + l)^-C, 

where m is a whole number, which will be known when \ is de- 
termined by either of the methods pointed out above. The quantity 
C evidently shews how much the distance between the open ex- 
tremity, and the nearest node differs from — , or it expresses the 

From the above equation, 

C={2m + l))-l, 


and the displacement is thus determined. 

36. The following table exhibits the magnitude of this displace- 
ment in a tube of given diameter, as determined experimentally for 

different values of - . The positions of the nodes were in these cases 

carefully ascertained by means of the membrane suspended in the 

Vol. V. Paet II. L i. 



Diameter of the tube = 1.35.* 

Value of — . 
at temp. 63". 

Computed dist. of a Node 
from B, (fig. 1.) 

Observed dist. of the 
same Node. 

Displacement of 
the Node. 


4. 82 

I 7.15 






} mean = .36.'i 



The above values of - were determined by means of a membrane 

2 •' 

and a tube closed at the upper end, nearly 100 inches in length. The 

distance of a node from the closed end being found = b, we must 

Or, if bf be the observed distance, sub- 

have n . - = o, or - = - , 
2 2 w 

ject to an error /3, and therefore b ± (i the true distance, we have 
- = - + —. The value of /3 will probably be less than ^ inch, and 

t^ Tt ft 

in the determination, for example, of the first of the above values 

of -, « was about 45, so that that value of - may probably not be 

subject to an error exceeding .001 inch. We may also remark, as an 
indication of accuracy in the numbers 10.88 and 6.78, given in the third 

* The measures are all given in inches. 

t In the determination of the quantity b, the temperature at the time of observation must 
be carefully noted, since the variation in the velocity of aerial undulations produced by a varia- 
tion of temperature of even less than 1°, is sufficient to make a very sensible difference in the 
value of h, this value being as much as nearly 100 inches. 

Since the distance of any proposed node from the upper end of the tube will be proportional 
to the velocity of the undulation, it is manifest that by observing the values of b, corresponding 
to different temperatures, we may estimate directly the effect of temperature on the velocitj' of 
sound. This method is capable of great accuracy. 



column, that 10.88-6.78 = 4.10 must =2.-, which gives us - = 2.05, 

differing but .006 from the more accurate value. The error in the 
two numbers above mentioned, 10.88 and 6.78, does not probably exceed 
.01 or .02, and cannot, I conceive, exceed .04, and consequently, I think, 
the utmost limit to the error in the corresponding numbers in the 
fourth column cannot exceed .05, and is probably considerably less. The 
same may be concluded respecting the numbers .47, .59, in the same 

The above results may, then, be considered sufficiently accurate to 
determine the fact of the magnitude of the displacement increasing 
with increased values of X, though not sufficiently so to determine with 
certainty the law of this corresponding increase. 

The displacement does not depend only on the value of \ ; it depends 
also on the area of the mouth of the tube, as appears from the following 

Values of ^ • 


Diameter of tube = 1.35. 

Diameter of tube = .8. 




These values of the displacement of the nodes have been obtained 
by the method mentioned in Art. 35, as that best applicable to small 
tubes. The results in the second column of this table ought to be 
the same as the two first in the last column of the former table; but 
this method is liable, I conceive, to greater error and uncertainty than 
the former, and to this, I doubt not, the discrepancy is due.' All 
these latter results, however, are probably subject to an error of the same 

L L2 


kind, and are too small both in the large and' small tube. They can 
leave no doubt of the fact of the magnitude of the displacement being 
dependent on the diameter of the tube. 

It is important to observe, that the values of X determined in the 
large tube and the small one, from the consideration that the distance 

between any two nodes must equal some multiple of - , was exactly 

the same, being for the first case in the table 2.05, very nearly agreeing 
with the accurate value 2.044. This proves that the distance between 
the nodes is independent of the diameter of the tube, provided the dis- 
turbance take place uniformly throughout its extreme section. 

37. I have before remarked, that there can be nothing arbitrary 
or indeterminate in the vibratory motion of the air at the extremity 
of the open tube when the vibrations in it are excited according to 
some known law ; and consequently, if our theoretical knowledge of 
the subject were complete, we should undoubtedly find in our investiga- 
tions the cause of the retardation of phase, of which I have spoken, 
in the reflected wave of the open tube, supposing it to be the actual 
cause of that displacement of the whole system of nodes which I have 
established as an experimental fact. Our knowledge at present, how- 
ever, is totally inadequate to this purpose, and therefore we can only 
conjecture what may be the probable cause of this retardation in the 
reflected wave; but at all events, our formulse, with the modifications 
1 have introduced into them, do become perfect representations of all 
those phenomena which we can distinctly determine by experiment, 
in the cases to which our mathematical investigations apply. The fact 
too, of a retardation of phase in the reflected wave may not be very 
difficult to conceive, or appear improbable, if we suppose the undulation 
proceeding from the open end of the tube to advance through a certain 
space before it assumes that form in diverging into free space, which 
it must ultimately assume when it sends back no reflected wave from 
any point of its path. Before it reaches this state, a partial wave may 
be reflected in its course from each point towards the tube; and an 
indefinite number of these reflected waves will form a general reflected 


wave, of which the period will be the same as that of each of its 
component waves, but the phase of which will be retarded as compared 
with that of a wave reflected immediately from the extremity of the 
tube. This is equivalent to our supposing a certain space beyond the 
extremity of the tube as subject to a disturbance (acting at consecutive 
instants along this space) such as to produce a wave diverging in all 
directions, and consequently sending a portion of this general wave 
back along the tube. 

To give generality to the investigations of the preceding section, 
I have considered the effect on the position of the nodes which would 
be produced by any retardation of the phase of the wave reflected from 
the stopped end of a tube. It appears, however, that there is not in 
this case any displacement of the nodes appreciable by the mode of 
experimenting I have described. The only reason, in fact, for supposing 
any retardation of phase in this case, is founded in the imperfect 
analogy between the cases of the open tube and the tube closed with 
an elastic substance. The cases are far too different, however, to admit 
of any thing but vague inferences from such analogy ; and it is 
manifest that no reasoning similar to that above applied to the open 
tube, can be applied to the closed one. If any retardation do exist in 
this case, I can only conceive it to arise from a cause similar to that 
suggested by Mr Willis*, viz. that time must be necessary for the 
action between the elastic stop and the air to produce its effect. This, 
however, appears much less probable in this case than in that which 
suggested the idea to Mr Willis, in which the action between the air 
and the vibrating body (a membrane) was lateral instead of being direct, 
as in the present instance. I have not been able to detect any indica- 
tion of such law of force in a displacement of the nodes in the closed 
tube, though I have examined the case with great care, conceiving 
that any facts bearing directly upon the nature of the mutual action 
of two elastic media at their common surface must necessarily be of 

* Cambridge Transactions, Vol. IV. Part III. p. 346. 


The experimental deductions in the preceding part of this section 
are based on the evidence afforded by the exploring membrane, because 
it is more direct than any other evidence which the phenomena appear 
to admit of, and therefore better calculated to supply those decisive 
and positive tests for ascertaining the accuracy or fallacy of our theoretical 
results, which it is my object to supply. We have seen the perfect 
accordance of these results with the general indications of the membrane, 
and also with the striking and well-defined phenomenon of the im- 
possibility of making the plate vibrate in a certain manner with tubes 
of certain lengths. It remains for us to consider also how far our 
theory agrees with the phenomena of resonance, in those cases in 
which the conditions assumed in our mathematical investigations are 
satisfied, viz. where the communication between the external air and 
that in the tube at the surface of the plate is prevented, and the 
disturbance extends uniformly over the whole orifice. In such cases 
it will appear from the following enunciation, that the intensity of 
the sound is proportional to that of the aerial vibrations, as indicated 
by the membrane, and by the difficulty or facility with which the 
vibrations of the plate may be maintained. (See Arts. 27, 31.) 

The resonance of the open tube is scarcely perceptible when the length 
of it does not differ much from something less than an even multiple 

of -, or 2m • j - C ; but as it approximates to something less than an odd 

multiple of that quantity, or (2m'+ 1)- — C, the resonance increases, and 

at length becomes of painful intensity, increasing till it is no longer possible 
to maintain the same mode of vibration of the plate. Whether the length 
of the tube be gradually increased or diminished in approximating to 
the above-mentioned lengths, the phenomena are precisely the same. 

I was the better pleased to obtain this result, inasmuch as those 
which I first obtained (when the precaution of preventing communication 
with the external air was not attended to*), as well as those of previous 

* In such cases the resonance was always greatest (as in the case considered in the text) 
when the difficulty of making the plate vibrate was greatest. The corresponding lengths of the 
tube may be seen in Art. 30. 


experimenters, appeared either to contradict theory, or at least to be 
altogether anomalous. According to our common notion on the subject, 
an open tube gives the strongest resonance when its length is nearly 

equal to an even multiple of 7, instead of an odd multiple, as above 

stated ; and Savart* has given this as the result of his own experiments 
for tubes of about the same diameter as those I have usually employedf ; 
but asserting also that the length is less as the diameter is increased, 
and this too whether the disturbance extend over the whole orifice of 
the tube or not. My results, however, are entirely at variance with 
this latter assertion, for I confidently conclude from them that if the 
disturbance extend uniformly and equably over the orifice of the tube, 
the phenomena will be independent of its diameter:]:, with the exception 
of the effect it may have on the displacement of the nodes |. If, however, 
the disturbance extend but partially over the orifice, I see no reason 
to doubt the accuracy of the last-mentioned results of M. Savart ; and 
this supposition will also account for the apparent discrepancy between 
his results and mine as respects the length of the open tube (of which 
the diameter does not much exceed an inch) producing the greatest 
resonance; for it is manifest that with this partial disturbance none 
of that condensation and rarefaction on the surface of the plate can 
take place, which in my experiments necessarily attends, and may be 
considered as causing, that powerful resonance of which I have spoken. 
It is easily seen, in fact, that when the length of the tube is neany 

equal to an odd multiple of -, the phase of the waves reflected from 

any considerable part of the orifice not occupied by the vibrating plate, 
will be directly opposite to that of the waves propagated by the plate 
itself; and that thus a great part of the vibration within the tube will 
be destroyed by interference. 

There is no difficulty, therefore, in explaining the non-existence of 
resonance in this case. If the tube, however, be lengthened or shortened 

by about - , (still supposing the disturbance at its mouth partial), a 

* Annates de Chimie, Tom. XXIV. p. 56. t See Art. S6. 

t See Art. 36, p. 264. § Art. 36. 


resonance will be heard, though extremely feeble as compared with 
that I have found in my experiments. This is, in fact, the kind of 
resonance which has been observed by all experimenters. It does not 
appear to me to admit of the same obvious explanation which the 
other admits of ; that which is usually received being, as I conceive, 
in itself insufficient, when subjected to those restrictions which must 
be imposed upon it by the general laws which govern the communication 
of motion from one particle of matter to another. At present, however, 
it is not my object to enter on the discussion of this and of some 
other points relative to this part of the subject. It is sufficient for 
me now to have shewn that that powerful resonance which I have 
observed in my experiments is exactly accordant with the results of 
our mathematical investigations, when the conditions assumed in those 
investigations are fully satisfied. I hope to return to the careful examina- 
tion of other cases at a future period. 

I have already alluded* to a paper by Mr Willis, published in the 
Transactions of this Society, in which he has described some experiments 
bearing on this subject, and affiarding a general corroboration of some 
of the results above stated. He fixed a reed to a sliding tube, and 
observed the intensity of the sound, when the reed was made to speak, 
produced by different lengths of the tube, and by means of a microscope 
carefully adjusted, he was able to observe the excursions of the reed 
in its vibration, and to obtain micrometer admeasurements of them. 
He thus found that when the length of the tube equalled about an 

even multiple of - , it gave the exact note of the reed with no perceptible 

resonance. As the tube was gradually lengthened, the tone was flattened, 

and as the length approximated to about an odd multiple of -, the 

extent of the reed's excursions was diminished, its vibrations became 
irregular and convulsive, till at length it ceased to produce any musical 
tone. When the tube, however, was a little lengthened beyond this 
point, the reed suddenly assumed its original form of vibration, 
producing a note of painful intensity, similar to that which I have 

* See page 260. 


described in my own experiments, although the extent of excursion of 
the reed was in this case less than in that in which no resonance was 

One discrepancy is observable between this experiment and mine, 
inasmuch as the intensity of the sound, instead of increasing as the 

length of the tube approximated to the odd multiple of - , as in my 

experiments, gradually decreased*. The explanation, however, of this 
fact, is easily found in the diminished excursion of the reed, and still 
more, I suspect, in the irregularity of its vibration, by which the 
undulations produced by it are probably rendered imperfectly sonorous^. 
With this explanation of this apparent discrepancy, the general results 
of Mr Willis's experiments afford as strong a corroboration of those 
Avhich I have obtained, as the difference between our modes of experi- 
menting will allow. The flexibility of the reed, however, and its 
consequent ready obedience to the vibrations of the air, as compared 
with the inflexible obstinacy of a glass plate, together with the partial 
disturbance produced by the reed, render it a totally unfit agent in 
obtaining experimental tests for our mathematical results, though it 
presents to us in its own motions many interesting points of enquiry. 

Our theory will also perfectly account for one of the most striking 
phenomena observable in wind instruments, viz. the rapidity with which 
different states of vibration are assumed within the tube, corresponding 
to different effective lengths of it, as determined by the opening or 
closing of the finger holes. We have seen (Art. 22, VII. VIII.) that 

* For a very clear and distinct account of these experiments, I must refer the reader to the 
excellent paper from which the above is taken. It will be observed, however, that the results 
mentioned in the text were not the direct objects of Mr Willis's investigations, but were such as 
naturally offered themselves in the course of his experiments on the production of the vowel 

t I think it very possible that \heform of the aerial vibrations may have more to do with our 
sense of the intensity of sound than has been generally supposed ; and perhaps some cases of 
resonance may admit the most satisfactory explanation on this hypothesis. 

Vol, V. Part XL M m 


according to theory, if the cause maintaining the vibratory motion in 
a tube be suddenly changed, (as in passing from one note to another), 
the effect of the former mode of disturbance on the form of the 
succeeding vibration will become inappreciable in an exceedingly short 
period of time. Now in the most rapid musical passages, the number 
of notes played in a second never probably exceeds ten or twelve, 
and these usually embrace only the higher notes of the scale, for 
which there must be many hundred vibrations in a second. Suppose 
this number, however, not greater than about two hundred ; any undula- 
tion transmitted from the reed or embouchure would still be reflected 
about twenty times at the open end in the interval between two 
consecutive notes in the most rapid musical passage. Now assuming 
unity to represent the intensity of a wave incident at the open extremity 
of the instrument*, let 1 — )3 represent that of the reflected wave, 
(1 — /3)". will represent (at least sufficiently approximately) its intensity 
after n reflections ; and consequently, as we have no reason to suppose /3 
very small as compared with unity, it is probable that after five or six 
reflections, the intensity of this wave will be quite inappreciable. Hence 
the apparently instantaneous cessation of sound after the exciting cause 
has ceased, and the most rapid transition from one note to another, 
are perfectly accordant with theory. 

M. Poisson, in the Memoir referred to in the early part of this 
paper, has also investigated the vibratory motion of air in two tubes 
of different diameters united together at one extremity. I hope to 
examine this case also experimentally. His results must necessarily be 
erroneous, as far as they depend on the physical condition he has assumed 
to exist at the extremity of the open tube, and which I have shewn to 
be inconsistent with observed phenomena in the uniform tube. 

* See Art. 14. 


St Peter's College, 
i\r««/ 20, 1833. 


XI. On the Latitude of Cambridge Observatory. By George Biddell 
Airy, M.A. late Fellow of Trinity College, Plumian Professor 
of Astronomy and Experimental Philosophy, and one of the 
Flce-Presidetits of the Society. 

[Read April 14, 1834.] 

The accurate determination of the latitude, with an instrument 
like the Mural Circle now in use at the Observatory, seems at first 
sight to be an easy business. In practice, however, it is not without 
difficulties. I do not here allude to the correction for refraction ; 
since, though there may be a trifling uncertainty in regard to its magni- 
tude, it is easy to leave , a result subject to that uncertainty, and 
admitting of correction without any trouble whenever a correction of 
the refraction shall be established. Nor do I allude to the uncertainty 
in the corrections by which, from a star's apparent place on any day 
of observation, its mean place at a fixed epoch can be inferred; since 
the uncertainty about any of these is far less than the smallest quantity 
for which we could pretend to answer in fixing the latitude of any 
place ; and its effects being periodical, would in a comparatively short 
series of observations, produce no sensible effect. The difficulties to 
which I allude are instrumental: they are not periodic in time, like 
the latter; nor do they admit of correction from posterior researches, 
like the former of the causes of uncertainty which I have mentioned ; 
they are moreover such as would scarcely be suspected to exist, until 
their effects are discovered from the discordance of the results of 

The Mural Circle is an instrument which gives simply the reading 
of that point of the graduated limb which is opposite to an imaginary 
fixed index when the telescope is pointed to the object of observation. 

M M2 


A single observation therefore gives us no tangible result. It is 
necessary to have one other observation, or a series of observations, 
by which the reading of that point of the limb can be found which 
is opposite to the same index when the telescope is directed to some 
point of reference; then the difference between this reading and the 
former is the angular distance of the object observed from the point 
of reference. It was intended originally by the maker that this point 
of reference should be the celestial pole. In practice, however, it is 
found necessary to descend one step nearer to terrestrial things, and to 
adopt for the point of reference the zenith ; a point which, though not 
marked any more than the pole by any obvious phenomena, can yet 
be discovered by a process which involves less of astronomical assump- 
tions, and requires a shorter time for the complete determination. 

The method of determining the zenith point from observations 
by reflexion at the surface of mercury, has been introduced into 
observatories almost entirely by the practice of the present Astronomer 
Royal at the Greenwich Observatory. The use of two similar circles 
(as at Greenwich) makes the process one of little labour, though requiring 
the co-operation of two observers. The same celestial objects being 
repeatedly observed by direct vision with both circles, the differences 
of the corresponding readings of the two circles are found ; and any 
observations made with one can be referred to the other. Then when 
any bright star passes the meridian, one circle is employed in observing 
it by direct vision, and the other at the same time is employed in 
observing it by reflexion at the surface of mercury ; the reading of the 
latter circle is referred to the former circle; and then the reading 
which is a mean between the reading for the direct observation and 
the referred reading for the reflected observation, is the reading that 
corresponds to a horizontal position of the telescope; and by adding 
or subtracting a quadrant, the reading which corresponds to a zenithal 
position of the telescope is obtained. 

With a single circle this process cannot be adopted. In some 
instances it has been imitated by observing a star directly on one 
night, and observing the same star by reflexion on another night. The 


calculation for the zenith point then relies on our perfect acquaintance 
with the variations of refraction and other corrections from one night 
to another ; and thus a cause of inaccuracy is introduced, which does 
not exist in the other method. In the Cambridge Observatory a different 
method is regularly employed (for the idea of which I am indebted 
to a suggestion of Mr Sheepshanks). When a star is to be observed 
by reflexion, the circle is set approximately for the reflected observation, 
and the six microscopes are read; when the star has entered the field, 
and before it has reached the center, it is bisected by the micrometer 
wire, (which in fact measures its distance from the fixed wire, and thus 
gives a correction to be applied to the mean of the six microscopes,) 
and then there is ample time to allow the circle to be turned to the 
position in which the star can be observed directly, shortly after it 
has passed the center of the field. Thus a direct and reflected observa- 
tion are obtained at the same transit. This method is, in my opinion, 
much preferable to the second that I have mentioned, and in some 
respects superior to the first. 

Either of the methods which applies to one circle enables us, as 
will shortly be seen, to examine severely into the consistency of the 
results obtained in different positions of the circle ; and this must be 
considered as a most valuable property of this method of determining 
the zenith point, and one which places it far above the use of a collimator 
or any similar instrument. 

I had hoped, on commencing observations with the Mural Circle 
at the beginning of the year 1833, to be able in a very short time to 
obtain a very approximate latitude. I proposed to observe some stars 
every night in the manner above described, as well as circumpolar stars 
(which might or might not be observed in the mercury): by the former 
I should obtain a very good zenith point; and then each observation 
of the latter, above and below the pole, would give me a value of the 

But after a few nights' observations, I found that the reading for 
the zenith point, as determined by different stars, was not the same. 


Had the discordance been wholly without regularity, this would have 
given me no anxiety. But the first Aveek's observations enabled me to 
see with certainty that one general rule could be laid down : the reading 
for the zenith point as determined by northern stars was invariably 
greater than that fovmd from southern stars. As the readings increase 
while the telescope is turned towards the south, this discordance is of 
the same kind as that which would be produced if the object end of 
the telescope dropped by its own weight. 

After much anxious thought and many fruitless attempts to explain 
this discordance, I was obliged to give it up entirely. The method 
which was adopted for approximate reduction of the observations, easily 
admitting of future correction, was the following. When in one night, 
or in several nights which it appeared practicable to group together, 
stars had been observed by reflexion in different parts of the meridian, 
1 took the three means of zenith points determined by stars far north, 
by stars far south, and by stars near the zenith, as three separate results ; 
and then I took the mean of these three for the zenith point. For an 
approximate co-latitude I used 37°. 47'. 6",83. 

At the beginning of March the telescope was moved about thirty 
degrees on the circle; at the beginning of August it was again moved 
thirty degrees, and on this occasion (as it appeared that the circle was 
not exactly balanced) a pound of lead was attached to the eye end of 
the telescope ; at the beginning of December it was again moved 
about thirty degrees. It does not appear however that the fact of the 
discordance has been affected, but its law seems to have been in some 
degree altered. 

A discordance of the same kind exists, I believe, in every circle 
that has been properly examined. I am informed by Mr Henderson 
(late Cape Astronomer) that he has found it in the Cape Circle. It 
was recognized as existing in the Greenwich Circles : and, though the 
system of observing there, which I have described, does not allow us 
to trace the unmixed faults of either circle, yet from the discordance 
in the places of stars as determined by the two circles, and its variation 


in different points of the meridian, I am inclined to think that the 
defect in one circle is different from that in the other. 

In vain have I endeavoured to discover the cause of this discordance. 
I once thought that it might be owing to the circumstance, that for 
the reflection-observation the circle is at rest for some minutes after 
the microscopes are read, and possibly it might (though clamped) have 
changed its position. A series of observations expressly made, showed, 
however, that there was no sensible change either in a few minutes 
or in many hours. I thought that the surface of the mercury might 
be sensibly curved, and that from a habit of observing in one part of 
the trough, an error might be produced. A set of experiments proved, 
however, that there was not the least sensible difference in the results 
found from observing at one or the other end of the trough. A flexure 
of the wire in the field of view would not explain it, as the discordance 
which that would produce is of the opposite kind. There appeared 
to be no reason for supposing an error in the determination of the 
coincidence of the micrometer wire with the fixed wire, in the value 
of the micrometer screw, or in the observation with the micrometer 
wire. The object glass, repeatedly examined by myself and once by 
Mr Simms, did not appear to be loose in its cell. I am driven at last 
to the supposition that the circle sensibly changes its figure ; but I 
have no proof of this, nor do I see distinctly how it should produce 
the discordance in question. Three sets of readings of every 10° under 
all the microscopes, have not assisted me to discover such change. 
My a priori opinion is, that a change in figure is hardly possible. The 
telescope, it must be remembered, is attached at its ends to the limb 
of the circle : the limb is in one piece (cast in several pieces and burnt 
together) ; and the whole arrangement of parts seems admirably adapted 
to prevent any change. If I had to fix on an astronomical instrument 
which appeared less likely to change than any other, I should certainly 
choose the Mural Circle. 

To discover experimentally the law of discordance, I proceeded 
as follows. The observations being reduced, and those of each star 
being digested under the heads of D, R, SP. D., and SP. R., I 


selected for the three first positions of the telescope all the un- 
exceptionable corresponding observations D and R. (The stormy 
weather of December made it impracticable to observe low stars by 
reflexion). In each case of a double observation, the difference of 
the results D and R would be double the difference between the 
zenith point as found from that star, and the zenith point adopted 
in the reductions. The mean of the differences of all the correspond- 
ing results D and R, would therefore be double the mean of all 
the differences between the zenith points found from the particular 
star, and the zenith points found from all by a tolerably uniform 
system : and thus it might be considered as double the difference 
between the zenith point found without error of observation from 
that star, and a certain imaginary well defined point. These values 
for all the stars, and for each position of the telescope, were arranged 
in tables (for which, as well as for some other numerical values, I 
must refer to the Cambridge Observations, Vol. VI.) 

The next step was, to connect these, approximately at least, by a 
law. I soon found that to attempt this by calculation was almost hope- 
less. Combinations of constants, sin Z.D., sin Z.D. cos^ Z.D., cos2Z.D., 
were tried in vain. I therefore adopted a graphical method similar 
to that used by Sir John Herschel, in the reduction of his sweeps, 
and described by him in the Phil. Trans. 1833. Taking the line of 
abscissae for zenith distances, and the ordinates to represent the mean 
of the differences above mentioned, I made a curve to pass among 
the points so determined, as well as I could, giving to each point 
an importance depending on the number of observations. From this 
curve I measured off" the ordinates for every 10° of zenith distances; 
half of this quantity I considered to be the correction to the ob- 
served zenith distance, to be applied with different signs to the 
direct and the reflected observation. The only respect in which 
theoretical consideration may be said to have assisted me is the 
following. Since the error in the relation between the position of the 
telescope and the reading of the circle, to which the discordance 
must be due, is periodical and never infinite, it may be expressed by 
sines and cosines of the Z. D. and its multiples. Now it is useless 


to take sines of even multiples, or cosines of odd multiples, because 
when 180° — Z.D. is substituted for Z. D., the result is equal in 
magnitude but opposite in sign ; and therefore when the two are 
added together, (as they are in finding the zenith point from each 
star), no trace of these terms would remain. Thus there may be 
sensible flexure in the circle which cannot be discovered from ob- 
servation by reflexion. The sines of odd multiples, and the cosines 
of even ones, (all which may be expressed in finite series of powers 
of sin Z.D.), will produce the same values with the same signs for 
180° — Z.D. as for Z. D., and these will affect the zenith point. 
Thus it appears that the terms which aff*ect the zenith point are 
the same for a direct observation and for the corresponding observation 
by reflexion, and it is this which justifies us in applying half the 
discordance to each. It appears also that when Z. D. = 90°, the 
function is maximum or minimum, and hence the curve in the 
graphical process above described must there be parallel to the line of 

The tables of corrections being thus formed, I now considered 
myself entitled to apply them to the reduced r^ults of all the 
observations, whether there were corresponding observations of the 
opposite kind or not. ' >»/ ^ i - 

The principal steps of the succeeding process may be gathered 
from the subjoined table. The first column contains the name of 
the star, its position with regard to the pole, (the lower transit being 
marked by S.P.), and the method of observing it (the letters D and 
R being always used for direct and reflected vision). Here it is to 
be observed that a star above the pole and the same star below the 
pole are reduced as separate stars, which is necessary, because the 
observations have been reduced with an assumed co-latitude, or an 
assumed place of the pole, the error in which assumption can be 
found only by comparing the separate results for the same star above 
and below. The second column contains the number of observations. 
The third contains its mean N.P.D. for Jan. 1, 1833, as found from 
the mean of all the results in each position and mode of observation. 
Vol. V. Part II. Nn 


and reduced with the assumed co-latitude 37° . 47' . 6,"83 : those de- 
termined from the lower transits of the star have the negative sign. 
For refraction, Bessel's tables are used. The fourth column contains 
the seconds only, as corrected for the errors above described ; this 
has been done by taking the number of observations in each position 
of the telescope on the circle, and finding the mean correction, 
supposing that to each observation the correction proper to that 
position was applied. The negative sign has still been retained for 
the lower observations. The fifth column contains the whole number 
of observations in each position of the star : and the sixth contains 
the mean N.P. D. for each position, as inferred from the combina- 
tion of direct and reflected observations. The seventh contains the 
whole number of observations for both positions. The eighth contains 
the algebraic sum of the two determinations of N.P.D., as the star 
is above or below the pole. If the assumed co-latitude were correct, 
this sum would = ; if the assumed co-latitude be increased by x, 
this sum would be increased by ^x, and therefore to make it now 
= 0, X must be taken = — i x sum in 8th column. The results, as 
might be expected, are however different for different stars, though 
the difference is much smaller than I could almost have hoped ; the 
extreme difference in the correction of latitude being 1,"3, and this 
being the difference between two results from stars nearly in the 
same parallel (shewing that it does not arise from error in the cor- 
rections above described), and which had been not much observed. 
It now becomes necessary to determine how the relative importance 
of these results shall be estimated. It would not be right to give 
a value proportionate to the number of observations, because part of 
the discordance may be produced by errors of division and other 
causes which, in the observations of a single star, produce constant 
errors. The ninth column contains the immbers by which (from my 
estimation of the comparative influence of constant and variable errors) 
I suppose the value of each result to be estimated. The tenth con- 
tains the product of the corresponding numbers in columns 8 and 9- 
The sum of the numbers in column 10 being divided by the sum of 
those in column 9, gives + 2",82 for the double correction, or + 1",41 


for the single correction, of the co-latitude ; and the co-latitude thus 
corrected is 37° . 47' . 8",24, or the latitude 52°. 12'. 51",76. This result 
I conceive to be correct within a small fraction of a second. The 
number of circumpolar stars used for this determination is 10, and 
the whole number of observations 917. 

In describing the process by which I have arrived at the above 
result, it has been my wish to present to the Society not only a 
determination possessing considerable local interest, but also an account 
of instrumental anomalies which are of general scientific importance. 
In further illustration of the latter point I will allude to the dis- 
cordances in the determinations of the obliquity of the ecliptic. It 
is well known that most astronomers have found the obliquity smaller 
from observations at the winter solstice than from those at the 
summer solstice. Now if I had used only the latitude found from 
direct observations of circumpolar stars, and had applied no correction 
to the observations of the Sun, I should also have found two values 
for the obliquity discordant by about 5", the winter obliquity being 
the smaller. With the corrections above described, (and which were 
formed entirely from observations of stars, and before I had even 
examined my sun observations) the two values of the obliquity 
agree within 1". I might have altered the corrections so as to re- 
move part of this discordance, but I prefer leaving them in 
the shape in which they were given by independent considerations. 
Indeed if I had confined myself to the January observations for the 
winter solstice, and omitted those of December when the correction is 
less certain, the discordance would wholly have disappeared. A very 
small alteration of the constant of refraction (such as would not alter 
the latitude much more than 0",1), or a very small alteration in the 
law of refraction (which would not be sensible in the latitude) would 
remove this difference. But I hardly venture to assume that obser- 
vations of the Sun, near the winter solstice, can be relied on to this 
degree of accuracy. 

I will only add, in conclusion, that I believe the method which 
I have used to be the only one of those in practice from which a 


good result can be obtained. Had I determined my zenith points 
by a floating collimator, the result of observations on Polaris and 
^ U. Minoris would have given the latitude more than a second 
wrong, and the polar distance of every southern body more than 
two seconds wrong : the result of observations on the Sun would have 
given nearly the same error in the latitude but with the opposite 
sign. If a circle reversible round a vertical axis had been used 
(as at Dublin, Palermo, &;c.) its errors would (supposing the mere 
circle exactly as good,) have been just as great as if a collimator 
were employed. The method adopted above appears most valuable, 
not only because it gives numerical conclusions more accurate than 
any other, but also because it enables us to observe discordances and 
to suspect faults which, though they confused our results, might 
otherwise have wholly eluded our discovery. 



March 23, 1834. 

Tratnsajclions of theCsanb.rhil.Soc-.Vol VTl. 7. 

W MOicalfi^. UefM^^CamJirui^e-. 

jTvnja-ciimus ofOu. Camji. f/vu. Jodefy, Vol V Tt S. 

Ft^. I 

jri^ Z. 

J^r^. 3. 

Jkfetcalfl, Ucho^'^ Cojnbnd^e'. 


Table exhibiting the Calculations j^r correcting the Latitude o/" Cambridge 
Obseuvatory ; the Observations having been reduced with the assumed 
Latitude 52M2' . 53",17. 

star's Name. 




Polaris D 

Polaris R 

Polaris S.P D 

Polaris S.P R 

8 Urste Minoris D 

S Ursa; Minoris R 

S Urs» Minoris S.P. ..D 
2 Ursae Minoris S.P. ..R 

/3 Ursae Minoris D 

/3 Ursae Minoris R 

/3 Ursifi Minoris S.P. ..D 
/3 Ursae Minoris S.P. ..R 

/3Cephei D 

/3 Cephei R 

ySCephei S.P D 

/3 Cephei S.P R 

2 Draconis D 

I Draconis R 

S Draconis S.P D 

S Draconis S.P R 

a Draconis D 

a Draconis R 

a Draconis S.P D 

a Draconis S.P R 

a Ursae Majoris D 

a Ursae Majoris R 

a Ursae Majoris S.P. ..D 
a Ursae Majoris S.P. ..R 

Uncorrected mean 




« Cephei D 

a Cephei R 

a Cephei S.P D 

a Cephei S.P R 

S Ursae Majoris D 

S Ursae Majoris R 

Ursae Majoris S.P. ..D 
g Ursae Majoris S.P. ..R 

a Cassiopeiae D 

a Cassiopeiae R 

o Cassiopeiae S.P D 

o Cassiopeiae S.P R 
































1 . 34 . 52,22 

1 . 34 . 53,77 

- 3 

24 . 46,34 

15. 9-41,65 


■15. 9.44,58 


20 . 10 . 15,91 

- 20 . 10 . 16,97 


22 . 37 . 54,40 

- 22 . 37 . 54,99 


24 . 49 . 24,80 

- 24 . 49 . 26,90 


27 . 20 . 55,75 

- 27 • 20 . 59,30 




7 . 12,67 

32. 2.18,12 


■32. 2.20,38 


34 . 22 . 45,38 


- 34 . 22 . 47,37 






- 54,62 

- 54,83 


- 47,40 

- 47,28 


- 45,52 


- 17,74 

- 16,52 


- 56,14 




- 27,90 







• 12,93 




- 22,45 


- 47,93 

- 46,54 












- 54,70 


- 47,35 


- 45,57 


- 17,29 


- 55,83 

• 28,04 

■ 59,86 


No. Algebraic 

of Sum of 

Obs. Determin. 








- 47,57 













- 1,55 





69 - 4,12 





- 16,05 










Vol. V. Part II. 





Vol. V. Part III. 








XII. On the Diffraction of an Ohject-glass with Circular Aperture. By 
George Biddell Airy, A.M. late Fellow of Trinity College, 
and Plumian Professor of Astronomy and Experimental Philosophy 
in the University of Cambridge. 

[Read Nov. 24, 1834.] 

The investigation of the form and brightness of the rings or rays 
surrounding the image of a star as seen in a good telescope, when a 
diaphragm bounded by a reetihnear contour is placed upon the object- 
glass, though sometimes tedious is never difficult. The expressions 
which it is necessary to integrate are always sines and cosines of mul- 
tiples of the independent variable, and the only trouble consists in 
taking properly the limits of integration. Several cases of this problem 
have been completely worked out, and the result, in every instance, 
has been entirely in accordance with observation. These experiments, 
I need scarcely remark, have seldom been made except by those whose 
immediate object was to illustrate the undulatory theory of light. 
There is however a case of a somewhat different kind; which in 
practice recurs perpetually, and which in theory requires for its com- 
plete investigation the value of a more difficult integral ; I mean the 
usual case of an object-glass with a circular aperture. The desire of 
submitting to mathematical investigation every optical phaenomenon of 
frequent occurrence has induced me to procure the computation of the 
numerical values of the integral that presents itself in this inquiry : 
and I now beg leave to lay before the Society tlie calculated table, 
with a few remarks upon its application. 

Let a be the radius of the aperture of the object-glass, f the focal 
length, h the lateral distance of a point (in the plane which is normal 
Vol. V. Part III. Pp 


to the axis of the telescope) from the focus. Then, the lens being 
supposed aplanatic, and a plane wave of light being supposed incident, 
the immediate effect of the lens is to give to this wave a spherical 
shape, its centre being the focus of the lens. Every small portion of 
the wave, as limited by the form of the object-glass, must now be 
supposed to be the origin of a little wave, whose intensity is propor- 
tional to the surface of that small portion ; and the phases of all these 
little waves, at the time of leaving the spherical surface above alluded 
to, must be the same. If then Sx x Sy be the area of a very small part 
of the object-glass, q the distance of that part from the point defined 
by the distance b, the displacement of the ether at that point, caused 
by this small wave, will be represented by 

Sx X. Sy X sin—- {vt — q — A) ; 


and the whole displacement caused by the small waves coming from 
every part of the spherical wave will be the integral of 

sin — (vt—q — A) 

through the whole surface of the object-glass, q being expressed in 
terms of the co-ordinates of any point of the spherical surface. 

Now let X be measured from the center of the lens in a direction 
parallel to i; y perpendicular to x and also perpendicular to the axis 
of the telescope; and % from the focus parallel to the axis of the 
telescope. Then 

q=.^{{x- by + y- + x} = -y/ix' +f+x'-2bx) 

omitting squares and superior powers of b. But x^ + y^ + z' —f^^ 
since the wave is part of a sphere whose centre is the focus ; therefore, 

q = VW^-^bx)=f-j,x nearly; 
and the quantity to be integrated is 

sm— \vt - f - A + -x). 
^ J 


The first integration with regard to y is simple, as y does not 
enter into the expression, which is therefore to be considered as con- 
stant. Putting y, and y^ for the smallest and greatest values of y 
corresponding to x, the first integral is 


To this point of the investigation the expressions are general, including 
every form of contour of the object-glass. 

We must now substitute the values of y^ and y^ in terms of x, 
before integrating with regard to x. For a circular aperture 

y, - y. — ^y/a^-x" 

where the sign of the radical is essentially positive. Hence the dis- 
placement of the ether at the point defined .by the distance A is re- 
presented by 

2 f, Va' - x" . sin — {vt-f- A + ^x) 
= 2sm -^{vt-/— A) f^\/a^-af .cos-— .^x 

\ Ay 

+ 2cos — - {vt —f— A) X a/«^ — x\sm—-.^x, 

A ^ J 

and the limits of integration are from x = — a to x = + a. Between 
these limits it is evident that 

;- /-: « . 2-ir b 

f^Va' — x^ .sm—- . ^x = 0, 

^ J 

(as every positive value is destroyed by an equal negative value) ; and 
the displacement is therefore represented by 

2sin— -(«^— /— ^) ji\/«^ — ar'.cos ^ .^x, 
^ ■ , ^ ./ 

the integral being taken between the limits x= -a, x— -^a. 

p p2 


If we make - = w, — — . -2r = n, the expression becomes 
2a^.sm-—{vt-f~A)J^V^-uf'Cosnw, fromw=— 1 tow=+\, 


or 4«^ sin — - {vt-f~ A) j„\/l — tt;''. cos nw, irom w = to w = l. 


It does not appear, so far as I am aware, that the value of this 
integral can be exhibited in a finite form either for general or for 
particular values of w. The definite integral 

J„^/\ — vf . cos nw (from w=-0 to w = \,) 

(which will be a function of 7i only) being expressed by N, it may be 
shewn that N satisfies the linear differential equation 

n ' dn dv? ' 

which may be depressed to an equation of the first order that does 
not appear to yield to any known methods of solution. 

If we solve the equation by assuming a series proceeding by powers 
of n, or if we expand cos nw and integrate each term separately, we 
arrive (by either method) at this expression for the integral 

TT . rf_ n^ _ _ "" Xr \ 

4 "" ^ 2.4"^2.4^6 ^:^\Q'.S^^^-' 

The table appended to this paper contains the values of the series 
in the bracket, for every 0,2 from w=0 to w = 12. Each value has 
been calculated separately, the logarithms used in the calculation have 
been systematically checked, and the whole process has been carefully 
examined. The calculations were carried to one place further than the 
numbers here exhibited. I believe that they will seldom be found in 
error more than a unit of the last place; except perhaps in some of 
the last values, where the rapid divergence of the series for the first 
five or six terms made it difficult to calculate them accurately by 


In the use of tins table n must be taken = — -.-^. If instead of 

using the linear distance h to define the point of the field at which 
we wish to ascertain the illumination, we use the number of seconds *, 

then A = /. *.sin 1", and n must be taken = — as sin 1". If \ be taken 

for mean rays = 0,000022 inch, n must be taken = 1,3846 x as, a being 
expressed in inches. From this expression, and from the numbers of 
the table, we draw the following inferences. 

1. The image of a star will not be a point but a bright circle 
surrounded by a series of bright rings. The angular diameters of these 
(or the value of s corresponding to a given value of n) will depend 
on nothing but the aperture of the telescope, and will be inversely as 
the aperture. 

2. The intensity of the light being expressed (on the principles 
of the undulatory theory) by the square of the coefficient of 

sin-^ivt-f- A), 

and the intensity at the center of the circle being taken as the standard, 
it appears that the central spot has lost half its light when « = l,6l6, 

I 17 
or s = — — ; that there is total privation of light, or a black ring, when 

2 76 
n = 3,832, or * = — — ; that the brightest part of the first bright ring 

Q WQ -I 

corresponds to w = 5,12, or * = — — , and that its intensity is about — of 

a Oi 

5 16 
that at the center; that there is a black ring when n = 7,14, or s= -- — ; 


that the brightest part of the second bright ring corresponds to ra = 8,43, 

or * = — — , and that its intensity is about — r of that of the center ; 

7 32 
that there is a black ring when w =10,17, or *= — — ; that the brightest 


part of the third bright ring corresponds to w = 11,63, or *= — — , and 
that its intensity is about ^— - of that of the center. 

The rapid decrease of light in the successive rings will sufficiently 
explain the visibility of two or three rings with a very bright star 
and the non-visibility of rings with a faint star. The difference of 
the diameters of the central spots (or spurious disks) of different stars 
(which has presented a difficulty to writers on Optics) is also fully 
explained. Thus the radius of the spurious disk of a faint star, where 
light of less than half the intensity of the central light makes no 

1 17 

impression on the eye, is determined by making /* = 1,616, or s=— — : 

whereas the radius of the spurious disk of a bright star, where light 
of — the intensity of the central light is sensible, is determined by 

1 97 
making n = 2,73, or * = — — . 

The general agreement of these results with observation is very 
satisfactory. It is not easy to obtain measures of the rings; since 
when a is made small enough to render them very distinct as to form 
and separation, the intensity of their light (which varies as a^) is so 
feeble that they will not bear sufficient illumination for the use of 
a micrometer. Fraunhofer however obtained measures agreeing pretty 
well (as to proportion of diameters, &c.) with the results above. 

For verification of the numbers it would probably be best to use 
an elliptic aperture. By an investigation of exactly the same kind as 
that above it will be found that the rings will then be ellipses exactly 
similar to the ellipse of the aperture, but in a transverse position ; that 
the major axes of the rings for the elliptic aperture will be the same 
as the diameters of the rings for a circular aperture whose diameter 
— minor axis of ellipse of aperture, but that the intensity will be 
greater in the proportion of the squares of the axes. I have not yet 
had an opportunity of examining this in practice. 


I shall now apply the numbers of the table to the solution of 
the following problem. To find the diameters, &c. of the rings when 
a circular patch, whose diameter is half the diameter of the object- 
glass, is applied to its center, so as to leave an annular aperture. 

The radius of the patch being -, it is easily seen that the dis- 
placement (using the same notation) is 

2sm-—-(vt—J'—A)fr\/a^-x'.cos—-.^a; (from a;—-a to x=+a) 
- 2sin ~(vt-f-A)J\/---af. cos-^ .^x (from x= -- to x= +^. 
Putting - =w, — = u, this becomes 

4a^ . sin -T-{vt —f — A) /„ \/l — vf . cos — .—^w 

A A / 

-4.^.sm yC^^-/- ^)/«vl-M'.cos— .— .M, 

the limits of integration both for w and for u being and 1. Omitting 
the factor oV, the intensity will be expressed by 


where (p{n) is the number given in the table. 

Upon forming the numerical values we find that the black rings 
correspond to values of w=3,15, 7,18, 10,97: and that the intensities 

of the bright rings (in terms of the intensity of the center) are — , — . 

Thus the magnitade of the central spot is diminished, and the bright- 
ness of the rings increased, by covering the central part of the object- 

In like manner, if the diameter of the circular patch = a ( 1 — />), the 
intensity of light would be proportional to {<p {n) — {l— pf .^{n—pn)}". 


The quantity under the bracket, if p is very small, is equal to 

X) ft 

2p .<p {fi) + pn<p' (n) = - . -J— {n^<p{n)}. 

In the case of a very narrow annulus therefore the diameters of the 
black rings will be determined by making ?i^(p (») maximum or 
minimum. It appears then that there ought to be only one black 
ring corresponding to each black ring with the full aperture, and that 
its diameter ought to be somewhat smaller. This conclusion does not 
agree with the experiments recorded by Sir J. Herschel, in the Encyc. 
Metrop. Article Light, page 488 : but it is acknowledged there that 
the results are discordant with Fraunhofer's : and I am inclined there- 
fore to attribvite the phasnomena observed by Sir J. Herschel to some 
other cause. 

The investigation of cases of diffraction similar to that discussed 
here appears to me a matter of great interest to those who are 
occupied with the examination of theories of light. The assumption 
of transversal vibrations is not necessary here as for the explanation 
of the phasnomena of polarization : and they therefore offer no argu- 
ments for the support of that principle. But they require absolutely 
the supposition of almost unlimited divergence of the waves coming 
not merely from a small aperture, but also from every point of a large 
wave : and the results to which they lead us, shew strikingly how 
small foundation there was for the original objection to the undulatory 
theory of light, viz. that if waves spread equally in all directions. 

there could be no such thing as darkness. 

Obsbrvatory, Cambridge, 

November 20, 1834. 

G. B. AIRY. 



4j r • 

Table of the values of 0(w) = — f^vl — nw from w = iow=\. 






+ 1,0000 


- 0,0922 


+ ,9950 


- ,0751 


+ ,9801 


- ,0568 


+ ,9557 


- ,0379 


+ ,9221 


- ,0192 


+ ,8801 


- ,0013 


+ ,8305 


+ ,0151 


+ ,7742 


+ ,0296 


+ ,7124 


+ ,0419 


+ ,6461 


+ ,0516 


+ ,5767 


+ ,0587 


+ ,6054 


+ ,0629 


+ ,4335 


+ ,0645 


+ ,3622 


+ ,0634 


+ ,2927 


+ ,0600 


+ ,2261 


+ ,0545 


+ ,1633 


+ ,0473 


+ ,1054 


+ ,0387 


+ ,0530 


+ ,0291 


+ ,0067 


+ ,0190 


- ,0330 


+ ,0087 


- ,0660 


- ,0013 


- ,0922 


- ,0107 


- ,1116 


- ,0191 


- ,1244 


- ,0263 


- ,1310 


- ,0321 


- ,1320 


- ,0364 


- ,1279 


- ,0390 


- ,1194 


- ,0400 


- ,1073 


- ,0394 


- ,0922 


- ,0372 

Vol. V. Paet III. 


» J » 0' { 


• JI.JL.A 4 4- ^li 

XIII. On the Equilibrium of the Arch. By the Rev. Henry Moseley, 
B.A. of St John's College; Professor of Natural Philosophy and 
Astronomy in King's College, London. 

[Read Dec. 9, 1833.] 

1. Let a mass acted upon by forces applied to any number of 
points in it be imagined to be intersected by an infinite number of planes, 
dividing it into exceedingly small laminse. Suppose the direction of the 
resultant of the forces acting upon one of these, having for its ex- 
ternal face a portion of the surface of the body, to be determined. 
Combining this force with those acting upon the different points of 
the next, contiguous lamina; let their common resultant be ascertained. 
Proceed similarly with the next, and with each succeeding lamina. 

These lines will then be the tangents to a curved line, called in 
the following paper the line of pressure, whose intersection with each 
lamina, marks the point where a single force might be applied so as 
to produce the same effect with all those impressed upon that lamina, 
this single force being impressed in the direction of a tangent to the 

If any of these imaginary intersecting planes be supposed to become 
real sections of the mass, so as to separate it into distinct parts, the 
conditions necessary that no one of these parts may slip or turn over 
on those contiguous to it, will manifestly be determined by the direc- 
tion of the line of pressure in reference to the plane of the section. 

In general it will be observed that forces applied to a system of 
variable form are, when in equilibrium, subject to the same conditions 
as though its form were invariable, together with certain other conditions, 
dependant upon the nature of the variation to which the form of the 
system is liable. In other words the conditions of the equilibrium of 
a system of invariable form are necessary to the equilibrium of a system 
of variable form ; but they are not sufficient. We shall first determine 



the form and position of the line of pressure on the hypothesis, that 
the form of the system is invariable, and then consider the modifica- 
tion to which these are subjected by the opposite hypothesis. 

2. Let there be conceived a mass, the connexion between the parts 

of which may be any whatever, and the nature of whose surface is 

determined by the equation 

^ xy% = 0. 

Let it be intersected by an imaginary plane whose position in reference 
to a given system of rectangular co-ordinates is determined by the arbi- 
trary constants A, B, C, and whose equation is 

z = Ax + By + C (1). 

Let Ml, Mi, Ms represent the sums of the forces acting upon one 
of the parts into which the mass is divided by the intersecting plane, 
resolved in directions parallel to the axes of x, y, z, respectively. Also 
let JVx, N'z, ^3 be the moments of these forces about the same axes. 
Then Mi, Mi, Ms; Ni, N^, iVs are given in terms of the arbitrary 
constants A, B, C — of the given forces — and of the constants involved 
in the given equation to the surface of the mass. 

Let the position of the intersecting plane be supposed to be such, 
that the forces acting upon the above mentioned portion of the mass 
may have a single resultant, an hypothesis which involves the known 

MiN, + M,N, + M,N, = (2). 

The equations to the resultant in any given position of the inter- 
secting plane, are 

Ml ^Ni 


Let the arbitrary constant C be eliminated from this equation, and 
from the equation to the intersecting plane by means of equation (2) ; 
and let the plane be then supposed to take up a series of positions, 
the law of which is fixed by its equation, and of which, each is im- 
mediately adjacent to the former. 



Further, let it be supposed that the resultant of the forces upon 
the portion of the mass, cut off by the plane, in each of its positions, 
intersects with the resultant similarly taken in its immediately previovis 
position — an hypothesis which introduces a new condition into the 
question and establishes a second relation between the quantities 
M„ M„ M,; A, B, C. 

That relation is determined as follows. 

Since x, y, as are to be considered as the co-ordinates of a point of 
intersection of two consecutive resultants; we may differentiate the 
equations (3) with respect to the arbitrary constants A and B, consi- 
dering X and y as constant. From this differentiation, the following 
equations are obtained: 

= ss 

= » 



■ dAv 









whence, eliminating z 




dA + r^ft— aJ? I 





dA + 







= 0...(5). 

This last equation determines the relation between A and B ne- 
cessary to the continual intersection of the consecutive resultants ; and 
the elimination of these quantities between equations (3) and (4), 
produces two equations in x, y, % which are those to the locus of 
that intersection. That is, they are the equations to the line of 



3. By the elimination of A,y B and C between the equations (2), (3) 
and (5), a relation is obtained between the co-ordinates of a point in 
the direction of the resultant force, applicable to every position of the 
intersecting plane. Being in fact, the equation to that developable 
surface which is the locus of the resultants, and, which has for its 
edge of regression, the line of pressure. This surface will be properly 
called the surface of pressure. 

It is evident that at that point where the line of pressure even- 
tually cuts the surface of the mass, there must be applied a force equal 
to the resultant of all the other forces impressed upon the system 
and in the direction of a tangent to the line of pressure at that point, 
or there must be applied to the surface of the last lamina cut off 
by the intersecting plane, forces whose resultant is of that magnitude 
and in that direction. 

4. These conditions may be expressed as follows. 

Let P' be the force — or the resultant of the forces — applied to the 
last lamina, x^, y,, ss, the co-ordinates of the intersection of the line of 
pressure with it, a, fi, y the inclinations of P' to the axps of x, y, %. 
^ Also let 

be the equations to the line of pressure. 

Since the point Xx, y^, %i is a point in the surface of the mass, 

.-. ^Xiy.z, = 0. 

Also, since it is a point in the line of pressure, 

.-. Xi = Fi%i 
y, = F^%x 

Since the direction of P' is that of a tangent to the line of 

tan a = 

tan /3 = 

d%, ' 



p = VM^VMr+~m, 

where M^, M.^, M-j are supposed to be taken throughout the tvhole 

Thus there are six equations of condition, which together with the 

cos^ a + cos^ /3 + cos^ 7 = 1. 
determine the seven quantities P', Xi, y^, %i\ a, /3, 7 in terms of the 
forces (other than P') which compose the system, and the constants 
which enter its equation. These fix the relations necessary to the equi- 
librium of the mass considered as one continued geometrical solid. 

Before proceeding to the discussion of the additional conditions 
requisite to the equilibrium when the mass passes from the invariable 
form here supposed, to a variable form, it will be well to give an 
example of the application of the principles which have been already 
laid down to the actual determination of the line of pressure in a par- 
ticular instance. 

5. Let then ABCD (fig. 1.) represent a heavy mass, bounded at its 
extremities by parallel planes AB and CD, and laterally, by the planes 
AC and BD inclined at any angle to one another. 

Let the mass be imagined to be intersected by an infinite number 
of planes parallel to AB, of which one is mn, and to be supported 
by forces acting at p and p' at angles cp and <f>' with the horizon. 

It is required under these circumstances to determine the form and 
position of the line of pressure. 

Let the line P'G bisect AB and CD. Draw P'E horizontal and 
PM vertical. 

Let P'M= A, CD = 2b, P'p = k, 
AB = 2a, P'G = h, Gp' = k'. 
Inclination of P'G to the horizon =7, - 
AB = /3. 



BP' - Pm _ BP - DG 
PP' ~ GP' ' 

2a — (mn) _ 2 (a — i) _ 
' ■ ^sec7 ~" ^ ' 

/ V « 2 (« - A) . 

.'. (mn) = 2a —r — - secy . A; 

.-. area (BAnm) = ^ {{AB) + {mn)\ . (P'P) . sin (PP'A) 

= sin {(3 +y) secy {2a A- ^-^ .secy .A"}; 

d\aYea(BAnm)} - /o , \ ia 2(a — b) 
.'. -^ V^ ^ = 8111 (/3 + 7) sec 7 {2a ^ , ^ sec7 . A\. 

- Now each element of the area has its centre of gravity in P'G ; 
,-, moment of area = 2^sin (/3 + 7) sec 7 ^{a^ ^— sec7^'} 

=^sin(/3 + 7)sec7{«^* ^"7 ^ sec7^H. 

iVs = moment of p + moment of area (BAmn) 

2 (a — b) 
= PffK sin (0 + /3) + ^sin (/3 + 7) sec7 {aA^ ^ , ' sec 7 A""). 

Mx — pg cos (p, Mz = 0, 

M^ =pgsm(j> - g sin (/3 + 7) sec 7 {2a^ ^ sec 7 ^^}, 

iv, = 0, iV3 = o. 

Calling therefore x and » the co-ordinates of any point in the re- 
sultant of the forces applied to the area (ABmn), we have for the 
equation to that resultant. 


or, zpgcoscp - pgsivKp + xgsm{(i + 7)sec7{2a^ -r—secy.A^} 

=pgK sin ((p + /3) +gsm{fi + 7)8607 {aA^ ^-— r — ^sec7 . A^}. 

Differentiating which equation with regard to the arbitrar}' constant 
A, we obtain 

A = x, 

whence by elimination and reduction, 

^^l( ^- ^ ] ({f3 + y) \ ^ 

^ \ pk J \ COS(p j ' 

_ /_a\ [ sec 7. sin (^ + 7) ] ^ 
\pj \ cos(p /■ 

+ tan (f) . X 

_^ sin((/> + /3) 


The above is the equation to the line of pressure. It indicates a 
point of contranj flexure corresponding to 


X =■ 7 cos <h. 

a — h ^ 

The curve is concave to the axis of x, between the origin and this 
point. It is afterwards continually convex. 

A minimum value of % is determined by the equation 

\Jta\ f / ip\^ sin>cos^7 ) 

^ \a-AV^ ^ ^ Ui •sinMi8 + 7)r * 

It will be observed that since all the forces applied to the system 
may be supposed to act in the same plane, the two conditions. 

First, " That in every position of the intersecting plane, the forces 
shall have a single resultant," and Secondly, "That the consecutive re- 
sultants shall intersect," are necessarily satisfied. 
Vol. V. Paet III. Re 


To simplify the question, let the planes AC, BD which bound the 
mass laterally be supposed to be parallel, the figure ABCD assuming 
the form of a rectangle. Fig. 6. 

This hypothesis will introduce the following conditions : 

« = *' ^ = J - 7- 

Hence, by substitution the equation to the line of pressure becomes 

% — . sec 7 . sec . a;^ 

+ tan ^ .X 

cos (7 - 0) 
Avhich may be put under the form 

, p . ^ 1 , P , ( cos (y — (j)) P sin^ cos 7 -, 

^x — -~ sm cos^r = - cos 7 .cos . \k ' + -7 ^-— ; — - — *|. 

* 2a ^ "a ' ^ ^ COS0 4« cos ' 

It is manifest therefore, that the line of pressure is in this case 

a parabola — having its axis vertical and at a distance = —- sin cos 7 

^ a 

from the origin — having its concavity downwards — its vertex at a height 

_ cos (7 — 0) p sin' cos 7 
~ cos 4 « cos ' 

above the axis of x — and having for its parameter the quantity 
• (^j . cos cos 7. 

Let us now seek to determine what relation must exist between 
the forces impressed upon the mass which we have hitherto considered 
of invariable form, that the equilibrium, may continue under the same 
circumstances when its form and dimensions are made to admit of 
variation. And let us suppose 


First. That certain of the sections, which we have imagined, be- 
come real sections of the mass, dividing it into separate and distinct 
parts, each of which retains the properties of a perfect solid. 

Secondly. Let us suppose every point in the system to admit of 
displacement, subject, within certain limits, to the law of perfect 

The determination of the conditions of the equilibrium in these 
two cases, will constitute a complete theory of construction. 

The discussion contained in the remainder of this paper will be 
confined to the first case. 

6. Let the mass AB (fig. 2.) have for its line of pressure the 
line PP'. Now it is clear, that if this line cut the plane QQ of any 
section of the mass in a point n' without the surface of the mass ; 
the tendency of the opposite resultants of the forces acting upon the 
two parts AQQ and SQQ', into which that section divides the mass, 
will be to cause them to revolve about the nearest point Q' of its 
intersection with the surface of the mass. And, this tendency being 
wholly unopposed, motion will ensue. And so in the mass represented 
(fig. 6.) the force p and with it the line of pressure pp' being given, 
it appears that, being cut transversely as shewn in the figure, the mass 
cannot be supported by any single force p if it extend beyond CD': 
any such force must, to produce equilibrium, be applied at q; and 
being applied there, the portion C'C"Z)"'iy will be wholly unsupported. 
The line of pressure being continued cuts the planes of the sections 
CD', CD', &c., without the surface of the mass. 

Thus then it is a condition of the equilibrium, that the line of 
pressure should intersect the plane of every section of the body within 
its mass. 

This condition will be satisfied if this line nowhere cut the surface 
of the mass except at the points P and P. Fig. 2. Or if the equation 

■^F^z, F^z, ■ « = 0, 



found by eliminating the values of a; and y between the equation to 
the surface and the equation to the line of pressure, involve only 
such possible values of z as correspond to the points P and i*, where 
the intersecting plane touches the surface, or to points where the line 
of pressure touches it. 

It is a further condition of the equilibrium that the line of pressure 
should not cut any section of the mass, at an angle with the perpen- 
dicular to that section greater than a certain given angle, dependant 
upon the friction of the surfaces in contact, and having for its tangent 
the coefficient of friction. 

The resistance of surfaces is not exerted exclusively in the direction 
of the normal, according to an hypothesis, which was probably in- 
troduced into the theory of Statics in order to simplify the investi- 
gations of those who originated that science, but which there seems 
no reason for retaining any longer. It is exerted in an infinity of 
different directions included within a certain angle to the normal, or 
rather within the surface of a certain right cone, having the normal 
for its axis and the point of resistance for its vertex. Any force, 
however great, applied within this conical surface will be sustained 
by the resistance of the surface of the mass — and no force however 
small, without it. 

Let R represent a single force on the resultant of any number of 
forces applied to a fixed surface, and let R' and R" be the resolved 
parts of R in the directions perpendicular and parallel to the surface. 
Also let p be the inclination of R to the vertical, and f the coefficient 
of friction. The friction of the surfaces in contact is therefore repre- 
sented by fR, and motion will, or will not, ensue according as R" is 


greater or is not greater than /R'. Or, according as -p, is greater or 

is not greater than f. Or, if y = tan (p, according as tan p is, or is 
not, greater than tan (p, or as p is greater or is not greater than (p. 


In the remainder of this paper the angle 0, or tan-'^ will be called 
the limiting angle of resistance*. 

From the above then it appears, that unless the tangent to the 
line of pressure at the point where it cuts any section of the mass, 
make with the perpendicular to the plane of that section an angle, 
which is not greater than the limiting angle of resistance, the surfaces 
there in contact will slip upon one another. 

This condition may be expressed analytically as follows : 

% = Ax + By + C 
is the equation to the plane of any section of the mass, therefore 

x-x^ = - Ai^-z), y-y,= -B {x-z), 

are the equations to the perpendicular to that section. And the angles 
which that perpendicular makes with the co-ordinate axes have for 
their cosines 

-A -B -1 

VA^TW+\' VA' + B'+l' VA' + B' + l' 

Also it appears from the given equations (3) to the resultant 
force, or tangent to the line of pressure, that this line makes angles 
with the co-ordinate axes which have for their cosines the quantities 

M, M, Ms 

Hence, therefore if / be the inclination of these lines to one 

* It is here supposed that the coefficient of friction f is constant for the saifie surfaces, 
whatever be the force B! by which they are pressed together. This is usually assumed 
to be the law of friction. It is only however an approximation to that law. The ex- 
periments of Mr Rennie shew that f must be considered a function of R' increasing con- 
tinually, but very slowly, up to the limits of abrasion. 



AM, + BM, + Ms 

cos /= — 

{{A' + B' + 1) {Mr' + Mi + Mi)}k ' 
in which expression M^, M^, M^, and B, are known functions of A. 

Now / must not exceed the limiting angle of resistance. Therefore 
cos / must not be less than the cosine of that angle. 

On the whole then we have these two conditions necessary to the 
equilibrium of a mass intersected by a series of planes, under the cir- 
cumstances supposed. 

1. That the equation 

-VF,%, F^z, » = 0, 

shall involve no possible roots, except such as correspond to the ex- 
tremities of the line of pressure, or to points where it touches the 
surface of the mass. 

2. That the fraction 

AM, + BM, + Ms 

shall for all values of A, corresponding to real sections of the mass, 
be not less than the cosine of that arc, whose tangent is the coefficient 
of friction. 

The first of these conditions being satisfied, the parts of the mass 
cannot turn upon one another. The second being satisfied, they can- 
not slip upon one another. 

We have supposed the whole of the forces impressed upon the 
system to be known excepting the force P', which has been deter- 
mined in terms of the rest. The force P' may be supplied by the 
resistance of a point in a fixed surface, in which case the amount and 
direction of that resistance will be known. 


If, however, there enter two or more resistances of surfaces among 
the forces which compose the equilibrium, since the magnitudes of 
these and also their directions may be any whatever, within the limits 
imposed by the friction of the surfaces; the problem remains, in so 
far as the known conditions of equilibrium are concerned, indeterminate, 
and recourse must be had for its solution to other principles. 

7. Suppose the mass AJS to be acted upon by any number of forces 
among which is the force P being the resultant of certain resistances, 
supplied by different points in a surface Sb, common to the inter- 
sected mass and to an immoveable obstacle SC. 

Now it is clear that under these circumstances we may vary the 
force P', both as to its amount, direction, and point of application, 
without disturbing the equilibrium, provided only the form and 
direction of the line of pressure continue to satisfy the conditions im- 
posed by the equilibrium of the system. 

These are manifestly, that it no where cut the surface of the mass, 
except at P" and within the space JSb, and that it no where cut a 
section of the mass or the common surface of the mass and obstacle, 
at any angle with the perpendicular greater than the limiting angle 
of resistance. " 

Thus, varying the force P', we may destroy the equilibrium, either, 
first, by causing the line of pressure to take a direction without the 
limits prescribed by the resistance of the section through which it 
passes ; or, secondly, by causing the point P to fall without the surface 
Bb, in which case no resistance can be opposed to the resultant force 
acting in that point ; or, thirdly, the point P lying within the surface 
Bb, we may destroy the equilibrium by causing the line of pressure 
to cut the surface of the mass somewhere between that point and P'. 

Let us suppose the limits of the variation of P' within which the 
first two conditions are satisfied, to be known ; and varying it, within 
those limits, let us consider what may be its least and greatest values 


so as to satisfy the third condition ; and where, and in what direction 
they must be applied. 

In the first place it will be observed, that by diminishing the force 
P', its direction and point of application remaining the same, the line 
of pressure is made continually to assume more nearly that direction 
which it would have, if P' were entirely removed. 

Provided then, that if P were thus removed, the line of pressure 
would cut the surface, that is, provided the force P' be necessary to 
the equilibrium ; it follows that by diminishing it, we may vary the 
direction and curvature of the line of pressure until we at length make 
it touch some point or other in the surface of the mass. 

And this is the limit; for if the diminution be carried further, it 
will cut the surface, and the equilibrium will be destroyed. It ap- 
pears then that under the circumstances supposed, when P' acting at 
a given point and in a given direction, is the least possible, the line 
of pressure touches the surface of the mass. 

In the same manner it may be shewn, that when it is the greatest 
possible, the line of pressure touches the surface of the mass. 

Now by varying the direction and point of application of P', as 
well as its amount, this contact may be made to take place in infinite 
variety of different points, and each such variety supplies a new value 
of P', producing the required contact. Among these, therefore, it 
remains to seek the absolute maximum and minimum values of that 

To express these conditions analytically, let Xi, y^, z.^ represent the 
co-ordinates of a point where the line of pressure touches the surface 
of the body. 

Since the point x^, y^, & is common to the line of pressure and 
to the surface of the body, 

.-. -^Xty^Xi = 0, Xi = F%o, y., = F^x^. 


Also, since it touches the surface in the point ar^ysSSj; 

dz-i i d-^x-iyiZj ^^ 

\ dXi J 

( d'^x^.yi%i \ 
dF,%., V dz, ) ^ ^ 
d%2 l d'^X2yi%i \ 

\ dyi I 

Eliminating x.^, y^, z.^ among these Jive equations two relations are 
established between the force P'*, the co-ordinates of its point of ap- 
plication, and the angles which fix its direction (see Art. 4) ; by elimi- 
nation between which a further relation is established between six of 
these seven quantities, and, finally, by the equations of condition 

COS^ a + COS^ /3 + COS* 7=1. 

a relation is obtained between four of them. 

Thus then we may obtain the value of P' in terms of three of 
the quantities x^, y^, s, ; a, /3, 7. 

Its maximum and minimum values are then at once determined by 
the known conditions of the maxima and minima of functions of 
three variables. 

8. It is evident that the minimum value of P', being that which 
just counteracts the tendency of the mass to revolve about the point 
where the line of pressure touches its surface, is also precisely that 
force which would be exerted there by another equal and similar mass, 
acted upon by equal forces, under the same circumstances, but placed 
in a contrary position, so that its line of pressure shall have, at P, 
a common tangent with the line of pressure of the first mass. 

* The line of pressure is here supposed to commence at P', and the force P" to enter 
among the other forces which determine its equation. 
Vol. V. Part III. S s 


Two masses, therefore, thus placed together would remain in equi- 
librium, without the aid of any external force, and by reason only of 
their mutual pressures and the resistance of their abutments. 

It is also evident that since the line of pressure is similarly situated 
in both, they cannot be thus placed together so that their lines of 
pressure may meet and have a common tangent at the point where 
they meet, unless both lines of pressure be perpendicular to the com- 
mon surface at that point. 

This condition throws two new equations into the system, and de- 
termines the value of P' in terms of a single variable. 

The value of P' is not in this case that which we have called 
the absolute minimum or minimum minimorum, but simply the greatest 
or least force, which applied at a given point, in a given direction will 
support the system. 

If however instead of a single point of contact we suppose the 
masses to be in contact throughout the whole surfaces of two planes, 
it is evident that the point P' * will take up for itself that position, 
which we have supposed to correspond with the absolute minimum ; 
a condition to which the form of the line of pressure, and the 
position of its point of contact with the surface of the mass, will also 
be subjected. 

Hence it appears that two masses, thus in contact throughout the 
surfaces of two planes, sustain a less aggregate of pressure, on their 
common surface of contact, than two similar masses in contact only 
by a single point, unless that point, and the position of the masses, 
be such as to correspond to the minimum minimorum. 

In the preceding pages we have supposed the form of the solid 
to be given, together with the positions of the different sections 
made through it, and we have thence deduced the form of its line 
of pressure and the direction of that line through its mass. 

. ?;* The point P is here the point of application of the resultant of the resistances on 
the different points of either plane. .:iorte«pi sJi -^..rnaHb 

..H .Ui T ■ 


It is manifest that the converse of this operation is possible. 

9. Having given the form and position of the line of pressure, and 
the positions of the different sections to be made through the mass, we 
may, for instance, enquire what form these conditions impose upon the 
surface which bounds it. 

Or we may make the direction of the line of pressure and the 
form of the bounding surface subject to certain conditions not abso- 
lutely determining either. '"* t'^ oxi'>»^i'« inn ouJ ifion 

For instance, if we suppose the form of the intrados of an arch to 
be given, and the direction of the intersecting plane to be always per- 
pendicular to it, and if we suppose the line of pressure to intersect this 
plane always at the same given angle with the perpendicular to it, 
so that the tendency of the pressure to thrust each from its place may 
be the same, — we may determine what under these circumstances must 
be the extrados of the arch. ,''^'^",' '"'^ 

If this angle equal constantly the limiting angle of resistance, the 
arch is in a state bordering upon motion, each voussoir being upon 
the point of slipping downwards or upwards, according as the constant 
angle is measured above or below the perpendicular to the surface of 
the voussoir. 

The systems of voussoirs which satisfy these two conditions are the 
greatest and least possible. 

If the constant angle be zero, the line of pressure being every- 
where perpendicular to the joints of the voussoirs, the arch would 
stand even if there were no friction of their surfaces. 

It is then technically said to be equilibrated. It is impossible to 
conceive any arrangement of the parts of an arch by which its stabi- 
lity can be more effectually secured*. 

10. The theory stated above readily explains the phenomena ob- 
served in the settlement and fall of the arch. 

* The great arches of late years erected by Mr Rennie, in this country, have for the 
most part been so loaded as very nearly to satisfy this condition. 

ss 2 


Thus let ABS" (fig. 3) represent an arch having the joints of its 
voussoirs perpendicular to the intrados as they are usually made. 

Let RQPQR' be the line of pressure, touching the intrados in the 
points Q and Q'. It is manifest that this curve is then perpendicular 
to the joints of the voussoirs at Q and Q, and inclined in respect to 
those above and below these points. The inclination being downwards, 
or towards the intrados, in reference to the former, and upwards, or 
from the intrados, in reference to the latter. 

Hence, therefore, it appears that the tendency of the pressure is 
to cause all the voussoirs above the points Q and Q' to slide down- 
wards, and those beneath those points, upwards. 

And that these effects may be expected to follow the striking of 
the centre of the arch ; the weight being then suddenly thrown upon 
the voussoirs, and these admitting of a certain degree of motion in 
the directions of the forces impressed upon them. 

Now this is precisely what was observed at the bridge of Nogent, 
of the construction of which Perronet has left a detailed account. 

Three straight lines were drawn upon the face of the arch before 
the striking of the centre, shewn in the figure 4, by the polygon 
nmm'n', mm' being horizontal, and the other two mn and m'n' stretch- 
ing from the extremities of mm' towards the springing of the arch. 

After the centre had been struck, the lines were observed to have 
assumed the curved forms indicated by the dotted lines MM', MN', 
M'N', indicating, in accordance with the theory, a downward motion 
in all the voussoirs above Q and Q', and an upward motion in those 
beneath those points. 

These observations have been confirmed by numerous others, and 
especially by those (made also by Perronet) at the Pont de Neuilly. 

The sinking of the voussoirs at the crown necessarily tends to pro- 
duce a separation of their joints at the intrados in the neighbourhood 
of that point, and thus to cause the actual contact of the key and 
adjacent voussoirs to take place only at their superior edges. 


If therefore the settlement be considerable, we may conclude that 
the line of pressure touches the extrados at the crown, and for 
some distance on either side of it. The material of the arch may 
therefore be expected to yield more particularly about that point and 
the points Q and Q' than any other; a great proportion of the 
pressure being there thrown upon the edges of the voussoirs. 

11. If by reason of such yielding, or from any other alteration in 
the forces impressed upon the mass, or in the circumstances of their ap- 
plication, the form of the line of pressure be altered, it may manifestly 
be expected to intersect the surface of the mass first about those points; 
the least possible alteration of form being there sufficient to produce 
the intersection. And this being the case, the portion of the arch above 
Q and Q' must separate into two portions, revolving at those points 
about the lower portions of the arch (see fig. 5) and at A, upon the 
extremities of one another. 

Nevertheless this revolution is manifestly impossible unless the 
points Q and Q yield outwards. And this can only take place by 
the yielding of the material at Q and Q', by the slipping back of 
the voussoirs there, or by the portions of the arch or its abutments 
beneath those points revolving outwards, in consequence of the inter- 
section of the extrados by the extremities QR and QR' of the line 
of pressure (fig. 3). 

The last is in point of fact the cause which leads, in the great 
majority of cases, to the fall of the arch. 

The extremity R of the line of pressure is made to cut the 
extrados of the arch, or the outer surface of the pier, by the 
diminution or removal of some force which acted there in opposition 
to the tendency of the arch to spread itself, and which kept the 
direction of the line of pressure within its mass, — the resistance of 
a mass of earth for instance, or the opposite thrust of some other 
arch springing from the same pier or abutment. 

On the whole, then, it appears that in the commencement of its 
fall the arch will divide itself into six distinct portions, of which four 


will revolve about the points S, S', Q, Q' and A, as represented in 
the figure 5. Now this is what is uniformly observed to take place 
in the fall of the arch. 

12. Gauthey, having occasion to "destroy a bridge, caused one of its 
arches to be insulated from the rest; and the adhesion of the cement 
being sufficient to counteract the tendency of the pressure to rupture 
the piers, he caused them to be cut across. The whole then at once 
fell, the falling portion separating itself into four parts. Having con- 
structed small arches of soft stone, and without cement he loaded them 
until they fell. Their fall was always observed to be attended with the 
same circumstances. Before the arch finally yielded the stone also was 
observed to chip at the intrados about the points Q and Q', round 
which the upper portions of it finally revolved. 

Some experiments made by Professor Robinson with chalk models 
were attended with slightly different results. Having loaded them at 
the crown until they fell, he observed first, that the points where 
the material began to yield were not precisely those where the rupture 
finally took place. 

This fact presents a remarkable confirmation of the theory expounded 
in this paper. 

It is manifest, that according to that theory, with any variation 
in the least force P', which would support the semi-arch if applied 
at its crown, there will be a corresponding change in the position of 
the point Q. 

Now as the load upon the crown is increased, this least force P' 
is manifestly increased. The result is a corresponding variation in the 
• form of the line of pressure, tending to carry its point of contact 
with the intrados lower down upon the arch. 

This is precisely what Professor Robinson observed. The arch 
began to chip at a point about half way between the crown and the 
point where the rupture finally took place. 


The existence of the points Q and Q', about which the two upper 
portions of the arch have a tendency to turn, and about which the 
material is first observed to yield, has long been known to practical 
men. The French engineers have named these points the points of 
rupture of the arch ; and the determination of their position by a 
tentative method forms an important feature in the very unsatisfactory 
theory which they have applied to this important branch of Statics. 

13. The theory of the equilibrium of the groin and that of the 
dome are precisely analogous to the theory of the arch. 

In the former case a mass springs from a small abutment spread- 
ing itself out symmetrically with regard to' a vertical plane passing 
through the centre of its abutment. It is in fact nothing more than 
an arch, whose voussoirs vary as well in breadth as in depth. The 
centres of gravity of the different elementary voussoirs of this mass 
lie all in its plane of symmetry. Its line of pressure is therefore in 
that plane, and its theory is embraced in that which has been already 
laid down. 

Four groins commonly spring from one abutment ; each opposite 
pair being addossed, and each adjacent pair uniting their margins. 
They thus lend one another mutual support, partake in the properties 
of a dome, and form a continued covering. 

The groined arch is of all arches the most stable ; and could ma- 
terials be found of sufficient strength to form its abutment and the 
parts about its springing, it might be safely built of any required 
degree of flatness, and spaces of enormous dimensions might readily 
be covered by it. 

It is remarkable that modern builders, whilst they have erected the 
common arch on a scale of magnitude nearly approaching perhaps the 
limits to which it can be safely carried, have been remarkably timid 
in the use of the groin. 


King's College, London, 
Ocl<^er 9, 1833. 

XIV. Third Memoir on the Inverse Method of Definite Integrals. 
By the Rev. R. Mukphy, M.A. F.R.S., Fellow of Cuius College, 
and of the Cambridge Philosophical Society. 

i;;Read March 2, 1835.] 


In the two preceding Memoirs on the Inverse Method of Definite 
Integrals, the limits of integration had been fixed throughout at and 
1, but in the sixth Section, which is the first of the present Memoir, 
the integrations terminated by arbitrary limits are fully considered; and 
when performed with respect to any function of the independant vari- 
able, the proper methods for discovering reciprocal functions are given, 
and it is remarkable that the forms thus obtained for the trigonome- 
trical functions, for Laplace's and an infinite variety of other reciprocal 
functions, are all similar, differing only by a constant. 

In identities obtained between the »"" differential coefficient of a 
function not containing n, and its expanded value, we may, generally, 
by changing the sign of n, obtain a corresponding identity between 
the ra"" successive integral and its expansion, abstracting from the ap- 
pendage of integration which ought to contain ?« arbitrary constants ; 
this property however extends also to certain reciprocal functions which 
contain n ; and this consideration leads in the same section to the com- 
plete resolution of Laplace's equation for the reciprocal functions of 
one variable, which are the coefficients in the developement of the reci- 
procal of the distance of two points; the w*"" coefficient when multiplied 
by an arbitrary constant, satisfies that equation, as is well known, but 
as the equation is of the second order, another function multiplied by 
■^ Vol. V. Part III. Tr 


an arbitrary constant must be also represented by the same equation, 
this function, which is here found, is altogether different in its form and 
properties from Laplace's coefficients. 

The great class of reciprocal functions above alluded to possess the 
remarkable property, that their integrals vanish between any of their 
own maxima or minima values. 

In this Section I have noticed some curious trigonometrical func- 
tions of which the properties are very elegant, particularly as affording 
simple means of representing by Definite Integrals the general differ- 
ential coefficients of rational and integral functions ; another applica- 
tion of trigonometrical functions is made, in representing the sum of 
the divisors of any given number, by means of a Definite Integral. 

The seventh Section is on Transient Functions. The way of forming 
reciprocal functions by means of arbitrary coefficients, when the form of 
the general term was given, has been shewn in the Second Memoir on 
this subject. To this I have here added the method of finding the 
functions which shall be reciprocal to any proposed one, and applied 
the method to the cases where the given function is r, (log. t)", and 
cos" {t) ; the reciprocal functions which thence resulted are transient, that 
is, they have but a momentary existence between the limits of inte- 
gration ; that existence is however sufficient to make their integrals 
finite, and to endow them with remarkable properties. They are capa- 
ble of representing the electrical state of a body when an electrical 
spark is infinitely near, and about to form a part of the system ; they 
are also capable of representing, under continuous forms, the state of a 
body considered as composed of absolute mathematical centres of forces, 
separated mutually by infinitesimal intervals. 

The eighth and last Section is on the Resolution of Equations which 
contain Definite Integrals; the first method for this purpose is to de- 
compose the integrals into elements, and then determine the unknown 
functions by elimination. This tedious process is useful in verifying 
results otherwise obtained, and in giving numerical approximations in 
the most difficult cases. Afterwards I have considered separately, 


Equations to Definite Integrals ; first, when they contain but one Defi- 
nite Integral and one parameter ; second, when they contain two or 
more Definite Integrals and as many parameters; third, Simultaneous 
Equations ; fourth, Definite Integral Equations of superior orders and 
degrees; besides which, the nature of the appendage analogous to the 
arbitrary constant of integration is discussed in the same Section. 

Throughout the whole of this Memoir, a considerable number of 
examples, illustrative of the corresponding theories, are dispersed. 




Method of' discovering Reciprocal Functions when the integrations are per- 
formed with respect to any Junction of the independant variable. 

(l) When the limits of integration are arbitrary. 

1. The investigations of reciprocal functions contained in the Second 
Memoir on the Inverse Method of Definite Integrals, are founded on the 
supposition that and 1 are always the limits of the independant 
variable, but it is often of importance to possess reciprocal functions in 
which the limits of integration are different from those quoted. The 
principle by which this is most easily accomplished, is to suppose the 
integrations performed relative to a function of the independant vari- 
able, which must be so chosen, that when the values and 1 are 
assigned to the independant variable, the corresponding values or the 
function may be the proposed limits of integration. 

2. Let Q„, R„, be functions of a variable (^), the limits of which 
are arbitrary, as a and h, between which limits f^Q^Rm always must 
vanish, except when the integers m and n are equal. 

Suppose that a function of <p, as t, is found such that when ^ = a 
t = 0, and when (p = h, t=l, conditions which it is always easy to satisfy. 

We may now conversely regard as a function of t, and then the 
preceding integral becomes fiQ„Rm-jr, the limits being now reduced to 

and 1. Suppose that -~ is separated into any two factors, X and X'; 

then since f,QnX x R,„\' = 0, except when 7n — n, it follows that Q„X, 
R„\' are mutually reciprocal, and may therefore be found in an inde- 
finite variety of modes by the principles explained in Section iv; and 
dividing these functions respectively by X, X', and substituting in the 
quotients the value of t expressed in terms of ^, the required functions 
Q„, Rm will be obtained. 


If it be desired that Q„, R^ should be functions of the same 
nature, differing only in the order expressed by m and n, that is 

self-reciprocal, put \ = W = \/{-~\, and having found any kind of 

self-reciprocal functions in which the limits are and 1, as for ex- 
ample, the functions denoted by P,„, P„ in the preceding Memoirs, we 
then obtain 

3. If a function V can he determined so that the quantity 

d°f(ttyV} dt^ 
dt" ■ d0 

may he of n dimensions in t, (where t' = 1 — t as in the former Memoirs), 
this quantity will he a self-reciprocal function when the integrations are 
performed relative to (p. 

Denote this quantity by Q„, and supposing m to be an integer 
less than n, it is necessary to show that f,pQmQn — 0, or that 


^'^- di" -^' 

the limits of t being and 1. 

Now Q„ being of m dimensions in t, let its general term be re- 
presented by Oj.f, where it is evident that p cannot exceed n — 1, 
since m<n; the part of the preceding integral dependant on this term is 

""'^'^ dv' — • 

The latter integral may by partial integration be put in the form, 

the last term being 

and therefore the index of differentiation never becomes negative. 


The first term, and 'a fortiori', all the succeeding terms of this 
series vanish between the limits ^=0, and t=\, or t' = 0, for 

d''-'{{tt'rV} _rr d"-'{tt'Y ,,^ ^,dV (f-^tty 
dt--' ~ dt^-' ^^ ' dt dt"-' 

{n-l){n~2) dT d^-^itt'f , 
"^ 1.2 dt' ■ dt"-' "^ ' 

the first term of this latter series contains a factor tt', the second a 
factor {tt'f, &iC., and therefore the whole vanishes between limits. 

The following exception to this theorem must however be attended 
to; V must not he of the form {tt')".Vj, where r is equal to, or 
greater than unity, for the above reasoning will not be applicable, 
since then 

d"-mtt'Yr\ _d'-'{{tty-^r,} 


dt"-^ dt' 

which being expanded as above, will not vanish unless r be less than 

4i. If a function V can he determined so that the quantity 

d''f(ttrvi d0 

at" ■ dt 

may he of n dimensions in t, then the factor hy which -^ is here 

multiplied, will he a self-reciprocal function when the integrations are 
performed relative to cp. 

Denote this coefficient by q„, then 

r -r ^0 _ /■ <^" (tf'Y ^ d<t> 

and as we may suppose m<n, the general term of qm-^, as a^t^ 

cannot be of greater dimensions than n — 1, and therefore the part of 
the whole integral dependant on this term vanishes, as has been 
shewn in the preceding article, hence f^qmq„ = 0, when m and n are 


We must except, as before, from the application of this theorem 
the case where V is of the form {tt')-\Vi, and r greater than, or 
equal to unity. 

5. If (f) be any of the transcendants contained in the indefinite 
integral jj (tt')", where m is between — 1 and + x exclusive, and if 

^"~ 1.2.3...ndt"'^"^ ' 
then shall Qn be a self-reciprocal function for integrations relative to <p. 

For Q„ is evidently of the form — ~rp: ■ ~TZ' ^'^d ^ is not 

of the form excepted in Art. 3., since m is between —1 and + oo. 
Moreover, by actual differentiation we get 

1 .^.S-.-ndt" 
where a, b, c, &c. are constant quantities. 


Q„ = at"' ^btt'"-' +ctH"'-^ + kc., 

which is of m dimensions in t, and therefore all the conditions re- 
quired in Art. 3. are here fulfilled; therefore Q„ is a self-reciprocal 
function relative to <p. 

6. If (p be any of the transcendants expressed by the indefinite 
integral jj (tt')"", where m is between + 1 and — oo exclusive, and if 

qn = 

- d° (tt')°- 


n» ■ 

then is qn a self-reciprocal function relative to (p. 

d' (tfy V 
For §-„ is here of the form i ' — , and V does not belong to 

the excepted cases, moreover 

# _ d'^.jtt'y-'" 

^'' dt ~ \.2...ndt"-^^^^ 
is evidently of n dimensions in t, therefore all the conditions of 
Art. 4. are here satisfied. 


7. For the purpose of convenience both in evaluating and using 
reciprocal functions, the knowledge of the functions which they generate 
is very useful. The generating function, for example, being the quan- 
tity denoted by q^, Art. (6), the process for finding in this case the 
function generated, will sufficiently exhibit the general principle, and 
therefore it is now proposed tb sum the series q^ + q^h + q^k' + q^h^, &c. 

Substituting for q„ its value given in the preceding article, and 
representing the required sum by S we have 

o /. 'V J. ditty-'" le d'itt'f-'" M dHtfy-'" , , 

But if we form the equation, u = t + ku (1 — u), and suppose y'(M) to 
be the derived function from J'{u), we have generally 

^^ r(«\-f'(A^h^it^^)-^A. *' d^{f'it).(ttj} 

,_f^_ d?\f{t).{tty\ 

+ 17273 • df *'''■ 

which is obtained by differentiating the value of /(«) given by La- 
grange's Theorem. 

The preceding series coincide by supposing 

f(f) = {ttf)-"' = t-" il-t)"", 

and therefore /'(«) = «"" (l-w)"" = j^-L 

by the assumed equation. 

(u-t)-'" du 
Hence 5- = -^^ . ^^ . 

Now the actual solution of the assumed quadratic equation gives 

u = 


, where R= {l-2h{l-2t) + h'}K 

, B-l-\-h{l-2t) . du 1 
whence u-t= ^ , and -^ = ^i 


therefore S = |/2 - 1 + A (1 - 2/)} -" 


Knowing thus the generated function S, we can conversely find q„ by 
taking the coefficient of A" in the quantity S, and substituting for t 
its value in terms of <^. 

An exactly similar process applied to the function Q„ of Art. (5), 
woxild give 

as the function generated, 

and observing that 

R'-{\-h{\-^t)\" ^ 4^h'tt', 

this quantity may be transformed to 

a III 


so that Q„ is the coefficient of h" in the expansion of this function. 

8. From the theorems given in Arts. (5) and (6), we can determine 
reciprocal functions relative to <p, which quantity may denote any 
transcendant contained in the formula Jt{tt'y, from m—-<xi to »« = + x ; 
circular arcs are amongst these transcendants, namely, when m = — ^, 
and since both theorems are true simultaneously, when m is between 
— 1 and + 1, we shall get in this instance the two species of circular 
self-reciprocal functions, namely, the sines and cosines of the multiples 
of the simple arc. 

I. To evaluate Q„ when «/ = — i- 

For the variable with respect to which the integrations must be 
performed, we have 

^ = jXtty^ = l ^y^r^ -■= COS- (1-2^), 

neglecting the constant which is unimportant. 
Vol. V. Part III. Uu 


By Art. (7), 

Q„ = coefficient of Jf in -^ {^ + 1 - A (1-2^)}^, 

in which R represents |1 — 2// (1 - 2/) + /i'}*. 

Putting for t its value in terms of 0, we obtain 

J? = {1-2/i cos^ + /i^}-i = (l-Ae*^^)i.(l-//e-'''^^)^^ 
and l-/i(l-2it) = l-/4cos0 = 1(1 -Ae*^^) + ^ (1 -//e"*^^). 

Hence, ^ + 1 -/i (1 - 2/f) = |{(l-^e*^^)^ + (1 -//e-*^^)-^''; 

therefore, Q„ = coefficient of /r in x . Jj " ^t-v-!!-^4' "r'^^! 

= ^ coefficient of A" in (l-/<e*^^)-* + (1 — Ae-*^'^)-* 

= c 

c . cos n<p] 

13 5 (2« — 1) 
where c = ' ' '"^ ^ , the limits of ^ are and w. 

2.4.6... 2ra 

II. To evaluate q„ when y« = — i. 

As above, we have (p = eos"' (1 — 2t), 

and q„ = coefficient of h" in ^-^ . {^- 1 +/< (1 — 2/)}''. 

But i? - 1 + /. (1 - 20 = i p-^;!:l'^>' - (i::.^-_:!:^^)H'^ . 

I V -1 V-l j 

-. q„ = ^ coefficient of /«" + ' in 

\/-i V-i 

c — . = c sin (1 + n) (p, 

* v — 1 

, 1.3.5...(2m + 1) 

^^^••^ ^' = 2.4.6■.■(2;^4-2) ' 


9. But whatever may be the value of m, the quantities Q„, q„ may 
always be simply expressed in terms of t by the theorem of T^eibnitz, 

d"(uv)_ cV'v clu d'"^v 7i.{n — l) dHi^ f/"'^ v 

after {yiplying which we may substitute for t its value in terms of (p. 
Thus when m= — ^ 

1.3.5....(2«-1) n 2n-l 

2.4.6....2W ^' ~1- 1 " 

• ' J. n{n-l) {2n-l){2n-3) , ,„_, _ „ . 

_ 1 .3.5.. ..{2n - 1) , 2n{2n-l) 

-" 2.4.6... .2» ^^ 1.2 " 

, 2^(2?^-l)(2>^-2)(2«-3) ,,^,„., . 

"^ ^^ "^''•^ 

•^ 2.4.D....2ra ^ ' ' ^ 

and in the same way we have 


^"~ rr2. 3. ...«<//" 

_ 3.5.7....(2» + l) » 2;» + l 

~ 2.4.6....2« ^^^ 1- 3 ^' " 

w(w-l) (2w + l)(2«-l) 5, 3 „ , 

= I 3.5.7.--.(2?? + l) , ^--- . ^ , — i /.h2»+2) 

2V"=l"2.4.6....(2« + 2) ^^^ +^^^) -(^ _V-1#0 \, 



and passing to the variable cp, since 1 — 2^=cos^; therefore ^ = sin — 


and /' = cos-^, whence #'^ +\/ — 1 #* = cos^ + \/ — 1 sin^ by substi- 
2 2 ~" 2 -^ 

tuting which we obtain 

„ 1.3.5....(2w-l) , 

^'-^ 2.4.6....2/. •^"^^'^' 

1.3.5....(2«4-J) . , ,, ^ 


which values are the same with those in Art. 8. 

The numerical coefficients in these formuhe may be rejected as having 
no importance in self-reciprocal functions ; it is also observable that q„ 
contains a different multiple arc from that in Q„, the reason of which 
is that Q„, <7„ are to be self-reciprocal functions for all entire values 
of n from to + oo, and then f,j,q„q,n = except when 7i = »i, this ex- 
ception (on which the main value of reciprocal functions depends) would 
not hold universally true if q„ were of the form sin(«0), for then 5-0 = 0, 
and therefore f^qo.qo=f> contrary to the principle of the exception, 
but in the form above found this irregularity does not occur. 

10. From the results found in Art. 9, it follows that if we put 
the real functions Q„, q„ possess a common property, viz. 

except when m = n, which exception does not apply to the last integral 
when m = ?i = 0. 

From the same results the following identities are obtained : 

, ff'^*!T\. 1. ■ («')^ = cos {n cos-' (1-20} 
1.3.5....(2tt-l)</^" ^ ' ' \ Ji 

(» + l)2''+'rf"(«T^ • J/ , -,. w, o*\\ 

i-3-^i-^^^-^,, =sm U« + 1) cos ' (1 -20}. 


We shall now consider whether analogous formulee hold true for negative 
values of n the index of differentiation. 

Generally if u and v be functions of t and fi'u denote the w"' suc- 
cessive integral of ti, then 

for if we take the w* differential coefficient of each term in this series, 
all the terms resulting mutually destroy each other except the first 
term tiv. 

Putting u-=t'-"-^, v = t""-^, and rejecting the constants of integration 
in the latter, we have 

also — - ^^L±lt'--l, ^ - (2« + l)(2« + 3) .,_„_! » 

Hence fiitf)'"-^ 
(-2)"(^0-- (.,-. n 2«+l ,_^_, n{n-l) (2^+l)(2« + 3) „_ „ „ , 

= i.3.5....(2«-i)^^ "i-~T~-^ ^+-r¥-- Ts ^ ^-^^-'^ 

«r ^327 • dt-'^ ■ ^^^ > 

= cos {mcos-' (1 — 2^)^, 

the appendage which contains all the arbitrary constants being 

{^o + ^it + ^.f+...A„_J"-'\ . {tt')K 


Dividing the last equation by {tty, and integrating witt respect 
to t, we get 

1.3.5...(2w-l) d-'^^-'iffy-i 1 . c i/i o.M 
7 — x; — ^ • TT — 1 = - Sin \ncos-Ul-2t)}. 

Putting » = ???- 1 , we get 

, ,, 1.3.b...(2m-S) <?-"(«') -'" + -i . <,, , ,,, ^,, 

('"-1) • {-2)'"-' • dt^ "''" {(i-»i) cos-' (l-2t)}, 

thus are obtained the corresponding formulae for negative indices. 

11. The two series of reciprocal functions arising from the theorems 
in Arts. 5. and 6., differ essentially, only in reference to the inde- 
pendant variable of integration, for in Art. 5., ni may be any quantity 
between —1, and +x, and in Art. 6. any quantity between +1 and 
— 00 ; change in the latter theorem m into — m, and the limits of w 
Avill then be the same in both ; for distinctness, also let 6 be used 
instead of (p in the value of §'„. 

d" itt'Y*'" 
Hence, Q^= i ^.s.'.ndf ■^*^'^'""' ^"'^ <t> = !^itfY, 

d" (ttY'^" 

^- lALndt" ^ ^"d ^ = ;(«')-'". 

Now the reciprocal functions of Art. 5., give the equation 
UQnQn=0, or feQM,. ^=0. 

But ^ =(«')", and ^ =(«')-'" ; therefore ^ = («7'». 

Hence, feQAtt'Y ^ QAtt'Y = 0. 

And since QAtt'T = qn, and QAtt'y = q„; it follows that UQnQn' is 
equivalent to [dq„qn; the only difference being with respect to the 
variables (f> and 9 employed for integration. 


If in the formulEe of Arts. 5. and 6., we assign to m all possible 
values between —1 and 4-1, we obtain two series of self-reciprocal 
functions, which when m = become identical with each other, and 
with the functions denominated P„ in the preceding memoirs. For 
every other value of m between those limits, there are two different 
kinds of reciprocal functions, one of which only is a rational and entire 
function of t, for instance when m= —\, we have found the functions 
cos n(p and sin {7i + l)<p, the former of which only is a rational fvmc- 
tion of cos <p. 

12. (1.) W/ien m= -i- 

To determine cp in this case, make sin 9 = ii — / ', squaring and ob- 
serving that t + f' = l, we get sin^ 6 = 1 -2 {tt')K whence » 

/■i + ?;'i = V^(2 - sin- 0), and 2'\tty = cose. 
Differentiate the assumed equation, and we get 

^os ^ = a ^**'\\ • ~7^ ' therefore — -^ . -r^ = 2 cos . -7 

;i 5 

2 {tty • cie' ^"-'-— ^tt')i ■ d0~ ■ti + t'i 

hence, (p = 2E {e)-F(d). 

The extreme values of the amplitude of these elliptic functions 

being — -, and + -; the limits of ^ are 0, and 4:Ei — 2JF\, where 

El and Ei denote the complete functions when the amplitude extends 
from zero to a right angle. 

The reciprocal functions for integrations relative to (p, are 



4.8.12 4.W 

4n{^n-l) 4n{4>n-l)(in-i){4u-5) „ , 

^* sTi * ^^ ^ (,^^-U 


5.913...(4« + 1) 


4.8.12 4« 

, (,0J {t-- (4>^ + l)-4^,.-.,^ (4. + l).4^».^(4«-3)(4»-4) ^,„.,^, ^^^^ 

(2.) When m= — 1. 

In this case <?„ = 1.2.3...,,^^, 

and <^ = j;(«')-'=h.l. (I). 

Hence, ^ = e*, ^ = 1 + e* ; 


_(« + l)_ ^^ ^+1 nin-X) {n^\){n) ^_^ , 

9» (1 + £</>)»+'= »^ 1- 2 -^ ^ 1.2 ■ 2.3 "^ '^'^■^' 

where the limits of ^ are — « and + w . 

13. To express the functions Q^ and qn z« terms of t alone. 
By Art. 6., we have 

^'~ 1.2.3...ndf 

= (w-?»)(w-OT-l)...(l-OT) _ „_ 71 n-m 

1 .2.3...W v ; • J j^ . ^ _^ 

«j^_l) (/^-m)(«->»-l) , 

^ 1.2 • (l-»»)(2-»/) ^ ^>&c.}. 


Suppose 1- 1 substituted for t' in each term between the brackets, 
then expanding each, the coefficient of t" in the whole will be 

n{n-'l)...{n-r + l) , ^y t-. , „ »-"^ , r(r-l) {n~m){n-m-l) , ,_ , 
1.2...r "^ ^^ ^^^^l-m^ 1.2 ' {l-m){2-m) ^^^'^ 

«(«-!). ..(M-r+l).(-l)' (^„-„ d't'-'" d.f-'" </'->. r-" 

1.2...rx{l-tn) (2-m)...{r-m) '^ dt' dp ' dt'-' 

r.(r-\) d^ . t"-" d'-' . p-'" 
■^ 1.2 • dt^ '~dF^' ^^•^' 

when t is put equal to unity after the differentiations. 

But by the theorem of Leibnitz, the part within the latter 
brackets is equivalent to 

fjr fr+n—2m 

— -^- — =(n-2m + l) (n-2m + 2)...{n — 2m + r).t"-'"", 

hence, the required coefficient of 

,_,_., «■(»- l)...{n-r+l) {n-2m+ 1) {n — 2m + 2)...{n — 2m + r) 
~^ '' 1.2...r ^ il-m) (2-m)...(r-m) 


« _ (»-ffi)(n-m-l) . . , n n-2m+l 
7"- 1.2.3...n ^'~^' ^^~T' l-m '^ 

n(n-l) (n-2m-\- 1) (n - 2m + 2) ^ , , 
"^ 1.2 • {l-'tn)i2-m) * > ^<^'h 

Again, by Art. 5., 

_ d"(ttT'" 

^"- l.2...ndf-^^^' 

(n + vi) (n + m—l)...(l+m) j^,„ n n + m ,„_, 

1.2...n *■ 1 1+m 

t'"-' t 

, n.jn-X) (n + m) (n + m -I) ^„.,^, \_, 
"*■ 1.2 • (l+m)(2 + m) ' ^'^^.j, 

Vol. V. Part III. Xx 


the reduction of which to the powers of t is effected as before, putting 
— m for m, whence 

(w + ot) (w + m-l)...(l +»?) ( n n + 2m + l 
^"~ 1.2...ra ^ 1* 1+m ■ 

When j» = 0, 

« o 1 ** " + i M«:il) (w + i)(w + 2) 

which is the same as the value of P„, Sect. ii. Art. 2. 
When m= — ^ 


and t = sin^ ^ 


{(w + ly-l^{(>^ + ly-2-} _,^.^,0 

Q»=2.4 2n •il-1.2'^ ''" 2 + -^ ''" 2 *'''-^- 

14. To express the quantities Q„, q„ by means of a differential 

Suppose /{t) is a function of #, subject to the condition 

t(l-i) ./"it) + (»»+ 1) (1-20 ./' (0 + « • (« + 2/»+ 1) ./{t) = 0, 

where /"(^ denotes the second, and /'(^) the first differential co- 
efficient of f{t) relatively to t ; differentiating this equation, we get 

t{l-t) .f"{t) + (»« + 2) (1 - 2^ ./"(/) + (»- 1) (« + 2»? + 2) ./' (^) = 0, 

^ (1 -0 ./"" {t) + (w + 3) (1 - 20 ./'" (0 + (« - 2) (W + 2»« + 3) ./" {t) = 0, 

and generally, 



Put # = in all these equations successively, thence we have 

{m + l)./'{0) =-n.{n + 2m + l).f{0), 

(m + 2) ./" (0) =- in~l){n + 2m + 2) ,/' (0), 

m + 3 ./'" (0) = - (ra - 2) . (« + 2w + 3) ./(O), 


it follows from this by Maclaurin's Theorem, that the preceding equa- 
tion will be satisfied, as a particular solution, by taking 

^•/^v ^/«v(i ^ n + 2m+l .n.{n-\) (w + 2w^ + 1) (w + 2>» + 2) ., , . 
./(0=/(0){l-i.-i-^^-.^ + ^^. (i + ^).(2 + «.) ^ ^^-&c-}» 

and ,/(0) being arbitrary if we put it equal to 

(1 + m) (2 + w<) (3 + ?») (n + m) 

i '. 2 ; 3 TTTT^ n ' 

this value oi f{t) will become the same as the value found for Q„ in 
the preceding article ; hence, replacing \ — thy its equal t', we get 

(«') ^ + (« + 1) (1-2^) . ^ + « . (w + 2m + 1) . Q„ = 0. 

But if in the value of /{() we change the sign of m, putting 

... _ (l-m)(2-?w) {n-m) 

'^^"^~ 1.2 n ' 

then y*(#) becomes equivalent to q„ {tt')" ; and if we put this for y (#) 
in the first supposed equation, and divide the result by {tf)"', we get 

{ti')^ + im + l){l-2t).^+{n+l){n-2m).q„ = 0. 

(2) Particular inferences resulting from the preceding theory. 

15. Denoting as before by <f> the" indefinite integral fi(tt')'", and 



then assigning to m all possible values from — oc to +00, the functions 
Q». qn will give an infinite series of reciprocal functions relative to all 
the transcendants contained in ^ considered as the variable of integra- 
tion ; and when m is between - 1 and + 1, pairs of reciprocal functions 
will be obtained, except when ?» = 0, when both coincide. 

In this series are included the trigonometrical functions, namely, 
when m-= —\'., and Laplace's functions, when /« = 0. 

In all the reciprocal functions thus arising, there exists one common 
property, namely, the definite integral always vanishes between the 
limits which make the functions themselves maxima and minima; this 
remarkable property I have had occasion in another place to notice, in 
the particular case of Laplace's functions.* 

To prove this generally take the equations of the preceding article, 

/#'^ + (»» + l)(l-20.-^'' + «(w + 2»« + l)Q„ = 0, 

«'^ + (?» + l)(l-20•-^+(« + l)(w-2»^)9„ = O. 

Multiply both equations by {tiy, and integrate reserving the con- 
stants under the integral sign ; hence, 

{tt'Y^^ ^ + « (« + 2m + 1) j: Q„ (tty = 0, 

' (^0'""''-^+ (« + l)(«-2»^)/,^„ («')"* = 0; 

and changing the independant variable by the condition 7;r =(^^')"'"» we 

(«')""+^^ + w (« + 2/» + 1) 4 Q„ = 0, 

(<0*"*' ^ + (» + 1) (»« - 2»w) /^ ^„ = 0. 

Electricity, Introduction. 


But when Q„, q„ are maxima and minima, -^ and -^ respectively 

vanish ; therefore, between the corresponding limits of (p, we must have 
U Q» = 0, f^qn = 0, which general property is easily verified when 

Q„ = «cos«0 and q„ = a sin {n + 1) (p. 
16. To find the complete integral of the differential equation 
tt' , -^-^ + (m + 1) (1 - 2t) ^ + n (n + 2m + 1) tr =s 0, 
where n is integer and m any constant. 

The differential equation for Q„ (Art. 15.) is of the same form as 
the above equation, and therefore u=cQ„ is a particular solution, c 
being an arbitrary constant. 

The form of the differential equation for q„ will become the same 
as that of the given equation, if — (w + 1) be written instead of n in 
the former; hence, another particular solution is c'q-^„+^y 

The complete solution is therefore 

u = cQ„ + c'q.^„+iy 

This solution fails first when m = 0, for then the functions Q„, 
5'-(«+i) in their expanded forms become both identical with Laplace's func- 
tion P„, and consequently the two constants c, c' merge into only one, 
viz. their sum ; but if we put generally 

b , , b 

c = a -\ — and c = , 

m m 

then M=«Q„ + &. ^"~ ^-'"^' 


And putting m = 0, the latter term becomes a vanishing fraction, and 

u = aP, + ^^{Qn- S'-(«4.i)} when m =0. 


The term by which b is here multiplied, is the coefficient of m in 
Qn-q-i,„^^), which is easily found from the expansions in Art. 13; hence, 

n{n-\) (« + ])(w + 2) r 1 1 I 1 1 Ux,. 

The general solution also fails when m is an integer, for then some 
of the terms in the expansion of Q„ or g-.^+u will become infinite, and 
the principle of vanishing fractions will simply enough in this case 
also be applicable in determining the complete solution ; but if we put 
for Q„, q„ their differential forms, the solution will never fail, for the 
failures arise from the entrance of logarithms into the result, and these 
will actually enter in the latter forms; changing our constants, the 
complete solution for all cases is 

it is therefore necessary to shew that the functions by which the ar- 
bitrary constants are multiplied, are particular solutions. 

Putting v-itty-"^, then -t- =(» + »») (1-2^) («')"+""-', 
and -^ =(m + »?) {n + m - 1) (1 - 2tf («')"+'"-^- 2 (?< +«?) {tt'f *''-'. 

Hence tt' .-^ -{n + m—\){l-2f) -r.-\-2{n-\-m) .v-0, 
and by successive differentiations the following equations arise: 
(«■').^-(« + '«-2)(l-20.^+2(2« + 2«^-l)^=0, 

(«')f^-(» + ^«-3)(l-20.^+2(3« + 3»e-3)g=0. 



and the law of the successive formation of these equations being very 
simple, we have generally 


Put k = n, hence 

Transpose n(n + 2m + l)-j—, and multiply by {tt')-", hence 

from which it follows that M = (it')"" .d". ' satisfies the equation of 

Art. 16. 

Again put «j' = («')"'"*'"^^ or tt' -j- +(n + m + l)(l-2t)v=0, and by 

successive integrations we obtain 

tt' . v' + (n + m) (1 -2t) ftv' + 2 (n -\-m) ft'v =0, 
tt' . ftv' + {n + m-l) (1 -2t) . ft' v -\-2(2n + 2m- 1) J^'v; = 0, 
and generally * 

tt' ft*-'v' + {n + m-k + l){l-2t) ft'v' + 2h{n + m) - ^ '^^~^^ \ . ft"-' v==0. 

Put k = n, hence 

tt' fr' V + (m + 1) {1 -2t) . ft" v' + 91 (n + 2m + 1) . ft"*' v' = ; 

from which it appears that m = jJ" +*(»') is also a particular solution, and 
therefore the complete solution of the general equation is 


Laplace's equation occurs when we put m = 0, and therefore 

the first term alone of which is the type of Laplace's functions, the 
equation is therefore more general than the functions it was used to 

The term ^"+' («')-<"+"'+') gives n + 1 constants of integration which 

enter as coefficients of the appendage which is a rational function of 

n dimensions, but this must be rejected, since the constants must be 

determined so that the rational function of n dimensions may satisfy 

the given equation, and this only identifies the appendage with the 

d' ift'Y^"' 
other term in u, viz. aitf)"" — , / — . 

17. To find explicitly the omitted part of the complete integral in 
Laplace's equation. 

The general equation of Art. 16. becomes in this instance 

and the complete solution is 

u = a^^^^+bfr'{tt')-^'-'\ 

the first term being Laplace's function, and the second the transcendant, 
it is required to find explicitly. 

Let a, /3 be any arbitrary quantities, then we have 

dar\t-a' fi-a) ~ ^-a da^\t-a)^'^da\fi-a] da'-'\t-al 

n{n-l) d^ / 1 X d'-^ / 1 N 

^ 1.2 Ma'[(i-a) da"''[t-a)' ^' 




^-/ 1 I 

+ 1 I 

_ \«-a)(/3-a)| ^ 1 r 1 W + 

, (n + mn + 2) 1 L_ + &cl 

(« + !)(« + 2) 

Commuting in this equation the quantities a and /3, we have 

(» + !)(» + 2) 1 1 

,., '^"{(f-/3)(a-/3)} _ 1 r . ^ ^ + 11 1 

(» + l)U + 2) 1 1 1 

1.2 ■ {a-(iy{t^(iy^ ^^■j 

If both equations be added observing that 

1 1 


(#-a)(/3-a) ^ (/-/3)(«-/3) {t-a){^-t)' 

the sum of the left-hand members 

fpn £ 


rf".-^ ef-. ^ 

^_1)„^ ^-« /3-^ 

1.2.3...wrfa"' 1 .2.3...W6?/3" 
1 1 

Vol. V. Part III. Yy 


Hence, we get the general identity, 

1 1 

K^-«)(/3-^)l"^' (/3-ar' 

i 1 ,^ + 11 1 (n + l){n + 2) 1 _JL_&cl 

(^-ar' "^ 1 -fi-a-it-ay^ 1.2 •(/?- a)^ ' (#-«)'- '[ 

^_^ ^^±1 ^^ 1 (« + l)(>^ + 2) 1 _J__£,c 

■^(/S-O'*""^ 1 •/3-a-(/3-0" 1.2 •(/3-af-(/3-#)'-' •] 

Put now a = 0, /3 = 1, and therefore (i - t = f, hence, 
>+' "^ 1 •/" "^ 1.2 ■^"-' 

[+F^"*" 1 V""^ 172 •^-^ + *'c.j 

in which identity n must be one of the natural numbers 0, 1, 2, 3, &c. 
and the number of terms in each series must be limited to w-f 1. 

Suppose the (ra + 1)* successive integral of each term of this expansion 

is taken after multiplying, for convenience, by 1.2.3 n, the result 

will consist, 

1st, of a logarithmic part, viz. . 

(-,)-.h.i.w{i-f.^.^.^^. <''";'.'r' '''-M 

where the part between brackets in the upper line is equivalent to 
the function P,„ and in the lower to (-1)".P„, and therefore the whole 

to (-l)".P„.h.l. ^,. 

* This method is applicable in every case to the decomposition of fractions, the denomi- 
nators of which contain equal factors. 


2d, of a rational and entire function />„ which satisfies the equation, 


since the term 2 ( - 1)". -^ is the result which arises if the logarithmic 

term ( — 1 )" P„ . h. 1. -> be put for u in the actual equation. 

3d, of an appendage containing n + 1 arbitrary constants, which as 
before remarked must be rejected altogether. 

Differentiating the equation for p„ above obtained, we get 

(«',^- + Mi-20.^-+>-i)(«+^)# + 2(-ir.^--o,. 

(«-l)(l-20.^" + 2(2«-l)^^" + 2(-l)»^^=0, 

• : ■ ^" df-^ + ^^^> ~dF~^' 

when these equations terminate, since j9„ is of « — 1 dimensions. 

Put ^ = 0, in all these equations beginning with the last, observing 
that then 

^ = (-l)-.(« + l)(» + 2)...(2«), 
^^ = - ( - 1)" . « (« + 1) (« + 2)...(2« - 1), 

'^=^-^)"-^^i^^-^'*''^)^'' + ^) (2«-2),&c. 

Y Y2 


Hence -^^ = - 2. (« + !)(» + 2)... (2»- 1), 

^^" = (m + 1)(« + 2)...(2m-1), &c. 
and the value of j9„ is the rational function 

^ 1.2...{n-l) ^^ +^'^ +A,f ...+A„^,], 

in which the coefficients are successively formed from the equation 

{n-m-lf.A„ + {m + 2){2n-m-l).A„^i 

+ 2(-ir "i^-'^)-("^ + ^) n{n-l)...(n-m-l) _ 
' '1.2...{n-m-l)' 2n{2n-l)...{2n-m) 

and the omitted part in the integral of the proposed equation is 

6|p„h.l. (I) + (-l)».^. 

18. When m = —^, the general equation of Art. 16. becomes 

and putting ^ = cos"^(l — 2#), we have Q„ = cosn<p, §'_,„+,) = sin ncp, the 
complete solution is therefore M = a cos«^ + 6 sin «^. 

Though the trigonometrical functions were the first used in analysis 
as reciprocals, for the purposes of expressing functions by means of 
definite integrals and of expanding them, in the former instance of 
their application there remain a few remarkable cases which do not 
seem to have been noticed, with which we shall conclude this Section. 

19. The two functions which possess the remarkable properties al- 
luded to, are 

e = e^'<«» . COS {x sin &), and 6' = e^ ""^^ sin {x sin B). 


The successive differential coefficients with respect to x of the func- 
tions 0, 6' follow simple and elegant laws, thus 

do dQ' 

= 6^'=°'* cos fa: sin + 0}, ^— = e^'=»*« sin {a; sin + 0}, 

d'Q d^Q' 

d^ = 6"^"°'' cos {ar sin + 20}, ^^= e^'^"** sin {« sin + 20}, 

and generally 

d" d° 0' 

• ^-; = e^'°'^ cos {x sin + «0}, -j— = e^ ">"* sin {x sin + «0} . ■ 

Again, the successive integrals relative to x, follow the same laws, 
omitting the arbitrary constants of integration, 

/^0 = e^cose cos {a; sin 0-0}, /,©' = e^<=°'» sin {« sin 0-0}, 

//e = e^cose cos 1^ sin - 20}, f,'Q' = e^'^"^* sin {x sin 0-20}, 

fj-Q = e^'^"'^ COS far sin0-w0}, f/O' = 6^<=<«« sin {a;sin0-«0}, 

for it will readily be seen by actual differentiation that 

d" d" 

= ^-;; {e^'="'''cos(xsin0-«0)}, 0' = T-^ i^icose sin (a; sin0- m0)}. 

Again, changing the forms of the proposed fimctions, we get 

= 1 {e-'^ + 6"-'^^}, 0' = -4== {e"'^^ - e"''"^'}, 

whence, expanding and passing from the exponential to trigonometrical 

a;* of 

= 1 + a; cos + - — - . cos 20 + , . ^, cos 30 + &c. 
1.2 1.2.3 

0' = a; sin + — — . sin 20 + , ^ „ sin 30 + &c. 

1.2 1.2.3 


jeW cos wy — -. j— g— g ^ the limits of being and ir, these formulee 

\ apply for all integer values of n, except 

Now e cos nO ± e' sin nO = e^^s* cos {x sin + «0|. 

Hence /ee"""^ cos {arsin^- w0| =7r . — — -— , 

l^^xcose cos {arsin0 + wej =0. 

The particular case where w = is included in the first of these two 

20. By the results thus obtained, we are enabled to represent any 
rational and integer function of a; in a form adapted to general differen- 

By applying Maclaurin's theorem, we first have 

(}>(x) = Ao + A,.x +^2- j^ + ^3 - ^ g 3 + &c.; 

and passing to definite integrals by the formulae of the last article, 

(h{x) = - /ge^cose 1^^ cos (x sin 9) + A^ cos {x sin 6-9) 

+ ^2 cos (.r sin - 20) + &c. J 
also if A^u A^i, A_3, &c. represent arbitrary constants, 

= - /ee^'=°^* {A-i cos(x sin + 0) + ^_2 cos (ar sin + 20) 

+ ^_3 cos (a; sin + 30) + &c.} 

both of which integrals must be added before <p (x) can be subjected in 
a complete form to general differentiation. 

We then obtain the w*'' differential coefficient by adding n9 under 
each cosine in this sum, that is. 


'-pp- = i /ee'-^o'S {Ao cos (x sin + n9) + A, cos [a; sin d + {n-l)e'\ 
(toe '^ 

+ ^a cos [ic sin +(» — 2)0] + &c.} 
+ _|ge'^cose |^_jCos[xsine + (re + l)0] + ^_,cos[xsin0 + (w + 2)0] 

+ ^_3 cos [a; sin + (w + 3) 6*] + &c4 . 

I. When n «'* a positive integer, the whole of the second line 
vanishes, there will then be no arbitrary constants; also, the first n 
terms of the upper line disappear. 

II. When n is a negative' integer, the first n terms of the second 
line remain, and these contain n arbitrary constants. 

III. When n is jractional, the whole of the second line remains, 
giving an infinite number of constants. 

21. The theory of numbers as connected with definite integrals, 
afibrds another remarkable application of reciprocal functions. 

Let n be any integer of which the divisors are n, Ji', n" 1; also 

let m be any intger, and d an arc of which the limits are 0, tt. 

Then, generally, 

1 -2Acos»?0 + A'' = (l - A6'»e^^)(l- Ae-"*^^); 
and hence, 
h. 1. (1 — 2 A cos ra + A") = - 2 { A cos m + ^ ^' cos 2 »J + ^ A^ cos 3 /w + &c. I . 

Suppose now that m is one of the numbers n, n', n" 1; this 

series must contain one term involving cos»0, viz. 

— A^cos w0: 

and therefore. 

Tit — 

j^cosw^h.l. (1 — 2Acos»»0 + A^) = — TT. — . A"". 

But when m is not a divisor of n, there will be no term in the 
expansion found to contain the arc n9, and therefore, 
^cos«0h.l. (1 -2ACOSJW0 + A^) = 0. 


Put now for m successively every integer from 1 to w inclusive, and 
take the sum of all the definite integrals thus resulting, hence 

/ecos»0h.l. {(l-2Acos0 + A^)(l-2Acos2e + A^)...(l-2Acos«0 + A*)} 

\n , w' -^, w" -4, 1,1 

= - ttX- .h + - .h" + — . A" + ...- . h"). 
\n n n » J 

Now the quantities -, — , — , &c. are the reciprocals of all the 

Tt ft Ti' 

possible divisors of n, and therefore this definite integral may also be 
expressed by 

"^ -^{k + -,h''' + \h"" + ...-h"}. 

'^ n' n n ' 

For 9 in the preceding equation write 20, the limits of the latter 

variable will be and - . 


Also put h = 1, and therefore, 

1 -2hcosd + h' = 2{l-cos2(p) = 4!sm^(f), 
1 - 2A cos 20 + ^2 = 2 (1 - cos 40) = 4 sin' 20, 
&c. ; 
.-. h.l. {{l-2heos9 + h') {I -2hcos2e + h'')...{l - 2hcosne + h')} 
= 2w h. 1, (2) + 2 h. 1. {sin sin 20. ..sin w0}. 

The integral of the constant multiplied by cos2«0 vanishes, and therefore 
7^ h.l. {sin sin 20 sin 30. .. sin »0} , cos2»0 = — t|~ + — +— , +...4-ll; 
and multiplying both sides by , we get this theorem. 

The sum of all the divisors of a given number n, including the 
number itself and unity, is expressed by the definite integral 


/^h. 1. {sin sin 20 sin 30. .. sin w0} . cos2w0. 


On Transient Functions. 

22. Let ^ (h, t) be such that when h has a particular value as- 
signed, the whole function vanishes whatever may be the value of t, 
except in one case ; (/^, t) under those circumstances, is a transient 
function having only a momentary existence. 

Thus the function _ , (^ —0.t\l.hH^ ' ^^*^" ^' ^^ P"^ equal to 
unity is a transient function, because its value is zero in every case 
except when t = 0, for then it becomes t- — j-^ when h is put equal 
to 1, that is, it acquires momentarily an infinite value. 

If the value of the function had been always zero, its definite inte- 
gral relative to t would also be zero; but if we actually integrate from 
^ = to t = \ without previously assigning a particular value to h, the 
definite integral 

2A \\-h \+h\~ ' 

thus this integral is independent of h, and therefore remains the same 
when h=\, that is, for the transient function. 

By the principles of the Second Memoir we can always form a 
self-reciprocal function in which the general term may be of any par- 
ticular kind ; thus if f{t, n) were the type of the general term, and 
if we put generally, 

Fit, n) = a,f{t, 0)+a,f{t, l) + a^f{t, 2)+ + a„fit, n), 

lastly, if we determine the coefficients a,, «2, a„ in terms of «„ and n, 

by the n equations (arising from the definite integrals) following, 

Vol. V. Part III. Zz 


!,F(t, n) ./{t, 1) = 0, 

SF{t,n).f(t,n-l) = 0; 
then the function F(t, n) will obviously be self-reciprocal. 

But if f{t, n) not containing arbitrary coefficients, but being abso- 
lutely given as P, (cos^% &c. is proposed as a function to which some 
unknown function is reciprocal, the discovery of the latter, which is 
effected in the next article, is of a more difficult nature than the pro- 
cess above mentioned; and in the particular cases quoted, as well as in 
many others, this required function is transient, it is therefore in this 
character that transient functions are here introduced. 

23. Given f (t, n) a Junction of known form with respect to the vari- 
able t and the integer n, it is required to find another Junction of t and 
n, as ^ (t, n), such that the definite integral jjf (t, n) ^(t, n') may always 
vanish when the integers n and n' are unequal. 

Begin with forming a self -reciprocal function F{t, n), the general 
term of which may be of the given form J{t, n) ; thus 

F{t, n) = a,f{t, 0) + a,f{t, l)+a,f{t, 2)+ +a„f{t, n), 

where the coefficients are determined in the manner indicated in the 
preceding article. 

Suppose next that the required function {t, n) is expanded in an 
infinite series of which the general term is of the form F (t, n), thus 

<p{t, n):=Ao.F(t,0)+A,F{t,l) + ...+A„F{t,n) + A„^,F{t, (n + l)}, &c. 

Multiply by f(t, 0), f(t, 1), f{t, 2) f(t, n - 1) successively, and 

integrate the products between the given limits of t, observing that 

f,F{t, 1) .fit, 0) = 0, f,F(t, 2) .fit, 0) = 0. ..J,F{t, n)f{t, 0) = 0, 

by the property of the functions F {t, n) ; 


and similarly, 

jlFit, 2) fit, 1) = 0, f^Fit, 3)/{t, l) = 0...f,F{t, n) .f{t, 1) = 0, 
&c. &c., 
we thus obtain the following equations ; 

j; {t, n) .fit, 0) = A, j,f{t, 0) . F (f, 0), 
^,<^{t, n) .fit, 1) = A,S,f{t, 1) . F{t, 1), 

^0 (A n) .fit, w - 1) = A._,!,f{t, n) . Fit, n-l); 

hence the imposed condition of reciprocity requires that the first n co- 
efficients Ao, Ai...A„-i in the expansion of 0(#, w), may be each equal 
to zero ; and therefore, 

0(^, n)=A„F{t, n) + A„^,F{t, n + 1) +A„+,F{t, n + 2), &c. ad inf. 

Multiply successively both sides by f{t, n + 1), f{t, n + 2), &;c., and 
integrate; and since n + \, n + 2, &c. are each > n, the definite integrals 
must vanish. 


AJ,F{t, n) .fit, n + l) + A„^J,Fit, n + 1) .fit, m + 1) = 0, 

AJtFit, n) .fit, » + 2) + An^,^,Fit, w + 1) .fit, n + 2) 

+ A„^2ftFit,n + 2).fit,n + 2) = 0, 
&c. &c., 

from whence the coefficients An+i, ^„+2, &c. are known in terms of A^ 
and ti, and therefore the required function ^ it, n) is known. 

24. To find the function which is reciprocal to t°. 

First, we must form a self-reciprocal function, of which the general 
term is of the form /"; this has been already effected in Section 
IV., namely, 

:,. n n + 1 ^ , nin-\) in + !)(» + 2) ^ 
^,,-1- j.-y- t+ ^^ . j-^ .t -&C.. 

z z2 


which has been also proved to be the coefficient of A", in the ex- 
pansion of {l — 2h{l — 2f) + h^\~^, (Section IV. Art. 9), and to be equal 

cl" (tt'Y 
*^ 1 — oQ — ~Tf^' where t' = \ — t, (Section iv. Art. 2.) 

Then representing by V„ the required function which is reciprocal 
to f, we have by the preceding article 

where it is obvious that when n' is less than w fiVj"' = 0, and it is 
only necessary that the coefficients may be so determined, that the 
same equation may remain true when n is greater than n ; and since 
one of these coefficients is arbitrary, we may put ^„ = 1. 

Now in general, we have by Section iv. Art. 2. 

x{x~l) {x — 2)...{x-n + \) 

f,Pj'' = {-iy. 

(a; + 1) (a: + 2) (a; + 3)...(;r + w + 1) ' 

hence, i F„ #" +^ = ( - 1 )" { 7 ^^ zr~r ^^-h-r^ — v^ 

■' ^ ^ \(w + a; + 1) (w + a; + 2)...(2« + ar + l) 

_. {n-¥x) {n+x-\)...x . {n+x) {n+x~\)...{x—l) . 1 

~ "*'■ (w+ar+l)(ra+ar+2)...(2w+a;+2)^ '"'"{n+x+\){n+x+9)...{'in+x+S)~ ] 

Therefore, when x is any integer from 1 to x , we must have 

A ^ A X{X — Y) 

2w + X + 2 "■" (2» + a; + 2) (2 w + j; + 3) 

. x{x-\) (;r-2) 

~ "*" (2» + x + 2) (2w + a; + 3) (2w + ar + 4) "^ ' 

and putting for x the successive integrals 1, 2, 3, &c. 

= l-^„+i. 

2m + 3' 

« , ^ 2 . 2.1 

0=1— .4„+i. X—-—: -r-4„+2 

2« + 4 "+'■ (2w + 4)(2m + 5)' 

^ , . 3 . 3.2 . 3.2.1 

2w+5 ^"""+^- (2w+5) (2w+6) "+" (2«+5) (2«+6) (2m+7)* 

&c. &c. 


From whence we obtain 

A„^, = 2n + 3, A„^,= ^ ^ .(2n + 5), Jn+3=~ fgg '.(2n + 7), 

and to prove that this law of formation is general, we may observe 
that since 

/ iy+2^+i_ 2n + 2x+l 1 (2w + 2a; + 1) (2n + 2x)...{x + 1) 

V~ h) ~ 1 ' h^ "^ 1.2...(2w + ^ + l) 

iy+^-^' I X 1 x.jx-l) i,^\ 

^ \ hi '\ 2n + x + 2' h {2n + x + 2){2n + x + S)' K" ]' 

Qfi 4- 2 
and (l-hy^'-^'Hl + h) = l + {2n+.3) . h + . {2n + 5).k' 

(2« + 2) (2w + 3) ,„ „. ,3 J 
+ ^ o 9 • ^^^ + 7).h^ + &c. 

Multiply both, and take the coefficient of ,^„^^^, in the products, 
and we get 

{2n + 2x + ]) {2n + 2x),..{x + 1) ^ x , 

\.2.3...{2n+x + \) * ~ 2n + x + 2'^ ' 

x{x-l) (2w + 2)(2w + 5) 

■*"(2w + a; + 2) (2« + a; + 3)' 1.2 ' ^^ 

= coefficient of -1- in (-IV (l+^Xl-^r'' 

= (-1)'. coefficient of A' in (1 +^) (1-^)^'-. 

Now the coefficient of h" in (1 +^) (1 -Af""', is evidently the sum 
of the coefficients of h'~\ and of h\ in the expansion of (1 — A)*'-'; 
that is, the sum of the coefficients of the two middle terms in a 
binomial raised to an odd power, and with alternate signs of + and 


— , hence the quantity we are considering must be zero, and there- 

2w + 3 X {2n + 2) (2w + 5) x(x—l) 

~ i '2ra + a; + 2'*' 1.2 ' (2n + x + 2) {2n + x + 3) 

_ (2» + 2) (2?^ + 3) i2n + 7) x{x-l) {x-2) „ 

~ 1.2.3 ' {2n + x + 2){2n + x + 3){2n + x + 4>) "' 

which shews the generality of the observed law of the coefficients ^„+i, 

Substituting now these values in the general formula for V,„ we 
get the required function which is reciprocal to t", namely, 

rr- r. /„ „v T, (2« + 2)(2w + 5) „ 
K = P„+{2n + 3).F„^, + ^ ^ ^ . P,+8 

(2w + 2) (2w + 3) (2w + 7) „ „ 
"- 17273 •^"*^' *'''• 

25. The Junction which is reciprocal to t" is transient. 
For in general 

d".(tt'Y _ d^itt'T 

\.2.3...ndt'' ^ ^ ' 1.2.3. ..ndt'"' 

and putting 1-t' for t, and expanding the binomial (l-t')", and 
lastly actually performing the differentiations indicated, we have 

^ A) .^n-A 11*^1.2 1.2 

and therefore 

V-l) ''n-JA ii'^+i2 1.2 * 

-(2w + 3)|i T" • ~T~ ~~T72~~ • 1T2 ' 

(2«+2) (2W+5) ,, »+2 n+3 - (w+2)_(^+l) (w+3) (w+4) ,,2_^ > 
+ 172 t^~ 11 1.2 • 1.2 ■ "^ 

-&c. &c. 


The term which is independent of t' is 

but in general we have 

(l-A)(l+A)-(^"+^' = l-(2« + 3).A + (^" + ^^)(|^ + 5) ^._^^ 

and putting h = \, we find that the term independent of if is zero. 
Again, multiplying the last equation by A"+'", we get 
(*"+"■ - A"*""*') (1 +^)-<'"+'* = A"+'» - (2 w + 3) . A"-^""*' + (^"+yv^"+^) . ^.+".. >^ _ &c. 
Now it is easily seen that when h = 1, we have 

-Tj^ (A"+'"-A"+"'+') = (w+w)(«+»?-l)...(«-/»+l)-(«+/» + l)(«+»»)...(«-w+2) 
= — 2»w . (w + >w) (w + TW — l)....(w — m + 2), 

-(A"+'"-A"+'"+')= - (2m-l)(«+m) («+m-l)....(«-»w+3). 

&c. &e. 

and therefore when A is put =1 after differentiation, we have 


_^^ {(A"+'»-A"+'"+>)(l+A)-'"+'} = -2-<""+''.2>«. («+»«) (w+»«-l)...(w-M + 2)x 

, , 2w+2 2OT-1 1 (2w + 2)(2w + 3) (2m - 1) (2?»-2) . . 

* ^* i ■«-»^^-2"^2^■ 1.2 ■(w-w + 2)(»-»» + 3) *^"» 

which series consists of only Im terms, and is equal to the infinite 
series obtained by differentiating the other side of the equation, viz. 

w(« — !)....(« -OT+l)x (w + l)(« + 2)....(w + »«) 

- (2«+3. («+!)». ...(«-»» + 2) X (w + 2)(» + 3)....(» + »« + l) 

+ (^^+^)(^" + 5) (n + 2)(w + l)...■(w-»^+3)x(w+3)(w + 4)....(w + »? + 2) 

— &c. oe? infinitum. 


Now it is obvious by putting m = 1, 2, &c. successively, that the 
finite series is always =0, and therefore the infinite series [which is 

( — t'Y 
the same as the coefficient of ^ ^ — ^ in the expression for ( — 1)°^„] 

vanishes also, so that if V„ be arranged according to the powers of t', 
it is + 0. ^' +0^'^ + &c,, nevertheless its value is in one instance infinite, 
namely, when t = 0, for then P„ = P„+i = &c. = 1, and therefore 

F„ = l + (2. + 8) + ("^ + f^f-^^^^^" + ^)^fV")(^^ + ^) + &c. 

= (l+A)(]-A)-«''+^ when h is put =1. 

= X . 

And if V„ did not possess this infinite element ft Vj", from i> = 
to t = 1 would vanish, whereas its actual value is the same as 

26. To express the transient Junction Vn in a finite form . 
Since by Art. (24.) K = P, + (2w + 3)P„+, 

, (2w + 2)(2w + 5) (2« + 2)(2w + 3)(2w+7) « „ 
"• j~^2 ■ '^^ 2 2 3 • "..+3, «c. 


1.2.3...(2w+l) r'„ = l .2. 3.. .2?? X (2« + l)P„ 

+ 2.3...(2w + l) X (2w + 3)P„+,./f + 3.4...(2w + 2) x (2« + 5)P,+s^\ &c. 
when A is put equal to unity. 

But in general, 

{1 - 2A (1 - 20 + K"] -^ = P„ + P,h + P^A^ + ...P„A" + P„+, A"*' + &c. ; 

= 1.2.3...2?«.P„ + 2.3,.,(2w4-l).P„+, A + 3.4...(2« + 2).P„+2AH&c. 


Multiply by 2A"+i, and diiFerentiating once more, we get 

^Thr djf^ 1 

= 1.2.3...2wx(2« + l)P„A'-i + 2.3...(2w + l)x(2» + 3)P„+,A"+J+&c. 

Hence, F„ = ^h-^^ . j. [jf^l ^^ j^ I "^ ' " "^> -^,f ^ . 

c?A\ 1.2.3...(2« + 1).</A^° j' 

when A is put = 1. 

Put for abridgment the radical {1 — 2A (1 - 2#) -f A^}-J = ^, then 

rf^ . {Rh") _ 2«.(2w-l)...(w + l) ^^B 

rfA^ ~ 1.2...« • • </A'' 

. 2w(2w- l)...w ^ ^ ,c?"+'J? . 
^ ' ' ' \c?A" W + 1 1 C?>&»+1 

w(w-l) _A^ c?»+^B 1 

(« + l)(« + 2)'1.2'rfA»*'" *''^7 

Whence 2 ~ {^-i ^^^^| = 2» (2«- !)...(« + 1) {(2« + 1) A«-4 ^ 

2w + 3 n ,^^ d'^'R 2n + 5 h"*^ .n .{n-l) d'^^R 
■^ 1 •« + ! • dh"^' "^ 1 . 2 * (» + l) (w + 2) •• d¥^ ■•■ *'*'• 

c?A"+' w + 1 c?A"+- J 

u 100 Tr d'R . 2w + 3 « .d^^'R 
Hence 1 .2.3.. .wF^, = -jT- + r — -^. .^ „ ,, 

2w + 5 w (w-1) ^^ rf»^^jB 

"'' 2» + 1 ■ (» + 1) (ra + 2) ■ 1 . 2 • rfF^ "*■ *'''• 

2 ^^ c?"^^Jg 2 n d'^'R 

2« + 1 * t/A"+' "^ 2w + l» + l ■ </A"+^ 

2 n(n-l) h^ d'^^R 

2w + l*(w + l)(« + 2)*1.2' <^A»^' ■•" *'^- 
Vol. V. Part HI. 3 A 


h being put = 1, after the differentiations; this value of 1.2,..?iF^, is 
expressed in two finite series, each containing only w + 1 terms. 

If we actually add the terms in this formula, which contain the 
same powers of A, we get 

V - 1 K:? w + 2 h d"^^R (n + 3) 

"~ 1 . 2...W \dh" "^ w + 1 ■ 1 ' dh"^' "^ (w + 1) (« 

, n 

h' d'^'R 


n + 2)' 1.2" c?A"+' 

(« + 4).ra(w-l) k" d'+^E 

^ SL 1 

{n + l){n + 2){n + 3)'l.2.3' dh 
when h is put equal to unity. 

27. -Discussion of the transient function N ^. 

Put « = in the general expression for ^„ in the preceding article, 


hence V. = R -^ 2h -y^- 


= U-2A(l-20+A^}-^ + 2A(l-2jf-A) {l-2A(l-2^) + A^}-i 
(1-A)(1+A) , , . 

This function, as has been observed in Section vii. (22), is in general 
zero, except in the particular case when ^ = 0, when its value is infinite. 

If we imagine a curve of which the equation is 

y {l-2A(l-2a;) + ^^}4' 

where h is less than unity but nearly equal to it, the limiting values of 
y as A approaches unity, will give the geometrical interpretation of 
the transient function V^. 



Take (Fig. 1.) AB = 1, AH = 1 - h, or 
BH = h both along the axis of x, and make 
A the origin, then putting x = 0, we have 

1 -i- h 

y — jz — T7^, which is very great, and tends to be 

infinite as h approaches unity, and is represented 
by AC\ next putting x = l — h = AH, we get the 

corresponding ordinate HE = (jZThf) ' (T+W' 
which also tends to infinity ; lastly, putting x = l 

we have y = 


= BD, which tends to vanish 

(1 + h)i 
in the ultimate case representing V^. 

Now varying the parameter h so as to make 
it approach unity, the points C and £1 recede 
indefinitely from the axis of x, and the point 
7> approaches it indefinitely. 

Yet the area DBACE remains constant (for 
the integral between x = 0, and x = \ of 

{l-2^(l-2x) + A''}J 
relative to x is evidently unity). 

And the altitude GN of the centre of gravity of this area is also 
constant, for 



and therefore is the same as that of the parallelogram HF, when 
AF=^AB, for the distance Gg from the axis of y 

x{\-h){\^h) \~h AH 

{l-^h(X-2x) + ¥\l 2 2' 

Hence G tends ultimately to the point g in the axis of y, which 
shews that the area DBH'E' tends absolutely to vanish, HE' being 

3 A 2 



an ordinate drawn near the origin at any small distance not varying 

with the parameter Ji, and since -r- has the same sign in the interval 

from B to H', H or A, it is evident that the portion of the curve 
BE' tends to coincide with the axis BH', the curve therefore which 
represents V„ coincides with AB, except infinitely near the origin 
A, when it suddenly mounts to an infinite height. 

Since the general function V„ is reciprocal to t", it follows that 
fi Vaf = 0, except when w = 0, and then the definite integral is unity ; 
hence if f{t) be any function containing only the positive and integer 
powers of t, the transient function Vo possesses the remarkable pro- 
perty expressed by the equation [tV'o ■/{f)=J^{Q). 

Fig. 2. Let 2a = AB, equal the 
length of the axis in a solid of revo- 
lution, the surface of which is covered 
with an indefinitely thin stratum of 
fluid, let any abscissa ON measured 
from the centre O be put equal to 
a (1 - 2#), the limits of t will evidently 
be O and 1 . 

Let the law of density or accumulation at any point P of a section 
perpendicular to the axis be expressed by the transient function \V^, 
X being constant, and let the total action of the fluid on any point Q 
in the axis be required, the law of force being capable of expansion 
according to the positive and integer powers of t. 

Put PA'' = y, then the whole quantity E of fluid is manifestly 

equal to ^Xtt ft T^^y -r, , s representing the arc AP. 

Now it is easily seen that the value of y-yr at the point A where 

y vanishes is iaR, R being the radius of curvature at that point, and 
by the nature of V^ this quantity is the value of the above integral, 
or E = SXttuR. 


Again, if we represent the distance PQ by r, and the law of force 
by y(/-) and put AQ = k the initial value of r, the total action is 

r rr ds „, ^ ON 


which by the property of F"o is equal to S\traRf{k), or to E .f{k). 

Let us now suppose an equal quantity of fluid, but of a contrary 
nature in its action, and therefore represented by — E to be collected 
in a single point C in the axis produced to a small distance AC- a. 

The total action of the compound system on Q will then be 

E{f(,k)-f(k + a)}, 

which tends to vanish as C approaches A. 

Lastly, suppose a unit of fluid when distributed over the surface 
according to a law expressed by {t), which depends on the figure of 
the solid, will exert no action on any point Q in the axis; then if 
the law of distribution of the fluid be expressed by X V^ + c (p {t), the 
total action on Q including that of C, will be still E {f{k) - f{k + a) ^ . 

From which it follows that when an electrical spark -Eh in- 
finitely near to the vertex of a conducting solid of revolution charged 
with a quantity of electricity E', the distribution of the latter under 
the influence of the former is expressed by the law 


\Va + c<i>{t) where \ — ^ ^, 


and where c is determined by the equation* 

Having thus given the geometrical and physical interpretations of 
Vo, it will not be necessary to discuss the transient functions V^, V^, 
&c., of which the properties are very analogous. 

* Vide First Memoir, Art. 35, the expression there obtained for a sphere being in- 
cluded in that obtained above, when the influencing point is infinitely near the sphere. 


28. To find the quantity to which V„ is the generating function. 

By Art. 26. 

_ d ^'iRh") ^hd''*^ (Rh") 

I .2.S.,.2ndh;"' 1 .2.3...(2» + l)rfA"^" 
where R= {l~2h(l — 2t) +h^}~K and h is ultimately equal to 1. 

Forming the equation u==h + &uK we have 

A^ d^f{h)h\ k^ d^f{h).m} 

hence {l-2«(l-2^) + «1-^ ■^^ = ^iZ + ^' ^^ ^ 1T2 <//^' +&c.; 



^l^^'-^^^^^' rS^F^ = '^' "°"^'^""' °^ ^"" ^" {l-2t.(l-20 + >/'M 
In like manner, 

«-i{l-2«*(l-2/) + «n-^.^ = 2?A-^+^'|f^ 

/fe^ d\{RM) ¥ d'jRh) 


_^M!!!1(^) = the coefficient of >P«- in ., , ^ f^^ .,, ; 
1 .2...(2« + l)rf^''+* {1-2m(1-20 + «'H 

(* + 2Am-^)^ 
consequently, V„ = the coefficient of ¥'^*' in v^ _ ^^(i _ 20 + «'}^ ' 


Now, by the assumed equation we have 

ui — hu~i = k, 

and by differentiation («J + Am"*) -jr = 2m ; 

but also (m^ + hu~^) = k + 2hu~i; 

hence, {k + 2hu~i)'Tr=2u; 

and therefore, F„ = 2 the coefficient of ^"+' in u {l-2u(l - 2t) + u"} -^ 
from which it follows that if we form the two equations, 

u'=h + (ku')i] . \U' =u' {l-2u' (l-2t) + u''}->^ 

} putting i 

u" = h + (ku")i] [t7" = «"{l-2«"a-20 + «"'i^^; 

then ^^j— = ^0 + J^^k + V^¥ +r,k'kc. ad inf. 

supposing that in the left-hand member h is finally put equal to unity. 
It may be observed that the quantities u', u" are the two roots of the 
equation u'^ — {2h + k)u' + h!' = 0. 

29. To expand a given ^function (pit), in terms of the transient 
function \ ^ . 

Let the general term of the expansion be A„V„, then by the nature 
of reciprocal functions we have 

= AJtPj", (Art. 24.) 

M.(W-l) 1 

= (-l)".^„ 

(w + !)(« + 2) (2« + l)' 



Examples : 

/» - -J_ V _ ^-3 V J. g-4-5 ^ . 

p_j^ 2«+3 „ (2w+4)(2«+5) (2m+5)(2« + 6)(2w+7) .^ , 

-rn—r„ J . ;'„+,+ I ^ . f'n+a J . f'n+g&C. 

the latter series would also result by reverting the series for V„, in 
Art. 24. 

30. To find a function U„ which shall he reciprocal to (h.l.ty. 

Following the steps indicated in Art. 23, we must first form a self- 
reciprocal function of which the general term is a constant multiplied 
by (h. 1. ty ; this has been already effected in Sect, v, namely, 

and then the form of the required function will be 

[/„ = t; + a 2;+ , + 6 1;^^ + c 7;+3 + &c. 

Multiply by (h. 1. ty, supposing m>n, and observing that 
j;r„(h.l.0" = 1.2.3...>^.(-ir. "-^'^-;)^:-f-("-'^-^^) bySect.v, 

and ir7„(h. 1. /)•" = 
by the nature of reciprocal functions, we get the general identity 

m{m — \){m — Q)...{m — n + l) , m — n , {m-n){m -n—\) 

® r.2.3...« •^^""•^TT+*- (« + i)(« + 2) — *'*'-^' 

but on the same supposition that m is greater than w, we also have 
= (l-l)'"-" = l-(m-w)+^^ 4p— ^-&c.; 


and by comparing the corresponding terms 

- ^ + 1 ^_ (« + l)(w + 2) _ {n + l)(w + 2)(w + 3) . 

rr-7--L'' + ^ T , (w +!)(« + 2) . (w + 1)(w + 2)(m + 3) 

Ly„ — J „ -I J . ^ „+i + r — . -I „+2 i r — - — . X „+3, Cue. 

31. To express the function Un which is reciprocal to (h. I. t)" in 
a finite form, and also the function which Un generates. 

l.2.3...nU„ = 1.2.3...nT„+2.3. 4...(w + l) T„+,+3 . 4 . 5...(w + 2) . T,.+,+&cc. 

=^ { r„ + r,A + 7;a^+ ... T^h'+T^^.h"^^ + &,c.], 

h being put equal to unity after the differentiation. 
But by Section v, we have 


J— ^ = T,+ T,h + T^h' + &c. ad inf. ; 


therefore, U„ = - — — — -jt when A = 1. 

1.2. 3...ndh" 

Now by Taylor's Theorem, this quantity is the coefficient of k" in 

the expansion of - — j~r^ the latter is therefore the function which 
U„ generates. 

32. Properties of Un- 

.1. jiUn (h. 1. ty = f,T„ (h. 1. ^)" = 1 . 2 . 3...W, by Sect. v. 

II. Changing the sign of k in the quantity which U„ generates, 
we get 

Vol. V. Paet III. sB 


pT'' = Uo- U,k + U,¥ - U^¥ + he. 

— f-f^§--} 

III. Since f =\ +x\i.\.t+ -^ . (h. 1. tf + j-^-j ■ (h. 1. <)' + ««:■ 

by means therefore of a single integral, x" may be adapted to general 

As this result is remarkable, we may confirm it by the general 
rule in the First Memoir. (Vide Sect, i.) Thus, 


put (x) = :j T— =l-\-xk + a^J^ + &c. 


then fit) = — -T =U,+ U,k+ U^k" &c. ; 

for all values of k, whence JiUnf — xf as before. 
33. Discussion of the Junction U^. 


By Art. 31. Uo = - — r when h is put eqifal to unity. Like the 

transient function Vo, discussed in Art. 27-, the quantity Uo is always 
zero for values of t between and 1 ; but when t = 1 its value is 
infinite, and thence its integral between the limits and 1 of / is finite, 
viz. unity. 



To prove this property, conceive a curve Y 
APC, of which the abscissa measured from 
A along AB is taken equal to /, and the 

corresponding ordinate y is equal to f''', 
and let us suppose h very nearly equal to 
unity, and at any point P draw a tangent 
PT; then since 



y = t'-\ 

therefore, 11^ is the limiting value of the tangent of the angle PTB. 

Take AB = 1 and the ordinate BC = 1, then it is evident that 
A and C are constantly points of the curve when the parameter h 
varies so as to approach unity. 

Again, for the entire area APCB the expression is Ji^'~*, from t=0 

1 — A 
to t=l, that is, - — Y, which evidently tends to vanish as the para- 

.« — fl 

meter k approaches unity ; and as no part of the area is negative, it 
follows that the curve APC tends ultimately to coincide with the two 
right lines AB, BC, and therefore when T is sensibly distant from 
B the tangent of the angle PTB tends to vanish, but when indefinitely 
near to B it tends to infinity ; and therefore Ug, which ultimately re- 
presents these tangents, is zero from A to indefinitely near to B where 
t is unity, when its value becomes infinite. 

In like manner the remaining functions C/i, U^, &c. may be dis- 
cussed with similar results. 

It may be observed that for values of t>l (which however do not 
enter the definite integral), the values of t/'^ are infinite. 

34. Expansion of given Junctions in terms of the functions Un . 
The general formula for this purpose is 

0(0 = V^kW) + UJt<p{t). 


^,<t>{f).ih.\.tf + 



.j;<^(0.(h.i. /)' + &c. 

3 b2 


T -u - ^±1 u + (^ + i)(^+^) r/ _ &c 

1 1.2 

which latter series is also produced by reverting to that which expresses 
C7„ in terms of 71 in Art. 30. 

35. To find a function reciprocal to t" when the limits of t are 0, 
and GO . 

Let M„ be the required function, and put t = e"', 

then ^lUj'^ = 0, from # = to ^ = x ; 

«» /V- 1 N™ ^ i>_ = to T = 1, 

therefore /"- (h. 1. t)-" = 0, from t = 

^» " 


hence m„ = t C/„ = t ; — ^ttt when h = \ 

1 . ^...ndh'' 

36. Tb ^«c? « function F„ ^t>A^cA *Aa?/ ie reciprocal to cos" ^, ^A^ 

- awa -. 

2 2 

ZmeV* o/*^ Je^?^ — - and - 

Following similar steps to those adopted in the preceding Articles 
we shall obtain, 

w + 2 
in cosines F„ = cos n<p — . cos {n + 2)(f> 

(« + l)(w + 4) , ,,,^ (w+l)(w + 2)(« + 6) , ,-,-, . 
+ —^-12 ^•cos(« + 4)0 -^ ^ g g ^cos(w + 6)<^, &c. 


in sines Fn = 2 sin (p {sin {n + l)(p — . sin {n + 3)(p 

+ ^^ -^ . sm (« + 5) <^ - &c. ] 

37. The Junction F„ is transient. 

Either of the preceding values of F„ give F„ = Fn—F", where 

F: = cos {n(t>) - ^±i . cos {n + 2).(p + ^^"^^^^^^^ • ^^^ (« + 4) <^ + &c. 
F„"= cos (w + 2) <^ - ^^ . cos (« + 4) . <^ + {n + l){n + 2) ^^^ („ ^ g^ ^ ^ ^^. 
passing from trigonometrical to exponential values, 

1 1.2' 

1 1.2" 

_ ("g.^vrr ^ g-</>vrT\-" 
= 2cos«0, 

2F„" = £("+2)*^^ - ""'"^ . e("+4)*v^ + (" + l)(^ + 2) _ ^(„^g)^^— _ ^^ 

1 1.2 

+ g-(n + 2),^V:rT _ ^jt2,e-(n + 4)0V^ ^ (w + 1) (W + 2) ^-(„ + 4)^vri _ j^^.^ 

= 2 COS «^, 
hence F„=-F';-F„"=0. 


However, if n be even, and our limits be — ^ and ^, the function 

becomes suddenly infinite at the limits, for the expansion of F„ is 
then identical with that of (1 -!)-'"+'•. 

38. To express infinite terms the transient function Fn. 

i?'„' = cos ra(^ - ^ . ^ cos (« + 2) + ^^i|^-^^ . A'' cos (« + 4) «/) - &c. 

F:'= hcos (n+2)(f> - ^ . A^ cos {n+4>).<l> + <"+!) (^ + ^) ^^ cos (w + 6)0-&c. 

1 1 • ^1 

Then F„ is the limit of F^—F" when h approaches unity. 

Put also 2 cos = a; + - , 

^ X 

hence 2-F„' 

x~^ X 

~ {l+A(a:^+a;-^) + *'}"+' 

W + 1 (W + X^ Tt W + 1 

COSW0+— j— .Acos(w-2)0+^— — ^^.A^cos(«-4)0...— — -.A"cosw0+A"+^cos(«+2)0 

~ fiTaFcosa^TFp^^ * 

the number of terms in the numerator being w + 2. 

In like manner, 
2F„" X 


h (aT-' + Aa;)"+i (x + Aa;-')"*' 

_ X (a; + Aa;-')"+' + a;-' (a;"' + Aa;)"+' 


eos{n+2)(l)+'^r— .h cosn(f)+- -^ .k^cosin-2)<p+ . . . —— .h" cos{n—2)(p+ h"*' cosn<p 

^ fTT2rcos2^1~Fp^' ' 


costKj) .{I -h"*^)+h\——cos{n-2)(p-cos{n+2)<p\+hH— — ~-cos(w-4)0 — — cosw0> 
^" {l + 2Acos20 + ^^}»+' ' ^' 

when h is put = 1. 

Thus F = (^-^)(^ + ^^ 
which is evidently a transient function, as its general value for A = 1 


is zero, except ^ is an odd multiple of -, when its value becomes 


And in general F„' and F„" are equal, when h is put equal to 
unity, and therefore F„ has a factor 1 — A in its numerator, which causes 


its general vanishing state, except when ^ = „, or an odd multiple of 

^, when the denominator becomes (1-A)^"'*"^ and as the numerator is 

of only n + 2 dimensions, it is evident F„ in this case is infinite, when 
k= I. 

In general f , ~i . r.^ 1: 2 = 2 tan"' . I^^^ . tan 0> + const., 

which taken from = to ^ = - is equal to tt, a quantity independent 

of h, a result similar to those already obtained from other transient 

39. When the sum of a series containing transient functions is 
required, the following process, with only such modifications as may 
simplify particular cases, will apply. 


Let S = UaVo + fli Fi . ss + (hV.i.%^ + ... + «x ^^K' + &c. 
be the series proposed. 

By the inverse method, put a^ = U/{t) . t' from t = to t = 1 . 

Then S = /^/(r) \V, + + V.t'z' + &c.| 

But V^+ V^k + VJe^, &;c. is the function which V„ generates, and may 
be represented by ^ {t, k), we have then 

S = X-/(t) . <p{t, T%), from T = to T = 1. 




On ike Resolution of Equations involvings Definite Integrals. 

(l) By the Decomposition of the Integrals into Elements. 

40. The utility of the method of decomposition consists principally 
in the verifications it offers to results obtained by other analytical pro- 
cesses, the difficulty in the eliminations which it requires. 

Pm-i-^ Pm+if Pa Put/ 

Suppose a cylindrical shell exerts no force on any point in its axis 
AB, the law of force tending to each particle of the shell being given, 
but the law of density of the shell unknown, then the application of 
the method of decomposition is this : 

Divide the shell into « + 1 equal portions by planes perpendicular 
to the axis PiQi, P2Q2, &c. 

Let the density throughout each portion be supposed uniform, and 
let the successive densities be pa, pi, p-> 

Let the total actions on the points of division Qj, Q2...Q„ be equated 
to zero, which will give n equations, and another will be obtained by 
considering the mass of the shell. 

From these n + 1 equations, let po, p^, p^, &cc. be determined in terms 
of ». .. 

Finally, make n infinite. 
Vol. V. Part III. 



41. General Calculus for the Cylinder with any law of force. 

Let «o, «i. a^.-Mn represent the total actions on the point A which 
would be exerted by the successive portions P^Pi, PiPz of the shell, 
if the density of each were unity; these quantities are given, since the 
law of force is supposed known. 

Then aopo, a^p^, a^p^, &c. represent the actual forces on A. 

Again, the action of any portion as P„+4P„+5 on any point Q„ of 
division in the axis, will be to the action of the similarly situated 
portion P^P^ on the point A in the ratio of the corresponding densi- 
ties, and in this case would be atpm+i. 

By this consideration the total actions on the points Qi, Q2...Q„ 
are easily estimated, and equating each to zero, we get the following 

system of n equations, which serve to determine the ratios — , — , &c. viz. 

pa Pa 

aopo- Ctopi — aipz — Chps — asPi — dn-lpn-l — (tn-iPn = 0, 

aipo + aopi — aop-i — aipi — aipi —ctn-spn-i — an-ip„ = 0, 

Oipo + aipi + ttopi- ttops—aipi — «n-4/'n-l — «n-3/'« = 0, 

a^po + a^pi + aip2 + aopa— Oopi —ctt,-ipn-i — (i„^ip„ = 0. 

a„-2po + an-3pl +«„_4p2 + «n-5/03 + «»-6P4 — «2/0«_l — fl!l/0» = 0, 

ffn-lpO + fin-2Pl + «»-3p2 + «»-4j03 + «n-5/04 +«l/On-l " aoPn = 0. 

Comparing the first equation with the »*'', the second with the (w-l)"", 
&c. it is obvious that po is involved in the same manner as /o„, p^ as 

Pn-l, &c. 

Hence, p„ = po, p„_i = pi, p„.i = p^, &c. 
Form now two functions in the following manner: 
a known function, M = «osin0 +«isin30 + a8sin50+ .,.«„_isin(2w-l)0, 
an unknown, aS'„ = jOoCOSw0+|OiCOs(w-2)0 + ^sCOs(«-4)0+ 


the first series may be continued to n terms or infinity indifferently, 
and the last term in the second series will be ^p^ when n is even, 


and p„_x . cos 9 when n is odd. 

Suppose now that the product 9,u.Sn is decomposed into the sines of 
the multiples of 9, and that all the multiples higher than the «'" are 
rejected from this product, the remaining part will evidently be, 

— {aopo—aopi — aip2 — a»-i/Oo} .sin(n — l)0, 

— {aipo + Uopi — aopi —a„.2po}.sm{n — 3)0, • 

— {(hpo + ctipi + aop2 —a„-3po}.sm{n — 5)$, &c. 

the whole of which by the given equations is equal to zero. 


2S.u = A„sm{n + l)9 + B„sin(n + 3).9 + C„sm{n + 5).B, &c. ; 
.-. 4 cos . S„u = A„ sin (nB) + {A„ + B„) sin in + 2)9 + {B„ + C) sin (« + 4)0, &c. 
and 2Sn.iU = A„^i sin {n9)+B„.i sin {n + 2).9 + C„_i sin (« + 4) . ; 

.-. 2{2cos9.Sn-j^S,.^} .u = i^A„ + B„-A„.^\.sm{n + 2) . 9, &c. 

Hence it follows that if we put So=po, S^ = po cos 9, 

and u = aoSin9 + a^ sin 3 + a^ sin 5 9 &c. ad inf., then. 

First, Supposing S^.^ and S,n known, form a quantity \„ by dividing 
the coefficient of sin(/» + l)0 in 2S,„u, by the coefficient of sin(/»0) 
in S/S'm.i .u. • 

Secondly, Form a quantity S^^^, by the equation 
-S'„+i = 2 COS0 . /y^ - X^iS*™.! , 
by which S^, S3 a^^ may be successively formed. 

Then it is obvious that the product 2S„u contains no multiple of 9 
below the «'\ and therefore the coefficients in S„ must be the required 

quantities po, p^, pi pn-j^ when n is odd, or p^, pi, pt ^p^ when 

2 3 

n is even. _ 



42. Applications, when the law of force is the inverse square of the 

(1) Let AB be the axis of a very broad cylindrical plate, the 
round side of which is covered with a fluid, attractive or repulsive, 
and so distributed as to exert no action on any point in the axis. 

Put AB = 1, APo = a the radius of the base. 

Let ab be one of the very small annuli into which the edge is 
divided, and put aPo = x. 

Then it is easy to prove that the action of the annulus a 5 on the 


point A is proportional to -^ -jr, or ultimately to the differential 

1 ■ X . 

coefficient of -7— with respect to x, that is, to -t-t? ryj, which quan- 

Aa {o'' + arj» 

tity expanded is proportional to a; — f rj + &c. ; and as b is very great 

compared with x, we need only take the first term of this expansion. 

In this case we may therefore put ao = l, a, = 2, ai = 3, &c., 
and therefore, M = sin0 + 2sin30 + 3sin50 + 4sin70 + &c. 

The calculus of S„ as indicated in the preceding article will be as 
follows : 

So = pay Si = |OoCOS&, 

f ^ coefficient of sin2g in ZS^u ^31 
\ ' "~ coefficient of sin0 in 2SoU ~ 2)' 

■ • Si = 2coseSi - x^So 

= Po Jcos20-^}; ■ • 


i _ coeffi cient of sin30 in 2SiU _2\ 
\ * ~ coefficient of sin29 in 2«S^m ~ 3 j 


= Po{cos39 — -cosO], .... 

J _ coefficient of sin 40 in 2S3U _ 5\ 
\ ' ~ coefficient of sin30 in ZSaU ""6/ 

*S. = 2cos9S3 - Xs.S^ . 

= po|cos40 — -cos 20 — ->, 

f coefficient of sin 50 in 2SiU _ 91 
\ '^ coefficient of sin4!0 in gAysM ~ lOj 

Si = 2cos9S^ - XiSi - 

= po |cos 50 - - cos 30 — -> . 

{2 2 11 

cos60 — ^cos40 — ^cos20 — ^> 

{2 2 2 1 

cos 70 — = COS 50 — - COS 30 — - cos 0> . 

Generally when n is an odd integer, suppose 
-^^ = cos(w-l)0-^— {cos(m-3)0 + cos(« - 5)0 + ... + cos20 + i}, 

and — = cos »0 {cos (w — 2) + cos (« — 4) + ... + cos 30 + cos 0}. 

po n - - 

The coefficient of sin w0 in 2»S'„_,« = -. p„, 

n — 1 ^ 

of sin(M+l)0 in 2S„u = .p^; 


therefore, x„ = <^±fi^ = 1 - ^ + ^ . 
n(n + l) n n + 1 


Hence, ^=2cos0.— -X^*^ 

pa pa po 

= cosln+l)9 -{cosin-l^e + cos(n-S)e + ...+cos29+i}, 

and by a repetition of the same process, 

O Q . ,■> .*,{ ,-; . •-:.. 

— ^ = cos(»+2)0 -{cbsw0 + cbs(«- 2) . + ... + cos30 + cos^}. 

pa n-TXt 

Hence the laws by which S^-x and S^ are expressed are uniform, 
and therefore we get for the required unknown quantities, 

2 2 2 _ 

po = Pa, pi=--pa> P8=~^Po P—i -P<>> Pn-Po- 

The positive values may be taken for the repulsive and the negative 
for the attractive parts of the fluid, and if E denote the excess of the 
former, we have 

[n n n n n] n [ n } n 

.-. po=»^-T — > which gives the complete solution of the problem. 

Thus the application of a process purely algebraical, conducts in this 
instance to a transient function, for if we suppose the final and equal 
densities po, p^ to be finite, all the intermediate values of the densities 
p^, pa pn-x become indefinitely small when n is made infinite; yet 

they are not to be rejected, for if so, the total charge would be 4nra^, 


whereas its actual value is only inra ~, an infinitesimal of the second 

as Its itutuiu vaiuc is fJiii-y •±'iru, — 





(2) Let AB be a right line perpendicular to the bounding planes, 
which terminate a solid composed of parallel strata of indefinite extent, 
but uniformly dense throughout that extent ; and let the law of den- 
sity of the different strata be such that there is no action on any 
point Q„ within. 

Let the solid be decomposed into n + 1 equal portions in which the 
densities are as before represented by po, pi, p% /o„. 

In this case the quantities ao, flj, 02 ci„ are all equal, and putting 

them equal to unity, we have 

u = sm9 + sin39 + sm56 + SiC. 

So = po, Si=poCOS9, \ = 1» 

S2 = 2cos9 . S^ — XiSo=poCos29, X8 = l, 

Ss=-2cos9.Sz — \2Si = poCOs39, X3=l, 

and generally, S„=pocosn9, and\„ = l. 

Hence the solution is pi = 0, p2 = p»_i = 0, pn = po' 

And if E be the whole mass and A the area of the bounding planes, 
which is supposed very great, we have 

E = 2iA.po. 

This result is analogous to the well-known fact, that electricity can 
reside only on the surfaces of bodies, and affords another instance of 
a transient function. 

The method of decomposition may always be applied to obtain 
numerical approximations in cases which involve Definite Integrals; 
for instance, in the distribution of electricity on bodies, and in esti- 
mating the forces between bodies which are electrised. 

(2) By means of Reciprocal Functions. 

43. Equations which contain only one definite integral. 

Let f(f, a) be a function involving a variable f, and an arbitrary 
parameter a; F{a) a function containing a only, and (p (t) a function 


containing t only, the first and second of these functions being given, 
it is required to find the third so as to satisfy the definite integral 

!,<l>{t).f{t,a) = F{a), 

the limits of t being given. 

Suppose (p {t) expanded according to any given class of self-reciprocal 
functions as P„, that is, 

^(^) = CoPo + CiPi + C2P2 + C3P3, &c. ad infinitum, 
where the coefficients Co, c,, Ca, &c. are unknown. 

Let J^{t, a) be expanded according to the same reciprocal functions, 
f{t, a) = AoPt, + A^P^ + A2P2 + A3P3, &c. ad infinitum. 

Then j?P„P„ = 0, and fiPnPn = a„ a known numerical quantity depend- 
ant on n, and on the particular species of reciprocal functions which 
are employed. 

Multiply both series and integrate between the given limits of /, 
and the proposed equation gives us 

F (a) = Aoao.Co + Ai ai.c^+ A2 as . C2 + ^303 . C3, &c. od infinitum. 

Now An being a known function of a and n, we can by Art. 23. 
Sect. VII., find another function of a and n, as An such that fiA„A„' = 0, 
when m and n are unequal integers. 

Multiply the equation successively by Ao, A^', Ai, &c. and take the 
definite integrals relative to a, hence 

jaA(s-P\a) = CoOojaAaAt, ', .'. Co ^ C A ' A ' 

f.A,'F{a) = e,aJ.A,A,'; •.: c, = ^^4^, 
and generally c„ = r'^'j • 
Hence <b(f) = ^ /^^°'--^(«) + ^ fgA^Fja) ^ P. fa-A^'Fja) ^ ^^ 

^ ' ao ' faAo'Ao a,' faAi'Ai aj ■ faAa'A^ 


44. Examples. 

In the following examples two things are to be observed. First, 
that the given functions are supposed to be continuous, and therefore 
the equation proposed must hold true for all values of the parameter a. 

Secondly, In the final equation for determining the unknown coeffi- 
cients, instead of using a reciprocal multiplier any means more simple 
may be occasionally employed. 

Ex. 1. Given ^^(/), cos («^) = 1 to determine <^{t\ ^he limits of t 
being and tt. 

Put (^{t) = Co + Ci cos/ + d cos (2/) + d cos (3/), &c. ml infinitum, 

and cos {at) = Ao + A^ cos t + A2 cos (2/) + A^ cos (3/), &c., 

where to determine Ao, A^, A,., &c. we multiply successively by 1, 
cos t, cos 2 A &c., and integrate from t = to t = ir, whence 

, _ sin {a-n) J _ 2asin«-7r , _ 2«sin«7r 

J n ^ / ,x„ 2« sin air , 
and generally A^ = ( — 1) . — ri ^ when n> 0. 

7r {a — n J 

Multiply both series and integrate, and we get by the proposed 


[Co a.€i , «C2 aCi . \ 

1 = sm a-K { — =- + ~„ — -„ z — -5 + &c.> 

{a a^—1 «^ — 2^ «^ — 3^ J 

Put a = 0, 1, 2, 3, &c. successively, and we get 
_ 1 2 3 . 

C(j — , Cj — , C2 = — , oZC. 


Hence ■tr(p{t) = 1 + 2cos/ + 2cos2# + 2cos3#, &c. 

The value of <t)(t) is therefore the transient function - . =^^ — ^-.'^ ^ — i-^ . 
^^ ^ TT I — Hh cos t + h^ 

{Vide Art. 38. Function Fo), when h is put equal to unity. 
Vol. V. Part III. 3D 


Ex. 2. Given fi(p(t) .cos {at) = cos (a 0). 

As before (p{t) = c„ + Cj cos t + c^ cos 2t + CaCosSt + &c. 

sin flTT fl 2«cos^ 2« cos 2^ 2a cos 3^ „ ] 

cos at = < J \ h &c.> 

therefore cos a0 = sin(a7r)l- - -~- + f^' „ ^^j + &c.l 

' [a a^ — 1 a^ — 2' a- — 3^ J 

But also by reciprocal functions we get 

sinaTrQ 2acos0 2acos20 2acos30 „ 1 
cosae = __ |- _ _,__ + --^-^^ -,__ + &c.} 

TT 1 2COS0 2cos20 2cos30 „ 
Hence Co = - , c, = , c, = , Cs — , &c. 


therefore 7r^(#) = 1 +2cos0 cos^ + 2 cos20 cos 2^ + 2 cos36 cos 3# + &c. 

or 27r(pt= 1 +2COS {9 + t) + 2 cos 2{e + 1) + 2COS 3{9 + 1) + Sic. 

+ l + 2cos{9-t) + 2cos2{9-t) + 2 cos 3{9-t) + &c. 

^ (1-A)(1+^) (1-A)(1+^) 

l-2h cos{9 + + A' 1 - 2A cos (0 - ^ + *' 

when A is put equal to unity. 

Ex. 3. Given ft <{> (t) : cos {at) = 27'(a). 

jP(a) must be such (in continuous functions) as not to change when 
— a is put for a, since cos (at) which is under the sign of integration 
will not then alter its value. 

Proceeding as in the former examples we get 

ET/ ^ • \<^o ac, ac, acs „ ] 

F{a) = sm«. |- _ -,_^ + -,_^ _ -^-^^ + &c.} 

Put successively a = 0, 1, 2, 3, &c. hence 

Co = - . 1^(0) , c. = - . F{1), c, = -. F{2), &c. 

ir IT TT 

hence 7r(p{t) = F(0) +2F(1). cos #+ 2F(2) . cos 2# + 2F(3) cos(30 + &c. 


Ex. 4. Given ft(p{t) . {/{a + t) +f{a-t)\ = F{a), 

where the forms of the functions f and F are known, and that of 

Put (f){t) = Co -1- Ci cos t + d cos %t + c-i cos St + kc. 
f{a) = oo + a, cos« + 02 cos 2«+ "3 cos 3a + &c. 
where a,„ a^, a^, &c. are known numerical quantities; hence 
J'{a + t)+J^(a-t)=^2ao+2ai cosacos t+2a2 cos 2a cos 2l + 2a3 cos 3« cos 3^-r&c. 

and JP(«) = 27raoCo + 7raiCi COS« + wa^d COS2« + TrogCs cos3« + &c. ; 

therefore Co = — — - , c, = —^— . fa F{a) . cos « , c^ = -^— L F{a) cos 2«, &c. 

J w.. 1 r t:t/ X f 1 2C0S«C0S^ 2cOS2«COS2# „ ] 

and 7r(p{t) = — f„F{a) \—- + + + &c.} 

Tr [Zao Oj as J 

the hmits of all the integrals being and tr. 

Ex. 5. /■*%=J-,. 
Jta — t a — h 

In this case we shall employ the functions V^ reciprocal to t". 

Put <^{t) = CflFo + Ci F", At C'^V-i + &c. «c? infinitum, 

1 1 !?;<'. , • ^ . 

and ;: = — \ — ; ^ — : + &c. «« infinitum; 

a-t a a' {^ "^ 

^, f 1 c„ 1 c, 2.1 6-2 3.2.1 Cs „ 

therefore r = ;r^ • -5 + ^ . e • ^ — ~. — ^ c <-, • -7 + &c- 

a-o a 2.3 e^ 3.4.5 a^ «' 

1 * A= 6' 

= -+—,+— + —4 , &c. 
a a^ ci^ a 

TJ 1 2.3 , 3.4.5 ,, ,3 „ 

Hence Co = 1, c, = ~ . *, c, = ^ ^ . V, c^ = .^ ^ ^ . h\ &c. 

and <^{t) = r,-^.br, + ^^.b^F.,-^^^^.b^r.. &c. ■ 
= r„ - 1. r,.(4i) + — •FAuy - 1^^. r3(4*r + &c. 



Put F{k) = Fo + F,A + V-it + kc. ad infinitum, 
as found in Art. 28. Sect. vii. 

Hence F{ - kr') = K-" + V.,kr' - &c. 

theretore j^ ^^^-— ^ - 2 ^^» a ^^*+ 2.4" '^'^ 2.4.6-'^'^''^ 

the limits of t being and 1 ; 

TT ^^ v'l-T^ »2' 2.4 2.4.6 

Ex. 6. j,<l>{t).f{a-t)=f{a-h). 

Denote by Pi,„ the reciprocal function P„ when ^ is the variable, 
by Pi_ri when 6 is the variable. 

Let/(«-0 = J,P,,, + A,Pt,, + A,P,,, + A^Pt,^ + &c. 
and (j>{t) = CoPf.o + c,Pu + CaP,,^ + c^.Pt.z + &c. 

.-. f{a-h) = ^oCo + g . AxCi + g . ^sC^ + \ .A^Ci + &c. 
but changing t into J in the expansion of J'{a — t) we get 

f{a-h) = AoP,,o + ^,Pm + A,P,,, + A,P,, + &c. 
which values are identical when Cc = Pi.,o, c, = SPs,,, c^ = SPh.n, &c. 
therefore (j){t) = PmP.o + SP^P., + 5P„,P,. + 7 P.^P.s + &c. 

45. Ow ^A^ appendage necessary to complete the Solution of' a 
Definite-integral Equation. 

In the examples in which f{a, t) = cosa^ given in the last article, 
the function F{a) is adapted to general differentiation relative to a, 
under the definite integral ; but besides the prime value thus obtained, 
there must be an appendage to represent the same operation on zero. 


which contains an infinite number of constants multiplied by functions 
of a, which may vanish or not, and be connected or unconnected ac- 
cording both to the nature of the particular operation and the nature 
of the calculus in which it is employed ; this has been already shewn 
by Mr Peacock*, and in Art. 20. Sect. vi. of this Memoir. The same 
remark applies to the value of (t) in the general equation 

to complete it we must add ^{t) where ft^{t) .f{t, a) = 0. 

To obtain \|/(/) in the equation ft<p{t) .cos (at) = F {a) above mentioned. 

Let us suppose (pi {t), (p-^ {t), found by the method of Art. 44., to 
satisfy the equations 

Jt(px (t) . cos (at) = 1 for continuity, 
ft(p2{t) . sin {at) = 1 for discontinuity, 

differentiating with respect to a, the first 2n times, the second 2«— 1 
times, we get 

f,<pi{t).f"' cos (at) = 0, 
ft (p.2 it), t"-' cos {at) ^ 0. 


^{t) = (p,{t) {At + Bf + Ct\ &c.} + 0,(0 {A'f + Bt^ + C't'^c.}, 
where A, B, C, &c. A', B', C, &;c. are absolute constants. 

When transient functions appear in the appendage or even in the 
prime solution, they must not be neglected (particularly in the mole- 
cular investigations) except they are inadmissible by the nature of the 
particular question, for they have a physical as well as a geometrical 
meaning, as they are capable of expressing in continuous analytical 
forms, the state of bodies and their mutual actions when they are com- 
posed of absolute mathematical centres of forces, all separated mutually by 
infinitesimal intervals. 

Q/ Q; a> 04 Qf Jr 

* Third Vol. Report of British Assoc, p. 212, &c. 


Thus let the ratio of the weight to the extent of an element P of 
a straight rod AB be expressed by the transient function 

(\-h){\+h) I, . , 

- — —^ /o -^N ■ J.2 ' when ^ = 1 ; 

and where AP=(p, and the whole length AJB = ir, and n is very great 
and integer. 

Then the whole weight is finite, viz. f - — ^—^ — '—- — - — '—n = 1, vet 
" J^l — 2hcos2n(p + h^ ^ 

this function has only an existence when = 0, -, — , — ...&c., and 

therefore the rod is actually composed of disjoint particles Q,, Qa, Q3, 

&c. which are separated by equal intervals, each infinitesimals, viz. -, 

when n is very great, and equal to the actual number of particles ; 
the action of such a system on another given one, may always be 
estimated by using the transient function in its general form, and lastly, 
putting h equal unity. 

46. Equations which contain two or more Definite Integrals. 

Given, jj cp (t) .f(t, a, b) + f,^l.{t) .F {t, a, b) = E {a, b), 

the forms of the functions^ F, E being known, the forms of and 
■<\f are required. 

Put /(#, a, b) = ^oPo + A,P, + A^P^ + A^Ps + &c. ad inf. 

where A^, Ai, A2, &c. are known functions of a and b, and Po, Pi, &c. 
any self-reciprocal functions of t, such that ftPr!^ = a„, which will be a 
known numerical quantity. 

Similarly, F {t, a, b) = B,Po + B,Pi + B,P, + B,P^ + &c. ad inf., 
where B^, Bi, B^, &c. are known functions of a and b. 

Again, let (p{t) =CoPo + c,Pi +CaP2 + C3P3, &c. ad inf. 
where Co, c,, Ca, &c. are unknown numerical quantities, 
and \l/{t) = eoPo + ejPj+e2P2 + e3P3,&iC. ad inf.. 


where eo, e^, e.,, &c. are also unknown. 
The proposed equation then becomes 

+ eoOoBo + eiUiBi + eia-iS^ + &c.) 

Now to the function A„ there may be found a function A„ reciprocal relative to a, 
and to B„ B„ b. 

Let f„AaB„ = U„ a function of b only, 
ftBoA,, = V„ a only. 

Hence, f^AoE (a, b)-CoaofaAoAo = eoaoUo + eia^Ui + 6.^0^112 + kc. ad inf. 
ft BoE {a, b) - e^a^ fiBoB^ = c^a^ K + c, a, F; + c^a^ F; + &c. ad inf. 

Let t/„ be the function of b, which is reciprocal to f7„, 
V„ of «, V^. 

\L k {Ao U,E {a, b) - c,a,A,A^ t7"„) = e„a„ /j t7„ U^ 
Hence, \ \, 

\fJ,{Bo KEia, b) - e„a,B,B,K) = c„«„/„F„rJ 

by which equations the constants Co, e,, are immediately determined. 

\fa fb (Ao U„E (a, b) - c^a^A^A^ U„) = e„ a„ /j Un UA 

Also, \ >; 

(/„ /, (^0 KE{a, b) - e,a,B,B, K) = c„a„ f„ V„ Vj\ 

and since c„, e^, have been found, the latter equations determine gene- 
rally the coefficients c„, e„, and therefore the required functions <p{t), 
^ {t) are known. 

In like manner by employing reciprocal functions relative to double 
integration, we may solve equations containing three unknown func- 
tions, &c. 

The problem of the distribution of electricity on bodies of which 
the surfaces are not . continuous, introduces equations of this nature. 


47. Simultaneous Equations to Definite Integrals. 

Given l-^*^^^^ ••^^^' *) +-'^^(^) • ^^^' «) = ^^^H 
\!i<p{t)Mt, a) +f,i.(t) . FM a) = EAa)i ' 

the forms of the functions j^ F, E,fi, Fi, Ei, being known, the forms 
of (p and \{^ are required. 

Multiply the second equation by an arbitrary quantity \, and adding 
to the first, put 

f(t, a) + X/ {t, a) = A,Po + A,P, + A,P, + &c. 

F{t,a) +xF^{t,a) = AoQo + A,Q, + A,Q2 + &c. 

(pit) = CoPo + c,P,' + c,P^ + &c. 

^^(0 = ^oQo' + e,Q, + e,Q: + &c. 

where P„, P,, P2, &c.\ « .. /? ^ i 

^ ^ .-k o f are functions of t only, 

Qo, Qi, Q2, &C.J ^ 

A^, Ai, Ai,\ known functions of a, X, and self-reciprocal relative 
to a, 

PI, Qn reciprocal to P„, Q„ respectively, hence 

(putting /,P„P„'=;),„ j;Q„Q„'=^„) E{a) + XEM 
= c^poAo + c,p,Ai + C2P2A2 + &c. + eoqoAo + e^q.A, + e^q^A-^, &c. ; 
.-. faAoEia) + X faAoE^ (a) = {copo + eoqo)faAo\ 
faA,E{a) + XlA,E,{a) = iCiP^ + e,q,)faA^ 

and giving to X any two values in each of these equations, the first 
will produce two equations which determine Co, eo, the second will 
similarly give Ci, e^, &c., and thence the functions (p{t), \l/t are known. 

The same method is applicable to any number (n) of simultaneous 
equations involving n unknown functions. 

48. Definite-integral Equations of superior orders and degrees. 

Methods similar to the preceding are applicable in most cases of 
the former class thus : 


Given fj,<t>{t, T)f(t, T, a) = F{a), 
the forms F and J" being known to determine (p. 

By Art. 16. Sect. iv. let a function Q„ be formed which shall be 
self-reciprocal, relative to double integration for t and t. 

Put ^(#,t) = Co Q„ + CiQi + C2Q2 + &C. 1 _rrri'i 

and/(^,T,«) = ^„Q„+^,Q, + ^,Q, + &c.r'''* ^^^ a»-i.^t^», 

hence F(a) = aoCo^o + aiCi^, +0203^2 + &c. 

Let ^„' be a function of a reciprocal to A„, 
then faA,'F{a) = c^aJaA.A^, 

faA,'F{a) = c,aJaA,A„ 
&c. &c. 

whence Co, Ci, &c. being determined, the function (p{f,T) is known. 

Equations of superior degrees must generally be converted into equa- 
tions of superior orders to be easily solved, thus; 

Given f,(p{t) .fit, a) x [,cp{t) . F{t, «) = >/.(«), 
the forms ^ F, and -^^f being given to find the function 0. 

Introduce another variable t having the same limits as t, then it is 
evident that 

J,<p{t) . F(t, a) = /^«^(t) . F{t, a) ; 

.-. U^cp{t) .(pi-r) ./{t, a) . F{t, a) = f (a), 

and since y(#, a) . F{t, a) is a given function of t, t and a, the unknown 
function (p{t).(p(T) will be determined as above, and representing it by 
<p^(t,T), let a be a root of the equation 0(t) = 1, then since (p{t).(p{T) 
= 0i(#, t), we get the required function (}>{t) = <pi{t, a), and again putting 
^ = a we get ^1 (a, a) = 1, from which equation a is known, and there- 
fore <p{t) = <pi{t,a) is also known. 

49. In researches on the subjects of electricity, and the phaenomena 
dependent on the molecular construction of bodies, the only data which 
can be furnished by experience are the total actions, and consequently 
Vol. V. JPart III. SE 


the analytical processes of calculation require the solution of definite 
integral equations: some of these have been resolved by Laplace and 
others, by means of particular artifices by which the unknown functions 
were subjected to differential equations ; but as no general method 
existed for this purpose, the resolution of such equations has been ex- 
tremely limited, and apparently simple physical problems, such as the 
distribution of electricity on surfaces, (with the exception of a very 
few cases) have consequently defied the powers of analysis. Besides, 
an abundance of facts connected with the interior arrangement of the 
molecules of bodies are of such a nature, that mathematics possessed 
but little power of reducing them to analytical forms, calculated to 
produce any valuable inferences ; these facts are daily increasing in 
number, and the analyst is far behind the cultivator of Experimental 
Physics. The Memoirs on the Inverse Method of Definite Integrals 
which are now concluded, and which have been pursued when the 
absence of ordinary engagements permitted, originated in the belief 
that by proceeding gradually from the simplest classes of Definite 
Integrals to the more complex, the general principles of an Inverse 
Method would be discoverable. The formation of all possible classes 
of Reciprocal Functions, and the Transient Functions included amongst 
them, have at length furnished means for the resolution of equations 
to Definite Integrals. The author is however well aware that there 
must exist numerous imperfections in the manner in which his design 
is executed, but believing also that by those endeavours, however weak, 
some fresh powers have accrued to analysis, as an instrument of investi- 
gation, he trusts they will deserve the approbation of the Society. 


Caius Colleob, 
Dec. 24, 1834. 


Analytical Table of Reference to the "Memoirs on the Inverse 
Method of Definite Integrals." 

FIRST MEMOIR, Vol. IV. Page 353, &c. 


Introduction 353 

Section I. Principles relative to Continuous Functions. 

Art. 1. Method of reducing the given limits of integration to and 1 in all cases 358 

Arts. 2, 3, 4. In the general equation ft f{t) .t''= <p(x), x is understood to lie between 
— 1 and + 00 , then cyj (x) converges to zero as x increases, when y(<) is any of the 
functions usually received in analysis; consequent division of the subject 35g 

Art. 5. Rule; When the known function <p{x) is rational, seek the coefficient of - in 


<p (x) . t~', dividing it by / we obtain fit) 362 

Art. 6. Examples s6S 

Arts. 7, 8. Means of facilitating the Calculus oi f{t) 3Q5 . 

Art. 9. and Note (A). When {x) is a logarithmic function 366, 400 

Art. 10. When (pix) is expressed by an equation to finite differences 367 

Art. 1 1 . When <f> (x) is a fraction, the denominator containing imaginary factors 369 

Art. 12. When {x) is irrational 37O 

Art. 13. Cases when equations of the form J',f(J,').(t'' ±t~')=if>x, may be resolved by 

the preceding method 37] 

Art. 14. Extension of the general rule to successive integration with respect to any 

number of variables , 373 

Section II. Principles relative to Discontinuous Functions. 

Art. 15. Cases of discontinuity in Physical Problems quoted 374 

Art. 16. To find a formula which shall represent the least of the two quantities a, /3. . 375 
Art. 18. To find a formula which shall represent_/(a) or y(/3) according as a is < or > /3. 376 

Art. 19. To find a formula which shall represent r^, or -= — j— , according as a is 

a — lip p — na 

< or > /3 377 

Art. 20. To find a formula which shall represent ■~^^ , or ^^^, , according as a is < or > /3 378 

Arts. 21, 22. Method of representing discontinuous functions of any number of breaks 380 

Arts. 23, 24. Geometrical Illustrations of the theory of discontinuity 382 




Section III. Application of the preceding principles to the Phaenomena of Developed 

Electricity/ 386 

Note (A), No. 2. On the general separation of the positive powers of the variable from 

the negative 402 

Note (B), No. 1 . On the apparently improper forms of (p (x) 404 

No. 2. Method of valuing the results of operative functions 406 

SECOND MEMOIR. Vol. V. Page 113, &c. 
Introduction 113 

Section IV. Inverse Method Jbr Defitiite Integrals which vanish, and theory of Reci- 
procal Functions. 

Arts. 1,2. X being restricted to the natural numbers 0, 1, 2, (»— 1) to &nd fit) so,f(t).f = 116 

Art 3. P„ denoting the function y(<) above-found, when m and n are unequal /,P„P„=0, 

and when equal /,P,P„ = 117 

Art. 4. To find a rational function _/(<) which may satisfy the equationy^y(<)'''=0, x being 

any number of the series p, ^+ 1,. . .p-{-n—l 118 

Art. 5. The general form of fit), when x is from to n— 1 inclusive, is 

d" (ft'" V\ 
At) = • ^^^„ ^ , where t' = \ - t 118 

Art. 6. In this case the equation^" (/) = 0, has n real roots lying between and 1 1 19 

Arts. 7, 8. To find a rational function of h. 1. /, such that /,y{h. 1. (<)}.<' = 0, when 

X is from to n— 1 inclusive 120 

Art. 9. Denoting this function by Z.„, the function which it generates is the value of 

u in the equation u {l — hh.\. u) = t 122 

Art. 10. If Q„ be the coefficient of *" in -r- , u being found from the equation 

m(1 — A £/) = <, where J7 is a function of u vanishing when u = \, and T the same 

function of t, then -^l = "("- V""^"~""^'^ 122 

Art. 11. If t/ be a rational and entire function of u vanishing when m=1, and Q, be 
the term independent of u in the product i/" I 1 I , then shaliy]Q„<'= 0, when 

X is from to «— 1 inclusive 12S 

Art. 12. To find (p, q)„ a rational and entire function of P" of n dimensions, which 

multiplied by a rational and entire function of f of less than n dimensions, the integral 

of the product may vanish from t — Otot=l 125 

Art. 13. Reciprocal Functions; such are ip,q)„, {<l>p\'; value of the integral of the 

product when n=n' 126 

Art. 14. To find a function A„ reciprocal to the function L„ found in Art. 8 128 



Art. 15. General principle for finding Reciprocal Functions to simple integration 130 

Art. l6. The same extended to integration for any number of variables 131 

Art. 17. Examples 132 

Section V. Inverse Method for Junctions which contain positive ponters of x, or are 
under any other form. 

Art. 18. An appendage must be annexed in all such cases 135 

Arts. 19, 20. When ^j; is a rational and entire function of x ; and particular example 

when ^(.r) = l 136 

Art. 21. To find /(<) when7(/(O-''=0W, an<i ^ is from to n—\ inclusive 138 

Arts. 22, 23. Various modes of determining y(<) in this case 141 


Arts. 24, 25. The coefficient of h" in the expansion of , is a self-reciprocal function 146 

THIRD MEMOIR. Vol. V. Page 315, &c. 
Introduction 315 

Section VI. Method of discovering Reciprocal Functions, when the integrations are 

performed with respect to any fonction of the variable. 

Arts. 1, 2. General principle for varying the limits 318 

d'.^t'f'V) dt 
Art. 3. If V can be found so that — "' — 'Tl "™^y "'^ ''^ " dimensions in t (where 

t' = \ — t) then this quantity will be self-reciprocal relative to (p $iq 

Art. 4. If V can be found so that — j^ ■ j? "^^y be of n dimensions in t, then 

the factor by which -j^ is multiplied will be self-reciprocal relative to 320 

Arts. 5, 6. If <p=f,(Jt(y indefinite, and m between —1 and -j-oo, and if 

d^ {(tt'Y'^"'\ 
Qn = 1 2 xdf-^" ' ("')""" *^" *^^^ ^ ^^ self-reciprocal relative to (p 321 

n — m 


If z= ftittfy indefinite, and m be between -1-1 and —00, and if a» = — ' 

^ -"^ ' > ' 1" i.2...ndt'" 

then shall q, be self-reciprocal relative to 321 

Art. 7. To find the functions which Q„, 5, generate 322 

Arts. 8, 9. When 7« = — §, Q„, q, are the trigonometrical reciprocals 323, 325 

Art 10. In the identities thus obtained, the sign of n may be changed so as to pass 

from differential coefficients to integrals 326 

Art. 1 1 . The two series of reciprocal functions obtained from the theorems of Arts. 5 81. 6. 

differ only with respect to the variable of integration 328 

Art. 12. Examples of the preceding theory 329 

Art. 13. To express Q„ and q„ in terms of t alone 330 

Art. 14. To express Q„ and q„ by means of differential equations 332 



Art. 15. The reciprocal functions expressed by the general formulae for Q,, q, all possess 
a common property, viz., their integrals vanish when taken between limits which 
render the functions maxima or minima 333 

Art. 16. To find the complete integral of the equation 

i<'^ + (m+l)(l-2/) -^^+n(n+2m + l)u = 335 

Art. 1 7. To find explicitly the omitted part of the complete integral in Laplace's equation, 
for the coefficients in the expansion of the reciprocal of the distance between two 
points in a plane 338 

Art. 18. When m = —\ the general equation of Art. l6. represents the trigonometrical 
functions 342 

Art. 19. Remarkable properties of the functions 

G = g^cose cos(a;sin0), 0'= £■>; ™s e sin (a; sin 0) 342 

Art. 20. Application to the general differentiation of rational and integer functions oi x . . . 344 
Art. 21. The sum of all the divisors of a number n, including itself and unity 

= fy^- 1- {sin^ sin 2^ sinn^} . cos 2n(^ 346 


Section VII. On Transient Functions. 

Art. 22. Nature of transient functions 347 

Art. 23. To find a function reciprocal to f(t, n) any given function of the variable t 

and integer n ' 348 

Art. 24. To find a function V„ reciprocal to /" 349 

Art. 25. The function V„ is transient 352 

Art. 26. To express the transient function F„ in a finite form 354 

Art. 27. Discussion of the transient function Fo ; it represents the state of a body which 

an electric spark is about to enter 356 

Art. 28. To find the quantity to which F„ is the generating function 360 

Art. 29. To expand a given function in terms of the functions F„ 36l 

Art. 30. To find a function t/„ reciprocal to (h. 1. t)' 362 

Art. 31. In a finite form U„ = H — 2_ when A = l 363 

1 .2...ndh" 

Art. 32. Properties of C7„ &&f,Uj' = x', &c S6S 

Art. 33. Discussion of the function Uq 364 

Art. 34. To expand a given function in terras of the functions I/„ 365 

Art 35. To find a function reciprocal to /" when the limits of t are and =0 366 

Art. 36. To find a function F„ reciprocal to cos"^ between the limits ^ = and <p=-ir. 366 

Art. 37. The function F„ is transient - 367 

Art. 38. To express F, in a finite form 368 

Art. 39. Means of summing a series expressed in transient functions 369 



Section VIII. On the Resolution of Equations which involve Definite Integrals. 

Art. 40. Method of decomposition into elements 371 

Art. 41. Density of a cylindric shell which exercises no action on any point in its axis 

with any law of force 372 

Art. 42. Examples when the law of force is the inverse square of the distance 374 

Art. 43. Resolution of equations which contain but one definite integral and one parameter 377 

Art. 44. Examples 379 

Art. 45. On the appendage necessary to complete the solution of a Definite Integral 

Equation 382 

Transient functions capable of representing in a continuous form the state of a body 

composed of mathematical centers of forces separated by infinitesimal intervals 383 

Art. 46. Equations which contain two or more Definite integrals and as many parameters 384 

Art. 47. Simultaneous equations to Definite integrals 386" 

Art. 48. Definite integral equations of superior orders and degrees 386 

Art. 49. Conclusion 387 


First Memoir. 359, line 9. 16. 18. dele y in the sign ooy. 
Vol. IV. 377, line 8. for A</3 read A-^ 1. 

406, lowest line and third from bottom, for terms read times. 

407, line 17. supply the word, equation. 

Second Memoir. Vol. V. 134, line 6, after -j^ supply (tf)*. 

line 7. after ,* supply (tf). 

136, line 3, 4. 18. for v, put v.. 

Third Memoir. Vol, V. 332, lowest line, for /'"(•-')(() read /'"<'-"(«). 
333, line 5, for /(O) read f"(0). 
for m+3 read (m+3). 

337, line 8, put (tf)-" before ~ in the last term. 
346, line 16, for intger read integer. 
357. line 8,/or (1-ft*) reod (1-A)*. 

XV. Oil the Determination of the Exterior and Interior Attractions of 
Ellipsoids of Variable Densities. By George Green, Esq., 
Caius College. 

[Read May 6, 1833.] 

The determination of the attractions of ellipsoids, even on the hypo- 
thesis of a uniform density, has, on account of the utility and difficulty 
of the problem, engaged the attention of the greatest mathematicians. 
Its solution, first attempted by Newton, has been improved by the suc- 
cessive labours of Maclaurin, d'Alembert, Lagrange, Legendre, Laplace, 
and Ivory. Before presenting a new solution of such a problem, it 
will naturally be expected that I should explain in some degree the 
nature of the method to be employed for that end, in the following 
paper; and this explanation will be the more requisite, because, from 
a fear of encroaching too much upon the Society's time, some very 
comprehensive analytical theorems have been in the first instance given 
in all their generality. 

It is well known, that when the attracted point p is situated within 
the ellipsoid, the solution of the problem is comparatively easy, but 
that from a breach of the law of continuity in the values of the 
attractions when p passes from the interior of the ellipsoid into the 
exterior space, the functions by which these attractions are given in the 
former case will not apply to the latter. As however this violation 
of the law of continuity may always be avoided by simply adding a 
positive quantity, u" for instance, to that under the radical signs in 
the original integrals, it seemed probable that some advantage might 
thus be obtained, and the attractions in both cases, deduced from one 
common formula which would only require the auxiliary variable u to 
become evanescent in the final result. The principal advantage how- 
ever which arises from the introduction of the new variable u, depends 
Vol. V. Part III. SF 


on the property which a certain function F'* then possesses of satisfy- 
ing a partial differential equation, whenever the law of the attraction 
is inversely as any power n of the distance. For by a proper applica- 
tion of this equation we may avoid all the difficulty usually presented 
by the integrations, and at the same time find the required attrac- 
tions when the density p is expressed by the product of two factors, 
one of which is a simple algebraic quantity, and the remaining one 
any rational and entire function of the rectangular co-ordinates of the 
element to which p belongs. 

The original problem being thus brought completely within the pale 
of analysis, is no longer confined as it were to the three dimensions of 
space. In fact, p' may represent a function of any number s, of in- 
dependent variables, each of which may be marked with an accent, in 
order to distinguish this first system from another system of s analo- 
gous and unaccented variables, to be afterwards noticed, and F' may 
represent the value of a multiple integral of s dimensions, of which every 
element is expressed by a fraction having for numerator the continued 
product of p into the elements of all the accented variables, and for 
denominator a quantity containing the whole of these, with the un- 
accented ones also formed exactly on the model of the corresponding 
one in the value of V belonging to the original problem. Supposing 
now the auxiliary variable u is introduced, and the s integrations are 
effected, then will the resulting value of ^ be a function of u and of 
the s unaccented variable to be determined. But after the introduction 

* This function in its original form is given by 

-. /• p' dx dy dz 

J {{X - xy + (/ - yf + (.' - 2)2}"-^' 

where dx dy dz represents the volume of any element of the attracting body of which p' 
is the density and x , y , z are the rectangular co-ordinates ; x, y, z being the co-ordinates 
of the attracted point p. But when we introduce the auxiliary variable u which is to be 
made equal to zero in the final result, 

jr _ r p dx dy dz 

J{(^a:'-xf-\.{y-yy + {z-zf + u^yr' 
■ - .YOii 
both integrals being supposed to extend over the whole volume of the attracting body. 


of w, the function V has the property of satisfying a partial differen- 
tial equation of the second order, and by an application of the Cal- 
culus of Variations it will be proved in the sequel that the required 
value of V may always be obtained by merely satisfying this equation, 
and certain other simple conditions when p is equal to the product 
of two factors, one of which may be any rational and entire function 
of the s accented variables, the remaining one being a simple algebraic 
function whose form continues unchanged, whatever that of the first 
factor may be. 

The chief object of the present paper is to resolve the problem 
in the more extended signification which we have endeavoured to ex- 
plain in the preceding paragraph, and, as is by no means unusual, the 
simplicity of the conclusions corresponds with the generality of the 
method employed in obtaining them. For when we introduce other 
variables connected with the original ones by the most simple rela- 
tions, the rational and entire factor in p still remains rational and 
entire of the same degree, and may vmder its altered form be ex- 
panded in a series of a finite number of similar quantities, to each of 
which there corresponds a term in V, expressed by the product of two 
factors; the first being a rational and entire function of s of the new 
variables entering into V, and the second a function of the remaining 
new variable h, whose differential coefficient is an algebraic quantity. 
Moreover the first is immediately deducible from the corresponding 
part of p without calculation. 

The solution of the problem in its extended signification being thus 
completed, no difficulties can arise in applying it to particular cases. 
We have therefore on the present occasion given two applications 
only. In the first, which relates to the attractions of ellipsoids, both 
the interior and exterior ones are comprised in a common formula 
agreeably to a preceding observation, and the discontinuity before 
noticed falls upon one of the independent variables, in functions of 
which both these attractions are expressed ; this variable being con- 
stantly equal to zero so long as the attracted point j) remains within 
the ellipsoid, but becoming equal to a determinate function of the co- 



ordinates of p, when p is situated in the exterior space. Instead too 
of seeking directly the value of V, all its differentials have first been 
deduced, and thence the value of V obtained by integration. This 
slight modification has been given to our method, both because it 
renders the determination of V in the case considered more easy, and 
may likewise be usefully employed in the more general one before 
mentioned. The other application is remarkable both on account of 
the simplicity of the results to which it leads, and of their analogy 
with those obtained by Laplace. (Mdc. C^. Liv. iii. Chap. 2.) In fact, 
it would be easy to shew that these last are only particular cases of 
the more general ones contained in the article now under notice. 

The general solution of the partial differential equation of the second 
order, deducible from the seventh and three following articles of this 
paper, and in which the principal variable 1^ is a function of # + 1 
independent variables, is capable of being applied with advantage to 
various interesting physico-mathematical enquiries. Indeed the law of 
the distribution of heat in a body of ellipsoidal figure, and that of the 
motion of a non-elastic fluid over a solid obstacle of similar form, 
may be thence almost immediately deduced; but the length of our 
paper entirely precludes any thing more than an allusion to these ap- 
plications on the present occasion. 

1. The object of the present paper will be to exhibit certain 
general analytical formulae, from which may be deduced as a very 
particular case the values of the attractions exerted by ellipsoids upon 
any exterior or interior point, supposing their densities to be represented 
by functions of great generality. 

Let us therefore begin with considering p as a function of the s 
independent variables 

»r J , x<i , x^ ••••• o/i, 

and let us afterwards form the function 

dxjdx^ dxj dxl . p .^. 

'{{x,-xiJ^{x,-xl)^^ ^(x.-xlJ^u'-S^ 




the sign / serving to indicate * integrations relative to the variables 

x^, x-i, X3', x/, and similar to the double and triple ones employed 

in the solution of geometrical and mechanical problems. Then it is 
easy to perceive that the function V will satisfy the partial differen- 
tial equation 

t/vr ^, d^ ^ n-s dV 
" ~ dx,^ "^ dxi "*" ^ dx^ '^ du^^ u du ^^' 

seeing that in consequence of the denominator of the expression (1), 
every one of its elements satisfies for V to the equation (2). 

To give an example of the manner in w^hich the multiple integral 
is to be taken, we may conceive it to comprise all the real values 

both positive and negative of the variables ar/, x^, x,, which satisfy 

the condition 

the symbol / , as is the case also in what follows, not excluding equality. 

2. In order to avoid the difficulties usually attendant on integra- 
tions like those of the formula (1), it will here be convenient to notice 
two or three very simple properties of the function F". 

In the first place, then, it is clear that the denominator of the 
formula (1) may always be expanded in an ascending series of the 

entire powers of the increments of the variables x^, x^, x„ u, and 

their various products by means of Taylor's Theorem, unless we have 

and therefore V may always be expanded in a series of like form, 
unless the s + 1 equations immediately preceding are all satisfied for 
one at least of the elements of V. It is thus evident that the func- 
tion V possesses the property in question, except only when the two 

4!fl# Mr green, on THE DETERMINATION OF THE 

% + %^%+ +% z 1 and u = .(3) 

are satisfied simultaneously, considering as we shall in what follows 
the limits of the multiple integral (1) to be determined by the conr 
dition (a)*. 

In like manner it is clear that when 

Z^2+ Jl+ + 77-2>^ (4)» 


the expansion of V in powers of u will contain none but the even 
powers of this variable. 

Again, it is quite evident from the form of the function f^ that 
when any one of the * + 1 independent variables therein contained be- 
comes infinite, this function will vanish of itself. 

3. The three foregoing properties of F combined with the equa- 
tion (2) will furnish some useful results. In fact, let us consider the 

fd.,d^,...d..duu-'.[[^)\ [^)\ + (g)\ (^^)] (5) 

where the multiple integral comprises all the real values whether posi- 
tive or negative of x^, x^, x,, with all the real and positive values 

of u which satisfy the condition 

/!« 2 A< 2 « 2 /|/2 

^■^^^ + -^^^^F^^ ^^^ 

* The necessity of this first property does not explicitly appear in what follows, but 
it must be understood in order to place the application of the method of integration by 
parts, in Nos. 3, 4, and 5, beyond the reach of objection. In fact, when V possesses this 
property, the theorems demonstrated in these Nos. are certainly correct: but they are not 
necessarily so for every form of the function V, as will be evident from what has been 
shewn in the third article of my Essay on the Application of Mathematical Analysis to 
the Theories of Electricity and Magnetism. 


«,, «2, a, and h being positive constant quantities; and such that 

we may have generally 

Ur > dr. 

In this case the multiple integral (5) wiU have two extreme limits, 
viz. one in which the conditions 

V ^ or IT 71 

-\ + -^ + + -^ + t; = 1 and u— a. positive quantity (7) 

are satisfied; and another defined by 

% + %+ +-, /I and « = 0. 

jVIoreover, for greater distinctness, we shall mark the quantities be- 
longing to the former with two accents, and those belonging to the 
latter with one only. 

Let us now suppose that J^" is completely given, and likewise F,' 
or that portion of f^' in which the condition (3) is satisfied ; then if 
we regard F/ or the rest of T^' as quite arbitrary, and afterwards endea- 
vour to make the quantity (5) a minimum, we shall get in the usual 
way, by applying the Calculus of Variations, 

-fdx.dx, clx,u"-^^r,'~- (8) 

seeing that ^V" = and SFj' = 0, because the quantities V" and F,' 
are supposed given. 

The first line of the expression immediately preceding gives generally 

= 2'+'— — ^Hf^ {^•\ 

' dxr du' u du ^ ' 

which is identical with the equation (2) No. 1, and the second line gives 

= u'"'' ~7-^(^' being evanescent) (9). 


From the nature pf the question de minimo just resolved, there can 
be little doubt but that the equations (2') and (9) will suffice for the 
complete determination of V, where V" and V-l are both given. But 
as the truth of this will be of consequence in what follows, we will, 
before proceeding farther, give a demonstration of it; and the more 
wiUingly because it is simple and very general. 

4. Now since in the expression (5) u is always positive, every one 
of the elements of this expression will therefore be positive; and as 
moreover V" and F"/ are given, there must necessarily exist a function 
Fo which will render the quantity (5) a proper minimum. But it 
follows, from the principles of the Calculus of Variations, that this 
function Va, whatever it may be, must moreover satisfy the equations 
(2') and (9). If then there exists any other function F", which satisfies 
the last-named equations, and the given values of V" and V^, it is easy 
to perceive that the function 

will do so likewise, whatever the value of the arbitrary constant quan- 
tity A may be. Suppose therefore that A originally equal to zero 
is augmented successively by the infinitely small increments SA, then 
the corresponding increment of V will be 


and the quantity (5) will remain constantly equal to its minimum 
value, however great A may become, seeing that by what precedes 
the variation of this quantity must be equal to zero whatever the 
variation of V may be, provided the foregoing conditions are all satis- 
fied. If then, besides F"o . there exists another function F"; satisfying 
them all, we might give to the partial differentials of F", any values 
however great, by augmenting the quantity A sufficiently, and thus 
cause the quantity (5) to exceed any finite positive one, contrary to 
what has just been proved. Hence no such value as F, exists. 

We thus see that when F"" and F"/ are both given, there is one 
and only one way of satisfying simultaneously the partial differential 
equation (2), and the condition (9). 


5. Again, it is clear that the condition (4) is satisfied for the whole 
of F"/; and it has before been observed (No. 2.) that when V is deter- 
mined by the formula (1), it may always be expanded in a series of 
the form 

r = ^ + J?«' + Cu' + &c. 

Hence the right side of the equation (9) is a quantity of the order 
?/"-'+' ; and v! being evanescent, this equation will then evidently be 
satisfied, provided we suppose, as we shall in what follows, that 

n — s \ \ is positive. 

If now we could by any means determine the values of V" and 
V( belonging to the expression (1), the value of V would be had 
without integration by simply satisfying (2') and (9), as is evident from 
what precedes. But by supposing all the constant quantities a,, «2> «3 
a, and h infinite, it is clear that we shall have 

= V", 

and then we have only to find V^, and thence deduce the general 
value of V. 

6. For this purpose let us consider the quantity 

w ^ ^7 n-AdVdU dVdU , dVdU dVdU\ 

jdxidx.i...dx,duvr '{-r—-j— + -f— -j— + ••• + i—n— + -i t-)\ (10^ 

{dxidx^ dx.dxi dx.dx, du du j ' ^ ' 

the limits of the multiple integral being the same as those of the 
expression (5), and U being a function of ;r,, x^, x, and u, satis- 
fying the condition 0= U" when «,, a^, a, and h are infinite. 

But the method of integration by parts reduces the quantity (10) to 

— fdXidxi dx,—j — u'"-' . V 


-/..........x..»».-.r|.,«^+^.^^} (H, 

since = V"\ and as we have likewise = U", the same quantity (10) 
may also be put under the form 
Vol. V. Part III. SG 


— fdxidXi dxi—r—u'"-' . U' 

.fdx,dx,...dx,duu''-'.u\^r'^,+^ + '^^^ (12). 

Supposing therefore that U like V also satisfies the equation (2'), 
each of the expressions (11) and (12) will be reduced to its upper line, 
and we shall get by equating these two forms of the same quantity : 

idx^ dx2...dxs-j~ u'"-' V = fdxi dXi...dxs -y- «'"* U' : 
au au 

the quantities bearing an accent belonging, as was before explained, to 
one of the extreme limits. 

Because V satisfies the condition (9), the equation immediately pre- 
ceding may be written 

dU' dV 

fdxidx2...dxs-j — u'"~' V = fdxidxi...dx,—y^u'"-' U,'. 
du du 

If now we give to the general function U the particular value 

u= {{x, - x,"y + {x, - x,y + + {x, - xjy + u']^-, 

which is admissible, since it satisfies for V to the equation (2), and gives 
U" = 0, the last formula will become 


/dxidx-i dxsu'"'' —j-^ 
{{x, - x^y + {x, - x:j + + (a;, - xlj + m'^}^ 

_r dxydx^ c?;g,.(l-w) «'"-'+' V , 

\{x, - xlj + (ar, - xij + + {x, - x:j + u''\'^ 

in which expression «' must be regarded as an evanescent positive 

In order now to effect the integrations indicated in the second 
member of this equation, let us make 


x^ — Xi" = u'p COS 6i ; x-i—x" = u'p sin Qx cos 0^ ; Xi—x^'—u'p sin 0, sin 02 cos 03> &c. 
until we arrive at the two last, viz., 

«,_! -x[^-^ = u'p sin^i sin ^^ sin0,_2 cos0,_i, 

X, — ar," = «'/o sin ^1 sin 02 sin 0,_2 sin 0,_i; 

u' being, as before, a vanishing quantity. 

Then by the ordinary formulas for the transformation of multiple 
integrals we get 

dxi dx-i dx, = u''f/~^ sin^i'"^ sin 02*"' smO^^.^dp d6i dOi...dd,.i, 

and the second number of the equation (13) by substitution will become 

f dp d9i de, d9,_,p'-' sm9r'' sin 9,'-' sin e,_2 . (1 - ») r' 

/ »+i (1*)- 

But since u' is evanescent, we shall have p infinite, whenever x^, Xi,...x, 
differ sensibly from x", x^',,..x"\ and as moreover w — * + l is positive, 
it is easy to perceive that we may neglect all the parts of the last 
integral for which these differences are sensible. Hence V may be 
replaced with the constant value VI in which we have generally 

Jbf ^^ vUf • 

Again, because the integrals in (14) ought to be taken from 0,_, = o 
to 0r-, = 27r, and afterwards from 0,. = O to 9r = -n-, whatever whole number 
less than 5—1 may be represented by r, we easily obtain by means of 
the well known function Gamma: 


/sin^i'-'' sin 02'"' sin 03'"' sin0,_2C?0,</02...c?0,., = ^ZL; 

and as by the aid of the same function we readily get 

r» — * + l> 

f P'~'dp _ V2/ V 2 ) 

Wi + ,f-^ 2r(^) 



the integral (14) will in consequence become 

and thus the equation (13) will take the form 

dx.dx, dx,u"^-^^ ~^'^-^ ^~V^ ) 

J {{x,-x; 

J + {x, - x;j +.... + (x, - x:j + m'^ } -^ r (■ 


In this equation V '\?, supposed to be such a function of x^, x.^ x, 

and u, that the equation (2) and condition (9) are both satisfied. More- 
over V'^O, and Vo is the particular value of F' for which 

Let us now make, for abridgment, 

P = u"-' -r-, {when u = 0) (A), 

and afterwards change x into x\ and x" into x in the expression im- 
mediately preceding, there will then result 

_- s f fi—s + V 

r dx^ dx2 . ...dx,'P,' ""^^'^ I 2 / „, ,,^^ 

/ ^^ rr — r; f^ •••U5), 

{{x,'-x,f+{x,'-x,y + ...+{'>':-^sY+u"]— r(^) 


P' being what P becomes by changing generally Xr into x,', the unit 
attached to the foot of P' indicating, as before, that the multiple 
integral comprises only the values admitted by the condition {a), and 
V being what V becomes when we make u = 0. 

The equation just given supposes u' evanescent; but if we were to 
replace u with the general value u in the first member, and make a 
corresponding change in the second by replacing F'' with the general 
value F, this equation would still be correct, and we should thus have 


r dx'dxi dxlP,' \ 2 J ,^ , „ 

f '■ '—^ ^, = -L-jl— ir...(l6). 

-^ \{x,'-x,y+{x,'-x,y+... + {x:-x.y+u'\— r(^^) 

For under the present form both its members evidently satisfy the 
equation (2), the condition (9), and give V" = 0. Moreover, when the 
condition (3) is satisfied, the same members are equal in consequence 
of (15). Hence by what has before been proved (No. 4), they are 
necessarily equal in general. 

By comparing the equation (16) with the formula (1), it will become 
evident, that whenever we can by any means obtain a value of V satis- 
fying the foregoing conditions, we shall always be able to asSgn a value 
of p which substituted in (1) shall reproduce this value of V. In fact, 
by omitting the unit at the foot of P", which only serves to indicate 

the limits of the integral, we readily see that the required value of p is 
p'= \ P' {c). 

r^ 'r. fn~S+l\ ' 

7. The foregoing results being obtained, it will now be convenient 
to introduce other independent variables in the place of the original 
ones, such that . 

^1 = «i?i» «a = 02^2j x, = as^„ u = hv, 

Oj, ttj, flj being functions of h, one of the new independent variables, 

determined by 

a,' = «;* + h', a,- = (h' + h\ a/ = aj' + /^^ 

and V a function of the remaining new variables, f,, ^2, ^3, ^s satis- 
fying the equation 

1 = v' + |;^ + e/+ + U; 

a,', a/, Os', 0/ being the same constant quantities as in the equation 

(a), No 1. Moreover, Oi, a.^, a, will take the values belonging to 

the extreme limit before marked with two accents, by simply assigning 
to h an infinite value. 


The easiest way of transforming the equation (2) will be to remark, 
that it is the general one which presents itself when we apply the 
Calculus of Variations to the quantity (5), in order to render it a 
minimum. We have therefore in the first place 

and by the ordinary formula for the transformation of multiple integrals, 
dx.dx, dx,du=^^^^ (l-2r' ^') d^,dl,...dldh. 

• But since 1 - 2,'+' ^^ = v + ^»S,'+' ^, 

a; ' Ur 

the expression (5) after substitution will become 

fd^^d^i d^sdhui tti ih a.A""'!/""'"'. 

Applying now the method of integration by parts to the variation of 
this quantity, by reduction, we get for the equivalent of (2) the equation 

^- dh^^ V" ,^ a;) hdh^^^ ^^''^ ar'dl' ^' "" ^^^a:-dlr' 
+ A^2^ X 2-^^ - A'22-Ml -^^ (2") 

where the finite integrals are all supposed taken from r = l to r = * + l, 
and from r' = 1 to r' = * + 1. 

The last equation may be put under the abridged form, 

d^ . ( ^«:^ dV 

provided we have generally 

o = -^+(»-s5-)^ + vr (n. 


coefficient o£^-mvV=~ {1 -^^'-2.'*' 1^ ^ "' + ^ ^"}> 
coefficient of ^, i» V ^ = - ^. lU 

coefficient of -j^ in vF=-^|-» + 2 ^ ^i. 

Moreover, when we employ the new variables 

du " y- ^ ^ . ; • Y' a? d^r dh ]' 

and therefore the condition (9) in like manner will become 

— -(>-^r"i^ff-^} «'»^ 

where the values of the variables ^1,^2, ?, must be such as satisfy 

the equation i;" = 0, whatever h may be; and as n-s-\-l is positive, it 
is clear that this condition will always be satisfied, provided the partial 
differentials of V relative to the new variables are all finite. 

8. Let us now try whether it is possible to satisfy the equation 
(2'") by means of a function of the form 

r^Hct> (/?); 

H depending on the variable h only, and cp being a rational and entire 
function of ^1, f^, f, of the degree 7, and quite independent of h. 

By substituting this value of V in (2'") and making 

^ d'H ( ^«:^ dH , „ ,,„^ 

we readily get 

= v<^ - '«P (18); 

where, in virtue of (17) k must necessarily be a function of h only; 
and as the required value of (p, if it exist, must be independent of k, 
we have, by making h = in the equation immediately preceding, 

= v'0 - ko(p (19); 

ko being the value k, and v'^ that of v^ when h = 0. 


We shall demonstrate almost immediately that every function ^ of 
the form (20), No. 9, which satisfies the equation (19), and which there- 
fore is independent of h, will likewise satisfy the equation (18); and 
the corresponding value of k obtained from the latter being substituted 
in the ordinary differential equation (17), we shall only have to integrate 
this last in order to have a proper value of V. 

9. To satisfy the equation (19) let us assume 

<^ = ^(e.^ ?/, ?3^ ?/)?,.?„ &c (20); 

F being the characteristic of a rational and entire function of the 
degree 2y', and the most general of its kind, and f,, ^„ &c. designating 
the variables in which are affected with odd exponents only; so that 
if their number be v we shall have 

7 = 27' + c, 

the remaining variables having none but even exponents. Then it is 
easy to perceive, that after substitution the second member of the 
equation (19) will be precisely of the same form as the assumed value 
of (p, and by equating separately to zero the coefficients of the various 

powers and products of ^1, |s, ^,, we shall obtain just the same 

number of linear algebraic equations as there are coefficients in <p, and 
consequently be enabled to determine the ratios of these coeflScients 
together with the constant quantity ^0. 

In fact, by writing the foregoing value of (p under the form 

</) = aS'^„„„, „„?.•"' ?."» ?»•" (20'); 

and proceeding as above described, the coefficient of ^ri ^/"t ^,'', 

will give the general equation 

K + 2)(m. + ]) 

^..i K+2)K + i) . 



the double finite integral comprising all the values of r and r, except 
those in which r = r , and consequently containing when completely 
expanded s (s - 1) terms. 

For the terms of the highest degree 7 and of which the number is 
7 + 1.7 + 2 y + s — l _ ^ 

the last line of the expression (21) evidently vanishes, and thus we 
obtain JV distinct linear equations between the coefficients of the degree 
7 in <p and ko. 

Moreover, from the form of these equations it is evident that we 
may obtain by elimination one equation in ko of the degree JV, of 
which each of the iV roots will give a distinct value of the function 
(p^'y\ having one arbitrary constant for factor; the homogeneous function 
^''1'' being composed of all the terms of the highest degree, 7 in (p. 
But the coefficients of (p'-^'' and kg being known, we may thence easily 
deduce all the remaining coefficients in (j>, by means of the formula (21). 

Now, since the A'' linear equations have no terms except those of 
which the coefficients of ^'^^ are factors, it follows that if ^0 were taken 
at will, the resulting values of all these coefficients would be equal to 
zero. If however we obtain the values of N' — 1 of the coefficients 
in terms of the remaining one A from iV- 1 of the equations, by the 
ordinary formulas, and substitute these in the remaining equation, we 
shall get a result of the form 


where jRT is a function of ka of the degree iV. We shall thus have 
only two cases to consider : First, that in which A = 0, and consequently 
also all the other coefficients of 0*^' together with the remaining ones 
in <p, as will be evident from the formulae (21). Hence, in this case 

= 0: 

Secondly, that in which kg is one of the iV roots of = K, as for 
instance, ko in this case all the coefficients of will become multiples 
of A, and we shall have 

Vol. V. Part III. S« 


(j) = Acpr. 

01 being a determinate function of ^,, ^a, E<- 

We thus see that when we consider functions of the form (20) 
only, the most general solution that the equation 

= v'^ - *o'0 (19') 

admits is 

or, (p = 0; or, (p = atp; 

a being a quantity independent of ^,, ^2, ^„ and (p any function 

which satisfies for <p to the equation (19'). But by affecting both sides 
of the equation 

with the symbol v, we get 

= V • v' - *o' . V ^ ; 

and we shall afterwards prove the operations indicated by v and v' 
to be such, that whatever may be, 

V v' = V' V 0- 

Hence, the last equation becomes 

v' (v ^) - k„' V (p; 

and as V like (p is of the form (20), it follows from what has just 
been shewn, that 

either = v cp, or, \7 (p = acp, 

a being a quantity independent of ^i, ^2, ?«• 

The first is inadmissible, since it would give ^ = 0; therefore when 
(p satisfies (19'), we have 

V 0' = a(p, i.e. = V — "0- 

But since a is independent of ^1, ^2, Bs, this last equation is 

evidently identical with (18), since the equation (18) merely requires that 
K should be independent of fi, ^2, ^s- 


Having thus proved that every function of the form (20) which 
satisfies (19) will likewise satisfy (18), it will be more simple to deter- 
mine the remaining coefficients of (j> from those of cp^^^ by means of 
tlie last equation, than to employ the formula (21) for that purpose. 

Making therefore h infinite in (18), and writing ~ in the place 
of K, we get 

where (22) comprises the — ^ — —!■ combinations which can be formed of 


the s indices taken in pairs. 

If now we substitute the value of before given (20'), and recol- 
lect that for the terms of the highest degree we have 2»«r = 7, we shall 
readily get 

0=(7-2»«,)(7+2»?r+»-l)^™,,»,,....,+(7».+2)(»w,+l)^„^, „^+2,...„^...(22), 

from which all the remaining coefficients in will readily be deduced, 
when those of the part 0'^' are known. 

10. It now remains, as was before observed, to integrate the ordi- 
nary differential equation (17) No. 8. But, by the known theory of 
linear equations, the integration of (17) will always become more simple 
when we have a particular value satisfying it, and fortunately in the 
present case such a value may always be obtained from by simply 

changing f, into ' , . In fact if we represent the value thus ob- 

tained by Ho we shall have 

cih ^' </e/«v(2«:')' 

and by a second differentiation 



(22) as before comprising all the ~ — — combinations of the * indices 
taken in pairs. 

Hence, the quantity on the right side of the equation (17), when 
we make H = Ho, becomes 

+ 2(22)^^.-^;^,+ («-2^)2^.--^ (23). 

d^rd^r a,«r'2«/ V «r / «^r o, 'v/(2«;^) ^ 

But if we recollect that we have generally 

it is easy to perceive that in consequence of the equation (18) the 
quantity (23) will vanish, and therefore the foregoing value of Ha 
will always satisfy the equation (17). 

Having thus a particular value of H, we immediately get the 
general one by assuming 

H= Hfzdh. 

In fact, there thence results 

H = KHj „„ '^"^^ . 

±l(,~ Ux, Oi, (h a, 

the two arbitrary constants which the general integral ought to con- 
tain being K, and that which enters implicitly into the indefinite in- 
tegral. But the condition = V" requires that H should vanish when 
h is infinite, and consequently the particular value adapted to the 
present investigation is 

n - jr rr f ^""dh 

J^ Mo'a^, «2 «» 

11. The values of (f> and H being known, we may readily find 
the corresponding values of V and p. For we have immediately 


r=i/^ = jir^//./-gj£2^ «6), 

and as the function (p is rational and entire, and the partial differen- 
tial of f^ relative to h is finite, it follows that all the partial differ- 
entials of F^ are finite; and consequently, by what precedes (No. 7.) 
the condition (9') is satisfied by the foregoing value of F', as well as 
the equation (2) and condition = F". Hence the equations {b) and 
(c) No. 6 will give, since 

du- "V ^' ~^) Y" ^■d^~~dh\' 

and h must be supposed equal to zero in these equations 

- r f^^ii) 

p' = — , A. .---V^-^^ (where h = 0); 

since where A = 0, a, = «/ ; and therefore, 

1 - 2/^' ^^ = 1 - ^r' V = ^'. 

If now we substitute for V its value (26), and recollect that « — * + 1 is 
always positive, we get 

-r(^) ^ 



since it is clear from the form of Ho that this quantity may always 
be expanded in a series of the entire powers of A^ In the preceding 
expression, (27), H^ indicates the value of Ho when h = 0, and (p! 
the corresponding value of or that which would be obtained by 

simply changing the unaccented letter fi, ^2, ^, into the accented 

ones ^1', f/, ?/ deduced from 

(7) x; = a,'?/ ; x.^ = «; ^/ ; x/ = «/ ^;. 


It will now be easy to obtain the value of V corresponding to 

without integrating the formula (1) No 1, where F is the character- 
istic of any rational and entire function. In fact it is easy to see from 
what precedes (No. 9), that we may always expand JF' in a finite series 
of the form 

F{xl, x-l xl) = bo^o + ii0i' + bo(p2 + 63^3' + &c. 

after a;/, x-J, &c. have been replaced with their values (7). Hence, we 
immediately get 

p' = „"-«-' . {bo(po' + b,<p! + h(p; + &c.} (29). 

By comparing the formulae (26) and (27) it is clear that any term, 
as 5,0/ for instance, of the series entering into p, will have for cor- 
responding term in the required value of V, the quantity 

^ ^ i^„'«/< a:.b.<pM.f-j j^^'"/^ ^ (30): 

''co -'^O "1 Ms (Is 

Ha being a particular value of H satisfying the equation (17), and 
immediately deducible from (p by the method before explained. 

12. AU that now remains, is to demonstrate that 

V'V0 = VV> (31), 

whatever <p may be. For this purpose let us here resume the value 
of A0, as immediately deduced from the equation (2") No. 7, viz. 

+ A^2^-A^2lx2i^ (32), 

P /w-1' 


where for simplicity the indices at the foot of the letters ^ and a have 
been omitted, and their accents transferred to the letters themselves. 
Moreover all the finite integrals are supposed taken from 1 to «+l. 

By making A = in the last expression we immediately get v'<^, 
and if for a moment, to prevent ambiguity, we write h, in the place 
of the original «'r and omit the lower indices as before, we obtain 

V>=(l-2a2^4p+(*-«-l)2|^ (33); 

where to avoid all risk of confusion r has been changed into r" , and 
the double accent of this index transferred to the letters ^ and h 

We will now conceive the expression (32) to be written in the 
abridged form 

the order of the terms remaining unchanged. 

If then we recollect that the accents have no other office to per- 
form than to keep the various finite integrations quite distinct, and 
consequently that in the final results they may be permuted in any 
way at will, we shall readily get 

V'Va^ — VaV'0 = 

(l-Sf)(4S2 ,^„,„ .-T^>-' +22-^, X 2^-^ I 
+ 42^'x22-^.-,p^^, +22^x2^x2^-^ 

^ ' [ a^b^ aa^\^^^ a^ c^V af (n) J 

(1 - 2P^|-422— i^^^^— _22-L ^ 1 

P g rf'0 P (/'^ 

2(l-2f)2JL.^ +22i:x2^^^ 

-2 (1-2^2^x2— ^ -2 2-x2^x2-^ 

all the finite integrals being taken from ?• = 1 to r = s + 1, and from 
/ = 1 to r' = s + 1. 

In order to obtain the required value 

v'v^ - w'<p, 

it is clear that we shall only have to add the first of the five preceding 
quantities to the sum of the four following ones multiplied by A', and 
to render this more easy, we have appended to each of the terms in the 
preceding quantities a number inclosed in a small parenthesis. 

Now since the accents may be permuted at will, and we have like- 
wise or = b^ + U, it is easy to see that the terms marked (1), (6) and 

(12) mutually destroy each other. In like manner, (2), (3), (7) and 
(18) mutually destroy each other; the same may evidently be said of 

(13) and (16), of (15) and (17), of (9) and (19), and of (8) and (14). 
Moreover the four quantities (4), (5), (10) and (11) will do so likewise, 
and consequently, we have 

V'V0 - VV> = 0. 
Hence the truth of the equation (31) is manifest. 


Application of the preceding General Theory to the Determination of the 

Attractions of Ellipsoids. 

13. Suppose it is required to determine the attractions exerted by 
an ellipsoid whose semi-axes are a', b', c' whether the attracted point 
p is situated within the ellipsoid or not, the law of the attraction being 
inversely as the w"*" power of the distance. Then it is well known 
that the required attractions may always be deduced from the function 

j^ _ r p' dx' dy' dx 

{{x ~x'f + {y-yj + {x-%jy^ ' 

p being the density of the element dx' dy' d%' of the ellipsoid, and 
X, y, % being the rectangular co-ordinates of p. 

We may avoid the breach of the law of continuity which takes 
place in the value of V, when the point p passes from the interior of 
the ellipsoid into the exterior space, by adding the positive quantity 
M* to that inclosed in the braces, and may afterwards suppose u eva- 
nescent in the final result. Let us therefore now consider the function. 


p' dx' dy' d%' 

{{X - x'y + (y- y'y + (z- zy + M^p ' ' 

this triple integral like the preceding including all the values of x', tf, »', 
admitted by the condition 

,/2 ^-^ 

— + — + — Z 1 

If now we suppose the density /o' is of the form 

f^'^i^-T^^-h-z^ ' /(^',y.«') (34). . 

which will simplify / {x', y, »') when p is constant and n' = 2, and then 
compare this value with the one immediately deducible from the general 
expression (28) by supposing for a moment n' = n, viz. 

Vol. V. Part III. 3 1 


we see that the function f will always be two degrees higher than F. 
But since our formula become more complicated in proportion as the 
degree of F is higher, it will be simpler to determine the differentials 
of V, because for these differentials the degree of F and f is the same. 
Let us therefore make 

, _ 1 dV _ r /o' (« — x) dx dy' d % 

~ 1 m' fir ~ J >rTi ' 

i,ia;-x'Y + iy-i/r + {z-zr + u''} — 

then this quantity naturally divides itself into two parts, such that 

A =xA' + A", 

, ,, /- p dx dy d% 
where A' — -^r J '^ ;;rr\ , 

{{x -x'Y + {y- yj + [%-%)' + u^}~ 

and A"=-f~ 

x'p dx' dy' dx 

{{x-xy + {y-yy + {z-%y + u^~ 

By comparing these with the general formula (1), it is clear that 
M — 1 = n' + 1, and consequently n = n + 2. In this way the expression 
(28) gives 

which coincides with (34) by supposing F=f. 

The simplest case of the present theory is where y(a;', y', x') = l, and 
then by No 11, we have 0o'= 1 and &„ = 1. when A is the quantity 
required, and as the general series (29), No 11, then reduces itself to 
its first term, we immediately obtain from the formula (30), the value 
of A! following, 

* A= , — -- (the \ — 7 — (35), 

because in the present case H^, = 1, « = 3, and n = n' ^ 2. 

Again, the same general theory being applied to the value of A" 
given above, we get 


F («', y', %') = - x'fix, y', a') = - ar' (when / = 1), 

and hence by Noll, F{x',y',%') — — a'l'. In this way the series (29) 
again reduces itself to a single term, in which 

and the particular value H^ corresponding thereto, by omitting the super- 
fluous constant /(fj2,i,'i — tk will be (No 10), 

Ho — a. 
These substituted in the general formula (30) as before, immediately give 

A" . ^""^^UJ ,3,, ,„ c ^'-'^dh 
A = -^ ; — 7- « o C Pa / — rr — , 

and consequently by reduction since a^ = x, 

A=xA' + A''=-^-l-^r^l a'b'c'x f^^ (36). 

r ( "^ ] " 

The value of A just given belongs to the density 

' _ fi _ ^" y! _ ?!\^ 

Hence we immediately obtain without calculation the corresponding 

1 dV_ ^^^^U) ,,, , f h^-'^dh 

1 dV *'"'' (2) ,-, , rU-'^'dh 

2^5 r 

C = :; , -y— = ;f — r" abc% / — , ^ . 


31 2 


If now we suppose moreover 

__ 1 d^ _ f f> dx dy d% 

the method before explained (No 11), will immediately give 

Z) = — , — T— a cu / T — , 

p/ w'+ 1 \ J^ abc 

and therefore if for abridgment we make 

-'^d') .,, 

the total differential of V may be written 
rfr=i»f{2^rf^/^-^^ + 2s.rfy/^-^^ + 2.t/./^— j-^ + 2«rf./^-^^}, 

which being integrated in the usual way by first supposing h constant, 
and then completing the integral with a function of h, to be after- 
wards determined by making every thing in F variable, we get 

A being a quantity absolutely constant, which is equal to zero when 
w' > 1. What has just been advanced will be quite clear if we recollect 
that h may be regarded as a function of x, y, % and u, determined by 
the equation 

» = ?^' + 4^ + ?4^ ^- 1 - f + -' + f' -^ ' <''^>' 

seeing that a' = a'*-\-h\ V = b'' + h\ and & = c" + h\ 

After what precedes, it seems needless to enter into an examination 
of the values of V belonging to other values of the density p, since 
it must be clear that the general method is equally applicable when 


where f is the characteristic of any rational and entire function. 

The quantity A before determined when we make u = 0, serves to 
express the attraction in the direction of the co-ordinate x of an ellipsoid 
on any point p, situated at will either within or without it. But by 
making « = in (37) we have 

, _ a^ y" z' ^ 

a" + h' "*■ b'-' + li' '^ c" + h' ^ h' ^^ ^' 

and it is thence easy to perceive that when p is within the ellipsoid, 
h must constantly remain equal to zero, and the equation (38) will always 

be satisfied by the indeterminate positive quantity — . When on the 

contrary p is exterior to it, h can no longer remain equal to zero, but 
must be such a function of x, y, %, as will satisfy the equation (38), of 
which the last term now evidently vanishes in consequence of the 
numerator o'. Thus the forms of the quantities A, B, C, D and F" 
all remain unchanged, and the discontinuity in each of them falls upon 
the quantity k. 

To compare the value of A here found with that obtained by the 
ordinary methods, we shall simply have to make n' = 2 in the expression 

(36), recollecting that r(l) = 1, and r (-] =i\/7r. In this way 

, .,,,,/- hdh ^ ,,, , r da 

A = — Aiiraoc X \ -rrr- = — 4nrab c x / -7^— 
J^ctbc J„drbc 

= + ^a'h'c'x f 4?- = 4-«'*'c' J . , f " ^ 

But the last quantity may easily be put under the form of a definite 

integral, by writing - in the place of a under the sign of integration, 

and again inverting the limits. Thus there wiU result 
J 47r«'J'c' /•! v"dv 

^ = 'n^~ J 

a •'o 

a + ^«^)(i + ^-/-^^) 

a' ' a- 


which agrees with the ordinary formula, since the mass of the ellipsoid 




and «^ = d^ + h\ 

Examination of a j)articular Case of the General Theory exposed in the 

former Part of this Paper. 

14. There is a particular case of the general theory first considered, 

which merits notice, in consequence of the simplicity of the results to 

which it leads. The case in question is that where we have generally 

Avhatever /• may be 

a/ = a. 

Then the equation (19) which serves to determine 0, becomes by 
supposing kn = k . a"' 

= il-'2r'^r')^r'^, + (.v-»-l)2/*'?.^-A<^ (39). 

If now we employ a transformation similar to that used in obtaining 
the formula (14), No 6, by making 

^i = P cos 9i, ^2 = p sin 9i cos On, ^3 = p sin 0, sin 9.^ cos 63, &c. 

and then conceive the equation (39) deduced from the condition that 

'"^'"i- "f-d-sf =)'^ {^■*' (f )' - r^l 

must be a minimum (vide No 8), we shall have 

rf^,rf^2 c?f,, = p'-' sin0/-^ sina/-^ sin 9^-, dp de,d9, «?0,_„ 

\d^rl \dp) p" ' sin0,^sin0/ sine^,-,' 

and 1 - 2^,= = 1 - p\ 

Proceeding now in the manner before explained, (No 8), we obtain 
for the equivalent of (39) by reduction 

d''(j) , ( _ _ , V cos g,. d(f> 
d'<p s-l-np' dcl> 1 d9;''^^^~'' Um9rd9,. k 

dp' ^ p{l-p') 'dp p' ' sm9,'sm9^' sin0^,., 1 _^2 9-V*U). 

But this equation may be satisfied by a function of the form 

= Pe,e,e3 e,_,; 

P being a function of p only, and afterwards generally 6, a function 
of dr only. In fact, if we substitute this value of (p in (40), and then 
divide the result by ^, it is clear that it will be satisfied by the system 


— ^<- 1 

d'Q,.^ , _ cos0,_2 o?e,_2 , X._, ^ .^^^ 

Os-2 de^._s ' sin 0,_2 9s_2</0s_2 sin 6^,-2 

+ 2 . -7— -p^ — 77 j7i h -; — —^ — = X,- 

B._3d0\.3 ' sin 9^-3 Qs-3d9,_3 sin0^ 

&;c. &c. &c. &c. 

combined with the following equation, 

d'P s-l-np' dP \, k ^ 

P^p" ^ /" (1 -p') ■ P«?/' />' 1 -p' 

where k, X,, X^, X3, &c. are constant quantities. 

In order to resolve the system (41), let us here consider the general 
type of the equations therein contained, viz, 

- ^'Q'- , (r-i\ ^"^^-- '^^^ + ( ^'-^' X "i ft 
d9\_,. ^ >sm9,_/d9,.r \sm9\., a,_,j «,.,. 

Now if we reflect on the nature of the results obtained in a preceding 
part of this paper, it will not be difficult to see that 6,_r is of the form 

e,_. = (sine._,)*;j = (1-M^)«;>; 

where j9 is a rational and entire fimction of m = cos0s_r, and / a whole 

By substituting this value in the general type and making 

\..r^i = - i{i + r - 2) (43) 

we readily obtain 

= {1-M.')^: -{2i + r)^^- {X._. + i{i + r-l)}p. 


To satisfy this equation, let us assume 

Then by substituting in the above and equating separately the coefficients 
of the various powers of yu, we have in the first place from the highest 

X._, = - e{e^-r—\) (44), 

and afterwards generally 

. e-i-9.t .e-i-M-\ . 

'*' ~ ~ 2/ + 2x2e + r-2#-3 " 

But the equation (43) may evidently be made to coincide with (44), by 
writing «*''' for i, and t^''+'^ for e, since then both will be comprised in 

\,_,+, = - e*--' {e<'-* + r-2| (45). 

Hence we readily get for the general solution of the system (41), 

"^ 2.4 X {2f<'-> + r-3|{2«"-' + r-5} " - &C.J ; 

where w = cos 9,_r, and i*''* represents any positive integer whatever, pro- 
vided ^''■' is never greater than ^*'■*". 

Though we have thus the solution of every equation in the system 
(41), yet that of the first may be obtained under a simpler form by 
writing therein for X^.i its value — i® deduced from (45). We shall 
then immediately perceive that it is satisfied by 

cos [ J 

In consequence of the formula (45), the equation (42) becomes 

^- dp' ^ pO-p') dp \ / '^T^'i^' 

which is satisfied by making ^= —\, -(«'*' + 2ft)) (e"*'' + 2a) + w — l), and 


p = „i") f„2. _ 2a>x2i'"+2^ + v?-2 o„_2 
t' -(H 2 ^ 2i''» + 4w + »-3 ^ 

, 2«) . 2a)-2 X 2i'~' + 2ai + *-2 .2i'^' + 2w + *-4 , - „ , 
2.4 X 2«<-' + 4to + «-3,2« + 4w + w-5 ^ ' 

where w represents any whole positive number. 

Having thus determined all the factors of (f>, it now only remains 
to deduce the corresponding value of H. But Ho the particular value 
satisfying the differential equation in H, will be had from by simply 
making therein 

since in the present case we have generally «/ = «'. 

Hence, it is clear that the proper values of 0,, di, 9^, &c. to be here 
employed are all constant, and consequently the factor 

0102 ©3 ©s-l 

entering into (f> is likewise constant. Neglecting therefore this factor 
as superfluous, we get for the particular value of H, 


since ,0^ = ?.= + ?/ + + ?/ = ^^ = fj , 

ga a 

and Pa represents what P becomes when p is clianged into — . 

a' ^ 

Substituting this value of Ho in the equation (25), No 10, there 
results since a' = a'^ + h^ 

H=K.P„ f ^'"""^^ , (46) 


K being an arbitrary constant quantity. 

Thus the complete value of V for the particular case considered in 
the present number is 

Vol. V. Part III. 3 K 


v= pe.e,......e..,.irp« f "^ , (47) 


and the equation (27), No 11, will give for the corresponding value of p', 

in - IN 




where P/, 9/, 62', &c. are the values which the functions P, 0i, 02, &c. 
take when we change the unaccented variables fi, ^2, ^, into the cor- 
responding accented ones ^/, ^/, f/, and 

p «-^ + l-w — ^ + 3 n — s + 2a}-l 

' ~ » + 2i + 2ft)-l .7^ + 2^ + 2(0 + 1 n + 2i + 4<w-3' 

or the value of P when p = 1 ; where as well as in what follows i 
is written in the place of i'''. 

The differential equation which serves to determine H when we 
introduce a instead of h as independent variable, may in the present 
case be written under the form 

. = a=(a^-«'^) Vr + «M»«'-(*- !)•«"} ^ 
^ ' dcf * ' ada 

+ {?■(« + *- 2) a'' -(« + 2ft.)(« + 2a) + w-l)a'} H, 

and the particular integral here required is that which vanishes when 
h is infinite. Moreover it is easy to prove, by expanding in series, that 
this particular integral is 


provided we make the variable r to which A" refers, vanish after all 
the operations have been effected. 

But the constant k' may be determined by comparing the coefficient 
of the highest power of a in the expansion of the last formula with 
the like coefficient in that of the expression (46), and thus we have 

" yfc' = Kd'^"" (-\Y « + 2^' + 2a)-l.w + 2? + 2a> + l ?^ + 2^^ + 4a,^.-3 

^ ^ 2.4 . 6 2o) . 


Hence we readily get for the equivalent of (47), 

rr vtc^ c^ ^ « + 2« + 2a)-l .M + 2e + 2a)4-l n + 2i + 4im-3 

2.4. 6 2w 

xKa'^'-'''"{-l)''a'A''a"-fdaa'-"'-'-"{a^-a"') ^ 

■ ■ GO 

In certain cases the value of V just obtained will be found more 
convenient than the foregoing one (47). Suppose for instance we repre- 
sent the value of f^ when h = 0, or a = a' by V^. Then we shall hence 
r^ i»c> o o n + 2i+2a)-l .n + 2i + 2w + l » + 2i + 4ft.-3 

2.4,6 2a> 

g — l~n~Suo 


which in consequence of the well known formula 


/"'a-'da (a' - a'')-" = - «''-"-^? x -J^ i , 

by reduction becomes 

fl+s — n\^[n + 2i + 4!w — l'^ 


2r(a,+ i)r( ^ + ^'^+^" ) 

since in the formula (5), r ought to be made equal to zero at the end 
of the process. 

By conceiving the auxiliary variable u to vanish, it will become clear 
from what has been advanced in the preceding number, that the values 
of the function P within circular planes and spheres, are only particular 
cases of the more general one, (49), which answer to * = 2 and s = 3 
respectively. We have thus by combining the expressions (48) and 
(49), the means of determining Vo when the density p is given, and 
vice versa; and the present method of resolving these problems seems 
more simple if possible than that contained in the articles (4) and (5) 
of my former paper. 



XVI. On the Position of the Axes of Optical Elasticity in Crystals 
belonging to the Ohlique-Prismatic System. By W. H. Millek, 
A.M. Fellow and Tutor of St John's College, and Professor of 

[Read Dec. 8, 1834.] 

1. Fresnel has proved that whatever be the regular arrangement 
of the medium which by its elasticity produces the optical properties 
of a crystal, there are always three directions at right angles to each 
other, which may be considered as axes of optical elasticity. This 
being understood, it is further already established, that crystals belong- 
ing to the tesseral system have three equal axes of optical elasticity ; 
that rhombohedral and pyramidal crystals have two axes of elasticity 
equal to each other and perpendicular to the crystallographic axis, 
which therefore is the third axis of elasticity and also an optic axis; 
and that crystals belonging to the remaining systems have three unequal 
axes of elasticity, and consequently two optic axes (that is, axes of 
optical phenomena) making with each other angles which are bisected 
by the axes of greatest and least elasticity. 

Sir David Brewster, who discovered the mutual dependence of the 
forms and optical properties of crystals, has determined the angles be- 
tween the optic axes of a great number of biaxal crystals; his obser- 
vations, however, do not contain any data from which the positions of 
the axes with respect to the faces of the crystals can be found. 

2. In the right prismatic system the axes of elasticity coincide (as 
might have been expected) with the rectangular crystallographic axes. 
In the oblique prismatic system, if the three axes be XX', YY', ZZ', 
the crystallographic axis {YY'), which is perpendicular to the other 
two {XX', ZZ'), is always one of the axes of elasticity. This, in 
Gypsum, at the ordinary temperature of the air, and in many other 
crj'stals, is the mean axis, or it is perpendicular to the optic axes; in 


Borax, Acetate of Soda, Felspar, Tartaric Acid and Gypsum, when heated 
to about lOO^C, as was first observed by Mitscherlich, it is the greatest 
or least axis of elasticity, and is therefore in the same plane with the 
optic axes and makes equal angles with them. 

The position of one axis of elasticity having thus an evident rela- 
tion to the crystallographic form, we are naturally led to inquire if 
any relation can be discovered between the other two axes of elasti- 
city and the crystallographic form. The only attempts to discover any 
such relation, with which I am acquainted, are those of M. Soret, 
(Memoires de la Socidte de Physique de Geneve, tome I.) and Pro- 
fessor Neumann of Konigsberg (Poggendorff's Annalen, B. xxvii. S. 
240). Neumann shews, that in Gypsum the axes of elasticity and also 
the thermal axes, or the three lines in the crystal which remain at 
right angles to each other at all temperatures, constitute a system of 
rectangular crystallographic axes. It appeared at first sight not im- 
probable that a similar relation might be found to exist between the 
form and axes of elasticity of other oblique-prismatic crystals. Though 
my observations appear to disprove the law which has thus been sug- 
gested, they do not establish any other in the place of it. The only 
general fact which I have noticed is, that in many instances, though not 
in all, one of the two axes of elasticity which are perpendicular to 
W, is also the axis of one of the principal zones of the crystal. 

3. To find the angle between a normal to any face {T) of a 
crystal, and the apparent direction of one of the optic axes as seen in 
air through any parallel faces of the crystal. 

Let the crystal be attached to an index, moveable on a graduated 
circle having its plane parallel to the axis of the polarizing instrument, 
or a table on which the position of the index may be marked by a 
line drawn along its edge with a tracing point. Let the crystal be 
placed in such a position, that the apparent direction of the optic axis 
in air and a normal to T may be parallel to the circle. Move the 
index till the center of the coloured rings coincides with a mark in 
the axis of the polarizing instrument, and observe the points in which 
it meets the circle. Turn the crystal half round in the plane of T, 


taking care not to alter the inclination of T to the index, (this may 
be effected by moving the crystal, the index being fixed, tiU the image 
of some well defined object seen by reflexion in T appears in the 
same direction after the crystal is turned as it did before.) If the 
index be now turned till the center of the coloured rings coincides 
with the mark, the angle it has described between the observations 
will be manifestly equal to twice the angle between the apparent 
direction of the optic axis in air and a normal to T. The angle 
between the optic axis in air and a normal to any other known face 
of the crystal being found in the same manner, the direction of the 
optic axis in air wiU be completely determined. 

4. To find the optic axes, their apparent directions in air being 

Let Qlt, Q'K (Fig. 1.) be tangents to the circular and eUiptic sec- 
tions of a wave diverging from O made by a plane through the optic 
axes, and therefore OQ, OQ', perpendiculars to QB, will be the optic 
axes; OP the direction in which the optic axis OQ is seen in air; 
OS a perpendicular to the faces through which it is seen. 

The vibrations in that part of the wave which has a circular sec- 
tion are perpendicular to the plane QOQ, consequently a ray polar- 
ized in the plane QOQ is refracted in that plane according to the 
law of sines. Let m be the ratio of the sine of incidence to the sine 
of refraction for such a ray out of air into the crystal, D the mini- 
mum division of the ray when refracted in the plane QOQ' through 
the prism formed by two natural or artificial planes meeting at an 
angle / in a line perpendicular to QOQ. Then ^ sin ^ / = sin ^ (Z) + /), 
and fM sin QOS = sin POS. Whence the direction of QO is known. 
0*0 being found in the same manner, the axes of elasticity O^, Oi[, 
which bisect the angles qOQ, QOQ, are also known. 

5. The diagram which accompanies the description of each crystal, 
is the representation of a sphere, to the surface of which the faces of 
the crystal are referred by means of perpendiculars drawn from the 
center of the sphere. The point in which the perpendicular to any 


face meets the surface of the sphere, will be called the pole of that 
face. The measurements express the angles between the perpendiculars 
to the faces, or the supplements to the angles between the faces them- 
selves. This method of representing crystalline forms appears to have 
been first employed by Neumann, in his Beitrage zur Krystallonomie, 
and afterwards by Grassmann and Uhde. It has the advantage of ex- 
hibiting all the faces of a crystal without confusion in one figure, 
each zone being distinguished by a great circle drawn through the 
poles of the faces composing it, and also of allowing all the requisite 
calculations to be performed by spherical trigonometry applied to the 

T cos PX = T cos PY = 7 cos PZ, 
h k I 

or to formulse deduced therefrom, X, Y, Z being the points in which 
radii parallel to the axes of the crystal meet the surface of the sphere, 
and P the pole of the face {h; k\ l), which is parallel to the plane 

h- + k\-^ I- = 0. 
a b c 

ad, /3/3', ^f, ^^' will be used to denote the extremities of diameters 
drawn parallel to the optic axes, and the two axes of elasticity which 
are perpendicular to YY' . In Figs. 5, 6, 7, 8 the faces are denoted 
by the same letters as in the treatises of Mohs and Naumann. The in- 
clinations of the faces of crystal (1) and (2) are deduced from a mean 
of the best measurements of thirty or forty crystals, and are probably 
within 1' of the truth. 

The chemical notation and atomic weights are those employed by 
Dr Turner, in the fifth edition of his Elements of Chemistry. 




(1). Sulphate of Oxide of Iron and Ammonia. According to Mit- 
scherlich (Jahresbericht 13), the composition of this salt, which belongs 
to an extensive plesiomorphous group, is expresssed by the formula 
H'^NS ■\- FeS -^ 1 H. Fig. 2. represents the poles of its faces. Their 
symbols are A{1; 0; 0), C(0; 0; 1), H{0; 1 ; 1 ;) M{1', 1; 0), 
P(l; 1; 1), Q(-l; 1; 1), T{2; 0; 1). 

AT 42", 14' 




129°, 18' 

MQ 42»,23' 

TC 64 ,34 


44 ,54 


25 ,21 

QC 34 ,20 

CA' 73 ,12 


109 ,36 


140 ,55 

TP 35,14 

AYP 68 ,12-1- 





MT 52 ,46. 

PYC 38 ,35i 




58 ,32 

When yellow light is refracted through the faces TC in the plane 
AC A', the minimum deviation of a ray polarized in the plane AC A', 
is 41", 26'. The apparent direction of the optic axis aa in air, when 
seen through the faces TT', makes an angle of 7°,10' with 2'2"; 
and the optic axes appear to be inclined to each other at an angle of 
79" when the crystal is immersed in oil, of which the index of re- 
fraction is 1,47. From these data we find Ta = 4'',47', Tfi = 71°,2', 
r^ = 33'',8', A^ = 9'',6'. 

Tan T^ is nearly equal to 4tan>4f. The value of A^ deduced from 
the equations tan Tf = 4tan^^, T^ + ^^ = 42°, 14' is 9°, 13'i. This 
would make Q = 82'',25'i. Now, 46" tan 9'^ 13'^ = tan 82'',22'| ; therefore, 
if we refer the faces T, A, C, to the rectangular axes ff, YV, ^^', 
neglecting the difference of 3' in the value of C^, their simplest symbols 
will be (1; 0; 1), (4; 0; -1), (2 ; ; -23). The magnitude of the last 
index renders the hypothesis that ^f , ^^' are crystallographic axes highly 

(2). The composition of Tartrate of Ammo7iia is expressed, according 
to Dulk, (Jahrbuch fiir Chemie und Physik, 1831. B. 1.) by the for- 
mula H^NT+^H. The poles of its faces are represented in Fig. 3. 

Vol. V. Paet III. sL 


A{i; 0; 0), C(0; 0; 1), H{0; 1; 1), 2^(1; 0; 1), L{-1', 0; 1), 
3f(l; 1; 0), P(l; 1; 1), Q(-l; 1 ; 1). 

Cleavage parallel to the face A. 

AK 520,31' AM 55»,2' Qd gr^a?' QA' 6o»,54'i 

KC 39,53 HH' 81, 4.6 QL 41 ,l6^ CP 55,34 

CL 38 CH 49 ,7 ^i* 63 ,22 PM 35 ,48^ 

L^' 49,36 PP' 94,55 PIT 28 ,12 JWQ' 34 ,53 

iIfil/'69,56 PX 42,32 /TQ 27,31i QC 53,44-i. 

Z> = 25'',17', the light being refracted through CK. The apparent 
angle in air between the optic axes aa and AA', is 4°,55'. In oil, 
the index of refraction of which is 1,741, the apparent angle between 
the optic axes =42'',20'. This gives ^0 = 8", 7', A(i = 35'',54>', A^=W,M', 
i'^ = 33°,12. 

In this case the positions of some of the faces A, K, C, L must be 
altered half a degree before they can be referred to the rectangular 
axes ff, W, ^^' with tolerably simple indices. 

(3). A solution of Benzoic acid in alcohol, when suffered to eva- 
porate, affords crystals of which the faces C, K, I (Fig. 4) alone are 
bright. Cifr= 69",25', C/=97'',20' nearly. Z) = 64°,45', refraction taking 
place through the faces CK. The apparent direction of ad in air 
when seen through CC makes with CC an angle of 4°, 30'. When 
immersed in oil of which the index of refraction is 1,471, the appa- 
rent angle between the optic axes is 75". Hence Ca =2'',47', C/3=59'',50', 
C^=28»,31', ^^=40^54'. 

tan K^, tan I^, tan C^ are nearly as the numbers 3, 1, 5. 

The equation 1 tan /iT^ = tan 7^ = |^ tan C^ is satisfied by making Cf= 
27°,56'^, JC = 97'',17'. Hence the faces C, I, K may be referred to the 
rectangular axes ^f, YY , l^ without greatly altering the observed 
angles, and their symbols will be (-1 ; 0; 5), (1 ; 0; 1), (1 ; ; 3) re- 


(4). In Felspar (Fig. 5.) the optic axes lie in the plane of the 
most perfect cleavage, and make with a normal to M, angles of about 
57" or 58°, (58^^ according to Sir David Brewster) which increase when 
the crystal is heated. Hence, ^^' is the axis of the zone PM. 

(5). The optic axes of Pyroxene (Fig. 6.) seen in air through a 
slice cut perpendicular to MM are in the plane Pr, and make angles 
of 16" with the axis of the zone MM. Hence, ^^' is the axis of the 
zone MM', a, /3 approach ^ when the crystal is heated. At ordinary 
temperatures a/3 is probably about 19"^. The best measurements of 
Pyroxene shew that Pr, tr are nearly but not exactly equal, and 
therefore, that its faces cannot be referred to ^^', YY', X,'C as crystallo- 
graphic axes. In all the crystals of Pyroxene which I have examined, 
the rings surrounding ad are brighter than the rings surrounding /3/3'. 

(6). The form of Borax (Fig. 7.) closely resembles that of Py- 
roxene ; its optic axes however are very differently situated. It was 
observed by Sir John Herschel and also by Professor Nbrrenberg, that 
the optic axes for different colours do not lie in the same plane. This 
being the case, we cannot expect to find any simple connexion between 
the form and the directions of the axes of elasticity. 

The mean directions of the axes seen in air through the faces 7'T" 
make angles of aO"^, with a normal to the faces TT', and a perpendi- 
cular to them makes an angle of 55° with MM'. The rings sur- 
rounding ad, /3/3' are indistinct on the sides towards M'P and MP' 
respectively, the extremities a, /3 of the axes being next to the eye of 
the observer. This shews that the positions of ^f', ^^' vary slightly 
with the colour of the light employed. 

(7). In Chromate of Oxide of Lead, as I have been informed by 
Professor Norrenberg of Tiibingen, the axis of the zone MM (see the 
figure in Phillips or Naumann) bisects the angle between the optic 
axes, and is therefore one of the axes of elasticity. The other two 
axes of elasticity are, without doubt, the lines which bisect the angles 
formed by normals to MM'. 



(8). In Epidote, (Fig. 8.) the optic axis aa seen in air through the 
faces r, r, makes with r r' an angle of 8^ 50', /3/3' seen in air through 
the faces M, M', makes with MM' an angle of Sl^SO'. The determina- 
tion of m is rendered difficult by the complete absorption of the light 
polarized in the plane MT. Assuming /x = 1,7, which is probably near 
the truth, we get /•a=5'',ll', M/3=18°,5'. According to Mohs 2V=51'',41', 
TM=64>'>,30', therefore, T'a=46",30', 2)3 = 46'\31'. Hence ^^' is the axis 
of the zone PT. The near approximation of the values of 7'a, Tfi to 
equality must be considered accidental, as the positions of the optic axes 
are usually uncertain to the amount of some minutes. 

The question whether any proposed lines are crystallographic axes 
must be decided, as has already been intimated, by the simplicity and 
symmetry of the numerical relations which the expression of the faces 
requires with reference to these axes. This according to the old Hauyian 
views of the structure of crystals, is equivalent to saying that the pri- 
mitive form must be such that the other forms can be derived from 
it by simple laws of decrement. Now, we find that by assuming the 
axes of elasticity to be crystallographical axes, we have in the crystal 
(1) a face (2; 0; —23), which though not very probable is not im- 
possible, and in (5) a face ( — 1; 0; 5); in (2) the observed and com- 
puted positions of some of the faces differ half a degree. 

In (6), the optical properties are not symmetrical. 

In (4), (5), (7), (8) one of the axes of elasticity f^' or ^^' is the 
axis of a zone. 

St John's Collegb, 
Dec. 8, 18S4. 


Tr^ff/.sojcHen.' (/^ Vu Cam^./'hjrZ. S&c. Vol.S /'i. 

Ik,. VI. 

Fi^. V. 

fi^. IT. 

AUuaiTe J.iVw9- ('■amfiritCj/e 

TntAMicbjonj o/'the.Qani. Thil.Soc. VfC A fl ji> 

Ft:^. I. 

rr^.M. ^^ 

Svg JI. i 

■Fi^ V 

X ^/A 

Ftg. iv: 

■V Mff nr. 

^ j'i^. im^. 

M^/jy^J]k. /iJfij firm/>rT/iii^.. 

^^ ^rvf^'N! 

B. » A c. ' ■ » k T*. ■ ■.■■■■ 


nnni ^^^^/ 

v.. AAAA^ ■ 


'. ^ v;^ 


•> uM 




v-.^'^.,^^ V 




A'^- . / o "> -ja , 




.>^^^^'^.^,^ r,^^^A''Y^' 

■--.' . .^ lY^^^^r:.^^-^^A^' 

"^ '..r^rn. 


■f^^^M'^-^^mr^^t&r^ ^...rm 

•^^ ' ^^W^C 


^Cf\A,^^/" Jt-^: