A geometrical height scale for sunspot penumbrae

Draft version July 19, 2010

Preprint typeset using WT^^ style emulateapj v. 04/20/08

O

1—5

Oh-

o

>

<N

o
o

^

A GEOMETRICAL HEIGHT SCALE FOR SUNSPOT PENUMBRAE

K. G. PUSCHMANN^'^, B. Ruiz COBO^'^ AND V. MARTINEZ PlLLET^
(1) Institute de Astrofisica de Canarias (lAC), E-38200 La Laguna, Tenerife, Spain
(2) Departamento de Astrofisica, Universidad de La Laguna (ULL), E-38205 La Laguna, Tenerife, Spain

Draft version July 19, 2010

ABSTRACT

Inversions of spectropolarimetric observations of penumbral filaments deliver the stratification of
different physical quantities in an optical depth scale. However, without establishing a geometrical
height scale their three-dimensional geometrical structure can not be derived. This is crucial in
understanding the correct spatial variation of physical properties in the penumbral atmosphere and
to provide insights into the mechanism capable of explaining the observed penumbral brightness. The
aim of this work is to determine a global geometrical height scale in the penumbra by minimizing the
divergence of the magnetic field vector and the deviations from static equilibrium as imposed by a
force balance equation that includes pressure gradients, gravity and the Lorentz force. Optical depth
models are derived from the SIR inversion of spectropolarimetric data of an active region observed
with SOT on-board the Hinode satellite. We use a genetic algorithm to determine the boundary
condition for the inference of geometrical heights. The retrieved geometrical height scale permits the
evaluation of the Wilson depression at each pixel and the correlation of physical quantities at each
height. Our results fit into the uncombed penumbral scenario, i.e., a penumbra composed of flux tubes
with channelled mass flow and with a weaker and more horizontal magnetic field as compared with
the background field. The ascending material is hotter and denser than their surroundings. We do
not find evidence of overturning convection or field free regions in the inner penumbral area analyzed.
The penumbral brightness can be explained by the energy transfer of the ascending mass carried by
the Evershed flow, if the physical quantities below z = -75 km are extrapolated from the results of the
inversion.
Subject headings: sun: sunspot, polarization, radiative transfer, methods: numerical

1. INTRODUCTION

Our understanding of the structuring of the penum-
bral magnetic field has increased significantly over the
last two decades. Three seminal works wer e published
back in 1993. Using filter magnetograms, iTitle et all
(|1993[ ) found that the magnetic field inclination fiuctu-
ated by as much as ±18 degrees (rnis) on azi muthal paths
around the spot center. Lites et aT] (| 19931 ). using spec-
tropolarimetric data, identified the same fiuctuations in
the magnetic field, but this time in both inclination and
field strength. The sense of the correlation found was
that horizontal field lines tended to a have a weaker field
strength. In both works, the well known Evershed fiow
was observed to be con centrated in the hori zontal compo-
nent. Simultaneously, iSolanki fc Montavon (1993) pro-
posed a model, commonly referred to as the uncombed
model, of vertically interlaced magnetic field components
to explain the properties of the observed net circular po-
larization.

The discovery of dark cored penumbral filaments by
iScharmer et al.l (J2002), once a resolution of '^0^1 was
reached, led to the development of new models try-
ing to explain the structure of penumbral filaments.
Among theni we can cite the fiux pumpi ng mechanism
(jWeisseTaT '2004'; 'Brummeh efalJllOOl), the MISMA
model (Sanchez Almeida 2005 ^), and the "g appy" model
fSpruit fc Scharmerll2006l : IScharmer fc SDruitl l2006). For
a detailed overvie w we refer to r ecent rev i ews o f
iBellot R^ubid (|2009l) . iBorrero I (|2009l) . IScharmerl (|2009h .

Electronic address: |kgp@iac.es, brc@iac.es, vmp@iacles]

ISchli chenmaierl (|2009[ ). and iTritschle rl (120091).

Many penumbral mode ls (ISchlichenmaier e t al.l

1998a,U; 'Marti nez PiTIetl 120001 : IBorrero et al.M2005u ■20'Q6[
Borrero 2007) share the geometry of the uncombed
penumbral model: nearly horizontal magnetic fiux tubes
embedded in a stronger and more vertical background
field. Throughout the article we will refer them as the
uncombed scenario. In the frame of this scenario, several
observational properties of sunspot penumbrae, like the
filamentary structure, the strong Evershed outfiow,
and the uncombed geometry can be easily explained.
Besides, it explains t he properties of the observed net

circular polarization (ISchlichenmaier fc ColladosI 120021 :
ISchlichenmaier et al.l 120021 : iMiiller et al.l I2002L I2006D .
Rec ently the uncomb e d scen ario has been modified,
see IBorrero fc SolanO (|2010f ). to make it compatible
with a possible existence of convective motions inside
penumbra l filamen ts as p ointe d out b y recent obser-
vation s (iBello Go nzalez e t al.l 120051 : llchimoto et al.l
l2007bl : IZakharov et al.. .2008f ). One open question in
the uncombed scenario is how the penumbral contin-
uum intensity can be as large as 75% of the normal
quiet Sun brightness (see, e.g., Solanki fc Riiedi 2003;
Schlichenmaier fc Solankil 120031 ISpruit fc ScharmeJ
2006). Howev er, it is interesting to no t e tha t the
simulations by iRuiz Cobo fc Bellot Rubia (J2008l ) can
explain up to 50 % of the quiet Sun intensity if the
Evershed fiow consists of horizontal fiow channels with
a length of several Mm that bring hot material from
deeper layers. Indeed, the proble m of the penumbral
heating was a strong motivation for IScharmer fc SpruitI

Puschmann et al.

(|2QQ6[ ) and ISpruit fc Scharmerl (|2QQ6[ ) to propose the
gappy penumbral model, in which field-free overturning
convection occurs inside the penumbra and efficiently
heats its atmosphere. However, this model has not yet
been confronted to spectropolarimetric observations. In
particular, it is unclear if this model is able to reproduce
the observed net circular polarization. However, for that
to happen, the magnetic field inside the filaments should
be around 1000 G, which seem s to be incompatible wit h
the concept of a field- free gap (iBorrero fc Sola nkill2010l ).
Recently, Scharmer (2008) proposed the possibility of
strongly reduced (although non-zero) magnetic field
inside the field-free gap. These modifications of both
scenarios, uncombed (including convection inside flux
tubes) and gappy (including significant magnetic field
inside the field- free gaps), could in the end produce
similar observational effects, although they are based on
different physical mechanisms.

Simulations based on radiative magneto-convection
in inclined magnetic fields can help to shed light on
open questions. First attempts to simulate penum-
bral stru cture in small slab- like sections of sunspots
(|Heinema nn et al.l 120071 : iRempel e t al. 2009a) resulted
in rather narrow penumbral regions. Recent numeri-
cal simulations by iRempel et al.l ()2009b.) seem to repro-
duce many observational properties of penumbral fila-
ments: e.g. the filamentary structure of the penum-
bra, the Evershed flow, some supersonic velocities (al-
though the observe d ones are certai n ly more com-
plex, s ee e.g. Bellot Rubio et al. 1 120041: llchimoto et al.l
I2007at iMartmez Pillet et a l. 2009), the correlation be-
tween inclination and strength of the magnetic field,
and the penumbral brightness. Nevertheless, the result-
ing penumbra is still somewhat immature, revealing too
short and too fragmented filaments. D ark lanes along
bright filaments observationally found bv lScharmer et al]
(J2002 ) form only occasionally. At present, these simula-
tions do not allow a direct comparison with spectropo-
larimetric observations, since up to our knowledge no
synthesized spectra are available. Surprisingly, the sim-
ulated penumbrae share some properties with both the
uncombed and the gappy scenarios: i.e. convective mo-
tions inside zones harboring nearly horizontal and rela-
tively strong (around IkG) magnetic fields. The vertical
slices across a penumbral filament taken from a snapshot
of Rempel et al's simulations and presented in Figure 1
of lBorrerol (J2009) point to the co-existence of significant
magnetic fields and overturning velocities.

Observationally, 3D visualizations of penumbral fila-
ments can only be obtained from inversions of spec-
tropolarimetric data (in the abs ence of rob ust re-
sults from local helioseismology, iGizon et al.l [2009).
While Milne-Eddington inversions only yield the
physical parameters at an average optical depth
(jSanchez Almeida et al.lll996r). more sophisticated inver -
sion codes lik e SIR (iRuiz Cobo fc del Toro Iniestalll992f )

or SPINOR ( Frutiger et all[2000f ) provide the stratifica-
tion of physical parameters in an optical depth scale.
Several 3D optical depth models have been published
in the re cent past (see , e.g;., IWestendorp Pl aza et al.
2001allbl: iM athew et aH 120031 : ISanchez Cuberes et al.
2001 iBec k 2006, 200"^

ing SpectroPolarimeter (CRISP, iScharmerl l2006f ) and
by the Hinode/SP spectropolarimeter (Lites et al.
2001, see also Tsuneta et al. 2008). In the
near future th e Gregor Fabry Perot interferometer
(GFPI, see e.g. Puschmann et al. 2006, 2007) and its
full Stokes polarimet er f Bello Gonzalez fc Kneer 20^81:
iBalthasar et al.l l2009f ) may reveal spectropolarimetric
data at even highe r spatial resolution once Gregor
([Volkmer et al.l [20071 ) is in operation. The ground based
2D observations take advantage of image reconst ruc-
tion techniques (see e.g. Lofdahl 2002; van Noort et al.l
20051: iPuschmann fc Sailer I2Q06I: iDenker etld] 120071:
Bello Gonzalez fc Kneer. .20081 : IWoger fc von der Liihd

20081 ) to improve spatial resolution and to reduce spa-

The highest spatial resolution input data for this
endeavor are currently provided by the CRisp Imag-

ti al crosstalk from va riabl e seeing.

iJurcak et all (J2007D and lJurcak fc Bellot R^ubiol (j2008r i
have carried out inversions of Hinode data and found
an atmosphere that is basically consistent with the un-
combed scenario: horizontal channels carrying the Ever-
shed flow are substantially magnetized. These channels
appear at large optical depths, but no information about
geometrical heights is obtained.

Given the highly fluctuating densities that are ex-
pected in the penumbra, the interpretation of an optical
depth scale as a good approach for geometrical heights
can be misleading and a reliable transformation to ge-
ometrical heights is mandatory. Besides, to establish a
geometric height scale is important for the determination
of the electrical current vector J that is crucial for the
determination of ohmic energy dissipation, and for a test
of the reliability of ah initio MHD simulations. While
traditionally in quiet Sun the transformation from an
optical depth scale to a geometric one has been done
by assuming hy drostatic equilibrium (see, for example,
iPuschmann et al...2005.) . this is not justified in the mag-
netized penumbra. The force balance in this case must
include magnetic forces which require the calculation of
horizontal and vertical spatial derivatives of the magnetic
field.

An absolute geomet rical height s cale for penumbrae
has been derived by Sa nchez Almeida (2005) from a spec-
tropolarimetric inversion under the MISMA hypothesis
by imposing equal total pressure between adjacent pixels,
although neglecting the magnetic tension in the Lorentz
force. Carroll fc Kopf (2008) obtained the stratification
of physical quantities in geometrical height applying a
neural network inversion technique based on MHD sim-
ulations of quiet Sun. The application of this technique
on sunspot penumbrae mig ht be str aight forward, using
e.g. the simulations of Rem pel et al . (2009b) as input.

The main objective of the work presented here is to
infer a 3D geometrical model of the penumbra by im-
posing the solenoidality of the magnetic field and the
dynamic equilibrium including the Lorentz force. This
model allows the determination of the Wilson depres-
sion at each pixel, the correlation of physical quantities
at each height, and the evaluation of the energy budget
transferred by mass motions.

2. OBSERVATIONS

The active region AR 10953 near solar disk center was
observed using the spectropolarimeter (SP) of the Solar
Optical Telescope (SOT) on-board the Hinode spacecraft

A geometrical height scale for sunspot penumbrae

10 20 30 40 50

10

20 30

[Mm]

40

50

Fig. 1. — Upper panel: Continuum intensity at 630.2 nm of a
fragment of the active region AR 10953 observed by SOT on-board
of the Hinode spacecraft. The white rectangle marks the area under
study in the present paper. The arrow indicates the direction to
disk center. The slit is vertically oriented. Bottom panel: Stokes
V at +286 mA from the line center.

on the 1"* of May 2007, between 10:46 and 12:25 UT. The
region was scanned with a step size of 0'/148 and a slit
width corresponding to 0'/158, recording the full Stokes
vector of the pair of the neutral iron lines at 630 nm with
a spectral sampling of 21.53 mA. The spatial resolution
was about 0'.'32.

The integration time was 4.8 s, resulting in a noise level
of approximately 1.2 x 10~^. The wavelength calibration
was done assuming that the average umbral profile does
not exhibit velocities.

For reasons of simplicity this article centers on a section
of the inner penumbra of the large sunspot presented in
Figure [TJ The analyzed area is marked by a white rect-
angle centered at an heliocentric angle ^ = 4.63°. The
arrow indicates the direction to disk center. In figure co-
ordinates, the disk center would be at X = -5.5Mm and

Y = 74 Mm. This area has been chosen because it con-
sists of homogeneous and radial aligned filaments. In the
bottom panel of Figure [1] we show the map of Stokes V
at +286 mA from the line center. In this map, the black
points correspond to pixels in which the magnetic field
shows a reverse polari ty and harbors strong downfiows
(|Ichimoto et al.ir2007al ). The area analyzed in this pa-
per does not include such points. Nevertheless, provided
that these strong downfiows are mainly located in the
mid and outer penumbra, the results of this paper can
be considered representative for the inner penumbra. In
a posterior paper we will apply the method described in
this article to the entire sunspot.

3. INVERSION PROCEDURE

To derive the physical parameters of the solar at-
mosphere as a function of continuum optical depth,
i.e. temperature T{r)^ magnetic field strength 5(r),
field inclination 7(r), field azimuth ^(r), and line
of si^ht velocity v\os(t), the S IR inversion code
(|Ruiz Cobo fc del Toro Iniestalll992"l ) was applied on the
spectropolarimetric data set. In the region of the inner
center-side penumbra under study, all Stokes profiles are
slightly asymmetric, are not multi-lobed and do not show
strong displacements. Consequently, there is not suffi-
cient information to obtain a two-component atmosphere
unambiguously. Therefore, we use an one-component in-
version to obtain the stratification of plasma parameters
in each pixel. The same procedure has already been ap-
plied on sunsp ot penumbral data obser ved with the Hin-
ode/SP by e.g. iJurcak fc Bellot Rubiol (^008). The spec-
tropolarimeter on-board Hinode allows us to distinguish
the fine structure of the penumbra with a resolution of
0^3. The d ark cored penumbral filaments discovered by
lScharmei~e t al. (2002), with a typical size less than O'.'l,
could indicate the existence of smaller physical struc-
tures, but could also be generated by temperature differ-
ences prese nt in iso-tau layers above large st ructures like
ffux tubes ("Ruiz Cobo fc Bellot Rubi3|2008D or field-free
gaps (Spruit fc Scharmer 2006). In any case, our results
have to be interpreted as the averaged properties of struc-
tures on a scale of 0^3.

The free physical parameters are calculated by SIR at
different optical depth points (nodes). We have chosen 5
nodes in T{r)^ 3 nodes in B{r) and v\os{t), and 2 nodes in
7(r) and 0(r). The final temperature stratification is ob-
tained by spline interpolation. We have used a parabolic
interpolation for B{r) and vios{t) and linear interpola-
tion in case of 7(r) and 0(r). We have neither considered
microturbulent velocities nor stray light contamination.
The final synthesized profiles have been convolved with
a macroturbulent velocity Vmac as an additional free pa-
rameter of the inversion.

In Figure [21 we present the results of the inversion at
an arbitrarily chosen pixel. Note that the synthesized
spectral lines even reproduce the asymmetries present in
the observational Stokes profiles (left four panels in Fig-
ure [2]). In the right four panels of this figure, we present
the stratification of T, v\os^ B^ 7, and (j) versus the con-
tinuum optical depth. The error bars calculated at the
different nodes are also shown. The sunspot has negative
polarity (7 > 90°). We do not perform any correction
of the azimuth to solve the 180° ambiguity, since all re-
sulting azimuth values are in the range between -90° and

Puschmann et al.

[h

630.07 630.21 630.34

wavelength [nm]

-5 -4 -3

-2 -1
log(T)

Fig. 2. — Left column: Normalized observed Stokes /, Q, U and
V profiles of one pixel (circles); The red solid lines correspond to
the best fits of synthesized profiles. Right column: Corresponding
stratification of temperature T, line-of-sight velocity ^'los5 magnetic
field strength B^ field inclination 7 and field azimuth cf) (dashed
line) vs. logarithm of continuum optical depth at 500 nm log(r);
over-plotted are the error bars calculated at the different nodes.

90° showing a smooth spatial variation across the field
of view (FOV).

4. DETERMINATION OF A GEOMETRICAL
HEIGHT SCALE

The inversion method delivers for each pixel the strat-
ification of an atmospheric model vs. continuum optical
depth, i.e. we obtain B{x^y,r), T{x^y^r), etc. .. (see
Figure [3j). Taking into account that the region un-
der investigation is not exactly placed at disk center
(/i = 0.997), our XF-plane is not parallel to the local
solar surface. A transformation of the reference system
can only be applied a posteriori, once the z-scale has
been obtained. Besides, it is impossible to perform this
transformation without the knowledge of horizontal ve-
locities. Consequently, we consider the z-axis to be par-
allel to the line-of-sight direction throughout this article.
We will term the z-axis as the "vertical" and accordingly
the perpendicular plane as the "horizontal" plane.

A geometrical height scale z{x,y,r) can be obtained
by integrating

T [K] at log(T) = B [G] at log(T) = -1.5

5400 5600 5800 6000 6200 1400 1600 1800 2000 2200

7 n at log(T) = -1.5

100 110 120 130 140

los
-0.2

km/s] log(T) = — 1.5
0.0 0.2 0.4

dz = dr .

(1)

Fig. 3. — Maps of temperature T at log(r) = 0, and maps of mag-
netic field strength B, field inclination 7, and line of sight velocity
-uios at log(r) =-1.5. Negative and positive velocities correspond to
upfiows and downfiows, respectively.

For the integration of equation ([T]), three ingredients
are needed: k (continuum absorption coefficient per
gram), p (mass density) and the boundary condition
Zw = —z{r = 1) (Wilson depression)^ . The first ingre-
dient, nix^y^r), is evaluated by SIR from T{x,y,r), the
gas pressure (Pg) and the abundances. However, as we
will see below, hz is nearly independent of Pg. The strat-
ification of Pg and K, are obtained under the assumption
of hydrostatic equilibrium (HE), i.e., by the integration
of

dPg/dr = g/K , (2)

where g represents gravity. In each step of the inte-
gration (i.e. at each r) hz and Pg are found iteratively
such that both variables are compatible with the ioniza-

^ We introduce a change of sign to be consistent with the usual
definition of the Wilson depression, i.e. positive values for de-
pressed r = 1 layers.

A geometrical height scale for sunspot penumbrae

tion equilibrium in LTE (Saha equation). For the gas
pressure boundary condition at the top layer (log r = -
4.8) we used 655 dyn/cm^ (taken from IVerna zza et al.l
1^81). Solving the HE (equation (|2])) with boundary con-
ditions differing by two orders of magnitude (Pg (log r = -
4.8) = 10 dyn/cm^ and Pg (log r = -4.8) = 1000 dyn/cm^)
the resulting stratification of the absorption coefficients
(^10 (t) and /^iooo(t)) is practically invariant. The ratio
between a>:iooo('?") and A>:io(r) approximately follows the
empirical law:

^1000 (t)
^io(t)

1

5-10-

(3)

i.e. at logr >-2 the difference is smaller than 0.5%.

The pressure stratification is used for the calculation
of the second ingredient, the density, using the equation
of state for an ideal gas considering partial ionization.

By choosing an arbitrary Wilson depression Zw at all
pixels, a geometric height scale z{x^y,r) can be con-
structed after integration of equation ([T]) (which would be

a HE height scale). P(x, y^ z) maps are then obtained by

the interpolation of the P(x, y^ r) maps resulting from

the SIR inversion. Such P(x, y^ z) maps have a diver-
gence different from zero. Furthermore, this model would
not be in mechanical equilibrium. An optimum choice of
Zw(^, y) should minimize at a given height both the di-
vergence of the magnetic field V • B and the error in the
equation of motion.

Neglecting viscosity, the equation of motion can be
written as

dv -^ -i
^ dt

pg-^Pg

(4)

where v stands for velocity and J for the current density.
The acceleration on the left hand side of equation (|4]) can
be decomposed in a temporal derivative and an advective
term

dv dv ,^ ^.^

Due to the extended lifet ime of p enumbral filaments of
^ 1 h our (see e.g. Siitterli n et al.l 2004; Langhans et al.
120071 ) compared to the sound travel time across a fila-
ment, their velocity can be considered stationary and we
thus neglect the partial temporal derivative ^ in equa-
tion ([5j). Since the photospheric magnetic field is nearly
frozen into the plasma, we can assume, as an order of
magnitude estimation, that the material moves along the
field lines and derive v from v^ = —v\o^^ 7, and (\). Note
that positive v\q^ corresponds to a red-shift while positive
'Uz are upfiows.

We also negelect the dynamic term (iT- V)iT, since it can
be shown to be much smaller than the advective term.
In Figure |4] we plot the histogram of the errors intro-
duced by neglecting the dynamic term: ratiOx, ratiOy,
ratioz, which represent the ratio between the advective
term (the second term on the right hand side of equa-
tion dSj) and the dynamic term (second and third terms
on the right hand side of equation (j4|)) without consider-
ing the Lorentz force for each of the three components:

ratiOx

-P-

(_^^VK

dPg/dx

Fig. 4. — Histograms of the ratio between the advective term
and the dynamic term without considering the Lorentz force for
each of the three components evaluated at a height of 200 km in
the penumbral region presented in Figure [T] ratiOx (black), ratiOy
(red), ratiOz (blue). Binsize of 5 x 10~^.

ratiov

ratio^

' dPg/dy

dPg/dz^ pg

(6)

The magnitudes of equation ([6]) have been evaluated at
the reference height of 200 km where the minimization
procedure will be applied. The neglection of the advec-
tive term is justified since most of the values are below
1%.

Let us define then the residual force F to be

J X B

VPn

pg-v±^g, (7)

with Pg inferred from equation (|2]) . In our approximation
F is zero if the equation of motion is fulfilled. To ensure
the physical meaningfulness of the solution also V • B
must be zero.
We define a merit function as

X

E ^l(i^x'+^y+^z')+^2(V-^)^

(8)

where i^i, W2 are coefficients introduced to adequately
weight the contribution of both V • B and F. By intro-
ducing vertical displacements of the atmospheric models
at each pixel, we minimize the merit function of equa-
tion (j8]). Therefore, we used a genetic algorithm that
iteratively changes an array Dz{x^y) containing the dis-
placements of the atmospheric models at each pixel. The
merit function is evaluated at the 200 km height level. At
this layer the response function of Stokes V of the visible
lines to the magnetic field strength reaches its maximum
and consequently the uncertainties of the inverted mag-
netic field are minimal. Besides, at this layer the error
introduced by neglecting the acceleration term in equa-
tion (|5j) is expected to be smaller than at deeper photo-
spheric layers since, in sunspot penumbrae, the velocity
field is confined to lower layers.

Owing the uncertainties of the inversion method, the
obtained stratification of the magnetic field is neither
solenoidal nor does it satisfy the equation of motion.
Thus, for the evaluation of x^ we use slightly modified
values for the magnetic field vector inside its uncertain-
ties, i.e. we define B'{x,y^ z) = B{x^y, z) -\- Nsix^y),
Y{x, y, z) = 7(x, y, z) + N^{x, y), and 0'(x, y, z) =

6

Puschmann et al.

(j){x,y,z)-\- N(f){x^y), with absolute values of noise
functions Nsix^y)^ N^ix^y)^ Nfj^ix^y) smaller than
the error uncertainties of the respective parameters.
The resulting synthesized Stokes profiles, considering
B' ^ 7' and 0', are still compatible with the observed
Stokes profiles. In summary, the solution found by the
genetic algorithm is constituted by the optimum values
of Dz(x,y), Nsix^y), N^{x,y), and Ncf){x,y). We have
performed several realizations of the genetic algorithm
to determine the optimum values of the weights i^i,
W2. If we set wi to zero (i.e. just minimizing V • B)
many different solutions are reached; with W2 equal to
zero (i.e., minimizing the components of F) the solution

reduces in particular the V • B values, however not sat-
isfactorily. The best solution is reached setting wi and
W2 such that both addends in equation (J8j) contribute
in an equal manner {w2 = 3.3-10^ • Wi, using cgs units).
In each realization, the code produces slightly different
results with an approximately Gaussian distribution
around a mean value at each pixel. Therefore we
adopted the average of 20 individual realizations as
the final solution. In the upper left panel of Figure [5]
we present the resulting displacement Dz{x^y). The
displacements span a range between -250 km and 100 km
and show a spatial distribution well correlated with
the x-y maps of several physical quantities, e.g. T,
5, or 7 (see Figure [3j). Negative numbers in this scale
correspond to deeper layers. The zero height of the scale
has been arbitrarily set at one of the pixels analyzed; its
comparison with an absolute height scale in the quiet
Sun photosphere is discussed in Sect. 5. The relative
noise Nsix^y), N^{x^y)^ and N(f){x,y) is presented in
the upper right and lower panels of Figure El Note that
the introduced noise functions account for errors below
3% in B and 7, and below 4° in case of 0.

After introducing Dz{x,y)^ Nsix^y)^ N^{x^y) and
N(f){x,y) and interpolating to a common z- scale we can
assume that the layer at 200 km approximately satis-
fies both, V • B and the equation of motion, although
the pressure stratification for each pixel continues be-
ing the HE one. We can obtain now a more accurate
Pg - stratification by the integration of the z - component
of equation Q. However, when the Pg - stratification
changes, the z - scale is also modified. Provided that T{r)
and B{r) are invariant, it is easier to integrate this equa-
tion in terms of optical depth. We can rewrite the vertical
component of equation Q in terms of r, setting P = 0:

dT

RT

KllPg

(J X B)^ .

(9)

After the integration of this equation, e.g. by Runge-
Kutta, we obtain Pg. From the equation of state we
evaluate p and subsequently the new z - scale from equa-
tion ([1]). The newly obtained Pg values produce slightly
modified stratification and the whole procedure starting
with equation ([7]) is again repeated. The modifications
are minimal and convergence is achieved after two iter-
ations. Figure [6] shows the histograms of the divergence
of the magnetic field and of the three components of the
residual force before and after the minimization. The
process clearly improves V • B (changing the standard
deviation of its histogram, a^ from 0.38 to 0.16 G/km),

Dz [km]
-200 -100

Nb/B [%1
100-3 -2-10123

12 3 4

X [Mm]

Nr/7[%1
-3-2-10 1 2 3-4

1 2 3

X [Mm]

Fig. 5. — Solutions of the genetic algorithm: Displacement of the
2; = 200 km layer Dz{x,y) and noise functions A^^, A^^, and N^f).
See text for details. We do not plot relative A^^ values because cf)
reaches null values at some pixels.

Px (cr from S-IO""^ to 4-10"Myn/cm'^), and Py (a from
8-10~^ to 6-10~^dyn/cm ) and slightly diminishes F^ {a

from G-IO""^ to S-lO^'^dyn/cm ).

The models at all pixels were then interpolated to a
common global z- scale. In Figure [7] we present the re-
sulting histograms of V • P at six different layers. To
check the significance of the results, we evaluated his-
tograms of the simulated magnetic field distributions
which deviate from solenoidal fields only by Gaussian
noise, with a sigma equal to the estimated errors of the
field components at each layer. We plot the mean his-
tograms of 1000 realizations shifted by 0.2 G/km as red
lines. Owing to the resulting coincidence between the
histograms of both V • B and the errors, we conclude
that in the range 50 km to 200 km, the residual value of
V • B can be entirely attributed to noise except for a sys-
tematic offset of 0.2 G/km. At lower and higher layers,

A geometrical height scale for sunspot penumbrae

&^

Iln

^

10

^

:

S 6

A

2

8

:

s^

r \

s^

1 ,

fX 4

-

a.

6

r

1

-

S 2

cr

r^

A,..:

Id

4
2

J

\

'^^^^,^;

-1

.5 -1.0 -0.5 0.0 0.5 1.0 1

5

-3

-2

-1

1

2 3

V-B [G/km]

^x

[10-

dyn^cm"''

s

^ 10

10

:

S 8

n

s

8

P^

b

_

a 6

Dh

6

fl

\

:

1 4

(U
r^

J\ :

(U

4

2

^

/

1

V

:

-

3-2-10 1 2 [

]

-3

-2

-1

1

2 3

F [10"' dyn*cm"']

''z

[10-

dyn^cm"''

Fig. 6. — Histograms of V • S and the three components of the
residual force F at z = 200 km before (red hnes) and after (black
lines) the application of the genetic algorithm. Binsize of 0.025

(V • B) and 10~^ (force components), respectively.

z= km

z= 50 km

7

6

5

-

4

jf\

3

M

2
1

n

/'■\i -

-1.5-1.0-0.5 0.0 0.5 1.0 1.5
V-B [G/km]

z= 150 km

-1.0-0.5 0.0 0.5 1.0 1.5

V-B [G/km]

Fig. 7. — Black: Histograms of the divergence of the magnetic
field strength at six different layers after the integration of the
vertical component of the equation of motion. Red: Ivlean his-
tograms of the divergence obtained from a random Gaussian field
distribution. The error histograms have been shifted by 0.2 G/km.
Blue: Histograms obtained before the application of the genetic
algorithm. Binsize of 0.025

not all V • 5 values can be regarded as noise. The blue
lines represent V • B calculated with a z - scale obtained
in hydrostatic equilibrium and considering as boundary
condition Zw = and P^(log(r) =-4.8) = 655dyncm~^.
Note that at layers below 250 km the divergence of the

magnetic field vector improves considerably after apply-
ing the genetic algorithm.

What remains elusive is the origin of the systematic
0.2 G/km offset. To understand how this value is dis-
tributed among the three components of V • 5, the hor-
izontal derivatives -^, -^ and the vertical derivative

^^ , in Figure [8] we show these components separately
(both before and after the minimization of the merit
function by the genetic algorithm). It turns out to be
somewhat surprising that the individual histograms of
the derivatives in the three Cartesian directions do not
change much with the application of the minimization al-
gorithm. It is only their addition in the form of the diver-
gence that is effectively minimized by the algorithm. But
in this figure, we readily see that the offset of 0.2 G/km
originates entirely from the vertical derivative . This
should be no surprise. Already Westend orp Plaza et al.l
()2001af ). who performed a similar analysis to the one
done in this work but without computing a geometrical
height scale, found an excess in the vertical derivative
after inverting d ata obtained with the Advanced Stokes
Polarimeter (see iLites et al.|[l993f ). These data had a
spatial resolution at least thr ee times wors e than that
used in our work. .Westendorp Plaza et al.l (|2Q01af ) at-
tributed the origin of this excess to the unresolved na-
ture of the structuring of the penumbral magnetic field.
Interestingly, they reported on a mean excess 5-10 times
larger than the 0.2 G/km found here. The question of
how the vertical derivative can be made compatible with
the null divergi:ence condit i on wa s debated by several
authors. iSanchez Almeida! ()1998f ) suggested that a mi-
crostructuring of the penumbral magnetic field in ranges
around 1-10 km is abl e to solve this problem. Following
iMartinez Pilletl ()2000l ). the azimuthal components gen-
erated by the background field wrapping around penum-
bral flux tubes cancel out in the resolution element lead-
ing; to the sniall ho rizontal gradients observed (see also
iBorrero et al.l l2008 f). In Figure 1 of 'Sanchez Almeidal
(0^998) one finds the magnitude of the horizontal deriva-
tives that enter the calculation of the divergence for ob-
servations with resolution worse than 0^2. The values
cited fluctuate with typical magnitudes of 0.15 G/km,
clearly smaller than the vertical gradients estimated with
1 arcsec resolution spectropolarimetry. The situation in
our present analysis has improved considerably as now
the horizontal derivatives fluctuate around values that
are of the same order of magnitude than the vertical one
(see Figure [8j). However, the excess of 0.2 G/km still
shows the presence of unresolved components that could
be present either in the vertical or in the horizontal di-
rections.

In order to estimate the compatibility of the re-
sulting models with the observed Stokes profiles
(r^^Q^^^U^^^V^^'), we have synthesized the Stokes

profiles (if^rial^Qfinal^^final^yfinal^^ f^^^ ^^^ ^^^^

model atmospheres. In the upper left panel of Fig-
ure [H we present the histogram of the rms difference

Alrms = ^^(if'^^^ -r^y between the final and the

observed profiles (red line). The histogram of rms differ-
ences between the best fit found by SIR (labelled in the
figure as SIR) and the observed profiles are plotted in

Puschmann et al.

-1 1

dB,/dx [G/km]

dB/dy [G/km]

S 6

X3

-

a 4

II

o

d „

CD 2

3

cr

nl llllWlWiM* iN^

j4^wrri™ " ifiij

.^

.....„...a^iiF^ ^ , II """H*^"

dB^dz [G/km]

V-B [G/km]

Fig. 8. — Horizontal (top) and vertical (bottom left) derivatives
both before (red) and after (black) the minimization by the genetic
algorithm. Vertical dashed lines correspond to the mean values of
the distributions. The bottom right panel shows the sum of the
three derivatives and is the same histogram as shown in Figure [6]
at other scale.

SIR-Obs.
Final-Obs.
Fmal-SIR '

^Q.^y^n

AU /a

Fig. 9. — Histograms of the rms differences between the inverted
(SIR) and the observed (Obs.) Stokes profiles in units of the stan-
dard deviation of the noise for the respective Stokes parameter
(black). Histograms of the rms differences between the Stokes pro-
files synthesized from the final models (Final) and the observed
profiles (red) and between the final and the SIR profiles (blue).

black. This rms is precisely the quantity minimized by
the inversion method. The histogram of rms differences
between the final and the SIR profiles (i.e. the error in-
troduced by the modification of the model through the
application of the present technique) is presented in blue.
The corresponding histograms for the other Stokes pa-
rameters are shown in the remaining panels of this figure.
Note that the rms differences between the SIR and the
observed profiles peak around 1.5a - 2a (where a is the
standard deviation of the noise in the respective observed
Stokes parameter), while the additional noise introduced
by the present technique has a distribution that peaks
(in case of Q, U, and V) at around la. This is com-

patible with the fact that the noise functions, Nsix^y),
N^{x^y)^ N(i){x^y) introduced, are within the error un-
certainties of the respective parameters. In case of Stokes
I, the introduced error is slightly larger, peaking around
2 a. This enhanced sensitivity of I is due to the change
of continuum intensity introduced by gas pressure mod-
ifications.

5. RESULTS AND DISCUSSION

In Figure [TOl we present horizontal slices of different
physical quantities at three different height layers, i.e.
T, B, 7, Vz, at km, 100 km, and 200 km, respectively
(the most right panels of each row will be described
later). White contour lines correspond to v^ larger than
0.3kms~^ (i.e., relatively strong upfiows) as estimated
at km. In all panels of this figure, the typical fila-
mentary structure of the penumbra is clearly observed,
showing a qualitatively good correlation between the dif-
ferent physical quantities. The temperature contrast
strongly diminishes with height, being nearly isothermal
at z = 200km; a (T) /T changes from 11% at z = 0km to
3.7% at z = 200km. The mean magnetic field strength
slightly diminishes from 1859 G to 1794 G in the same
height range, while the mean field inclination increases by
3°. Uz changes from an upfiow of 0.22 km s~^ to zero. Ar-
eas with ascending velocities larger than 0.3kms~^ cor-
respond to places with higher temperatures and weaker
and more horizontal magnetic fields.

In Table [H we summarize the mean value and standard
deviation of some quantities grouped in terms of vertical
velocity at z = 0. If we interpret the ascending velocities
as the line-of-sight component of the Evershed fiow, the
behavior of T, 14, B and 7 confirms the main proper-
ties of the uncombed scenario: ascending hotter material
shows weaker and more horizontal magnetic fields. In-
terestingly, the upfiowing material shows enhanced gas
pressure and densities by around 86% and 71% respec-
tively at all layers. Our results do not show any field- free
region but rather indicate only a weak reduction of the
magnetic field strength (from the nearly vertical back-
ground field of ~ 1900 G to the more horizontal field of
bright penumbral filaments of ^ 1750 G) in the observed
penumbral region. The gradient of gas pressure, appear-
ing at the borders of ascending areas, can not be com-
pensated by the small variation of the magnetic pressure
generated by the magnetic field weakening: it is balanced
by the magnetic tension, which is produced by the strong
variation of the magnetic field inclination. This is the
reason why the method to establish an absolute geomet-
rical height based on the equal total pressure between
adjacent pixels (Sanchez Almeida 2005) can produce er-
roneous solutions. In Figure [TTI we show a map of the gas
pressure at z = 0; the contour lines, enclosing areas har-
boring strong upfiows, are clearly related to areas with
increased gas pressure. The horizontal component of the
Lorentz force (blue arrows) nearly balances the gas pres-
sure gradient.

The weak reduction of the magnetic field strength is in
contradiction with the gappy model (Scharmer & Spruit
2006 and Spruit & Scharmer 2006) that predicts "nearly"
field-free gaps harboring convective motions and reach-
ing photospheric layers. A possible explanation could
be a lack of spatial resolution: .Scharmer et a l., (|2008f )

A geometrical height scale for sunspot penumbrae

TABLE 1

Mean value and standard deviation of temperature T, magnetic field strength B, field inclination 7, gas pressure Pg,

DENSITY p, AT THREE DIFFERENT HEIGHT LAYERS (0 KM, 100 KM, 200 KM) FOR POINTS HARBORING VERTICAL VELOCITIES AT 2; = KM (vq)

SMALLER OR LARGER THAN 0.3 KM S

RESPECTIVELY.

z [km]

Weak up/downflows
Strong upflows

T[K]

BIG]

7[°]

Pg [10^ dyncm-2]

p [10"'^ gcm~3]

vo < 0.3kms~-'^
Vq > 0.3kms~-'^

5175 ±471
5769 ±614

1895 ±200

1777 ±177

119 ±11
110±7

66 ±49
125 ±47

1.9±1.2
3.3 ±1.0

100

vo <0.3kms~^
vo >0.3kms~^

4802 ± 203
5001 ± 204

1851 ±231
1763 ± 166

120 ±9
112±6

30 ±23

57 ±24

1.0±0.7
1.8±0.7

200

vo <0.3kms~^
Vo > 0.3 km s"-*^

4557 ±172
4695 ±123

1810 ±207
1755 ± 163

121 ±9
115±6

14 ±10
25 ±11

0.5±0.3
0.8±0.3

estimate that the minimum magnetic field strength in
a dark cored penumbral filament is reduced by 25%
when the spatial resolution is increased from 0^/3 to
0'/15. Following this order of magnitude estimation, the
strength of our more horizontal fields would reduce to
^13QQG, i.e. a figure still in th e kG range like also
found by Borrero & Solanki (2010). The simulations of
iRempel et al. (2009a. b) present reduced field strengths
down to 700-1000 G.

Besides, we do not find any signature of overturning
convection in the penumbral area observed (at our reso-
lution) . In Table [21 we present the net mass flow evalu-
ated at three different geometrical heights: km, 100 km,
200 km. The uncertainties have been calculated after the
introduction of an artificial velocity offset of ±100 ms~^
considering possible uncertainties in the wavelength cal-
ibration. The average mass fiow is strongly ascending
and cannot be considered as the residual of the can-
cellation between ascending and descending parcels: At
km and 100 km, the upward directed mass fiow is ~ 5
times larger than the downward one and ~ 2 times larger
at 200 km. Our results appear to be compatible with
mass motions dominated by the Evershed fiow which
are mainly upwards directed in the inner penumbrae
(jRimmele & IMarino 2006). In the panels of the 4*^ col-
umn of Figure [TOl the weak downfiows almost exclusively
appear in areas harboring stronger and more vertical
(background) magnetic field. According to the gappy
scenario, we should find downfiows only at the borders of,
but still inside, features with reduced and more horizon-
tal magnetic fields. Perhaps, the lack of these downfiows
mi^h t be explained a ^ain by insufficient spatial resolu-
tion: IZakharov et al.l (2008), analyzing data obtained at
the 1-m Swedish Solar telescope, found weak downffows
of around 90 m/s (certainly at the limit of the veloc-
ity accuracy) at the side of one filament. Although, it
is questionable wether these weak downfiows can be sig-
nificant enough to counterbalance the factor 5 between
< P'^z > up and < pVz > down found in the present work.

The uncombed penumbral model and the gappy model
differ mainly in the way the energy is transferred. In the
unco mbed model the energy is carried by the Evershed
flow (iRuiz Co bo fc Bellot Rubio"2008\ while the gappy
model r oots on convective energ y transfer in fleld free
regions (jScharmer fc Spruitll2006f ). For the flrst time we
are able to estimate the energy flux stratiflcation directly
from observational data. The stratiflcation of physical
quantities in common geometrical height scale delivers

TABLE 2

Mean value of net mass flow over the full observed area

(< pVz >), MEAN VALUE OF MASS FLOW IN ASCENDING (< pVz > up)
AND DESCENDING {< pVz > down) AREAS. MASS FLOW GIVEN IN

UNITS OF [gs~^ m~2]. Percentage of ascending areas (N[%]).

z[km]

<pvz>

< pVz > up

< P'^z > down

N[%]

72 ±23

96±8

-17±4

78 ±22

100

20 ±12

32 ±2

-6 ±4

70 ±25

200

2±5

9±2

-5±1

52 ±28

all the ingredients needed for this calculation. In the
right panels of Figure [10] we show the energy flux (power
per surface unity) Fe carried by the ascending mate-
rial between two height layers in units of the solar flux
Fq = crT^f f . We have evaluated Fe by the integration of

dFE
dz

-pCpVz

dT

dz

dT

dz

where Cp is the speciflc heat at constant pressure
adiabatic temperature gradient is evaluated as

dz J ^ dz

(10)
and the

(11)

with Va is the double logarithmic isentropic temperature
gradient. Both Cp and Va have been evaluated taking
into account partial ionization of hydrogen.

In the upper right panel of Figure [TOl Fe integrated
from z = -75 km to z = km is depicted in units of solar
flux. Only in areas with strong upflows (white contours),
does the energy flux reach signiflcant values and even ex-
ceed (in some points) the solar flux up to 5 times (the
right hand panels of Figure 10 are not scaled to the re-
spective min/max value for contrast enhancement). A
strong decrease of the energy flux with height is appar-
ent. The contribution of the km to 100 km layer to
the energy flux reaches maximum values equal to the so-
lar flux (medium right panel), while Fe integrated from
z = 100 km and z = 200 km, (bottom right panel) changes
its sign, indicating that the ascending plasma is increas-
ing its energy from the environment.

Between z = -75 km and z = km the flux averaged over
the area under study amounts to 23 % of the solar flux,
while the medium and upper layers contribute with 0.9 %
and -0.2 %, respectively. We determine the total amount
of energy flux carried by the ascending material in our

10

Puschmann et al.

T [K] at z = km B [G] at z = km

4500 5000 5500 6000 6500 7000 7500 1600 1800 2000 2200 2400

7 [°] at z = km

, [km/s] at z = km

Fp/F^ from z = -75 to km

100 110 120 130 1.2 1.0 0.8 0.6 0.4 0.2 0.0-0.2 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

B 3

^^^ ^^

' ^ -. —

T [K] at z=100 km

I [G] at z=100 kn

7 ["] at z=100 km

V [km/s] at z=100 km

Fp/F^ from z = to 100 km

4500 5000 5500 6000 6500 7000 7500 1600 1800 2000 2200 2400 90 100 110 120 130 1.2 1.0 0.8 0.6 0.4 0.2 0.0-0.2 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4

T [K] at z = 200 km B [G] at z = 200 km 7 ["] at z = 200 km v^ [km/s] at z = 200 km

4500 5000 5500 6000 6500 7000 7500 1600 1800 2000 2200 2400 90 100 110 120 130 1.2 1.0 0.8 0.6 0.4 0.2 0.0-0.2

F^/Fq from z=100 to 200 km
-0.02 -0.01 0.00 0.01

S 3

X [Mm]

1 2

X [Mm]

X [Mm]

2
X [Mm]

Fig. 10. — Maps of temperature T, magnetic field strength B, field inclination 7, and vertical component of velocity Vz at three different
geometrical height s : z = 0km (upper panels), 2; = 100 km (middle panels), 2; = 200 km (lower panels). The power per surface unity, Fe,
transferred by plasma motions in vertical direction between 2: = -75 km and 2; = 0km (upper right panel), between 2; = 0km and 2; = 100 km
(middle right panel), and between 2; = 100 km and 2; = 200 km (bottom right panel) is presented in units of the solar fiux, respectively.
White iso-contours enclose areas with velocities at the 2; = 0km level larger than 0.3 kms"-*^ (strong upfiows).

full height range to be 24 % which is well below the ex-
pected 75 % in sunspot penumbrae. Owing to the strong
increase of the energy flux with depth one can expect that
the contribution of layers below z = -75 km might be suf-
ficient to explain the observed brightness of penumbra as
channeled by the upfiows.

As an order of magnitude estimate we have evaluated
the contribution to the energy fiux Fe of the deep layers
located between z = -225km and z = -75km (for reasons
that will become evident later). Clearly, the visible lines

used in this work are nearly insensible to physical param-
eters at layers deeper than -75km, i.e., the stratification
of temperature, density, pressure and velocity at such
layers result from linear extrapolations (in the original
log(r) scale); and so the results at these layers can be
considered only qualitatively. In panel (a) of Figure [12]
we plot the energy fiux carried by the vertical motion
of the solar plasma between -225 km and 200 km, aver-
aged over the penumbral region studied here. The total
contribution of the energy fiux in the full height range

A geometrical height scale for sunspot penumbrae

11

log(^P„

^0.6

0.2
0.0

-200 -100 100 200

z [kml

-200 -100 100 200

z [km]

0.002

q

0.000

/ "
/

/

\ •

^*_j- 1 1 1 1 1 1 1

■ ,,/ ,.

-200 -100 100 200

z [km]

-200 -100 100 200

z [km]

Fig. 12. — Panel (a): Energy flux < Fg > in solar flux units
evaluated from the integration of the function plotted in panel
(d) from -225 km until z. Contribution to the energy flux of lay-
ers higher than -75km together with their uncertainties (blue).
Panel (b): mean temperature gradient (black) and mean adiabatic
temperature gradient (red) averaged over the region under study.
The error bars represent the errors of the mean temperature gra-
dients above -75 km. Panel (c): Height derivative of the energy
flux carried by the vertical component of the plasma motion under
constant pressure conditions < —pcpVz -^ > in units of the solar
flux (black), and the correction (red) in the height gradient of the
energy flux which must be subtracted to account for pressure vari-
ations <—pCpVz (^) >• Panel (d): height derivative of the en-
ergy flux carried by the vertical component of the plasma motion
in units of the solar flux.

Fig. 11. — Map of gas pressure [dyn cm~^] and horizontal com-
ponent of the Lorentz force (blue arro ws) at z = km. White iso-
contours are the same as in Figure [TO]

amounts to 78.7% of the solar flux, a value near the aver-
age brightness of a sunspot penumbra. The contribution
to the energy flux of layers probed by the visible lines
is also plotted (in blue). At these layers, the error bars
have been calculated by a Monte Carlo simulation from
the uncertainties of the physical quantities produced by
the inversion method and the genetic algorithm. In or-
der to understand why only layers between -180 km and
20 km appreciatively contribute to the energy flux, we
plot in panel (b) of Figure [12] the mean temperature gra-
dient (black) and the mean adiabatic temperature gra-
dient (red) averaged over the region under study. The
strong increase of the uncertainties of these gradients
with depth reflects the above mentioned insensitivity of
the lines under study at deeper layers. Nevertheless, be-
low km the temperature gradient is seen to be steeper
than the adiabatic one, thus at these layers the ascend-
ing matter transfers energy to the surroundings. Apart
from temperature gradients, it is necessary to take into
account density, vertical velocity and specific heat strati-

dT

>

fication: in panel (c) of Figure [T2| we plot < —pCpVr
(black), i.e. the height derivative of the energy flux car-
ried by the vertical component of the plasma motion un-
der constant pressure conditions. The red line repre-
sents the correction in the gradient of energy flux which
must be subtracted to account for pressure variations
< —pCpVz (^) >• Under the caveat mentioned above

Fig. 13. — Map of Wilson depression Zw- White iso-contours
enclose areas with velocities at z = Okm larger than 0.3 km s"-*^
(strong upflows). The origin of the Zw-scale is arbitrary and
~ 100 km should be added.

we can understand in the light of this figure why layers
above -200 km are heated by the ascending material.

In panel (d) of Figure [121 we present the height deriva-
tive of the energy flux carried by the vertical component
of the plasma motion in units of the solar flux evaluated
as the difference between the black and red line of panel

12

Puschmann et al.

7500 r"
7000 -^
6500 :
' 6000 :
5500 -
5000 :
4500 :

300 200 100
Z [km]

7500 p
7000 :
6500 -
\ 6000 :
5500 :
5000 :
4500 - >•

300 200 100 -100
\ [km]

7500
7000

Z =

-100km

6500

^ 6000

^ 5500

5000

^•4^^^

4500

_,,^4^m^

300 200 100
Z, [km]

^800

2600

2400

2200

■'■r--vl^&:^-^ ':.■•■■:■ ■

2000

^ ■;:Vf *'*fe*jSyiJiv.>r';}' -

1800
1600

■■:■■■ -rsm^m^-'-^

1400

300 200 100
\ [km]

2800
2600
2400
, 2200
' 2000
1800

iftp'

300 200 100 -100
\ [km]

2800

2600

2400

^ 2200

o

^ 2000
1800
1600

1400

300 200 100
Zw [km]

150

140

K

130

X '

iL^

120

r N. '

f^

110
100

qo

' v\

300 200 100
\ [km]

300 200 100
Z„ [km]

150

140

130

^ 120

110

100

■■■ >f ■■■•■■■ ■

90

300 200 100
Z [km]

300 200 100
Z, [km]

300 200 100

\ [km]

Fig. 14. — Scatter plots of temperature T, magnetic field strength B, field inclination 7, and vertical component of velocity Vz vs. Wilson
depression; at 2; = 0km (upper panels), 2; = 100 km (middle panels), and 2: = 200 km (bottom panels). Points with vertical velocities at
2; = 0km larger than 0.3 km s"-*^ (strong upflows) are presented in all panels by red colour. The origin of the Zw-scale is arbitrary and
~ 100 km should be added.

(c), which is essentiahy the average of equation (p!Q|) .

Another important parameter, which can be evaluated
once the absolute geometrical height scale is known, is
the Wilson depression, i.e. the depth where the con-
tinuum optical depth equals 1. The knowledge of Zw
allows us to study which penumbral features are ele-
vated or depressed. In Figure [13] we present the map
of Zw- Areas harboring significant up-flows (enclosed
by the iso-velocity lines) correspond to elevated struc-
tures (Zw < 100 km). When comparing with Figure [TOl
a good correlation with the different physical quantities is
apparent. For a quantitative analysis of this correlation,
scatter plots of T, 5, 7, 'Uios vs. the Wilson depression
are depicted in Figure [14] for several geometrical heights.
We observe a strong correlation between temperature T
and Zw- in places with higher T, log(r) =0 is displaced
to upper layers due to the strong dependence of opacity
wit h temperature (i n agreement with the results found
bv iSchmidt fc Frltzl [20 04) . Note that, as stated above,
the z = 200km layer is nearly isotherm. Furthermore,
for depressed areas (e.g. Zw > 150 km) the difference in
temperature between the heights of km and 200 km is
only ~500K, while for elevated areas (e.g. Zw <Okm),
temperature differences of the order of 2000 K between
both layers are reached. The magnetic field strength B
shows a very weak correlation with Zy/: lower Zw tends
to be related to weaker magnetic fields. Nevertheless,
we find clear trends between field inclination 7 and the
Wilson depression, at all layers. The same behavior is
observed in case of the vertical component of velocity Vz.
In Figure [10] we have plotted the points corresponding
to pixels harboring significant upflows (in red). Focusing

on the distribution of red points throughout the panels,
we can conclude that the Evershed flow corresponds to
areas with decreased Wilson depression (elevated struc-
tures), hotter temperatures, and weaker and more hor-
izontal magnetic fields. The mean value and the stan-
dard deviation of the Wilson depression in areas with
stronger ascending velocities ('Uz > 0.3kms~^), amount
to 34 km ±57 km, while in the remaining areas values of
122 km ±82 km are retrieved: i.e. features harboring the
Evershed flow are in average clearly elevated.

The zero in the z-scale (and consequently the zero of
the Wilson depression) is arbitrary. Since the area where
the genetic algorithm has been applied did not include
the surroundings of the spot, an absolute scale with re-
spect to the quiet photosphere was not directly inferred
from our analysis. However, such an absolute scale can
be estabhshed in the following way. If we assume that the
central part of the umbra has a Wilson depression of 600
km (Solan3 l2003[ ) with an inclination of 180 degrees, a
linear continuation of the data points in the third panel of
the upper row in Figure [H] (see linear fit) establishes our
mean umbral r = 1 level at 500 km, suggesting that our
km reference level is ~ 100 km depressed with respect
to the normal quiet Sun. This indicates that our mean
Wilson depression over the FOV (94 ± 86 km in our scale)
corresponds to a depression of ~ 194 km with respect to
the surrounding quiet Sun atmosphere. This value is in
agreement with previous esti mates of the Wilso n depres-
sion at the inner penumbra (|Ruedi et al.lll995l ).

6. CONCLUSIONS

A geometrical height scale for sunspot penumbrae

13

Using a genetic algorithm we have established a com-
mon geometrical height scale for photospheric penumbral
models resulting from SIR inversions of Hinode data. Al-
though the genetic algorithm only minimizes V • B and
the deviations from static equilibrium in the equation of
motion at a height of 200 km, the resulting atmospheric
model, interpolated to a common z- scale, shows very
small V • B values (and in any case, always inside the
calculated error bars) in a height range from 50 km to
200 km. The horizontal derivatives of the divergence are
of similar magnitude to that of the vertical one. However,
the latter has a mean value (0.2 G/km) that cannot be
compensated by the mean values of the horizontal deriva-
tives. This shows that, even at Hinode resolutions of 0.3''
(200 km), the penumbra displays unresolved structure.
This unresolved structure can be present in both hori-
zontal and vertical directions. In the first case, improved
spatial resolution should reduce the mean value of the
vertical derivative. In the second case, the only way to
achieve better resolution in the z-direction, would be by
including more spectral lines in the inversion.

Having at our disposal a common geometrical height
scale, we can evaluate e.g. electric currents (which will
be analyzed in a forthcoming paper), the Wilson depres-
sion, and in general the three dimensional structure of
penumbral features. The results show a significant cor-
relation between the different physical quantities. Owing
to the visible photospheric lines used, we are only able to
reliably probe the penumbra in the range from ~ km to
~ 200 km. At deep layers, strong spatial contrast in all
parameters is found, whereas at the highest layers both
velocities and temperatures are nearly spatially homo-
geneous. Penumbral filaments with strong upfiows have
a weaker and more horizontal magnetic field, are hotter
and more dense and exhibit a r = 1 surface displaced to

higher layers. All these properties fit into the picture
proposed by the uncombed scenario.

We point out that at the resolution of the Hinode/SP
there is no trace of field-free regions in the penum-
bral volume retrieved by observations (pixel volume of
~200^km^). Our inversions show kG field strength at
all points inside this penumbral volume. We do not find
traces of overturning convection in the inner penumbral
area analyzed: the mass fiow evaluated at three differ-
ent height layers is clearly dominated by upfiows, the
amount of ascending material being five times larger
than the descending one. This results might certainly
change when analyzing the mid and outer penumbra due
to the presence of stro ng downflow patches as found by
llchimoto et al.| (|2007aD.

Besides, the observed upward motion, here identified
as the vertical component of the Evershed fiow, seems
to carry enough energy to explain the brightness of the
penumbra if the physical quantities below z = -75 km are
extrapolated from the results of the inversion.

As the method presented here has been applied so
far only on a small penumbral region its extension on
the entire sunspot penumbra, including the umbra and
moat regions, will be addressed in a future work.

This work has been supported by the Spanish Ministe-
rio de Ciencia e Innovacion through projects ESP 2006-
13030-C06-01, AYA2007-65602, AYA2007-63881, and the
European Commission through the SOLAIRE Network
(MTRN-CT-2006-035484). The genetic algorithm has
been kindly provided by E. Paez Maha. We thank C.
Beck, R. Schlichenmaier, and S. K. Solanki for fruitfull
discussion.

REFERENCES

Balthasar, H., Bello Gonzalez, N., Collados, M., et al. 2009, in:

Strassmeier, K.G., Kosovichev, A.G., Beckmann, J.E. (eds.).

Cosmic Magnetic Fields: From Planets, to Stars and Galaxies,

lAU Symp. 259, p. 665
Beck, C. 2006, Ph.D. Thesis, Albert-Ludwigs-University, Freiburg
Beck, C. 2008, A&A, 480, 825

Bello Gonzalez, N. &; Kneer, F. 2008, A&A, 480, 265
Bello Gonzalez, N., Okunev, O. V., Dommguez Cerdeha, I.,

Kneer, F., & Puschmann, K. G. 2005, A&A, 434, 317
Bellot Rubio, L., in Magnetic coupling between the Interior and

the Atmosphere of the Sun, S.S. Hassan and Rutten (eds.),

ASP Ser., 2009, in press, arXiv:0903.3619
Bellot Rubio, L. R., Balthasar, H., &; Collados, M. 2004, A&A,

427, 319
Borrero, J. M. 2007, A&A, 471, 967
Borrero, J. M. 2009, Sci. China Ser. G, 52, 1670
Borrero, J. M., Lagg, A., Solanki, S. K., & Collados, M. 2005,

A&A, 436, 333
Borrero, J. M., Lites, B. W., &; Solanki, S. 2008, A&A, 481, L13
Borrero, J. M. & Solanki, S. K. 2010, ApJ, 709, 349
Borrero, J. M., Solanki, S. K., Lagg, A., Socas-Navarro, H., &;

Lites, B. 2006, A&A, 436, 333
Brummell, N. H., Tobias, S. M., Thomas, J. H., & Weiss, N. O.

2008, ApJ, 686, 1454
Carroll, T. A. & Kopf, M. 2008, A&A, 481, L37
Denker, C. Deng, N. Rimmele, T. R. Tritschler, A. Verdoni, A.

2007, Sol. Phys., 241, 411D
Frutiger, C, Solanki, S. K., Fligge, M., &; Bruls, J. H. M. J. 2000,

A&A, 358, 1109
Gizon, L., Schunker, H., Baldner, C. S., et al. 2009, Space Sci.

Reviews, 144, 249
Heinemann, T., Nordlund, A., Scharmer, G. B., &; Spruit, H. C.

2007, ApJ, 669, 1390
Ichimoto, K., Shine, R. A., Lites, B., et al. 2007a, PAS J, 59, 593

Ichimoto, K., Suematsu, Y., Tsuneta, S., et al. 2007b, Science,

318,1597
Jurcak, J. Sz Bellot Rubio, L. 2008, A&A, 481, L17
Jurcak, J., Bellot Rubio, L., Ichimoto, K., et al. 2007, PAS J, 59,

601
Langhans, S., Scharmer, G. B., Kiselman, D., & Lofdahl, M. G.

2007, A&A, 464, 763
Lites, B. W., Elmore, D. F., Seagraves, P., & Skumanich, A. 1993,

ApJ, 418, 928
Lites, B. W., Elmore, D. F., &; Streander, K. V. 2001, ASPC, 236,

33L
Lofdahl, M. G. 2002, Proc. SPIE, 4792, 146
Mathew, S. K., Lagg, A., Solanki, S. K., et al. 2003, A&A, 410,

695
Martinez Fillet, V. 2000, A&A, 361, 734
Martinez Fillet, V., Katsukawa, Y., Puschmann, K. G., Sz Ruiz

Cobo, B. 2009, ApJ, 701, L79
Miiller, D. A. N., Schlichenmaier, R., Fritz, G., & Beck, C. 2006,

A&A, 460, 925
Miiller, D. A. N., Schlichenmaier, R., Steiner, O., & Stix, M.

2002 ASzA 393 305
Puschmann, K. G.', Kneer, F., Nicklas, H., Wittmann, A. D. 2007,

in: Kneer, F., Puschmann, K. G., Wittmann, A.D. (eds.)

Modern Solar Facilities - Advanced Solar Science, p. 45
Puschmann, K. G., Kneer, F., Seelemann, T., Wittmann, A. D.

2006, A&A, 451, 1151
Puschmann, K. G., Ruiz Cobo, B., Vazquez, M., Bonet, J. A., Sz

Hanslmaier, A. 2005, A&A, 441, 1157
Puschmann, K. G. & Sailer, M. 2006, A&A, 454, 1011
Rempel, M., Schiissler, M., Sz Knolker, M. 2009a, ApJ, 691, 640
Rempel, M., Schiissler, M., Cameron, R.H., &; Knolker, M. 2009b,

Science 325 171
Rimmele,'T. Sz Marino, J. 2006, ApJ, 646, 593
Ruedi, I., Solanki, S. K., & Livingston, W. 1995, A&A, 302, 543
Ruiz Cobo, B. Sz Bellot Rubio, L. R. 2008, ASzA, 488, 749

14

Puschmann et al.

Ruiz Cobo, B. &; del Toro Iniesta, J. C. 1992, ApJ, 398, 375

Sanchez Almeida, J. 1998, ApJ, 497, 967

Sanchez Almeida, J. 2005, ApJ, 622, 1292

Sanchez Almeida, J., Ruiz Cobo, B., &; del Toro Iniesta, J. C.

1996, A&A, 314, 295
Sanchez Cuberes, M., Puschmann, K. G., &; Wiehr, E. 2005,

A&A, 440, 345
Scharmer, G. B. 2006, A&A, 447, 1111
Scharmer, G. B. 2008, Phys. Scr., 133, 014015
Scharmer, G. B. 2009, Space Science Reviews, 144, 229
Scharmer, G. B., Gudiksen, B. V., Kiselman, D., Lofdahl, M. G.,

& Rouppe van der Voort, L. H. M. 2002, Nature, Volume 420,

Issue 6912, pp. 151
Scharmer, G. B., Narayan, G., Hillberg, T., et al. 2008, ApJ, 689,

L69
Scharmer, G. B. &; Spruit, H. C. 2006, A&A, 460, 605
Schlichenmaier, R. 2009, Space Science Reviews, 144, 213
Schlichenmaier, R. & Collados, M. 2002, A&A, 381, 668
Schlichenmaier, R., Jahn, K., &; Schmidt, H. U. 1998a, A&A, 337,

897
Schlichenmaier, R., Jahn, K., & Schmidt, H. U. 1998b, ApJ, 493,

121
Schlichenmaier, R., Miiller, D. A. N., Steiner, O., &; Stix, M. 2002

A&A, 381, L77
Schlichenmaier, R. &; Solanki, S. K. 2003, A&A, 411, 257
Schmidt, W. &; Fritz, G. 2004, A&A, 421, 735

Solanki, S. K. 2003, Astronomy and Astrophysics Review, 11, 153
Solanki, S. K. &; Montavon, C. A. P. 1993, A&A, 275, 283
Solanki, S. K. & Riiedi, I. 2003, A&A, 411, 249

Spruit, H. C. & Scharmer, G. B. 2006, A&A, 447, 343
Siitterhn, P., Bellot Rubio, L. R., &; Schlichenmaier, R. 2004,

A&A, 424, 1049
Title, A. M., Frank, Z., Shine, R. A., Tarbell, T. D., Topka, K. P.,

Scharmer, G. & Schmidt, W. 1993, ApJ, 403, 780
Tritschler, A. 2009, In: Proceedings of the Second Hinode Science

Meeting, M. Cheung et al. (eds.), ASP Conf Series, in press,

arXiv:0903.1300
Tsuneta, S., Ichimoto, K., Katsukawa, Y., et al. 2008, Sol. Phys.,

249, 167
Vernazza, J. E., Avrett E. H., &; Loeser, R. 1981, ApJS, 45, 635

(VAL-C)
van Noort, M., Rouppe van der Voort, L., &; Lofdahl, M. G. 2005,

Sol. Phys., 228, 191
Volkmer, R., von der Liihe, O., Kneer, F., et al. 2007, in: Kneer,

F., Puschmann, K.G., Wittmann, A.D. (eds.) Modern Solar

Facilities - Advanced Solar Science, p. 39
Weiss, N. O., Thomas, J. H., Brummell, N. H., & Tobias, S. M.

2004, ApJ, 600, 1073
Westendorp Plaza, C, del Toro Iniesta, J. C, Ruiz Cobo, B.,

Martinez Fillet, V., Lites, B. W. &; Skumanich, A. 2001, ApJ,

547, 1130
Westendorp Plaza, C, del Toro Iniesta, J. C, Ruiz Cobo, B., &;

Martinez Fillet, V. 2001, ApJ, 547, 1148
Woger, F. & von der Liihe, O. 2008, Proc. SPIE, 7019, 46
Zakharov, V., Hirzberger, J., Riethmiiller, T. L., Solanki, S. K., &;

Kobel, P. 2008, A&A, 488, L17