EDP Sciences
Free Access
Issue
A&A
Volume 601, May 2017
Article Number L8
Number of page(s) 4
Section Letters
DOI https://doi.org/10.1051/0004-6361/201730753
Published online 23 May 2017

© ESO, 2017

1. Introduction

The thermal structure of sunspots, in particular below the visible surface, is of great importance for a number of scientific studies. It determines the sound speed and therefore the inference of the subsurface structure of sunspots through the calculation of travel times through local helioseismology (see Couvidat et al. 2006; Moradi et al. 2010; Gizon et al. 2009; Khomenko & Collados 2015,and references therein). It also establishes the location of the β = 1-surface (i.e., the region where magnetic and gas pressure are equal) where most of the mode conversion of magnetoacoustic waves takes place (Cally 2005; Schunker & Cally 2006; Khomenko & Cally 2012). Furthermore, a proper knowledge of the thermal stratification can also be employed to assess the realism of three-dimensional magnetohydrodynamic simulations of sunspots (Heinemann et al. 2007; Rempel et al. 2009; Rempel 2012). It also imposes strong observational constraints on the different models that aim at explaining the onset of convection and heat transport in sunspot penumbrae: convective rolls (Danielson 1961), flows along magnetic flux tubes (Schlichenmaier et al. 1999), or overturning convection in field-free gaps (Spruit & Scharmer 2006; Scharmer & Spruit 2006). Finally, it can help elucidate whether the subsurface structure of the penumbra is better explained in terms of the monolithic (Deinzer 1965; Meyer et al. 1974) or cluster/spaghetti (Parker 1979) models.

thumbnail Fig. 1

Temperatures in two different regions: quiet Sun (top) and penumbra (bottom). Results at optical depths τ5 = 1 and τ5 = 3 are displayed in the left and middle panels, respectively. Right panels show the normalized continuum intensity images. White contours in the bottom right panel (continuum intensity in the penumbra) enclose bright penumbral filaments, whereas white regions in the penumbral temperature maps (left and middle bottom panels) indicate the umbra. All values of the temperature are given in thousands of Kelvin.

Open with DEXTER

Unfortunately, the subsurface layers are not easily accessible. For instance, the analysis of spectral lines and their polarization signals (i.e., spectropolarimetry) allows inferring the physical properties of solar and stellar atmospheres (Chandrasekhar 1960; del Toro Iniesta 2003) as a function of the continuum optical depth τc, where τc = 1 is considered the deepest observable layer and is oftentimes referred to as the continuum-forming layer. However, the wavelength dependence of the continuum optical depth offers the possibility of observing slightly deeper or higher layers of the solar atmosphere by observing different wavelengths (Chandrasekhar & Breen 1946). The observed surface of the Sun (see Fig. 9 in Borrero et al. 2016) is located about 70−80 km deeper at 1565 nm (near-infrared) than at 630 nm (visible): z(τ15 = 1)−z(τ6 = 1) ≈ 70−80 km. Here, the subindexes 15 and 6 indicate wavelengths of 1565 nm and 630 nm, respectively. If instead of using two different reference wavelengths we employ only the commonly used reference wavelength of 500 nm, it can be shown (see Fig. 10 in Borrero et al. 2016) that the response function to the temperature for spectral lines at 630 nm peaks at τ5 ≈ 1, while for spectral lines at 1565 nm, it peaks at around τ5 ≈ 3. In this work we exploit this property to investigate the penumbral thermal stratification T(τ5) in the optical depth range τ5 ∈ [1,3], that is, in the first 70−80 km below its visible surface.

2. Observations and inversion results

The observations employed in this work correspond to spectropolarimetric data (i.e., the Stokes vector as a function of wavelength I(λ) = (I,Q,U,V)) in three Fe i lines at 1565 nm. Although the spectral region contains six neutral iron spectral lines with a spectral resolution of 40 mÅ pix-1, here we analyze only the three with the largest Landé factor and cleanest continuum: Fe i 1564.8 nm, Fe i 1565.2 nm, and Fe i 1566.2 nm. The atomic parameters of these spectral lines can be found in Table 1 in Borrero et al. (2016). The data were acquired with the GREGOR Infrared Spectrograph (GRIS; Collados et al. 2012) coupled with the Tenerife Infrared Polarimeter (TIP2; Collados 2007) that is attached to the 1.5-m GREGOR telescope (Schmidt et al. 2012). The target, NOAA 12049, was observed on May 3, 2014, very close to disk center (heliocentric angle Θ = 6.5°). All details of the data reduction and post-processing can be found in Borrero et al. (2016), Franz et al. (2016), Lagg et al. (2016), Martínez González et al. (2016). In addition, a PCA-based deconvolution (principal component analysis; see Ruiz Cobo & Asensio Ramos 2013) was performed in order to account for an estimated 20% of wide-angle scattered light in the observations. The noise in the final data is about 10-3 (in units of the quiet-Sun continuum intensity), and the spatial resolution is estimated to be about 0.40.45′′.

While the size of the scanned region is 53 × 45 arcsec2 (see bottom panel in Fig. 1 in Borrero et al. 2016), in this work we focus on two distinct smaller regions. The first region is a quiet-Sun patch with an area of 4.9 × 8.3 arcsec2 that contains about 2200 pixels. The second region is a 9.6 × 16.5 arcsec2 penumbral patch located on the limb side that contains about 8600 pixels. The right panels in Fig. 1 display the normalized continuum intensity in these regions. The normalization is done such that the average continuum intensity in the quiet-Sun region is unity.

The PCA-deconvolved spectropolarimetric data have been analyzed with the SIR (Ruiz Cobo & del Toro Iniesta 1992) inversion code in order to retrieve the physical parameters (temperature, magnetic field, line-of-sight velocity, etc.) as a function of optical depth τ5. Owing to the removal of scattered light, we consider only one atmospheric component within each pixel (Borrero et al. 2016). In order to minimize the probability of arriving at a local minimum, the inversion at each pixel has been repeated ten times employing randomly generated initial guess models. These initial models feature τ-independent values (randomly generated using a uniform distribution of values) in the kinematic and magnetic parameters. The thermal stratification is generated by performing random perturbations (also using a uniform distribution) at different optical depth locations on the HSRA model (Gingerich et al. 1971) and reinterpolating T(τ5) in between these locations. Out of the ten inversion results we kept only the solution with the smaller χ2. Because in this work we are mostly interested in the temperature stratification T(τ5), which is mostly encoded in Stokes I, we gave all four Stokes parameters equal weight in the inversion wi,q,u,v = 1 (cf. Borrero et al. 2016). During the inversion, the temperature was allowed to change in three nodes, while the components of the magnetic field (B,γ,φ) and the line-of-sight velocity vlos were allowed to change in two nodes when inverting the penumbral area, but only one node in the quiet-Sun area.

Figure 1 shows the resulting temperatures at two optical depths τ5 = 1 (left) and τ5 = 3 (middle) in the quiet-Sun (upper panels) and penumbral regions (bottom panels). The temperatures clearly closely follow the continuum intensity maps in the right panels of the same figure. For visualization purposes, the umbral region, defined as those regions where the continuum intensity is lower than 0.85, is drawn in white on the temperature maps. Regions where the temperature is below the lower limit of the color bars are also shown in white. On the other hand, white contours in the continuum intensity map of the penumbra (bottom right panel) indicate the location of bright penumbral filaments. Pixels belonging to bright filaments are selected as those penumbral pixels where the continuum intensity is above the mean continuum intensity at the same radial distance from the sunspot center as the pixel subjected to scrutiny. As expected, the temperature in these bright filaments is generally higher than in the dark penumbral regions. Table 1 summarizes the mean temperatures at optical depths τ5 = 1 and 3, as well as their difference, in different regions: average quiet Sun, granules only, intergranules only, average penumbra, and bright penumbral filaments.

Table 1

Mean temperatures (in Kelvin) at two different optical depths and their difference.

3. Discussion and interpretation

3.1. Penumbra and bright filaments vs. average quiet Sun

A particularly interesting feature is the finding that the difference of the mean temperatures between the quiet Sun and the penumbra is larger at τ5 = 3 than at τ5 = 1 (see Table 1). At τ5 = 1, the average quiet Sun is about (374 ± 95) K hotter than the average penumbra, while this difference increases to (701 ± 113 K at τ5 = 3. This also occurs when comparing only the bright penumbral filaments to the quiet Sun: at τ5 = 1, the average quiet Sun is about (257 ± 95) K hotter than the average over the bright filaments, while this difference increases to (490 ± 113) K at τ5 = 3. This feature is not only seen in the average temperatures, but also in the individual temperatures inferred at each pixel. Figure 2 shows the histograms of the difference between the mean quiet-Sun temperature at τ5 = 3 and τ5 = 1, minus the same difference at each pixel of the penumbra (red) or at each pixel of the bright filaments (blue). Hereafter, this quantity is referred to as ξ(T), (1)Locations where ξ(T) < 0 are regions where the temperature stratification T(τ5) converges to the mean quiet-Sun stratification as we probe below the visible surface (τ5 ∈ [1,3] or about 70−80 km below the surface), while pixels where ξ(T) > 0 correspond to regions where T(τ5) diverges from the mean quiet-Sun stratification. The determination of the temperature in this region has been possible because the analyzed spectral lines (i.e., three Fe i lines located at 1565 nm) convey reliable information about the temperature at this depth. A similar study employing data from Hinode/SP is not possible because of the large uncertainties in the retrieval of the temperature at τ5 = 3 using the Fe i spectral lines at 630 nm.

Most of the penumbra and bright filaments are formed by pixels where ξ(T) > 0, with pixels where ξ(T) < 0 representing only 1.3% and 2.8% of the total, respectively. As in Borrero et al. (2016), we have checked that the results we present here are independent of the amount of wide-angle scattered light considered during the deconvolution process: the very same results are obtained when considering 40% of the wide-angle scattered light as well as when analyzing the original (i.e., not deconvolved) Stokes vector.

The vertical dashed lines in Fig. 2 indicate the mean value of the histograms. We refer to it as . They are located at at K for the entire penumbra (dashed red) and at K for the bright penumbral filaments (dashed-blue).

thumbnail Fig. 2

Histograms of ξ(T) as defined in Eq. (1). Red lines correspond to the penumbral region (bottom panels in Fig. 1). Blue lines correspond to the bright penumbral filaments (white contours in the bottom panels of Fig. 1). The vertical dashed lines indicate the mean values of the histograms, .

Open with DEXTER

We have found that both on average and in 98.7% of the studied penumbral area, ξ(T) > 0, which implies that the temperature gradient below the visible surface of the penumbra is lower than the mean temperature gradient of the quiet Sun. A particular interpretation that can be drawn from these results is related to the penumbral thickness below the visible surface. The concepts of thin and thick penumbrae have been discussed on theoretical grounds (see, e.g., Jahn & Schmidt 1994; Solanki 2003), but so far it was not possible to distinguish them observationally. While in a thin penumbra the magnetopause (defined as the current sheet separating the penumbral magnetic field from the surrounding convection) is located at τ5 ≈ 1, in a thick penumbra it lies at greater depth. Therefore, if the penumbra were thin, the temperature between τ5 ∈ [1,3] would converge toward quiet -Sun values, and therefore feature ξ(T) < 0. However, our results show that the opposite occurs: ξ(T) > 0, thereby favoring a thick penumbra.

It has also been suggested that while in the penumbral spines (regions of stronger and less strongly inclined magnetic field) the magnetic field reaches farther down, in the penumbral intraspines (regions of weaker and more inclined magnetic field) the magnetic field quickly yields to field-free convection below τ5 = 1 (see Fig. 3 in Spruit & Scharmer 2006). Interestingly, this latter possibility can be ruled out because the high spatial resolution (0.4−0.45′′) spectropolarimetric data of GREGOR allow us to establish that ξ(T) > 0 also below bright penumbral filaments (i.e., intraspines), both on average and in 97.2% of the area covered by bright filaments.

3.2. Bright filaments vs. granules

Notwithstanding the previous results about the penumbral thickness, there is compelling evidence for the existence of convective-like motions inside the magnetized bright penumbral filaments (Zakharov et al. 2008; Rempel 2012; Tiwari et al. 2013), in particular of some form of convective upflow at their center. In the optically thick region we are studying (τ5 ≥ 1), the thermal stratification T(τ5) is dominated by the adiabatic expansion (i.e., cooling) of the plasma as it rises. Because of this, we might surmise that the temperature gradient in the quiet-Sun granules and bright penumbral filaments would be very similar. Our results indicate, however, that the opposite occurs (see Table 1). A possible solution to this seemingly contradictory results can be offered by considering that while in the quiet Sun this adiabatic expansion can take place in all spatial directions, in bright penumbral filaments (i.e., intraspines) it will be limited by the magnetic pressure and tension of the external field (i.e., spines). Consequently, the expansion of the magnetized upflow inside the bright filament will have fewer degrees of freedom, thereby leading to a decreased cooling and thus the smaller temperature gradient observed.

Acknowledgments

The 1.5-m GREGOR solar telescope was built by a German consortium under the leadership of the Kiepenheuer-Institut für Sonnenphysik in Freiburg with the Leibniz-Institut für Astrophysik Potsdam, the Institut für Astrophysik Göttingen, and the Max-Planck-Institut für Sonnensystemforschung in Göttingen as partners, and with contributions by the Instituto de Astrofísica de Canarias and the Astronomical Institute of the Academy of Sciences of the Czech Republic. Financial support by the Spanish Ministry of Economy and Competitiveness through projects AYA2014-60476-P and Consolider-Ingenio 2010 CSD2009-00038 are gratefully acknowledged. We would like to thank Matthias Rempel, Fernando Moreno-Insertis, and Oskar Steiner for fruitful discussions. This research has made use of NASA’s Astrophysics Data System.

References

All Tables

Table 1

Mean temperatures (in Kelvin) at two different optical depths and their difference.

All Figures

thumbnail Fig. 1

Temperatures in two different regions: quiet Sun (top) and penumbra (bottom). Results at optical depths τ5 = 1 and τ5 = 3 are displayed in the left and middle panels, respectively. Right panels show the normalized continuum intensity images. White contours in the bottom right panel (continuum intensity in the penumbra) enclose bright penumbral filaments, whereas white regions in the penumbral temperature maps (left and middle bottom panels) indicate the umbra. All values of the temperature are given in thousands of Kelvin.

Open with DEXTER
In the text
thumbnail Fig. 2

Histograms of ξ(T) as defined in Eq. (1). Red lines correspond to the penumbral region (bottom panels in Fig. 1). Blue lines correspond to the bright penumbral filaments (white contours in the bottom panels of Fig. 1). The vertical dashed lines indicate the mean values of the histograms, .

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.