Search for photospheric footpoints of quiet Sun transition region loops
J. Sanchez Almeida, L. Teriaca, P. Suetterlin, D. Spadaro, U. Schuehle, R. J. Rutten
aa r X i v : . [ a s t r o - ph ] S e p Astronomy&Astrophysicsmanuscript no. 8124ms c (cid:13)
ESO 2018November 6, 2018
Search for photospheric footpoints of quiet Sun transition regionloops
J. S´anchez Almeida , L. Teriaca , P. S ¨utterlin , D. Spadaro , U. Sch¨uhle , and R. J. Rutten Instituto de Astrof´ısica de Canarias, E-38205 La Laguna, Tenerife, Spaine-mail: [email protected] Max-Planck-Institut f¨ur Sonnensystemforschung, Max-Plank Str. 2, 37191 Katlenburg-Lindau, Germanye-mail: [email protected], [email protected] Sterrenkundig Instituut, Universiteit Utrecht, P.O. Box 80000, NL-3508 TA Utrecht, The Netherlandse-mail:
[email protected], [email protected] INAF-Osservatorio Astrofisico di Catania, I-95123 Catania, Italye-mail: [email protected]
Received 20 June 2007 ; accepted 18 September 2007
ABSTRACT
Context.
The footpoints of quiet Sun Transition Region (TR) loops do not seem to coincide with the photospheric magnetic structuresappearing in traditional low-sensitivity magnetograms.
Aims.
To look for the so-far unidentified photospheric footpoints of TR loops using G-band bright points (BPs) as proxies for photo-spheric magnetic field concentrations.
Methods.
Comparison of TR measurements with SoHO / SUMER and photospheric magnetic field observations obtained with theDutch Open Telescope.
Results.
Photospheric BPs are associated with bright TR structures, but they seem to avoid the brightest parts of the structure. BPsappear in regions that are globally redshifted, but they avoid extreme velocities. TR explosive events are not clearly associated withBPs.
Conclusions.
The observations are not inconsistent with the BPs being footpoints of TR loops, although we have not succeeded to uniquely identifyparticular BPs with specific TR loops.
Key words.
Sun: activity – Sun: magnetic fields – Sun: photosphere – Sun: transition region
1. Rationale
The solar Transition Region (TR) is defined as the part of the so-lar atmosphere characterized by temperatures from 2 × K(upper chromosphere) to 10 K (corona), and densities from ∼ cm − to ∼ cm − (e.g., Mariska 1992). In classi-cal 1-D coronal models (e.g., Gabriel 1976), the TR is a thin( ∼
100 km) thermal interface between the cooler chromosphereand the hotter corona. Although such interface regions must ex-ist at the footpoints of large active region coronal loops, theyappear to be responsible of only a small fraction of the TR emis-sion (e.g., Athay 1982; Feldman 1983). The quiet Sun TR, inparticular, does not represent a continuous transition betweenthe chromosphere and the corona. Rather, sensitive UV obser-vations show the upper solar atmosphere to consist of a hierar-chy of loop structures with di ff erent temperatures and extents(e.g., Dowdy et al. 1986; Feldman et al. 2000; Feldman 2002).Small cool looplike structures fill most of the quiet Sun imagesand spectroheliograms obtained in lines formed at TR temper-atures (see Feldman et al. 1999, Fig. 7). The footpoints of suchloops do not seem to be associated with known traditional mag-netic structures. They lie across network boundaries with thefootpoints presumably in the interior of supergranulation cells.Feldman et al. (2001) find no chromospheric counterpart near Send o ff print requests to : [email protected] the apparent footpoints of the structures. Warren & Winebarger(2000) find that the loops do not connect magnetic struc-tures in full-disk magnetograms obtained with the MichelsonDoppler Imager (MDI) aboard SoHO. This kind of magne-togram, however, does not have enough spatial resolution andsensitivity to reveal magnetic structures in supergranulationcell interiors. In fact, it has been known for a long time thatsuch structures do exist (Livingston & Harvey 1975; Smithson1975). They show up as weak Hanle depolarization signals(e.g., Stenflo 1982; Faurobert-Scholl 1993; Trujillo Bueno et al.2004), weak Zeeman polarization signals (e.g., Wang et al.1995; Lin & Rimmele 1999; S´anchez Almeida & Lites 2000;Dom´ınguez Cerde˜na et al. 2003), and small bright points in in-tergranular lanes (S´anchez Almeida et al. 2004; de Wijn et al.2005). According to numerical simulations (e.g., Cattaneo 1999;V¨ogler et al. 2005; V¨ogler & Sch¨ussler 2007; Stein & Nordlund2006) and observations (e.g., S´anchez Almeida & Lites 2000;S´anchez Almeida et al. 2003; Dom´ınguez Cerde˜na et al. 2006),a complex magnetic field pervades the seemingly non-magnetic quiet photosphere. Theoretical arguments suggest that a sig-nificant part of such photospheric magnetic field actuallyreaches the TR and the corona (Schrijver & Title 2003;Jendersie & Peter 2006). Obviously, these ubiquitous magneticfields seem to be the natural candidates for the so-far unidenti-fied quiet Sun TR loop footpoints. If such conjecture turns outto be correct, it o ff ers a new standpoint for studying and under- S´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops
Fig. 1.
Spectral range of the SUMER observations aver-aged over the full FOV. Spectral radiances are given inW m − s − sr − nm − . All relevant lines are indicated with theknown rest wavelengths (O i from Kelly 1987; C iii and H i Ly γ from Morton 2003). The vertical dashed line marks the bound-ary between the KBr-coated and the bare sections of the SUMERdetector.standing the nature of the TR and its loops. In addition, it wouldprovide a new scientific rationale for studying the magnetismof the quiet Sun. Guided by these ideas, we undertook a firstexploratory study to identify the photospheric footpoints of thequiet Sun TR loops. Such work is described in the present paper.The study requires simultaneous observations of the quietSun TR and the photospheric magnetic fields. The SolarUltraviolet Measurements of Emitted Radiation (SUMER) spec-trometer (Wilhelm et al. 1995) aboard SoHO was used to recordthe 97 nm to 98 nm spectral range (Fig. 1) that includes theH i Ly γ T ∼ . × K) and the C iii T ∼ × K). The latter is one of the brightestlines of the solar VUV spectrum, allowing low-noise spectrato be recorded with exposure times of a few seconds. Beingformed in the middle TR, the C iii ff raction limitedimages in a routine fashion (Rutten et al. 2004).The paper is organized as follows. The observations, reduc-tion, and alignment are described in §
2. Aligning the ground-based images with the satellite images is critical, which justifies the details given in §
2. The observational results are put forwardin §
3. The implications of such results are analyzed in §
2. Observations, data reduction and co-alignment
The observations were carried out on March 25, 2006, from9:00 UTC to 11:00 UTC. Due to poor weather conditions atthe DOT site, it turned out to be the only useful time slot outof three attempts in a coordinated campaign involving SUMER,MDI (Scherrer et al. 1995), and DOT. This section describes thedata sets plus the procedure that allowed us to bring the satel-lite images and the ground-based images to a common referencesystem.
The SUMER instrument is a slit spectrometer and, therefore, im-ages are obtained by raster scanning across the region of interest.Our data consist of six rasters taken sequentially and formingthree pairs. Due to a failure of the SUMER A detector, only astrip of about 20 ′′ along the slit could be imaged and, therefore,a second raster scan was placed 25 ′′ towards the south of theprevious raster. The same region near the center of the solar diskwas observed three times – always in east-west direction – yield-ing the six scans mentioned above. Each single raster is made of99 step positions taken with a cadence of 13 s (12 s exposure)and therefore lasting about 22 minutes. Between rasters the solarrotation was compensated automatically by displacement of thefield-of-view (FOV) towards west. One pair of rasters renders aFOV of about 100 ′′ × ′′ (see Fig. 2). After accounting for solarrotation, the step size of the scan turns out to be 1 . ′′ . ′′ ′′ (Lemaire et al.1997), corresponding to about 1000 km on the Sun at the dis-tance from SoHO to the Sun.The spectra were corrected for flat-field inhomogeneities andfor the geometrical distortion induced by the electronic read-out scheme by using the standard routines available to SUMERusers. We corrected for residual geometrical distortion in thespectral direction, that is significant at the edges of the detec-tor. This was done by straightening the averaged (over the entiredataset) slit spectrum around the O i lines present in the observedspectral range (see Fig. 1). Lines from neutrals and single-ionized species are known to show very small average Dopplershifts and can be used to obtain a wavelength calibration (e.g.,Teriaca et al. 1999, and references therein). Finally, the datawere calibrated to yield spectral radiances in W m − sr − nm − .The recorded spectral range is shown in Fig. 1. The spectra wereused to compute integrated radiances at selected bandpasses to-gether with Doppler velocities, and line widths. Three band-passes are mentioned in the paper: C iii i Ly γ (97.2 nm – 97.3 nm). In order to obtain velocities and linewidths, the line profiles of C iii − and 0.5 pm, respectively. When a rest wave-length of 97.702 nm is used (Morton 2003), the average shiftover the entire dataset is 11 km s − . This value is comparable tothe values found in the literature for lines formed at similar tem-peratures (see e.g., Teriaca et al. 1999, and references therein).The structures appearing in the SUMER C iii ´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops 3 Fig. 2.
SUMER C iii iii ′′ × ′′ ). This fact, together with the lack of an objectivedefinition of C iii elongatedbright structure to describe the elongated patchy structures inFig. 2, keeping in mind, however, that they should be identifiedwith the structures termed loops by Feldman et al. (1999).In addition to the SUMER data, we also use full diskSoHO / MDI magnetograms taken from 9:00 UTC to 11:30 UTCwith a cadence of 1 min. In this case the pixel is 1 . ′′
98 square,with the noise level corresponding to some 16 G (Scherrer et al.1995; Liu & Norton 2001).
This work employs only part of the series of images routinelyprovided by the DOT (Rutten et al. 2004), in particular, we an-alyze G-band images, and Ca ii H line core images. They arerestored using speckle techniques (Rutten et al. 2004) which, un-der good seeing conditions, render di ff raction limited images( ∼ . ′′ . ′′ ′′ × ′′ (see Fig. 3). The timeseries has an irregular cadence equivalent to one image everytwo minutes. The best seeing occurs by the end of the time se-quence and, therefore, it is associated with SUMER rasters 5 and6 (Fried’s parameter 8.5 cm). As expected, the G-band images are full of BPs in intergran-ular lanes (Fig. 3). Automatic identification of BPs is not trivialsince a simple threshold criterion does not su ffi ce. Sometimesgranules are brighter than BPs clearly visible in the intergranularlanes (e.g., Bovelet & Wiehr 2003). We applied a simple algo-rithm consisting of three steps: (1) construction of a smoothedversion of the G-band image which is computed after removalof the brightest features in the image, (2) subtraction of thesmoothed image from the original one to enhance the smallbright features, and (3) selection of the bright features in thesubtracted image, but only when they are localized in dark ar-eas of the smoothed image. The first step, where the brightestfeatures are removed, produces a smoothed version of the imagewhich is not contaminated by the presence of BPs. The di ff er-ence between the full image and this smoothed version enhancesthe contrast of the G-band BPs, and this high-contrast image isused to select the bright features existing in the dark intergran-ular lanes. The algorithm does a good job, in the sense that itagrees with the visual identification of the BPs. It is not perfect,and a few bright borders of granules are misidentified as BP,and some BPs are overlooked. However, the identification suf-fices for the exploratory analysis carried out in this paper. Theuse of a di ff erent algorithm would slightly modify the numberof selected BPs. As we explain in §
3, the trends that we obtainremain the same for all the three SUMER raster pairs, whichcorrespond to di ff erent seeing conditions at the DOT site. Sinceseeing modifies the number of detectable BPs, and it does notchange the trends, the details on how the BPs are detected donot seem to alter our results. S´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops
Fig. 3.
Example of speckle reconstructed G-band image. It shows many G-band bright points tracing magnetic concentrations inintergranular lanes. The spatial coordinates are referred to the center of the image.
The quiet Sun magnetic network is well defined in both stan-dard magnetograms and Ca ii line core filtergrams (e.g., Beckers1977). Our method uses this property to align satellite data(SoHO / MDI magnetograms) with ground-based data (DOTCa ii H images). We bring both SUMER raster scans and DOTG-band images to the spatial coordinates of MDI. Then SUMERraster scans and DOT G-band images can be superposed directly.The actual procedure is explained here in some detail.Although SUMER data and MDI data come from a singlesatellite, they are not co-aligned. Errors in the raster mecha-nism and thermal excursions of the payload produce an unpre-dictable o ff set of up to 10 ′′ . The SUMER to MDI co-alignmenthas been accomplished by comparing the pseudo-continuumSUMER image coming from the bandpass between 97.4 nmand 97.6 nm (Fig. 1), with the average among the MDI magne-tograms taken during the time-span of the SUMER raster scan.The co-alignment is carried by trial and error, blinking on a com-puter screen the images of the SUMER rasters and the absolutevalue of the average MDI magnetogram. One of the images isthen shifted with respect to the other to get the best match. Theresults are illustrated in Fig. 4. Repetitions of the exercise al-ways yield the same o ff set within ± ′′ (optimistic view), and ± ′′ (conservative view). If the procedure uses Ly γ photons in- stead of pseudo-continuum, then the alignment remains the samewithin the quoted uncertainties. We also tried aligning with andwithout removal of the solar rotation when averaging the magne-tograms, and using a logarithm grayscale to represent SUMERradiances. No significant change is observed. The three couplesof SUMER raster scans give the same o ff set.The alignment between DOT and MDI is carried out bymeans of the Ca ii H images. DOT images are both rotated andshifted with respect to SoHO images. The rotation is given bythe angle between the geocentric North and the solar rotationalNorth, which we set according to the ephemeris. As for the rela-tive displacement, we use the same trial and error approach de-scribed above for the SUMER to MDI alignment. Since the reso-lution of the MDI magnetograms is much lower than the specklerestored DOT Ca ii H images, the average among the burst ofimages gathered for speckle restoration is used for comparison.Errors are smaller than the SUMER to MDI alignment since thestructures observed in DOT Ca ii H and MDI are quite simi-lar and therefore easy to identify. After repeating the trial anderror process several times, one finds the displacements to beconsistent within 1 ′′ . DOT G-band images and DOT Ca ii H arealso misaligned. We find the shift between the images by cross-correlation. This method does not correct for the slight di ff erentorientation of the two images, and for a small di ff erence of the ´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops 5 Fig. 4.
Example of alignment between a SUMER pseudo-continuum image and the absolute value of the average MDI magne-togram. The latter is shown as a contour plot, with contours at 7 G and 15 G. The spatial coordinates are referred to the solar diskcenter according to the MDI scale.spatial scales. However, the two e ff ects leave a residual error al-ways well below 1 ′′ .In short, the critical part of the alignment has been carriedout by trial and error and, therefore, its uncertainty is di ffi cult toestimate. However, judging the errors by the consistency of thetrial and error process, they should be smaller than 2 ′′ . This un-certainty is mostly set by the SUMER to MDI alignment, makingall other errors negligible.
3. Observational results
The best DOT seeing occurred by the end of the time series,corresponding to SUMER rasters 5 and 6. The number of G-band BPs in an image depends critically on the spatial resolu-tion (Title & Berger 1996; S´anchez Almeida et al. 2004), there-fore, our analysis is focused on these last rasters and the bestG-band image taken together with them. We also analyzed theother raster pairs, and other snapshots of the time series. Theresults are always consistent with those reported below.
The G-band images show a significant number of BPs with noobvious counterpart in the MDI magnetograms (Fig. 5). As wepoint out in §
1, a significant number of photospheric magneticstructures, and so of TR loop footpoint candidates, does notshow up in traditional measurements. The G-band BPs existingin the DOT image of best angular resolution (Fig. 3) are over-plotted on the SUMER C iii iii iii X ≃ ′′ and Y ≃ ′′ (see Fig. 2). In order to quantify this im-pression, we computed the histogram of C iii Fig. 6. (a) Histogram (PDF) of the logarithm of the SUMERC iii iii iii
S´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops
Fig. 5.
MDI magnetogram (the image) together with the BPs existing in the G-band image (red contours). The magnetogram hasbeen scaled from -20 G to 20 G, so that only the black and white pixels are well above the noise level ( ∼
16 G). The coordinatesare referred to the MDI solar disk center. The rectangle indicates the DOT G-band FOV. Note that these two images have not beenaligned directly but via Ca ii H imagesradiances at the G-band BPs turns out to be shifted toward largeradiances (see the dashed line). This shift is a feature commonto all three pairs of raster scans. Note also the lack of very brightfeatures in the G-band histogram – the dashed line in Fig. 6adrops down for radiances larger than 2500 mW m − sr − . Thisdrop may unveil the tendency for the BPs to avoid the core ofthe bright C iii Figure 7 contains the map of velocities derived fromC iii . ± . − (full FOV) and 12 . ± . − (BPs), which renders a relative shift of the order of 1.9 km s − .The existence of such excess of redshift is a very robust result.The velocity histograms are made of hundreds of points, andthey are well represented by Gaussians with widths of about5 km s − . The histograms of the means are necessarily muchnarrower, with widths of the order of 5 km s − divided by thesquare root of the number of individual velocities used to com-pute the means (e.g., Martin 1971). In our case this uncertaintyof the means ( ≤ − ) is much smaller than the separationbetween the means. Two final comments are in order. First, thedi ff erence of velocities cannot be caused by systematic errors bi-asing our wavelength calibration, since they would a ff ect the twohistograms in the same way. Second, similar shifts are present inall raster scans, reinforcing the results.Nothing special seems to be associated with the maps ofC iii iii − , but they really contain muchlarger spatially unresolved upflows and downflows. The line ´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops 7 Fig. 7. C iii Fig. 8. (a) Histogram (PDF) of the distribution of C iii − at C iii Explosive Events (EEs) are believed to be the result of mag-netic reconnection. They can be identified because the line pro-files are strongly non-Gaussian, and they often coincide with lo-cations of very large Doppler shifts and / or line widths derivedby single-Gaussian fitting (Dere et al. 1989; Teriaca et al. 2004).We run a procedure to identify EEs by comparing the result ofa single-Gaussian fit with the observed profiles. All profiles forwhich at least three contiguous spectral pixels consistently devi-ating by more than two sigma from the fit are flagged as EE. Inour case we find 26 such spectra clustered in at least 7 patches.Unfortunately, most EE are outside the DOT FOV – the EEs se-lected in rasters 5 and 6 are represented in Fig. 10. From this veryreduced statistics, we find no clear overlapping between EEs andG-band BPs, although each EE is not very far from a BP either.The spatial separation between BPs and EEs does not seem to bean artifact due to the lack of simultaneity between the SUMERspectra and the DOT image chosen to represent the photosphere.(Both EEs and BPs are transitory events lasting shorter than theSUMER scans. The DOT snapshot of best seeing may miss BPsexisting during the individual EEs, masking a putative relation-ship between EEs and BPs.) BPs do not coincide with EEs evenwhen the G-band image closest in time to each EE is used tosearch for BPs. Another possible bias has to do with the criterionto select EEs, which may overlook some of the weaker events orevents that happen to produce a very broad but fairly Gaussianprofile. However, as pointed out in § S´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops
Fig. 9. C iii
4. Discussion
As far as we are aware of, this work represents the first attemptto find footpoints of quiet Sun TR loops in the interior of super-granulation cells. G-band bright points (BPs) are used as proxiesfor photospheric magnetic field concentrations which may an-chor and guide the TR C iii ff ort to provide SUMERC iii ffi culties have to do with iden-tifying intrinsically fuzzy forms in our reduced FOV ( § loop to denote the elongatedstructures present in the C iii iii iii X ≃ ′′ and Y ≃ ′′ ) is surroundedby two chains of BPs. It is tempting to think of the C iii iii iii iii ∼ × K) are systemati-cally redshifted. The corresponding plasma velocities turn out tobe in the range between 8 and 15 km s − . We find that the posi-tion of the C iii ± − (Fig. 6), in goodagreement with the predictions of the TR loop model. Moreover,C iii − with respect to the average TR ( § ´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops 9 Fig. 10.
Image of the pixels with EEs (non-Gaussian line profiles) together with the G-band BPs. The EE pixels are represented inblack. The rest of the layout remains as for Fig. 2.results are consistent with the BPs being footpoints of loop-likestructures.Explosive Events (EEs) are believed to be the result of mag-netic reconnection. Where the reconnection takes place is underdebate. There are authors proposing for reconnection in eitherthe photosphere / chromosphere (the reconnection in the photo-sphere leads to shocks that accelerate the plasma at TR tem-peratures, Tarbell et al. 2000), the TR (direct formation of bi-directional jets at the reconnection site, e.g., Innes et al. 1997),and the corona (reconnection high in the corona generates highenergy particle beams that heat and accelerate the chromosphericplasma leading to the TR signature, Benz & Krucker 1999). Wehave identified several of those events in our FOV, finding a ten-dency to avoid BPs. This is a new result suggesting that EEsare not located low in the chromosphere (at least not lower thanthe point where the field starts expanding significantly). If BPsare footpoints of loops undergoing reconnection, then the factthat the EEs do not coincide with them indicates that site ofTR plasma acceleration is far from the photospheric footpoints.This would exclude the hypothesis of a flare-like mechanism, asthe particle beams reaching the loop footpoints would result inplasma accelerated at, or very close to, the footpoints. In case ofshocks formed by reconnection in the photosphere, these shockstravel more than one Mm (distance between the observed EEsand the closest BPs) before being dissipated. Reconnection andconsequent heating and plasma acceleration in the low chro-mosphere seems also excluded. However, it should be men-tioned that EEs have a weak signature in chromospheric lines(Teriaca et al. 2002) and that there is some evidence that EEsare first observed in chromospheric lines, and then in TR lines(Madjarska & Doyle 2002).As mentioned at the beginning of the section, this work rep-resents a first attempt to outline research avenues for identify-ing the footpoints of quiet Sun TR loops. Clearly, the temporaland the angular resolution of both the visible and the UV data must be improved to proceed further. Observations obtained withSUMER, DOT and instruments on the satellite Hinode are ex-pected to yield such improvements. Acknowledgements.
The authors acknowledge the use of the Solar Soft pack-age for data reduction and analysis. This work has been partly funded bythe Spanish Ministry of Education and Science (AYA2004-05792), and by theItalian Space Agency (ASI I / / / References
Athay, R. G. 1982, ApJ, 263, 982Beckers, J. M. 1977, in Illustrated Glossary for Solar and Solar-TerrestrialPhysics, ed. A. Bruzek & C. J. Durrant (Dordrecht: Reidel), 21Benz, A. O. & Krucker, S. 1999, A&A, 341, 286Berger, T. E., L¨ofdahl, M. G., Shine, R. A., & Title, A. M. 1998, ApJ, 495, 973Berger, T. E., Schrijver, C. J., Shine, R. A., et al. 1995, ApJ, 454, 531Bovelet, B. & Wiehr, E. 2003, A&A, 412, 249Cattaneo, F. 1999, ApJ, 515, L39de Wijn, A. G., Rutten, R. J., Haverkamp, E. M. W. P., & S¨utterlin, P. 2005,A&A, 441, 1183Dere, K. P., Bartoe, J.-D. F., & Brueckner, G. E. 1989, Sol. Phys., 123, 41Dom´ınguez Cerde˜na, I., Kneer, F., & S´anchez Almeida, J. 2003, ApJ, 582, L55Dom´ınguez Cerde˜na, I., S´anchez Almeida, J., & Kneer, F. 2006, ApJ, 636, 496Dowdy, J. F., Rabin, D., & Moore, R. L. 1986, Sol. Phys., 105, 35Faurobert-Scholl, M. 1993, A&A, 268, 765Feldman, U. 1983, ApJ, 275, 367Feldman, U. 2002, in From Solar Min to Max: Half a Solar Cycle with SOHO,ed. A. Wilson, ESA SP-508 (Noordwijk: ESA Publications Division), 531Feldman, U., Dammasch, I. E., & Wilhelm, K. 2000, Space Sci. Rev., 93, 411Feldman, U., Dammasch, I. E., & Wilhelm, K. 2001, ApJ, 558, 423Feldman, U., Widing, K. G., & Warren, H. P. 1999, ApJ, 522, 1133Gabriel, A. H. 1976, Phil. Trans. Roy. Soc. Lond. Series A, 281, 339 http://solar-b.nao.ac.jp/index e.shtml0 S´anchez Almeida et al.: Photospheric footpoints of quiet Sun TR loops