The Chromospheric Telescope
C. Bethge, H. Peter, T. J. Kentischer, C. Halbgewachs, D. F. Elmore, C. Beck
AAstronomy & Astrophysics manuscript no. chrotel c (cid:13)
ESO 2018October 15, 2018
The Chromospheric Telescope
C. Bethge , , H. Peter , , T. J. Kentischer , C. Halbgewachs , D. F. Elmore , and C. Beck Kiepenheuer-Institut f¨ur Sonnenphysik, Sch¨oneckstr. 6, 79104 Freiburg, Germany High Altitude Observatory, National Center for Atmospheric Research (cid:63) , P.O. Box 3000, Boulder, CO 80307, USAe-mail: [email protected] Max-Planck-Institut f¨ur Sonnensystemforschung, 37191 Katlenburg-Lindau, Germany National Solar Observatory, 3010 Coronal Loop, Sunspot, NM 88349, USA Instituto de Astrof´ısica de Canarias (CSIC), Via Lactea, E-38205 La Laguna, Tenerife, SpainReceived 9 June 2011 / Accepted 27 July 2011
ABSTRACT
Aims.
We introduce the
Chro mospheric
Tel escope (ChroTel) at the Observatorio del Teide in Iza˜na on Tenerife as a new multi-wavelength imaging telescope for full-disk synoptic observations of the solar chromosphere. We describe the design of the instrumentand summarize its performance during the first one and a half years of operation. We present a method to derive line-of-sight velocitymaps of the full solar disk from filtergrams taken in and near the He i infrared line at 10830 Å. Methods.
ChroTel observations are conducted using Lyot-type filters for the chromospheric lines of Ca ii K, H α , and He i i line allows us to determine line-shifts by calibrating the line-of-sight velocity maps derived from thefiltergram intensities with spectrographic data from the Tenerife Infrared Polarimeter at high spatial and spectral resolution. Results.
The robotic operation and automated data reduction have proven to operate reliably in the first one and and half years. Theachieved spatial resolution of the data is close to the theoretical limit of 2 arcsec in H α and Ca ii K and 3 arcsec in He i . Line-of-sightvelocities in He i can be determined with a precision of better than 3-4 km s − when co-temporal spectrographic maps are availablefor calibration. Conclusions.
ChroTel o ff ers a unique combination of imaging in the most important chromospheric lines, along with the possibilityto determine line-of-sight velocities in one of the lines. This is of interest for scientific investigations of large-scale structures in thesolar chromosphere, as well as for context imaging of high-resolution solar observations. Key words.
Sun: chromosphere - Telescopes
1. Introduction
Full-disk observations of the Sun, from both the ground andspace, have always played an important role in solar physics be-cause they are not only essential to global solar studies, e.g.,for helioseismology and solar activity, but also provide contex-tual information for observations at high resolution that studyonly small parts of the solar disk. Among the most prominentexamples are the GONG network (Harvey et al. 1996) and theMDI instrument onboard SOHO (Scherrer et al. 1995). It iswidely believed that key insight emerges predominantly fromhigh-resolution observations advancing to smaller and smallerstructures in the Sun’s atmosphere. Ground-based full-disk in-struments commonly share the disadvantage of a comparablylow spatial resolution. They must be operated seeing-limited be-cause adaptive optics can correct only for a small portion of thesolar disk. However, despite this disadvantage, they are still vi-tal in solar research - first and foremost for structures on largerscales up to the complete disk, but also for rarely occurringevents that are often missed when examining only a small frac-tion of the disk. Here, we present such an instrument for obser-vations of the solar chromosphere.Chromospheric full-disk observations are often performed inthe spectral lines of Ca ii K (3933.7 Å) and H α (6562.8 Å) be- (cid:63) The National Center for Atmospheric Research is sponsored by theNational Science Foundation. cause both lines are known to indicate the presence and the struc-ture of magnetic fields. Prominent examples include the RISE-PSPT instruments observing in Ca ii K (Ermolli et al. 1998) atthe Osservatorio Astronomico di Roma (OAR) and the MaunaLoa Solar Observatory (MLSO), as well as the H α patrol instru-ment of the Kanzelh¨ohe Observatory (Otruba 1999) in Austria,which is part of the global H α network (Steinegger et al. 2000).The He i triplet at 10830 Å has become an increasinglypopular diagnostic of the chromosphere, although its forma-tion process is complicated. In spectropolarimetry, this is dueto the sensitivity of the polarization signal to both the Hanle andthe Zeeman e ff ect, rendering the line(s) an ideal tool for bothstrong and weak magnetic field regions (e.g., R¨uedi et al. 1995;Trujillo Bueno et al. 2005). However, it is also very useful forfull-disk observations because the line is formed entirely in theupper chromosphere (that is at chromospheric temperatures) orthe lower transition region with no photospheric contributions.In addition, the line formation is strongly influenced by coro-nal emission because the triplet states are mainly populated bythe photoionization-recombination mechanism (Centeno et al.2008). This permits an estimation of coronal activity in ground-based observations - coronal holes for example are clearly vis-ible in He i ff er from photospheric contamination inregions with little photospheric activity and / or very low coronal a r X i v : . [ a s t r o - ph . S R ] A ug . Bethge et al.: The Chromospheric Telescope illumination because the line becomes so optically thin that thephotospheric continuum shines through.Telescopes observing in He i are CHIP (Elmore et al.1998) at MLSO and the OSPAN instrument (formerly ISOON,see Neidig et al. 1998) at the National Solar Observatory atSacramento Peak. The latter also provides observations in H α .The Chromospheric Telescope ChroTel introduced hererecords data in all three spectral lines (Ca ii K, H α , He i ff erent lines sample di ff erent tempera-ture regimes (and therefore heights), this gives a more holisticview of phenomena in the chromosphere and their connectionto higher layers. In addition, the data in He i are taken at sevenwavelength positions in and around the He i triplet with a tunableLyot-type filter. As we later demonstrate, the line-of-sight (LOS)velocities across the full disk can be determined with these fil-tergrams, which is essential for investigating the dynamics inthe chromosphere. Since it is located on Tenerife, i.e., in a dif-ferent time-zone from both the CHIP and OSPAN instruments,ChroTel will also increase the available temporal coverage offull-disk data in He i i . Finally, we makesome concluding remarks on the data quality and the potentialof the method for the velocity determination.
2. Scientific scope
ChroTel observes a wide range of solar phenomena of scientificimportance, some of which are presented here. This is of courseneither complete nor representative of what can be achieved withthe data, but gives an impression of the potential of the instru-ment.
Filaments (and prominences) are naturally a subject of chromo-spheric full-disk observations both for statistical reasons and be-cause they often occupy a significant fraction of the solar disk.The measurement of their LOS velocities permits the examina-tion of large-scale flows within filaments. Along with informa-tion about the magnetic field configuration, this can serve as animportant ingredient of filament models, especially for the onsetof flares and coronal mass ejections (CMEs).Related to this are Moreton waves (Moreton & Ramsey1960), which are horizontal waves traditionally observed inH α that propagate away from flare sites at speeds of up to2000 km s − . Authors have tried to establish connections bet-ween EIT waves (the name stems from observations of the waveswith the Extreme ultraviolet Imaging Telescope, see Thompsonet al. 1998) higher up in the atmosphere and Moreton waves,primarily on the basis of the co-spatiality of the two phenom-ena. While some authors indeed find co-spatial features (e.g.,Thompson et al. 2000; Warmuth et al. 2002), others come to adi ff erent result (e.g., Eto et al. 2002; Okamoto et al. 2004) andquestion whether a connection exists.Investigations of observations in an intermediate layer seemfavorable and have indeed been done with data in He i α , He i , and EIT maps,with a forerunner in He i with respect to H α . They conclude that the waves are of a mechanical nature, whereas Gilbert et al.(2004) suggest that the co-spatial features they find in He i withrespect to EIT data indicate a ‘chromospheric imprint’ of a wavein the corona triggered by the enhanced coronal EUV emission,which then again leads to an increase in the He i absorption. Withalmost co-temporal data in both H α and He i , this controversycan be investigated using ChroTel data. As Gilbert et al. (2004)conclude in their paper: “In any case, a high-time cadence He i λ − .This would have to be done in a dedicated mode of operationlooking for these events, e.g., by taking images with a high ca-dence only in the line cores of H α and He i . Since Moreton wavestravel a significant distance during the acquisition time of about8 s for the whole He i filtergram sequence, the determination ofLOS velocities in He i is of limited use for this purpose, but thepropagation of the waves can also be traced in high-cadence in-tensity images, e.g., using the di ff erences between consecutiveimages. Magnetic flux in the quiet Sun appears in two di ff erent shapes:di ff use and weak magnetic flux in the internetwork, and con-centrated magnetic field in both the network and isolated fluxtubes. A possible method for the transition from the di ff use tothe concentrated state was proposed by Parker (1978), which hecalled the superadiabatic e ff ect : when the magnetic field linesare shu ffl ed to and ‘squeezed’ at the boundaries of supergran-ules, the magnetic field strength rises and suppresses convection.In these regions, this leads to a cooling of the plasma which is al-ready in a general downdraft region. The downdraft is further en-hanced by the cooling, through which plasma in the upper partsof a filled flux tube can lose its support from below and the fluxtube becomes evacuated rapidly. The plasma from above shouldthen (nearly) undergo a free-fall motion in this picture, which isusually referred to as convective collapse . These motions indeedwere observed in a plage region by Schmidt et al. (2000) in H β and He i − . The motions lasted for about 90 s, whichagrees with the results of Grossmann-Doerth et al. (1998) whoinferred a duration of several minutes for convective collapsesfrom 2D simulations. It is puzzling, however, that the reports ofthese events are very sparse, although Parker’s scenario shouldbe rather general and occur essentially on an everyday basis forthe Sun.Lagg et al. (2007) also report high-velocity downflows inHe i , but in the vicinity of a growing pore. The downflows werelasting (and increasing) for over an hour, so the convective col-lapse scenario in this case is rather unlikely. The authors inter-preted their results as a continuous drainage of a slowly risingmagnetic flux tube, leading to increasing downflows in the foot-points of the loop.With ChroTel, it will be possible to improve the quality of thestatistics of the rates and properties of these events. The down-flows in the aforementioned observations were seen in regionsextending from about 5 to 30 arcseconds, with velocities up to42 km s − . In Sect. 5, we demonstrate that this should be easilyaccessible with ChroTel data in He i .
2. Bethge et al.: The Chromospheric Telescope
It has been known for many decades that the solar wind exhibitsa slow and a fast component, the latter having average speeds of700-800 km s − (e.g. Phillips et al. 1995). It is also known thatthe fast component emerges from coronal holes (e.g., Kriegeret al. 1973), i.e., regions appearing darker than their surround-ings at EUV and X-ray wavelengths, although there are indica-tions that it might also emerge from quiet Sun regions (Habbalet al. 1997).It is still unclear, however, at which height and tempera-ture the acceleration of the fast solar wind sets in. Dupree et al.(1996) investigated wing asymmetries in He i ff ect for the velocities, i.e.,the amplitudes of the velocities were found to be dependent onthe cosine of the heliocentric angle. This was interpreted as a ra-dial outflow in the coronal holes and as the onset of the fast solarwind at chromospheric heights.Tu et al. (2005) investigated SUMER (Wilhelm et al. 1995)scans of a polar coronal hole in the emission lines of Si ii , C iv ,and Ne viii . The Ne viii emission forming in the lower corona at6 × K was found to be blueshifted on average by 9 km s − ,whereas in the C iv emission that is formed at 1 × K,only weak signatures of outflows were seen at the disk center(Dere et al. 1989) and at the poles (Peter 1999; Peter & Judge1999). The question of course arises of how the blueshifts inHe i iv higher up in the atmosphere.Along with co-temporal velocity information about the tran-sition region, ChroTel can help us to determine whether the out-flows seen by Dupree et al. (1996) are indeed a signature of thefast solar wind in the chromosphere or should be interpreted ina di ff erent way. As we show in Sect. 5, coronal holes pose achallenge to the velocity determination with ChroTel filtergramsbecause the intrinsic He i absorption is very weak in these re-gions. However, this might be addressed with the averaging ofdata because coronal holes are long-lived phenomena.
3. The instrument
The configuration of the instrument is described in detail inKentischer et al. (2008) and Halbgewachs et al. (2008), so onlya brief overview is given here.
The outdoor structure of the telescope is a domeless turret inan alt-azimuth mount. The turret is sealed to protect the opti-cal elements from wind, humidity, and extreme temperatures.Figure 1 a) shows the telescope structure attached to the build-ing of the Vacuum Tower Telescope (VTT) on Tenerife.Because of the alt-azimuth mount both the azimuth and theelevation axis have to be rotated in order to follow the positionof the Sun during the day. The guiding is conducted in threeconsecutive steps. First, the values for azimuth and elevation ofthe Sun on the sky are calculated from ephemerides, not takinginto account atmospheric refraction. Second, a position-sensingdetector (PSD) corrects small and slowly varying deviations ofthe Sun’s designated position within a field-of-view (FOV) of2.2 ◦ . The last step is the compensation of fast image movementswith a tip-tilt mirror (M5 in Fig. 1 b)). It operates with a closed control loop at a cycle time of 108 µ s and within a FOV of 0.7 ◦ .The pointing accuracy of this system is better than 0.5 arcsec. The light enters through a wedged and anti-reflection-coatedentrance window (W1). The window is protected by a shutterwhen the instrument is not operating to minimize contamina-tions by dust and humidity. The light is redirected to the opticslab inside the VTT building by the flat telescope mirrors M1-3and imaged onto a single CCD by a combination of achromaticlenses (L1: f = =
800 mm; L5:f =
600 mm; L6: f =
500 mm; L7: f =
800 mm) and field lenses(L2-3). The aperture of 100 mm leads to a theoretical resolutionlimit of 1 arcsec for the calcium channel, 1.7 arcsec for H α , and2.7 arcsec for He i . The spatial sampling on the 2048x2048 pixelCCD is about 1 arcsec per pixel, i.e., in the first two channels,the spatial resolution is sampling limited to about 2 arcsec bythe detector.All channels are recorded with one camera, hence the CCDmust be sensitive over a large wavelength range from 393 nmto 1083 nm. A Kodak KAF-4320E CCD was selected, operatedin a water-cooled Spectral Instruments Series 800 camera. Thedetector has a dynamic range of 14 bit, a pixel size of 24x24 µ m,and a quantum e ffi ciency (QE) of around 35% for Ca ii K, 63%for H α , and 2% for He i .To perform observations in all three spectral lines, the beamhas to be redirected successively to three Lyot-type filters (seeSection 3.3). For this, an in-house developed filter exchange unit(FEU) is used: the filters are mounted in an equilateral trianglesetup with static flat mirrors both at the entrance and the exit ofthe filters (M8-9a,b,c in Fig. 1 b)). These mirrors are tilted by45 ◦ with respect to the optical axis of the filters. Two rotatablemirrors (M7 and M10), also tilted by 45 ◦ , redirect the light toone of the static mirrors sets, hence through one of the Lyot fil-ters. Rotating the (coupled) mirrors M7 and M10 by 120 ◦ aroundthe axis coinciding with the direction of the incident light thenallows a di ff erent filter to be selected. The mirrors M7-10 havea reflectivity of 90% at the relevant wavelengths, i.e., the FEUentails a loss of more than one-third of the incident intensity.Along with the low QE of the detector in the Ca ii K and He i wavelengths, this involves comparably long exposure times ofup to 1 s in these channels as shown in Table 1.Flatfielding is done using a di ff user plate that can be insertedinto the beam when necessary. The di ff user plate is custom-madeand generates a uniform intensity over a larger angular rangethan regular holographic di ff user plates. It is mounted within theoutside telescope structure to take care of contaminations on allsubsequent imaging optics. The di ff user plate rotates during theflatfield exposures to compensate for possible inhomogeneitiesin the manufacturing process. To do this, the period of the ro-tation is chosen as an integer multiple of the exposure time.Flatfields are currently taken once per hour in every channel. For each of the spectral lines, a Lyot-type birefringent narrow-band filter is used for the observations. All filters containwide-fielded elements for a larger monochromatic acceptanceangle. We now provide a short description of each Lyot filter:Ca ii K: The wavelength of the filter passband is adjustedto the central minimum of the Ca ii K line at 3933.7 Å. TheFWHM of the filter passband in the narrowest possible mode is
3. Bethge et al.: The Chromospheric Telescope
Diffuser W1Weather sensorsL3 L2 M3M2 M1 a) L1PSD
L1 L6CCD F3 P1 L2 F1L3 P1 L5F2M8a,b,cL7a,b,cM9a,b,cPSDW1M1 M2M3 L4 M5 (Tip−Tilt) GuiderFineM4 Lyot Filter M7M10 M6P2 b) Filter Exchange UnitDiffuser
Fig. 1. a) Telescope structure outside the optical lab. b) Schematic overview of the optical path (not to scale): W1 entrance window, M1-3 telescopemirrors, L1 telescope lens with aperture, PSD position sensing detector, L2-3 field lenses, L4-7a,b,c achromatic lenses, M5 tip-tilt mirror, M7 / M10rotatable mirrors, M8-9a,b,c static mirrors, F1-3 focal planes, and P1-2 pupil planes.
Table 1.
Overview of the observed spectral lines. The values in italics indicate the default settings for the synoptic observations.Spectral line wavelength Lyot filter widths typical exposure minimum cadence [s][Å] FWHM [Å] time [ms] single channel all channelsCa ii K 3933.7 ; 0.6; 1.2 1000 10 60H α ; 1.0 100 10 60He i r and K v emission peak up to about 0.25 Å on thered and blue side of core of the line. Detachable polarizers at theentrance and the exit of the filter also allow filter passbands witha FWHM of 0.6 Å and 1.2 Å. Figure 2 shows a measurementfrom 2011 of the narrowest possible filter passband. The FWHMof the central transmission peak determined from a Gaussian fitto the curve is 0.29 Å. The measurement also revealed a slightlyenhanced contribution from a side lobe in the red wing of the Ca ii K line.H α : The FWHM of the filter passband in the default settingis 0.5 Å. It can be widened to 1.0 Å when a detachable entrancepolarizer is removed. An additional contrast element can beused to suppress side lobes of the transmission curve. Figure 3shows a measurement of the transmission curve from 2009without and with the contrast element. In the latter case, the sidelobes are significantly reduced, whereas the central wavelengthof the transmission peak is shifted slightly towards the blue by0.1 Å. This is nevertheless the preferable (and default) setting
4. Bethge et al.: The Chromospheric Telescope re l a t i v e t r a n s m i ss i o n Δλ [Å]
Fig. 2.
Measured passband (grey dashed line) of the Lyot filter used forChroTel Ca ii K observations. The zero point marks the minimum ofthe Ca ii K line at 3933.7 Å. For comparison, a spectrum around theCa ii K line is overplotted (solid line, recorded with the spectrograph ofthe VTT). because the wings of the line are so bright compared to the corethat the fraction of the intensity coming from the side lobesis over 70% when the contrast element is not used. With thecontrast element, this drops to 40%, i.e., despite the shift, thelargest fraction of the signal comes from the core of the line.The FWHM of the central transmission peak determined from aGaussian fit is 0.48 Å.He i i infrared triplet. It was assembled by the High AltitudeObservatory based on the design of the filter for the CHIP instru-ment (Elmore et al. 1998). The filter contains liquid crystal vari-able retarders (LCVR) between the birefringent stages, which al-low a rapid adjustment of the central wavelength position of thetransmission peak. CHIP acquires filtergrams in and around theHe i line at seven fixed wavelength positions in a non-equidistantorder. These wavelengths were also chosen for ChroTel to en-sure the comparability of the data of the two instruments. InSect. 5, we show that these filtergrams can be used to determineline-shifts in He i i Lyot filter from 2006. The overplotted He i profile comes from a plage region. With its enhanced line depth,it shows a favorable case for the determination of velocities. Inthe quiet Sun, the line depth is substantially smaller than for theprofile shown here.The mean FWHM of the passbands obtained from polyno-mial fits to the data is 1.29 Å. Shifting the central wavelengthposition with the LCVRs is done in less than 60 ms, i.e., witha typical exposure time of 1000 ms, all filtergrams are obtainedwithin less than 8 s.
4. Operation
ChroTel was designed as a monitoring instrument for synop-tic observations. It operates autonomously and continuously,weather permitting, and can also be remote-controlled. The in-strument starts operating automatically in the morning and shutsdown in the evening, or in case of unsafe conditions, which arebeing recognized based on data from weather sensors for wind,humidity, and brightness. As soon as the conditions are safeagain after the instrument has shut down, observations continueautomatically.
Δλ [Å] re l a t i v e t r a n s m i ss i o n -6 -4 -2 0 2 4 61.00.80.60.40.20.0 Fig. 3.
Measured passbands (grey dashed lines) of the Lyot filter usedfor ChroTel H α observations. The zero point marks the minimum ofthe H α line at 6562.8 Å. Top: filter passband without contrast element.
Bottom: filter passband with contrast element. For comparison, a spec-trum around the H α line is overplotted (solid lines, recorded with thespectrograph of the Schauinsland Observatory). re l a t i v e t r a n s m i ss i o n wavelength [Å] He I tripletSi I H O i = Fig. 4.
Measured passbands of the tunable Lyot filter used for ChroTelHe i Grey dotted lines: measured values.
Greysolid lines: polynomial fit to the measurements. Central wavelength po-sitions for the passbands in Ångstroms (Å), from left to right: 10827.45,10828.47, 10829.60, 10830.30, 10831.00, 10832.13, 10833.15. Forcomparison, a spectrum around the He i triplet from a plage region isoverplotted in black (smoothed; observed with the Tenerife InfraredPolarimeter). All involved hardware components are controlled by a tele-scope control software written in LabView. The software allowsone to initiate actions such as taking flatfields, choosing chan-nels, and defining exposure times etc. within a text script. By de-fault, a standard script is run for the regular operation describedin Sect. 4.1. Other scripts can be used to customize the modeof operation, e.g., to coordinate observations with the nearbyspectrograph at the VTT. Requests for specific modes of oper-ation can be addressed to the Kiepenheuer Institute (Freiburg,Germany). While no limitations on observation requests are ine ff ect at the moment, requests for high-cadence observationsover extended periods of time might be denied or limited in ordernot to endanger the usefulness of the synoptic program. Send observation requests to: [email protected]. Bethge et al.: The Chromospheric Telescope
Table 2.
Overview of the data products from ChroTel.Data type Cadence Resolution Format Availability[min] [pixels]Scientific 3 2048x2048 FITS Next dayPreview and 3 1024x1024 JPG Few minuteslive pictures after recordingMovies 3 512x512 AVI Next day
An image is acquired in each channel once every three minutesin the regular mode of operation. Small variations in the timeconsumed by computer communications limit the precision ofthe cadence within one channel to ± ff set of about 10-15 s between the channels.Single-channel observations can be done upon request witha minimum cadence of 10 s in Ca ii K and H α and 30 s in He i .The minimum cadence for observations in all three channels isone minute. All data reduction, processing, and transfer is automatized.Every acquired image is instantly corrected for bias and gainwith the most recent dark and flatfield exposures. In practice,this means that first the dark current is subtracted from both theflatfield and the observational data. All zero values in the flat-field are then subsequently replaced by unity to avoid a divisionby zero. A gain table is computed then from the flatfield, i.e., thepixel values are rescaled with a division by the median value to arange around unity to maintain the dynamic range of the obser-vational data. As a last step, the observational data are dividedby the gain table. The reduced data are then saved as 16-bit in-teger files in gzipped FITS format. As a byproduct, live imagesare created as 1024x1024 pixel JPG files, which are available afew minutes after recording on the website of the KiepenheuerInstitute. They are also used as a guiding aid for the nearby VTTand Gregor telescopes and later on as preview images for thedata. Observations acquired 300 days per year and 7 hours perday would generate a yearly amount of data of about 5 TB.Overnight, the data are transferred from Tenerife to Freiburgto make them accessible for public use the next day. This in-cludes the automated creation of daily overview movies andwebpages for all channels. Table 2 gives an overview of the pro-vided data products and the times they are available . The robotic operation and automatic data processing has provento be reliable in the first one and a half years of operation. Any All data are open for scientific usage and can be accessed at thefollowing addresses:Live / most recent images:http: // / index.php?id = = // archive.kis.uni-freiburg.de / pub / chrotel / Overview of available data:http: // / ˜chrotel / index.html outages were caused mainly by bad weather, maintenance, oroverall facility shutdowns in the winter.Figure 5 presents some ChroTel full-disk images in all chan-nels acquired on 31 August 2010. In addition, cutouts are shownto give a visual impression of the full spatial resolution of theimages. The bottom row demonstrates that the smallest visiblestructures in H α and He i are close in size to the theoretical reso-lution limits of 2 arcsec in H α and 2.7 arcsec in He i . In Ca ii K,the theoretical limit of 2 arcsec is not entirely achieved owing tothe long exposure time that is necessary in this channel and thehigher susceptibility to seeing at shorter wavelengths.The mean spatial resolution of the data was computed forthree sample days in summer 2010 (July 26, July 31, August 3).In a box of 800x800 pixels covering both quiet Sun and activeregions, the Fourier power as a function of spatial frequency wasdetermined. A flat part in this curve, i.e., when the Fourier poweris the same for all frequencies, represents the Fourier transfor-mation of white noise where no information is left in the data.The point where the curve becomes flat therefore marks the res-olution limit. The determination of this point is not exact, butnevertheless gives a rough estimate of the spatial resolution ofthe data. For the data of the aforementioned three days, the meanspatial resolution in all channels turned out to be about two timesworse than the theoretical resolution limit, i.e., the smallest re-solved structures are about twice as large. This is predominantlycaused by deteriorating seeing conditions in the afternoon.For the same days, the mean scattered light at 1.1 · R (cid:12) was de-termined as 0.13 in Ca ii K and 0.04 in both H α and He i relativeto the mean intensity on the solar disk.
5. Line-of-sight velocities in He i Figure 4 illustrates that a range of ± i triplet iscovered by the seven filtergrams. Theoretically, this should en-able us to detect LOS velocities of up to about ±
80 km s − . Onepossible way to determine the position λ line of the line in everypixel is a center-of-gravity approach using the filtergram inten-sities I i and the central wavelengths λ i of the filter passbandsknown from measurements λ line = (cid:80) i (1 − I i ) λ i (cid:80) i (1 − I i ) . (1)Strong absorption in a filtergram therefore leads to a strongweighting of its central wavelength. Equation (1) is applicablefor an undisturbed single spectral line and filtergrams that arenormalized to the continuum value. For real data however, astraightforward determination of the line-shift is a ff ected by theinfluence of other spectral lines, di ff erent continuum levels, dif-ferent line widths and depths, a varying ratio of the blue to redpart of the He i triplet, multiple velocity components in a resolu-tion element, and instrumental e ff ects. The theoretical approachdescribed in Eq. (1) therefore has to be extended to account forthese (partially unknown) influences. This was addressed by cal-ibrating the line-shift maps generated from the filtergrams withmeasured line-shift maps from a spectrograph.In December 2007, parallel observations in He i
6. Bethge et al.: The Chromospheric Telescope d i s t an c e [ a r cs e c ] d i s t an c e [ a r cs e c ] d i s t an c e [ a r cs e c ] d i s t an c e [ a r cs e c ] a r csec d i s t an c e [ a r cs e c ] d i s t an c e [ a r cs e c ] Fig. 5.
Sample images from ChroTel on 31 August 2010; full disk and full resolution cutouts. The black squares indicate a size of 2 arcsecondsin the bottom row.
From left to right: Ca ii K (12:07:39 UTC), H α (10:37:16 UTC), and He i i image to enhance the contrast. performed solely in the He i channel with the highest possiblecadence of 30 s.Line-shift maps were compiled from the TIP-II scansby determining the position of the minimum intensity of asmoothed spectrum in the wavelength range between 10828.5and 10831.9 Å in every pixel, thus between the photospheric Si i line and the water vapor line indicated in Fig. 4. This was donewith a precision of one pixel in the spectra, which correspondsto a precision in velocity of ∼
300 m s − . The maps exhibit a largerange of velocities, i.e., blueshifts up to −
14 km s − and redshiftsup to 43 km s − .The regions scanned with TIP-II were cut out from theChroTel full-disk filtergrams and carefully aligned with theTIP-II maps. To simulate the scanning procedure and the tem-poral evolution of the observed structures during the scanning procedure, we first tried to take only stripes from the ChroTelfiltergrams closest in time to the actual slit step and to composean artificial filtergram set from these stripes (cf. Beck et al. 2007,Appendix B). The varying seeing conditions in the ChroTel im-ages however prohibited a reliable composition. We thereforedecided to take only the exposure with the best seeing condi-tions during the 15 minutes of the TIP-II scan (at UTC 11:15:52,11:25:15, and 12:07:08). This is justified by the steadiness ofmost structures seen in the scans, which was verified usingmovies compiled from the ChroTel images.To perform a first order correction of the possibly varying in-fluence of the prefilter transmission curve on the measured inten-sities at di ff erent wavelengths, the median filtergram intensities I i were normalized to the median intensity in the second filter-gram. The latter has the least contribution from any spectral line
7. Bethge et al.: The Chromospheric Telescope
C. Bethge et al.: The Chromospheric Telescope
Map 1 - 3 filtergrams
Map 1 - 5 filtergrams
Map 1 - 7 filtergrams
Map 2 - 3 filtergrams
Map 2 - 5 filtergrams
Map 2 - 7 filtergrams
Map 3 - 3 filtergrams
Map 3 - 5 filtergrams
Map 3 - 7 filtergrams
Fig.7. α i factors for the reconstruction of the three velocity maps with a di ff erent number of used filtergrams. The abscissa denotes the number ofthe filtergram according to Fig. 3. Each run was repeated ten times to see if the α i converge to a single set. plement to spectropolarimetric observations, as will be demon-strated in a future paper.The inner 5 filtergrams seem to be su ffi cient for creating theDopplergrams, as the spread of the α i and the correlation coe ffi -cients for 7 filtergrams indicate that the fits become too arbitrary.It might be valuable in the future to combine Dopplergrams cre-ated with 3 and with more filtergrams to optimize the results overthe whole velocity range.It is unclear to what extent the velocity calibration per-formed within the scope of this work can be extrapolated andused for other regions on the Sun than the ones considered here.However, the fact that the solar rotation can be seen in full-diskDopplergrams created with the obtained calibration factors us-ing data from another day builds confidence that these might beof a general nature and are not only valid for the specific datasetused for the calibration. Acknowledgements.
ChroTel was built and put into operation with the help ofmany people from the workshops of the participating institutes. Sincere thanksare given to A. Bernert, H.P. Doerr, R. Friedlein, R. Hammer, T. Hederer,T. Keller, A. Lecinski, C. Prahl, M. Sigwarth, W. Schmidt, A. Tritschler,M. Weissch¨adel, and O. Wiloth. The ChroTel project was funded in part by the DeutscheForschungsgemeinschaft (DFG) under grants Pe782-4 and Pe782-10.
References
Centeno, R., Trujillo Bueno, J., Uitenbroek, H., & Collados, M. 2008, ApJ, 677,742Charbonneau, P. 1995, ApJS, 101, 309Dere, K. P., Bartoe, J., Brueckner, G. E., & Recely, F. 1989, ApJ, 345, L95Dupree, A. K., Penn, M. J., & Jones, H. P. 1996, ApJ, 467, L121 + Elmore, D. F., Card, G. L., Chambellan, C. W., et al. 1998, Appl. Opt., 37, 4270Ermolli, I., Fofi, M., Bernacchia, C., et al. 1998, Sol. Phys., 177, 1Eto, S., Isobe, H., Narukage, N., et al. 2002, PASJ, 54, 481Gilbert, H. R., Holzer, T. E., Thompson, B. J., & Burkepile, J. T. 2004, ApJ, 607,540Grossmann-Doerth, U., Schuessler, M., & Steiner, O. 1998, A&A, 337, 928Habbal, S. R., Woo, R., Fineschi, S., et al. 1997, ApJ, 489, L103 + Halbgewachs, C., Bethge, C., Caligari, P., et al. 2008, in Presented at theSociety of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol.7019, Society ofPhoto-Optical Instrumentation Engineers (SPIE) ConferenceSeriesHarvey, J. W., Hill, F., Hubbard, R. P., et al. 1996, Science, 272, 1284Kentischer, T. J., Bethge, C., Elmore, D. F., et al. 2008, in Presented at theSociety of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol. Fig. 6. α i factors for the reconstruction of thethree velocity maps with a di ff erent number ofused filtergrams. The abscissa denotes the num-ber of the filtergram according to Fig. 4. Eachrun was repeated ten times to see whether the α i converge to a single set (see Sect. 5.2). and therefore comes closest to a continuum value. The normal-ization was done by computing the median value in a centeredsquare covering about 37% of the solar disk in each filtergram.Each filtergram was subsequently multiplied by the ratio of themedian intensity of the second to the actual filtergram˜ I i = I I i I i . (2)With the obtained intensities ˜ I i , the line-shift maps were thencomputed based on Eq. (1) according to ∆ λ recon = (cid:80) i ( I c − α i ˜ I i ) λ i (cid:80) i ( I c − α i ˜ I i ) − λ , (3)where λ is the rest wavelength of the He i line at 10830.3 Å, and I c represents the continuum intensity in every pixel, which wasestimated to be I c = I . . (4)The α i in Eq. (3) are arbitrary factors that assign an enhanced ordecreased weighting to a specific filtergram, hence to a specificwavelength. A variation in these factors thus leads to a di ff er-ent result for the calculated line-shift map. A specific set of α i can be considered optimal when the sum over all pixels of thesquared di ff erence of the real line-shifts ∆ λ real computed fromthe TIP-II scans and the reconstructed line-shifts ∆ λ recon fromthe filtergrams is smallest, i.e., when (cid:88) pixels ( ∆ λ real − ∆ λ recon ) (5)is minimal or (cid:88) pixels ( v real − v recon ) (6) when the line-shifts are expressed as Doppler velocities accord-ing to v = ( ∆ λ/λ ) · c .Finding this set of α i is an optimization problem that wasaddressed with the genetic algorithm PIKAIA (Charbonneau1995). Genetic algorithms are comparably slow, but are knownto find global minima more reliably because they employ a ran-dom mutation of the fit parameters, which allows them to escapefrom local minima. Calibration runs were conducted with all three TIP-II scans us-ing three, five, and seven filtergrams for the reconstruction. Ifless than seven filtergrams were used, those that are closest toline center were chosen, e.g., ff erent number of filtergrams were intendedto show whether the additional information from the outer filter-grams improves the fit results or might be obsolete.The values of the α i were allowed to vary between 0 and 2.By default, every pixel was treated the same. In additional runs,enhanced weightings were applied to pixels with high absorptiondepth and to pixels showing small velocities because the lattercontribute the largest fraction. Each run was repeated ten timesto see whether the results of the optimization algorithm (ideally)converge to a single set of α i . The α i from the ten runs and fromall three maps were then averaged to compile velocity maps fora quantitative analysis. α i Figure 6 shows the α i parameters for all three calibration mapsapplying a di ff erent number of filtergrams for ten runs each. Thedistribution of the parameters is consistent for the three mapswithin a fixed number of used filtergrams. Map 1 shows slightlyenhanced weightings of filtergram 5 and decreased weightingsof filtergram 3 for the use of five and seven filtergrams, which is
8. Bethge et al.: The Chromospheric Telescope C on t i nuu m T o t a l i n t en s i t y [ c oun t s ] y [ a r cs e c ] T o t a l i n t en s i t y [ c oun t s ] y [ a r cs e c ] T o t a l i n t en s i t y [ c oun t s ] y [ a r cs e c ] H e I ab s o r p t i on dep t h T I P - II D opp l e r v e l o c i t i e s -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] C h r o T e l D opp l e r v e l o c i t i e s -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] -10-50510 L O S v e l o c i t y [ k m / s ] y [ a r cs e c ] Fig. 7.
Rows from top to bottom:
Continuum maps of the three TIP-II scans used for the velocity calibration for ChroTel. He i i LOS velocities as determined from the TIP-II spectra. He i LOS velocities as reconstructed from the ChroTel filtergrams (only filter-grams α i , same weighting for all pixels. See Sect. 5.1.). because of the high redshifts occurring in this map. For three fil-tergrams, the code always applies almost the same weighting tofiltergram 3 and 5, so the velocities are apparently not reflectedin the ratio of α to α , but in the ratio of both α and α to α .As the number of filtergrams used increases, the spread inthe α i parameters also increases for the ten runs, indicating thatthe fits become more arbitrary as the number of fit parametersincreases. While the α i are mostly confined to within a range of0.5 for five filtergrams, we find a spread over almost the wholerange for seven parameters. This indicates that the application ofonly three filtergrams seems to be the best choice as long as thevelocities are not too large. The two top rows in Fig. 7 show the absorption depth in He i andthe intensity in the nearby continuum of the three TIP-II scansthat were used for the velocity calibration. In the third row, theLOS velocities are shown as determined from the TIP-II spectra.For each of these velocity maps, an example of a correspondingvelocity map reconstructed from the three innermost ChroTel fil-tergrams with one set of α i is shown for comparison in the low-ermost row.As can be expected from both the lower spatial and spec-tral resolution, the ChroTel velocity maps lack the fine struc-tures exhibited in the TIP-II maps. This seems to be the case
9. Bethge et al.: The Chromospheric Telescope
Fig. 8.
Left column:
True vs. reconstructed ve-locities for the use of three, five, and seven fil-tergrams. The light grey and dark grey linesdepict linear-regression fits to the velocityregimes − < v < − and v (cid:62) − ,respectively. Right column:
Histograms of thevelocity di ff erence for the true and recon-structed velocities. The FWHM is calculatedfrom a Gaussian fit to the histograms (seeSect. 5.3.1). primarily in regions of weak He i absorption, as indicated by thetwo blueshifted streaks in the second map at about x = (cid:48)(cid:48) . Incontrast to the fine structures on the right-hand side of the map(where the He i absorption is low), the two blueshifted streaksare clearly visible in the ChroTel velocity map, although theyare not larger in size. For very low absorption depths, the spectralresolution of the ChroTel filtergrams does not appear to providethe sensitivity to capture the location of the helium-line core ac-curately, even if it can still be determined in the high-resolutionTIP-II spectra.The overall velocity structure is recovered remarkably wellin regions of high absorption depth, although the reconstructedvelocities significantly underestimate the real values. This isconfirmed in Fig. 8, where the left column shows a scatter plot ofthe true versus the reconstructed velocities for the use of three,five, and seven filtergrams. While the true velocities reach valuesof 40 km s − and larger, the values of the reconstructed velocitiesnever exceed 10 km s − .The scatter plots should ideally display a linear relation overthe whole velocity range. We can indeed separate the velo-city distribution into two distinct ranges: − < v < − and v (cid:62) − . For these regimes, the Pearson’s correlation coe ffi - Table 3.
Correlation coe ffi cients for the Doppler shift calibration (seeSect. 5.3).calibration number of correlation coe ffi cientrun filtergrams − < v < − v (cid:62) − regular 3 0.44 0.475 0.39 0.497 0.30 0.49small velocities 3 0.44 0.475 0.39 0.497 0.30 0.50high abs. depth 3 0.44 0.475 0.38 0.497 0.29 0.49 cient was computed to see how closely the distribution follows alinear relation (cf. Table 3). In addition, histograms of the di ff er-ence between the true and the reconstructed velocities were com-puted and fitted with a Gaussian profile (right column in Fig. 8).
10. Bethge et al.: The Chromospheric Telescope
The FWHMs of the histograms in Fig. 8 suggest that the calibra-tion gets successively better with the number of filtergrams used.While this seems to be confirmed by the correlation coe ffi cientsfor redshifts larger than 5 km s − , the opposite is seen for ve-locities of − < v < − . The coe ffi cients drop significantlyfor seven filtergrams, i.e., the code apparently focusses on thelarge velocities. Using only the inner three filtergrams thereforeseems to be the most e ff ective choice for the reconstruction ofvelocities up to about 20 km s − . From this point on, the veloci-ties exceed the wavelength coverage of these filtergrams and theouter filtergrams have to be employed.An FWHM of 8.1 km s − results in σ = − for theuse of three filtergrams. This however does not imply thatsmaller velocities can not intrinsically be detected. We recall thata lot of fine structure is less likely to be resolved with the lowerspatial and spectal resolution of ChroTel, and that we insteadobserve a ‘time gradient’ in the TIP-II scan, while the ChroTelfiltergrams represent a snapshot. In a full-disk Dopplergram (seeFig. 9) created with the mean α i from several averaged ChroTelfiltergrams, the solar rotation is clearly visible. This means thatthe accuracy of the LOS velocities must be more precise than2 km s − for small velocities on large scales. Two additional calibration runs were performed in attempting toimprove the precision of the velocity reconstruction. In the firstrun, only velocities of − < v < − were considered forthe determination of the α i (“small velocities” in Table 3). In thesecond run, the fit results were weighted with the square of theabsorption depth in every pixel (“high abs. depth” in Table 3).For the first run, the FWHM of the histogram for three fil-tergrams does not change at all, and the correlation coe ffi cientsalso vary only very little compared to the regular run where allpixels were treated the same. For the second run focussing onhigh absorption depths, the FWHM drops insignificantly from8.1 to 8.0 km s − , and again the correlation coe ffi cients displayvery little variation. The di ff erent weightings therefore do notseem to have any significant impact on the calibration quality.
6. Discussion and conclusions
The spatial resolution determined in the ChroTel images showsthat they are at least comparable in this respect to data from sim-ilar instruments such as CHIP for He i , PSPT for Ca ii K, or theH α telescope of the Kanzelh¨ohe Observatory. The main advan-tage of ChroTel however is the almost simultaneous acquisitionof all three spectral lines. Combining this information permits amore accurate estimation of the temperatures and height profilesof structures.The data are taken with a cadence of three minutes and arealso publicly available for long-term in this cadence , which isnot the case for most other full-disk instruments observing inthese lines. Although a cadence up to 30 s would be possible,three minutes were chosen as a compromise between the quan-tity of data and also trying to catch scarce and short-lived eventsin the regular mode of observation. The adjustable observingmode nevertheless gives observers the opportunity to take dataat a much higher cadence if desired.The determination of LOS velocities from filtergrams in He i o ff ers an additional tool to study the dynamics of the chromo-sphere. It has been shown that a precision of at least 3-4 km s − is L O S v e l o c i t y [ k m / s ] Fig. 9. He i LOS velocity for the full disk on 6 July 2009, determinedfrom averaged ChroTel data (UTC 15:58-16:39) using three filtergramsand the mean α i parameters from all three calibration maps. possible with the method presented here, i.e., when co-temporalspectrographic maps are available for calibration. Once cali-brated, the ChroTel Dopplergrams are a valuable supplement tospectropolarimetric observations because they provide the chro-mospheric velocity field in the full FOV with a high cadence.This will be demonstrated in a future paper.The inner five filtergrams seem to be su ffi cient for creatingthe Dopplergrams, as the spread of the α i and the correlation co-e ffi cients for seven filtergrams indicate that the fits become arbi-trary. It might be valuable in the future to combine Dopplergramscreated with three and more filtergrams to optimize the resultsover the whole velocity range.It is unclear to which extent the velocity calibration per-formed within the scope of this work can be extrapolated andused for regions on the Sun other than the ones considered here.However, that the solar rotation can be detected in full-diskDopplergrams created with the obtained calibration factors us-ing data from another day provides confidence that these factorsmight be of a general nature and are not only valid for the spe-cific dataset used for the calibration. Acknowledgements.
ChroTel was built and put into operation with the help ofmany people from the workshops of the participating institutes. Sincere thanksare given to A. Bernert, G. Card, C. Chambellan, H.P. Doerr, R. Friedlein,R. Hammer, T. Hederer, T. Keller, M. Knobloch, A. Lecinski, R. Lull, C. Prahl,W. Schmidt, M. Sigwarth, T. Sonner, A. Tritschler, M. Weissch¨adel, andO. Wiloth.The ChroTel project was funded in part by the Deutsche Forschungs-gemeinschaft (DFG).
References
Beck, C., Bellot Rubio, L. R., Schlichenmaier, R., & S¨utterlin, P. 2007, A&A,472, 607Centeno, R., Trujillo Bueno, J., Uitenbroek, H., & Collados, M. 2008, ApJ, 677,742Charbonneau, P. 1995, ApJS, 101, 309Collados, M., Lagg, A., D´ıaz Garc´ıa, J. J., et al. 2007, in The Physics ofChromospheric Plasmas, ed. P. Heinzel, I. Dorotoviˇc, & R. J. Rutten, ASPConference Series, 368, 611Dere, K. P., Bartoe, J., Brueckner, G. E., & Recely, F. 1989, ApJ, 345, L95Dupree, A. K., Penn, M. J., & Jones, H. P. 1996, ApJ, 467, L121Elmore, D. F., Card, G. L., Chambellan, C. W., et al. 1998, Appl. Opt., 37, 4270Ermolli, I., Fofi, M., Bernacchia, C., et al. 1998, Sol. Phys., 177, 1
11. Bethge et al.: The Chromospheric Telescope