A Unique Resource for Solar Flare Diagnostic Studies: the SMM Bent Crystal Spectrometer
J. Sylwester, B. Sylwester, K. J. H. Phillips. A. Kepa, C. G. Rapley
aa r X i v : . [ a s t r o - ph . S R ] A p r Draft version April 8, 2020
Typeset using L A TEX default style in AASTeX63
A Unique Resource for Solar Flare Diagnostic Studies: the
SMM
Bent Crystal Spectrometer
J. Sylwester, B. Sylwester, K. J. H. Phillips, A. Kępa, and C. G. Rapley Space Research Centre, Polish Academy of Sciences (CBK PAN), Warsaw, Bartycka 18A, Poland Scientific Associate, Earth Sciences Dept., Natural History Museum, Cromwell Road, London SW7 5BD, UK Dept. of Earth Sciences, University College London, Gower Street, London WC1E 6BT, UK (Received ...; Revised ...; Accepted ...)
Submitted to ApJABSTRACTThe
Bent Crystal Spectrometer (BCS) on the NASA
Solar Maximum Mission spacecraft observedthe X-ray spectra of numerous solar flares during the periods 1980 February to November and 1984 –1989. The instrument, the first of its kind to use curved crystal technology, observed the resonancelines of He-like Ca (Ca xix ) and Fe (Fe xxv ) and neighboring satellite lines, allowing the study ofthe rapid evolution of flare plasma temperature, turbulence, mass motions etc. To date there has notbeen a solar X-ray spectrometer with comparable spectral and time resolution, while subsequent solarcycles have delivered far fewer and less intense flares. The BCS data archive thus offers an unparalleledresource for flare studies. A recent re-assessment of the BCS calibration and its operations is extendedhere by using data during a spacecraft scan in the course of a flare on 1980 November 6 that highlightssmall deformations in the crystal curvature of the important channel 1 (viewing lines of Ca xix andsatellites). The results explain long-standing anomalies in spectral line ratios which have been widelydiscussed in the past. We also provide an in-flight estimation of the BCS collimator field of view whichimproves the absolute intensity calibration of the BCS. The BCS channel 1 background is shown tobe entirely due to solar continuum radiation, confirming earlier analyses implying a time-variable flareabundance of Ca. We suggest that BCS high-resolution Ca xix and Fe xxv line spectra be used astemplates for the analysis of X-ray spectra of non-solar sources.
Keywords: atomic data – Sun: abundances — Sun: corona — Sun: flares — Sun: X-rays, gamma rays INTRODUCTIONThe NASA
Solar Maximum Mission ( SMM ) spacecraft, launched on 1980 February 14 at the peak of solar cycle 21,carried a battery of instruments to observe the active, particularly the flaring, Sun at ultraviolet, X-ray, and gamma-ray wavelengths. It remained operational until 1989, although there was a loss of fine pointing between 1980 Novemberand 1984 April due to the failure of an attitude control unit which was replaced by astronauts on the
Space ShuttleSMM Repair Mission (STS-41-C).The
Bent Crystal Spectrometer on SMM was a high-resolution spectrometer designed to observe the X-ray spectraof solar flares in the region of highly ionized Ca and Fe emission lines. The intensity ratios of the lines determine arange of physical parameters for the hot flare plasma including temperature, emission measure, and plasma motions.The instrument formed a key component of the X-ray Polychromator (Acton et al. 1980). It had eight channels eachconsisting of a bent germanium crystal wafer and a position-sensitive proportional counter, enabling spectra in narrowranges to be instantaneously formed by Bragg diffraction over time intervals short enough to follow the changing flareintensity, particularly during the impulsive stage. The BCS pioneered the concept of curved crystal technology for
Corresponding author: K. J. H. [email protected]
J. Sylwester et al. space instrumentation, the advantage of which was the lack of a need for moving mechanical parts as with scanning flatcrystal spectrometers as well as encoders to read the scanning angle. The BCS was fixed on the spacecraft baseplate,so that particular active regions could be selected through spacecraft pointing, which was done on a day-by-daypreplanned basis. A grid collimator confined the emission from a single active region to be viewed, avoiding spectralconfusion when two or more flares occurred simultaneously in different regions.An extensive review of the BCS instrumental parameters was made by Rapley et al. (2017) so that more accurateanalysis of the data can now be made. There is continued interest in the BCS data as the spectral resolution (resolvingpower λ/ ∆ λ ∼ for its long-wavelength channel viewing Ca xix lines but up to for three of its high-resolutionFe xxv channels) for such a long-operating instrument was unprecedented and continues to be unsurpassed for anysolar X-ray spectrometer. To our knowledge, only the SOX2 X-ray spectrometer on Hinotori (Tanaka 1986), whichviewed Fe xxv and Fe xxvi spectra during flares over a short period in 1981 – 1982, had better spectral resolution. TheBCS was used to observe the upflowing plasma and turbulence occurring at the flare impulsive stage, leading to flaremodels invoking chromospheric evaporation processes (Antonucci et al. 1982; Culhane et al. 1981). The replacementof the
SMM attitude control unit and other repairs during the 1984 Space Shuttle repair mission enabled the frontpanels of the various
SMM instruments to be inspected, and it was evident, from photographs taken by the astronauts,that the thermal filter of the BCS had disintegrated at some point since launch in 1980. Careful analysis of data fromvarious flares by Rapley et al. (2017) now indicates that this failure occurred on or around 1980 October 14, and isprobably associated with the large (
GOES class X3) flare that occurred on this date. Calibration corrections havebeen calculated and are applied to the spectra to account for this change.In this work, we extend the analysis of Rapley et al. (2017) by considering BCS data collected during a uniquedata set made over the daylight period of an orbit (approximately one hour) on 1980 November 6 when the
SMM spacecraft executed a series of maneuvers in which a flaring active region was repeatedly scanned in a square pattern,approximately six arcminutes on a side. The flare emission thus passed over the BCS collimator both aligned withand perpendicular to the BCS dispersion direction. In each case, the motions modulated the intensity of the spectraobserved according to the angular response function of the collimator. The motions aligned with the dispersion axisgave rise to spectra that appeared to shift along the BCS detectors as a result of the changing angle of incidence(Section 5). The analysis reported here is particularly for BCS channel 1, which viewed a group of lines due to He-likeCa (Ca xix ) and associated satellite lines, because these spectra have the highest count rates of all eight channelsand the Ca xix line emission persisted for much longer than the Fe xxv emission. From channel 1 spectra we foundevidence for irregularities in the crystal curvature that then explained a long-standing problem in the intensity ratioof the Ca xix x and y intercombination lines (see Table 1 and Figure 7 for identification of the spectral lines of theBCS channel 1 spectra). The angular extent of the BCS collimator was also estimated, confirming and improvingon pre-flare measurements. Fits to BCS channel 1 spectra using theoretical spectra, similar to those used in recentanalysis of CORONAS-F
DIOGENESS spectra (Phillips et al. 2018), depend on electron temperature, mainly throughthe ratio of the Ca xviii satellite line k to the Ca xix resonance line w , and it was found that this temperature issimilar to that estimated from the two channels of GOES , so offering a useful proxy for temperatures using Ca xix lineratios. This in turn helped us to derive a new absolute intensity calibration for channel 1 from the comparison of BCSemission with the fluxes of the two channels of
GOES . We also deduced that the channel 1 background pedestal can beattributed to solar continuum with negligible amounts of instrumental (e.g. crystal fluorescence) background, whichconfirms an assumption made in previous studies of the time-varying flare abundance of calcium (Sylwester et al. 1984,1998). THE BENT CRYSTAL SPECTROMETERAn outline of the principle of the BCS is shown in Figure 1. Incoming X-rays from a solar flare were incident throughthe BCS collimator and thermal filter on to one of eight curved crystal wafers mounted on substrates. The rays werethen diffracted according to Bragg’s law, λ = 2 d sin θ ( λ = diffracted ray wavelength, θ = grazing angle of incidence, d = crystal lattice spacing). Because of the crystal curvature, slightly different values of θ occur for different pointsalong each crystal’s length. Eight position-sensitive proportional counters received the diffracted radiation from eachcrystal, so over short data gathering intervals, typically a few seconds, complete spectra were formed as the detectedphotons were read off from the anode wire of each detector. Figure 1 illustrates the scheme for BCS channel 1, thewavelength range of which includes the Ca xix resonance line and associated satellites and for which the diffractingcrystal was a thin wafer of Ge 220 cylindrically bent crystal with d = 4 . Å and nominal radius of curvature equal
MM Bent Crystal Spectrometer Collimator Figure 1.
Ray path scheme for BCS channel 1 (Ca xix ) generated by Computer Aided Design (CAD). Solar X-rays (parallelgray bars, upper left) are incident through a thermal filter and the × arcmin (FWHM) square multi-grid collimator (frontand rear grids indicated) on to the Ge 220 cylindrically bent crystal wafer. The diffracted rays (pale gray bars), incident on thebent crystal at angles θ and θ , have different wavelengths ( λ , λ ) according to their position along the crystal. The dispersedradiation received by the position-sensitive detector forms a complete spectrum over the wavelength range 3.169 – 3.227 Å fora flare on the BCS optical axis defined by the direction of the collimator’s peak transmission. to 5.83 m. A source like a solar flare or hot active region that was off-axis by an angle ∆ θ along the crystal dispersiondirection, or solar E – W during nearly all spacecraft operations, would give rise to a spectrum shifted in wavelengthby an amount ∆ λ given by ∆ λ = 2 d cos θ ∆ θ. (1)The BCS collimator, operating on the standard Oda principle (Oda 1965), consisted of eight arrays of co-alignedsquare holes ( . × . mm) evenly located to form a ∼ × arcmin (FWHM) square pyramid transmission pattern(Section 6). Only the front and rear grids are shown in Figure 1. Based on its design and on pre-flight optical and X-raycalibrations, the estimated on-axis transmission is ± %. Offsets of the flare source from the instrument opticalaxis defined by the direction of the peak transmission of the collimator accordingly result in intensity modulations. J. Sylwester et al.
Figure 2.
Top left: White-light image of the Sun from Kislovodsk Observatory taken on 1980 November 7. Solar North is up,east to the left. Top middle: Sunspots of AR 2779 with fields of view of
SMM instruments: the red square box shows the BCScollimator field of view (red cross is the BCS optical axis), the blue square the FCS scan dimensions, and the green square theUVSP scan dimensions. Bottom left: FCS white-light sensor image of the sunspots of AR 2779. Bottom middle: Kislovodsksunspot image spatially degraded to match the FCS image. Right-hand panel: The Kislovodsk sunspot image with overlaidFCS flare images (blue: O viii contours; red: S xv ) and UVSP Fe xxi flare image (green contours). The position of the BCSoptical axis is at the intersection of the continuous lines. BCS data from the M3.5 flare on 1980 November 6 (SOL1980-11-06T22:27) are of particular interest since the
SMM spacecraft performed raster-like scans over active region AR 2779 during the flare enabling BCS line shifts to beprecisely related to the spatial shifts. We estimated the solar location of the flare and the host active region usingimaging data from two
SMM instruments: the
Flat Crystal Spectrometer (FCS) observing soft X-rays and white-lightemission using an in-built sensor; and the
Ultraviolet Spectrometer and Polarimeter (UVSP), which was an imagingspectrometer observing in the 1750 – 3600 Å range. White-light images from the Kislovodsk Observatory (part ofthe Pulkovo Observatory, St Petersburg, Russia), using the Debrecen catalogue (see Györi et al. (2017)) are given inFigure 2, taken at around 06:00 UT on November 7, some eight hours after the M3.5 flare peak. A full-Sun image and amore detailed image around the sunspots of AR 2779 are shown (top left panels and right panel). The FCS white-lightsensor images (spatial resolution ∼ arcseconds) were correlated with the Kislovodsk sunspot image by degrading thespatial resolution of the latter (lower panels). The flare maximum X-ray and ultraviolet emission are shown relative tothe Kislovodsk sunspot image in the right panel. The FCS showed significant emission during the flare rise from fiveof its seven X-ray channels, the lowest-temperature ( ∼ . MK) emission being from H-like oxygen (O viii , 18.97 Å,blue contour, right panel), and the highest-temperature emission from He-like sulfur (S xv , temperature ∼ MK,red contour). The UVSP obtained images in the region of 1300 Å; the wavelength range included the forbidden line ofC-like Fe (Fe xxi , 1354 Å, green contours: see Doschek et al. (1975)), with similar emitting temperature to the S xv X-ray line. The Fe xxi and S xv emission regions are, as expected from their similar temperature dependence, nearlyidentical in extent. SMM
SPACECRAFT MANEUVERSThroughout the
SMM spacecraft maneuvers during the 1980 November 6 flare, BCS spectra were obtained withdata gathering interval equal to 19 s. In a pre-planned sequence, the hour-long daylight portion of the
SMM orbitstarted with the spacecraft pointed at AR 2779 for a period of 13 minutes starting at 22:02 UT. The FCS performedtwenty-three rasters, × arcmin square, over AR 2779. From GOES light curves, the M3.5 flare began its rise overthis period, referred to here as Phase A . At 22:15 UT, the spacecraft started its series of raster maneuvers. The GOES light curves indicate that over this period (Phase B ) the flare reached its maximum at about 22:26 UT, with a furthersmaller peak at 22:32 UT. The spacecraft raster motion had angular excursions of ± arcseconds in the E – W MM Bent Crystal Spectrometer Collimator ± x direction according to the usual solar coordinate convention, with + x being the W direction), i.e. thespectrometer’s dispersion axis, and with each excursion taking 240 seconds. As explained in Section 6, the spacecraftalso moved northwards at the end of each E – W scan, forming a “boustrophedonic” (“as the ox ploughs”) pattern.Although official documentation on the nature of the spacecraft maneuvers is no longer available, it is known thatthey occurred through the activation of on-board reaction wheels in the SMM attitude control unit. (A failure in thisunit later in 1980 November was the reason that the
SMM pointed instruments were unable to perform until the unit’sreplacement during the 1984 Space Shuttle Repair Mission.) The reaction wheel controlling the E – W motion wouldhave spun up over a short time, moved uniformly for most of the excursion, then decelerated to a stop at the end ofthe E – W motion. The reaction wheel operating in the N – S direction would then have spun up over a short timeinterval, followed by a spinning up of the E – W reaction wheel in the reverse direction. Thus, another part of theraster pattern would have been formed, and so on till the raster pattern was complete, a total time of approximately22 minutes. Because of the short but non-zero amounts of time for the reaction wheels to spin up, the motion wouldnot have been exactly uniform. These non-uniformities were hardly apparent apart from a slight spectral smearing insome cases.After the spacecraft scanning maneuvers, there was a period (22:37 – 22:57 UT, Phase C ) when the spacecraftattitude was undetermined and no meaningful BCS spectra were obtained. During a two-minute-long final phasebefore the sunlit portion of the orbit ended, starting at 22:57 (Phase D ), the spacecraft was pointed to another flare-productive active region, AR 2776, near disk centre, and remained there until the end of the orbit. A flare was also inprogress in this region, and BCS spectra showed a displacement of only around 1 arcmin. BCS SPECTRA DURING THE
SMM
SPACECRAFT SCANNING MANEUVERS5.1.
Detection of BCS Crystal Non-uniformities
The BCS spectra were collected in bins that are linearly related to wavelength. For BCS channel 1, bin numbersran from 1 to 254; the range sensitive to solar X-rays was 33 – 220 while the end bins registered emission from thespectrometer’s calibration source. The dispersion, calculated from the bin numbers corresponding to the peaks ofprominent line features due to Ca xix lines w and z + j (Table 1), is 0.3040 mÅ/bin. Using this and the mid-channelwavelength of 3.198 Å (Rapley et al. 1977), the observable wavelength range is 3.169 – 3.227 Å for an on-axis source.Owing to a peculiarity of the data readout system, bin numbers in the raw data stream decreased with increasingwavelengths (see equation (4) of Rapley et al. (2017) giving the relation between the two). A more intuitive conventionis used here in which BCS bin numbers B n increase with wavelength, and are related to those of Rapley et al. (2017)( B RSP ) by B n = (253 − B RSP ) + 1 .Figure 3 shows the time history of BCS channel 1 observations during the November 6 flare. BCS channel 1 spectraare shown as color intensity (red temperature scale) plots in the top and fourth panels with wavelength increasingfrom approximately 3.169 Å to 3.227 Å; the spectra in the fourth panel are normalized to the emission in this range.
GOES and BCS channel 1 light curves are shown in the second panel, and temperature ( T G ) and emission measures( EM G ) derived from the intensity ratio of the two GOES channels in the third panel. The four phases of spacecraftpointing ( A , B , C , D ) are indicated in the top panel. The GOES light curves in Figure 3 (second panel) show anumber of peaks. Since
GOES has no spatial information, the sources cannot be unambiguously identified, but theBCS channel 1 total photon count rates have maxima that correspond to when the spacecraft was aligned with emissionfrom AR 2779, so it can be presumed that the repeated maxima up till 22:35 UT are from AR 2779. The increase inthe BCS count rate after 22:57 UT is due to the AR 2776 flare near disk center, which the spacecraft was pointed tojust before spacecraft night.The location of the BCS spectra are unaffected by spacecraft motion perpendicular to the dispersion axis, but forangular offsets in the plane of the dispersion axis, the spectra appear to “slide” back and forth along the detector asthe Bragg angles of incidence vary on the curved crystal. The bottom panel of Figure 3 illustrates the shift in binsof channel 1 spectra, in particular the oscillatory pattern made when the spacecraft performed its scanning motionduring Phase B . During the flare rise (Phase A ), some 41 spectra were obtained with the BCS optical axis pointed atthe hot flaring kernel; prominent spectral line features, identified in Table 1, are indicated. During the spacecraft scans(Phase B ), 66 spectra were collected with the line features showing a wavy pattern reflecting the way the spacecraftE – W or W – E scanning motion moved the spectral lines along the dispersion axis – apparent wavelengths increasedas the spacecraft boresight moved toward solar east ( − x ) and decreased toward solar west ( + x ). In Phase C , there J. Sylwester et al.
Table 1.
Line features in BCS channel 1.Ion and line identification Transition Wavelength(Å) a Ca xix w s S − s p P xviii n = 3 1 s p P − s p p ( P ) D ( d
13 + d ) 3.18180Ca xix x s S − s p P xix y s S − s p P xviii q s s S / − s s p ( P ) P / xviii r s s S / − s s p ( P ) P / xviii k s p P / − s p D / xviii j + Ca xix z Ca xviii ( j ) s p P / − s p D / xix ( z ) s S − s s S a From DIOGENESS: Phillips et al. (2018). are no recognizable features in channel 1 spectra. During Phase D , with the spacecraft now pointed at AR 2776 nearSun center, channel 1 spectra showed no significant displacements, the flare being close to the BCS optical axis.The shifts in BCS channel 1 spectra during Phase A and B (converted to bin units) were determined as follows:Each normalized spectrum in the bin range 33 – 220 was cross-compared with a normalized reference average spectrumto determine an optimum shift in BCS bins; this was done by interpolating the reference spectrum with a step sizeof 0.1 bin, moving each analyzed spectrum in steps of 0.1 bin to get the best overlap. Two criteria were applied:firstly, a “multiplicative” approach (black points in the bottom panel) in which the shift corresponding to a maximumin the product of the analyzed and reference spectra was obtained; secondly (red points), the shift correspondingto where the traditionally defined χ difference attained a minimum value. The red points coincide with the blackpoints to within 0.05 bin. Note that during Phase A the line features are not exactly stationary in bin space; thereare slight fluctuations with a variation of σ = 0 . bin, representing the upper limit to the accuracy of the bin shiftdetermination.In Figure 4, spectral shifts in BCS channel 1 are illustrated by spectra plotted against bin numbers that are averagesover seven periods. The first (top panel) corresponds to Phase A , and is the average of 41 spectra during the AR 2779flare rise (Phase A ), with a wavelength displacement of zero ( . ± . bin). The following five panels show spectraduring Phase B , and illustrate how the spectra are displaced to shorter or longer wavelengths according as SMM ispointed to respectively west or east of AR 2779 (offsets are indicated in each panel). The bottom panel shows averagedspectra during Phase D when SMM was pointed at AR 2776 near disk centre. Spectra taken in intervals near theextreme ends of the spacecraft scans have slightly smaller line widths than those between the ends attributable to thefact that BCS were accumulated in data gathering intervals (19 s) that are significant fractions of the time that
SMM took to do a full E – W excursion or the reverse.In Figure 5, displacement of eight recognizable BCS channel 1 spectral line features before (Phase A ), during(Phase B ), and after (Phase D ) the SMM scan maneuvers are plotted against spacecraft offset (arcseconds). The linepositions were determined by eye (automated routines were unreliable for weak line features like the d
13 + d satellitefeature). There are no strong indications of any Doppler shifts in Phase A indicative of the flare impulsive stage, buteven so we excluded Phase A observations from the determination of line feature positions (yellow points in Figure 5).Likewise, points in Phase D (red points) are excluded because they are from the AR 2776 flare. Thus, the points forPhase B (after the AR 2779 flare impulsive stage) are the most reliable. For five of the line features ( w , x , q , k , j + z ),the bin numbers of the points are distributed linearly with spacecraft offset. The bin numbers for two of the remainingfive ( d
13 + d , r ) line features are slightly non-linear with spacecraft offset, although these line features are relativelyweak or partly blended. The points for the fairly strong line feature y are also clearly non-linear. The non-linearity ismost likely to be instrumental in origin, as discussed below. MM Bent Crystal Spectrometer Collimator w x y q r k z + j A B C D r a t e s [ ] & GO ES T G [ M K ], E M G [ c m − ] w x y q r k z + j [ b i n ] No spectra
Figure 3.
Time history of
GOES , BCS Channel 1 (Ca xix ) spectra, and spectral line shifts during the 1980 November 6 flare(22:02 – 22:59 UT). Top panel: Channel 1 spectra plotted (vertical scale) on a red temperature intensity scale (yellow for highintensities, blue low). Second panel: BCS channel 1 photon count rates in bins 33 – 220 (3.169 – 3.227 Å for an on-axis flare)in units of s − with GOES . − Å (W m − , multiplied by × , blue) and − Å (W m − , multiplied by × ,red) light curves. Third panel: temperature ( T G in MK, red) and emission measure ( EM G in units of cm − , blue) fromthe emission ratio of the two GOES channels. Fourth panel: Channel 1 spectra normalized to the BCS total count rate in the3.169 – 3.227 Å range. Bottom panel: wavelength shifts (expressed as relative bin number: 1 bin equals 0.3040 mÅ) determinedby two different methods (red and black points: see text).
J. Sylwester et al.
Normalized spectra
Offset= 0.00 ± 0.14 [bin]
Phase A
Offset= −6.70 ± 0.35 [bin]
Phase B
Offset= −3.77 ± 0.70 [bin]
Phase B
Offset= 0.13 ± 0.54 [bin]
Phase B
Offset= 3.80 ± 0.73 [bin]
Phase B
Offset= 7.10 ± 0.30 [bin]
Phase B
50 100 150 200Spectral bin0.000.010.020.03 N o r m a li z ed i n t en s i t y pe r b i n Offset= 0.96 ± 0.05 [bin]
Phase D
Figure 4.
Normalized BCS spectra during the 1980 November 6 flare referred to BCS bin numbers. Top panel: average of 41spectra in Phase A (zero offset). Second to sixth panels: Phase B spectra averaged over short intervals at averaged spacecraftoffsets indicated in the legend (+ = offsets to spacecraft west; - = spacecraft east). Bottom panel: average of five spectra with SMM pointed at AR 2776. The vertical dotted lines show the undisplaced bin positions of the Ca xix lines w and z + j . Correction of BCS Spectra for Crystal Non-uniformities
Long experience with the detector design, including laboratory development and pre-flight calibrations (Rapley et al.2017), assure that anode wire resistivity non-uniformities are not the origin of the channel 1 wavelength non-uniformitiesfor Ca xix line y , and instead they are to be ascribed to slight deformities in the crystal bend radii. Because of thelatter, the bin widths do not exactly correspond to wavelength intervals, ∆ λ , and so the number of photon countsper bin does not quite correspond to photon counts per unit ∆ λ . The effect of deviations from the crystal’s perfect MM Bent Crystal Spectrometer Collimator Line centre positions vs. offset −200 −100 0 100 200Offset [arcsec]6080100120140160180200 b i n po s i t i on z+jkrqyxd13w Figure 5.
Measured bin positions of eight line features (labeled) in BCS channel 1 spectra, estimated by eye, plotted againstspacecraft offset from boresight position in arcsec. The lines are: Ca xix line w , Ca xviii d
13 + d (labelled d ), Ca xix x and y ; Ca xviii q , a , k ; and Ca xix z and Ca xviii j (blended). The black points are bin positions from spectra during thePhase B spacecraft scans; yellow points those from Phase A ; red points those from Phase D . curvature is to stretch or compress the spectrum with the same effective collecting area per detector bin, giving riseto intensity changes which therefore need adjusting.These intensity changes affect satellite-to-line ratios which are important diagnostics of solar flare plasmas, as wasdiscussed by Gabriel (1972) and more recently summarized by Porquet et al. (2010). For Fe xxv and Ca xix X-rayspectra, intensity ratios of dielectronic satellites such as k and j to the resonance line w (Table 1: see also Figure 7,discussed below) inversely depends on electron temperature T e . With BCS channel 1 (Ca xix ) spectra, the relevantratio is k/w as satellite j is blended with Ca xix line z . For satellites q and r , the upper levels of the transitions areprimarily excited by electron collisions of the Li-like ion, and so the q/w and r/w intensity ratios allow the ratio of0 J. Sylwester et al.
Table 2.
BCS Channel 1 spectra and
SMM spacecraft scan motions.Spectra no. Time range (UT) Phase BCS sp. offset Spacecraft (S/c) Scanch. 1 (bins) E – W offset x in arcsec: W = + , E = − A S/c repointing to AR 2779;spectrum rejected.2 - 42 22:01:57 - 22:15:08 A . ± . S/c not scanning; boresightat AR 277943 - 45 22:15:08 - 22:16:05 S/c repointing to begin scans46 - 111 22:16:05 - 22:37:00 B Spectra obtained every 19 swhile S/c scans46 22:16:05 - 22:16:24 B +7.33 S/c scans begin; S/c at min. x ( x = − )51 22:17:40 - 22:17:59 B -6.97 S/c at max. x ( x = +162 )57 22:19:34 - 22:19:53 B +7.18 S/c at min. x ( x = − )65 22:22:07 - 22:22:36 B -6.71 S/c at max. x ( x = +174 )69 22:23:23 - 22:23:42 B +7.02 S/c at min. x ( x = − )75 22:25:17 - 22:25:36 B -7.02 S/c at max. x ( x = +162 )82 22:27:30 - 22:27:49 B +7.40 S/c at min. x ( x = − )88 22:29:24 - 22:29:43 B -6.74 S/c at max. x ( x = +156 )93 22:30:59 - 22:31:18 B +7.20 S/c at min. x ( x = − )100 22:33:12 - 22:33:31 B -6.60 S/c at max. x ( x = +156 )105 22:34:47 - 22:35:06 B +7.50 S/c at min. x ( x = − )111 22:36:41 - 22:37:00 B -6.32 S/c at max. x ( x = +150 )112 - 173 22:37:19 - 22:56:57 C
174 - 175 22:56:38 - 22:57:16 S/C repointing176 - 180 22:57:16 - 22:58:51 D . ± .
181 22:58:51 - 22:59:04 S/C pointed at AR 2776 the Li-like ions to the He-like ions. This then enables checks on the presence or departure from ionization equilibriumin hot flare plasmas. For isothermal plasmas, significant differences in the derived temperatures from the ratios ( d
13 + d /w and k/w in Ca xix spectra indicate the presence of non-thermal electrons in the excitation processes(see Gabriel & Phillips (1979), relevant for Fe xxv ).To evaluate the intensity changes to the 112 BCS spectra collected during Phases A , B , and D of the November 6flare, we took each spectrum and found, using GOES data, the value of T G . As was established in an earlier work(Phillips et al. 2018), this is nearly equal to the temperature from the dielectronic satellite-to-resonance line ratio k/w ,or T k/w (see Table 1 for identifications). We then defined factors F for each wavelength giving the intensity ratio ofeach BCS spectrum to the calculated spectrum defined by T G , as described by Phillips et al. (2018). The matrix of F values for Phase B are a function of the offset of the SMM boresight from the flare, and can be used to investigatecrystal non-uniformities. It was found that there are non-uniformities at the wavelengths of the Ca xix line y and therelatively weak and partly blended satellites r and d . Non-uniformities at the position of line y are of considerableimportance as the intensity ratio Ca xix x/y has been problematical in previous analyses of BCS spectra, with line MM Bent Crystal Spectrometer Collimator −200 −100 0 100 200Source offset E−W [arcsec]0.81.01.21.41.6 F − f a c t o r v a l ue x−liney−linew−line Figure 6.
Values of correction factors F for the intensities of Ca xix lines w , x , and y as a function of SMM offset. Departuresfrom 1 indicate anomalies in the crystal curvature. A large anomaly (1.54) is indicated for zero offset for Ca xix line y . y over-intense compared with theory (e.g., Bely-Dubau et al. (1982); Phillips et al. (2004)). This might ordinarilybe attributed to additional atomic processes by which the upper level of line y is populated. Even with improvedcollisional rates the discrepancy remained, and Phillips et al. (2004) concluded that it must be an instrumental effect,which is supported by the fact that the intensity anomaly has not been noted for other high-resolution solar flarespectra (Phillips et al. 2018) and laboratory spectra from the Alcator tokamak (Rice et al. 2014). The variation of theintensity ratio F for the wavelengths of the Ca xix lines w , x , and y as a function of SMM offset is given in Figure 6.While there is little variation for line w , there is a considerable variation for lines x (large offsets) and y (zero offset).The relatively large value (1.54) of F for zero offset for line y is of particular interest as it indicates an anomaly in thecrystal radius at this location; BCS spectra at this wavelength should be divided by this value to obtain true intensity.Since the vast majority of flares seen by the BCS in its operational history were for zero offset, the intensity of the y line needs to be corrected for all such spectra.The correction factors F including those for Ca xix line y were applied to averaged spectra over Phase A (BCS opticalaxis directed at the AR 2779 M3.5 flare), Phase B (during the spacecraft scans), and Phase D (flare in AR 2776).The corrected spectra are shown in Figure 7. Error bars represent statistical uncertainties in the observed spectra.Each spectrum is compared with spectra (red lines) calculated as in Phillips et al. (2018). The calculated spectra area function of T G which as mentioned is nearly equal to T k/w . The theoretical spectra in Figure 7 take into accountthe contribution of ionized Ar lines, in particular the Ar xvii s − s p line at the wavelength of the diagnosticallyimportant Ca xviii q line, as noted by Doschek & Feldman (1981); the blue line in each spectrum shows the totalcontribution made by the Ar xvii lines plus the continuum. The abundance ratio of Ca to Ar is assumed to be 3.0,based on the average Ca abundance from Sylwester et al. (1998) and the Ar abundance from Sylwester et al. (2010).As can be seen from Figure 7, the relative intensities of the Ca xix x and y lines in the theoretical spectra are nowvery close to the observed, and therefore this well-known intensity anomaly is now explained by the non-uniformity ofthe BCS channel 1 dispersion.The theoretical spectra include free–free and free–bound continua which were taken from routines in the chianti database and code (Dere et al. 1997; Del Zanna et al. 2015). As can be seen from the three spectra in Figure 7, not2 J. Sylwester et al.
Aver. normalized Phase A spectrum & fit (red) N o r m a li z ed i rr ad i an c e w d13 x y q r k z+j T = 10.96 [MK]A Ca = 6.93A Ar = 6.45 χ = 1.29 Aver. normalized Phase B spectrum & fit (red) N o r m a li z ed i rr ad i an c e T = 13.18 [MK]A Ca = 6.93A Ar = 6.45 χ = 4.63 w d13 x y q r k z+j Aver. normalized Phase D spectrum & fit (red) N o r m a li z ed i rr ad i an c e T = 9.55 [MK]A Ca = 6.93A Ar = 6.45 χ = 0.32 w d13 x y q r k z+j Figure 7.
Normalized BCS channel 1 spectra (left to right) averaged over all spectra during Phase A , Phase B , and Phase D (shown as points with error bars representing photon count statistical uncertainties). The red curve is the theoretical spectrum(including continuum) as calculated by Phillips et al. (2018), and the blue curve the Ar xvii lines plus the continuum. Theassumed Ca/Ar abundance is 3.0. Values of χ are indicated in each plot (the large value for the middle panel is due to thevariation in temperature over Phase B). only the line emission but also the continuum in BCS channel 1 matches the theoretical continua seen in line-freeregions to within only a few percent. This was assumed in analyses (Sylwester et al. 1984, 1998) of the Ca lines inBCS channel 1 to obtain the flare abundance of Ca which was found to vary from flare to flare. The present work thusconfirms the earlier work. BCS COLLIMATOR ANGULAR RESPONSEIn previous work (Rapley et al. 2017), the BCS collimator transmission pattern was described as being of triangularshape in E – W ( x ) and N – S ( y ) directions with a nominal value of the FWHM equal to 360 arcsec and a peaktransmission of 0.33 for a source on the BCS optical axis. The three-dimensional pattern was approximated by apyramid shape in Figure 12 of Rapley et al. (2017), which is correct for the collimator transmission near the opticalaxis position but away from it ( & arcmin) the pattern of the collimator transmission has a trapezoidal, not triangular,shape, and so requires a small correction. For distances in the E – W ( x ) and N – S ( y ) directions of the source fromthe optical axis, the transmission C T is given by C T = 0 . (cid:16) − x FWHM (cid:17) (cid:16) − y FWHM (cid:17) . (2)We used BCS channel 1 data during the 1980 November 6 flare to determine the in-flight transmission of the BCScollimator and obtain an empirical value for the collimator transmission FWHM. Rather than assume that T G = T k/w as earlier, we calculated the emission in the entire wavelength BCS channel 1 range (lines and continuum) and relatedit to the emission from an isothermal source in the two GOES channels ( . − Å, − Å, called here f G4 , f G8 )separately. The emission is equivalent to the contribution or G ( T e ) curves that are generally given for individualspectral lines, and is defined here for a volume emission measure of cm − . The GOES and BCS channel 1 curvesare plotted in Figure 8 (upper panel). We empirically determined a function f , given by log f = log( f G4 + 0 . × f G8 ) − . , (3)which over the observed temperature range of the November 6 flare ( . − MK) is more nearly equal to the emissionin BCS channel 1 than either f G4 or f G8 . In the lower panel of Figure 8, the ratio of the BCS channel 1 emission to f G4 , f G8 , and the f functions are plotted against temperature; there is only a 13.3% change in the ratio BCS channel 1emission to f over the . − MK range.During the time period of the spacecraft scan (22:16:05 – 22:37:00), X-ray emission from the M3.5 flare variedconsiderably (Figure 3), so that emission at each time step of the scan is a convolution of the flare time variationsand the movement of the spacecraft across the flare. However, having established that the BCS channel 1 emission isvery similar to the
GOES emission function f , it is possible to reduce or nearly eliminate the flare time variations byplotting the ratio ( R ) of BCS channel 1 emission to f ; a normalized version of R is plotted in Figure 9 (top panel), MM Bent Crystal Spectrometer Collimator
10 11 12 13 14 15 16Temperature [MK]1.01.52.02.53.03.54.04.5
Log f l u x BCS
Log f l u x r a t i o GOES 1−8 ÅGOES 0.5−4 Åchange: 13.3 % log(f)=log(f G4 + 0.07 x f G8 ) − 0.2 Figure 8.
Upper panel: Temperature dependence of the logarithm of the flux from an isothermal source in the two
GOES channels ( . − Å, − Å, W m − multiplied by and respectively) and the total emission in BCS channel 1 shown aslog (flux) with flux calculated for an emission measure cm − . Lower panel: The ratio of BCS channel 1 flux to f G (bluepoints and curve), to f G (red), and to the function f where log f = log( f G4 + 0 . × f G8 ) − . (green). and shows modulations as the BCS optical axis repeatedly passes either directly north or directly south of the flareemission. There are eleven maxima (numbered in the figure) corresponding to times when the BCS optical axis wasaligned with the flare. The progressive increase then decrease in the maxima is due to the BCS optical axis (and sospacecraft axis) scanning in an E – W or W – E direction. The direction of the spacecraft motion across the Sun insteps at the end of each E – W or W – E scan is determined to be towards the north from the fact that the BCSoptical axis is to the south of the X-ray source (Figure 2) before the scans began. In addition, the spacecraft scans (1to 6) took a longer time to reach the X-ray source than to travel from it (scans 7 to 11).4 J. Sylwester et al.
60 80 100 120Spectrum No.0.00.20.40.60.81.0 N o r m a li s ed r a t i o / GO ES S ou r c e on a x i s N − S phase B scans −300 −200 −100 0 100 200 300Boresight position of E−W scan [arcsec]0.00.20.40.60.81.01.2 A tt enua t i on f a c t o r A tt enua t i on f a c t o r FWHM
E−W = 352.9 ± 7.9 [arcsec]
Figure 9.
Top panel: Ratio of emission in BCS channel 1 to the function f plotted against the spectrum number (Table 2).The ratios are normalized to one for a hypothetical source along the BCS optical axis. The diagonal lines are drawn throughthe maxima of corresponding E – W scans performed at fixed positions along the N – S direction. Center panel: Points fromeach of the E – W scans assembled using scaling factors corresponding to respective maxima (1 – 11) in the top panel. Bottompanel: Points from left side of dotted line in center panel overplotted on the right side points. The straight line is the best-fitline defining the collimator FWHM in the E – W direction, determined to be . ± . arcsec. Diagonal lines through the maxima in Figure 9 (top panel) meet at the time when the BCS optical axis was nearlyexactly aligned with the M3.5 flare. For each of the eleven E – W scans, the ratio of the BCS channel 1 emission to f (normalized to the ratio for a hypothetical source along the BCS optical axis) is plotted in Figure 9 (center panel).The points in this plot trace out the BCS collimator transmission along the dispersion direction. The central regionis flattened which is probably due to the ∼ arcsec E – W extent of the flare source; this is indicated by the UVSPimage in Figure 2 (green contours) which show that the flare extends ∼ arcsec in a N – S direction. In the bottompanel, both “arms” from the central panel are overplotted by mirroring in the dotted vertical line in the center panel.The width (FWHM) of the collimator response can be found from the slope of the diagonal lines, and has the value . ± . arcsec, only 2% different from the pre-launch value of 360 arcsec (Rapley et al. 2017).The SMM scanning pattern deduced from the BCS spectral shifts in the E – W direction, as well as the diagonal linesin Figure 9, is shown in Figure 10. It is superimposed on the BCS collimator transmission response using Equation 2.Spectra were taken over the intervals indicated by the numbers (see Table 2).
MM Bent Crystal Spectrometer Collimator
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99100101 102 103 104 105106107108109110111
Figure 10.
SMM scanning pattern (green line) superimposed on the BCS collimator transmission response (colored background)with spectrum numbers shown (Table 2).
In summary, the ratio of the BCS channel 1 total emission to the function f , which combines the emission in the two GOES channels, allows the in-flight determination of the BCS collimator profile, and that the profile width (FWHM)is within 2% of the pre-launch value. SUMMARY AND BCS DATA AVAILABILITYArchived data from the
SMM
Bent Crystal Spectrometer are a unique resource for studying the X-ray spectra ofthe diagnostically important He-like Ca (Ca xix ) and Fe (Fe xxv ) line groups emitted by hot solar flare plasmas inthat the spectral resolution was and remains (apart from the short-lived
Hinotori
SOX2 spectrometer) the highest ofany instrument, solar or non-solar. The BCS operated over the periods 1980 and 1984–1989, which included periods ofvery high solar activity (higher than any observed since) and through the solar minimum of 1986. In this work, usinga data set during an M3.5 flare when the SMM spacecraft performed E – W scans over the active region, we measuredthe in-flight angular response of the BCS grid collimator. We used the fact that, as the BCS scanned over the flare,the flare intensity variations can be estimated and compensated for by taking the ratio of the total emission in BCSchannel 1 to a function f (Equation 3) that combines the emission in the two channels of GOES rather than the BCSemission alone. The absolute intensity calibration, given by Rapley et al. (2017) (Table 3), can now be refined foroff-axis flares using Equation 2 with the collimator response width (FWHM) determined to be 352.9 arcsec.Two other conclusions from BCS channel 1 Ca xix spectra have been drawn. The first is the finding that an apparentanomaly in the Ca xix x and y line intensity ratios in BCS spectra is due to a non-uniformity in the Ge 220 crystalcurvature, deduced from measuring the bin positions of prominent line features (Figure 5). The second is that thebackground emission in BCS channel 1 is almost entirely due to solar continuum observed in the portion of the Sundefined by the BCS collimator, so vindicating earlier work (Sylwester et al. 1984, 1998) of flare-to-flare changes in theabundance of Ca. This indicates a more complex picture of observed variations in the abundances of elements withlow first ionization potential than a simple multiple as has been previously suggested.A few non-solar X-ray spectrometers have operated in the region covered by the SMM
BCS including the gratingspectrometers on the
Chandra X-ray Observatory (Canizares et al. 2005) and the microcalorimeter Soft X-ray Spec-trometer (SXS) on
Hitomi (Hitomi Collaboration 2018). While these and other spectrometers have the advantage ofa large spectral range, the spectral resolution is modest (resolving powers up to 1000) compared with the BCS for allits channels (resolving powers 4000 to 15000). The time resolution is, owing to the faintness of the targets, generally6
J. Sylwester et al. measured in tens of kiloseconds, so such instruments are unable to follow the temperature evolution of flares on activestars. A comparison of the HETG (High Energy Transmission Grating Spectrometer) spectra of two coronally activebinary stars with solar flare spectra from the solar X-ray RESIK (Rentgenovsky Spekrometr s Izognutymi Kristalami)instrument on the
CORONAS-F spacecraft (Huenemoerder et al. 2013) shows the presence of He-like and H-like ionlines in the stellar spectra that are similar to those in solar flare spectra. The HETG spectra of both stars show theresonance lines of Fe xxv and Ca xix but only a hint of the satellite line structure. The deep exposures of the Perseusgalaxy cluster reported by Hitomi Collaboration (2018) show several lines to the long-wavelength side of the Ca xix and Fe xxv resonance lines, some of which can probably be identified with dielectronic lines. Detailed analyses of theseand similar line groups could possibly take advantage of solar flare BCS spectra as “templates” since the excitation ofthe lines is unlikely to be very different for particularly stars with X-ray-emitting coronae.BCS data for the solar flares in the NASA
SMM archive cover a range extending from small flares ( GOES ratingB) to the very large flares seen in 1984 April, at the onset of Cycle 22. Software is already available for preliminaryanalysis of the spectra, written in the Interactive Data Language (IDL), with documentation. However, the correctionsdiscussed here for the small deformations in the crystal curvature for BCS channel 1 and those needed for flares thatwere offset from the BCS optical axis have yet to be applied, but programming is underway. For BCS channelsother than channel 1, viewing lines of highly ionized iron (wavelengths from 1.77 – 1.95 Å), corrections for crystalfluorescence are needed which for some wavelength ranges are large and so estimates of the continuum may not bereliably determined. Work is in progress for these channels and will be the subject of future publications.ACKNOWLEDGMENTSWe acknowledge financial support from the Polish National Science Centre (grant number UMO-2017/25/B/ST9/01821). We thank Jarosław Bąkała for his considerable help with the drawing of figures, and CraigTheobold (University College London Mullard Space Science Laboratory) for assistance in recovering original BCSdrawings.
Facilities:
Solar Maximum Mission (BCS)
Software:
SolarSoft Interactive Data Language (Freeland & Handy 1998), chianti (Del Zanna et al. 2015) Description of the
SMM spacecraft and links to the BCS data archive are available from https://umbra.nascom.nasa.gov/smm/
MM Bent Crystal Spectrometer Collimator17REFERENCES