The SPEAR Instrument and On-Orbit Performance

0.2 MB·English

by  Jerry Edelstein

Draftversion February5,2008
PreprinttypesetusingLATEXstyleemulateapjv.6/22/04
THE SPEAR INSTRUMENT AND ON-ORBIT PERFORMANCE
Jerry Edelstein1, Eric Korpela1, Joe Adolfo1, Mark Bowen1, Michael Feuerstein1, Jeffrey Hull1,
Sharon Jelinsky1, Kaori Nishikida1, Ken McKee1, Peter Berg1, Ray Chung1, Jorg Fischer1,
Kyoung-Wook Min2, Seung-Han Oh2, Jin-Guen Rhee2, Kwangsun Ryu2, Jong-Ho Shinn2,
Wonyong Han3, Ho Jin3, Dae-Hee Lee3, Uk-Won Nam3, Jang-Hyun Park3, Kwang-Il Seon3, In-Soo Yuk3
1SpaceSciences Laboratory,UniversityofCalifornia,Berkeley,CA94720
2KoreaAdvanced InstituteofScienceandTechnology, Dajeon,Korea305-701and
3KoreaAstronomyObservatory,Dajeon,Korea305-348
Draft versionFebruary 5, 2008
ABSTRACT
6
′′
0 The SPEAR (or “FIMS ) instrumentation has been used to conduct the first large-scale spectral
0 mappingofdiffusecosmicfarultraviolet(FUV,900-1750˚A)emission,includingimportantdiagnostics
2 of interstellar hot (104 K – 106 K) and photoionized plasmas, H , and dust scattered starlight. The
2
instrumentation’s performance has allowed for the unprecedented detection of astrophysical diffuse
n
a farUVemissionlines. Aspectralresolutionofλ/δλ∼550andanimagingresolutionof5’isachieved
J on-orbit in the Short (900 – 1175 ˚A) and Long (1335 – 1750 ˚A) bandpass channels within their
◦ ◦
6 respective 7.4 x 4.3’ and 4.0 x 4.6’ fields of view. We describe the SPEAR imaging spectrographs,
2 their performance, and the nature and handling of their data.
Subject headings: ultraviolet: ISM, instrumentation: spectrographs
1
v
8 1. INTRODUCTION parabolic-cylinder collecting mirror that focuses plane-
8 parallellightto aslit. Cylindricalradiationfromthe slit
The Spectroscopy of Plasma Evolution from
5 thenstrikesaconstant-ruledgratingwithanellipticalfig-
Astrophysical Radiation (hereafter SPEAR and
1
′′ ure that corrects on-axis aberrations to the third order.
0 a.k.a“FIMS ) payload is surveying the sky for
Diffractedlightisimagedasaspectrumonaplanarposi-
6 cosmic FUV emission. The SPEAR mission, launched
tion sensitive photon counting detector. Radiation from
0 27 Sep. 2003, its science objectives, and the mission
off-plane angles is imaged along the detector perpendic-
/ profile and performance are described in Edelstein et al.
h (2005). SPEAR’s bandpass (λλ 900 – 1750 ˚A), spectral ular to the dispersion plane, analogous to a slit-imaging
p resolution (λ/δλ ∼550), large field of view (7.4o x 4.5’), spectrograph. Thecylindrical-sourceschemedoublesthe
o- and imaging resolution(∼5’) facilitate the measurement usable imaging angle in comparisonwith classicalpoint-
source spectrographs. The result is a large solid angle
r of energetic interstellar gas while rejecting air-glow
t – collecting area product, a determining factor for dif-
s emission and stellar contamination that have plagued
a earlier measurements attempts. This paper describes fuse source sensitivity. The design is an extension of the
: EURD instrument (Bowyer et al. 1997)
v the SPEAR instrument, its on-orbit performance, and
Diffuse FUV spectrograph performance is a function
i the basic processing of instrument data. Measurements
X of optical quality, dispersion, efficiency, and scattering
of faint diffuse FUV emission lines are a difficult
due to surface or grating imperfections. The collecting
r undertaking wherein instrumental and systematic er-
a mirrors were measured to have a focal line width of <
rors can dominate the results. Therefore, we discuss
100µmFWHMatthe focaldistanceof15cmandasur-
instrumental facets in detail in order to allow users to
faceroughnessof10˚ARMS. The gratingswerepolished
understand the nature and quality of SPEAR data.
to < 35 ˚A RMS surface roughness and holographically
Further technical description regarding various aspects
ruled with a blazed profile formed by ion-beam abla-
of the SPEAR instrument are found in the literature
tion of chemically etched grooves, providing 65% peak
(Edelstein et al. 2003; Korpela et al. 2003; Nam et al.
groove diffraction efficiency. The Long and Short chan-
2003, 2002; Ryu et al. 2003).
nels(hereafterreferredtoas“L/S”)usethesamegrating
2. THESPECTROGRAPHICMETHOD filinguermewmh−i1leatnhdetLhoengShcohratncnhealnwnoerlkwsoinrkfisrsintosredceorndato3r0d0e0r
TheSPEARinstrumentconsistsofdualimagingspec- at 2250 line mm−1. The L/S optics use coatings, MgF
2
trographs optimized for the detection of faint diffuse on Al and B C on a thin Cr base (Keski-Kuha et al.
4
FUV radiation. The two spectral channels are desig- 1995), were chosen to optimize bandpass efficiency and
nated as the “Short” (λλ 900 – 1175 ˚A) and “Long” resist degradationby atomic oxygen.
(λλ 1335 – 1750 ˚A) bandpasses. The bands were cho- Carewastakentoreducebackgroundnoisefromstrong
sen to include astrophysically important emission lines airglowlines,stars,detectorandionbackground,andop-
fromabundant ionic species while avoidinginstrumental tics fluorescence. The Long channel includes a flexure-
contamination by the intense H I λ 1216 ˚A and O I λ mounted, zero-power CaF cylindrical meniscus filter
2
1304 ˚A geocoronal radiation. Each channel uses an op- that excludes geocoronal Lyman α before the 150 µm
tical configuration (see Fig. 1) consisting of a collecting slit. TheLongbandresponseislimitedatshortλbythe
mirror, a slit, a diffraction grating, and a detector. The CaF filter and at long λ by the detector photocathode
2
unique two-element f/2.2 optical system has an off-axis
2 EDELSTEIN, et al.
(see below) such that high-order diffraction is rejected. thresholds, and shutter timing. On-board software con-
The Short band is un-filtered and its response is lim- trols the instrument operation, data packetization and
ited by falling optical coating and diffraction efficiency flow, and spacecraft-time synchronization. Commands
at short λ, and by the detector photocathode at long λ, can synchronize the instrument and spacecraft clocks to
although high-orderresponse to the bright He I λ 537 is 250 ms. Detector events (photons, stimulation events,
found in flight data. The Short channel uses MgF win- noise events) are queued into packets interleaved with
2
dowsthattransmitλ>1150˚Ato detectorareasadjacent a 10 Hz time reference and other marks used during
to the science field so that instrumentally scattered Ly- datareductionforattitudesynchronization. Sciencedata
manαairglowcanbemonitored. Consequently,the L/S are transmitted to a mass memory system at 200 kbps
science fields of view are 7.4o x 4.3’ and 4.0o x 4.6’. We for downlink. The data system throughput impacts the
estimate the radiation induced filter fluorescence from event rate resulting in a net dead-time of ∼133 µs per
the filters’ fluorescence rate due to a predicted orbital event, corresponding to a throughput loss of 5% for a
radiation environment, the detector’s sensitivity to the 1kHz event rate. Dead-time was determined by on-orbit
fluorescencespectrum, and the detector’sview ofthe fil- observationsofidenticalbrightskyregionsusingthe10%
ters. The shutter can be selected to admit 0%, 1%, 10% and 100% shutter positions.
or 100 % of the available light for safe and photometric
3. PHOTONEVENTPROCESSINGANDCALIBRATION
observations of bright sources.
Individualphotondataeventsaresubjecttoautomatic
2.1. Sensor and Systems processingin-orbitandtoground-basedpipelineprocess-
Each channel’s spectrum is focused upon a separate ing. Photon events are selected for a valid pulse height
position-sensitive photon-counting microchannel plate based upon the nominal distribution (65% FWHM).
(MCP)Z-stackthatistopcoatedwithanL/Soptimized Low-amplitude noise events are discarded by the on-
photocathodeofCsIandKBr. Thestacks’23mmsquare boardelectronics. Typically,∼7%oftelemeteredevents,
active fields share a single event position encoding sys- presumably caused by ions and cosmic rays, are dis-
tembyusingauniquecrossdelay-lineanode(Rhee et al. carded due to excessive pulse amplitude. Photon events
2002)withabifurcatedlinetosensethespectralaxisand are marked with the concurrent total count rate so that
a continuous line to sense the imaging axis. The anode dead-timecorrectionscanbecalculated. Photondetector
axesarerotated15◦ withrespectto thedispersionplane coordinatesarepipelinecorrectedfordetectorelectronics
tomitigatetheappearanceoffalsespectralfeaturesfrom thermaldriftanddistortionevery5susinga2-dgradient
anodeorelectronicsdifferentialnon-linearity. Eventaxes distortionreferencedtothestimulationmarks’centroids.
positionsareindependentlydeterminedbypreciselytim- Thedriftcorrection,almostentirelyinthe dispersiondi-
ingthearrivaltimesofamplifiedanodepulses. Theposi- rection,decaysfrom∼0.6˚Atoinsignificancewithin120s
tionconversionsystemhasafixeddead-timeof86µsper ofanobservation’sstart. Thestimcentroidpositionsare
event. Astimulationunitinjectsanartificialeventsignal stableto∼0.15˚A.Photoneventsoccurringoutsideofan
for each field corner at 10 Hz so that thermal drifts in active boundary, defined using deep detector exposures,
the position encoding system can be calibrated in flight. are rejected including a 1 – 10 s−1 event “warm-spot”
The amplifiers also produce a signal proportional to ev- near the Long channel’s detector edge. Other photon
eryevent’schargeamplitude toallowforthe rejectionof data integrity checks reject < 0.5% of the data.
low amplitude noise or high amplitude ion and cosmic Photon wavelength, λ, and field angle, φ, are derived
ray events. A count-rate monitor turns off the detector fromdetector coordinates using simple polynomialplate
in case of excessive count-rates. Observation restart is scalesinthespectralandangulardimensions. Theimag-
automatically attempted within 2 to 30 s so that entire ing angular scale, 0.31◦ mm−1, was measured pre-flight
surveysweepswillnotbe losttomomentarybrightstars using a collimated FUV spectral illumination at precise
transits. Timingmarksareinsertedintothedatastream field angles. Imaging L/S angular resolution, averaged
to accurately account for such interruptions. overthe bandpass,was measuredfrom bright-starcross-
The spectrographs are contained in an enclosure in- ings and are 6.5’ and 4.5’ HEW. The L/S spectral dis-
cluding the gratings, order baffling, the detector, a persion scales, 17.9 ˚A mm−1 and 12.0 ˚A mm−1, were
shutter-slit unit, a mirror unit with field baffling, and a determined by comparing the Gaussian-fit centroids of
deployable dust cover. Each channel is optically baffled measuredauroralemissionlinestocentroidsderivedfrom
fromtheother. Thermaliondetectornoiseissuppressed predictedair-glowemissions(Strickland et al.1999)that
by positively biased wire grids and baffles on all enclo- have been smoothed by a 2 ˚A HEW Gaussian.
sure openings and the detector face, and by high-energy A λ distortion correction was constructed using com-
magnets at the slit. The 5x8 cm optics are bonded to positeemissionline spectrafrom20orbitsofauroralob-
thermally-matched metal flexures using a low-modulus servation. The L/S central λ error correction for each
adhesive,and then attached to three-point mounts. The of 15 and 9 lines was determined at 45’ intervals. The
gratingswere alignedwith divergentFUV radiationillu- correctionappliesa2-dgradienttoλandφwitharesid-
minating the slit and mirror-slit alignment used visible ual angle-integrated line λ centroid error of 0.5 ˚A and
collimatedlight. Full-systemgroundcalibrationwithcol- 0.75˚A.Theresultingspectrashowsremaininghigheror-
limated FUV light established the field width and angu- derdistortions. Consequently,a secondλcorrectionwas
lar scale. The complete 22 kg instrument is 45×45×15.5 constructedusingcompositelinespectrafrom155orbits
cm in size. of auroral observations. The L/S λ error-correction for
The instrument uses ground command to set payload 22 and 15 lines was determined in 5’ angular steps and
operatingmodes andto tune criticalengineeringparam- thenboxcarsmoothedat15’angularwidth. Corrections
eters such as detector voltage levels, detector electronics for λ’s between the measured lines were estimated by
SPEAR Instrumentation 3
a low-tension spline fit. The second correction improves 4.5’ field of view crossed a star, and using regular sky-
spectralresolutionandcentroidaccuracy,butintroduces survey operations. Data were acquired over a range of
flat-field non-linearity as described below. The result field-angle positions on the detector for both cases. For
is an L/S angle-integrated median λ centroid errors of thecalibrationdata,a’timed-sweep’methodwasusedto
0.11˚A and 0.081 ˚A, with mean errors of 0.24 ± 0.41 ˚A derivestarexposuretimesbydividingthespacecraftroll
and 0.21 ± 0.24 ˚A. Residual systematic errors of ∼1 ˚A rate by the ground-calibratedangular slit width. About
which we intend to correct in future work, persist in the 15% of the sweeps were rejected due to data faults. The
S channel for λ<1000 ˚A. net stellar signal was determined after subtracting adja-
Spectral resolution is measured from the observed au- cent sky background (∼5–10% of the stellar flux). The
roral spectra, together with airglow spectra for the S sweeps’stellarcountratesshowanearlyGaussiandistri-
channel, after applying the distortion corrections. Be- bution with a ∼15% width. For the sky-survey data, we
causethe sourcesarediffuse,the resultsareindependent used the sky mapping method (Edelstein et al. 2005) to
of spacecraft pointing knowledge. The first correction obtain source and background spectra of the field stars.
gives an L/S spectral half-energy line width, averaged The sky-map derived flux was verified to reproduce the
overtheangularfield,of3.2˚Aand1.9˚Awhilethesecond more comprehensive timed-sweep calibration results to
correction widths are 2.95 ± 0.76 ˚A and 1.71 ± 0.33 ˚A. within 10–15%.
These values are upper limits to the resolution because For the L channel, calibration observations of
any intrinsic width of multiplets in the source spectra G191B2BandHD188665provided253sand19sofstel-
were not accounted. larexposure,and13.2kand26.1kphotons(background
Detector differential non-linearity (DNL) arises from subtracted) for comparison to the spectra of Kruk et al.
the λ correction process from its application to dis- (1995) and Buss et al. (1995), respectively. The field
tributed noise sources such as intrinsic detector noise, starsHD 74753,HD 74273,andHD 72014were mapped
events from ions and penetrating cosmic rays, or instru- and referenced to their IUE spectra. The Aeff for each
mentallyscatteredairglowlines,asignificantnoisesource star was derived by dividing the observed count rate by
in the Short channel. The processing-induced DNL can thereferenceflux,withatypicalstatisticalerrorper1.5˚A
becorrectedandeliminatedifthedistributednoisespec- spectralbinof∼10%. The stars’Aeff wereaveragedus-
tra is known because the DNL is fixed in detector co- inginversecounting-errorweightingandfitwithathird-
ordinate space. Whether uncorrected DNL can charade orderpolynomialtoobtainthespectralcalibrationcurve,
as a spectral feature depends on the relative intensity showninFig. 2. Theindividualstar’sAeff deviatefrom
of the distributed noise to the true external signal and theweightedaverageby∼20–30%overthebandwithno
on whether both the DNL and the feature have similar discerniblein-flighttimetrend. AttheLbandcenter,we
spectralprofiles. WeusesimulationsoftheinducedDNL find Aeff= 0.20 cm2, corresponding to a full-field grasp
by processing a uniform random distribution of photons of Γ=3.2×10−5cm2sr−1.
on the detector to estimate that completely uncorrected For the S channel, a calibration observation of HD
DNL canappear as line emissionfeatures with an inten- 93521, conducted March to April 2004, provided 146 s
sity of ∼1.5 % of the distributed noise level. and 9300 photons (background subtracted) for compar-
An on-orbit ’dark’ background integration of events ison to the spectra of Buss et al. (1995). Also, the field
due to intrinsic detector, particle and penetrating radi- starHD 1337wasobservedinthe surveyduring Novem-
ation noise was accumulated in 42 ksec from 250 orbits ber2003for57secwith3500countsandreferencedtoits
that include 0% shutter observations which were con- HUT spectra. Eachstars’Aeff wasderivedusingeightλ
firmed to exclude occasional mis-positioned shutter air- 10–30binsthatavoidtheairglowLymanseriesandhave
glow leaks. The detector dark background rate is from . 6% statistical error,except for the λ 915-925bin with
0.02–.04countss−1˚A−1overthefullsciencefieldangular a 37% error. The Aeff bins were fit with a second-order
height. The dark background shows the expected λ cor- polynomial to obtain spectral calibration curves, shown
rection DNL features. In the S channel, instrumentally- inFig. 2. AttheSbandcenterwefindthatAeff=0.035
scattered Lyman α 1216 ˚A airglow radiation is a sig- cm2 for HD 1337, and Aeff= 0.016 cm2 for HD 93521.
nificant contributor to the intrinsic background. The S There appears to be a significant S-channel degradation
channelscatteringfunction,measuredbycomparinginte- with time. The earlier A corresponds to a full-field
eff
gratedclosed-slitandopen-slitbackgroundobservations, height S grasp of 3.0×10−6cm−2sr−1. HD 1337 (AO
iswellfitby anexponentialwith190˚Awidthandanin- Cas) is an eclipsing binary with a 3.4 day period and a
tensity of 0.016 counts s−1 ˚A−1 at the band center. The ∼±7.5% photometric variation in FUV resonance lines,
airglowcorrectioncanbeestimatedforobservationssub- such as NVλλ 1240 and CIVλλ 1550 (Gies & Wiggs
ject to time-variable airglow by using the Short channel 1991). OurmeasurementofAOCasshouldbeofamean
MgF2 filter regions to measure the relative intensity of intensity because it broadly samples the binary phase in
scattered Lyman α radiation. λbinsthatarelargecomparedtoobservedphasedveloc-
4. SENSITIVITY CALIBRATION ity shifts. We intend to refine our sensitivity calibration
and its temporal behavior in future work by measuring
The sensitivity to diffuse radiation is determined by more stars.
thegrasp,Γ,aproductofthesolid-angle,Ω,andeffective Systematic errors in the determination of exposure
area,Aeff. ThesolidangleforL/S,aproductoftheslit’s times and in the reference stellar spectra are likely to
angularwidthandtheviewedangleperλ,constrainedby
dominate our flux calibration accuracy. We estimate
thedetectorboundary,are1.6×10−4srand8.4×10−5sr, theseerrorssystematicoveralltobe∼25%. Therelative
respectively. TheA wasdeterminedusingspecialcal-
eff angular sensitivity variation from the mean response is
ibration observations where repetitive roll-sweeps of the
4 EDELSTEIN, et al.
6.5% as found from the off-axis response to bright in- sion line 3-σ sensitivity of 55 and 750 photons s−1 cm−2
band telluric diffuse night glow emission lines (e.g. O sr−1 for diffuse illumination filling the full angular field,
I 1356˚A, H I 1026˚A) accumulated over all observations. presuming a 10 ksec observation, and a background de-
Wepresumethatthe nightglowlinesareeffectivelycon- termination over 100˚A of bandpass.
stant in intensity over the sky because these data are Limits to the measurement of emission lines that are
taken over a wide range of zenith angles. Compared to not fully isolated from airglow lines, such as OVI 1032
its pre-launch performance, SPEAR’s on-orbit sensitiv- ˚A near Lyman β 1027 ˚A and CIII 977 ˚A near Lyman γ
itywasunfortunatelydegradedfromitsdeliveredperfor- 972 ˚A , depends on the intensity of the airglow line and
mance due to contamination caused by spacecraft and theaccuracyofspectralprofilingusedincompoundspec-
launch ground operations. tral line-fit modeling. We leave a systematic treatment
of such modeling and its limitations to papers specifi-
5. CONCLUSION:PERFORMANCE cally observing these spectral lines. As an example of
SPEAR’s sensitivity can be estimated from the on- the sensitivity, however, we provide that 7000 LU of O
orbitperformancevalues. Theultimate capabilitytode- VI λ 1032was measuredwith 5 σ significance in a 31 ks
tect diffuse emission with SPEAR depends on mission observationstowardtheEridanusLoop(Kregenow et al.
factors such as sky exposure, attitude knowledge and 2005).
backgroundtemporalbehavior. Thesefactorsaretreated
elsewhere in this issue (Edelstein et al. 2005). We find
thatthe dominantcauseofbackgroundis a combination
of detector dark noise and instrumentally scattered air- SPEAR / FIMS is a joint project of KASI & KAIST
glowfor the Shortchannel, anddarknoise andinterstel- (Korea) and U.C., Berkeley (USA), funded by the Ko-
lar continuum for the Long channel. In-orbit measured reaMOST andNASA GrantNAG5-5355. We thank the
L/Sbackgroundrateshavetypicalvaluesof∼0.02counts team for their dedication in producing the instrument,
s−1 ˚A−1 for diffuse high latitude sky fields (excluding and Jerry thanks Liz and Dan for immeasurable contri-
stars). These values correspondto anL/Sisolatedemis- butions.
REFERENCES
Bowyer,S.,Edelstein,J.,&Lampton, M.1997,ApJ,485,523 Nam,U.etal.2003,Proc.SPIE,4854,602
Buss,R.H.,Kruk,J.W.,&Ferguson,H.C.1995,ApJ,454,L55+ Nam,U.-W.etal.2002,JournalofAstronomyandSpaceSciences,
Edelstein,J.etal.2003, Proc.SPIE,4854,329 19,273
—.2005,ApJ,thisissue Rhee,J.G.etal.2002,JournalofAstronomyandSpaceSciences,
Gies,D.R.&Wiggs,M.S.1991, ApJ,375,321 19,57
Keski-Kuha,R. A.,Osantowski, J.F., Blumenstock, G. M.,et al. Ryu,K.etal.2003, Proc.SPIE,4854,457
1995, Proc.SPIE,2428,294 Strickland,D.etal.1999,J.Quant.Spectrosc.Radiat.Transf.,62,
Korpela,E.J.etal.2003,Proc.SPIE,4854,665 689
Kregenow,J.etal.2005, ApJ,thisissue
Kruk,J.W.,Durrance,S.T.,Kriss,G.A.,Davidsen,A.F.,Blair,
W.P.,Espey,B.R.,&Finley,D.S.1995,ApJ,454,L1+
SPEAR Instrumentation 5
Fig. 1.— TheSPEARdual-spectrographopticallayout. Lightiscollected bythemirrors(right) toamotorizedslitshutter. TheLong
channel (upper) is filtered by a CaF2 meniscus after the slit. The light then diffracts from gratings (left) to photon counting detectors.
TheShortchannel(lower) isun-filtered,althoughMgF2 windowsonthedetector abutthesciencefieldtomonitorairglow. Forscale,the
gratingis∼8cmlong.
6 EDELSTEIN, et al.
Fig. 2.— TheSPEAR Aeff calibrationfitto stellarobservations for(top panel) the Short bandinNovember 2003 (upper curve)and
April2004(lower curve),andfortheLongband(bottom panel)fittotheaveragehistogramfromfivestellarobservations.