ACCESS: Enabling an Improved Flux Scale for Astrophysics

1.2 MB·English

by  Mary Elizabeth Kaiser

ACCESS: Enabling an Improved Flux Scale for
Astrophysics
Mary Elizabeth Kaiser, Jeffrey W. Kruk, Stephan R. McCandliss, David J. Sahnow,
Robert H. Barkhouser, W. Van Dixon, Paul D. Feldman, H. Warren Moos, Joseph
Orndorff, Russell Pelton, Adam G. Riess1,3
1
Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles
Street, Baltimore, MD, USA 21218
0
1
Bernard J. Rauscher, Randy A. Kimble, Dominic J. Benford, Jonathan P. Gardner, Robert
0
2 J. Hill, Bruce E. Woodgate
n 2NASA Goddard Space Flight Center, Greenbelt, MD USA 20771;
a
J Ralph C. Bohlin, Susana E. Deustua
2 3Space Telescope Science Institute, San Martin Drive, Baltimore, MD, USA 21218
2
Robert Kurucz
]
M Harvard Smithsonian Center for Astrophysics, Garden Street, Cambridge, MD, USA 02139;
I
Michael Lampton
.
h
Space Sciences Laboratory, 7 Gauss Way, Berkeley, CA 94720;
p
-
o Saul Perlmutter
r
t University of California, Berkeley, Berkeley, CA 94720;
s
a
Edward L. Wright
[

University of California, Los Angeles, Los Angeles, CA 90095
1
v
5 ABSTRACT
2
9 Improvementsin the precisionofthe astrophysicalflux scale areneeded to answerfundamentalscientific
3 questions ranging from cosmologyto stellar physics. The unexpected discovery1,2 that the expansionof the
.
1 universe is accelerating was based upon the measurement of astrophysical standard candles that appeared
0 fainterthanexpected. Tocharacterizethe underlying physicalmechanismofthe “DarkEnergy”responsible
0
for this phenomenon requires an improvement in the visible-NIR flux calibration of astrophysical sources
1 to 1% precision. These improvements will also enable large surveys of white dwarf stars, e.g. GAIA,3 to
:
v advance stellar astrophysics by testing and providing constraints for the mass-radius relationship of these
i
X stars.
r ACCESS (Absolute Color Calibration Experiment for Standard Stars)4 is a rocket-borne payload that
a
willenablethetransferofabsolutelaboratorydetectorstandardsfromNISTtoanetworkofstellarstandards
withacalibrationaccuracyof1%andaspectralresolvingpowerofR=500acrossthe0.35-1.7µmbandpass.
Amongthe strategiesbeing employedto minimize calibrationuncertaintiesare: (1)judicious selectionof
standard stars (previous calibration heritage, minimal spectral features, robust stellar atmosphere models),
(2) execution of observations above the Earth’s atmosphere (eliminates atmospheric contamination of the
stellar spectrum), (3) a single opticalpath and detector (to minimize visible to NIR cross-calibrationuncer-
tainties), (4) establishment of an a priori errorbudget, (5) on-boardmonitoring of instrument performance,
and (6) fitting stellar atmosphere models to the data to search for discrepancies and confirm performance.
Keywords: ACCESS, standard stars, calibration, photometry, spectrophotometry, dark energy, Vega, Sir-
◦
ius, BD+17 4708,HD37725, NIST, sub-orbital, rocket, optical spectroscopy, NIR spectroscopy
Send correspondence to M. E. Kaiser: E-mail: [email protected]
Kaiser, M.E. 1 18th Annual CALCON Technical Conference
1. INTRODUCTION
Current astrophysical problems need a precise
(better than1%)networkofastrophysicalflux stan-
dardsspanningawidedynamicrangeacrossthevis-
ible and near-infrared (NIR) bandpass to address
fundamental questions (e.g. the characterof the ex-
pansion history of our universe) with the required
uncertainty.5
However, overall uncertainties in the astrophys-
ical flux scale exceed 1% in the spectral range ex-
tending from the ultraviolet through the NIR. And,
although isolated spectral regions may achieve 1%
levelsofprecision,itis uncertainifthe absolutecal-
ibration in those spectral regions is accurate to 1%.
Technological advances in detectors, instrumen-
tation, andthe precisionofthe fundamentallabora- Figure 1. Differential magnitude-redshift diagram for
tory standards used to calibrate these instruments dark energy models with Ω, w0, and w′ = xwa.11 Note
havenot beentransferredto the fundamental astro- that the models do not begin to distinguish themselves
physical flux scale across the visible to NIR band- from one another until z∼1 and the difference between
pass. Furthermore, the absolute normalization of models is of order 0.02 magnitudes (roughly 2%) at z∼
the current astrophysical flux scale is tied to a sin- 2.
glestar,Vega,astarthatistoobrighttobeobserved
with today’s premier optical telescopes. z,isplottedagainstitsrest-frameB-bandfluxtode-
termine the SNeIa Hubble brightness-redshift rela-
Systematic errorsassociatedwith problems such
tionship. Cosmologicaland dark energy parameters
as dark energy now compete with the statistical er-
aredeterminedfromtheshape,nottheabsolutenor-
rorsandthuslimitourabilitytoanswerfundamental
malization,ofthisrelationship. Sincetherest-frame
questions in astrophysics.
B-band is seen in different bands at different red-
The scientific impetus for the ACCESS program
shifts,the relativezero-pointsofallbandsfrom0.35
arose from the discovery of the accelerated expan- µm to 1.7 µm must be cross-calibrated to trace the
sion of the universe. These results6,7,8,9 compared supernovae from z=0 to z∼1.5. The term “abso-
the standardized brightness of high redshift (0.18<
lute color calibration” is defined as the slope of the
z <1.6)TypeIasupernovae(SNeIa)tolow-redshift
absolute flux distribution versus wavelength. This
SNeIa,10,1,2 showing that at a given redshift the
color calibration must be precise enough to clearly
peak brightness of SNeIa is fainter than predicted.
reveal the differences between dark energy models
The most plausible explanation for the unexpected
(Figure 1) over this range of redshifts.
faintness of these standard candles is that they are
Using SNeIa to distinguish dark energy models
further away than expected, indicating a period of
from one another levies a requirement for 1% preci-
acceleratedexpansionoftheuniverseand,hence,the
sion in the cross-color calibration of the SNeIa flux
presence of a new, unknown, negative-pressure en-
across a bandpass extending from 0.35−1.7µm.
ergy component - dark energy. Using SNeIa to dis-
tinguishdarkenergymodelsfromoneanotherlevies With the reduction ofstatistical uncertainties in
arequirementfor1%precisioninthecross-colorcal- supernovacosmology,theimportanceofunderstand-
ibrationoftheSNeIafluxacrossabandpassextend- ingandcontrollingsystematicuncertaintieshasgained
ing from 0.35−1.7µm. prominenceandisnowkeytoinvestigatingthedark
energy properties.12 Observations of higher redshift
Sincethen,severalclassesofobservationallytest-
SNeIa are needed to discriminate between differ-
able models have been proposed to explain the na-
ent dark energy models, thus motivating an abso-
ture of the dark energy. Accurate testing of models
lute color calibration to 1% precision in the NIR
throughobservationofSNeIadependsontheprecise
(1−1.7µm). Controlling the systematic errors to
determinationoftherelativebrightnessoftheSNeIa
this level of accuracy and precision is required, not
standard candles. For each supernova, its redshift,
Kaiser, M.E. 2 18th Annual CALCON Technical Conference
only for the absolute color calibration, but also for sphere and avoiding both atmospheric emission and
other sources of systematic errors which themselves absorptionthatpresentsaseverecontaminanttothe
depend on the color calibration (e.g. extinction cor- stellarspectrumlongwardof0.85µmatbothground
rections due to the Milky Way, the SN host galaxy, andballoonaltitudes. Measurementrobustnessalso
and the intergalactic medium, in addition to the benefits from (4) obtaining observations with an in-
K-corrections which provide the transformation be- strument that uses a single optical path and detec-
tween fluxes in the observed and rest-frame pass- tor across its full bandpass of 0.35−1.7µm, thus
bands). eliminating cross-calibration systematic errors. Es-
tablishing and tracking (5) an a priori error bud-
2. ACCESS OVERVIEW get,maintainsfocusonthemagnitudeoferrorsthat
canbetoleratedatthesub-systemlevel. Performing
This program, ACCESS - “Absolute Color Cal- NIST traceable sub-system and end-to-end payload
ibration Experiment for Standard Stars”, is a se- calibrations, yields an absolute calibration in addi-
riesofrocket-bornesub-orbitalmissionsandground- tion to the relative calibration that is the focus of
basedexperimentsthatwillenablethe absoluteflux many scientific applications. Furthermore, (6) mon-
for a limited set of primary standard stars to be itoring and tracking payload performance with an
established using calibrated detectors as the funda- on-boardmonitoringsourcethatutilizesallelements
mental metrology reference. These experiments are oftheopticalpathenablesknowledgeofsystemper-
designed to obtain an absolute spectrophotometric formance prior to and during payload flight.
calibration accuracy of <1% in the 0.35−1.7µm
Current technology has enabled increased cali-
bandpass at a spectral resolution greater than 500
brationprecisionfordetector-basedirradiancestan-
bydirectlytracingthe observedstellarfluxesto Na-
dards, providing a factor of two reduction in un-
tionalInstituteofStandardsandTechnology(NIST)
certaintiesfromthe previous,source-based,spectral
irradiancestandards. Transferofthe NIST detector
irradiance scale in the visible and even greater im-
standards to our target stars will produce an abso-
provements are realized in the NIR,13,14,15 making
lute calibrationof these standardsin physicalunits,
standard detectors the fundamental reference cali-
including the historic absolute standard Vega.
brator of choice.
ACCESSwillreduceuncertaintiesinthe current
And while the full end-to-end calibration of an
standardstar calibrationsystemthroughcarefulat-
instrument in physical units with an absolute labo-
tention to details, both large and small, that can
ratorystandardispreferred,itisnotalwaysfeasible.
impact the success of this project. In designing this
Consequently the essence of the ACCESS program
program we have sought to identify and ameliorate
istoestablishthiscalibrationforthefullyintegrated
potentialsourcesoferrorthatcouldprohibitachiev-
telescopeandspectrographandtransferthiscalibra-
ing a 1% calibration. As a result, we have (1) tar-
tion to the stars using both NIST irradiance stan-
getedthejudiciousselectionofstandardstars. Only
dards and the end-to-end NIST SIRCUS16,17 cali-
existing (known) standard stars with previous cal-
brationfacility. Asaresult,theabsolutefoundation
ibration heritage will be observed. Selection crite-
of the existing network of standards stars, includ-
ria include restrictions to spectral classes that ex-
ing the standards Vega and Sirius, will be strength-
hibitminimalspectralfeaturesandforwhichrobust
enedthroughaccurate,higherresolution,higherpre-
stellar atmospheres models are available (e.g. pure
cision,broaderbandpassmeasurementsextendingto
DA white dwarf stars, AV stars). (2) Stellar at-
lowerfluxlevels. Thisimprovednetworkofstandard
mosphere modeling of our measurements will pro-
stars,extendingto10th magnitude,willbeavailable
vide an important cross-check on the observations
to all telescopes as standard sources.
and enable the extension of these measurements to
wavelengthsoutside ourobservedbandpass. Within
3. FLUXSTANDARDS&CALIBRATION
the signal-to-noise constraints of the observations,
standard stars with flux levels extending to ∼10th
Ultimately, observed astrophysical fluxes must
magnitude will be selected to enable observationby
be converted to physical units. Three of the most
large aperture telescopes and thus eliminate addi-
common methods of determining the absolute color
tional calibration transfers.
calibration of stellar fluxes are solar analog stars,
Uncertainties in the absolute flux will be further stellar atmospheremodels, andcomparisonto certi-
minimizedby(3)observingabovetheEarth’satmo- fied laboratory standards. The existing precision in
Kaiser, M.E. 3 18th Annual CALCON Technical Conference
eachofthese methods is inadequate for darkenergy
SNe cosmology.
3.1 Solar analog stars
Useofsolaranalogstarsasastandardsourcere-
lies upon the star having the same intrinsic spectral
energy density (SED) as the sun. Unfortunately, no
star is a true solar analog. Even G-type stars with
the most-closely matching visible spectra can differ
by a few percent. In the NIR, differences in mag-
18
netic activity can restrict the accuracy to 5%. In
addition, uncertainties in the solar SED itself are
2-3%.19
3.2 Stellar atmosphere models
Stellaratmospheremodels arecurrentlythe pre- Figure 2. Uncertainties in the absolute flux for Vega:
ferred method for calibrating stellar fluxes due to HST/STIS observations18,20 (black line), the Kurucz
stellar model with T =9400K (green), and the Ku-
the agreementbetweenthe models and the observa- eff
rucz stellar model at 9550K (red) are compared. The
tionsandtheincreasedresolutionofboththemodels
observations exhibit better agreement with the cooler
and the data. In the ultraviolet and visible region model at the longer and shorter wavelengths. The hot-
of the spectrum, this calibration network is based termodelagreesbetterwiththemeasuredfluxby∼1%
at4200–4700˚A.Vega,apole-onrotatingstar,presentsa
on the relatively featureless spectra of unreddened
range of temperatures, which introduce added complex-
hot white dwarf (WD) stars with pure hydrogenat- ity to obtaining a robust stellar model.
mospheres. Absolute photometry of Vega is used to
normalize the spectral energy distributions of these 3.3 Certified Laboratory Standard Sources
stars and their stellar models to an absolute flux
To reduce uncertainties to <1%, the NIST fun-
scale.
damentallaboratorystandardsmustbedirectlytied
Currently, the three primary WD standards of
to astrophysical sources, i.e. stars, at this level of
the HST CALSPEC network are internally consis-
precision.
tent to an uncertainty level of 0.5% in the visible
Photometric measurements of Vega have been
withlocalizeddeviationsfrommodelsrisingto∼1%
overthe4200−4700˚Aspectralrange,18 anda±1% absolutelycalibratedagainstterrestrialobservations
uncertaintyintheNIR(1−2µm).18 Thus,thelevel ofcertifiedlaboratorystandards(e.g.tungstenstrip
lamps, freezing-point black bodies).25 These abso-
of agreement between the model and the data is a
lutephotometryexperimentsprovidethenormaliza-
function of wavelength. Any systematic modeling
tionfor the networkofstellar models andtemplates
errors that affect the shape of the flux distributions
thatareusedaspracticalabsolutestandards. These
of all three WD stars equally cannot be ruled out
absolute calibrations to standard sources were dif-
and would make the actual error larger.
ficult. Observation of the laboratory source using
CurrentuncertaintiesintheextensiveNIR(1.0<
the full telescope was typically achieved by placing
λ <1.7µm)networkofstandardstarsare∼2%.21,22,23
the source at a nearby distance (∼200m) and cor-
A recent compilation24 of IR standard star cali- recting for the intervening, non-negligible, air mass.
brationsbasedondirectabsolutemeasurementsand This resulted in errors due to the large and vari-
indirect calibrations through modelling and extrap- able opacity of the atmosphere. This methodology
olation/interpolationofobservationstests the inter- is being re-visited using current technology.26,27,28
nal consistency of these measurements and exam- Other programs (e.g. Pan-STARRS, LSST) plan to
ines the impact of extrapolating the IR data into use dedicated telescopes to simultaneously monitor
thevisible. Thedataarefoundtobeconsistent,but theatmospheretoprovidecorrectionsforthescience
adjustmentsof∼±2%topublishedcalibrationsare observations at the neighboring telescope.29,30,31
recommendedandthedeviationsofVegafromatyp-
In the IR, introducing the absolute calibrator at
ical A0V star between the visible and IR are noted
the focal plane of the telescope minimized atmo-
and quantified.
Kaiser, M.E. 4 18th Annual CALCON Technical Conference
sphericabsorptionofthelaboratorystandard,32 but In order to improve on the limited precision of
required corrections due to differences in the opti- modelsanddirectlytieourfluxesforthebrightstars
cal train and atmospheric corrections to the stellar Vega and Sirius to the fainter stars needed for large
observations. telescopes, the V=8.4mag Spitzer/IRAC standard
40
HD37725 and the V=9.5mag Sloan Digital Sky
Discrepancies of >10% in Vega’s flux exist at
◦
Survey (SDSS) standard BD+17 4708 will be ob-
0.9 − 1µm, whereas the measurements from 0.5 −
0.8µm agree to ∼1%.20,25 Beyond 1µm, windows served as primary targets. Using a rocket platform,
observingtime aboveatmosphereis limitedto ∼400
of low water vapor absorption have been used for
absolute photometry.33,34 seconds. Consequently, standard star selection was
constrainedtotargetsbrighterthan 10thmagnitude.
Theuncertaintyinthestandardstarfluxcalibra-
◦
BD+17 4708 has precisely established fluxes on
tion networkrelative to the fundamental laboratory
the HST WD scale41 andwill tie both the HST and
standards currently exceeds 1%.
SDSS networksdirectly to the NIST flux scale. The
Certified Detectors: ◦
fluxes for Vega and BD+17 4708 were measured by
Acalibrationmethodologybasedonpreciselycal- STIS; and the sensitivity of our payloadis sufficient
ibratedphotodiodedetectorstandardsisadvocated35 to confirm the flux ratio of BD+17◦4708/Vegawith
given the limitations and challenges imposed by at- a S/N of 1%. NIR spectrophotometry of the his-
mospheric transmission and radiance standards in torically fundamental standard, Vega,whose flux at
achieving 1% photometry from the ground. The 5556 ˚A sets the absolute level for all standard star
calibrationprecision and stability of photodetectors networks, has not yet been done. Observations of
hasgreatlyimprovedsinceearlypioneeringmeasure- VegaandSiriusarealsoessentialfortyingtheNIST
ments.36,37 NIST ∼2σ uncertainties in the absolute fluxes to the extensive Cohen IR network.21,22,23
responsivity of standard detectors are ∼0.2% for Si Models for all the targets observed by ACCESS
photodiodes and 0.5% for NIR photodiodes.38 This will be computed to confirm the consistency of our
increasedprecisioninthephotodetectorcalibration, observations over 0.35−1.7 µm and to provide an
ease of use, and repeatability, now make standard extension to other wavelengths.
detectors the calibrator of choice.
Two flights are required for each of the two ob-
◦
servingfields(Vega+BD+17 4708;Sirius+HD37725)
3.4 Observing Strategy
toverifyrepeatabilityto<1%,whichisessentialfor
A rocket platform was selected for the ACCESS provingtheestablishmentofstandardswith1%pre-
observationsbecausetherocketfliescompletelyabove cision.
the Earth’s atmosphere, thus eliminating the chal-
lengingproblemofmeasuringtheresidualatmospheric 4. ACCESSTELESCOPEANDSPECTRO-
absorptionandstrongatmosphericemissionseenby GRAPH
ground-basedobservationsandevenbyobservations
TheACCESStelescopeisaDall-Kirkhamcassegrain
conductedat balloonaltitudes. Ahigh-altitude bal-
design with aluminum and fused silica over-coated
loonflyingat39kmisabove99%oftheatmospheric
Zerodur mirrors. The telescope feeds a low-order
water vapor (the primary source of absorption at
echellespectrographwithacooled,substrate-removed,
these wavelengths), but this is still well below most
HgCdTe detector.
of the source of the time-variable OH airglow emis-
sion, which originates in a 6−10 km layer at an The telescope optical bench has flown a number
altitude of ∼89 km.39 This forest of emission lines, oftimes. Itconsistsofaninvarprimarymirrorbase-
extendingfrom0.85µmto2.25µm,ismuchstronger platewithacantileveredinvartube, whichservesas
than the continuum flux from a 13th mag star. The both a heat shield and mounting structure for the
number, strength, and variability of these lines has secondarymirrorand a co-alignedstar tracker. The
implications for increased statistical noise and sys- optical bench sits on thermal insulating pads and
tematic effects resulting fromsubtractionofthe OH is bolted to a radex joint to which the outer rocket
backgroundin additionto the challengeofeliminat- skins are attached. The “thermos bottle” configu-
ing the strong scattered light within the instrument ration of rocket skin and inner heat shield provides
arising from these lines. the thermal isolation required to keep the primary
andsecondaryvertex-to-vertexdistancefixedtoless
than 0.001 inch for the duration of the flight.
Kaiser, M.E. 5 18th Annual CALCON Technical Conference
Image Spots for 5 Wavelengths per Order
Each Spot is Referenced to the Ch ief Ray
Y Detector [microns]556666990112494051048260 0.9000µm Y Detector [microns]555555778899273849482604 1.1600 µm Y Detector [microns]555555667788172738604826 1.4000 µm Y Detector [microns]555555455667506172482604 1.6800 µm Y Detector [microns]5555555344556640506176048260 1.9000 µm
−X85 D68e−te8c5to3r2 [−m8i4c9ro6n−s8]460 −39X60 D−3et9e2c4to−r3 [8m8i8c−ro3n8s5]2 X3 D24etec3to6r0 [mi3c9ro6ns]432 X52 D92ete5c3to2r8 [m5i3c6ro4ns5]400 918X0 D9e2t1ec6to9r 2[m52icr9o2n8s8]

Y Detector [microns]777777122334839495260482 0.4500 µm Y Detector [microns]666666455667838495048260 0.5800 µm Y Detector [microns]666666223344161728048260 0.7000 µm Y Detector [microns]566666901122940516482604 0.8400 µm Y Detector [microns]555666899011849405604826 0.9500 µm
−X8 5D6e8t−ec8t5o3r 2[m−8ic4r9o6n−s]8460 −39X6 0D−e3t9ec2t4o−r 3[m88i8cr−o3n8s5]2 X 3D2e4tect3o6r0 [mi3cr9o6ns]432 X5 2D9e2tec5t3o2r8 [m5i3cr6o4ns5]400 918X0 D9e2te1c6tor9 [2m52icro9n2s8]8

99565182 0.3200 µm 8100 0.3870 µm 7290 0.4670 µm 66880548 0.5600 µm 66568482 0.6330 µm
Ftlehifgetusartneadr3.eanrRteeariynttchriaedceDenavtlileo-wKniotrfhkheAaCpmrCimCESaasrSsy.ePgmraairrianrlolertell(recaseycnostpefreroamotf Y Detector [microns]9999334549502604 Y Detector [microns]7778899083944826 Y Detector [microns]7777011272834826 Y Detector [microns]6666566784958260 Y Detector [microns]6666344572832604
7830 7020 6318
figure). The telescope secondary is at left in the figure −75X24 D−e7t4e8c8to−r7 [4m5i2c−ro7n4s1]6 −39X6 D0−et3e9c2to4r− [3m8i8c8ro−n3s8]52 288X 3D2e4tect3o6r 0[mic3r9o6ns]432 X5 2D9e2tec5t3o2r 8[m5ic3r6o4ns]5400 91X80 De9t2e1c6tor9 [2m5i2cro9n2s8]8
andthegratingistheopticalsurfaceattheextremeright
Figure 5. Spot diagrams for an optical layout at a se-
in thefigure.
lection of wavelengths spanningtheACCESS bandpass.
Each box is labelled in microns and each spot is refer-
enced to the chief ray. In general, the ray trace yields
1×4 pixel images at the blaze wavelength, with 18×18
µm pixels.
time viewing and control by the operator on the
ground.
Theopticalelementsaresealedinastainlesssteel
vacuum housing to provide for contamination con-
Figure 4. Raytrace view of the ACCESS spectrograph trol,thermalstability,andcalibration. Afusedsilica
illustratingthegratingontheright,thecross-dispersing
entrance window sits behind the slit jaw. The de-
prism,andthefirstthreeordersdispersedbythegrating
tectorismountedonafocusadjustmentmechanism,
and incident on the detectorat left.
and the grating and cross disperser are mounted in-
side along with a set of baffles. The spectrograph
Thespectrographisconfiguredasanechelle(Fig.4)
vacuum is maintained by a non-evaporable getter
and used in 1st (9000−19000˚A), 2nd (4500−9500
and is monitored by an ion gauge. The typical vac-
˚A)and3rdorders(3000−6333˚A).Itconsistsofjust uum is < 10−6Torr.
two optical elements, a concave diffraction grating
The focal plane array will be a 1K×1K HgCdTe
with a low ruling density, and a prism with spheri-
device, with the composition tailored to produce a
callyfiguredsurfacesplacedintheconvergingbeam.
long-wavelengthcutoffat∼1.7µm,andtheCdZnTe
The separation of the three orders on the detector
growthsubstrateremovedtoprovidehighNIRquan-
is ∼1mm. The resolution of the spectrograph de-
tumefficiency(QE)andresponsethroughthevisible
pends on the telescope point spread function (PSF)
to the near-ultraviolet (Fig. 6). Pioneered by Tele-
andthesizeofthedetectorpixels. Forthetelescope
′′ dyneImagingSensors(TIS)toenabletheHST/Wide
PSF of 1.17 (as achieved on recent flights with a
Field Camera 3 (WFC3) to operate without a cry-
similar design) the 18µm pixels of the detector are
ocooler,thesedetectorsrequireamuchsimplercool-
critically sampled and produce constant wavelength
ing system than that required by standard HgCdTe
resolutionelements in eachorder,giving a resolving
detectors with cutoffs at longer wavelengths. The
power ranging between 500 and 1000 (Fig. 5).
band gapcorrespondingto the 1.7µmcutoff yields
The spectrograph housing42 will be evacuated lowdarkcurrentatoperatingtemperaturesnear145
andmountedtothebackoftheprimarymirrorbase- −150Kandmakesthedetectorsrelativelyinsensitive
plate. An angled mirrored plate with a 1mm aper- to thermal background radiation, though moderate
tureinthecenterlocatedatthetelescopefocusserves cooling of the detector surroundings (the evacuated
as the slit jaw, allowing light to enter the spectro- spectrograph) will be required. While the detector
graph while reflecting the region surrounding the does have a view factor to the warmer telescope op-
targetintoanimage-intensifiedvideocameraforreal- ticsupstream,theseareseenonlythroughthesmall
Kaiser, M.E. 6 18th Annual CALCON Technical Conference
1.0
)
0.8 FPA 104 esel−1 Å)−1 10−8 Sirius 0th mag in V
0 rs −1 10−10 Vega
QE 0.6 N = 20s cm −2 10−12 BD+17 4708 10th mag in V
Sg
0.4 x(er 10−14 400. sec integration
u(
l
F 5000 10000 15000 20000
0.2
Wavelength (Å)
0.0
400 600 800 1000 1200 1400 1600 1800
Wavelength [nm] Figure 7. The flux limit for a signal-to-noise of 200 in
a 400 second rocket flight is shown for each of thethree
ACCESS orders. The flux for three selected targets is
Figure 6. Quantum Efficiency of a flight candidate de- overplotted for comparison. Except at theveryshortest
tector. The increase in signal below 400nm is probably wavelengths for BD+17◦4708, a signal-to-noise ratio of
theresultofaredleak. Theseinitialmeasurementswere 200 is achievable at a resolving power of 500 in a single
madewhenthedetectorwasslatedforaNIR-onlyinstru- rocket flight.
ment.
1. Establishastandardcandlethatcanbetraced
slit jaw aperture, reducing their background contri- toa NIST detector-basedirradiancestandard.
bution to acceptable levels.
2. Transfer the NIST calibrated standard(s) to
The detector array is indium bump-bonded to a
the ACCESS payload- calibrate the ACCESS
Hawaii 1-R multiplexer. The resulting device has a
payloadwithNISTcertifieddetector-basedlab-
formatof1024×1024pixels,each18µm×18µmwith
oratory irradiance standards.
1014×1014 active imaging pixels and 5 rows and
columns of reference pixels at each edge. The refer- 3. Transfer the NIST calibrated standard(s) to
ence pixels are connected to capacitive loads rather the stars- observethe standardstarswith the
than active imaging pixels; they track the effects of calibrated ACCESS payload.
thermal drift and low frequency noise that plagued
4. Monitor the ACCESS sensitivity - track sys-
earlier generations of such devices.
tem performance in the field prior to launch,
Usingrealisticvaluestocharacterizethethrough- while parachuting to the ground, and in the
put of the optical components, the grating, and the laboratory, to monitor for changes in instru-
detector indicates that subsecond integration times ment sensitivity.
will be required to avoid saturation of the detector
5. Fit stellar atmosphere models to the flux cali-
forthe brightstarsSirius andVega(Fig.7). Proven
bratedobservations-confirmperformance,val-
algorithmsforsubarrayreadoutsofthedetectorwill
idate and extend standard star models.
be used to accomplish this.
For the fainter targets, a 400 secondobservation Determination of the ACCESS instrument sen-
yields a S/N of 200 per spectral resolution element sitivity is, in principle, a simple process of knowing
down to the Balmer edge. Additional binning can theratioofthetotalnumberofphotonsenteringthe
further increase the background subtracted signal- telescope aperture to the total number of photons
to-noise ratio of the acquired spectrum. detected by the spectrographdetector as a function
of wavelength.
5. ACCESS CALIBRATION
Quantificationofthenumberofphotonsentering
5.1 Calibration Overview the telescope requires a source with a known num-
ber of photons in a beam matched to the entrance
TheACCESScalibrationprogramconsistsoffive aperture of the telescope. For ACCESS, this source
principal components. (Fig.8) will consist of a stellar simulator and a
Kaiser, M.E. 7 18th Annual CALCON Technical Conference
Figure 8. Ground calibration configuration including the light source feeding a dual-monochromator, which is fiber
fed to an integrating sphere. The output of the integrating sphere is baffled to match the collimator f-ratio. The
double pass configuration for the collimator is shown. The collimated beam, with the flat mirror removed, is the
calibrated light source for the ACCESS instrument (telescope with spectrograph).4
Kaiser, M.E. 8 18th Annual CALCON Technical Conference
collimator to provide the “star-at-infinity” required The relative expanded uncertainty (∼2σ) error
by the telescope. The stellar simulator (Fig.8) will of the absolute responsivity of the Si photodiodes
becomprisedofapinholeplacedatthecollimatorfo- is ∼0.2%.38 With the NIST Spectral Comparator
cusandfedbyanintegratingspherebaffledtomatch Facility (SCF), the spectral responsivity of the NIR
the focal ratio of the collimator. To ensure spectral photodetectors can be measured with a combined
purity, the integrating sphere is fed by a fiber optic relativestandarduncertainty ofless than 0.4%.43,38
coupled to the output of a dual monochromator.
Knowledgeofthe totalnumberofphotonsinthe Collimator Avacuumgradecollimator,underni-
output beam of the collimator, which is the input trogenpurge, will be used to illuminate and charac-
beamto the telescope,willbe providedbytwomea- terize the ACCESS instrument in a darkroom envi-
surements. The firstis a simple measurementof the ronment. Determination of the collimator through-
intensity of the light passing through the pinhole of putwillbeachievedbyusingthecollimatorindouble-
the stellar simulator by a NIST calibrated photodi- pass with the stellar simulator. The double pass
ode transfer standard and the second is a measure- configuration(DPC)consistsoftheprimaryandsec-
ment of the reflectivity of the collimator (§ 5.2). ondarycollimatormirrorsincombinationwithahigh
The end-to-end calibration of the telescope with quality flat mirror (Fig8). Measurement of the re-
spectrograph may then be performed as a function flectivityoftheflatmirrorandtheinputandoutput
of wavelength by simply measuring the intensity of signalofthestellarsimulatorwillresultinthedeter-
the simulated star, measuring the count rate at the mination of the reflectivity product of the primary
spectrograph detector, and correcting for the colli- and secondary mirrors.
mator attenuation of the simulated star.
The signal measured from the stellar simulator MeasurementoftheReflectivityoftheFlat
by the photodiode is a radiantflux (power)and has The relative reflectivity of the flat mirror will be
unitsofergs−1. Thesignalfromthestarisanirradi- measuredasafunctionofwavelengthusingamono-
ance, and has units of ergs−1cm−2. The calibrated chromator. Thisrelativemeasurementistransferred
irradiance is then obtained after precise measure- to anabsolutereflectivitymeasurementthroughthe
mentofthe telescopeprimaryandsecondarymirror measurementofthe witness samplesinconcertwith
dimensions and dividing the calibrated radiant flux the flat at each wavelength. These measurements
by the illuminated area of the primary mirror. areperformedinsideacleantentundernitrogengas
purge in a dark room to maintain cleanliness and
Although simple in principle, systematic effects,
a low light level environment. The photodiodes are
such as the uniformity of reflective coatings, match-
cooledforstabilityandlowdarkcurrent. Thesource
ing of the collimator and telescope apertures, the
is operatedin a thermally controlledhousing with a
spatial uniformity of the photodiode detectors, the
radiometric power supply which controls the light
transmission of the slit, the scattered light determi-
rippleto≤0.4%. Amonitoringdiodewilltrackany
nation,the determinationofthe areaoftheprimary
variationsinthesourceoutputduringthereflectivity
andsecondarytelescope mirrors,the stability ofthe
measurements and the data will be correspondingly
lightsource,etc.,mustbecloselytrackedifthispro-
corrected. Background signal measurements, with
cess is to yield the required precision and accuracy.
the source shuttered, will be taken before and after
5.2 NIST Absolute Calibration Transfer each reflectivity measurement.
StandardDetectors TwotypesofNIST-calibrated
Collimator Throughput The second step in
standard photodiodes will be required to calibrate
the collimator calibration establishes the reflectiv-
the spectral range of the ACCESS instrument from
ity product of the combined primary and secondary
3500˚A – 1.7µm. A thirteen year pedigree of sta-
mirrors. For this measurement a light source, dual
bility dictates our choice of a Si photodiode in the
monochromator,spectralon-coatedintegratingsphere,
visible. In the NIR, an InGaAs photodiode will be
andF/12baffleboxwithpinhole(Fig.8)combineto
used. NIST will measure the absolute spectral re-
generate a stellar simulator, or artificial star, which
sponsivity and map the spatial uniformity for each
directly illuminates the DPC. To avoid polarization
photodiode.
of the beam, no mirrors fold the beam. Instead, the
Kaiser, M.E. 9 18th Annual CALCON Technical Conference
Ground Calibration Steps
1. Reflectivity of flat as a function of wavelength.
Relative of entire surface, cross-calibrate to witness
Absolute calibration of witness
2. Relative calibration of stellar simulator (input beam to telescope).
Measure F/12 output of pinhole (artificial star)
Measure the return beam from F/12 collimator in double pass off flat.
3. Check uniformity of collimator beam.
Scan sub-aperture in auto-collimated configuration
4. Characterize collimator to telescope pupil match.
5. Measure slit losses.
Slit-in, slit-out method
Direct characterization of PSF with flight array detector in focal plane
6. Measure PSF of telescope spectrograph at spectrograph focal plane.
7. Characterize flat-field response of spectrograph detector.
8. Characterize linearity of spectrograph detector.
9. Characterize linearity of absolute calibration standards.
10.Characterize read noise of the detector.
11.Characterize readout properties of the detector.
Figure 9. The annular distribution of rays in the im-
12.Absolute calibration of telescope and spectrograph.
age show the collimator beam spot on the photodiode Measure monochromatic pinhole output into collimator.
Measure response of flight array detector at spectrograph focal plane.
detector. The black circle enclosing these rays depicts
theexpectedsizeoftheNISTbeamduringtheabsolute
responsivity calibration. Figure 10. Outlineof theground calibration steps.
direct measurement of the artificial star and the re- End-to-end ACCESS instrument throughput
turn measurement of the image after the DPC are To check for losses at the slit, the instrument will
accomplishedusingarotationstagetoplacethepho- be illuminated with and without the slit and the
todiode into the incident and return beams. We ex- throughput measured at the telescope focal plane
pect an 85% overlap(Fig.9) in the photodiode area with a photodiode for eachcase. The results will be
sampled by the stellar simulator and the NIST cali- comparedtoverifythatslitlossesareinsignificantas
bration beam. expected with the 1mm (∼32 arcsecond) ACCESS
slit aperture. This will be checked against a direct
A reference photodiode at an output port of the
measurement of the telescope PSF using the array
integrating sphere will monitor the source for signal
variations. The source will be shuttered and mea- detector.
surements will be taken before and after each mea- Next the PSF will be measured with the array
surement to correct for the backgroundsignal. detector at the spectrograph focal plane, again im-
ageswillbecheckedforsystematiceffects. Measure-
Thef-numberofthe artificialstarsystemwillbe
matchedtothe f-numberofthe collimatorto ensure ments will be made to check for, characterize, and
eliminate sources of scattered light.
that the beam slightly underfills the full aperture of
the collimator. From measurments of the artificial Theabsolutecalibrationofthetelescopecanthen
star input and return signals without an aperture bedeterminedbymeasuringtheartificialstarsignal
stop we determine the efficiency of the collimator. at the pinhole with a NIST calibrated photodiode,
then measuring the signal at the ACCESS instru-
To ensure that the telescope will be underfilled
mentfocalplanewiththeinstrument(spectrograph)
and no light lost, an aperture stop will be inserted
detector.
intothecollimatedbeam. Theinputandreturnsig-
nals will be measured and the size of the aperture
stop will be known from prior measurements. Thus Cross-Checks Theprimarycalibration,described
the input signal to the telescope is determined.
above, uses a NIST precision calibrated photodiode
Theuniformityoftheilluminationattheflatwill detector standard to map the instrument’s sensitiv-
be checked using a rotating mask with a small sub- ity in a series of monochromatic wavelength steps.
apertureinfrontoftheflatandmonitoringfordiode A brief outline of calibrationthat will be performed
signal fluctuations. The flat is then removed from is presentedin Figure 10. After this calibration,the
the system and the collimator is now used in single instrument will be transported to NIST where the
pass with the artificial star to illuminate the tele- throughput will be determined using two additional
scope and determine its end-to-end sensitivity. methods. The first method will calibrate the end-
to-end sensitivity using the Spectral Irradiance and
Radiance Responsivity Calibrations with Uniform
Kaiser, M.E. 10 18th Annual CALCON Technical Conference