ARCADE 2 Measurement of the Extra-Galactic Sky Temperature at 3-90 GHz

0.33 MB·English

by  D. J. Fixsen

Draftversion January21,2009
PreprinttypesetusingLATEXstyleemulateapjv.10/09/06
ARCADE 2 MEASUREMENT OF THE EXTRA-GALACTIC SKY TEMPERATURE AT 3-90 GHZ
D.J. Fixsen1, A. Kogut2, S. Levin3, M. Limon4, P. Lubin5,P. Mirel6, M. Seiffert3, J. Singal7, E. Wollack2, T.
Villela8 and C. A. Wuensche8
Draft version January 21, 2009
ABSTRACT
The ARCADE 2 instrument has measured the absolute temperature of the sky at frequencies 3,
9 8, 10, 30, and 90 GHz, using an open-aperture cryogenic instrument observing at balloon altitudes
0 with no emissive windows between the beam-forming optics and the sky. An external blackbody
0 calibrator provides an in situ reference. Systematic errors were greatly reduced by using differential
2 radiometers and cooling all critical components to physical temperatures approximating the CMB
temperature. Alinearmodelisusedtocomparetheoutputofeachradiometertoasetofthermometers
n
onthe instrument. Smallcorrectionsaremadeforthe residualemissionfromthe flighttrain,balloon,
a
J atmosphere, and foreground Galactic emission. The ARCADE 2 data alone show an extragalactic
rise of 50±7 mK at 3.3 GHz in addition to a CMB temperature of 2.730±.004 K. Combining the
5
ARCADE2datawithdatafromtheliteratureshowsabackgroundpowerlawspectrumofT =1.26±
] 0.09 [K] (ν/ν0)−2.60±0.04 from 22 MHz to 10 GHz (ν0 = 1 GHz) in addition to a CMB temperature
O of 2.725±.001 K.
C Subject headings: cosmology: Cosmic Microwave Background— cosmology: Observations
.
h
p 1. INTRODUCTION The ARCADE 2 is a balloon-borne double nulled in-
- strument with seven radiometers at frequencies ranging
o The standard big bang model places the formation of
r theCMBatz ≈6×106 withanearlyperfectblackbody from 3 to 90 GHz mounted in a liquid helium bucket
t dewar. Each radiometer consists of cryogenic and room
s spectrum. The black body spectrum remains in ther-
a mal equilibrium with the electrons and ions in the early temperature components. A cryogenic Dicke switch op-
[ erating at 75 Hz alternately connects the amplification
universe until the surface of last scattering at z=1089.

1 Measurementsbythe FarInfraredAbsoluteSpectropho- chain to either a corrugated horn antenna (Singal et al.
2005)oraninternalreferenceload(Wollacket al.2007).
v tometer(FIRAS)instrumentacrossthepeakoftheCMB
The temperature of the reference load can be adjusted
5 spectrum (∼ 60 GHz to ∼ 600 GHz) limit deviations
toproducezerodifferentialsignal,nullingtheradiometer
5 from a blackbody, with temperature 2.725±.001 K, to
output. The horn in turn views either the sky or an ex-
5 belessthan50partspermillion(Fixsen&Mather2002,
0 Fixsen et al. 1996), but measurements at centimeter or ternalblackbodycalibrator. Theblackbodytemperature
. longerwavelengthsare less restrictive. Plausible energy- can be adjusted to match the sky temperature, nulling
1
instrumental offsets.
0 releasing processes including star formationand particle
The calibrator (Fixsen et al. 2006) has reflection of
9 decay or annihilation could produce observable distor-
less than -45 dB and can be positioned to fully cover
0 tions at centimeter or longer wavelengths while evading
the aperture of any of the horns. Residual reflections
: constraints at millimeter wavelengths.
v from the calibrator are effectively trapped within the
Atradiofrequenciesbelow10GHz,the radiationfrom
i horn/calibrator system. Within this system the calibra-
X the sky is increasingly dominated by synchrotron and
tor absorbs almost all of the radiation because the horn
free-free emission both from our own Galaxy and dis-
r
emissivity is low and the calibrator emissivity is high.
a tant point sources and perhaps distant diffuse sources.
Theeffectofreflectionfromthecalibratorsystemispro-
The ARCADE 2 (Absolute Radiometer for Cosmology,
portional to the difference in temperature between the
Astrophysics and Diffuse Emission) instrument observes
calibrator and the horn antenna throat.
theCMBspectrumatfrequenciesadecadebelowFIRAS
To minimize instrumental systematic effects the horns
to observe the spectrum where the crossoveroccurs.
are cooled and maintained at a nearly constant temper-
2. THEINSTRUMENT ature (∼1.5 K). The horns have a 12◦ full-width-half-
maximum beam and are pointed 30◦ from the zenith to
1University of Maryland, Code 665, NASA/GSFC, Greenbelt minimizeacceptanceofballoonandflighttrainemission.
MD20771. e-mail: dale.j.fi[email protected] Aheliumcooledflarereducescontaminationfromground
2Code665,GoddardSpaceFlightCenter,Greenbelt,MD20771 emission. No windows are used. The ambient atmo-
3JetPropulsionLaboratory, CaliforniaInstitute ofTechnology,
sphere during flight is kept from the instrument by the
4800OakGroveDrive,Pasadena,CA91109
4Columbia Astrophysics Laboratory, 550W 120th St., Mail efflux of helium gas.
Code5247,NewYork,NY10027-6902 After the switch, a GaAs high electron mobility tran-
5UniversityofCaliforniaatSantaBarbara
6WyleInformationSystems sistor (HEMT) amplifier boosts the signal. The signal
7KavliInstituteforParticleAstrophysicsandCosmology,SLAC passes then through a thermal break to a 280 K sec-
NationalAcceleratorLaboratory, MenloPark,CA94025 tion where it is further amplified and separatedinto two
8Instituto Nacional de Pesquisas Espaciais, Divisa˜o de As- sub-bands followed by diode detectors, making fourteen
trof´ısica,CaixaPostal515,12245-970-Sa˜oJos´edosCampos,SP,
channels in all. Helium pumps and heaters allow ther-
Brazil
2 Fixsen et al.
mal control of the cryogenic components which are kept 4. SKYTEMPERATUREESTIMATION
at2-3.5Kduringtheobservations. Thesignalfromeach
Conceptually, the calibration process is straightfor-
detector is demodulated with a lockin amplifier operat-
ward. The calibrator is placed over a radiometer horn,
ing synchronously with the Dicke switch. The critical
and the reference and calibrator are each warmed and
parameters of the radiometers and a full discussion of
cooled to allow the measurement of the emission cou-
the instrument are given in a companion paper (Singal
pling from each component into the radiometer. The
et al. 2008).
radiometeroutput is modeled as a linear combinationof
The sky temperature measurement depends critically
component temperatures:
on the calibrator temperature determination. The other
components (horn, switch, cold reference, and ampli- R≈A·T (1)
fier)becomeatransferstandardwhilecomparingthesky
whereT amatrixoftherelevanttemperatureswitheach
measurementstothecalibratormeasurements. Thereare
rowacomponentandeachcolumnatime. Theradiome-
24 thermometers embedded in the calibrator, from the
teroutputsareavector,R. Thesolution,A,containsthe
tips of the calibrator cones to the back of the calibrator,
couplings to the various parts of the radiometer. Since
along with 9 other thermometers on other parts of the
thelockinhasbothapositiveandnegativephasethecou-
calibrator to measure the temperature of its surround-
plings can be either positive or negative. The coupling
ings. Thesewereincludedtolookforgradientsandother
parameters include the gain and well as the emissivity.
artifactsaswellastoprovideredundancyinthecaseofa
Thecalibratoristhenmovedawaysotheradiometerob-
thermometer failure. These ruthenium oxide thermome-
servesthe sky;theparametersmeasuredwhile observing
ters (Fixsen et al. 2002) are read out at 0.9375 Hz with
the calibrator can then be used to deduce the temper-
2mKprecisionand1mKaccuracy. Identicalthermome-
ature of the sky. As the mathematics are developed, it
ters have been calibrated on separate occasions over 5
is important to remember that the essential comparison
yearswithabsolutecalibrationsstabletolessthan2mK.
is betweenthe sky andthe calibratorwhichbracketsthe
3. THEOBSERVATIONS sky temperature while the rest of the radiometer is in a
similar state.
The ARCADE 2 instrument was launched fromPales-
Themostefficientuseofthedatausesallofthecompo-
tine TX on a 29 MCF balloon 2006 Jul 22 at 1:15 UT.
nent temperature variations to obtain the best coupling
The instrument reached a float altitude of 37 km at
estimates. However, some of the variations occur while
4:41UT.Thecoverprotectingthecryogeniccomponents
the radiometer is observing the sky, which because of
was opened at 5:08 UT. The calibrator was moved 28
the beam scanning the sky includes the Galactic vari-
times from 5:30 to 8:11 providing at least 8 cycles be-
ation as well. The data could be binned by sky pixel
tween calibrator and sky for each of the radiometers.
and a solution made for each point, but that would not
During this time the entire gondola with the instrument
take advantage of the instrument variations happening
was rotated at ∼ 0.6 rpm, observing 8.4% of the entire
while observing other pixels. Instead, a Galaxy model
sky.
developed in a companion paper (Kogut et al. 2008) is
The5GHzswitchfailedinflight,sotherearenouseful
subtractedateachpixelso onlythe uniformbackground
datafromthatradiometer. The30GHzradiometerwith
sky remains. Then a general least squares fit is used to
the narrower beam is not matched to the beams of the
solve for the emissivities, gain and the sky background
other radiometers and has much higher noise than the
temperature simultaneously, including all of the sky ob-
other 30 GHz radiometer so its data are not used here.
servations.
The most useful observations were from 5:35 to
The Galaxy model is derived iteratively. For the first
7:40UT and all ofthe followingderivations will use var-
iteration the Galactic model is zero. The residual cali-
ioussubsetsofthis data. Duringthis time the calibrator
brated time series, combined with pointing data, is then
temperaturewascontrolledbetween2.2and3.1Kwitha
used to generate a map as in Figure 1 of Kogut et al.
meantemperatureof2.72K.Aselectionof25minutesof
(2008). The multi-frequency sky maps are combined to
rawdatafromtheflight(approximately10%oftheuseful
formamodelofGalacticemission,whichis thenusedto
data) are shown in Figure 1. Only one of the channels
correct the time-ordered data for subsequent iterations
for each radiometer is shown. The other channel is sim-
of the sky solution. The Galaxy model depends on the
ilar. Full scale on these plots are ∼1 K for the 3,8, and
spatial variation while the sky background temperature
10 GHz radiometers, ∼3 K for the 30 GHz radiometer
dependsonthezerolevelsotheconvergenceisvaryrapid.
and ∼2 K for the 90 GHz radiometer. While the 3 GHz
Sincethemodelhereimplicitlyassumesauniformskyno
radiometerobservesthecalibrator,the8GHzradiometer
variations are injected into the Galactic model from the
observesthe sky andthe 10,30and 90GHz radiometers
fit. However, the background sky temperature depends
“observe” the flat aluminum underside of the carousel.
on the absolute level of the Galactic model.
The 10,30 and 90 GHz radiometersobserve the sky and
calibrator together. The Galactic crossings are clearly
4.1. Galactic Foreground Subtraction
evident in the 3 GHz and 8 GHz sky data. Reference
load changes can be seen in both the sky and calibrator The Galacticforegroundis complicated. Acompanion
inthe10GHzand90GHzdata. The30GHzradiometer paper (Kogut et al. 2008) describes the detailed model
datahavemuchhigherintrinsicnoise. Afirstapproxima- producedusingtheARCADE2dataat3,8,and10GHz
tiontotheskytemperaturecanbe obtainedbyselecting along with published results from sky surveys at higher
aninterestingskydatumonthefigureandmovingacross and lower frequencies. The spatial structure of Galac-
toanappropriatecalibrationdatumthenreadingoffthe tic radio emission is modeled as a linear combination of
calibrator temperature for that datum. template maps based on the 408 MHz survey (Haslam
The Extra-Galactic Sky Temperature at 3-90 GHz 3
Fig. 1.— A subset of raw data from the flight. Full scale on these plots are ∼1 K for the 3,8, and 10 GHz radiometers, ∼3 K for the
30 GHz radiometer and ∼2 K for the 90 GHz radiometer. While the 3 GHz radiometer observes the calibrator, the 8 GHz radiometer
observes the sky etc. The 10, 30, and 90 GHz radiometers view the sky or calibrator as a group. Intermediate transients have been
suppressed. TheGalacticcrossingsareclearlyevident inthe3GHzand8GHzskydata. Referenceloadchanges canbeseeninboththe
skyandcalibratorinthe10GHzand90GHzdata.
et al. 1981) and the full sky Cii map from Fixsen et al. 4.2. Calibrator Thermometry
(1996). The offset of the template model is then ad-
Selecting which thermometer to use for the calibrator
justed to match the totalGalactic emissiontowarda set
intheleastsquaressolutionwouldbetrivialiftherewere
of reference lines of sight.
only one thermometer or all of the thermometers read
ThedistinctionbetweenGalacticandextragalacticra-
the sametemperature. Using manythermometersinthe
diation varies from author to author. We define the to-
equation allows the radiometer data themselves to se-
tal Galactic emission along selected lines of sight using
lectthebestlinearcombinationtodescribetheradiome-
twoindependenttechniques. Thefirstmethodtreatsthe
ter data. However the best differentiation of the various
Galaxy as a plane-parallel structure, and bins the tem-
thermal modes of the calibrator is data from the times
peratureofthe ARCADE 2 skymaps andlow-frequency
that the calibrator is moving or rapidly changing tem-
radio surveys by the csc|b| to determine the Galactic
perature. Unfortunately these data can not be used as
emission at the north or south galactic poles. A sec-
they are taken during the calibrator movement or when
ond technique uses atomic line emission to trace Galac-
the thermal state of the calibrator is poorly determined.
ticstructure. WecorrelatetheARCADE2orradiodata
One way out of this dilemma is to make a thermal
against the map of Cii emission to determine the ratio
model of the calibrator using all of the data, then use
of radio to line emission in the interstellar medium. We
that model during the times when the thermal and ra-
thenmultiply this ratiobythe observedCiiintensity to-
diometric state of the calibrator is best understood to
wardselectedlinesofsight(northorsouthgalacticpoles
calibrate the radiometer. This has the advantage that
plus the coldest patch in the northern sky) to estimate
the data canbe usedfor the calibratormodelevenwhen
thetotalradioemissionassociatedwiththeGalaxyalong
thecalibratorisbeingobservedbyadifferentradiometer.
each line of sight. The two techniques agree well along
A straight-forward way of generating a model of the
each independent line of sight. We then compare the
calibrator is to use principle component analysis (PCA)
templatemodeltotheestimateoftotalGalacticemission
to extract the most essential calibrator modes. The cal-
along each line of sight to derive a single offset required
ibrator has 24 thermometers embedded in the absorber
to force the template model to match the model of to-
to measure its temperature. A set of 21 of these ther-
tal Galactic emission. The three lines of sight provide
mometerswascarefullyrecalibratedaftertheflight. The
consistentestimatesfortheoffsetinthetemplatemodel,
data from these thermometers is arranged as a matrix
with scatter 5 mK at 3 GHz and less than 1 mK at 8 or
T of 21 rows of measurements each 6996 measurements
10 GHz.
long, corresponding to over 2 hours of observation. The
4 Fixsen et al.
tions of the thermometers within the cones. The total
front to back gradient is approximately 600 mK; how-
ever,sincemostofthegradientoccursneartheconetips,
97%oftheabsorbervolumeremainswithin10mKofthe
base temperature. The thermometer locations were cho-
sen using a simple static thermal model (Fixsen et al.
2006)andareconcentratedneartheconetips. Thether-
mometersareapproximatelyuniformlydistributedalong
the actual gradient so that in-flight gradients are well
sampled throughout the absorber volume.
The details of the metal surface under the calibrator
change as the calibrator moves from one position to an-
other. We observe with the calibrator in one of three
positions. While over the high frequency horns, most of
thecalibratorisoveraflataluminumplatelying1.5mm
below the cone tips. A few individual cones lie over the
high-frequencyhornantennas,withalargergapbetween
the cone tip and the aluminum wall of the horn. The
3 GHz horn is nearly the same size as the calibrator.
While over the 3 GHz horn, most of the cone tips are
Fig. 2.— The temperature within the calibrator averaged over
thedataperiodversesthedistancefromthetip(point). Thelineis ∼ 150 mm from the aluminum wall of the horn. The
thepreflightpredictionoftheshapeofthegradient. Thepreflight third position has roughly half the cone tips near the
predictionwasusedtoselecttheplacementofthethermometers. It aluminum aperture plate and the other half suspended
isnotusedintheanalysis. The21thermometersareconcentrated
above the 5 and 8 GHz horns. The third and fourth
near thetips tofullysamplethegradient. Someof thedispersion
of the measured points is due to the radial gradient which is not eigenmodesshowtheimprintofthelargechangeinther-
shownhere. mal conductivity between the cones and the 1.5 K aper-
eigenvector decomposition is ture correspondingtodifferences inthe heightofthe gas
column below each cone. The thermal imprint of the
V ·D·VT =T ·TT (2)
threepositionscanbeconsideredthreevectors. Sincethe
where D is a diagonal matrix of 21 eigenvalues and V is mean gradient has already been removed, there remain
asetof21eigenvectorssuchthatV ·VT =I. Thissetof only two dimensions. The third and fourth eigenmodes
eigenvectorsdescribesandorganizesthedataintomodes. spanthis space. Togetherthe first4eigenmodesaccount
The corresponding eigenvalue calibrates the importance for 99.98% of the temperature variance.
of each mode. Smaller modes are more difficult to identify with
One of the critical differences between these modes is knownthermalconditionsbutmayreflectchangesinhe-
the response time of each mode. The best data to dis- liumflow or changesin the aperturetemperature during
tinguish the time constants is the first few seconds after differenttimesoftheflight. Weincludethenext6modes
a thermal shock such as the movement of the carousel. to be conservative. This accounts for 99.996% of the
The firsteigenmodesexhibitthe crucialthermaldynam- variation of the thermometers. The residual is roughly
ics while the last eigenmodes are residual noise. consistentwith noise. Thus we can describe the thermal
If the calibrator were perfectly isothermal the first state of the calibrator with
(largest)eigenvaluewouldcontainallofthevarianceand ′
U =V ·T (3)
thecorrespondingeigenvectorwouldequallyweighteach
of the thermometers. In fact the largest eigenvalue has where V′ is V truncated to 10 rows.
99.9% of the variance and the weights for the various
4.3. The Solution
thermometers only varies by a ∼5% in the eigenvector.
The second largest eigenmode (0.08% of the tempera- A linear model can then be used to predict the ra-
ture variance) is a front to back gradient in the calibra- diometer output:
tor. Such a gradient was anticipated based on heat flow R=gE·X (4)
from the 2.7 K calibrator to the colder (1.5 K) aperture
where g is the responsivity of the radiometer, X is the
below. A wrap-around tank of superfluid liquid helium
matrix of ten thermal modes, U,(each row is a mode,
surrounds the back and sides of the calibrator structure
eachcolumnatime)augmentedbyrowsfortheskytem-
to intercept heat from the outside. A thermally con-
perature, the reference load temperature, the horn tem-
ductive aluminum shield lies between the tank and the
perature,theswitchtemperature,andafifthorderpoly-
absorbing cones to provide an isothermal surface at ap-
nomial. E is a vector of emissivities and R is a vector of
proximately 2.7 K. The face of the calibrator opens to
radiometerreadings. Since the radiometeris followedby
an aluminum plate maintained at the bath temperature
alockinthesignofE ispositiveforthehornandcalibra-
(1.5 K). While the calibrator does not touch the plate,
torbutnegativeforthereferenceload. Somecomponents
the diffuse (∼ 300Pa)helium gascolumncantransport
(eg. the switchorsome ofthe gradients)canhaveeither
heat from the absorbing cones to the plate. The result-
sign depending on the details of the unwanted asymme-
ing heat flow creates a thermal gradient within the cali-
tries. Since neither g nor E is known a priori, they are
brator,whichwemeasureusingthermometersembedded
combined; A=gE. A least squares fit:
within the absorbing cones.
Figure 2 compares the measured gradient to the loca- A=(X ·W ·XT)−1·XT ·W ·R (5)
The Extra-Galactic Sky Temperature at 3-90 GHz 5
produces the weighted solution to the optimum parame- in the least squares fit, no further correction is needed.
terization A, where W is a weight matrix. The solution The temperature of these elements is set by the vapor
contains the sky background temperature as well as the pressureofsuperfluidhelium. Sincethehornsarecoated
gain and other parameters of the fit. withafilmofsuperfluidheliumtheyareisothermal. The
To minimize extrapolation and possible nonlinearities smallvariationsintemperaturearedrivenbythechanges
the calibrator temperature range should match the sky in balloon altitude. The reference load temperatures for
temperature range. The reference and horn tempera- 3 and 8 GHz are low compared to the sky and calibra-
turesshouldeithermatchtheskytemperatureorremain tor, leading to a large signal in the radiometer output.
stable. Table 1 shows the mean temperatures and their Thiscanallowinstrumentgainvariationtoaffectthein-
variations for the flight times when the data are used. ferred sky signal. Gain variations are measured through
The calibrator temperature listed is the simple average the temperature variation of the calibrator. They are
of the calibrator thermometers. removed with a fifth order polynomial.
Allofthedataareweightedequallyandassumedtobe
MeanTemperaturesandRMSVariations independent,exceptsomedataareexcisedbymakingthe
Radiometer Calibrator HornAntenna Reference HEMTAmpweight, W, zero. This excising is done to eliminate the
3GHzRad 2731±134 1486±3 1987±48 1439±3 data that are obviously bad or suspect on grounds other
8GHzRad 2710±116 1414±3 1474±3 1440±3
than their position within the residual distribution. For
10GHzRad 2728±111 1470±3 2840±158 1403±3
example, during part of the flight the lower 3 GHz band
30GHzRad 2728±111 1635±379 2290±737 1436±3
shows a signal at one azimuth. This signal is not seen
90GHzRad 2724±108 2775±173 2970±349 2961±784
in any of the other channels. This is precisely the type
TABLE 1 of signal one would expect from a radar watching the
Mean temperaturesand RMSvariationsof themajor balloonwitha narrowfrequency beam. Thesedata were
components ofthe radiometers. Differentradiometers
excised. An additional 9% of the data were excised as
observed thecalibratoratdifferenttimes. All
outliers;mostofthesemeasurementsweretakennearthe
temperaturesin thetablearein milliKelvin.
edgeofthe Galaxywherethereisahighspatialgradient
and pointing errors as well as the details of the beam
shape are most important. These data have a minimal
The averagecalibratortemperatureis wellmatchedto effect on the background estimation.
the temperature of the sky, and the variation in tem- In addition to the thermometers in the calibrator,
perature of the calibrator covers the same range as the references, and horns there were thermometers on the
variation in sky temperature, thus the estimation of the HEMT amplifiers. Since the HEMT amplifiers follow
sky temperature is an interpolation rather than an ex- the Dicke switches, they cannot affect the offset of the
trapolation. This also means that the thermodynamic radiometers, but their gain can affect the output. To
temperature scale of the calibrator is carried to the sky correct for any temperature dependence in the gain, the
andno further correctionsare neededto make the result rowfor the amplifier inthe temperaturematrix contains
a thermodynamic temperature. R ∗ δTamp where δTamp = Tamp − hTamp.i The mean
Equation(5)assumesalinearmodelofradiometercou- of the temperature is removed to improve the condition
pling. Suchanassumptioniswelljustified. Thepowerat of the matrix which would otherwise have a row nearly
the detector diode for each radiometer is dominated by identical to the data being fit.
thenoisetemperatureofthecoldamplifier. Evenforthe Figure3showstheresidualsinthedataafterthemodel
radiometer with the lowestnoise temperature (8 K), the hasbeenremoved. ComparingFig.1withFig.3itcanbe
largestGalacticsignalvariationsof0.15Krepresentless seen that while the radiometer component temperatures
than 2% of the total detector loading. Furthermore, the vary by ∼ 150 mK, the residuals of the fit vary by only
mean difference in power between the calibrator obser- tens of milliKelvins. This demonstrates that the system
vations and the sky observations is less than 1%. Gain is close to linear, and that all of the major components
compressionandnon-lineareffectsarenegligibleforsuch are accounted for.
small variations.
4.4. Instrumental Foreground Estimation
As can be seen from Table 1, the mean temperatures
of the major components are near the CMB tempera- Most of the instrument was in the far sidelobes of the
ture. This minimizes the effects of reflections, unmod- antenna beams so its thermal emission to the radiome-
eled emission and responsivity variations. The cold ref- ter is negligible. However the flight train, consisting of
erence has as much impact on the radiometer signal as the parachute, ladder, FAA transmitter and balloon is
the sky or calibrator. But the sky temperature estima- directly abovethe instrument 30◦ from the center of the
tion does not depend on the absolute accuracy of the beam. Itsemissioncouldnotbeignored. Sincetheflight
reference thermometer. The reference load and the rest train is complicated and moves with the balloon rather
oftheinstrumentaremerelyatransferstandardtocom- than the gondola, a reflector constructed of aluminum
parethecalibratortothesky. Neverthelessthereference foil covered foam board was attached to the gondola to
thermometer and all of the other thermometers are read hide these components and instead reflect the sky into
out to a precision of 2 mK and have been calibrated to the radiometers. The signal from these local sources is
2 mK against an absolute NIST standard. calculated by Singal et al. (2008).
While the horns are considerably cooler than the sky, One of the principal advantages of a balloon flight is
their temperatures are very stable. Hence whatever sig- thatitputsthe instrumentaboveabout99.7%oftheat-
nals they contribute during sky measurements are re- mosphereandanevenlargerfractionofthewatervapor.
peated during the calibration, and as they are included The residual atmosphere contributes less than 1 mK to
6 Fixsen et al.
Fig. 3.— The residuals of the fit to the data from figure 1. Full scale is +/- 50 mK for all radiometers except the 30 GHz which is
+/-500mK.Exciseddataarenotshown. Skyandcalibratorobservationsarelabeled.
the10GHzchannel(Staggs,1996)andasmalleramount 50dB.The inducedsignalonthe radiometerisless than
forlowerfrequencies. Theatmosphericsignalistoosmall .1 mK. At higher frequencies the gap is a larger faction
to be seenin ourtipping scans;a correctionis made and ofawavelengthbuttheedgeismuchfurtherawaysothe
a 30% uncertainty is included in the final uncertainty neteffectisamuchsmallerleakageathigherfrequencies.
estimate.
5. UNCERTAINTYESTIMATION
Component 3L 3H 8L 8H 10L 10H 30L 30H 90L 90H
Operating the radiometer in a near null condition al-
Instrument 10.8 5.8 36.6 42.2 2.9 2.3 4.4 4.8 17.2 16.9
Atmosphere 0.7 0.7 0.7 0.7 0.7 0.7 2.2 2.2 l5o.8ws p5r.e8cise measurements to be made with greatly re-
Galaxy 23.2 19.0 2.2 1.9 1.2 1.1 .08 .07 la0xed co0nstraints on the gain, linearity and reflection of
thesystem. Itisinstructivetoimagineanidealsituation
TABLE 2 whereallofthecomponentsoftheradiometer(horn,cal-
Estimates of foregroundradiation. All estimatesarein
ibrator,switch, reference,amplifier, etc) are atthe same
mKfromSingalet al.(2008).
temperatureasthesky. Inthiscasethereisnochangein
radiometer output when switching from sky to calibra-
tor, and the gain, offset and linearity of the radiometer
Asecondadvantageofaballoonflightisitgetsthein- are irrelevant. Instrumental reflections do not matter
strument to 35 km, well above the nearest source of any since signals reflectedinto the radiometer havethe same
radio transmitters. Although the instrument was sensi- temperature as the sky. What matters in this ideal case
tivetonearbyradiotransmittersthesensitivitywassub- is only the calibrator temperature and the contributions
stantially reduced when the transmitter was below the of foregrounds.
plane of aperture. While the radio noise in a city might ARCADE2wasoperatedwithin0.1Kofideal,except
be significant the balloon trajectory explicitly avoids for the horns and the references of the 3 and 8 GHz ra-
cities and spends only a few minutes over small towns. diometers. The horns have a mean temperature 1.22 K
Other than the radarsignature in one channel we see no below the CMB temperature, but the horn temperature
evidenceforanygroundbasedradiointerference,soother is very stable. The reference temperatures of the 3 and
thanthrowing outthe apparently contaminateddata we 8 GHz are too low, requiring a polynomial fit to allow
make no correction for radio interference. for gain variation during the observations. The range of
Thehornsarecorrugatedtolimittheelectricfieldnear calibratortemperaturesmeasuredthroughoutthe obser-
the wall. The gap between the calibrator and the aper- vationsincludessignificantoverlapwiththe2.72Kofthe
ture plate is only 2% of a wavelength at 3 GHz. The CMB.
radiationleakingthroughthisgapwasmeasuredtobe∼- The overall uncertainty in the radiometric tempera-
The Extra-Galactic Sky Temperature at 3-90 GHz 7
ture is a combination of the uncertainties of the parts gas over the entire front of the calibrator so the ther-
of the model that go into the radiometric temperature mal dynamics admit only large scale gradients. Second,
estimate. Each of the uncertainties listed in Table 3 will the natural frequency of gradients is proportional to the
be discussed in turn. scale size to the inverse second power. Hence any high
spatial frequency gradients will be quickly damped. We
Source 3L 3H 8L 8H 10L 10H 30L 30H co9n0sLerv9a0tHively excise all data taken within 20 seconds of
ThermometerCal 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 a c1.a0libra1.t0or move. Third, the low frequency channels
Radiometer Cal 6.7 5.7 4.2 4.4 4.3 4.2 153 75 obs3e5rve20a.0large section of the calibrator (the 3 GHz ra-
Statistics 5.0 4.7 7.7 8.6 3.9 4.1 27.3 13.5 13.8 6.9
diometer observes the entire calibrator) so any residual
Galaxy 5.3 4.9 0.6 0.6 0.4 0.3 0.0 0.0 0.0 0.0
gradientswillbesmoothedinthelowfrequencychannels.
InstEmiss 3.2 1.7 11.0 12.7 0.9 0.7 1.3 1.4 5.2 5.1
Althoughthegradientsarewellmeasuredthecoupling
Atmosphere 0.2 0.2 0.2 0.2 0.2 0.2 0.7 0.7 1.4 1.4
to the various modes for eachradiometer must be deter-
Total 10.5 9.1 14.1 16.0 6.0 6.0 155 76.2 38 21.8
mined from the data. The radiometer noise enters into
TABLE 3 the determination of the coupling to the various modes.
Uncertaintyestimates arediscussed in §5. Uncertainties Hence the uncertainty is large for the radiometers with
areadded in quadrature. All estimatesarein mK.
high radiometer noise (eg the 30 GHz radiometer).
Estimatingthe uncertaintydue to the gradientsin the
calibrator presents a challenge. Since the gradients are
notstatictherelativeweightsofthevariousmodesinthe
5.1. Absolute Thermometer Calibration Uncertainty
fit will compensate for the gradients. To estimate the
The sky temperature can not be determined to better uncertainty in this compensation, the sky temperature
accuracythanthe absolutecalibrationofthe thermome- solution is repeated replacing the ten most significant
ters in the calibrator. The thermometer calibration was calibrator thermal modes with the temperatures of ten
tested several times before the flight. After the flight randomly selected thermometers. Solutions using 1000
21 thermometers from the calibrator, still embedded in differentrandomthermometerselectionsareusedtogen-
theircones,werecarefullycomparedtoaNISTcalibrated erate 1000 sky temperatures. The standard deviation of
thermometer over the 2.2 to 3.6 K range (Singal et al. thedistributionofderivedskytemperaturesisrelatedto
2008). The flight electronics and flight cables were used the uncertainty of the sky temperature. Since there are
inthetestandthetestwasrepeated3timestoverifythe 21 well calibrated thermometers and only ten are used
uncertainties. In addition, the lambda transition to su- to generate the trial estimates, the dispersion in the sky
perfluidheliumat2.17KwasclearlyseenwiththeNIST temperatureestimatesisanoverestimateofthe finalun-
standard thermometer in the calibration data providing certainty. On the other hand, the normal estimate of
an absolute reference. An estimate of the uncertainty the uncertainty of the mean of the 1000 selections is an
for this test is 1 mK. By using all 21 of the thermome- underestimate of the uncertainty as the samples are not
ters, the individual errorsare further suppressed, except independent. Weconservativelyestimatetheuncertainty
for the uncertaintyof the NIST calibratedthermometer, due to thermal gradients is half of the dispersion or 16
which is common to all the thermometers. times the standard uncertainty of the mean.
5.2. Temperature Gradient Uncertainty 5.3. Statistical Uncertainty
The uncertainty in the sky temperature is dominated The statisticaluncertainty is deriveddirectly fromthe
by thermal gradients in the calibrator. If the calibrator data. After the residuals are computed the χ2 is renor-
wereisothermal,itsonlycontributiontotheskytemper- malized so that the χ2/DOF is one. The ideal would
ature uncertainty would be the absolute calibration un- have the same amount of data for each radiometer, but
certainty of the embedded thermometers. Spatial gradi- the vagariesofballoonflightandrandomnoiseleave the
entsareobservedwithinthe absorbercones. The largest remaining degrees of freedom with a range of 4312 to
gradient averages 600 mK front-to-back, with the ab- 8169 with an averageof 6932.
sorber tips cooler than the back. Transverse gradients The in-flight noise is roughly in agreement with the
aremuchsmaller,withameangradientof20mK.These preflight data for the 3 GHz and 10 GHz radiometers.
gradients are not stable in time, but vary slowly with The in-flight noise is an order of magnitude higher than
scatter comparable to the mean amplitude. the preflight measurements for the 30 GHz data. In lab-
Theradiometrictemperatureofthecalibratordepends oratorytests we noted that there weremany settings for
ontheintegralofthetemperaturedistributionwithinthe which the HEMT’s would oscillate. We had attempted
absorber, weighted by the electric field at the antenna tosetthecontrolvoltagesoftheHEMTamplifiersinval-
aperture. This integralis approximatedas a linear com- leys of stability. Apparently conditions drifted in flight
bination of ten thermal modes of the calibrator. The leading to oscillations and excessive noise in the 30 GHz
time variation in the temperatures and radiometer out- radiometer. Thenoiseintheskydataandcalibratordata
put is used to derive a single time-averaged weight for are identical within their uncertainties. The 8 GHz ra-
each of the eigenmodes. The procedure is insensitive to diometer in-flight noise is higher than the preflightnoise
thermal gradientsin directions not sampled by the ther- by a factor of 4. But the in-flight noise is dominated
mometers, or on spatial scales smaller than the spacing by low frequency noise due to the gain variations on the
between thermometers. large signal due to the low reference load temperature.
Small scale gradients are not likely to be significant The in-flightnoise of the 90GHz radiometeris alsohigh
for three reasons. First, heating is spread over the en- duetodriftsinthewarmmixerandlocaloscillator. The
tire backofthe calibrator,andcoolingis done by diffuse measuredin-flight variance is statistically propagatedto
8 Fixsen et al.
the sky temperature uncertainty in table 3.
5.4. Galactic Emission Uncertainty
The extragalactic sky temperature is defined as the
residual remaining after subtracting a model of Galactic
emission. We define the absolute temperature of Galac-
tic emission along 3 independent lines of sight using the
meanofthetemperaturesderivedfromtheplane-parallel
spatialmorphologyoftheGalaxyortheobservedcorrela-
tion between radio and atomic line emission. The three
reference lines of sight provide consistent estimates for
thetotalGalacticemission,withscatter5mKat3GHz,
0.5 mK at 8 GHz, and 0.4 mK at 10 GHz.
5.5. Instrument and Atmosphere Emission Uncertainty
The emission from the reflector and the flight train
wasmodeledandmeasuredandthetwoagreewithinthe
measurementuncertainties. The carefulmeasurementof
the beam from the mouth of the antenna allows a very
complete model. The tipping tests demonstratethat the Fig. 4.— The thermodynamic temperature as a function of
frequency. ThesolidlineisthebestfittotheARCADE2datawith
modelisessentiallycorrect. Themajoruncertaintyinthe
a constant CMB temperature plus a synchrotron like component
modelistheemissivityofthealuminumfoilonthefoam. with an assumed -2.62 index. The vertical lines are ±1σ. The
Bycalibratingagainstthe measurementthis uncertainty dottedlineistheFIRASCMBtemperature.
is reduced. must remain close to the horn temperature. Its effects
The minimal contribution of the atmosphere at 37 km arelimited to changingthe dielectric surface of the horn
isfromDanese&Partridge(1989)andLiebe(1981). We near its mouth, where the electric field is largely decou-
assume a 30% uncertainty in both of these sources. pled fromthe walls. Finally, whatever minimal effect ni-
trogensnow might havewill cancelin the sky/calibrator
5.6. Instrument Drifts
comparison.
Of some concern are the possible drifts of the instru- Itis difficultfornitrogento collectonthe insideofthe
ment gain and offset in the instrument. The offset of calibrator as the back of the calibrator is sealed and the
the high frequency amplifiers is effectively canceled by front of the calibrator is closed by the aperture plane or
chopping between the reference and the sky/calibrator the horn. Bothofthesearewellbelow the freezingpoint
at 75 Hz. The gain of these amplifiers might be temper- of nitrogen so any nitrogen getting to the surface of the
ature dependent. The cold HEMT amplifier is expressly horn or aperture plane will freeze immediately and stay
checked in the model, although excluding the amplifier whereitfirstcomesintocontactwiththe apertureplane
temperature from the fit does not result in a significant or flare. The calibrator looks down so no nitrogen snow
change in the final temperature estimation. The warm or oxygen rain can fall into it.
amplifiers were cooling very slowly during the observa- Many voltage, current, and other temperatures (both
tions. Including a linear gain drift did not significantly cryogenic and ambient) were tested for correlation with
improvethefitoralterthefinaltemperature,andthusit residuals. Such a test can show some connection even if
was not included in the final fit. The helium liquid level the connection is not immediately understood. None of
changed over the course of the observations. But all of the auxiliary sensors showed any significant correlations
the components of the radiometers were well above the with the residuals.
liquid helium for the entire set of observations.
A camera showing the exposed parts of the ARCADE 6. EXTRA-GALACTICSPECTRUM
instrument at ∼2.5 K shows some nitrogen ice buildup Figure4showstheextragalacticspectrummeasuredby
on the insulators and outer parts of the ARCADE 2 in- theARCADE2instrument. Althoughthedataat10,30,
strument. Ice here has no radiometric effects. The ra- and 90 GHz are consistent with the CMB temperature
diometriccontributionfromnitrogenicecollectingonthe 2.725±0.001 K measured by the COBE/FIRAS instru-
aperture plane and flares visible in our camera is negli- ment at frequencies above 60 GHz (Fixsen & Mather,
gible as it remains at the temperature of the aperture 2002),the dataat8and3GHz showaclearexcess. The
plate or flare. excess is statistically significant, with both 3 GHz chan-
The efflux of 5 m3 s−1 of boiloff helium gas prevented nels lying more than 5 standard deviations above the
nitrogenicefromaccumulatingontheoptics. Noconden- FIRAS value.
sation is visible on the horn antennas. If small amounts TheARCADE2dataalonecannotconstrainthespec-
were to collect within the horns, it would freeze out at tral dependence of the excess signal to extrapolate to
the horn mouth since the flow of helium gas from the other frequencies. Additional data from the literature
horn impedes nitrogen flow into the interior. Any nitro- were selected to compare to the ARCADE 2 data. Al-
gen freezing on the horn will have negligible radiometric though there are many published measurements at fre-
impact since solid nitrogen has no rotational modes and quencies below 3 GHz, only a few have small enough
hence no emission lines at centimeter wavelengths. Fur- beam and sufficient sky coverage to separate the Galac-
thermore, since it is in thermal contact with the horn it tic component from the extragalactic component. We
The Extra-Galactic Sky Temperature at 3-90 GHz 9
use surveys at 22 MHZ (Roger et al. 1999), 45 MHz
(Maeda et al. 1999),408 MHz (Haslam et al. 1981),and
1420MHz(Reich&Reich1986)toestimatetheGalactic
and extragalactictemperature. As with the ARCADE 2
data,thetotalGalacticemissionisestimatedalongthree
reference lines of sight (north or south Galactic poles
plus the coldest patch in the northern Galactic hemi-
sphere) using both a csc|b| model of the plane-parallel
spatialstructureorthemeasuredcorrelationbetweenra-
dio emission from each survey and atomic line emission
tracedbytheCiisurvey. Thetwomethodsagreewellfor
the total Galactic emission along each independent line
of sight. The residual remaining after subtracting the
model Galactic emission from the measured radio emis-
sion along each line of sight constitutes an extragalac-
tic background. The scatter in this estimate from the
three independent lines of sight provides an estimate of
the uncertainty in the background temperature. Uncer-
tainties in the gain and offset for each survey contribute
additionalsources of uncertainty,which are combined in
Fig. 5.— The excess antenna temperature as a function of fre-
quadrature. Table 4 summarizes the extragalactic tem- quency. Thelineisthebestfitlinewitha-2.62index. Diamonds
perature derived from the ARCADE 2 data and lower- are low frequency points from the literature. Squares are AR-
frequency radio surveys. The largest uncertainty in the CADE 2 data. The 30 GHz data point is included in the fit but
sinceitsexcesstemperaturecomesoutnegativeitdoesnotappear
low frequency data is the gain uncertainty, but the un-
ontheplot. The90GHzerrorbarjustappears atthelowerright
certainty in table 4 also includes the offset uncertainty corneroftheplot.
and the uncertainty in the Galaxy subtraction added in 2.729±0.004K,T =1.19±0.14Kandβ =−2.62±0.04
R
quadrature. for reference frequency ν = 1 GHz with χ2 = 14.5
0
for 10 DOF. Figure 5 shows the radio background after
Frequency Temperature Uncertainty subtracting off the best-fit CMB temperature. The AR-
Source GHz K K CADE 2 data are in excellent agreement with the radio
Roger 0.022 21200 5125
backgroundderived from the low-frequency surveys.
Maeda 0.045 4355 520
Haslam 0.408 16.24 3.4 7. DISCUSSION
Reich 1.42 3.213 .53
ARCADE2 3.20 2.792 0.010
ARCADE2 3.41 2.771 0.009 DataSets TR(K) Index T0(K) χ2/DOF
ARCADE2 7.97 2.765 0.014 LF+ARC+FR 1.26±0.09 −2.60±0.04 2.725±0.001 15.5/11
ARCADE2 8.33 2.741 0.016 LF+ARC 1.26±0.09 −2.62±0.04 2.729±0.004 14.5/10
ARCADE2 9.72 2.732 0.006 LF+FR 1.44±0.41 −2.56±0.10 2.725±0.001 1.0/2
ARCADE2 10.49 2.732 0.006 ARC+FR 1.24±0.15 −2.60 2.725±0.001 14.2/8
ARCADE2 29.5 2.529 0.155 LF 1.48±0.53 −2.55±0.10 2.6±0.6 1.0/1
ARCADE2 31 2.573 0.076 ARC 1.13±0.19 −2.60 2.730±0.004 13.0/7
ARCADE2 90 2.706 0.019
TABLE 5
TABLE 4 Various combinationsof low frequencydata(LF),
Dataused inthe determination ofthe CMBandlow ARCADE2 data(ARC), andFIRAS data(FR) areused to
frequency riseestimates. TheARCADE 2final determine theradiobackgroundandthetemperatureof
measurementsarelisted herealong with their theCMB. TheFIRAS dataistreated asa singlepoint
uncertainties. Thelowfrequency measurementsare with an effectivefrequencyof 250GHz.
antennatemperaturewhiletheARCADE 2 results are
thermodynamictemperature.
The ARCADE 2 measurement of the CMB tempera-
ture is in excellent agreement with the FIRAS measure-
Inclusion of low-frequency radio surveys allows unam-
ment at higher frequencies. The double-nulled design
biguous characterization of the excess signal in the AR-
and novel open-aperture cryogenic optics demonstrate
CADE 2data. ThedatafromTable4arefitto the form
significant improvements in both calibration accuracy
T(ν)=T0+TR(ν/ν0)β (6) and control of systematic errors compared to previous
measurementsatthesefrequencies. Withonlytwohours
where T is the CMB thermodynamic temperature and
0 of balloon flight observations, ARCADE 2 approaches
T is the normalization for a radio background. The
R the absolute accuracy of long-duration space missions.
radio background is expressed in units of antenna tem-
The absolute temperature scale for ARCADE 2 is set
perature, related to the thermodynamic temperature T
by the calibration of thermometers embedded in the ex-
by
T =Tx/(ex−1), (7) ternal blackbody calibrator, and is cross-checked using
A observationsofthesuperfluidtransitioninliquidhelium.
where x = hν/kT, h is Planck’s constant, and k is The largest uncertainties in the ARCADE 2 measure-
Boltzmann’s constant. We obtain best-fit values T = ments result from thermal gradients within the black-
0
10 Fixsen et al.
body calibrator. Thesegradientsaredrivenbyheatflow uncertainties with previous measurements at 10 GHz
from the 2.7 K calibratorthrough the diffuse helium gas and 30 GHz (Staggs et al. 1996, Fixsen et al. 2004).
to the colder (1.5 K) aperture plate below. The tem- Theagreementbetweenthehigh-frequencychannelsand
peratures within the calibrator are monitored using 21 the FIRAS CMB temperature argues against any unde-
thermometers suitably spaced to uniformly sample the tected systematic errors associated with thermal gradi-
calibrator gradient. The gradient is largely confined to ents within the calibrator.
thetipsoftheabsorberconeswithinthecalibrator: 97% Further evidence against an error in the ARCADE 2
of the calibrator volume lies within 10 mK of the base data comes from comparing the ARCADE 2 data to a
temperature. Aprincipalcomponentanalysisofthether- similar analysis using independent radio surveys (Table
mometerdatademonstratesthatthethermalstateofthe 4). Repeating the CMB/radio fit from Eqn. 6 but using
calibratoratanypointintimecanbecharacterizedusing various combinations of data yields parameters in table
onlyafewmodesformedfromlinearcombinationsofthe 5. None of the combinations are in serious disagreement
thermometers. The first 4 modes are clearly related to with any of the others. The FIRAS data constrain the
the expected heat flow from the calibrator to the aper- CMB temperature. The low frequency data constrain
ture, and account for 99.98% of the thermometer vari- theindex. TheARCADE2databestconstraintheradio
ance. We conservatively model the calibrator thermal background amplitude. The ARCADE 2 data alone do
state using the first 10 modes, accounting for 99.996% not have sufficient low-frequency coverage to determine
of the thermal variance. Tests comparing the sky tem- the spectral index of the 3 GHz excess. We assume a
peraturederivedafterdroppingindividualthermometers spectralindex -2.6 and fit the ARCADE 2 data alone or
demonstrate that the calibrator has more than enough the ARCADE 2 and FIRAS data. The two independent
thermometers to adequately sense the in-flight thermal datasetsagreeonboththeCMB temperatureandradio
gradients. Infact,4or5wellplacedthermometerswould amplitude, reducing the likelihood of serious systematic
havebeensufficienttomeasurethekeythermalgradients error in either data set.
within the calibrator. The extragalactic radio background is 2–3 times as
Further improvements in the calibrator performance brightastheGalacticemissiontowardthenorthorsouth
are possible. For operational simplicity, the instrument polarcaps. Hasforegroundemissionbeenmisinterpreted
design forces the temperature of the aperture plate and as background radiation? Relativistic electrons trapped
hornantennas to the helium bath temperature by flood- in the Earth’s magnetic field emit synchrotron radia-
ingreservoirsattachedtothesestructureswithsuperfluid tion (Dyce & Nakada 1959). This emission is polarized
liquidhelium. Itwouldbe straightforwardtomodify the andanisotropic,however,withintensitypeakingnearthe
thermal design so that the aperture plate and horns are Earth’smagneticequatorandfallingtozeroatthepoles.
only weakly coupled to the bath, allowing thermal con- Further,itismeasuredtobelessthan3Kat30MHzand
trolofthese surfaces analogousto the successfulcalibra- expectedtobemuchlessthan1mKat3GHz(Ochsetal.
tor design. Even modest thermal control could reduce 1963, Peterson & Hower 1963). It therefore is unlikely
the temperature difference between the calibrator and to be a significant contaminant of our results. A spher-
the aperture, thereby reducing heat flow and associated ical halo around the Galaxy with diameter comparable
thermal gradients within the calibrator by an order of tothediskwouldnotshowupintheplane-parallelcsc|b|
magnitude or more. analysis, and would be identified with an extragalactic
The detected extragalactic radio background is background. However,theCiicorrelationanalysiswould
brighter than expected. Low frequency Galactic radia- identify any halo containing ionized carbon. The agree-
tion,andbyextensionextragalacticradiation,isthought mentofthecsc|b|andCiitechniqueswouldindicatethat
to be a mixture of synchrotron and free-free emission. any halo contribution is faint compared to emission as-
Our analysis shows the detected background to be con- sociated with the dominant plane-parallel structure. A
sistent with a single power-law with spectral index β = similar radio/atomic line correlation using a simultane-
−2.60±0.04from22MHzto10GHz. Estimatesofradio ous fit to the Cii, Hα, and Hi lines shows no significant
point sources (Windhorst et al 1993, Gervasi et al 2008) shiftin the results,nordoes infrareddustemissionshow
indicate a similar spectrum, but the radio background evidencefor asignificanthalo. Aradio-brighthalowith-
from ARCADE 2 and radio surveys is a factor of ∼ 5 out dust, hydrogen, or carbon would be very peculiar.
brighter than the estimated contribution of radio point Radio observations of edge-on spiral galaxies typically
sources. show modest high-latitude structure more typical of ra-
Itisdifficulttoreconcilethedetectedbackgroundwith dio spurs extending from the disk than a true spheri-
thecontributionfromapopulationofradiopointsources cal halo (Irwin et al. 2000, Irwin et al. 1999, Hummel
(Seiffert et al. 2008). Could the detected signal be in et al. 1991). Hα observations of edge-on spirals indi-
error? The thermal gradient in the ARCADE 2 calibra- cate that diffuse high-latitude gas contributes only 12%
tor is an obvioussourceof concernfor systematic errors. ofthe totalHα intensity (MillerandVeilleux 2003),well
However, the gradient is well sampled and the uncer- short of the factor of 2–3 required to explain the results
tainties associated with the calibrator thermal state are of this paper. Finally, we note that estimates of the
included in the ARCADE 2 uncertainties. Furthermore, Galactic/extragalactic separation along three indepen-
the bulk of the gradient is concentrated at the tips of dent lines of sight agree on the background amplitude
the cones. The skin depth for absorptionwithin the cal- within 10%, despite Galactic foregrounds that vary by
ibrator is a function of frequency: the high-frequency morethanafactorof2fromonelineofsighttoanother.
channelspreferentiallysample the tips andouter surface
of the absorber cones, while the 3 GHz channel samples
the entire absorber volume. The results agree within This research is based upon work supported by the