Table Of ContentA&A manuscript no.
(will be inserted by hand later) ASTRONOMY
AND
Your thesaurus codes are:
ASTROPHYSICS
09.07.1, 09.04.1, 12.04.2, 13.09.3, 13.18.3, 13.09.2 1.2.2008
First detection of the Warm Ionised Medium dust emission.
Implication for the Cosmic Far-Infrared Background
G. Lagache1, A. Abergel1, F. Boulanger1, F.X. D´esert2, and J.-L. Puget1
1 Institut d’AstrophysiqueSpatiale, Bˆat. 121, Universit´e Paris XI, F-91405 Orsay Cedex
2 Laboratoire d’Astrophysique,Observatoirede Grenoble, BP 53, 414 rue dela piscine, F-38041 Grenoble Cedex 9
9 Received 15 July 1998, Accepted 29 september 1998
9
9
1
Abstract. We present a new analysis of the far-IR emis- the ISM far InfraRed (far-IR) emission.
n sion at high Galactic latitude based on COBE and HI
a data. A decomposition of the Far-IR emission over the Boulanger et al. (1996) have extensively studied
J
HI, H+ and H2 Galactic gas components and the Cosmic the emission of the dust associated with the HI com-
6
Far InfraRed Background (CFIRB) is described. ponent using the spatial correlation between the far-
1 IR dust emission as measured by DIRBE and FIRAS
v For the first time the far-IR emission of dust asso- and the 21 cm HI emission as measured by the Lei-
9 ciated with the Warm Ionised Medium (WIM) is evi- den/Dwingeloo survey of the northern hemisphere. The
5 denced. This component determined on about 25% of dust emission spectrum derived from this correlation(for
0 the sky is detected at a 10σ level in the [200, 350] µm N <4.51020 cm−2) can be quite well represented by a
1 HI
band.The best representationofthe WIMdust spectrum single modified Planck curve characterized by T=17.5K
0
9 is obtained for a temperature of 29.1 K and an emissiv- and τ/NHI = 10−25 (λ/250µm)−2 cm2. This emissivity
9 itylawτ/N+ = 3.8±0.810−26 (λ/250µm)−1 cm2.With lawisveryclosetotheonepredictedbytheDraine&Lee
H
h/ a spectral index equal to 2, the emissivity law becomes (1984) dust model.
p τ/N+ = 1.0±0.210−25 (λ/250µm)−2 cm2, with a tem-
H
- perature of 20K, which is significantly higher than the Dust emission associated with molecular clouds has
o
temperatureofdustassociatedwithHIgas.Thevariation been recently studied through Far-IR and submillime-
r
t inthe dustspectrumfromthe HI to the WIMcomponent ter (submm) observations with the DIRBE, FIRAS and
s
a canbe explainedbyonly changingthe upper cutoff ofthe SPM/PRONAOS instruments. In a previous paper (La-
:
v Big Grain size distribution from 0.1 µm to 30 nm. gache et al., 1998), we have extensively studied the spa-
i The detection of IR emission of dust in the WIM signif- tial distribution of the temperature of the dust at ther-
X
icantly decreases the intensity of the CFIRB, especially mal equilibrium using the DIRBE and FIRAS experi-
r
a around 200 µm which corresponds to the peak of energy. ment. We have found at large angular scale the presence
of a cold dust component (with a median temperature
of 15 K), very well correlated with molecular complexes
with low star forming activity such as Taurus. The low-
est values of the temperature found in the cold regions
(∼ 13K) are comparable with that obtained for dense
1. Introduction clouds in star forming regions by the balloon-borne ex-
periment SPM/PRONAOS (Ristorcelli et al., 1996, 1998,
The extraction of the Cosmic Far Infrared Background
Serra et al., 1997).The associationbetween the cold dust
(CFIRB), induced by the emission of light from distant
component and molecular clouds is further demonstrated
galaxies (Partridge & Peebles, 1967; Bond et al., 1986
bythefactthatallskypixelswithsignificantcoldemission
and references therein), requires an accurate subtraction
have an excess IR emission with respect to the high lati-
of the Interstellar Medium (ISM) foreground emissions.
tude IR/HI correlation. A threshold value of the column
The two instruments DIRBE and FIRAS on board the density, N =2.51020 H cm−2, below which cold dust is
COBEsatelliteprovideactuallythebestavailabledatato HI ◦
notdetectedwithintheFIRASbeamof∼7 hasbeende-
study,onthewholesky,thedistributionandpropertiesof
duced. This knowledge on the spatial distribution of the
Send offprint requests to: G. Lagache dust associated with cold molecular clouds is important
2 First detection of the WIMdust emission. Implication for theCFIRB
for the searchof the CFIRB since it allowsto select parts tic foregrounds. First, upper limits have been reported:
of the sky for which cold dust is not detected. Hauser et al. (1995) from DIRBE data and Mather et al.
(1994)fromFIRASdata.LowerlimitsontheCFIRBhave
On the other hand, the knowledge of the dust emis- beenobtainedfromthedeepestIRASandKgalaxycounts
sion associatedwith the H+ component is very poor. Ob- (Hauser et al., 1994 and references therein). The first di-
servations of H emission at high Galactic latitudes and rect detection of the CFIRB has been reported by Puget
α
of dispersion measures in the direction of pulsars at high et al. (1996). All the Galactic foregrounds were modeled
|z| indicate that the low-density ionised gas (the Warm and removed using independant dataset in addition to
Interstellar Medium, WIM) accounts for some 30%of the the FIRAS data. Its spectrum indicates the presence of
gas in the solar neighborhood (Reynolds, 1989). There is sources at large redshift. The main uncertainty on the
also evidence that partof the WIM is spatially correlated CFIRB comes from Galactic foregrounds. Therefore, we
with the HI gas (Reynolds et al., 1995). Consequently, a stress that the Puget et al. (1996) results were confirmed
significantfractionofthe Far-IRemissionassociatedwith in the cleanest parts (NHI < 1020 cm−2 in a 7◦ beam) of
the WIM may contribute to the spectrum of the dust as- thesky(Guiderdonietal.,1997).Morerecently,Fixsenet
sociatedwith the HI gas.However,the scale heightof the al.(1998)andHauseretal.(1998)havealsoconfirmedthe
H+ medium is much larger than the HI one, so a signif- detection of the CFIRB using FIRAS and DIRBE data.
icant part of the H+ is completely uncorrelated with the
HI. Since most of the grain destruction is expected to oc- Ageneralproblemwithallthesedeterminationsofthe
cur in the low-density component of the ISM (Mc Kee CFIRB comes from a potential Far-IR emission from the
1989), the WIM could also be dust poor. Depletion stud- WIM which has never been determined. The goal of this
ies of elements that form the grains show that grains are paper is to push our knowledge of the Galactic emission
indeed partly destroyed in the low density phases of the one step forward by deriving the far-IR spectrum of the
ISM (review by Savage & Sembach,1996).Measuring the WIM dust emission. Then, we use our understanding of
dust emission from the WIM could allow to understand the interstellardust emissions associatedwith the HI and
the evolution of the dust in the low-density gas. How- H+ components to give a more accurate estimate of the
ever, this measure is difficult because one can not easily CFIRB spectrum. The paper is organised as follow: in
separate the contribution of the H+ gas from that of the Sect. 2, we present the data we have used. The variations
HI. Boulanger & Perault (1988) unsuccessfully searched of the dust emission spectrum associated with different
in the 100 µm IRAS all-sky map for such a contribu- HI gas column densities are studied in Sect. 3. In Sect. 4,
tion. The unfruitful analysis may be due to the spatial we show that the spatial variations of the dust emission
correlation between the HI and H+ emissions. Boulanger spectrum in the low HI column density regions can be
et al. (1996) have searched for such a component in the due to the presence of the non-correlatedH+ component.
residualFIRASemissionafter the removalofthe HI com- After the removal of the dust HI component, we detect
ponent. They found that the residual emission is consis- a residual Galactic emission which is attributed to the
tent withanemissionspectrumlike thatofthe HI gasfor WIM(Sect.5).ThisisthefirstdetectionoftheWIMdust
N ∼ 41019 cm−2.However,theyconsiderthisasanup- emission. We show (Sect. 6.1) that the FIRAS spectra in
HI
per limit for the contribution of the H+ component since the very low HI column density regions exhibit a large
they could have measured emission from low contrasted excess over the emission of dust associated with HI and
molecular clouds. Arendt et al. (1998) have also inves- H+ components. In these regions, the CFIBR dominates
tigated potential IR WIM dust emission. They conclude the FIRAS emission. We test its isotropy in Sect. 6.2. All
thattheywereunabletodetectanyIRemissionassociated the results are summarised in Sect. 7.
withlowdensityionisedgasathighGalacticlatitude(the
fraction of the sky used is less than 0.3%). However,very
2. Data presentation and preparation
recently, Howk & Savage (1999)have pointed out, for the
first time, the existence of Al- and Fe-bearing dust grains The FIRAS instrument is a polarising Michelson interfer-
towardstwohigh-zstars.Theyhaveshownthatthedegree ometer with 7◦ resolution and two separate bands which
ofgraindestructioninthe ionisedmedium, throughthese have a fixed spectral resolution of 0.57cm−1 (Fixsen et
two stars, is not much higher than in the warm neutral al. 1994). The low frequency band (2.2 to 20cm−1) was
medium. If dust is presentin the WIM, one shoulddetect designed to study the CMB (Cosmic Microwave Back-
its infrared emission. ground) and the high frequency band (20 to 96 cm−1)
to measurethe dustemissionspectruminthe Galaxy.We
The CFIRB is expected to have two components: di- use the so-called LLSS (Left Low Short Slow) and RHSS
rectstarlightredshiftedinthefar-IRandsubmm,andthe (Right High Short Slow) data from the ”pass 3” release
stellar radiation absorbed by dust. We concentrate here whichcoverthelowandhighfrequencybandsrespectively
on the submm part of this background. Its detection is (see the FIRAS explanatory supplement).
verydifficultbecauseofthe strongandfluctuating Galac- DIRBEisaphotometerwithtenbandscoveringtherange
First detection of the WIMdust emission. Implication for theCFIRB 3
from1.25to240 µmwith40arcminresolution(Silverberg mostly coming from the diffuse medium; in Sect. 6.2, we
et al. 1993). We choose to use annual averaged maps be- take x=3, since we need a fraction of the sky as large as
cause they have a higher signal to noise ratio than maps possible.
◦
interpolatedatthesolarelongationof90 (seetheDIRBE We also use the 240 µm/HI map excess of Reach et al.
explanatorysupplement).Inouranalysis,weonlyuse the (1998). Regions for which this excess is greater than 3σ
140 and 240 µm maps. The present study is based on are systematically discarded.
”pass 2” data. From Lagache et al. (1998), we have for each line of sight
Since in our analysis we combine the FIRAS and DIRBE the spectrum of the cold component of the dust emission
data at 140 and 240 µm, we convolve the DIRBE maps (if cold dust is detected) and the cirrus spectrum. These
with the FIRAS Point Spread Function (PSF). The PSF spectra are used to compute the contribution of the cold
is not precisely known for all wavelengths, so we use the dust emission in Sect. 3 (Table 1).
approximationsuggestedby Mather (private communica-
◦ ◦
tion) of a 7 diameter circle convolved with a line of 3
3. Emission spectrum of dust associated with HI
length perpendicular to the ecliptic plane (Mather et al.
gas
1986).
Before studying the Far-IR emission, we have subtracted In this part, we first concentrate on the spatial variations
the CMB and its dipole emission from the FIRAS data ofthedustemissionspectrumwiththeHIgascolumnden-
using the parameters given by Mather et al. (1994) and sities.We deduceacolumndensitythresholdabovewhich
Fixsen et al. (1994). To remove the InterPlanetary (IP) thecontributionofthecolddustcomponentinducesasig-
dust emission, we first consider the 25 µm map as a spa- nificant submm excess with respect to the ν2 emissivity
tial template for the IP dust. Then, we compute the IP law. Then, we investigate the spectrum of the dust asso-
dustemissionat100 µmusingthezodiacalemissionratio ciated with regions containing HI column densities below
given by Boulanger et al. (1996): Iν(100)/Iν(25)=0.167. this threshold.
We remove the IP dust emission at λ≥140µm using the
IP emission template at 100µm and considering a zodia-
cal spectrum I ∝ν3 (Reach et al., 1995).
ν
3.1. Variation of the dust emission spectrum with the HI
column densities
The HI data we used are those of the Lei-
den/Dwingeloo survey, which covers the entire sky down
Tocomputetheemissionspectrumofdustassociatedwith
◦
to δ = −30 with a grid spacing of 30’ in both l and b
HIgasfromlowtolargecolumndensities,weusethesame
(Hartmann & Burton, 1997). The 36’ Half Power Beam
method as in Boulanger et al. (1996). We first select sky
Width (HPBW) of the Dwingeloo 25-m telescope pro-
pixelsaccordingtotheirHIcolumndensityandsortthem
vides 21-cm maps at an angular resolution which closely
into 10 sets of pixels bracketed by the following values
matches that of the DIRBE maps. These data represent
of N : [1.1, 2.7, 3.7, 4.6, 5.5, 7.3, 9.1, 10.9, 12.7, 16.4,
HI
an improvement over earlier large scale surveys by an or- 20.1]×1020 Hcm−2 which correspond to W : [60, 150,
HI
der of magnitude or more in at least one of the principal 200, 250, 400, 500, 600, 700, 900, 1100]K kms−1. These
parametersofsensitivity,spatialcoverage,orspectralres-
different sets cover between 1.2% (for the highest column
olution. Details of the observationnal and correction pro-
density) and 16% (for the lowest) of the sky.
cedures are given by Hartmann (1994) and by Hartmann
Dust emission spectra are computed for each set k using
& Burton (1997). The 21cm-HI data are convolved with
the equation:
the FIRAS PSF. We derive the HI column densities with
1 Kkms−1=1.821018 Hcm−2 (optically thin emission). dF <F > −<F >
k l
(k,l)= (1)
dN <N > −<N >
HI HI k HI l
Throughoutthis paper,diffuse parts ofthe skyare se-
lected following Lagache et al. (1998). To remove molec- where < F > corresponds to the mean FIRAS spectra
i
ular clouds and HII regions, we use the DIRBE map of computed for the set of pixels i, and <N > the mean
HI i
the 240 µm excess with respect to the 60µm emission: HI column density for the same setof pixels.Forming the
∆S=S (240)-C×S (60) with C=4±0.7. This map shows differenceinEq.1removes,withinstatisticalvariance,any
ν ν
as positive flux regions, the cold component of the dust residual IR emission which is not correlated with the HI
emission, and as negative flux regions, regions where the gas such as an isotropic component.
dust is locally heated by nearby stars (like the HII re- In this part, < F > correspond to the FIRAS spectra
k
gions).Therefore,diffuse emissionpixelsareselectedwith computedforthe10setsofpixelsgivenaboveand<F >
l
|∆S| < xσ, σ being evaluated from the width of the his- tothespectrumderivedintheverylowHIcolumndensity
togram of ∆S and x being chosen for our different pur- regions, N ≤ 1.11020 Hcm−2 (representing ∼ 2% of
HI
poses. For example, in Sect. 5, we take x = 1, which the sky).
is very restrictive, to ensure that the selected pixels are
4 First detection of the WIMdust emission. Implication for theCFIRB
Table1.MeanNHI computedfordifferentsetsofpixels(seeSect.3.1).Opticaldepths(τ)andtemperatures(T)oftheHIdust
emissionspectra(derivedfollowingEq.1).Residualfluxintegratedinthe[609-981] µmbandareobtainedbyremovingasingle
componentν2 modifiedPlanckcurve(definedbyτ andT) from eachHIdustspectrum.Thefifthcolumn givesthepercentage,
in flux in the[609-981] µm band,of thedetected cold emission with respect tothecirrus emission for thecorresponding set of
pixels (see Lagache et al., 1998 for more details).
Mean NHI 10−6×τ T (K) Residual flux % of detected
(H cm−2) (normalized at 250 µm) (10−10 W m−2 sr−1) cold emission
for NHI =1020 H cm−2
1.8 1020 6.8 ± 1.8 18.0±1.2 -1.0 ± 8.6 0.0
3.2 1020 9.8 ± 1.4 17.3±0.6 3.0 ±3.7 0.0
4.1 1020 9.7 ± 1.1 17.7±0.5 4.0 ± 2.7 0.9
5.0 1020 10.2 ± 1.0 17.6±0.4 4.4 ± 2.2 0.6
6.3 1020 11.3 ± 0.9 17.5±0.3 6.4 ± 1.6 3.4
8.1 1020 12.1 ± 0.8 17.8±0.3 9.5 ± 1.3 6.6
9.9 1020 13.4 ± 0.7 17.9±0.2 10.9 ± 1.1 7.1
1.2 1021 17.1 ± 0.6 18.0±0.2 13.1 ± 1.0 11.9
1.4 1021 16.3 ± 0.6 18.0±0.2 13.9 ± 0.8 12.3
1.8 1021 17.1 ± 0.6 17.6±0.2 13.9 ± 0.7 8.0
For each set of pixels k, the dust spectrum dF (k) is
dNHI
fitted by a modified Planck curve:
−2
νI = τ λ B (T),
ν (cid:16) 250 µm (cid:17) ν
where τ is the optical depth at 250 µm, B (T) is the
ν
Planck curve (in W/m2/sr) and λ the wavelengthin µm.
The α=2 emissivity index corresponds to standard in-
terstellar grains (Draine & Lee, 1984). The mean tem-
perature is equal to 17.7K with no significant variations
from set to set (Table 1, column 3). The optical depth
(τ) increases linearly with the HI column density up to
N =1021 Hcm−2. Then, it levels off at ∼1.710−5 (Ta-
HI
ble 1, column 2).
We compute the residual spectrum for each set:
ν R = dF − νI ,
ν dNHI ν
andstudyitbyintegratingνR inthe[609-981] µmband.
ν
Results are presented column 4 of Table 1 and Fig. 1.
For N larger than 5 1020 Hcm−2, we detect a signif- Fig. 1. Residual flux integrated in the [609-981] µm band
HI
icant residual emission (> 3σ). Spectra at high column afterremovingasinglecomponentmodifiedPlanckcurvewith
density contain a positive submm excess with respect to α=2 from dust spectra normalised at NHI=1020 Hcm−2 and
a single temperature ν2 emissivity law Planck curve for computed for different HI column densities.
λ > 600µm. In Lagache et al. (1998), we have shown,
using DIRBE data (at the FIRAS resolution), that cold
emission from interstellar dust associated with H2 gas is tic dust (T∼15K). We can note that the residual flux
only detected where HI column densities are larger than as well as the optical depth become nearly constant for
∼ 2.51020 Hcm−2. In the Galaxy, at |b| > 10◦, 90% of NHI ≥1.21021 H cm−2 since the contributionofthe cold
this cold emission is distributed over regions with HI col- emission does not increase anymore.
umn densities larger than 51020 Hcm−2. Column 5 of
Table 1 gives the percentage, in flux in the [609-981] µm In conclusion, in the diffuse medium, the spectrum of
band, of detected cold emission for the different sets, de- thedustassociatedwiththeHIgascanbeaccuratlyrepre-
rivedfromtheanalysisofLagacheetal.(1998).Weclearly sentedbyamodifiedPlanckcurvewithaν2emissivitylaw
see that the submm excess increases with the contribu- and a single temperature (the residual emissionis around
tion of the cold emission. Therefore, we conclude that 0,seeTable1)asitwasalreadyshownbyBoulangeretal.
this positive excess is due to the emission of cold Galac- (1996). For N larger than 5 1020 Hcm−2 (A ≥0.25),
HI v
First detection of the WIMdust emission. Implication for theCFIRB 5
the cold dust (T∼15K) induces a significant submm ex- dispersion measures in the direction of high altitude pul-
cess with respect to the ν2 emissivity law. sars,most ofwhich located at|b|<10◦ (Reynolds, 1989).
However,thesemeasures,whichgiveN /N ∼ 30%,
H+ HI
do not necessarily apply to the local ISM. A more di-
rect measurement of the WIM at high Galactic lati-
3.2. Dust emission spectrum in low HI column density re-
tude comes from Jahoda et al. (1990). They have mea-
gions
sured in Ursa Major the HI emission through the 21 cm
In order to assess the reliability of the spectrum associ- line and the Hα intensity. They have obtained NHI =
ated with very diffuse regions, we explore now in detail, 0.71020 Hcm−2 and Hα = 0.2R. The emission measure
the variations of the HI spectrum in the low HI column may be convertedinto column density assuminga certain
density medium, where the contribution of the cold dust electron temperature and density. For Te = 8000K and
to the total emission is negligible (NHI≤4.51020 H cm−2 ne = 0.08cm−3 (Reynolds, 1991), we find for Ursa Ma-
i.e WHI≤250Kkms−1, see Table 1). This threshold is jor the same NH+/NHI ratio (∼ 30%) as with dispersion
the same as in Boulanger et al. (1996). We construct the measures. In the following, we thus consider that this ra-
spectrum of dust associated with HI gas following Eq. 1, tiodoappliedtothehighlatitudesky.These30%contain
forvariouscouples(l,k)ofsmallerHIsetsofpixels(Table the correlated (NcH+) and the non-correlated (NncH+)
2). H+ components (by correlated, we mean spatially corre-
lated with the HI gas).
Fig.2ashowsthat the spectralshapeofthe fourspec-
tra is relatively constant but the absolute level varies by The total hydrogen column density is defined as:
a factor ∼1.6 between the two extremes. The two spec- NH = NHI + NcH+ + NncH+.
tra obtained in regions where N < 2.51020 H cm−2 We neglect the contribution of the molecular hydro-
HI
(W < 140Kkms−1) show a significant variation of gen since (1) it represents less than 1% of the emis-
HI
about 30%. This variation cannot be explained by the sion in regions where WHI < 250Kkms−1 i.e.
contribution of the cold dust since (1) it is not detected NHI <4.51020 Hcm−2 (Tables1 and2)and(2) the frac-
intheseverylowHIcolumndensityregions(Table2)and tion of H2 gas is less than 1% in the very diffuse medium
(2) the residual emission (ν R , see Sect. 3.1 for the defi- (Bohlin et al., 1978). The correlated H+ component can
ν
nition)isaroundzero.Moreover,colddustwouldproduce be written: NcH+ = a × NHI. Since our knowledge on
an excess at long wavelengthrather than a constant vari- the non-correlatedH+ componentis verypoor,wechoose
ation on the whole spectra. In the next section, we show to simulate it using two assumptions:
someevidencethatthese variationsofthe diffusemedium - (1) It is distributed in the sky like the 240 µm emission
spectracanbeduetodustassociatedwiththeionisedgas, atthesamelongitudebutoppositelatitude.Thisassump-
more specifically from small scale structures in the WIM tionisarbitrarybutensuresthatthenon-correlatedWIM
uncorrelated with the distribution of HI gas. gas has the same scale properties and dependance with
the Galactic latitude as the diffuse interstellar medium.
We note that it is essential to use, for Nnc , a map
H+
with small scale structures to ensure the non-correlation
withtheHI componentatsmallscales.Thismapiscalled
Table 2. HI bins for the set of pixels l and k ˜I240.
- (2) The non-correlated part of the WIM has an IR
WHI (l) WHI (k)
K km s−1 K km s−1 emissivity per hydrogen equal to that of the HI gas,
τ/N = 10−25 (λ/250µm)−2 (Boulanger et al., 1996).
[0,50] [50,250] HI
[0,90] [90,140] Computing τ, defined as ˜I240/B240(T = 17.5K), gives
[0,50] [50,140] the corresponding column density.
[0,140] [140,250]
The map obtained with assumptions (1) and (2) is
written N˜ . Thus, the uncorrelated H+ component is
HI
represented by: Nnc = b×N˜ . The constraint on a
H+ HI
and b is: a+b=30%.
4. Effects induced by the diffuse ionized gas on
Dust spectra are computed following Eq. 1, using
the Far-IR/HI correlation
< N > rather than < N > , for the same HI sets
H l,k HI l,k
To simulate the potential effect of the WIM Far-IR emis- as in the previous section (Table 2) and for different val-
sionontheFar-IR/HIcorrelation,wehavefirsttoevaluate ues of a and b with a step equal to 1% and the constraint
thehydrogencolumndensityoftheWIM.Theonlydirect a+b=30%.Wequantifythedispersionbetweenthespec-
determination of the WIM column density comes from tra by computing the total rms of the difference between
6 First detection of the WIMdust emission. Implication for theCFIRB
Fig.2.(a)SpectraofdustassociatedwithHIgas(normalisedatNHI=10 20 Hcm−2)computedfollowingtheEq.1fordifferent
setsofpixels(Table2).(b)Spectraofthedustcomputedinthesamewayas(a)andforthesamesets.Thenormalisation isno
morebyHIatombutbythetotalhydrogencolumndensitydefinedasNH = NHI + NcH+ + NncH+ withNncH+ =b ×N˜HI,
NcH+ =a ×NHI,a=0.09 andb=0.21 (seeSect.4).ThedottedlinecorrespondstotheWHI setsofpixels[0, 140]and[140,
250], the continuous line to [0, 90] and [90, 140], the dashed line to [0, 50] and [50, 140] and the dashed-dotted line to [0, 50]
and [50, 250] K kms−1. Spectra have been truncated below 135 µm because of the poor signal to noise ratio. They havebeen
smoothed to a resolution of 4 cm−1.
the four spectra and the mean spectrum, in the wave- 5. Emission spectrum of the diffuse H+ compo-
length range [200, 300] µm. Fig. 2b shows the result for nent
the minimal dispersion, obtained with b∼21%. We see no
5.1. Detection
moredifferencesbetweentheabsolutelevelsofthespectra.
They are all stabilized on the mean spectrum. The same The IR emission from any dust associated with the WIM
conclusionis reachedwithotherHI sets ofpixelasforex- would follow a csc(b) variation, like that from any diffuse
ample: l=[0, 50] and k=[50, 190], l=[0, 190] and k=[190, component in the Galactic disk. Based on such an hy-
250], l=[0, 70] and k=[70, 160], l=[0, 100] and k=[100, pothesis,Boulangeretal.(1996)havefoundintheFIRAS
180]Kkms−1.This simulationis only illustrativebut we dataaresidualGalacticemission(aftertheremovalofthe
think thatit has a quantitativesignificanceprovidedthat HI component) consistent with an emission spectrum like
the spatial distribution of the uncorrelated WIM shares that of the HI gas for N ∼4 1019 Hcm−2. They con-
HI
the same morphological properties (ratio between small siderthis asanupper limitforthe contributionofthe H+
scalestructuresandlargescaleGalacticgradient)thanour componentsincethey couldhavemeasuredemissionfrom
“DIRBE”template.Thisisdemonstratedbythefactthat low contrasted molecular clouds.
the optimalb value is very stable whenwe rotatethe H+
nc
template with the Galactic longitude as long as this tem- To searchfor any residualdiffuse component after the
plate stays uncorrelated with the HI component at small removaloftheHIcontribution,weuseallskyFIRASmaps
scales. This simulation leads to the following proportion at|b|>20◦.Wekeeponlypixelsforwhich|∆S|<1σ(see
of uncorrelated and correlated H+ gas of 70% and 30% Sect. 2). This criterium is over restrictive (we keep only
which are in agreement with the Reynolds et al. (1995) 25.6%ofthesky)butensurethattheremainingpixelsare
determination in a small region. mostly coming from the diffuse medium.
Theemissivityofthespectrumofdustassociatedwith ForeachselectedFIRASspectrumF (i,j),weremove
ν
the HI gas(including theH+ component)thatwededuce the HI related emission following the equation:
c
is:
∆F (i,j)=F (i,j)−N (i,j)×I (ν) (3)
ν ν HI HI
τ/N =8.7±0.9 10−26(λ/250µm)−2 cm2 (2)
HI
whereN theHIcolumndensityandI (ν)iscomputed
HI HI
with a temperature of 17.5 K. fromνI (ν)= τ/N × B (17.5K)whereτ/N isthe
HI HI ν HI
This emissivity value is compatible with the one de- HI dust emissivity givenEq.2 andB (17.5K)the Planck
ν
rived in Boulanger et al. (1996). curve at 17.5K in Wm−2 sr−1.
First detection of the WIMdust emission. Implication for theCFIRB 7
We first search for any residual emission by correlat-
ing the Galactic latitude emission profiles of ∆F at each
ν Table3.Cosecantslopes(ν Dν)oftheresidualGalacticemis-
FIRAS wavelength, with the mean latitude emission pro-
sion averaged in the wavelength band [200,350] µm and com-
filecomputedinthewavelengthrange[200-350] µm.This
puted in different Galactic longitude areas.
correlation method is probably more accurate because it
Galactic longitude Residual Galactic emission
avoids the problems with the cosecant method linked to
range (in ◦) (109 W m−2 sr−1 )
thelocalbubble.Thespectrumobtainedwiththesecorre-
0 - 85 9.4±0.5
lationsis presentedonFig.3.This emissionisclearlynon
85 - 170 6.4±0.4
zero and is detected at a 10σ level in the [200-350] µm 170 - 235 9.9±0.6
band. The spectrum can be fitted by a modified Planck 235 - 360 10.7±0.7
curve with a να emissivity law. The best fit is obtained
forα=1andT=29.1K.Thevalueα=2is alsocompatible
with the data. In that case, we have T=20.0K, which is
significantly higher than the temperature of dust associ- 5.2. Normalisation and emissivity
ated with HI gas.
To normalise the WIM far-IR spectrum, we make use
To check the Galactic origin of this residual emission, of the cosecant law measured for Hα emission at high
we fit the latitude profiles computed over the selected re- Galactic latitude. The slope of the Hα cosecant law is
gions and for each FIRAS wavelengthwith cosecant vari- egal to 1 R (Reynolds et al., 1984). This gives (follow-
ations: ing the results obtained in Sect. 4: 70% of 1R) 0.7 R
for the non-correlated H+ gas. Using a conversion factor
∆Fν(|b|)∝Dνcosec|b| (4) I (R) = 1.15 × N+/1020 cm−2 (derived from Oster-
Hα H
brock, 1989), we obtain Nnc = 6.11019 cm−2. This
The slopes Dν measured at each wavelengthform a spec- H+
column density is close to that derive from the pulsar ob-
trum which has the same temperature and shape as the
servations (Reynolds, 1989).
spectrum obtained with the correlation (Fig. 3). There-
Using this normalisation, the emissivity per H+ ions is:
fore,we conclude that the spectrum derivedwith the cor-
τ/N+ = 3.8±0.810−26 (λ/250µm)−1 cm2. The uncer-
relation has a Galactic origin. H
tainty correspondsto the 21 % dispersion observedin the
different longitude bins (Table 3). With a ν2 emissivity
The ISM is known to contain structures over a very
law,wefindτ/N = 1.0±0.210−25 (λ/250µm)−2 cm2.
wide range of scales with inhomogeneous physical con- H+
◦ This emissivity value agrees within error bars with the
ditions. Within the large FIRAS beam (7 ), small scale
emissivity of dust associated with HI gas.
molecular clouds or HII regions can be diluted and thus
notdetectableusingourcriteria.Inordertocheckwhether
suchsmallscalestructurescontaminatethe residualspec- We have not taken into account until now of
trum of Fig. 3, we use the same two methods (correlation the uncertainty on the HI dust emission spectrum.
andcosecantvariations)to derivethe spectrum using dif- The removal of the HI maximal spectrum, defined by
ferent cuts for the selection of the pixels (either more or τ/NH+ = 9.6 10−25 (λ/250µm)−2 (see Eq. 2) and
less restrictive than above). The spectra obtained with T=17.5K,only decreasesthe emissivity ofthe WIM dust
correlations or cosecant laws are very stable. This sup- component by 30%; the removal of the minimal spec-
ports the idea that the residual Galactic emission cannot trum, defined by τ/NH+ = 7.810−26 (λ/250µm)−2 and
be due to low contrastedmolecular clouds or HII regions. T=17.5 K, increases the emissivity by 24%.
Moreover, to check whether this residual Galactic emis- The spectrum derived from the slopes of the cosecant
sion is not dominated by a particular region of the sky, law (which measures the Galactic emission in a direc-
werepeatthesamecosecantfitanalysisin4binsoflongi- tionperpendiculartotheGalacticdiskoverasignificantly
tudeofroughlyequalsize(6%oftheskyeach).Thesmall largearea)correspondsto thesubmmdustemissionofan
numberofpixelsdoesnotallowtohaveabettersampling. HI equivalent column density of 1020 cm−2 (this value is
We see, in Table 3, that the residual Galactic component estimatedusingD andI (ν)),whichishigherthanthe
ν HI
is present in the 4 different bins with small variations. tentative detection of Boulanger et al. (1996) due to the
Therefore, we can exclude this residual Galactic compo- difference in the HI dust spectrum used for the normali-
nent to come from a particular region of the sky. sation. Since this column density of 1020 cm−2 is roughly
the Reynolds’s estimate of the total H+ column density,
The residual Galactic spectrum corresponds to mate- ourresultisconsistentwithadustabundanceintheWIM
rial which is uncorrelated with the HI gas (by construc- which does not differ much from that of the HI gas. We
tion). However, it is Galactic because it follows a cose- reach the same conclusion using the minimal or maxi-
cantlaw.Weproposethatthis isthe firstdetectionofthe mal HI spectrum. However, this conclusion is still pre-
submm emission of dust associated with the WIM. liminarysince the Reynolds’sestimateis madeinpartsof
8 First detection of the WIMdust emission. Implication for theCFIRB
Fig. 3. Correlations slopes of the residual Galactic emission Fig. 4. Upward curves: HI dust emission spectrum (dotted
line), HI synthetic spectrum (corresponding to the emissivity
(dotted line) after the main HI component has been removed
andtemperaturegiveninSect.4,dashedline)andHIemission
(slopes have been computed using the residual Galactic emis-
sion in Iν).Thedash-dottedlinerepresentsamodified Planck model(square)for NHI = 1020 cm−2.Downwardcurves:H+
curve with α=1, T=29.1 K and the dot-dot-dot-dashed line a dust emission spectrum (dotted line, divided by 1.5 103), H+
modified Planck curvewith α=2 and T=20 K.This spectrum synthetic spectrum (corresponding to the emissivity and tem-
peraturegiveninSect.5,dashedline)andH+ emission model
is the first detection of the WIMdust emission
(triangle) for NH+ = 1020 cm−2.
the sky which are different from those used here. Future
comparisonof the Far-IR dust emission with the Wiscon- amax=30nm, which is compatible with what is predicted
sin H-Alpha Mapper survey (WHAM, Tufte et al., 1996; inJonesetal.,1996)isneededtoreproducethe spectrum
Reynolds et al., 1998) will allow to give a more precise of the WIM (Fig. 4, downward curve). This comparative
estimate on the dust abundance in the WIM. analysis suggests that the difference of temperature be-
tween the HI and H+ diffuse media can be explained by
the erosion of the dust grains due to the collisions.
5.3. First interpretation
6. The extragalactic component
It is extremely difficult to disentangle whether the WIM
spectrumfollowsaν2 oraν1 emissivitylawbecauseofthe 6.1. FIRAS spectra in low HI column density regions: ev-
poor signal to noise ratio of our spectrum at long wave- idence for an extragalactic component
length (Fig. 3). However, this spectrum has a definitively
Pugetetal.(1996)havefoundinthe residualemissionaf-
higher temperature than the spectrum of dust associated
terthe removalofthe HI correlatedemission,anisotropic
with HI gas.
componentwhichcouldbe the CosmicFarInfraredBack-
DustgrainsintheWIMareexpectedtohavesmallersizes ◦
ground (CFIRB) due to distant Galaxies. At |b| > 40 ,
due to grain shattering in grain-grain collisions (Jones et
at least one third of the emission at 500 µm comes from
al., 1996, Fig. 17). We have used the dust model devel-
this isotropic component. Thus, at high Galactic latitude
opped by D´esert et al. (1990) to test the effect of the
in very low HI column density regions, this component
size distribution on the Far-IR spectrum. First, a stan-
should dominate the FIRAS emission.
dard Big Grains compositionand size distribution (abun- We have selected pixels for which N <9.11019 Hcm−2
dance in mass m/mH=5 10−3 1, silicates with dark re- and b>40◦ (33 pixels). These pixelsHaIre located near and
fractory mantle composition, distribution in size follow-
in the Lockman Hole and represent 0.54% of the sky. For
ing a 2.9powerlawwitha =15nmanda =110nm,
min max this set of pixels, the mean HI column density is equal to
andadensityof3g/cm3)remarquablyreproducethedust 8 1019 H cm−2. In this region,the non-correlatedH+ col-
emissionspectrumassociatedwith the HI gas(Fig.4,up-
umn densityis estimatedusingN /N =30%(Jahoda
H+ HI
wardcurve).Asmallercutoffinthesizedistribution(with
etal.,1990)whichgivesNnc /N =21%(seeSect.4).
H+ HI
1 This abundance in mass is lower than the one given in Weclearlysee(Fig.5)thattheemissionofdustassociated
Desertetal.,1990duetothedifferenceofcalibration between with the HI and H+ components cannot account for the
theIRAS and DIRBE/FIRASinstruments FIRAS spectrum, especially at λ>300 µm. The presence
First detection of the WIMdust emission. Implication for theCFIRB 9
ofanothercomponentis evident.The levelofthis compo- the dispersions measured for the selected pixels, uncer-
nentis the sameasthe residualFIRAS emissionobtained tainties on the HI dust emissions are given by Eq. 2, and
in Puget et al. (1996) before they subtract the estimated uncertainties on the WIM dust emissions correspond to
H+ and is in very good agreement with the Fixsen et the 21% dispersion observed in the four bins of Galactic
al. (1998)determinations. Pugetet al.(1996)have shown longitude (Table 3). Our residual emissions (Res2 in Ta-
that the most likely interpretation for this component is ble 4) are significantly lower than the ones reported in
the Cosmic Far-InfraredBackground(CFIRB) due to the Hauser et al. (1998) and Schlegel et al. (1997) since they
integrated light of distant galaxies. Despite the fact that haveneglectedthecontributionofthedustassociatedwith
itisnotphysicaltoconsiderauniquetemperatureforthis the WIM. We clearly see that it is essential to take into
background,itcanbeanalyticallyrepresentedinthesame account the dust emission associated to the WIM below
way as in Fixsen et al.(1998).The best fit, valid between 240 µm. In FIRAS data at 240 µm, the CFIRB emission
200 and 2000 µm, (Fig.5)isveryclosetoourDIRBEvalue.At140 µm,the
I(ν)=8.80×10−5(ν/ν0)1.4Bν(13.6K) (5) csiobmlepdaruiesotnobthetewceoennsiDdIeRraBbEleainncdreFaIsReAoSf tdhaetaFIiRs AnoStdpaotsa-
where ν0=100cm−1,is presentedonFig.6,togetherwith noise (that is why we have prefered to cut the spectrum
the range of determination of Fixsen et al. (1998). The at 200 µm in Fig. 5).
uncertainties on the fit will be discussed in a forthcoming
paper (Gispert et al., 1999).
6.2. Isotropy
We test in this part the isotropy of the CFIRB on large
scalessincewedonotknowthe spatialdistributionofthe
WIM dust emission at small angular scales. This test at
large scales is important to detect potential systematic
effects caused by an inacurate subtraction of ISM dust
emissions. As expected from our detection of the WIM
dust emission, we show that the CFIRB is isotropic only
ifweconsidertheemissionofdustassociatedwiththeH+
component.
Fig. 5. Mean FIRAS (continuous line), HI (dash-dot)
and H+ (dash-dot-dot-dot) spectra in regions where
NHI <9.1 1019 Hcm−2. The residual emission (dot line),
obtained by removing from the mean FIRAS spectrum
the HI and H+ components, dominates the FIRAS emis-
sion at high latitude in low HI column density regions.
The typical uncertainties for this residual emission are
around 2.6 10−9 W m−2 sr−1 for 200 < λ < 400 µm,
1.75 10−9 W m−2 sr−1 for 400 < λ < 600 µm, and
2.5 10−10 W m−2 sr−1 for λ > 600µm. Also reported are our
determinations of the DIRBE CFIRB at 140 and 240 µm
(cross points, see Table 4).
Fig. 6. CFIRB emission computed on 51% (dashed line) and
on 0.54% of the sky (dotted line). The continuous and dot-
TheCFIRBcanalsobedeterminedat240and140 µm ted-dashedlinescorrespondtotheanalyticformsoftheCFIRB
using DIRBE data for the same part of the sky. Results given by Eq. 5 and Fixsen et al. (1998) respectively. Cross
are presented in Table 4 together with previous determi- points correspond to our DIRBE CFIRB values at 140 and
nations.InthisTable,uncertaintiesontotalemissionsare 240 µm (Res2 of Table 4).
10 First detection of the WIMdust emission. Implication for theCFIRB
Table4.ContributionofthedifferentcomponentstotheDIRBE140and240 µmemissionsintheLockmanHoleregionwhere
themean HI column density is 81019 H cm−2. Res2 corresponds to theCFIRB values.
Component 140µm 240µm
108 W m−2 sr−1 108 W m−2 sr−1
Total 4.53±0.57 2.40±0.15
HI+H+c 1.94±0.19 0.91±0.09
Res1=Total - (HI + H+c) 2.59±0.60 1.49±0.17
H+nc 1.06±0.22 0.35±0.07
Res2=Total - (HI + H+c) - H+nc 1.53±0.64 1.14±0.19
Residual 1 2.50±0.70 1.36±0.25
Residual 2 3.19±0.43 1.67±0.17
1 Hauseret al., 1998
2 Schlegel et al., 1998
Thus, we have to compute the CFIRB spectrum on a
fraction of the sky as large as possible. For that, we pre-
fer to use a Galactic template based on far-IR emissions
rather than HI gas, (1) to take into account variations
in the dust temperature from place to place and (2) to
avoidthelargeholeinthesouthernhemisphereoftheLei-
den/Dwingeloo survey. We combine the DIRBE 140 and
240 µm data with FIRAS spectra. First, we extract from
◦
DIRBE data (at 7 resolution) a Galactic emission tem-
plate.Then,theDIRBEGalacticemissionisextrapolated
tolongerwavelengthsandsubtractedfromeachindividual
selected FIRAS spectrum to derive the CFIRB emission
and test its isotropy.
Thetwo(140and240 µm)GalacticDIRBEtemplates,
I ,arecomputedbyremovingthe CFIRB (Table 4)from
G
theDIRBEemissions.Assumingaν2 emissivitylaw,tem-
perature and optical depth of each pixel are estimated
using the two DIRBE Galactic templates. The residual
FIRAS emission is computed in the following way. First, Fig. 7. Variation of the FIRAS residual emission averaged in
we discard pixels located in known molecular clouds or the[300,609] µmbandwiththeGalacticlongitude.Theerror
HII regions.We keeppixels with |∆S|<3σ andworkat bars are statistical errors (1σ).
◦
|b| > 15 (51% of the sky). Then, for each selected pixel
with its associated temperature, we derive the ratio R
ν
between the modified Planck curve computed at each FI- (in the [300, 609] µm band) in 8 different longitude bins
RASwavelengthandthemodifiedPlanckcurvecomputed (each bin representing around 3.5% of the sky). The pro-
at 240 µm. Finally, the residual emission is computed in file, shown in Fig. 7, does not show any particular Galac-
the following way: tic structure. To test the isotropy with the Galactic lati-
tude, we compute the emission profile versus the latitude
Res (i,j)=F (i,j)−R ×I (i,j) (6)
ν ν ν G (cosecantvariation)ofourresidualemission(Fig.8a).We
clearly see a residual Galactic component. The slope of
The mean residual emission spectrum is shown Fig. 6 to-
the fitted cosecantlaw is ∼ 710−10 W m−2 sr−1 which is
gether with the spectrum obtained on 0.54% of the sky
4 times smaller than the slope of the H+ cosecant com-
and its analyticaldetermination (Eq. 5). The two spectra nc
ponent.
agree very well.
The isotropy of our residual emission is addressed by
lookingatits variationwiththe Galacticlongitudeorlat- The pixels selected before represent only the diffuse
itude.The variationofourresidualwiththe Galactic lon- medium but contain both the neutral and ionised emis-
gitudeisderivedbycomputingthemeanresidualemission sion and thus cannot be very well represented by a sin-
