astro-ph0005365/ir.tex
1:  %\documentstyle[12pt,aasms4]{article}
2: 
3: %\begin{document}
4: 
5: %\title{Contribution of Extragalactic Infrared Sources
6: %     to\\ CMB Foreground Anisotropy}
7: 
8: \section{Motivation}
9: 
10: 	The COBE detection of large-angular scale anisotropy 
11: in the Cosmic Microwave Background radiation 
12: \citep{smootetal92} 
13: has sparked a drive 
14: to measure the anisotropy on smaller angular scales with the goal 
15: of determining crucial information about the density and 
16: expansion rate of the universe, the 
17: nature of dark matter, 
18: and the spectrum of primordial density perturbations.  
19: Current anisotropy observations look at  sub-degree angular scales 
20: which correspond to observable structures
21: in the present universe.  Improved instrumentation
22: and the MAP (Microwave Anisotropy Probe) and Max Planck Surveyor
23: (formerly COBRAS/SAMBA) satellite missions focus attention on 
24: angular scales between
25: one-half and one-sixth of a degree.
26: 
27: Due to its large beam size, COBE was basically unaffected by 
28: extragalactic foreground sources \citep{bandayetal96, kogutetal94}.
29: Because the antenna temperature 
30: contribution of a point source increases with the inverse 
31: of the solid angle of the beam,
32: observations at higher 
33: angular resolution 
34: are more sensitive to extragalactic foregrounds, including 
35: the low-redshift infrared-bright galaxies examined
36: here.  At frequencies above 200 GHz, these infrared galaxies are the 
37: dominant extragalactic foreground.  Predictions and simulations for 
38: high-redshift infrared galaxies were presented in 
39: Section \ref{sect:wombat_simulations} and limits on their 
40: anisotropy will 
41: be discussed in Chapter \ref{chap:sources}.  
42:    
43: 	Previous work in this area 
44: \citep{toffolattietal95, franceschinietal89, wang91}
45:  used galactic evolution models with specific 
46: assumptions about dust
47: temperatures to predict the level of extragalactic foreground.  
48: We choose instead a phenomenological approach using 
49: the infrared-bright galaxies 
50: detected by the Infrared Astronomical Satellite (IRAS) and the Galactic 
51: emission detected by the COBE (Cosmic Background Explorer) satellite. 
52: Section \ref{sect:ir_results} 
53:  compares our results with those from galaxy-evolution models.  
54: 
55: 	The FIRAS (Far-Infrared Absolute Spectrophotometer) instrument of COBE
56: gives evidence for the existence of 
57: Cold ($<15$K) Dust in the Galactic plane \citep{reachetal95}.  
58: If the Milky Way has Cold Dust, then it is likely present in 
59: other dusty spirals, which comprise the majority of bright 
60: low-redshift infrared sources.  Some observations 
61: \citep{chinietal95, blocketal94, devereauxy92}
62: indicate the presence of Cold Dust in other
63: galaxies.  Neither galactic evolution models nor pre-FIRAS observations 
64: \citep[see][]{ealeswd89}
65: were able to set tight constraints on emission from Cold 
66: Dust, but the FIRAS observations do.  
67: Emission from dust close in temperature to 
68: the 2.73 K background radiation
69: is difficult to separate from real CMB anisotropies. 
70: If Cold Dust in spiral galaxies 
71: is typically accompanied by  
72: the Warm Dust to which IRAS is sensitive, 
73: we can use
74: the FIRAS information about the total dust emission 
75: spectrum of the Galaxy to overcome
76: this spectral similarity and learn about the amount of Cold Dust in 
77: a galaxy by 
78: measuring its amount of Warm Dust.
79: 
80: 
81: \section{Extragalactic Infrared Sources}	
82: 
83: 	The far-infrared discrete sources 
84: detected by IRAS are typically inactive spiral 
85: galaxies, although some are quasars, starburst galaxies, and Seyfert galaxies.
86:   The IRAS 1.2 Jy 
87: catalog \citep{fisheretal95}
88: provides flux measurements of 5319 galaxies at 12, 25, 60, and 100 $\mu$m, 
89: where interstellar dust emission is dominant.  
90: We compared the locations of these galaxies 
91: with those of a thousand of
92: the brightest radio sources, and only 7 possible coincidences resulted.  
93: This lack of coincidence shows that radio-loud galaxies can
94: be treated separately (see Chapter \ref{chap:radio}).  
95: The IRAS sources are roughly isotropic in distribution, except for
96: a clear pattern of the Supergalactic Plane.               
97: To reduce the
98: possibility of residual galactic contamination, we restrict our analysis
99: to  galactic latitude $|b| > 30^{\circ}$, which includes  
100: contributions from 2979 galaxies for a $0\fdg5$ beam.\footnote{The dependence
101: on beamsize is mild, but for a larger beam more sources centered outside 
102: of this area are convolved into it.}  
103: 
104: The nature of dust in spiral galaxies is still an open question.
105: It seems likely 
106: that there is dust at widely 
107: varying temperatures and possibly with different emissivities 
108: \citep{rowan-robinson92, franceschinia95}.
109: Attempts to fit observational data have yielded a variety of 
110: results; it is unclear if far-infrared luminous dust is well described by 
111: a one-component or a two-component model, and the emissivity power-law 
112: index is 
113: only known to be between 1 and 2.  We avoid specifying 
114: the nature of this dust by using the observed Galactic far-infrared emission
115: spectrum as a template for IRAS galaxies.
116: To check the accuracy of this template, we fit 
117: a two-component dust model to IRAS galaxies and
118: to the integrated 12, 25, 60, and 100 $\mu$m 
119: fluxes of the Milky Way measured by the DIRBE (Diffuse Infrared Background
120: Experiment) instrument of COBE.  This produces similar 
121: results for the Warm (15-40 K) Dust component to which IRAS and 
122: DIRBE are most 
123: sensitive; for an emissivity power-law index of 1.5, DIRBE gives
124: a Warm Dust temperature of 28K for the Milky Way, while the 425 
125: IRAS galaxies with highest-quality flux measurements are 
126: collectively fit to a Warm Dust temperature of 33K.  This Warm Dust 
127: accounts for the majority of the far-infrared emission of spiral galaxies.
128: 
129: There is, however, observational evidence that the far-infrared emission
130: of inactive 
131: spirals is dominated by dust slightly colder than 20K 
132: \citep{neiningerg96, chinik93}.
133: Fitting the FIRAS 
134: spectrum of the Milky Way also leads to a Warm Dust temperature close to
135: 20K.  These fits appear to conflict with the temperatures found above using
136: IRAS and DIRBE fluxes at $\lambda \leq 100 \mu$m.  
137: Using 
138: 60, 100, 140, and 240 $\mu$m DIRBE fluxes, however, indicates a
139: Warm Dust temperature for the Galaxy of 24K.  This shows that temperature
140: fits to data on one side of 
141: the peak of a graybody (modified blackbody) spectrum can be 
142: inaccurate.  
143: Figure \ref{fig:iras_fig1} shows that
144: the spectra of the Milky Way found by DIRBE and FIRAS
145: are indeed
146: compatible.  It may be an oversimplification to represent the 
147: Warm Dust in a galaxy by a single temperature.  
148: 
149: 
150: \begin{figure}
151: \centerline{\psfig{file=iras_fig1.ps,width=5in}} 
152: \caption{Galactic Far-IR Emission Spectrum} 
153: \mycaption{The FIRAS Galactic Dust spectrum, including emission lines, is 
154: shown by $+$ symbols.  
155: The smooth curve is a fit to this spectrum 
156: based upon a two-component dust model with synchrotron and free-free emission 
157: included using DMR results.  
158: The FIRAS error bars are not shown because they are
159: extremely small on this scale.  
160: The open circles are 
161: DIRBE integrated Galactic fluxes at 12, 25, 60, 100, 140, and 240 $\mu$m,
162: normalized to the FIRAS measurements.
163: }
164: \label{fig:iras_fig1}
165: \end{figure}
166: 
167: 
168: We recognize that 
169: not all IRAS galaxies have the same far-IR spectrum as the Milky Way.  
170: Active galaxies are warmer, 
171: with an average Warm Dust temperature of 33K (for emissivity index 2, 
172: \citealt{chinietal95}).
173: However, the
174: cirrus emission which dominates Galactic dust is consistent
175: with the emission from the majority of inactive spirals 
176: \citep{andreanif96, pearsonr96}.
177: Some observations indicate that our Galaxy is slightly warmer than 
178: the average inactive spiral \citep{chinietal95}.  None of these observations
179: includes enough frequencies to provide a 
180: template microwave emission spectrum, and
181: their results range by a factor of 3 depending on the choice of beam
182: corrections \citep{franceschinia95}.
183: The Milky Way is a 
184: good middle-of-the-road choice for a microwave template spectrum;
185: the 
186: DIRBE and IRAS dust 
187: temperature fits given above 
188: agree rather well.  For extrapolation to microwave frequencies, uncertainty 
189: in emissivity is of much greater importance than this level of temperature 
190: uncertainty, anyway.  
191: 
192: After removing Galactic emission lines (as in \citealt{reachetal95}), 
193: we fit a two-component dust model to the 
194: FIRAS dust spectrum.  The CO 1-0 emission line at 115 GHz is not clearly
195: detected by FIRAS but could be responsible for increased
196: emission at that frequency. 
197: With a $\nu^2$ emissivity law
198: assumed, the best fit is Warm Dust at $19.4$K and Cold Dust at $4.3$K with an 
199: optical depth 12.1 times that of the Warm Dust.  
200: It is possible to vary the parameters of the 
201: dust model significantly and still have an acceptable fit, so we refrain 
202: from assigning any physical importance to the parameters of the fit.  
203: We add synchrotron and free-free components with microwave-range
204: spectral indices of 
205: $-1.0$ and $-0.15$, 
206: respectively, so that these sources of microwave emission match COBE DMR
207: (Differential Microwave Radiometer)
208: observations below 100 GHz 
209: \citep{kogutetal96a, reachetal95, bennettetal92}.
210: Free-free emission is stronger than dust beyond the low-frequency end
211: of the FIRAS spectrum.     
212: 
213: We combine data from DIRBE, FIRAS, and DMR to form the  
214: broad Galactic spectrum shown in Figure \ref{fig:iras_fig1}.  
215: Each IRAS 1.2 Jy source is fit to 
216: the DIRBE end of the spectrum and extrapolated to the desired frequency using 
217: this template. 
218: In fitting each IRAS galaxy to the DIRBE fluxes of the Milky Way,
219: we give more weight to the 60 and 100 $\mu$m fluxes, which are most 
220: sensitive to Warm Dust, than to the 12 and 25 $\mu$m fluxes, 
221: which are also sensitive to Hot (100-300 K) Dust.  
222: The 1.2 Jy catalog gives redshifts for these galaxies.  
223: Most have $z < 0.05$ and all have $z < 0.3$. 
224: We take these redshifts into account while fitting 
225: and extrapolating.
226: 
227: It would be advantageous to 
228: fit each type of galaxy
229: to a specialized far-IR to microwave spectrum, but no other 
230: trustworthy template spectrum is currently available, 
231: so we use the Galactic far-infrared emission spectrum for all sources.
232: The Galactic spectrum agrees well with observed correlations between radio
233: and IR fluxes of IRAS galaxies 
234: \citep{condonb91, crawfordetal96}.
235: Our template spectrum is consistent 
236: with detections and 
237: upper limits for bright infrared galaxies
238: from DIRBE \citep{odenwaldns98}.  
239: This is helpful because DIRBE used 140 and 240 $\mu$m
240: channels, which IRAS lacks, allowing it to probe much cooler
241: dust temperatures than IRAS.
242: DIRBE rules out the 
243: possibility of extremely bright sources occurring in the 2\% of the high 
244: Galactic latitude sky not surveyed by IRAS and sees no evidence for 
245: sources whose emission comes predominantly from Cold Dust.
246: 
247: \section{Results}
248: 
249: \label{sect:ir_results}
250: 
251: We use the Galactic far-infrared emission spectrum to predict the 
252: microwave flux of each IRAS galaxy in Jy ($1$ Jy $= 10^{-26} W/m^2/$Hz).  
253: To convert from flux $S$ to antenna temperature $T_A$, we use
254: 
255: \begin{equation}
256: \label{eq:flux_to_temp}
257:  T_A = S \frac{\lambda^2}{2 k_B \Omega}\; \;,  
258: \end{equation}
259: 
260: \noindent 
261: where $k_B$ is Boltzmann's constant, $\lambda$ is the wavelength, and 
262: $\Omega$
263: is the effective beam size of the observing instrument.  Antenna temperature
264: is related to thermodynamic temperature by
265: 
266: \begin{equation}
267: \label{eq:t_to_ta}
268:  T_A = \frac {x}{e^x - 1} \; \; T,  
269: \end{equation}
270: 
271: \noindent
272: defining $x \equiv h \nu / k T$.  Small fluctuations in antenna 
273: temperature can be converted to effective thermodynamic 
274: temperature fluctuations using
275: 
276: \begin{equation}
277: \label{eq:dt_to_dta}
278:  \frac{dT_A}{dT} = \frac { x^2 e^x} {(e^x - 1)^2} \; \; .  
279: \end{equation}
280: 
281: Analysis of source counts indicates that the 1.2 Jy sample is complete down to 
282: an extrapolated flux of 3 mJy at 100 GHz.  We divide 
283: the sources logarithmically 
284: into groups of similar flux and 
285: find a gradual 
286: decrease in anisotropy as flux decreases, indicating that dimmer sources
287: will not generate significant anisotropy.  This may not hold true 
288: for high-redshift infrared galaxies, however, as the k-correction makes 
289: sources which are dim at 100$\mu$m quite bright in the sub-millimeter.  
290: \citet{toffolattietal95}
291: found a negligible contribution from non-Poissonian fluctuations.  
292: Poissonian fluctuations should
293: be dominated by those sources prevalent enough to have roughly one source
294: per pixel.  
295: If all sources have roughly the luminosity of the Milky Way, then 
296: for an instrument with a resolution of $10'$ to have one source
297: per beam, we must look at sources with $ z \simeq 0.24$.  
298: Assuming $(1+z)^3 $ luminosity evolution and including 
299: k-correction  
300: (see \citealt{pearsonr96, beichmanh91}), 
301: these sources will generate a 
302: temperature anisotropy only 2\% of that caused by IRAS 1.2 Jy galaxies.
303: We therefore expect the anisotropy generated 
304: by low-redshift 
305: sources too dim to make the 1.2 Jy catalog to be a small part of the total
306: anisotropy; the brightest sources are generating most of the fluctuations.
307: 
308: 
309:           
310: \begin{figure}
311: \centerline{\psfig{file=iras_fig2.ps,width=4in,angle=90}} 
312: \caption{Microwave anisotropy from low-redshift infrared galaxies.}
313: \mycaption{A log-log-log contour plot 
314: of equivalent thermodynamic temperature fluctuations  
315: due to extragalactic infrared sources as a function of frequency in GHz 
316: and angular resolution (FWHM) in degrees.  
317: The temperature anisotropy shown is $\log_{10} \frac{\Delta T}{T}$
318: where $\Delta T$ is the root mean square equivalent thermodynamic
319: temperature generated
320: by extragalactic infrared sources in Kelvin and $T$ is the 
321: temperature (2.73K) of the CMB.  The increase in anisotropy at low frequencies
322: occurs because synchrotron and free-free emission are included in our 
323: template spectrum. 
324: }
325: \label{fig:iras_fig2}
326: \end{figure}
327: 
328: 
329: 	To simulate observations, we 
330: convolved all sources on pixelized skymaps (2X oversampled) of
331: resolution varying from $10'$ to $10^{\circ}$. 
332: The resulting maps, covering a range of frequencies from 30 to 900 GHz, 
333: were analyzed to 
334: determine the expected contribution of IRAS galaxies to foreground 
335: confusion of CMB temperature anisotropy.
336: The information contained in these skymaps can be used 
337: to choose regions of the sky in which 
338: to observe \citep{smoot95}.  
339: The contour plot in 
340: Figure \ref{fig:iras_fig2} shows the 
341: rms thermodynamic temperature anisotropy
342: produced by extragalactic infrared sources 
343: over the full range of frequencies and instrument resolution.  
344: The minimum value of $\frac{\Delta T}{T}$
345: is $1.3\times 10^{-8}$ at large FWHM and medium frequency 
346: and the maximum value is $0.092$ 
347: at small FWHM and high frequency.  For frequency in GHz and FWHM in degrees, 
348: our results for temperature
349: anisotropy are fit to within 10\% by
350: \begin{equation}
351: \log_{10}\frac {\Delta T}{T} = 2.0 (\log_{10}\nu)^3 - 8.6 (\log_{10}\nu)^2
352: + 10.3 \log_{10}\nu - 0.98 \log_{10} (FWHM) - 9.2 \; . 
353: \end{equation}
354: The inverse linear relationship between anisotropy and
355: FWHM results from the combined effects of beam convolving 
356: and map pixelization (see Chapter \ref{chap:sources} for a derivation).  
357: Anisotropy from extragalactic infrared sources dominates 
358: expected CMB anisotropy at frequencies above 500 GHz.  This makes effective
359: foreground discrimination possible for instruments with a frequency range 
360: sufficiently wide to detect the extragalactic infrared 
361: foreground directly.
362: 
363: \begin{figure}
364: \centerline{\psfig{file=iras_fig3.ps,width=4in}} 
365: \caption{Confusion from low-redshift infrared galaxies}
366: \mycaption{Log-log plot of $\frac{\Delta T}{T}$
367: versus frequency for instrument resolutions of
368: $10'$, $30'$, $1^{\circ}$, and $10^{\circ}$,
369: showing window where foreground confusion should be
370: less than $10^{-6}$.  
371: Solid lines are for no pixel subtraction.  The 
372: dotted, dashed, and long-dashed lines show the results 
373: with pixels at a level of $5\sigma$ removed for resolutions of 
374: $10'$,$30'$, and $1^{\circ}$, respectively.  This $5\sigma$ subtraction 
375: makes no difference at any frequency for $10^{\circ}$.  
376: }
377: \label{fig:iras_fig3}
378: \end{figure}
379: 
380: 
381: 
382: 
383: Figure \ref{fig:iras_fig3} 
384: shows a summary of our results for several benchmark instrument
385: resolutions.  The dashed lines represent 
386: the results of subtracting pixels where
387: the fluctuations from extragalactic infrared sources are five times
388: times greater than the quadrature sum of the rms CMB anisotropy and the 
389: expected
390: instrument noise for the Planck Surveyor at that frequency.
391: These $5 \sigma$ pixels can be
392: assumed to contain bright point sources. 
393: Our results agree closely with those of \citet{toffolattietal98}
394: for their
395: model of moderate cosmological evolution of all galaxies.  
396: Our predictions for anisotropy are about a factor of three 
397: lower than those
398: of \citet{franceschinietal89},
399: who assume strong evolution of the 
400: brightest IR sources and include early galaxies with heavy starburst activity. 
401: \citet{wang91}
402: ignores the possibility of cold dust and uses galaxy evolution 
403: models to predict anisotropy levels somewhat lower than those found 
404: with our phenomenological approach.  
405: 
406: The $5 \sigma$ subtraction has a significant  
407: effect for small FWHM at frequencies below 500 GHz.   
408: The maximum effect is to subtract 0.002\% of the pixels, 
409: leading to a factor of 5 reduction in foreground 
410: temperature anisotropy.  
411: This is
412: further 
413: evidence that temperature anisotropy from extragalactic infrared sources 
414: is dominated by the brightest sources.
415:   The bright sources are a mixture of Local Group galaxies and 
416: more distant infrared-luminous galaxies such as starburst galaxies.  Optimal
417: subtraction of the extragalactic infrared foreground requires the contribution
418: from each bright source to be predicted accurately.
419: 
420: 
421: Figure \ref{fig:iras} shows a  
422: skymap of our extrapolated IRAS 1.2 Jy catalog at 100 GHz.  
423: Very few of 
424: these sources are significant at frequencies below 200 GHz, although the 
425: entire 5319 sources should be detected by Planck HFI due to its high 
426: resolution at high frequencies.  
427: The skymap shows the structure of the Supergalactic Plane, 
428: as well as the regions of the sky not included 
429: due to Galactic contamination or the IRAS satellite's failure to observe them.
430:   Several 
431: sources are considerably brighter than the maximum of the color 
432: table, which has been set 
433: very low to show all of the catalog, even though most sources 
434: are quite dim at 100 GHz.
435:  
436: \section{Discussion}
437: 
438: Our usage of the Galactic far-infrared emission spectrum as a template
439: causes systematic errors on a galaxy-by-galaxy basis.  
440: Our method can be improved in the future 
441: to account for the spectral difference 
442: between Ultraluminous Infrared Galaxies and normal spirals.  
443: It is easy to place constraints on our results; if all
444: galaxies had only 33K dust as is typical for active galaxies, the resulting 
445: anisotropy would be a factor of 100 lower.  This is highly unlikely, because we
446: know that most IRAS galaxies are inactive spirals, and galaxies
447: with colder dust will dominate the anisotropy at mm-wavelengths because 
448: of the selection effect favoring sources with flatter spectra.  A robust upper
449: limit on microwave anisotropy from infrared galaxies can be set by assuming
450: that these IRAS 1.2 Jy galaxies cause the full cosmological far-infrared
451: background 
452: \citep{pugetetal96, buriganap98, fixsenetal98, schlegelfd98, 
453: hauseretal98}.  
454: In this case we have underestimated
455: the anisotropy by a factor of 100, but no predictions of the IR background
456: expect these nearby galaxies to produce more than a few percent of it.  
457: A more realistic check on our results comes from Andreani \& Franceschini
458: (1995), who measured a complete sample of IRAS galaxies at 1300 $\mu$m 
459: (240 GHz).  Their average flux ratio of 1300 $\mu$m over 100 $\mu$m is half
460: that of the Galaxy, but one of their beam correction methods brings their ratio
461: into agreement with the Milky Way.  They find that the 60 $\mu$m 
462: emission of spiral galaxies receives enough contribution from a starburst
463: dust component mostly absent in the Galaxy that including 60 $\mu$m fluxes
464: in our fits may have caused a factor of 2 overestimate.  
465: Combined, these corrections
466: give us a possible systematic overestimate of anisotropy by a factor of 4.  If
467: typical IR-bright galaxies have dust colder than the Milky Way, our
468: results could instead be an underestimate by a factor of a few, but this
469: appears less likely. 
470: 
471: We estimate an overall systematic uncertainty of a factor of 2 at 300 GHz, 
472: increasing to a factor of 3 at 100 GHz and to a factor of 5 at 30 GHz.  
473: As it is hard to predict the typical dust temperature and emissivity in 
474: a given galaxy as well as its relative amount of free-free and synchrotron 
475: emission, we estimate that our predictions for the microwave 
476: spectrum of each source have an overall factor of 5 uncertainty 
477: and an independent  
478: factor of 1.3 uncertainty at each frequency.  
479: The factor of 5 uncertainty preserves a source's spectral shape whereas the 
480: factor of 1.3 uncertainties allow for errors in the predicted 
481: spectral shape.  These uncertainties are quite large, but the extrapolation 
482: we have performed is over a factor of 10-100 in frequency.  We expect 
483: that forthcoming microwave observations will give us better information, 
484: especially about the brightest IRAS 1.2 Jy sources.   
485: 
486: 
487: 	The recently obtained spectral knowledge of our Galaxy has enabled
488: us to take into account the possible 
489: presence of Cold Dust.  
490: Our predicted level of temperature anisotropy makes the extragalactic
491: foreground from low-redshift infrared galaxies 
492: dominant over the Galactic foregrounds
493: of dust, free-free, and synchrotron for angular resolutions 
494: near $10'$ and frequencies above 100 GHz.  
495: Below 150 GHz, radio sources are
496: expected to be the dominant extragalactic
497: foreground.  
498: The extragalactic low-redshift 
499: infrared foreground will not be significant in comparison to
500: CMB anisotropies around 100 GHz but will be dominant above 500 GHz.  
501:   Despite the possible presence of Cold Dust in infrared-bright galaxies,
502: our results leave a window at intermediate frequencies
503: for the measurement of 
504: CMB anisotropies without significant confusion
505: from extragalactic infrared sources.  
506: 
507: 
508: 
509: 
510: 
511:   
512:             
513: %\begin{figure}
514: %\centerline{\psfig{file=5sigma.ps}, width=5in} 
515: %\caption{
516: %A log-log surface plot of the number of pixels reaching the $5\sigma$ 
517: %level due to IRAS point source emission as a function of frequency 
518: %and instrument resolution.}  
519: %\label{fig:5sigma}
520: %\end{figure}
521: 
522: \cleardoublepage
523:   
524: \begin{figure}
525: \centerline{\psfig{file=iras.ps,width=6in,angle=90}} 
526: \caption{
527: Skymap at 100 GHz of predicted contribution from IRAS 1.2 Jy sources.}
528: \label{fig:iras}
529: \end{figure}
530: 
531: 
532: 
533: %\end{document}
534: