0712.2238/ms.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: % Gregory Dobler (started July 6, 2007) %
3: %                                       %
4: %    Halpha-correlated spinning dust    %
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: 
7: 
8: %
9: % Classes, packages, and comments
10: %
11: \documentclass[12pt,preprint]{emulateapj}
12: \usepackage{graphicx}
13: %\slugcomment{Draft --- July 25, 2007}
14: 
15: %
16: % Comments
17: %
18: 
19: 
20: %
21: % Equations, figures, tables, and references
22: %
23: \newcommand{\be}{\begin{equation}}
24: \newcommand{\bel}[1]{\be\label{eq:#1}}
25: \newcommand{\ee}{\end{equation}}
26: \newcommand{\ba}{\begin{eqnarray}}
27: \newcommand{\ea}{\end{eqnarray}}
28: \newcommand\bp{\begin{figure}}
29: \newcommand\ep{\end{figure}}
30: \newcommand\bpm{\begin{figure*}}
31: \newcommand\epm{\end{figure*}}
32: \newcommand{\btab}{\begin{tabular}}
33: \newcommand{\etab}{\end{tabular}}
34: \newcommand{\bt}{\begin{table}}
35: \newcommand{\et}{\end{table}}
36: \newcommand{\ben}{\begin{enumerate}}
37: \newcommand{\een}{\end{enumerate}}
38: \newcommand\reffig[1]{Figure \ref{fig:#1}}
39: \newcommand\refeq[1]{Equation \ref{eq:#1}}
40: \newcommand\refsec[1]{\S \ref{sec:#1}}
41: \newcommand\reftbl[1]{Table \ref{tbl:#1}}
42: \newcommand{\cm}{\rm{cm}}
43: \newcommand{\pc}{\rm{pc}}
44: \newcommand{\Ray}{\rm{R}}
45: \newcommand{\kjysr}{~\rm{kJy\,sr}^{-1}}
46: 
47: %
48: % Alignment
49: %
50: \newcommand{\bcn}{\begin{center}}
51: \newcommand{\ecn}{\end{center}}
52: 
53: 
54: %
55: % Symbols
56: %
57: \newcommand{\delnu}{\Delta\nu_{\rm p}}
58: \newcommand{\nup}{\nu_{\rm p}}
59: \newcommand{\hal}{H$\alpha$}
60: 
61: 
62: 
63: \begin{document}
64: 
65: \title{Identification of Spinning Dust in H$\alpha$--Correlated Microwave 
66: Emission}
67: 
68: 
69: \author{Gregory Dobler\altaffilmark{1,2} \& Douglas P. Finkbeiner\altaffilmark{1,3}}
70: 
71: \altaffiltext{1}{
72: Institute for Theory and Computation, 
73: Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, MS-51,
74: Cambridge, MA 02138 USA
75: }
76: \altaffiltext{2}{gdobler@cfa.harvard.edu}
77: \altaffiltext{3}{dfinkbeiner@cfa.harvard.edu}
78:   
79: 
80: \begin{abstract}
81: CMB experiments commonly use maps of H$\alpha$ intensity as a spatial
82: template for Galactic free--free emission, assuming a power law
83: $I_{\nu} \propto \nu^{-0.15}$ for the spectrum.  Any departure from
84: the assumed free--free spectrum could have a detrimental effect on
85: determination of the primary CMB anisotropy.  We show that the
86: H$\alpha$-correlated emission spectrum in the diffuse WIM is
87: \emph{not} the expected free--free spectrum at WMAP frequencies.
88: Instead, there is a broad bump in the spectrum at $\sim$ 50 GHz which
89: is consistent with emission from spinning dust grains.  Spectra from
90: both the full sky and smaller regions of interest are well fit
91: by a superposition of a free--free and ``warm ionized medium'' Draine
92: \& Lazarian (1998) spinning dust model, shifted in frequency.  
93: The spinning dust emission is $\sim$
94: 5 times weaker than the free--free component at 50 GHz, with the null
95: hypothesis that the H$\alpha$-correlated spectrum is pure free--free,
96: ruled out at $\geq
97: 8\sigma$ in all regions and $>100\sigma$ for the full sky fit. 
98: \end{abstract}
99: \keywords{ 
100: diffuse radiation ---
101: dust, extinction --- 
102: ISM: clouds --- 
103: radiation mechanisms: non-thermal --- 
104: radio continuum: ISM 
105: }
106: 
107: %------------------------------------------------------------------------------
108: % SECTION: Introduction
109: %------------------------------------------------------------------------------
110: \section{Introduction}
111: 
112: During the course of measuring the cosmic microwave background (CMB)
113: anisotropy, the \emph{Wilkinson Microwave Anisotropy Probe} (WMAP) has
114: produced the most detailed and sensitive maps of interstellar medium
115: (ISM) emission between 20 and 100 GHz taken to date.  The quality of
116: the data has profoundly influenced our understanding of the four main
117: ISM continuum emission mechanisms in this frequency range
118: \citep{bennett03,finkbeiner03,hinshaw07,DF07}.  At the highest WMAP
119: frequencies, the Galaxy is dominated by thermal dust emission, traced
120: by far IR maps \citep{finkbeiner99}.  At lower
121: frequencies, synchrotron radiation from supernova shock accelerated
122: electrons spiraling in the Galactic magnetic field makes up the
123: majority of the emission.  In the inner Galaxy, the synchrotron
124: emission is brighter relative to low-frequency maps, and has a harder
125: spectrum.\footnote{\citet{finkbeiner04} has previously interpreted this as a 
126: separate component, the synchrotron ``haze'', but this interpretation is
127: controversial \citep{finkbeiner04,DF07}.  This haze component is
128: spatially orthogonal to the other templates by construction, and
129: though we include it in our fits below, its existence has negligible
130: effect on the \hal-correlated spectrum.}  Next, free--free emission
131: from unbound electrons interacting with ions in $\sim 10^4$K ionized
132: gas has a harder spectrum than synchrotron, and is detectable in all bands.
133: Finally, there is a bump in the spectrum of dust-correlated emission
134: at $\sim 20$ GHz, sometimes called ``Foreground X'' \citep{deO02}.
135: Over the last decade, it has been recognized that this could be
136: electric dipole emission from rapidly rotating dust grains \cite[``spinning
137: dust'';][]{DL98b}.
138: 
139: \subsection{Foreground X: spinning dust?}
140: Spinning dust is the electric dipole emission from the smallest dust
141: grains which have a non-zero electric dipole moment and are spun
142: up by several different interactions with the interstellar medium
143: \citep[e.g., ion collisions, the interstellar radiation field, etc.; see][]{DL98b}.  
144: Spinning dust was considered as a potential source of
145: radio and microwave emission in other contexts \citep{erickson57,ferrara94}.
146:   Because the model depends on the grain size
147: distribution, ionization fraction, electric dipole moments, etc., the
148: peak frequency can be anywhere from $10-50$ GHz, but is usually in the
149: $20-40$ GHz range.  This
150: dependence of the spectrum on the ISM parameters has made spinning
151: dust difficult to definitively identify, but may provide the qualitative
152: behaviors necessary to make a fool-proof detection.
153: 
154: In the COBE/DMR data, \citet{kogut96} found that the dust-correlated
155: emission does not simply fall off monotonically with decreasing
156: frequency as would be expected for a thermal dust tail, but instead was greater at 
157: 31 GHz than at 53 GHz.  This behavior
158: was confirmed by the Saskatoon CMB experiment \citep{deO97},
159: the 14.5 and 32 GHz OVRO data \citep{leitch97},
160: the Cottingham \& Boughn 19.2 GHz survey \citep{deO98}, 
161: the Tenerife 10 \& 15 GHz survey \citep{deO99}, and the
162: QMAP 30 GHz data \citep{deO00}. 
163: A small survey of dust clouds with the Green Bank 42m telescope at 5,
164: 8, \& 10 GHz \citep{F02} revealed two candidates for spinning dust
165: clouds: L1622, later
166: confirmed \citep{finkbeiner04}; and LPH~201.663+1.643, which turned
167: out to be incorrect \citep{dickinson06}. 
168: 
169: Given this evidence, it was already clear prior to the WMAP data
170: release in early 2003, that there was some non-standard emission
171: mechanism, but it was unclear whether this mechanism was spinning dust
172: \citep{DL98b} or some other process.
173: 
174: WMAP \citep{bennett03} found that the dust-correlated emission falls
175: from 94 to 61 GHz and then rises steadily to 23 GHz.  The fact that
176: synchrotron typically dominates at lower frequencies led \citet{bennett03} to
177: interpret the low frequency rise as ``dust-correlated synchrotron''.
178: The reasoning was that because dusty regions correlate with regions of
179: star formation activity, electrons in the vicinity could be
180: accelerated by supernova shocks leading to synchrotron emission.
181: \citet{finkbeiner04} pointed out that WMAP is also consistent with
182: \citet{DL98b} if their model could be shifted down in frequency a bit.
183: 
184: Because the WMAP frequency coverage ($23-94$ GHz) coincides with the
185: range where the spinning dust and synchrotron spectra are similar,
186: data at lower frequencies, where the spectra diverge, have been used to
187: test the dust-correlated synchrotron hypothesis.  The WMAP ``synchrotron''
188: template has a much weaker cross-correlation at $10-15$ GHz than
189: expected for actual synchrotron \citep{deO04}.  Also, 8 \& 14 GHz data from
190: the Green Bank Galactic Plane Survey \citep{langston00} show a rising
191: spectrum consistent with spinning dust but not synchrotron
192: \citep{gb04} and implying that the majority of dust-correlated
193: emission at 23 GHz is Foreground X.  Comparing the Cottingham \&
194: Boughn survey to WMAP also shows a rise from 19 to 23 GHz
195: \citep{boughn07}.  The polarization properties \citep{page07,hinshaw07} in the
196: 3-yr WMAP data also favor the spinning dust hypothesis.
197: 
198: In spite of all this evidence, there are weaknesses in each of these
199: arguments.  Some use only low-latitude data and may not apply to the
200: diffuse ISM \citep{gb04}.  Others apply to only one cloud \citep{F02},
201: have poor signal to noise \citep{deO04}, or a short lever arm in
202: frequency \citep{boughn07}.  However, taken together, they are very
203: strong evidence against the dust-correlated synchrotron hypothesis,
204: but they are not specifically evidence \emph{for} spinning dust.
205: Other mechanisms such as magnetic dipole dust emission, caused by the
206: thermal fluctuation of the magnetization in the dust grains
207: \citep{DL99}, could conceivably explain all these results.  This
208: ambiguity motivated our search for further clues in the WMAP 3-yr
209: data.
210: 
211: \subsection{WMAP Multilinear Regression}
212: In \citet[hereafter, DF07]{DF07} we describe our method for
213: determining the spectral shape of each of the four primary WMAP
214: foreground components: free--free, dust, soft synchrotron, and hard synchrotron 
215: \citep[termed the ``haze'' by][]{finkbeiner04} emission.  We model the WMAP data 
216: as a linear combination of the CMB
217: and the 4 foregrounds, each described by a specific spatial template.
218: For free--free emission
219: we use the \hal\ map described in \refsec{halpha-em}, for dust (thermal
220: and spinning) emission we use the \citet{schlegel98} dust map
221: evaluated at 94 GHz by \citet{finkbeiner99}, for soft synchrotron we
222: use the \citet{haslam82} 408 MHz map, and for the haze we use a simple
223: $1/r$ template where $r$ is the distance from the Galactic center.
224: The individual WMAP bands were completely decoupled in our fit, and we
225: assumed \emph{nothing} about the spectrum of each foreground, only
226: that its morphology was traced by the template in all bands.  We
227: performed our fits over both the (nearly) full sky and also on smaller
228: regions of interest.  Point sources and regions where dust extinction
229: makes the \hal\ map unreliable (where $A(\mbox{\hal}) =
230: 2.65\,E(B-V) > 1\mbox{ mag}$) were masked out from the fits.
231: 
232: We showed that the spectrum of \hal-correlated emission is
233: \emph{not} the usual free--free spectrum.  Rather, there is
234: a bump around $\nu \sim 50$ GHz.  In \refsec{sdust-em} we explain why
235: the \hal\ map should actually trace spinning dust excited by
236: collisions with ions.  In \refsec{spectra} we present \hal-correlated
237: emission spectra for both the (nearly) full sky as well as smaller
238: regions of interest and show that, indeed, the spectra are well
239: characterized as a superposition of free--free and spinning dust type
240: spectra.  We summarize this work and discuss implications for future
241: observations in \refsec{discussion}.
242: 
243: %------------------------------------------------------------------------------
244: % SECTION: EM as a tracer of spinning dust
245: %------------------------------------------------------------------------------
246: \section{Emission measure as a tracer of spinning dust}
247: 
248: 
249: In this section we argue that in fully ionized environments, the emissivity 
250: density ($j_\nu$) of spinning dust emission is proportional to dust density times 
251: some excitation function related to the ion density.  The emission comes from the 
252: very smallest grains, which are spun up by a collision with an ion and spin down 
253: as they radiate.  In the limit where the spin down time is small compared to the 
254: time between collisions, the rotation is episodic and the emissivity also scales 
255: linearly with the ion number density.  In \S \ref{sec:spinem}, we assume the episodic limit in order to build some intuition about how WIM-correlated spinning dust might behave. 
256: 
257: 
258: \subsection{Free--free emission measure}
259: \label{sec:halpha-em}
260: 
261: The map assembled by \citet{finkbeiner03} from the VTSS
262: \citep{dennison98}, SHASSA \citep{gaustad01}, and WHAM
263: \citep{haffner03} surveys, traces the \hal\ recombination line
264: emission from hot interstellar gas.  At WMAP frequencies, this \hal\
265: map\footnote{available at \texttt{http://www.skymaps.info}} is used to
266: estimate the thermal bremsstrahlung, or free--free emission, resulting
267: from the interaction of the electrons and protons from ionized
268: hydrogen \citep{bennett03,finkbeiner04,hinshaw07}.  The \hal\ map is
269: corrected for dust absorption using the prescription given by Finkbeiner
270: (2003; but see Dickinson et al. 2003 for an alternative).  In this
271: frequency range the free--free specific intensity $I_\nu$ (in Jy/sr) is
272: often approximated as a power law, \bel{ff-spec} I_{\nu} \propto
273: \nu^{\alpha}, \ee with $\alpha \approx -0.15$.  Although this is a
274: good approximation, in our analysis we use the expression given by
275: \citet{spitzer} for the free--free spectrum.
276: 
277: Recombination line emission is a two particle process (involving the
278: capture of free electrons by a proton nucleus) and so the emissivity
279: is proportional to $n_e n_i$, which for fully ionized H is simply the
280: square of the number density of gas particles, $j_{\mbox{\hal}}\propto
281: n^2$.  Therefore, the \hal\ intensity corresponds to an \emph{emission
282:   measure} (EM) defined as $E_m \equiv \int n^2 dl$ where the integral
283: is taken along the line of sight.  The \hal\ intensity is usually
284: measured in Rayleighs (1 R $\equiv 10^6/4\pi$ photons s$^{-1}$
285: cm$^{-2}$ sr$^{-1}$) with an EM of 1 cm$^{-6}$pc $\approx 0.6$ R for
286: gas at $\sim 8000$ K.
287: 
288: 
289: \subsection{Spinning dust emission measure}
290: \label{sec:spinem}
291: 
292: \citet[hereafter, DL98]{DL98b} present a model for spinning dust
293: emission in which tiny dust grains with non-negligible electric dipole
294: moment are spun up by ion collisions; polycyclic aromatic hydrocarbons (PAHs) 
295: generally have a non-zero dipole, because perfect symmetry is rare in large 
296: molecules.  DL98 estimated the astronomical value of this parameter (actually the 
297: electric dipole moment per root mass) from measured electric dipole moments for 
298: dozens of known PAH molecules.  In the limit where ion interactions provide the 
299: dominant spin-up mechanism (see DL98 for details) and assuming full ionization, 
300: constant gas/dust ratio, and the same grain size distribution everywhere, the 
301: resulting \emph{total} spinning dust emission is $I \equiv \int I_{\nu} d\nu 
302: \propto \int n_i n_d dl \propto \int n_i^2$.  
303: With the additional assumption that the emission is episodic, the spectrum of the emission does not vary with $n_i$, but the amplitude does. 
304: That is, the spinning dust emission 
305: should look like an EM, and be traced by \hal.  This approximation is relevant 
306: where the collisional excitation is primarily due to ions (at least for the smallest grains), e.g. for the ``warm ionized medium'' (WIM) parameters used by DL98.
307: 
308: The WIM-correlated emission claimed here augments the ``usual''
309: spinning dust emission in the ``cold neutral medium'' (CNM) which has
310: a temperature $T_{\rm CNM} \sim 100$ K and is traced by the
311: \citet{schlegel98} dust map evaluated at 94 GHz by
312: \citet{finkbeiner99} (FSD99).
313: 
314: 
315: \subsection{Estimating the specific intensity}
316: \label{sec:sdust-em}
317: 
318: For a spherical dust grain of size $a$, rotating with angular frequency $\omega$, 
319: and with a dipole moment $p_0$, the rotational energy is,
320: \bel{rotenergy}
321:   E = \frac{1}{2} I \omega^2 = \frac{1}{5} m a^2 \omega^2,
322: \ee
323: where $m$ is the grain mass.  The energy loss rate is the total power emitted by 
324: the grain,
325: \be
326:   \dot{E} = \frac{1}{12\pi}\frac{\mu_0}{c} p_0^2 \omega^4
327: \ee
328: and the spin down time is,
329: \be
330:   \tau_s \equiv E/\dot{E} \propto a^2 p_0^{-2} \omega^{-2}.
331: \ee
332: The time between collisions is just,
333: \be
334:   \tau_c \sim \lambda/v \propto a^{-2} n_H^{-1}
335: \ee
336: where $v$ is the thermal velocity of the ion and $\lambda=(\pi a^2
337: n_H)^{-1}$ is the mean free path.
338: 
339: In the limit where the spin down time is much \emph{smaller} than the time between 
340: collisions ($r \equiv \tau_s/\tau_c \ll 1$), the excitation is episodic.  For the 
341: WIM parameters in DL98, this ratio is in fact $\sim 1$; however, we point out that 
342: \be
343:   r = \tau_s/\tau_c \propto a^4 n_H p_0^{-2} \omega^{-2}
344: \ee
345: so that very small changes to the grain size distribution can substantially affect 
346: the spin down time compared to the time between collisions.  In fact, if each ion 
347: collision imparts equal energy, then from \refeq{rotenergy}, $\omega \propto 
348: a^{-2}$.  Since $p_0 \propto N^{1/2} \propto a^{3/2}$ (see DL98) 
349: then in this case, $1/r$ diverges as $a^{-5}$.  The bottom line is that the 
350: episodic approximation is sensitively dependent on the ion density, the dipole 
351: moment, and 
352: \emph{especially} the grain size distribution. For the purposes of estimating a 
353: specific intensity, we will now assume the episodic approximation with the caveat 
354: that it may break down in dense environments, \citep[e.g., HII 
355: regions.  In fact,][find no evidence for spinning dust at 14.2 and 17.9 
356: GHZ in 16 Galactic HII regions]{scaife07} or for very small dipole moments.  
357: We emphasize that this does not affect the fact the the \emph{total intensity} $I$ 
358: should be proportional to $\int n_d n_H d\ell$ so that spinning dust emission 
359: should be traced by an EM map.
360: 
361: Draine \& Lazarian evaluate $j_\nu/n_H$ for parameters representative
362: of various interstellar environments.  In the limit that $j_\nu/n_H$
363: is a constant with $n_H$, then it equals $I_\nu/N(H)$.  However, in
364: our case it is not.  As a reference model, we take the DL98 WIM model
365: evaluated for $n_H=0.1~\cm^{-3}$ which peaks near 25 GHz with a
366: peak emissivity\footnote{The units of the ordinate, $j_\nu/n_H$, 
367: of Figs. 9-11 in DL98 is incorrectly labeled (Jy sr$^{-1}$) 
368: rather than (Jy cm$^2$\,sr$^{-1}$), an error propagated by Finkbeiner
369: et al. (2002).}
370: of $j_{p, {\rm ref}}/n_H = 5\times 10^{-21} \kjysr \cm^2$
371: or $j_{p, {\rm ref}} = 5\times10^{-22} \kjysr \cm^{-1}$. 
372: Multiplying the dependence on dust and gas density we have\footnote{
373: Again, the dependence on $n_H$ in \refeq{episod} is only strictly correct in the 
374: episodic limit, though in reality the deviation is likely small.
375: }
376: \bel{episod}
377: j_{p} = j_{p, {\rm ref}} \left(\frac{n_H}{n_{H,{\rm ref}}}\right)
378: \left(\frac{n_d}{n_{d,{\rm ref}}}\right)
379: \ee
380: where $n_{H,{\rm ref}}$ and $n_{d,{\rm ref}}$ are the density of H and small dust
381: grains, respectively, used in the reference model.  For full
382: ionization of H ($n_H = n_i$) and constant gas/dust ratio
383: ($n_H/n_{H,{\rm ref}} = n_d/n_{d,{\rm ref}}$) we have
384: \begin{eqnarray}
385: j_{p} = j_{p, {\rm ref}} \left(\frac{n_H}{0.1\,\cm^{-3}}\right)^2 \\
386: =5\times10^{-20} n_H^2 \kjysr \cm^5 \\
387: =0.15 n_H^2 \kjysr \cm^6\pc^{-1}
388: \end{eqnarray}
389: and the peak specific intensity is then 
390: \begin{equation}
391: I_{p}=\int j_{p} dl = 0.15 \kjysr \left(\frac{E_m}{1\,\cm^{-6}\pc}\right)
392: \end{equation}
393: or
394: \begin{equation}
395: I_{\rm p}/I_{\mbox{\hal}} = 0.09\,\kjysr\,\Ray^{-1},
396: \end{equation}
397: where $I_{\mbox{\hal}}$ is measured in R.  For comparison, the ratio
398: of free--free to \hal\ for a temperature of $T_e\sim 8000$K at 41 GHz
399: (WMAP Q-band) is $I_{41}/I_{\mbox{\hal}} \sim
400: 0.15\,\kjysr\,\Ray^{-1}$.
401: 
402: Although this is a rough estimate, it suggests that the intensity of
403: the free--free emission at 41 GHz and the peak of the spinning dust
404: emission should be of the same order of magnitude.
405: Throughout this discussion we make the assumption that the DL98 WIM
406: spectrum can be shifted by a factor of 2 in frequency.  A more
407: thorough analysis would require that the DL98 model be evaluated for
408: some set of parameters (ion density, electric dipole moment per root
409: mass, etc.) that give the actual spectrum observed.  For our purposes
410: this naive shift is adequate. 
411: 
412: \bp
413: \begin{center}
414:   \includegraphics[width=0.375\textwidth]{f1a.eps}
415:   \includegraphics[width=0.375\textwidth]{f1b.eps}
416:   \includegraphics[width=0.375\textwidth]{f1c.eps}
417: \caption{
418: \emph{Top panel:} The best $F+D+C$ (free--free plus spinning dust plus CMB) 
419: model fit to the \hal-correlated emission over the full sky.  The peak frequency 
420: is fixed to $\nu_{\rm peak} = 50$ GHz.  This two component plus CMB 
421: contamination model works surprisingly well with a spinning dust component that 
422: is 6.23 times weaker than the free--free (see \reftbl{fit-res}).  \emph{Middle 
423: panel:} $\Delta\chi^2$ contours for the best fit free--free coefficient in the 
424: spinning dust/CMB plane.  The null hypothesis that the \hal-correlated emission 
425: can be well fit by only a $F+C$ model (i.e., $D_0 = 0$) is ruled at very high 
426: confidence ($335\sigma$).  \emph{Bottom panel:} $\chi^2$ contours in the 
427: spinning dust/$\nu_{\rm peak}$ plane.  Though the peak frequency is not very 
428: well constrained, we can rule out $\nu_{\rm peak} < 40$ GHz and $\nu_{\rm peak} 
429: > 60$ GHz at $>4\sigma$ confidence.
430: }\label{fig:hdc-fs}
431: \end{center}
432: \ep
433: 
434: 
435: 
436: \bpm
437: \centerline{
438:   \includegraphics[width=0.23\textwidth]{f2a.eps}
439:   \includegraphics[width=0.23\textwidth]{f2b.eps}
440:   \includegraphics[width=0.23\textwidth]{f2c.eps}
441:   \includegraphics[width=0.23\textwidth]{f2d.eps}
442: }
443: \centerline{
444:   \includegraphics[width=0.23\textwidth]{f2e.eps}
445:   \includegraphics[width=0.23\textwidth]{f2f.eps}
446:   \includegraphics[width=0.23\textwidth]{f2g.eps}
447:   \includegraphics[width=0.23\textwidth]{f2h.eps}
448: }
449: \centerline{
450:   \includegraphics[width=0.23\textwidth]{f2i.eps}
451:   \includegraphics[width=0.23\textwidth]{f2j.eps}
452:   \includegraphics[width=0.23\textwidth]{f2k.eps}
453:   \includegraphics[width=0.23\textwidth]{f2l.eps}
454: }
455: \caption{
456: The same as \reffig{hdc-fs} except for the four regions of the Gum Nebula (see 
457: \reftbl{fit-res}).  Although the peak frequency is only loosely constrained, the 
458: $D_0=0$ null hypothesis is ruled out at $\geq 12\sigma$ in all regions.
459: }\label{fig:hdc-rg}
460: \epm
461: 
462: 
463: 
464: 
465: %------------------------------------------------------------------------------
466: % SECTION: Spectrum
467: %------------------------------------------------------------------------------
468: \section{Spectrum of \hal-correlated emission}
469: \label{sec:spectra}
470: 
471: As noted above, in DF07 we showed that the spectrum of \hal-correlated 
472: emission in the WMAP data has a bump around $\nu \sim 50$ GHz that we now argue 
473: is consistent with a spinning dust contamination of the familiar \refeq{ff-spec} 
474: free--free spectrum.  We also point out in DF07 that any measurement of a 
475: foreground spectrum is necessarily contaminated by a component which has the 
476: frequency dependence of the CMB due to imperfect foreground removal when 
477: constructing a CMB template.\footnote{Throughout this paper, we will work with 
478: the results found using our CMB5 estimator for the CMB -- see DF07 for details.}  
479: However, as we will show shortly, this contamination is incapable of producing 
480: the observed bump.
481: 
482: \reffig{hdc-fs} shows our full sky fit for \hal-correlated emission.  Our 
483: hypothesis is that this spectrum can be well fit by a linear combination of 
484: free--free, spinning dust, and CMB spectra.  Thus, our model spectrum is
485: \bel{mod}
486:   I_{\nu}^{\rm mod} = F + D + C,
487: \ee
488: where
489: \bel{ff-comp}
490:   F = F_0 \left(\frac{\nu}{23\mbox{ GHz}}\right)^{-0.15},
491: \ee
492: \bel{sd-comp}
493:   D = D_0 \times \mbox{ (DL98 WIM model with $\nu_{p} = 50$ GHz)},
494: \ee
495: and
496: \bel{cmb-comp}
497:   C = C_0 \left(\frac{\nu}{23\mbox{ GHz}}\right)^2.
498: \ee
499: We minimize $\chi^2 = \sum_{i} (I_{\nu_i}^{\rm mod} - I_{\nu_i})^2/\sigma_i^2$ 
500: over the parameters $F_0$, $D_0$, and $C_0$ which are scaled to be of order 
501: unity.  Here $\sigma_i$ are the formal errors in band $i$ of the spectrum from 
502: our DF07 template fit, and we have one degree of freedom.
503: 
504: As shown in \reffig{hdc-fs}, the spectral model in \refeq{mod} fits the data 
505: astonishingly well.  The free--free and spinning dust amplitudes are comparable 
506: while the contamination from the CMB is minimal.  The very low value of 
507: $\chi^2=1.84$ reflects the quality of the fit. \reffig{hdc-fs} also shows 
508: $\Delta\chi^2=$ 1,2,3,4, and 5 contours in the $D_0$-$C_0$ plane minimizing 
509: $\chi^2$ over $F_0$ at each point, and holding $\nup$ fixed at 50 GHz.  The 
510: null hypothesis that only a free--free plus CMB spectrum (i.e., $D_0 = 0$) fit 
511: the data is ruled out to a very high significance.  Thus, either free--free 
512: emission does not follow \refeq{ff-spec}, or the \hal\ map is tracing
513: an additional emission component.  This other emission component seems
514: to be well fit by a shifted WIM spinning dust spectrum.
515: 
516: Lastly, the $\chi^2$ contours in the $D_0$-$\nu_{\rm peak}$ plane in the bottom 
517: panel of \reffig{hdc-fs} indicate that, although the peak frequency of the 
518: spinning dust is not very well constrained, values of $\sim 45-55$ GHz are 
519: consistent with the data.  Peak frequencies less than 40 GHz and greater than 
520: 60 GHz are ruled out at $4\sigma$.
521: 
522: \bp
523: \centerline{
524:   \includegraphics[width=0.23\textwidth]{f3a.eps}
525:   \includegraphics[width=0.23\textwidth]{f3b.eps}
526: }
527: \centerline{
528:   \includegraphics[width=0.23\textwidth]{f3c.eps}
529:   \includegraphics[width=0.23\textwidth]{f3d.eps}
530: }
531: \centerline{
532:   \includegraphics[width=0.23\textwidth]{f3e.eps}
533:   \includegraphics[width=0.23\textwidth]{f3f.eps}
534: }
535: \caption{
536: The same as \reffig{hdc-fs} except for a region which encloses the Ophiuchus 
537: Complex (left column) and one which is near the Orion Nebula (right column).  
538: See \reftbl{fit-res} for details.
539: }\label{fig:hdc-op-or}
540: \ep
541: 
542: \begin{deluxetable*}{|c|c|c|c|c|c|c|c|}
543: \tablehead{
544:  & Full sky & \multicolumn{4}{c|}{Gum Nebula} & Ophiuchus & Orion \\
545:  & & 1 & 2 & 3 & 4 & &
546: }
547: \startdata
548:  lrange & 0:360 & 235:256 & 256:280 & 235:256 & 256:280 & 334:13 & 195:230 \\ 
549:  \hline
550:  brange & -90:90 & -30:0 & -30:0 & 0:30 & 0:30 & 0:90 & -35:0 \\ 
551:  \hline
552:  $\chi^2$ & 1.84 & 0.17 & 1.10 & 1.08 & 0.26 & 0.17 & 1.52 \\ 
553:  \hline
554:  $\Delta\chi^2_{D=0}$ & 335.02 & 43.84 & 87.32 & 17.59 & 12.09 & 11.89 & 8.30 \\ 
555:  \hline
556:  $ I_{F}/I_{D} (\nu=\nu_{\rm peak}) $ & 6.23 & 4.70 & 6.07 & 5.38 & 6.80 & 4.75 & 12.22
557: \enddata
558: \tablecomments{
559: Region definitions (in Galactic coordinates $l$ and $b$) and results of the fits 
560: presented in Figures \ref{fig:hdc-fs}-\ref{fig:hdc-op-or}.  The very low values 
561: of $\chi^2$ reflect both the quality of the fits and the small number of 
562: constraints (number of degrees of freedom = 2).  The $D_0=0$ null hypothesis is 
563: ruled out at $> 8\sigma$ in \emph{all} regions, while the relative intensity of 
564: free--free to spinning dust emission at $\nu_{\rm peak} = 50$ GHz is in the 
565: range 4.7-12.2.
566: }\label{tbl:fit-res}
567: \end{deluxetable*}
568: 
569: \bpm
570: \centerline{
571:   \includegraphics[width=0.75\textwidth]{f4.eps}
572: }
573: \caption{
574: WMAP K band data (\emph{top}) and the \hal\ map (\emph{bottom}) with our different 
575: fit regions shown as dotted boxes.  In addition to the full sky fit, the regions 
576: are the Gum Nebula (1,2,3,4), the Ophiuchus complex (5), and the Orion Nebula (6).  
577: See \reftbl{fit-res} for details.
578: }\label{fig:regions}
579: \epm
580: 
581: The next step is to consider smaller individual regions of the sky to show that 
582: the signal persists.  Figures \ref{fig:hdc-rg} and \ref{fig:hdc-op-or} show fits 
583: for four different regions of the Gum Nebula, the north central Galactic sky 
584: (including the Ophiuchus complex), and a region near the Orion Nebula (see 
585: \reftbl{fit-res} and \reffig{regions} for details). Though the fits are 
586: significantly noisier as a 
587: result of the smaller number of pixels used, the bump in 
588: the \hal-correlated emission remains in all regions, with the free--free
589: amplitude roughly 4-7 
590: times larger than the spinning dust.  Furthermore, the case of no spinning dust 
591: $D_0 = 0$ is ruled out in \emph{all} regions at $\geq 8\sigma$.  Although the 
592: $\chi^2$ contours in the $D_0$-$\nu_{\rm peak}$ plane offer a tantalizing hint 
593: that the peak frequency of the spinning dust is lower in the Orion Nebula region 
594: than the others, the present data are simply too noisy to make a statistically 
595: significant statement.
596: 
597: \begin{deluxetable}{|c|c|c|c|c|}
598: \tablehead{
599:    & H$\alpha$ & Dust & Haslam & Haze
600: }
601: \startdata
602:   H$\alpha$ &  1.00 &  0.45 &  0.18 &  0.10 \\
603:   Dust      &  0.45 &  1.00 &  0.56 &  0.26 \\
604:   Haslam    &  0.18 &  0.56 &  1.00 &  0.58 \\
605:   Haze      &  0.10 &  0.26 &  0.58 &  1.00
606: \enddata
607: \tablecomments{
608: The correlation matrix for the templates presented in \citet{DF07}.  These are 
609: the templates that were used in the multi-linear regression fit which revealed 
610: the bump in the \hal-correlated emission presented in Figures 
611: \ref{fig:hdc-fs}-\ref{fig:hdc-op-or}.
612: }\label{tbl:corrmat}
613: \end{deluxetable}
614: 
615: Lastly, we address the possibility that the bump in the \hal-correlated 
616: emission represents a cross-correlation bias between the different templates 
617: used in DF07.  That is to say, perhaps the \hal\ spectrum is absorbing power from 
618: other templates in the fit.  Table \ref{tbl:corrmat} shows the correlation 
619: matrix, defined as
620: \bel{corrmat}
621:   \Gamma_{i,j} = \frac{\langle T_i T_j \rangle}
622: 								{\sqrt{\langle T_i^2 \rangle\langle T_j^2 \rangle}}
623: \ee
624: where $\Gamma_{i,j}$ is the correlation between the $i$th and $j$th template.  
625: The correlation between the templates is rather small.  In particular, the dust 
626: map is the template which is most highly correlated with the \hal\ map, and even 
627: then, the variance in the \hal-correlated emission can be at most 
628: $\Gamma_{\mbox{\hal, dust}}^2 = 20\%$ due to correlation with dust.
629: 
630: \bp
631: \centerline{
632:   \includegraphics[width=0.24\textwidth]{f5a.eps}
633:   \includegraphics[width=0.24\textwidth]{f5b.eps}
634: }
635: \centerline{
636:   \includegraphics[width=0.24\textwidth]{f5c.eps}
637:   \includegraphics[width=0.24\textwidth]{f5d.eps}
638: }
639: \caption{
640: The intensity of the emission correlated with the four templates presented in 
641: \citet{DF07}.  The solid lines indicate fits for which the \hal-correlated 
642: emission was constrained to follow a $I_{ff} \propto \nu^{-0.15}$ (see 
643: \refeq{ff-spec}) frequency dependence, while the dotted lines indicate that the 
644: \hal-correlated spectrum was fit explicitly.  Both the insensitivity of the 
645: inferred dust, Haslam, and haze -correlated spectra to fixing the \hal\ spectrum 
646: as well as the relatively low correlation among the templates (\reftbl{corrmat}) 
647: indicates that the spinning dust bump is indeed correlated primarily with \hal\ 
648: emission.
649: }\label{fig:corr-hal}
650: \ep
651: 
652: In fact, \reffig{corr-hal} shows that the effects are much smaller.  Here, the 
653: solid lines show the resultant spectra for all foreground components both with 
654: and without fitting the \hal\ spectrum.  In the latter case the \hal\ emission was 
655: assumed to follow the free--free spectrum in \refeq{ff-spec}.  The features of 
656: these spectra are discussed at length in DF07, but the important point is that 
657: the other foregrounds are not significantly affected when the \hal\ spectrum is 
658: fixed.  This is strong evidence that the emission mechanism causing the bump in 
659: the \hal\ spectrum is highly correlated with the \hal\ template.
660: 
661: 
662: %------------------------------------------------------------------------------
663: % SECTION: Discussion
664: %------------------------------------------------------------------------------
665: \section{Discussion}
666: \label{sec:discussion}
667: 
668: In \citet{DF07} we identified a deviation from the classical
669: free--free (thermal bremsstrahlung) spectrum in the \hal-correlated
670: emission of the 3-year WMAP data.  In this paper we have argued that
671: the spectral data are consistent with a superposition of free--free
672: and spinning dust emission from a ``warm ionized medium'' \citep[WIM;
673: see][]{DL98b}.  Additionally, there is contamination by a component with the 
674: spectrum of the CMB due to chance morphological correlation between the CMB and 
675: the \hal\ map \citep[see][for details]{DF07}.
676: 
677: Since the intensity of both free--free and WIM spinning dust emission scale 
678: proportionately with the number density squared (free--free emission is 
679: generated by the collision of free electrons with ions while the WIM spinning 
680: dust emission comes from collision of ions with tiny dust grains), both should 
681: be traced by an emission measure map like the \hal\ map.
682: 
683: We have assumed throughout that the dust/gas ratio is constant and the
684: grain size distribution is the same everywhere, so that $I \propto \int
685: n_i n_d dl \propto \int n_i^2$. 
686: 
687: We have studied both the (nearly) full sky and smaller independent regions of 
688: interest including the Gum Nebula, the Oph complex, and the Orion Nebula.  
689: For all regions, the \hal-correlated emission is well fit by a linear 
690: combination of an $I_{\nu} \propto \nu^{-0.15}$ power law for free--free and a 
691: \citet{DL98b} WIM spinning dust model shifted in frequency by a factor of 2 and 
692: with only a minimal adjustment in amplitude.  Specifically, the spinning 
693: dust component is $\sim$ 4-7 times weaker than the free--free component.
694: 
695: The null hypothesis that a free--free only spectrum is consistent with the data 
696: is ruled out at $\geq 8\sigma$ confidence for all regions considered here, 
697: despite the substantial noise in the data.  Furthermore, we have shown that 
698: the bump in the \hal-correlated emission spectrum is \emph{not} due to chance 
699: cross correlation between the \hal\ map and the additional foreground templates 
700: presented in \citet{DF07}, indicating that it is indeed correlated primarily 
701: with \hal\ emission.
702: 
703: \citet{F02} listed a number of criteria by which the existence of
704: spinning dust could be established.  Among them were that the peak
705: frequency must shift as a function of the environment in a way at
706: least qualitatively consistent with DL98.  In this work we find in the
707: WMAP data a WIM-correlated component with a peak frequency of 50 GHz,
708: about a factor of two higher than the peak frequency of the spinning 
709: dust component that is spatially correlated with thermal dust emission (i.e. WNM 
710: or CNM spinning dust).  That this
711: frequency is higher is in accordance with the DL98 models, strongly
712: supporting the spinning dust hypothesis.  However, because the shift
713: is larger than expected, a careful assessment of the situation
714: requires that the reason for the size of the shift be determined.  If,
715: for example, the WIM dust has smaller grains, then a correlated
716: measurement of $R_V\equiv A_V/E_{B-V}$ using e.g. background stars
717: would complete the case for spinning dust. 
718: 
719: In summary, there is clearly an additional foreground component
720: correlated with \hal\ emission and its spectral shape is consistent
721: with spinning dust, though at a somewhat higher frequency than
722: expected.  However, the increased sensitivity, frequency coverage, and
723: finer spatial resolution of future missions like \emph{Planck} may provide the
724: necessary evidence.
725: 
726: We acknowledge helpful discussions with Bruce Draine, Gary Hinshaw,
727: Joanna Dunkley, et al.  Some of the results in this paper were derived
728: using HEALPix\footnote{The HEALPix web page is 
729: \texttt{http://healpix.jpl.nasa.gov}} \citep{gorski05}.
730: This research made use of the NASA Astrophysics Data System (ADS) and
731: the IDL Astronomy User's Library at Goddard\footnote{Available at
732: \texttt{http://idlastro.gsfc.nasa.gov}}.  DPF and GD are supported in
733: part by NASA LTSA grant NAG5-12972.
734: 
735: \begin{thebibliography}{}
736: 
737:   \bibitem[\protect\citeauthoryear{Bennett et al.}{2003}]{bennett03}
738:     Bennett C.L. et al., 2003, ApJS, 148, 97
739: 
740:   \bibitem[\protect\citeauthoryear{Boughn \& Pober}{2007}]{boughn07}
741:     Boughn S. P.\& Pober J. C. 2007, ApJ, 661, 938
742: 
743:   \bibitem[\protect\citeauthoryear{Broderick et al.}{2007}]{broderick07}
744:     Broderick A.E., Finkbeiner D.P, Afshordi N., \& Dobler G., in preparation
745: 
746:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{1997}]{deO97}
747:     de Oliveira-Costa A., Kogut A., Devlin M.J., Netterfield C.B., Page L.A., \& 
748: Wollack E.J., 1997, ApJ, 482, L17
749: 
750:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{1998}]{deO98}
751:     de Oliveira-Costa A., Tegmark M., Page L., \& Boughn S., 1998, ApJ, 509, L9
752: 
753:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{1999}]{deO99}
754:     de Oliveira-Costa A. et al., 1999, ApJ, 527, L9
755: 
756:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{2000}]{deO00}
757:     % ''Galactic Contamination in the QMAP Experiment''
758:     de Oliveira-Costa A. et al., 2000, ApJ, 542, L5
759: 
760:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{2002}]{deO02}
761:     de Oliveira-Costa A. et al., 2002, ApJ, 567, 363
762: 
763:   \bibitem[\protect\citeauthoryear{de Oliveira-Costa et al.}{2004}]{deO04}
764:     de Oliveira-Costa A. et al., 2004, ApJ, 606, L89
765: 
766:   \bibitem[\protect\citeauthoryear{Dennison et al.}{1998}]{dennison98}
767:     Dennison B., Simonetti J.H., \& Topasna G., 1998,
768:     Publ. Astron. Soc. Australia, 15, 147
769: 
770:   \bibitem[\protect\citeauthoryear{Dickinson et al.}{2003}]{dickinson03}
771:     % ``Towards a free--free template for CMB foregrounds''
772:     Dickinson C., Davies R. D., \& Davis R. J. 2003, MNRAS, 341, 369
773: 
774:   \bibitem[\protect\citeauthoryear{Dickinson et al.}{2006}]{dickinson06}
775:     Dickinson C., Casassus S., Pineda J. L., Pearson T. J.,
776:     Readhead A. C. S., \& Davies, R. D. 2006, ApJ, 643, L111
777: 
778:   \bibitem[\protect\citeauthoryear{Dobler \& Finkbeiner}{2007}]{DF07}
779:     Dobler G. \& Finkbeiner D.P, 2007, submitted to ApJ, arXiv:0712.1038
780: 
781:   \bibitem[\protect\citeauthoryear{Draine \& Lazarian}{1998a}]{DL98a}
782:     Draine B.T. \& Lazarian A., 1998a, ApJ, 494, L19
783: 
784:   \bibitem[\protect\citeauthoryear{Draine \& Lazarian}{1998b}]{DL98b}
785:     Draine B.T. \& Lazarian A., 1998b, ApJ, 508, 157
786: 
787:   \bibitem[\protect\citeauthoryear{Draine \& Lazarian}{1999}]{DL99}
788:     Draine B.T. \& Lazarian A., 1999, ApJ, 512, 740
789: 
790:   \bibitem[\protect\citeauthoryear{Erickson}{1957}]{erickson57}
791:     % ``A Mechanism of Non-Thermal Radio-Noise Origin''
792:     Erickson W. C. 1957, \apj, 126, 480
793:     
794:   \bibitem[\protect\citeauthoryear{Ferrara \& Dettmar}{1994}]{ferrara94}
795:     % ``Radio-emitting dust in the free electron layer of spiral galaxies:
796:     % Testing the disk/halo interface''
797:     Ferrara A., \& Dettmar R. -J. 1994, \apj, 427, 155
798:  
799:   \bibitem[\protect\citeauthoryear{Finkbeiner et al.}{2002}]{F02}
800:     Finkbeiner D.P., Schlegel D.J, Frank, C., \& Heiles C.,
801:     2002, ApJ, 566, 898
802: 
803:   \bibitem[\protect\citeauthoryear{Finkbeiner}{2003}]{finkbeiner03}
804:     Finkbeiner D.P., 2003, ApJS, 146, 407
805: 
806:   \bibitem[\protect\citeauthoryear{Finkbeiner}{2004}]{finkbeiner04}
807:     Finkbeiner D.P., 2004, ApJ, 614, 186
808: 
809:   \bibitem[\protect\citeauthoryear{Finkbeiner et al.}{1999}]{finkbeiner99}
810:     Finkbeiner D.P., Davis M., \& Schlegel D.J., 1999, ApJ, 524, 867
811: 
812:   \bibitem[\protect\citeauthoryear{Finkbeiner, Langston, \& Minter}{2004}]{gb04}
813:     % ``Microwave ISM Emission in the Green Bank Galactic Plane
814:     % Survey: Evidence for Spinning Dust''
815:     Finkbeiner D. P., Langston G. I., \& Minter A. H. 2004, ApJ, 617, 350
816: 
817:   \bibitem[\protect\citeauthoryear{Gaustad et al.}{2001}]{gaustad01}
818:     Gaustad J.E., McCullough P.R., Rosing W., \& Van Buren D., 2001, 
819:     PASP, 113, 1326
820: 
821:   \bibitem[\protect\citeauthoryear{G\'{o}rski et al.}{2005}]{gorski05}
822:     G\'{o}rski K.M. et al., 2005, ApJ, 622, 759
823: 
824:   \bibitem[\protect\citeauthoryear{Haslam et al.}{1982}]{haslam82}
825:     Haslam C.G.T., Stoffel H., Salter C.J., \& Wilson W.E., 1982, A\&AS, 47, 1
826: 
827:   \bibitem[\protect\citeauthoryear{Haffner et al.}{2003}]{haffner03}
828:     Haffner L.M., Reynolds R.J., Tufte S.L., Madsen G.J., Jaehnig K.P., \& 
829:     Percival J.W., 2003, ApJS, 149, 405
830: 
831:   \bibitem[\protect\citeauthoryear{Hinshaw et al.}{2007}]{hinshaw07}
832:     Hinshaw G. et al., 2007, ApJS, 170, 288
833: 
834:   \bibitem[\protect\citeauthoryear{Hooper et al.}{2007}]{hooper07}
835:     Hooper D., Finkbeiner D.P., \& Dobler G., 2007, arXiv:0705.3655
836: 
837:   \bibitem[\protect\citeauthoryear{Kogut et al.}{1996}]{kogut96}
838:     Kogut A. et al., 1996, ApJ, 464, L5
839: 
840:   \bibitem[\protect\citeauthoryear{Langston et al.}{2000}]{langston00}
841:     % ``The First Galactic Plane Survey at 8.35 and 14.35 GHz''
842:     Langston, G., Minter, A., D'Addario, L., Eberhardt, K., Koski, K., \&
843:     Zuber, J. 2000, AJ, 119, 2801
844:     
845:   \bibitem[\protect\citeauthoryear{Leitch et al.}{1997}]{leitch97}
846:   % ``An Anomalous Component of Galactic Emission''
847:     Leitch, E. M., Readhead, A. C. S., Pearson, T. J., \& Myers, S. T.
848:     1997, \apj, 486, L23
849: 
850:   \bibitem[\protect\citeauthoryear{Page et al.}{2007}]{page07}
851:     Page, L. 2007
852: 
853:   \bibitem[\protect\citeauthoryear{Scaife et al.}{2007}]{scaife07}
854:     Scaife A.M.M et al., 2007, arXiv:0711.3120
855: 
856:   \bibitem[\protect\citeauthoryear{Schlegel et al.}{1998}]{schlegel98}
857:     Schlegel D.J., Finkbeiner D.P., \& Davis M., 1998, ApJ, 500, 525
858: 
859:   \bibitem[\protect\citeauthoryear{Spergel et al.}{2003}]{spergel03}
860:     Spergel D.N. et al., 2003, ApJS, 148, 175
861: 
862:   \bibitem[\protect\citeauthoryear{Spergel et al.}{2007}]{spergel07}
863:     Spergel D.N. et al., 2007, ApJS, 170, 377
864: 
865:   \bibitem[\protect\citeauthoryear{Spitzer}{1978}]{spitzer}
866:     % ``Physical Processes in the Interstellar Medium''
867:     Spitzer, L. 1978, \emph{Physical Processes in the Interstellar Medium},
868:     Wiley, New York
869: 
870: 
871: \end{thebibliography}
872: 
873: \end{document}
874: