astro-ph0605362/ms.tex
1: \documentclass[prd,preprint,nofootinbib]{revtex4}
2: %\documentclass[prd,twocolumn,nofootinbib]{revtex4}
3: \usepackage{graphicx}
4: \usepackage{amssymb}
5: 
6: \begin{document}
7: %\begin{flushright}
8: %11 May 2006 Draft
9: %\end{flushright}
10: 
11: \title{Limit on the Neutrino Mass from the WMAP Three Year Data}
12: 
13: \author{  Masataka Fukugita, Kazuhide Ichikawa, Masahiro Kawasaki}
14: \affiliation{
15: Institute for Cosmic Ray Research, University of Tokyo,
16: Kashiwa 277 8582, Japan}
17: \author{Ofer Lahav}
18: \affiliation{Department of Physics and Astronomy, University College London, Gower Street, 
19: London WC1E 6BT, United Kingdom}
20: 
21: \date{\today}
22: 
23: \vskip10mm
24: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
25: %: abstract
26: \begin{abstract}
27: We derive an upper limit on the neutrino mass from the WMAP three-year
28: data alone by employing a deterministic minimisation method based on a
29: grid search in multi-parameter space. Assuming the flat $\Lambda$CDM
30: model with power-law adiabatic perturbations, we find $\sum m_{\nu} < 2.0$ 
31: eV in agreement with the result of the WMAP team.  This
32: result, the limit being nearly the same as that from the WMAP
33: first-year data, means that the fundamental limit on the neutrino mass
34: obtainable from the cosmic microwave background alone is already nearly
35: met, as anticipated from the previous analysis. We also clarify the
36: role of the polarisation data in deriving the limit on the neutrino
37: mass.
38: \end{abstract}
39: \maketitle
40: 
41: %%%%%%%%%%%%%%%%%%%%%%%%%%%%
42: 
43: %In the previous publication we pointed out that a valid limit on the
44: %neutrino mass can be derived from the cosmic microwave background
45: %(CMB) experiment alone \cite{Ichikawa:2004zi}. 
46: In \cite{Ichikawa:2004zi} it was shown that a valid limit on the
47: neutrino mass can be derived from the cosmic microwave background
48: (CMB) experiment alone.
49: %%
50: This contrasted to the then believed view
51: that such a limit is obtained only when the CMB data are combined with
52: those from large-scale clustering of galaxies 
53: \cite{Tegmark:2003ud,Spergel:2003cb}: 
54: the limit given
55: by Tegmark et al \cite{Tegmark:2003ud} derived from the first year  
56: data of the {\it
57: Wilkinson Microwave Anisotropy Probe} (WMAP-1) \cite{Spergel:2003cb}  
58: is $\sum m_\nu<11$
59: eV at a 95\% confidence. This implies that massive neutrinos alone
60: may constitute the entire dark matter of the Universe. This conclusion
61: was supported by the analysis of Elgar\o y and Lahav \cite 
62: {Elgaroy:2003yh}.
63: %We argued that if neutrinos are already
64: %nonrelativistic at the time of recombination, there appears a
65: %shift induced by massive neutrinos in the CMB observables that
66: %cannot be absorbed into those of cosmological parameters.
67: Our limit from WMAP-1
68: %the first year data of the {\it Wilkinson Microwave
69: %Anisotropy Probe} (WMAP) data [4]
70: alone, $\sum m_\nu<2.0$ eV (95\% CL)
71: obtained from a deterministic grid search, contradicts with that
72: of \cite{Tegmark:2003ud}. We ascribed this discrepancy to the  
73: possibility that
74: Markov chain Monte Carlo used in \cite{Tegmark:2003ud} was not long  
75: enough to sample
76: the likelihood function in the presence of strong parameter degeneracy.
77: 
78: The argument given by \cite{Elgaroy:2003yh} is based on the well-known 
79: fact that massive
80: neutrinos diminishes the small-scale power by free-streaming and this
81: leads to a limit on the neutrino mass. In this argument the fact was
82: not taken into account that massive neutrinos affect 
83: the CMB perturbations not only
84: by the modification of the power spectrum but also in some 
85: not so obvious ways: characteristically, the acoustic length scale
86: and the relative
87: heights of peaks are modified.  The shift of the position of the
88: acoustic peaks and the height of the first peak alone can be absorbed  into
89: the change of cosmological parameters, but
90: when the heights of the second and third peaks relative to the first
91: are modified, the shifts cannot be absorbed into that of the
92: cosmological parameters, leading to a limit on the neutrino mass.
93: This means that a limit on neutrino mass
94: could be obtained from the observed CMB multipoles only if neutrinos are
95: nonrelativistic at the recombination epoch, i.e., $m_\nu>0.6$ eV.
96: Several numerical analyses made recently seem to support this conclusion
97: \cite{MacTavish:2005yk,Hannestad:2006zg,Lesgourgues:2006nd}.
98: 
99: At the data release of the WMAP team for their three year observations
100: (WMAP-3) \cite{Spergel:2006hy}, they gave a limit on the  
101: neutrino mass $\sum
102: m_\nu<2.0$ eV (at a 95\% confidence) from their CMB data alone, using
103: a Markov chain Monte Carlo analysis. This agrees with our conclusion
104: in \cite{Ichikawa:2004zi}, where it was shown that the limit from  
105: WMAP-1, $\sum
106: m_\nu<2.0$ eV, could hardly be improved even with the increasing
107: quality of the CMB data.  In this report we attempt to verify this
108: explicitly by repeating the previous analysis with the WMAP-3 data using
109: the deterministic grid search algorithm. Incidentally, we try to
110: elucidate the role of the polarisation data if they could tighten the
111: constraint on the neutrino mass.
112: 
113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
114: 
115: \begin{figure}
116: \begin{center}
117: \includegraphics[width=15cm] {chi2min}
118: \end{center}
119: \caption{The curve of $\chi^2(\omega_\nu)$ as a function
120: of the neutrino mass summed over generations. The solid
121: curve uses all available data from WMAP including
122: polarisations. The dotted curve is a constraint from only 
123: temperature-temperature correlations.} 
124: \label{fig:chi2}
125: \end{figure}
126: 
127: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
128: 
129: We take the flat $\Lambda$CDM model with adiabatic perturbations of
130: the power law spectrum. The model contains 6 cosmological parameters,
131: the mass density parameters ($\omega_i=\Omega_i h^2$) for cold dark
132: matter, baryonic matter and massive neutrinos, the normalised Hubble
133: constant $h$, the reionization optical depth $ \tau$, the scalar
134: spectral index of primordial mass perturbations $n_s$ and their
135: amplitude $A$. The neutrino mass density parameter $\omega_\nu$ is
136: related to neutrino masses by $\omega_ \nu = \sum m_\nu/$(94.1 eV) 
137: (the Fermi distribution is assumed) and
138: we assume three generations of massive neutrinos with a degenerate
139: mass.  Theoretical CMB power spectrum is calculated using the CMBFAST
140: code \cite{Seljak:1996is} and $\chi^2$ by the likelihood code of the
141: WMAP three-year data release 
142: \cite{Jarosik:2006ib,Hinshaw:2006ia,Page:2006hz}.
143: We calculate the $\chi^2$ function for a given 
144: $\omega_\nu$ by minimizing it over the 6 parameters. The
145: minimization is carried out with the grid search using the Brent
146: method \cite{brent} extended to multi-parameters, as described in
147: \cite{Ichikawa:2004zi}.
148: 
149: The result is shown in Fig.~\ref{fig:chi2}. The logarithm of the
150: likelihood  $-2\ln\cal{L}$ for the power spectra
151: including all TT, TE, EE and BB with the pixel-based method for low  
152: multipoles is shown by the solid line.
153: The $\chi^2_{\rm eff}$ for 3162 degrees of freedom at the vanishing  
154: neutrino mass is
155: 1.037 for the parameters $\omega_b=0.0221$, $\omega_m=0.127$, $h=0.725$,
156:  $ \tau=0.091$ and $n_s=0.957$ in agreement with the WMAP-3  
157: analysis.
158: 
159: The limit corresponding to
160: $\Delta \chi^2 = 4$ (measured from the minimum) is $\omega_\nu \le
161: 0.024$ or $\sum m_\nu \le 2.3$ eV. We obtain the upper limits at a
162: 95\% confidence by an integration of the likelihood function ${\cal L}
163: = \exp \{-\Delta \chi^2/2 \}$,
164: \begin{eqnarray}
165: \omega_\nu \le 0.0215 \qquad {\rm or} \qquad \sum m_\nu \le
166: 2.0 \ {\rm eV},
167: \label{eq:limit}
168: \end{eqnarray}
169: which corresponds to $m_\nu<0.67$ eV. A slight difference between
170: the limit from $\Delta \chi^2 = 4$ and likelihood integral is due to
171: the $\chi^2$ curve flatter than quadratic at small neutrino masses.
172: The actual minimum of $\chi^2$ occurs at a non-zero neutrino mass
173: $\sum m_\nu \approx 1.3$ eV, but $\chi^2$ relative to the vanishing
174: neutrino mass is less than one, meaning that the preference of
175: a finite neutrino mass is insignificant.
176: 
177: The limit we obtained agrees with that from the WMAP team, endorsing  
178: the validity of their
179: Markov chain Monte Carlo analysis. We may also compare
180: eq.~(\ref{eq:limit}) with that from WMAP-1 \cite{Ichikawa:2004zi}: $  
181: \omega_\nu \le 0.023$.
182: The limit changes little in spite of the significantly increased accuracy
183: of the CMB data, which agrees with the analysis given in \cite 
184: {Ichikawa:2004zi}.
185: This is close to the fundamental limit that could be obtained from
186: the CMB alone even if the data quality would be higher. A further
187: improvement of the limit needs external constraints, such as those 
188: from galaxy clustering.
189: 
190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
191: \begin{figure}
192: \begin{center}
193: \includegraphics[width=12cm] {Cl_shift}
194: \end{center}
195: \caption{TT, TE and EE correlators
196: for three cases: $\omega_\nu =$0 (the solid line) and  0.02 (the dashed
197: line) with the other parameters
198: fixed to the WMAP mean values. In the right panels
199: the abscissa is rescaled according to the
200: modification of the acoustic length that varies
201: with the neutrino mass.}
202: \label{fig:TE}
203: \end{figure}
204: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
205: 
206: We now briefly discuss the role of the polarisation data in deriving
207: the limit on the neutrino mass. We show in Figure 1 above the
208: $\chi^2$ curve for the TT correlation only (the dotted curve), which
209: gives $\omega_\nu \le 0.0250$ or $\sum m_\nu \le 2.4$ {\rm eV}, slightly
210: weaker than eq. (\ref{eq:limit}).  We
211: note, however, that there is no particular sensitivity in 
212: the TE or EE correlation to
213: the inclusion of the neutrino mass beyond that of the TT
214: correlation data.
215: The left column of the six panel figure, Figure \ref{fig:TE}, 
216: shows the TT, TE and EE correlators for $\omega_\nu=0$ and 0.02 with
217: the other cosmological parameters fixed. 
218: In the right-column panels we rescaled the abscissa in the way
219: that accounts for the modification of the acoustic length scale 
220: by a finite neutrino mass. The two curves almost coincide 
221: for all correlators, TT, TE and EE.  The WMAP data for TT are accurate
222: enough to marginally distinguish the two curves at approximately 
223: a 95\% confidence level. The data involving polarisation, however, 
224: do not reach this accuracy, so that they do not give
225: new information as to the neutrino mass. We do not expect
226: any effect of the neutrino mass that modifies polarisation
227: other than those through the scales in acoustic dynamics.  
228: The slight improvement on
229: the limit of the neutrino mass upon the inclusion of the polarisation
230: data, as we saw above, 
231: arises from the tightened constraint on $\tau$.  There is a
232: significant negative correlation between $\omega_\nu$ and $\tau$
233: (see  Figure 2d in \cite{Ichikawa:2004zi}). 
234: The absence of the polarisation data makes
235: $\tau$ more uncertain and rather
236: drives  
237: the $\chi^2$ minimum
238: to $\tau\approx 0$, which in turn pushes the neutrino mass towards a larger
239: value. In this way, the polarisation data serve to 
240: slightly improve the limit on the neutrino
241: mass through a constraint on the reionisation optical depth.
242: 
243: Our final remark is that the limit from the CMB data alone is the most
244: robust result in view of the 
245: uncertainties in the current large-scale
246: galaxy clustering data. They drive the best fit parameter of the
247: matter density either way depending upon whether one takes the SDSS 
248: or 2dFGRS data \cite{Spergel:2006hy}: the limit on the neutrino 
249: mass is also sensitive to
250: the choice of the galaxy clustering data between the two \cite{Fukugita06}
251: as well as how much
252: weight is given to these data.
253: To go beyond the present neutrino mass limit it would be imperative 
254: to combine external data, but this needs proper understanding 
255: of systematic errors involved in them.
256: 
257: 
258: 
259: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
260: %\section{Summary}
261: %\label{sec:summary}
262: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
263: 
264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
265: \begin{thebibliography}{}
266: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
267: 
268: %\cite{Ichikawa:2004zi}
269: \bibitem{Ichikawa:2004zi}
270:     K.~Ichikawa, M.~Fukugita and M.~Kawasaki,
271:     %``Constraining neutrino masses by CMB experiments alone,''
272:     Phys.\ Rev.\ D {\bf 71}, 043001 (2005)
273: %  [arXiv:astro-ph/0409768].
274: 
275: %\cite{Tegmark:2003ud}
276: \bibitem{Tegmark:2003ud}
277: M.~Tegmark {\rm et al.}  [SDSS Collaboration],
278: Phys.\ Rev.\ D {\bf 69}, 103501 (2004)
279: %[arXiv:astro-ph/0310723].
280: 
281: %\cite{Spergel:2003cb}
282: \bibitem{Spergel:2003cb}
283: D.~N.~Spergel {\rm et al.},
284: Astrophys.\ J.\ Suppl.\  {\bf 148}, 175 (2003)
285: %[arXiv:astro-ph/0302209].
286: 
287: %\cite{Elgaroy:2003yh}
288: \bibitem{Elgaroy:2003yh}
289: \O.~Elgar\o y and O.~Lahav,
290: JCAP {\bf 0304}, 004 (2003)
291: %[arXiv:astro-ph/0303089].
292: 
293: %\cite{MacTavish:2005yk}
294: \bibitem{MacTavish:2005yk}
295:     C.~J.~MacTavish {\it et al.},
296:     %``Cosmological parameters from the 2003 flight of BOOMERANG,''
297:     arXiv:astro-ph/0507503.
298: 
299: %\cite{Hannestad:2006zg}
300: \bibitem{Hannestad:2006zg}
301:     S.~Hannestad,
302:     %``Primordial neutrinos,''
303:     arXiv:hep-ph/0602058.
304: 
305: %\cite{Lesgourgues:2006nd}
306: \bibitem{Lesgourgues:2006nd}
307:     J.~Lesgourgues and S.~Pastor,
308:     %``Massive neutrinos and cosmology,''
309:     arXiv:astro-ph/0603494.
310: 
311: %\cite{Spergel:2006hy}
312: \bibitem{Spergel:2006hy}
313:     D.~N.~Spergel {\it et al.},
314:     %``Wilkinson Microwave Anisotropy Probe (WMAP) three year results:
315:     %Implications for cosmology,''
316:     arXiv:astro-ph/0603449.
317: 
318: %\cite{Seljak:1996is}
319: \bibitem{Seljak:1996is}
320: U.~Seljak and M.~Zaldarriaga,
321: Astrophys.\ J.\  {\bf 469}, 437 (1996)
322: %[arXiv:astro-ph/9603033].
323: 
324: %\cite{Jarosik:2006ib}
325: \bibitem{Jarosik:2006ib}
326:     N.~Jarosik {\it et al.},
327:     %``Three-year Wilkinson Microwave Anisotropy Probe (WMAP) 
328:     %profiles, data processing, radiometer characterization and   
329:     %limits,''
330:     arXiv:astro-ph/0603452.
331: 
332: %\cite{Hinshaw:2006ia}
333: \bibitem{Hinshaw:2006ia}
334:     G.~Hinshaw {\it et al.},
335:     %``Three-year Wilkinson Microwave Anisotropy Probe (WMAP)   
336:     %Temperature analysis,''
337:     arXiv:astro-ph/0603451.
338: 
339: %\cite{Page:2006hz}
340: \bibitem{Page:2006hz}
341:     L.~Page {\it et al.},
342:     %``Three year Wilkinson Microwave Anisotropy Probe (WMAP)   
343:     %Polarization analysis,''
344:     arXiv:astro-ph/0603450.
345: 
346: \bibitem{brent}
347: R. P. Brent, {\it Algorithms for Minimization without Derivatives}
348: (Prentice-Hall, Englewood Clifs, NJ, U.S.A. 1973); see also
349: W.~H.~Press, B.~P.~Flannery, S.~A.~Teukolsky and W.~T.~Vetterling,
350: {\it Numerical Recipes} (Cambridge University Press, New York, 1986)
351: 
352: \bibitem{Fukugita06}
353: M.~Fukugita, Nucl.Phys.Proc.Suppl. {\bf 155}, 10-17 (2006). 
354: 
355: 
356: \end{thebibliography}
357: 
358: \end{document}
359: 
360: 
361: 
362: 
363: 
364: