1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{epsfig}
3: \usepackage{bm}
4:
5: \slugcomment{KSUPT-04/2 \hspace{0.5truecm} March 2004}
6:
7: \shorttitle{Alfven waves and the CMB}
8: \shortauthors{Chen et al.}
9:
10: %\def\baselinestretch{2}
11:
12: \begin{document}
13:
14: \title{Looking for Cosmological Alfv\'en Waves in WMAP Data}
15:
16: \author{Gang Chen\altaffilmark{1}, Pia Mukherjee\altaffilmark{2},
17: Tina Kahniashvili\altaffilmark{1,3},
18: Bharat Ratra\altaffilmark{1}, and Yun Wang\altaffilmark{2}}
19: \altaffiltext{1}{Department of Physics, Kansas State University, 116
20: Cardwell Hall, Manhattan, KS 66506.}
21: \altaffiltext{2}{Department of Physics and Astronomy, University of Oklahoma,
22: 440 W. Brooks Street, Norman, OK 73019.}
23: \altaffiltext{3}{Center for Plasma Astrophysics, Abastumani Astrophysical
24: Observatory, A.~Kazbegi ave 2a, Tbilisi, 380060,
25: Republic of Georgia.}
26:
27: \begin{abstract}
28: A primordial cosmological magnetic field induces and supports vorticity
29: or Alfv\'en waves, which in turn generate cosmic microwave background (CMB)
30: anisotropies. A homogeneous primordial magnetic field with fixed direction
31: induces correlations between the $a_{l-1,m}$ and $a_{l+1,m}$ multipole
32: coefficients of the CMB temperature anisotropy field. We discuss the
33: constraints that can be placed on the strength of such a primordial
34: magnetic field using CMB anisotropy data from the WMAP experiment. We place
35: 3 $\sigma$ upper limits on the strength of the magnetic field of $B < 15$ nG
36: for vector perturbation spectral index $n=-5$ and $B<1.7$ nG for $n=-7$.
37: \end{abstract}
38:
39: \keywords{cosmic microwave background --- cosmology: observation ---
40: methods: statistical}
41:
42: \section{Introduction}
43:
44: The origin of the large-scale part of observed galactic magnetic fields, of
45: $\sim \mu$G (microGauss) strength and apparently coherent over $\sim 10$ kpc
46: scales, is unknown. They could be the consequence of nonlinear amplification
47: of a tiny seed field by galactic dynamo processes. An alternate possibility
48: is amplification of a weak seed field through anisotropic protogalactic
49: collapse and subsequent further amplification via galactic differential
50: rotation. In both cases a primordial seed field of strength exceeding
51: $10^{-13}$ to $10^{-12}$ G, coherent over $\sim $ Mpc scales, is apparently
52: needed, and it is often suggested that upto a $\sim $ nG strength seed
53: field might be required. See Kulsrud (1999), Widrow (2002), and Giovannini
54: (2003) for reviews of the state of the art in this area. A primordial
55: magnetic field of present strength $\sim $ nG can leave observable
56: signatures in the cosmic microwave background (CMB) anisotropy.
57:
58: In standard cosmologies vorticity perturbations decay and so do not
59: contribute to CMB temperature or polarization anisotropies. The
60: presence of a cosmological magnetic field generated during an earlier
61: epoch\footnote{
62: Quantum fluctuations during an early epoch of inflation can generate
63: a primordial nG magnetic field, coherent over very large scales (see,
64: e.g., Ratra 1992; Bamba \& Yokoyama 2004)}
65: changes this situation: a primordial magnetic field induces and supports
66: vorticity or Alfv\'en waves (Adams et al.~1996; Durrer, Kahniashvili,
67: \& Yates 1998, hereafter DKY). These vector perturbations generate CMB
68: anisotropies.\footnote{
69: In the future one can hope to constrain vector modes through their effect
70: on CMB anisotropy polarization anisotropy spectra. CMB polarization spectra
71: that result from vector perturbations due to a primordial magnetic field have
72: been discussed by Seshadri \& Subramanian (2001), Pogosian, Vachaspati, \&
73: Winitzki (2002), Mack, Kahniashvili, \& Kosowsky (2002), and Subramanian,
74: Seshadri, \& Barrow (2003), while Lewis (2004) considers the case of vector
75: modes supported by free-streaming neutrinos.}
76: The presence of a preferred direction due to a homogeneous magnetic field
77: background leads to an $m$ dependence of $\langle a_{lm}a_{lm}^*\rangle$,
78: and induces correlations between the $a_{l+1,m}$ and $a_{l-1,m}$ multipole
79: coefficients of the CMB temperature anisotropy field. Since the CMB
80: anisotropies are observed to be random
81: Gaussian\footnote{
82: Colley, Gott, \& Park (1996), Mukherjee, Hobson, \& Lasenby (2000), and
83: Park et al.~(2001) are some early discussions of the Gaussianity of the CMB
84: anisotropy. More recent discussions of the Gaussianity of the WMAP CMB
85: anisotropy data are in Komatsu et al.~(2003), Colley \& Gott (2003),
86: Chiang et al.~(2003), Park (2004), Eriksen et al.~(2004a, 2004b),
87: Coles et al.~(2004), Vielva et
88: al.~(2004), Copi, Huterer, \& Starkman (2003), Hansen et al.~(2004),
89: Gurzadyan et al.~(2004), and Mukherjee \& Wang (2004). The simplest inflation
90: models predict Gaussian fluctuations (see, e.g., Fischler, Ratra, \&
91: Susskind 1985; Ratra 1985), and this is consistent with most
92: observational indications (see, e.g., Peebles \& Ratra 2003). While there
93: are indications of mild peculiarities in some subsets of the WMAP data,
94: for example the apparent paucity of large-scale power (Spergel et al.~2003;
95: see G\'orski et al.~1998a for a similar indication from COBE data) and the
96: differences between data from different parts of the sky (see papers cited
97: above), foreground contamination (see, e.g., Park, Park, \& Ratra 2002;
98: Mukherjee et al.~2002, 2003; de Oliveira-Costa et al.~2003; Bennett et
99: al.~2003b; Tegmark, de Oliveira-Costa, \& Hamilton 2003) and other
100: systematics might be responsible for part of this.}, it is known that such a
101: contribution can only be subdominant.
102:
103: We use the observed $\overline{\langle a_{l-1,m}a_{l+1,m}^* \rangle}$
104: correlations measured by WMAP to place constraints on the strength of a
105: homogeneous primordial magnetic field. The angular brackets here denote
106: an ensemble average, and the overbar indicates an average over $m$ for
107: each $l$. Limited by cosmic variance uncertainties, this would be the useful
108: measure to characterize the signature of a homogeneous primordial magnetic
109: field.
110:
111: The model on which we base our analysis is introduced in \S 2. Our analysis
112: of the WMAP data and our results are discussed in \S 3. We conclude in \S 4.
113:
114:
115: \section{CMB temperature anisotropies generated by Alfv\'en waves}
116:
117: We assume that the homogeneous magnetic field ${\mathbf B}$ is generated
118: prior to the time of recombination. Such a field could be generated during
119: the electroweak phase transition (see, e.g., Vachaspati 1991; Sigl, Olinto,
120: \& Jedamzik 1997; Giovaninni \& Shaposhnikov 1998), or by an $\alpha$-effect
121: dynamo driven by collective neutrino-plasma interactions (Semikoz \&
122: Sokoloff 2004). The energy density of this field, $B^2/(4\pi)$, must be
123: small, to prevent a violation of the
124: cosmological principle, and so may be treated as a first order perturbation.
125: Accounting for the high conductivity of the primordial non-relativistic
126: plasma (with $v \ll 1$ where ${\bf v}$ is the velocity field of the plasma),
127: we may use the infinite conductivity, frozen-in condition, ${\bf E} + {\bf v
128: \times B} = 0$. We also assume that charged particles are tightly coupled
129: to the radiation. We write ${\bf B}={\bf B_0} + {\bf B_1}$, where ${\bf B_1}$
130: denotes the first order vector perturbation in the magnetic field
131: (where ${\bf \nabla \cdot B_1}=0$), and ${\bf v}= 0 + {\bf \Omega}$ with
132: ${\bf \Omega}$ being the first order vector perturbation in the fluid
133: velocity (where ${\bf \nabla \cdot \Omega}=0$).
134:
135: As a consequence of magnetic flux conservation the field lines in an
136: expanding universe are conformally diluted, $B_0 \propto 1/a^2$, and
137: the Alfv\'en velocity in the photon-baryon plasma during the photon
138: dominated epoch when the energy density $\rho_R=\rho_{\gamma} + \rho_b
139: \simeq \rho_{\gamma}$ until recombination, is $v_A=B_0/\sqrt{4\pi(\rho_R+p_R)}
140: = 4 \times 10^{-4} (B_0/10^{-9}\mbox{G})$, and is time independent.
141: Rescaling physical quantities according to the expansion of the universe,
142: the MHD equations result in an equation describing Alfv\'en wave propagation
143: with velocity $v_A ({\bf b \cdot {\hat k}}) \equiv v_A \mu$ (DKY)
144: \begin{equation}
145: \ddot{\bf {\Omega}}=v_A^2 ({\bf b \cdot k})^2 \bf {\Omega} .
146: \label{alfven-eq}
147: \end{equation}
148: Here ${\bf b}\equiv {\bf B_0}/B_0$ is the unit vector in the direction of
149: the magnetic field and an overdot denotes a derivative with respect to
150: conformal time $\eta$. Choosing only the sine mode, to satisfy the initial
151: condition ${\bf\Omega}({\bf k}, \eta=0)=0$, we have
152: \begin{equation}
153: {\bf \Omega}({\bf k}, \eta) \simeq {\bf \Omega_0} v_A k \mu \eta,
154: ~~~~~~~~~|{\bf \Omega}_0| = \frac{v_A}{B_0} |{\bf B}_1| .
155: \label{1alfven-sol}
156: \end{equation}
157: In eq. (\ref{alfven-eq}) we have neglected viscosity so the vorticity
158: solution in eq. (\ref{1alfven-sol}) is applicable only on scales bigger
159: than the damping scale. Assuming that the initial vector perturbation is
160: generated by a random process, the two-point correlation function of the
161: vorticity field can be written as (Pogosian et al.~2002)
162: \begin{equation}
163: \langle \Omega_{0i}^* ({\bf k}) \Omega_{0j}({\bf k^\prime}) \rangle
164: = \left[(\delta_{ij}-{\hat k}_i{\hat k}_j)S(k)
165: + i\epsilon_{ijl}{\hat k}_l A(k)\right]
166: \delta ({\bf k} - {\bf k^\prime}) .
167: \label{two-point}
168: \end{equation}
169: Here $\epsilon_{ijl}$ is the totally antisymmetric tensor, and the $S(k)$
170: and $A(k)$ power spectra describe the symmetric and helical parts of the
171: two-point correlation function. We assume that the spectra
172: $S(k)$ $(= |{\Omega}_0 (k)|^2)$ and $A(k)$ are given by simple power laws
173: of the scale $1/k$ on scales larger than the perturbation damping scale
174: $1/k_D$, i.e., for $k < k_D$
175: we take $S(k) = S_0 {k^n}/{k_D^{n+3}}$ and $A(k) = A_0 {k^m}/{k_D^{m+4}}$.
176: Here $S_0$ and $A_0$ are dimensionless normalization constants with
177: $S_0\geq A_0$, and $n$ and $m$ are spectral indexes. The cutoff scale $1/k_D$
178: is the scale below which the magnetic field is damped away, due to the
179: non-infinite value of the conductivity.
180: Using $T_{\rm dec} \sim 0.3\,{\rm eV}$
181: and $t_{\rm dec} \sim 10^{23}~$cm, the comoving magnetic field damping wave
182: number at decoupling is $k_D(t=t_{\rm dec}) \sim 3 \times 10^{-10}
183: {\mbox{cm}}^{-1}$ (DKY).
184:
185: The CMB fractional temperature anisotropy, in direction $\bf n$ on the sky,
186: induced by a vorticity perturbation (ignoring a possible dipole contribution
187: from the vector perturbation) is (DKY)
188: \begin{equation}
189: \frac{\Delta T}{T}({\bf n}, {\bf k}) \simeq {\bf n} \cdot {\bf \Omega}
190: ({\bf k}, \eta) = {\bf n} \cdot {\bf \Omega}_0 v_A \mu (k \eta_{dec}) .
191: \label{delta-T}
192: \end{equation}
193: Decomposing the CMB fractional temperature anisotropy in a spherical
194: harmonic expansion,
195: \begin{equation}
196: \frac{\Delta T}{T}({\bf n})=\sum_{l = 2}^\infty \sum_{m = -l}^l
197: a_{lm}Y_{lm}({\bf n})
198: \end{equation}
199: and using the definition of the power spectrum $C_l$,
200: \begin{equation}
201: \langle \frac{\Delta T}{T} ({\bf n}) \frac{\Delta T}{T} ({\bf n^\prime})
202: \rangle = \frac{1}{4\pi}\sum_{l = 2}^\infty (2l+1)C_lP_l({\bf n} \cdot
203: {\bf n^\prime}),
204: \label{power-spectra-def}
205: \end{equation}
206: where $P_l$ is the Legendre polynomial, we obtain, in the isotropic case,
207: $C_l = \langle a_{lm}^* a_{lm} \rangle$. Here the angular brackets
208: denote a theoretical (averaging) expectation value over an ensemble of
209: statistically identical universes. In Fourier space this expectation
210: value can be replaced by integration over all possible wavenumbers,
211: i.e., $\langle ... \rangle \rightarrow \int d^3 {\bf k}/(2\pi)^3$.
212:
213: Computing $\langle a_{lm}^* a_{l^\prime m^\prime} \rangle$, it can
214: be shown that the helical part of vorticity does not contribute (see
215: Pogosian et al.~2002). Hence in what follows we consider only the
216: symmetric part of the spectrum. Detailed computation of $C_l$'s for
217: vorticity perturbations are described in DKY where it has been shown
218: that the presence of a homogeneous magnetic field induces off-diagonal
219: correlations in multipole space, in particular correlations between $l$
220: and $l \pm 2$ multipole coefficients. To quantify this we introduce
221: a second power spectrum defined by
222: \begin{equation}
223: D_l(m) = \langle a_{l-1,m}^* a_{l+1, m} \rangle =
224: \langle a_{l+1,m}^* a_{l-1, m} \rangle .
225: \label{dlm}
226: \end{equation}
227: The two power spectra, $C_l(m)$ and $D_l(m)$, depend on the spectral
228: index $n$, the normalization constant $S_0$, the Alfv\'en velocity $v_A$
229: and the perturbation damping wavenumber $k_D$. The power spectra are
230: defined only for $n > -7$ (the quadrupole diverges at small $k$ for
231: $n \leq -7$), and for $n > -1$ the results are determined by
232: the damping wavenumber $k_D$. The case $n = -5$ corresponds to
233: the Harrison-Peebles-Yu-Zel'dovich scale-invariant spectrum ($C_l, D_l
234: \sim l^2$). See DKY for a more detailed discussion.
235:
236: The non-zero correlation of temperature for unequal $l$'s has a simple
237: physical explanation: The presence of a preferred spatial direction, that
238: of the magnetic field ${\bf B}_0$, breaks the spatial isotropy of the CMB
239: map, leading not only to an $m$ dependence of the correlators, but also
240: non-zero off-diagonal (in $l$ space) correlations.\footnote{
241: In a recent paper, Bershadskii \& Sreenivasan (2004) show that collisions
242: between Alfv\'en wave packets and their cascades could generate
243: arcminute-scale CMB temperature anisotropies, and argue that this is
244: consistent with the WMAP data.}
245: The temperature perturbation correlation between two points on the sky
246: depends not only on the angular separation between the two points, but also
247: on their orientation with respect to the magnetic field.
248:
249: A simple observational test to detect (or constrain) the presence of a
250: homogeneous magnetic field in the Universe is based on computing the $D_l$
251: spectrum of CMB anisotropy data.\footnote
252: {Other statistics, related to the $D_l$'s here, could also provide
253: useful tests (e.g., Hajian \& Souradeep 2003).}
254: For this it is useful to introduce the
255: arithmetic mean over $m$ of the two power spectra,
256: \begin{eqnarray}
257: \overline{C}_l &\equiv &\overline{\langle a_{lm}^* a_{lm} \rangle}=
258: \frac{1}{2l+1}\sum_{m=-l}^l \langle a_{lm}^* a_{lm} \rangle
259: \nonumber \\
260: \overline{D}_l &\equiv &\overline{\langle a_{l-1,m}^* a_{l+1,m}
261: \rangle}= \frac{1}{2l+1}\sum_{m=-l}^l \langle a_{l-1,m}^* a_{l+1,m}
262: \rangle .
263: \label{cl-dl-mean}
264: \end{eqnarray}
265: According to DKY,
266: \begin{eqnarray}
267: {\overline C}_l &\simeq& S_0\left({\eta_{\rm dec}\over
268: \eta_0}\right)^2(k_D \eta_0)^{-(n+3)}v_A^2{2^{n+1}\Gamma(-n-1)\over
269: 3\Gamma(-n/2)^2}l^{n+3} , ~~~n<-1 \label{Cbar}\\
270: {\overline C}_l/{\overline D}_l&=&|n+1|
271: \left[\frac{\Gamma\left(-\frac{n+1}{2}\right)}
272: {\Gamma\left(-\frac{n}{2}\right)}\right]^2 ~~~~
273: n<-1.
274: \label{Char++}\\
275: \overline{C}_l &\simeq& \overline{D}_l \simeq S_0\left({\eta_{\rm dec}\over
276: \eta_0}\right)^2(k_D \eta_0)^{-2}v_A^2{1\over n+1}l^2 ,
277: ~~~n>-1 \label{Cbar+}
278: \end{eqnarray}
279:
280: Using ${\bf B}_1\leq {\bf B}_0$, we have $|{\bf \Omega}_0|^2 k^3 \leq v_A^2$
281: (see eq.~[2]). This inequality must hold on all scales inside the Hubble
282: radius at decoupling, $k\ge 1/\eta_{\rm dec}$. With the $S(k)$ spectrum
283: definition in eq.~(3) we therefore get $2S_0(k/k_D)^{n+3}\leq v_A^2$ for
284: $1/\eta_{\rm dec} \le k \le k_D$, implying
285: \begin{eqnarray}
286: 2S_0(k_D \eta_{\rm dec})^{-(n+3)} &\leq& v_A^2 ~~~~~ n\le -3
287: \label{lim1},\\
288: 2S_0 &\leq& v_A^2 ~~~~~ n\ge -3 ~.\label{lim2}
289: \end{eqnarray}
290: So for $n\leq -3$ the result is independent of the damping wavenumber
291: $k_D$.
292:
293: We now estimate an upper limiting value of $l$, $l_C$, beyond which
294: our approximation is no longer valid. Just like scalar perturbations
295: (Peebles 1980), vector perturbations are affected by collisionless damping.
296: Adding a photon drag force term on the right hand side of the vorticity
297: equation (\ref{alfven-eq}), we can see that there are no
298: oscillations in the vector perturbation case, and that damping occurs
299: on scales slightly larger than the damping scale for scalar perturbations,
300: when $k_C\eta_{\rm dec} \sim 10$ (DKY), corresponding to $l_C \sim 500$,
301: beyond which our approximation breaks down.
302:
303: Inserting the limiting values given for $S_0$ in eqs.~(\ref{lim1}) and
304: (\ref{lim2}) in eq.~(\ref{Cbar}), we find\\
305: for $n = -5$,
306: \begin{eqnarray}
307: \overline{C}_l = 9.04 \times 10^{-16} l^{-2} \left(\frac{B}{1 {\rm nG}}
308: \right)^4 , ~~~~~~~~~ \overline{D}_l = \overline{C}_l/2.26 ,
309: \label{sol1}
310: \end{eqnarray}
311: and, for $n = -7$,
312: \begin{eqnarray}
313: \overline{C}_l = 8.61 \times 10^{-10} l^{-4} \left(\frac{B}{1 {\rm nG}}
314: \right)^4 , ~~~~~~~~~
315: \overline{D}_l = \overline{C}_l/2.17 .
316: \label{sol2}
317: \end{eqnarray}
318: In this paper, we use $n=-5$ and $n=-7$ as two illustrative cases.
319: These two cases are interesting as they correspond to a
320: Harrison-Peebles-Yu-Zel'dovich scale-invariant spectrum result for
321: the $C_l$'s and $D_l$'s, and to a possible inflation model primordial
322: vorticity field perturbation spectrum, respectively. These two cases also
323: span the range of constraints that can be placed on $B$ using this
324: method, in the range $-3 \geq n \geq -7$, i.e., $n=-5$ gives the weakest
325: and $n=-7$ the strongest constraint on $B$. With $k_D \eta_0 \sim 0.4 \times
326: 10^{14}$, from eqs.~(9) and (11) $B$ is not meaningfully constrained for
327: $n > -3$.
328:
329:
330: \section{Analysis and results}
331:
332: We use the foreground cleaned Q, V, and, W band co-added WMAP data (Bennett
333: et al.~2003a) to determine the off-diagonal correlations. The data are
334: available in the Healpix format (Gorski, Hivon, \& Wandelt 1998b) at
335: resolution $N_{\rm side}=512$. For each value of the
336: magnetic field amplitude $B$ we generate 5000 simulations of the CMB sky,
337: and each time apply the $Kp2$ Galactic cut mask prior to computing the
338: model $D_l$'s. The expected value of the model $D_l$'s, obtained from the
339: mean of these simulations, is then compared with the $D_l$'s obtained
340: similarly from the WMAP data, using the $\chi^2$ statistic. Confidence
341: levels on the field strength are derived from the resulting likelihood
342: function.
343:
344: Specifically, each simulation is a realization of the CMB with power spectrum
345: $C_l$ given by the best fit flat-$\Lambda$ CDM model with power-law primordial
346: power spectrum (Spergel et al.~2003), and with $D_l$ the same as that
347: predicted by eq.~(10) (or more specifically eqs.~[14] or [15]) for a given
348: value of $B$. In other words, we generate $a_{lm}$'s such that they satisfy
349: \begin{equation}
350: \langle a_{lm}^* a_{l^\prime m^\prime} \rangle =
351: \delta_{m,m^\prime}\left[ \delta_{l,l^\prime} C_l + (\delta_{l+1,l^\prime-1}
352: + \delta_{l-1,l^\prime+1}) \overline{D}_{l} \right],
353: \end{equation}
354: instead of
355: \begin{equation}
356: \langle a_{lm}^* a_{l^\prime m^\prime}\rangle = \delta_{m,m^\prime}
357: \delta_{l,l^\prime} C_l.
358: \end{equation}
359:
360: The $a_{lm}$'s are generated upto an $l_{\rm max}$ of 512 (corresponding to
361: Healpix resolution $N_{\rm side}=256$; since the WMAP data are expected
362: to contain useful cosmological information upto such an $l_{\rm max}$).
363: These are then convolved with the beam functions of each of the Q, V, and W
364: radiometer channels (8 in all) to produce 8 maps at Healpix resolution
365: $N_{\rm side}=512$ (because the noise maps have this resolution). Independent
366: Gaussian noise realizations of rms $\sigma_0/\sqrt{N_{\rm obs}}$ from WMAP are added
367: to the maps, and the 8 maps are co-added weighted by $N_{\rm obs}/\sigma_0^2$,
368: where the effective number of observations $N_{\rm obs}$ varies across the
369: sky, and $\sigma_0$ is different for each radiometer channel. This is how
370: each simulation is created. Hereafter the same analysis procedure that is
371: applied to the data map is applied to each of the simulations. This consists
372: of bringing the co-added map down to Healpix resolution $N_{\rm side}=256$
373: (since we mostly use $D_l$'s only upto an $l_{\rm max}$ of 300 in the
374: subsequent analysis), and applying the $Kp2$ sky cut prior to computing the
375: $D_l$'s.
376:
377: The whole analysis can be repeated for different values of the spectral
378: index $n$ that characterizes the spectrum of the magnetic field perturbations.
379:
380: Fig.~1 shows the $D_l$'s obtained from the WMAP data (crosses) and the median
381: and 68\% confidence range contours from simulations for two illustrative values of $B$.
382: The spread of about $10^{-3}$ mK$^2$ in the 68\% confidence contours for
383: $l (l+1) D_{l}$ is consistent with what is expected from cosmic variance alone.
384:
385: To compare the likelihoods for different $B$ values, we use the diagonal
386: $\chi^2$ statistic
387: \begin{equation}
388: \chi^2 = \sum_{l=2}^{300} \frac{ ({D^W}_l - \overline{{D^S}_l})^2 }
389: {\sigma_l^2},
390: \end{equation}
391: where $D^W_l$ is the WMAP data value, and $\overline{{D^S}_l}$ the
392: average and $\sigma_l$ the standard deviation of the ${D^S}_l$,
393: both obtained from 5000 model simulations, for each value of $B$. Note
394: that we do not use the full covariance matrix for the $D_l$'s but rather
395: just the diagonal terms. This is because 5000 simulations are not
396: sufficient to produce a reliably converged full covariance matrix for
397: the $D_l$'s. As also noted by Eriksen et al.~(2004b), for example,
398: even the above $\chi^2$ test can provide a just comparison between
399: data and simulations. The likelihood is proportional to $e^{- \chi^2 /2}$.
400: We calculate the likelihood for a few different values of $B$.
401:
402: The likelihood function obtained for the $n=-5$ case is shown in Fig.~2.
403: After integration, we get a 3 $\sigma$ confidence upper limit of $B < 15$
404: nG. As we can see from Fig.~2, $B$=0 G, which corresponds to pure Gaussian
405: primordial fluctuations, is within the 1 $\sigma$ confidence range from the
406: peak of the likelihood function (this 1 $\sigma$ range corresponds to a
407: $\delta B$ of 3.9 nG). The jaggedness in the likelihood function is from
408: the fluctuation of mean $D_l$'s used in calculating $\chi^2$ (see Fig.~1).
409: With 5000 simulations, the fluctuation of mean $D_l$'s should be of the order
410: of $\sigma_{l}/\sqrt{5000}$. This results in a fluctuation of 0.1 in the
411: $\chi^2$ values, or a 5\% uncertainty in the estimated likelihood. This
412: does not much affect the results of our analysis (for example for the 3
413: $\sigma$ limit on $B$ we look for a $\Delta\chi^2$ of 9, which is not very
414: sensitive to a 5\% uncertainty in the estimated likelihoods).
415:
416: For the $n = -7$ case (see Fig.~3), the 3 $\sigma$ confidence limit is
417: $B < 1.7$ nG, and
418: again the pure Gaussian primordial fluctuation case with $B=0$ G is not far
419: from the 1 $\sigma$ confidence range from the peak. In this case 1 $\sigma$
420: corresponds to a $\delta B= 0.4$ nG. More stringent limits are obtained in
421: this case as expected (eq. [15] shows that larger $D_l$'s with a stronger
422: $l$-dependence are predicted by the model for $n=-7$).
423:
424: The conditions of homogeneity and unidirectionality of the primordial
425: magnetic field may be a better approximation on some scales rather than
426: others. In each of the above cases for $n$, other ranges in $l$, such as
427: $2-100$, $101-200$, $201-300$, or $2-500$, did not indicate anything
428: qualitatively different, i.e., $B = 0$ G remains a satisfactory fit.
429:
430:
431: \section{Conclusions}
432:
433: We study off-diagonal correlations of the form
434: $D_l=\overline{\langle a_{l-1,m}a_{l+1,m}^* \rangle}$ in the first
435: year WMAP CMB anisotropy data. Such correlations can result from a
436: homogeneous primordial magnetic field. We do not find significant
437: off-diagonal correlations in the data, which appear to be satisfactorily
438: fit by a zero primordial magnetic field hypothesis. We place 3 $\sigma$ upper
439: limits on the strength of the magnetic field of $B < 15$ nG for spectral
440: index $n=-5$ and $B<1.7$ nG for $n=-7$. These two cases are interesting
441: as they correspond to a Harrison-Peebles-Yu-Zel'dovich scale-invariant
442: spectrum result for the $C_l$'s and $D_l$'s, and to a possible inflation
443: model primordial magnetic field perturbation spectrum, respectively.
444: These two cases also span the range of constraints that can be placed on
445: $B$ using this method. Future CMB anisotropy data should allow for tighter
446: constraints on a primordial cosmological magnetic field.
447:
448: \bigskip
449:
450: We acknowledge useful discussions with R.~Durrer and A.~Kosowsky. GC, TK,
451: and BR acknowledge support from NSF CAREER grant AST-9875031 and DOE
452: EPSCoR grant DE-FG02-00ER45824. TK also acknowledges CRDF-GRDF grant 3316.
453: PM and YW acknowledge support from NSF CAREER grant AST-0094335.
454:
455:
456: \begin{thebibliography}{}
457:
458: \bibitem[Adams et al. (1996)]{adams96}
459: Adams, J., Danielsson, U.~H., Grasso, D., \& Rubinstein, H.~1996,
460: Phys.~Lett.~B, 388, 253
461:
462: \bibitem[Bamba \& Yokoyama (2004)]{bamba04}
463: Bamba, K., \& Yokoyama, J.~2004, \prd, 69, 043507
464:
465: \bibitem[Bennett et al. (2003a)]{bennett03a}
466: Bennett, C. L., et al.~2003a, ApJS, 148, 1
467:
468: \bibitem[Bennett et al. (2003b)]{bennett03b}
469: Bennett, C. L., et al.~2003b, ApJS, 148, 97
470:
471: \bibitem[Bershadskii \& Sreenivasan (2004)]{bershadskii04}
472: Bershadskii, A., \& Sreenivasan, K. R.~2004, astro-ph/0403702
473:
474: \bibitem[Chiang et al.(2003)]{chiang03}
475: Chiang, L.-Y., Naselsky, P.~D., Verkhodanov, O.~V., \& Way, M.~J.~2003,
476: ApJ, 590, L65
477:
478: \bibitem[Coles et al.(2004)]{coles04}
479: Coles, P., Dineen, P., Earl, J., \& Wright, D.~2004, \mnras, in press,
480: astro-ph/0310252
481:
482: \bibitem[Colley \& Gott (2003)]{colley03}
483: Colley, W.~N., \& Gott, J.~R.~2003, \mnras, 344, 686
484:
485: \bibitem[Colley et al. (1996)]{colley96}
486: Colley, W.~N., Gott, J.~R., \& Park, J.~R.~1996, \mnras, 281, L82
487:
488: \bibitem[Copi et al. (2003)]{copi03}
489: Copi, C.~J., Huterer, D., \& Starkman, G.~D.~2003, astro-ph/0310511
490:
491: \bibitem[de Oliveira-Costa et al. (2003)]{deoliveiracosta03}
492: de Oliveira-Costa, A., Tegmark, M., Davies, R.~D., Gutierrez, C.~M.,
493: Lasenby, A.~N., Rebolo, R., \& Watson, R.~A.~2003, astro-ph/0312039
494:
495: \bibitem[Durrer et al. (1998)]{durrer98}
496: Durrer, R., Kahniashvili, T., \& Yates, A.~1998, \prd, 58, 123004 (DKY)
497:
498: \bibitem[Eriksen et al. (2004a)]{eriksen04a}
499: Eriksen, H.~K., Hansen, F.~K., Banday, A.~J., G\'orski, K.~M., \&
500: Lilje, P.~B.~2004a, \apj, in press, astro-ph/0307507
501:
502: \bibitem[Eriksen et al. (2004b)]{eriksen04b}
503: Eriksen, H.~K., Novikov, D.~I., Lilje, P.~B., Banday, A.~J., G\'orski,
504: K.~M.~2004b, astro-ph/0401276
505:
506: \bibitem[Fischler et al. (1985)]{fischler85}
507: Fischler, W., Ratra, B., \& Susskind, L.~1985, Nucl.~Phys.~B, 259, 730
508:
509: \bibitem[Giovannini (2003)]{giovannini03}
510: Giovannini, M.~2003, astro-ph/0312614
511:
512: \bibitem[Giovannini \& Shaposhnikov (1998)]{giovannini98}
513: Giovannini, M., \& Shaposhnikov, M.~1998, Phys.~Rev.~Lett., 80, 22
514:
515: \bibitem[Gorski et al. (1998b)]{gorski98b}
516: G\'orski, K.~M., Hivon, E., \& Wandelt, B.~D.~1998b, in Evolution of
517: Large-Scale Structure: From Recombination to Garching, ed. A.~J.~Banday,
518: R.~K.~Sheth, \& L.~A.~N.~da Costa (Enschede: Print Partners Iskamp), 37
519:
520: \bibitem[Gorski et al. (1998a)]{gorski98a}
521: G\'orski, K.~M., Ratra, B., Stompor, R., Sugiyama, N., \& Banday,
522: A.~J.~1998a, ApJS, 114, 1
523:
524: \bibitem[Gurzadyan et al. (2004)]{gurzadyan04}
525: Gurzadyan, V.~G., et al.~2004, astro-ph/0402399
526:
527: \bibitem[Hajian \& Souradeep (2003)]{hajian03}
528: Hajian, A. \& Souradeep, T.~2003, ApJ, 597, L5
529:
530: \bibitem[Hansen et al. (2004)]{hansen04}
531: Hansen, F.~K., Cabella, P., Marinucci, D., \& Vittorio, N.~2004,
532: astro-ph/0402396
533:
534: \bibitem[Komatsu et al. (2003)]{komatsu03}
535: Komatsu, E., et al.~2003, ApJS, 148, 119
536:
537: \bibitem[Kulsrud (1999)]{kulsrud99}
538: Kulsrud, R.~M.~1999, Ann.~Rev.~Astron.~Astrophys., 37, 37
539:
540: \bibitem[Lewis (2004)]{lewis04}
541: Lewis, A.~2004, astro-ph/0403583
542:
543: \bibitem[Mack et al. (2002)]{mack02}
544: Mack, A., Kahniashvili, T., \& Kosowsky, A.~2002, \prd, 65, 123004
545:
546: \bibitem[Mukherjee et al. (2003)]{mukherjee03}
547: Mukherjee, P., Coble, K., Dragovan, M., Ganga, K., Kovac, J., Ratra, B.,
548: \& Souradeep, T.~2003, \apj, 592, 692
549:
550: \bibitem[Mukherjee et al. (2002)]{mukherjee02}
551: Mukherjee, P., Dennison, B., Ratra, B., Simonetti, J.~H., Ganga, K., \&
552: Hamilton, J.-Ch.~2002, \apj, 579, 83
553:
554: \bibitem[Mukherjee et al. (2000)]{mukherjee00}
555: Mukherjee, P., Hobson, M.~P., \& Lasenby, A.~N. 2000, \mnras, 318, 1157
556:
557: \bibitem[Mukherjee \& Wang(2004)]{mukherjee04}
558: Mukherjee, P., \& Wang, Y.~2004, astro-ph/0402602
559:
560: \bibitem[Park (2004)]{park04}
561: Park, C.-G.~2004, \mnras, 349, 313
562:
563: \bibitem[Park et al. (2002)]{park02}
564: Park, C.-G., Park, C., \& Ratra, B.~2002, \apj, 568, 9
565:
566: \bibitem[Park et al. (2001)]{park01}
567: Park, C.-G., Park, C., Ratra, B., \& Tegmark, M.~2001, \apj, 556, 582
568:
569: \bibitem[Peebles (1980)]{peebles80}
570: Peebles, P.~J.~E.~1980, The Large-Scale Structure of the Universe
571: (Princeton: Princeton University Press), 356
572:
573: \bibitem[Peebles \& Ratra (2003)]{peebles03}
574: Peebles, P.~J.~E., \&\ Ratra, B.~2003, Rev.~Mod.~Phys., 75, 559
575:
576: \bibitem[Pogosian et al. (2002)]{pogosian02}
577: Pogosian, L., Vachaspati, T., \& Winitzki, S.~2002, \prd, 65, 083502
578:
579: \bibitem[Ratra (1985)]{ratra85}
580: Ratra, B.~1985, \prd, 31, 1931
581:
582: \bibitem[Ratra (1992)]{ratra92}
583: Ratra, B.~1992, \apj, 391, L1
584:
585: \bibitem[Semikoz \& Sokoloff (2004)]{semikoz04}
586: Semikoz, V. B., \& Sokoloff, D. D.~2004, \prl, 92, 131301
587:
588: \bibitem[Seshadri \& Subramanian (2001)]{seshadri01}
589: Seshadri, T.~R., \& Subramanian, K.~2001, \prl, 87, 101301
590:
591: \bibitem[Sigl et al. (1997)]{sigl97}
592: Sigl, G., Olinto, A.~V., \& Jedamzik, K.~1997, \prd, 55, 4582
593:
594: \bibitem[Spergel et al.~(2003)]{spergel03}
595: Spergel, D., et al.~2003, ApJS, 148, 175
596:
597: \bibitem[Subramanian et al. (2003)]{subramanian03}
598: Subramanian, K., Seshadri, T.~R., \& Barrow, J.~D.~2003, \mnras, 344, L31
599:
600: \bibitem[Tegmark et al. (2003)]{tegmark03}
601: Tegmark, M., de Oliveira-Costa, A., \& Hamilton, A.~J.~S.~2003, \prd, 68,
602: 123523
603:
604: \bibitem[Vachaspati (1991)]{vachaspati91}
605: Vachaspati, T.~1991, Phys.~Lett.~B, 265, 258
606:
607: \bibitem[Vielva et al. (2004)]{vielva04}
608: Vielva, P., Mart\'inez-Gonz\'alez, E., Barreiro, R.~B., Sanz, J.~L., \&
609: Cay\'on, L.~2004, ApJ, in press, astro-ph/0310273
610:
611: \bibitem[Widrow (2002)]{widrow02}
612: Widrow, L.~A.~2002, Rev.~Mod.~Phys., 74, 775
613:
614: \end{thebibliography}
615:
616:
617: \begin{figure}
618: \epsscale{0.8}\plotone{f1.eps}
619: \figcaption{Off-diagonal power spectra obtained from the WMAP data (crosses)
620: and the median and 68\% confidence contours obtained from model simulations
621: with magnetic field strengths $B = $ 0 (solid lines) and 16 (dotted lines)
622: nG, for a magnetic field perturbation spectral index $n=-5$.}
623: \end{figure}
624:
625: \begin{figure}
626: \epsscale{0.8}\plotone{f2.eps}
627: \figcaption{The likelihood as a function of the strength of the magnetic
628: field for a magnetic field perturbation spectral index $n=-5$.}
629: \end{figure}
630:
631: \begin{figure}
632: \epsscale{0.8}\plotone{f3.eps}
633: \figcaption{The likelihood as a function of the strength of the magnetic
634: field for a magnetic field perturbation spectral index $n=-7$.}
635: \end{figure}
636:
637:
638:
639: \end{document}
640: