1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \lefthyphenmin=1000
4: \sloppy
5: \usepackage{apjfonts}
6: \def\et{{\it et al.}}
7:
8: \shorttitle{First CMB Polarization Results from CAPMAP}
9: \shortauthors{Barkats et al.}
10:
11: %\received{2004 September 15}
12: %\accepted{2004 December 17}
13: \journalinfo{{\sc The Astrophysical Journal, 619: L127-L130, 2005 February 1}}
14:
15: \begin{document}
16:
17: \submitted{}
18: \title{First Measurements of the Polarization of the Cosmic Microwave Background Radiation at Small Angular Scales from CAPMAP}
19:
20: \author{D.~Barkats\altaffilmark{1,2}, C. Bischoff\altaffilmark{3},
21: P.~Farese\altaffilmark{1,4}, L. Fitzpatrick\altaffilmark{3,5},
22: T.~Gaier\altaffilmark{6}, J.~O.~Gundersen\altaffilmark{7},
23: M.~M.~Hedman\altaffilmark{3,8,9}, L.~Hyatt\altaffilmark{1},
24: J.~J.~McMahon\altaffilmark{1}, D.~Samtleben\altaffilmark{3,9,10},
25: S.~T.~Staggs\altaffilmark{1}, K.~Vanderlinde\altaffilmark{3}, and
26: B.~Winstein\altaffilmark{3}}
27:
28: \altaffiltext{1}{Department of Physics, Princeton University, Joseph Henry Laboratories, Jadwin Hall, P.O. Box 708, Princeton, NJ 08544-0708.}
29: \altaffiltext{2}{Current address: Department of Physics, California Institute of
30: Technology, 1200 East California Boulevard, MS 59-33, Pasadena, CA 91125.}
31: \altaffiltext{3}{Kavli Institute for Cosmological Physics and Enrico Fermi
32: Institute, University of Chicago, 5641 South Ingleside Avenue, Chicago, IL, 60637.}
33: \altaffiltext{4}{Dicke Fellow.}
34: \altaffiltext{5}{Current address: Department of Physics, Harvard University,
35: 17A Oxford Street, Cambridge, MA 02138.}
36: \altaffiltext{6}{Jet Propulsion Laboratory, California Institute of
37: Technology, 4800 Oak Grove Drive, Pasadena, CA 91109.}
38: \altaffiltext{7}{Department of Physics, University of Miami,
39: James L. Knight Physics Building, 1320 Campo Sano Drive, Coral Gables,
40: FL 33146.}
41: \altaffiltext{8}{Current address: Department of Astronomy, Cornell University,
42: Space Sciences Building, Ithaca, NY 14853-6801.}
43: \altaffiltext{9}{Kavli Fellow.}
44: \altaffiltext{10}{Corresponding author: dorothea@kicp.uchicago.edu.}
45:
46: \begin{abstract}
47: Polarization results from the Cosmic Anisotropy
48: Polarization MAPper (CAPMAP) experiment are reported. These
49: are based on 433 hr, after cuts, observing a 2 deg$^2$
50: patch around the north celestial pole with four 90 GHz
51: correlation polarimeters coupled to optics defining $4\arcmin$
52: beams. The {\it E}-mode flat band-power anisotropy within
53: $l=940^{+330}_{-300}$ is measured as 66$^{+53}_{-39}~\mu$K$^2$; the
54: 95$\%$ confidence level upper limit for {\it B}-mode power within
55: $l=1050^{+590}_{-520}$ is measured as 38~$\mu$K$^2$.
56: \end{abstract}
57:
58: \keywords{cosmology: cosmic microwave background --- cosmology: observations}
59:
60: \section{INTRODUCTION}
61: %\nopagebreak
62: The cosmic microwave background (CMB) is arguably the most fruitful
63: source of cosmological information. Its spectrum has been measured with extraordinary
64: precision \citep{FIRAS} and its spatial anisotropy has been very well
65: characterized, with the {\it Wilkinson Microwave Anisotropy Probe} \citep[WMAP;][]{WMAP:map} providing the most
66: comprehensive results. There are two distinct patterns to the
67: polarization of the CMB, conventionally termed {\it E}- and {\it B}-modes. The
68: former have recently been detected at the
69: degree scale \citep{Kovac02,WMAP:TE}; they
70: arise from the same density perturbations that dominate the
71: temperature anisotropies. The latter, which can result from
72: the lensing of {\it E}-modes due to the intervening
73: matter distribution or from primordial gravity waves,
74: are expected to be far smaller. Characterization
75: of the {\it E}-mode power spectrum is important for testing the
76: understanding of the origin of the CMB as well as for breaking
77: degeneracies in cosmological parameter determinations
78: \citep[e.g.,][]{turok:I,Hu}.
79:
80:
81: In this Letter we present results from the first season of CAPMAP (Cosmic Anisotropy Polarization MAPper),
82: an effort to characterize the {\it E}-mode polarization at
83: 90~GHz, where foregrounds are expected to be small, and at 4\arcmin,
84: where the polarization is expected to be near its
85: maximum. In $\S\S$ \ref{sec:inst} and \ref{sec:obs} we
86: describe the instrument, its characterization and calibration, and the
87: observations. In subsequent sections we summarize the reduction of the data
88: ($\S$ \ref{sec:red}), the analysis ($\S$ \ref{sec:an}), and the systematic
89: studies ($\S$ \ref{sec:sys}). We conclude this Letter with a
90: discussion of the results and their implications ($\S$ \ref{sec:disc}).
91: %\nopagebreak
92: %=================================================================================
93: \section{INSTRUMENT}
94: %\nopagebreak
95: \label{sec:inst}
96: %=================================================================================
97: %\nopagebreak
98: The CAPMAP instrument is described in detail
99: in \citet{instrument}; here its most important features are given.
100: In the winter of 2003, four W-band
101: (84$-$100 GHz) polarimeters (A$-$D) were installed in the
102: focal plane of the 7~m Cassegrain antenna \citep{bellabs} at the
103: Lucent Technologies facility in Crawford Hill, New Jersey (W$74^\circ11\arcmin11\arcsec\!\!$,
104: N$40^\circ23\arcmin31\arcsec\!\!$). The horn and meniscus lens system attached to each
105: polarimeter underilluminated the telescope mirrors with an edge taper
106: of $-35$~dB on the primary, resulting in a final beam of
107: $3\arcmin\!\!.88$ FWHM ($\sigma_B=1\arcmin\!\!.65$) with an
108: ellipticity less than 3$\%$. The polarimeters were housed inside a
109: single cryostat, nominally arranged so that A (C) observed $0^\circ\!\!.25$
110: below (above) and B (D) $0^\circ\!\!.25$ left (right) of the optical axis of
111: the telescope.
112:
113:
114: Heterodyne analog correlation polarimeters that multiply
115: two orthogonal components of the incident electric field produced an
116: output proportional to one linear combination of the Stokes
117: parameters {\it Q} and {\it U}. The two components derive from an ortho-mode
118: transducer (OMT) oriented with one axis horizontal. The polarimeter
119: output is thus proportional to the difference in power for field
120: components aligned $\pm45^\circ$ to the horizontal in the focal plane.
121: The first-stage low noise amplifiers (LNAs) in each arm are InP HEMT
122: monolithic microwave-integrated circuits developed by JPL and Northrop
123: Grumman Space Technology \citep{northrop}. A mechanical refrigerator
124: cools the horns, the OMT, and the LNAs to $\leq$40 K. One local
125: oscillator inside the cryostat supplies the polarimeters with the 82~GHz
126: to down-convert the 84$-$100~GHz signals to an intermediate
127: frequency (IF) of 2$-$18~GHz. The oscillator signal in one arm of each
128: polarimeter is phase-switched at 4~kHz, well above the $1/f$ knee of
129: the amplifiers. A detector diode monitors the total power in each
130: arm. The phase-switched IF radiation is split into three 4~GHz-wide
131: subbands (S0, S1, S2) prior to multiplication, yielding then 12 polarization and eight total power channels.
132:
133:
134: The polarimeter outputs are digitized (24~bits, 100~kHz)
135: with a $\Sigma\Delta$ analog-to-digital converter (ICS-610); digitally
136: demodulated data are recorded at 100~Hz. The 100~kHz samples, averaged over 40 phase switches, are recorded to produce
137: quadrature data sets, demodulated out of phase with the phase switch,
138: which contain no celestial signal.
139: %\nopagebreak
140: %=================================================================================
141: \section{OBSERVATIONS AND CALIBRATION}
142: \label{sec:obs}
143: %=================================================================================
144: \nopagebreak
145: Between 2003 February 18 and April 6, 541 hr of CMB data were
146: recorded. Calibration data were taken before, regularly throughout,
147: and after the season.
148:
149:
150: For the CMB observation the telescope's
151: optical axis moved through the north celestial pole (NCP) with an 8~s period by $\pm0^\circ\!\!.53$ at constant elevation. Polarimeter
152: offsets from the optical axis are given in Table \ref{tab:horns}; the
153: sampled region is a circle of diameter $1^\circ\!\!.6$
154: centered on the NCP, with the inner circle of diameter $0^\circ\!\!.14$
155: missing. This scan strategy yields quite uniform {\it Q/U} coverage over the
156: region.
157: %---------------------------------
158: \begin{table}[h]
159: \caption{Polarimeter Position Offsets$^a$}
160: \begin{center}
161: \begin{tabular}{lllll}
162: \hline & A & B & C & D
163: \\ \hline \hline co-elevation & -0.022 & -0.247 & -0.025 & 0.200
164: \\ cross-elevation & -0.286 & -0.066 & 0.162 & -0.059
165: \end{tabular}
166: \end{center}
167: \label{tab:horns}
168: \footnotesize
169: $^a$ The polarimeter offsets from the optical axis are given in degrees. For the CMB observations, the constant-elevation scan was centered with the optical axis at $(-0^\circ\!\!.03, -0^\circ\!\!.09)$ from the NCP.
170: \end{table}
171: %---------------------------------
172:
173:
174: The pointing solution, determined with Jupiter observations, was
175: refined to $\pm30\arcsec$ using 150 observations of
176: astrophysical sources. The relative offsets, beam sizes and
177: the total power calibration were determined from the observations of Jupiter
178: in the total power channels.
179:
180:
181: Noise temperatures were estimated from several elevation scans
182: ($20^\circ-90^\circ$). They ranged between 60 and 70~K and agreed
183: with laboratory cold load tests.
184:
185:
186: For the determination of polarized gains, a nutating plate was
187: installed in front of the secondary mirror to produce a modulated
188: polarized signal; this gave calibrations with an overall uncertainty
189: of 10$\%$ and a 3$\%$ relative uncertainty among the
190: channels. Frequent scans of the polarized source Tau A near
191: parallactic angle 110$^\circ$ (where the signal in the receivers is
192: maximal) provided a good calibration check
193: \citep{instrument}. Polarimeter sensitivities are summarized in Table
194: \ref{tab:sens}.
195:
196:
197: \begin{table}[h]
198: \begin{center}
199: \caption{Polarimeter Characteristics}
200: \label{tab:sens}
201: \begin{tabular}{rccccc}\hline
202: & \multicolumn{4}{c}{Polarimeter$^a$} & $\nu_{eff}^b$\\
203: Channel & A & B & C & D & \\ \hline \hline
204: S0 & 3.2 & 3.4 & 2.9 & 3.1 & 87 \\
205: S1 & 2.1 & 2.1 & 2.1 & 2.2 & 92 \\
206: S2 & 2.7 & 2.0 & 3.5 & 2.5 & 96 \\ \hline
207: \end{tabular}
208: \end{center}
209: \footnotesize
210: $^a$ Sensitivities in mK$\sqrt{s}$ in thermodynamic units
211: \\ $^b$ Typical central frequency in GHz
212: \end{table}
213: %---------------------------------
214: The calibration was corrected for
215: slow few kelvin variations in the IF electronics and
216: for a varying phase between the digitization and phase
217: switch clocks.
218: The IF gains changed by up to 3$\%$~K$^{-1}$ but the temperature was stable to 1 K for 70$\%$ of the
219: data. The clock correction ranged from 0.9 to 1 and varied
220: by at most 0.003~hr$^{-1}$. The calibration for two of the channels (DS0, DS2) was adjusted by $\sim10\%$ to compensate for gain
221: compression.
222: %=================================================================================
223: \section{DATA REDUCTION}
224: \label{sec:red}
225: %=================================================================================
226: %\nopagebreak
227: The CMB data were divided into 40~s periods; data within
228: each were divided into 20 $\sim3\arcmin$-wide, nearly uniformly populated azimuthal
229: bins, discarding the $9\%$ of the data at the turnarounds. Means and standard errors calculated in each
230: bin were carried forward in the analysis. Varying the period
231: between 16 and 40~s did not alter the final results.
232:
233:
234: A second-order polynomial was fitted to the
235: azimuthal samples in each 40~s period and removed. The polarimetry channels reject the
236: common mode intensity by more than 23~dB, but their residual
237: sensitivity to the atmospheric temperature, which varied daily by 10~K, dominated the intercept terms. The linear terms were evident when
238: co-added over an hour, revealing slopes as large
239: as $500~\mu$K~deg$^{-1}$, with variations up to $200~\mu$K~deg$^{-1}$day$^{-1}$.
240: The quadratic terms were even smaller and more stable.
241:
242:
243: Abnormally high
244: fluctuations ($>4\sigma$) in the polarized channels or unusual changes
245: in the ratios between different total power channels were observed for 20 hr of data.
246: Both symptoms result from ice formation on the cryostat
247: windows so these data were removed. The five most sensitive total power channels were used to identify bad weather with a cut on their fractional deviations. Only periods that passed the
248: cut for all five channels were retained, eliminating 88 hr and
249: leaving 433 hr for the final data sample.
250: Ten percent variations in this cut level changed the result within statistics.
251:
252:
253: Collapsing all polarization data into the
254: 20 azimuthal bins reveals scan synchronous structures
255: (SSSs) with rms levels from 10 to 26~$\mu$K; from a 90~GHz side lobe beam map, these are likely due to
256: ground pickup. Dependences on variables like system temperature, local sidereal
257: time (LST), or ambient temperature were studied. Only BS1 and BS2 showed such
258: a dependence, varying with the ambient temperature.
259: The mean of the data in each azimuthal bin was
260: removed, and residual effects that could arise from
261: correlations to ambient temperature are modeled as described below.
262: The channel with the largest structure and variation (BS2) is dropped
263: with negligible change in the results but with a significant reduction
264: in the systematic uncertainty.
265:
266:
267: After structure removal, corrections for atmospheric
268: absorption (typically 20$\%$) derived from the total power
269: channels, were applied. The corrected periods,
270: already parsed into azimuth bins, were binned into 72 LST bins, yielding 72$\times$20=1440 element data vectors for each of the
271: 12 channels.
272: The weighted averages
273: of the three frequency channels (which have correlations less than
274: 1$\%$) were used in the likelihood analysis. One resulting data vector
275: is displayed in Figure \ref{fig:map}.
276: %-----------------------------------------
277: \begin{figure}[t]
278: \plotone{f1.eps}
279: \caption{The 20$\times$72 data vector for receiver C. The scale
280: ranges between $\pm 210$~$\mu$K; the average unchopped error for each
281: element is 64~$\mu$K. These pixels
282: are spaced approximately 0.8 beamwidths in azimuth; in the LST
283: direction, their spacing ranges from about 0.1 to 1 beam width
284: depending on the azimuthal position.}
285: \label{fig:map}
286: \end{figure}
287: %=================================================================================
288: \section{DATA ANALYSIS}
289: \label{sec:an}
290: %=================================================================================
291: %\nopagebreak
292: A Bayesian likelihood analysis was performed by maximizing the
293: likelihood,
294: \begin{equation} \mathcal{L} = { 1\over \sqrt{\det(C)}}
295: \exp \{ -{\bf x}^T C^{-1} {\bf x}/2 \},
296: \end{equation}
297: where ${\bf x}$ is the data vector with $4 \times 1440$ entries and $C$ is the
298: covariance matrix. The vector ${\bf x}$ sums the signal ($s$) and
299: receiver noise ($n$) terms: $x_i=A_{ij}s_j+n_i$, where $A_{ij}$
300: determines the linear combination of {\it Q} ($s_1$) and {\it U} ($s_2$) in pixel
301: $i$. Similarly, the covariance matrix comprises signal ($C_T$) added
302: to noise $N$: $C_{ij} = <x_i x_j> = C_{T,ij} + N_{ij}$, where $N_{ij}=n_i^2\delta_{ij}$, with typical noise
303: $n_i\approx 60\,\mu$K. The signal covariance depends on the multipole moments for {\it E}- and {\it B}-modes, $C_l^E$ and
304: $C_l^B$:
305: \begin{equation}
306: C_{T,ij} = \sum_l \frac{2l+1}{4\pi} (C_l^E W_{l,ij}^E + C_l^B W_{l,ij}^B ).
307: \label{eq:win}
308: \end{equation}
309: Here, the window function matrices $W_l^X$ account for (1) beam
310: functions $B_l^2\approx \exp(-l^2\sigma_B^2$), (2) the different
311: pixel areas in the data vector, and (3) the conversion from ({\it Q, U}) to $X=(E, B)$
312: \citep{ZS}. Figure \ref{fig:result} shows the trace of
313: receiver C's {\it E}-mode window function, similar to
314: those of the other receivers. The likelihood is maximized with respect
315: to one to three variables that parameterize $C^E_l$ and/or $C^B_l$,
316: as described below.
317:
318:
319: The signal-to-noise ratio eigenmode method was used \citep[e.g.,][]{bond98}: transformations into the basis where the noise matrix is
320: white and then into the eigenbasis of the
321: transformed $C_T$ are performed. Removing offsets every 40~s and
322: subtracting SSSs transform ${\bf x}$ to $D {\bf x}$. Accordingly, $C$ is
323: transformed to $DCD^T$, and the analysis is continued in the subspace
324: onto which $D$ projects.
325:
326:
327: The code was exercised extensively with simulated data
328: sets. Realizations with noise levels from that of the data to 20 times
329: smaller were superposed on CMB realizations of the concordance model
330: \citep{WMAP:parameters}. The likelihood estimator was found to be
331: unbiased; however, at the noise level of the data, the loss of
332: sensitivity from the offset removals is a factor of 2.
333:
334:
335: Four likelihood evaluations were made: first by taking $C^E_l$ as a
336: multiplier of the concordance model, with $C^B_l \equiv 0$; second
337: by taking $l(l+1)C^E_l$ to be a constant (flat band-power),
338: with $C^B_l \equiv 0$; third by simultaneously taking $C^E_l$ as
339: a multiplier to the concordance model and $l(l+1)C^B_l$ as a constant;
340: and fourth by taking $C^E_l$ to comprise three flat band-power levels,
341: in the $l$ bins (2-500/501-1500/1501-2900), with $C^B_l \equiv
342: 0$ (three-band analysis). Fit results are summarized in Table
343: \ref{tab:result1}. Effective $l$s are calculated according to
344: \citet{Steinhardt} and quoted with the central 68$\%$ region of
345: $W_l/l$.
346:
347:
348: The various likelihood evaluations yield a 2~$\sigma$ detection of
349: the {\it E}-mode signal, with no {\it B}-mode signal or significant
350: dependence on frequency observed. Results from the
351: three-band analysis are shown in Figure
352: \ref{fig:result}. The upper band has negligible correlation with
353: the other two. For the lower two bands the marginalized results
354: from integrating over their likelihood contours are displayed. The
355: marginalization shifted the peak of the middle band down by 6$\%$,
356: increasing its width by 12$\%$.
357: A fit of the middle $l$-band's likelihood to an offset
358: lognormal distribution \citep{bond} gives
359: 66~$\mu$K$^2$ with a variance of (44~$\mu$K$^2$)$^2$ and a noise related
360: offset of 93~$\mu$K$^2$.
361: %-----------------------------------------
362: \begin{figure}[t]
363: \plotone{f2.eps}
364: \caption{Likelihood results. {\it Top}: Results from the three-band analysis, the concordance model ({\it long dashed line}), and the trace of the window function $W_l^E$, as
365: defined in equation (\ref{eq:win}), for receiver C, in arbitrary units (dotted line). For the outer bands, 95$\%$ confidence level upper limits are
366: displayed. {\it Lower left}:
367: Individual marginalized likelihood curves for the three bands.
368: {\it Lower right}: Concordance model fitted to the data and
369: several null maps as described in the text.}
370: \label{fig:result}
371: \end{figure}
372: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
373: \section{EVALUATION OF SYSTEMATIC UNCERTAINTIES}
374: \label{sec:sys}
375: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
376: The three most important sources of systematic uncertainty were
377: associated with the telescope pointing, the scan synchronous residual
378: structure and the relative gains. Their impact was
379: evaluated on fits of {\it E}-mode power to a multiplier of the concordance
380: model; selected studies confirmed that uncertainties estimated with
381: those fits apply to the three other fitting methods.
382:
383:
384: The effect of possible mispointing was determined by varying the
385: global pointing solution within its uncertainty, $\sim10\%$ of a beamwidth.
386: The $1~\sigma$ shifts in azimuth and elevation were added in
387: quadrature, resulting in an uncertainty of $12\%$.
388:
389:
390: Effects of the SSSs were simulated. It was determined separately in each channel for two
391: halves of the data, divided according to ambient temperature and its extracted (linear) variation.
392: Multiple simulations were made with structure added to
393: the CMB according to the ambient temperature in each period. This shifted the result upward by on average
394: 0.1 (times concordance), and the spread, as the underlying CMB was
395: varied, was $\pm$0.1. The uncertainty is then conservatively taken to
396: be $\pm0.2$ (corresponding to a 10$\%$ uncertainty).
397:
398:
399: The effect of relative gain uncertainties was estimated from the
400: scatter in the result with the gain of each of the 11 polarimeters
401: chosen randomly within the estimated $\pm 3$\% uncertainty, leading
402: to an uncertainty of $6\%$.
403:
404:
405: Three other sources of uncertainty were studied. The beam size was
406: uncertain to 2$\%$; varying it by even 10$\%$ leads to less than a 3$\%$
407: change in the result. As mentioned earlier, the polarimeters have a
408: weak response to temperature anisotropies ($-23$~dB), and the optics
409: produces a small quadrupole response at scales smaller than the beam
410: ($-12$~dB). Simulations show that these also lead to negligible effects
411: \citep[see also][]{HHZ}.
412:
413:
414: Several checks of the robustness of the result were performed. Excluding single radiometers or 25$\%$
415: data samples gave consistent results. Difference maps were created by
416: splitting the data into two parts according
417: to high and low system and ambient temperatures and time. Data vectors
418: were produced for each half (D$_1$ and D$_2$) and the likelihood was
419: evaluated on (D$_1-$D$_2$)/2. These likelihoods, together with that
420: for the quadrature data, are shown in Figure \ref{fig:result}; none
421: show a signal. Finally, a parallel analysis was
422: performed. Offsets were removed on a 4~s timescale, and an
423: independent code implemented a different pixelization
424: scheme and a different means of evaluating the associated covariance matrix and nulling
425: out the appropriate modes. It was used to obtain a multiplier to concordance
426: and gave a consistent result.
427:
428:
429: Adding the three main uncertainties in quadrature, the overall
430: systematic uncertainty becomes 17$\%$. The 10$\%$ overall gain
431: uncertainty contributes an additional 20$\%$ without reducing the
432: significance of the detection.
433: %-----------------------------------------------------------------
434: \begin{table}[h]
435: \begin{center}
436: \caption{CAPMAP results$^a$}
437: \label{tab:result1}
438: \begin{tabular}{lccccc}\hline \hline
439: \multicolumn{6}{c}{{\it E}-mode multiplier to concordance, $l$=1120$^{+560}_{-520}$, $C_l^B\equiv$0}
440: \\ \hline frequency & all$^a$ & S0 & S1 & S2
441: \\ result$^b$ & 2.0$^{+2.1}_{-1.5}$ & 3.9$^{+7.3}_{-4.8}$ & -0.1$^{+2.8}_{-1.8}$ & 9.4$^{+8.8}_{-6.1}$
442: \\ \hline \multicolumn{6}{c}{Fits to flat band-power, $C_l^B\equiv$0 except for fit for {\it B}}
443: \\ \hline & B$^c$ & E & E1$^d$ & E2$^d$ & E3$^d$
444: \\ $l$ & 1050$^{+590}_{-520}$ & 1120$^{+560}_{-520}$ & 410$^{+70}_{-80}$ & 940$^{+330}_{-300}$ & 1940$^{+400}_{-350}$
445: \\ result$^a$& $<38\mu$K$^2$ & 31$^{+39}_{-27}\mu$K$^2$ & $<324\mu$K$^2$ & 66$^{+53}_{-39}\mu$K$^2$ & $<195\mu$K$^2$
446: \end{tabular}
447: \end{center}
448: \footnotesize
449: $^{a}$ Simultaneous fit with {\it B} also gives 2.0$^{+2.1}_{-1.5}$
450: \\ $^b$ With 68$\%$ interval of highest posterior density or 95$\%$ upper limit
451: \\ $^{c}$ Fitted simultaneously with {\it E}
452: \\ $^{d}$ Three-band analysis
453: \end{table}
454: %-----------------------------------------------------------------
455: %\nopagebreak
456: \section{DISCUSSION}
457: \label{sec:disc}
458: %\nopagebreak
459: Evidence of polarization anisotropy in the CMB in the multipole
460: region around $l=1000$, where the {\it E}-mode power spectrum is expected to
461: peak, has been presented. Signals of increasing significance are
462: seen when fitting for (1) a flat band-power, (2) a multiplier of the
463: expected concordance power spectrum, and (3) power in a band centered
464: at these high $l$-values. The {\it E}-mode signal does not change when
465: fitting simultaneously for both {\it E}- and {\it B}-modes, and the latter are not
466: detected. Here the possible contaminations to these results are
467: discussed together with their significance.
468:
469:
470: Foregrounds are not expected to be significant. From
471: \citet{FDS} dust maps and {\it WMAP} synchrotron maps \citep{WMAP:foregrounds}
472: around the NCP, temperature anisotropies of 4.4
473: and 1.3~$\mu$K, respectively, are expected. Even with pessimistic
474: assumptions about polarization fractions and how these foregrounds
475: extrapolate to CAPMAP's angular scales, their contributions are well
476: below the observed signal.
477:
478:
479: There are no deep source maps exactly at the NCP. A survey at
480: 90 GHz \citep{MMA} extrapolates 4400 sources brighter
481: than 100 mJy over the entire sky, implying a 20$\%$ chance for
482: one to be in the CAPMAP field. From the instrument response
483: (3~$\mu$K~mJy$^{-1}$), such a source would contribute at the 20~$\mu$K level in
484: 1 pixel if it had the same high polarization as Tau A (7.5$\%$).
485: Effects on the polarization power spectra, along with those from weaker but more
486: copious sources, are negligible. Conclusions from a dedicated source survey
487: near the NCP \citep{Leitch00} are similar.
488:
489: Adding the systematic uncertainty in quadrature to the result in the
490: middle l-band, a null result is excluded at just over the 2~$\sigma$
491: level. Including the overall calibration uncertainty, this result is
492: consistent with the concordance model within 72$\%$ of the highest posterior
493: density. This result at 90 GHz is consistent with the very recent
494: results from the Degree Angular Scale Interferometer and Cosmic Background Imager at 30 GHz \citep{Leitch04, Readhead}.
495:
496:
497: \acknowledgements We thank Tom Crawford, Wayne Hu, Norm Jarosik, Stephan Meyer,
498: Lyman Page, and David Spergel for many helpful discussions and
499: Michelle Yeh and Eugenia Stefanescu for help in the data collection. We thank
500: Lucent Technologies for use of the 7~m telescope and Bob Wilson, Greg Wright and Tod Sizer for their assistance with it.
501: This work was supported by NSF grants PHY-9984440,
502: PHY-0099493, PHY-0355328, AST-0206241 and PHY-0114422 and by the Kavli Foundation.
503: Portions of this work were carried out at the Jet Propulsion
504: Laboratory, California Institute of Technology, operating under a
505: contract from the National Aeronautics and Space Administration.
506:
507: \begin{thebibliography}{19}
508: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
509:
510: \bibitem[{{Barkats} {et~al.}(2005){Barkats}, {Bischoff}, {Farese}, {Gaier},
511: {Gundersen}, {Hedman}, {Hyatt}, {McMahon}, {Samtleben}, {Stefanescu},
512: {Staggs}, {Vanderlinde}, \& {Winstein}}]{instrument}
513: {Barkats}, D., {et~al.} 2005,
514: {\apjs, submitted}
515:
516: \bibitem[{{Bennett} {et~al.}(2003{\natexlab{a}}){Bennett}, {Halpern},
517: {Hinshaw}, {Jarosik}, {Kogut}, {Limon}, {Meyer}, {Page}, {Spergel}, {Tucker},
518: {Wollack}, {Wright}, {Barnes}, {Greason}, {Hill}, {Komatsu}, {Nolta},
519: {Odegard}, {Peiris}, {Verde}, \& {Weiland}}]{WMAP:map}
520: {Bennett}, C.~L., {et~al.} 2003{\natexlab{a}}, \apjs, 148, 1
521:
522: \bibitem[{{Bennett} {et~al.}(2003{\natexlab{b}}){Bennett}, {Hill}, {Hinshaw},
523: {Nolta}, {Odegard}, {Page}, {Spergel}, {Weiland}, {Wright}, {Halpern},
524: {Jarosik}, {Kogut}, {Limon}, {Meyer}, {Tucker}, \&
525: {Wollack}}]{WMAP:foregrounds}
526: {Bennett}, C.~L., {et~al.} 2003{\natexlab{b}}, \apjs, 148, 97
527:
528: \bibitem[{{Bond} {et~al.}(1998){Bond}, {Jaffe}, \& {Knox}}]{bond98}
529: {Bond}, J.~R., {Jaffe}, A.~H., \& {Knox}, L. 1998, \prd, 57, 2117
530:
531: \bibitem[{{Bond} {et~al.}(2000){Bond}, {Jaffe}, \& {Knox}}]{bond}
532: ---. 2000, \apj, 533, 19
533:
534: \bibitem[{{Bucher} {et~al.}(2001){Bucher}, {Moodley}, \& {Turok}}]{turok:I}
535: {Bucher}, M., {Moodley}, K., \& {Turok}, N. 2001, Phys. Rev. Lett., 87,
536: 191301
537:
538: \bibitem[{{Chu} {et~al.}(1978){Chu}, {Wilson}, {England}, {Gray}, \&
539: {Legg}}]{bellabs}
540: {Chu}, T.~S., {Wilson}, R.~W., {England}, R.~W., {Gray}, D.~A., \& {Legg},
541: W.~E. 1978, The Bell System Technical Journal, 57, 1257
542:
543: \bibitem[{{Finkbeiner} {et~al.}(1999){Finkbeiner}, {Davis}, \&
544: {Schlegel}}]{FDS}
545: {Finkbeiner}, D.~P., {Davis}, M., \& {Schlegel}, D.~J. 1999, \apj, 524, 867
546:
547: \bibitem[{{Fixsen} {et~al.}(1997){Fixsen}, {Hinshaw}, {Bennett}, \&
548: {Mather}}]{FIRAS}
549: {Fixsen}, D.~J., {Hinshaw}, G., {Bennett}, C.~L., \& {Mather}, J.~C. 1997,
550: \apj, 486, 623
551:
552: \bibitem[{{Holdaway} {et~al.}(1994){Holdaway}, {Owen}, \& {Rupen}}]{MMA}
553: {Holdaway}, M.~A., {Owen}, F.~N., \& {Rupen}, M.~P. 1994, {MMA Memo 123},
554: {http://www.alma.nrao.edu/memos/html-memos/alma123/}
555:
556: \bibitem[{{Hu}(2003)}]{Hu}
557: {Hu}, W. 2003, Annals of Physics, 303, 203
558:
559: \bibitem[{{Hu} {et~al.}(2003){Hu}, {Hedman}, \& {Zaldarriaga}}]{HHZ}
560: {Hu}, W., {Hedman}, M.~M., \& {Zaldarriaga}, M. 2003, \prd, 67, 043004
561:
562: \bibitem[{{Kogut} {et~al.}(2003){Kogut}, {Spergel}, {Barnes}, {Bennett},
563: {Halpern}, {Hinshaw}, {Jarosik}, {Limon}, {Meyer}, {Page}, {Tucker},
564: {Wollack}, \& {Wright}}]{WMAP:TE}
565: {Kogut}, A. , {et~al.} 2003, \apjs, 148, 161
566:
567: \bibitem[{{Kovac} {et~al.}(2002){Kovac}, {Leitch}, {Pryke}, {Carlstrom},
568: {Halverson}, \& {Holzapfel}}]{Kovac02}
569: {Kovac}, J.~M., {Leitch}, E.~M., {Pryke}, C., {Carlstrom}, J.~E., {Halverson},
570: N.~W., \& {Holzapfel}, W.~L. 2002, \nat, 420, 772
571:
572: \bibitem[{{Leitch} {et~al.}(2004){Leitch}, {Kovac}, {Halverson}, {Carlstrom},
573: {Pryke}, \& {Smith}}]{Leitch04}
574: {Leitch}, E.~M., {Kovac}, J.~M., {Halverson}, N.~W., {Carlstrom}, J.~E.,
575: {Pryke}, C., \& {Smith}, M.~W.~E. 2004, \apj, submitted (astro-ph/0409357)
576:
577: \bibitem[{{Leitch} {et~al.}(2000){Leitch}, {Readhead}, {Pearson}, {Myers},
578: {Gulkis}, \& {Lawrence}}]{Leitch00}
579: {Leitch}, E.~M., {Readhead}, A.~C.~S., {Pearson}, T.~J., {Myers}, S.~T.,
580: {Gulkis}, S., \& {Lawrence}, C.~R. 2000, \apj, 532, 37
581:
582: \bibitem[{{Readhead} {et~al.}(2004){Readhead}, {Myers}, {Pearson},
583: {Sievers}, {Mason}, {Contaldi}, {Bond}, {Bustos}, {Altamirano},
584: {Achermann}, {Bronfman}, {Carlstrom}, {Cartwright}, {Casassus},
585: {Dickinson}, {Holzapfel}, {Kovac}, {Leitch}, {May}, {Padin},
586: {Pogosyan}, {Pospieszalski}, {Pryke}, {Reeves}, {Shepherd}, \& {Torres}}]{Readhead}
587: {Readhead}, A.~C.~S., {et~al.} 2004, Science, 306, 836
588:
589: \bibitem[{{Spergel} {et~al.}(2003){Spergel}, {Verde}, {Peiris}, {Komatsu},
590: {Nolta}, {Bennett}, {Halpern}, {Hinshaw}, {Jarosik}, {Kogut}, {Limon},
591: {Meyer}, {Page}, {Tucker}, {Weiland}, {Wollack}, \&
592: {Wright}}]{WMAP:parameters}
593: {Spergel}, D.~N., {et~al.} 2003, \apjs, 148, 175
594:
595: \bibitem[{{Steinhardt}(1995)}]{Steinhardt}
596: {Steinhardt}, P.~J. 1995, in Particle and Nuclear Astrophysics and Cosmology in
597: the Next Millenium, ed. E.~W. {Kolb} \& R.~D. {Peccei} (Singapore: World
598: Scientific), 51
599:
600: \bibitem[{{Weinreb} {et~al.}(1999){Weinreb}, {Lai}, {Erickson}, {Gaier}, \&
601: {Wielgus}}]{northrop}
602: {Weinreb}, S., {Lai}, R., {Erickson}, N., {Gaier}, T., \& {Wielgus}, J. 1999,
603: in {IEEE MTT-S, Int. Microwave Symp. Digest}, 101
604:
605: \bibitem[{{Zaldarriaga} \& {Seljak}(1997)}]{ZS}
606: {Zaldarriaga}, M., \& {Seljak}, U. 1997, \prd, 55, 1830
607:
608: \end{thebibliography}
609:
610:
611: \end{document}
612: