WMAP 
FiveYear Wilkinson Microwave Anisotropy Probe (WMAP) Observations:

WMAP Fiveyear Paper Figures, Hill, et al.  
Individual figures are provided for use in talks. Proper display of PNG transparency in PowerPoint requires saving files to your computer before Inserting them. Please acknowledge the WMAP Science Team when using these images. Image Credit: NASA / WMAP Science Team 

Fig. 1 Flow chart of beam and window function processing. Boxes are color coded as follows: green, starting data; pink, intermediate data; blue, computations or processing steps; magenta, the result, b_{l}, which is the beam transfer function. The window function, w_{l}, for power spectra C_{l} involving a single DA, is b^{2}_{l} . Dashed boxes enclose iterative algorithms. 
PNG (28 kb  603x1024 pixels) PNG (33 kb  1207x2048 pixels) EPS (145 kb) 

Fig. 2 Beams in the WMAP focal plane for side A (left) and side B (right). The top panels show the measured beams, the middle panels show the beam models, and the bottom panels show the residuals. In the top four panels, each beam is scaled to its maximum (red) and plotted logarithmically to a level of 40 dB (blue). For the bottom panels, each beam’s residual is shown linearly as 100(datamodel)/beam peak. The scales are ±10% for K1; ±5% for Ka1, Q1, and Q2; ±3% for V1 and V2; and ±2.5% for W1–W4. A similar depiction of the A side only for three years of data is in Figure 9 of Jarosik et al. (2007). 
PNG (381 kb  1024x1520 pixels) PNG (453 kb  1380x2048 pixels) EPS (480 kb) 

Fig. 3 Fitted distortions of the Aside (left) and Bside (right) mirrors with respect to nominal shapes. Y axis is negative in the sky direction and positive toward the main spacecraft structure. Top row: primary mirrors. The dominant feature of the primary mirror distortions is the central rectangle, corresponding to a frame that is part of the backing structure. Hints of the stiffening lugs in the backing structure may also be seen around the edges. Bottom row: secondary mirrors. The mirrors are constrained only where they are substantially illuminated by the feed horns (Page et al. 2003b). Thus, for example, the secondary mirror for the B side appears as a bull’seye partly because the fit is only constrained near the center. Gray line: contour of the mean W band illumination function 15 dB from the peak. 
PNG (179 kb  943x1024 pixels) PNG (341 kb  1885x2048 pixels) EPS (337 kb) 

Fig. 4 Growth of solid angle in model beams as a function of kmax, for the V2 DA on the A side. Top: logarithmically scaled images of the model beam pattern as fitted for k_{max} = 12, 14, 16, 18, 20, 22, and 24, respectively, together with a beam that combines the k_{max} = 24 fit with randomphase modes extrapolated to k = 250. Axis tick marks are at 1° intervals. Bottom left: azimuthally averaged beam profiles for the models pictured, in units of gain relative to isotropic. Indigo–red: profiles of the seven beam models from k_{max} = 12 through k_{max} = 24, respectively. Gray: k_{max} = 250 extrapolation result. Black: Upper limit on the main beam sensitivity from Moon observations, obtained for side A by integrating over positive pixels in the differential sidelobe response pattern. The fiveyear model tail, which is a feature of the twodimensional beam pattern, is the part of the beam that is both inside the transition radius and below the hybridization threshold (dotted line; see ξ3.3). The hybridization threshold and transition radius from the threeyear analysis are indicated by the dashed line. Bottom right: model tail solid angle as a function of k_{max}, relative to the total solid angle inside the transition radius; squares, fiveyear; triangles, modified tail of fiveyear models, using threeyear threshold and transition radius. Fitting to k_{max} = 24 rather than k_{max} = 12 increases by a factor of ~ 2 the solid angle of the model tail as defined by threeyear main beam limits. But note that the difference between the various fits is ~ 0.1% of the total beam solid angle for the 5 year data. 
PNG (158 kb  1040x2048 pixels) PNG (214 kb  2080x4096 pixels) EPS (374 kb) 

Fig. 5 Power spectra of Aside (top) and Bside (bottom) primary mirror distortions as a function of spatial frequency on the primary mirror surface, f = k/(280 cm), where k is the spatial frequency index used in the physical optics fits. Vertical bars on the f axis indicate f corresponding to k = 12 and k = 24. Solid lines: extrapolated powerlaw distortion spectra with slopes fitted by comparison to Moon sidelobe data, namely, α = 4.95 for the A side and α = 4.43 for the B side. In practice, these extrapolated distortions are used to update the sidelobe response patterns, but not to model the main beams. The error bars and upper limits show the mean absolute deviation about the mean of points with length scales less than 15 cm, indicated by the dotted vertical line. Dashed curves: power spectra of primary mirror distortions from groundbased laboratory measurements of the surface, assuming a Gaussian form for the twopoint correlation function, with correlation length 9.3 cm for the A side and 11.3 cm for the B side; normalized to points with f < 0.05 cm^{1} (top) and f < 0.04 cm^{1} (bottom). 
PNG (17 kb  1024x1414 pixels) PNG (42 kb  2048x2828 pixels) EPS (51 kb) 

Fig. 6 Aside augmented beam profiles (green) compared to Moon sidelobe data (red). K1 and Ka1 appear to be dominated by diffuse reflection rather than the extended main beam, and so are excluded from the fit. Conspicuous DAtoDA differences are seen in the quality of the fit, e.g., W1 and W2 as compared to V1 and V2. Contamination of the fit by diffuse reflected light cannot be ruled out even in DAs other than K1 and Ka1; thus, the Moon data are best considered as upper limits. Vertical line: maximum radius of Moon data included in fit, for DAs Q1–W4. 
PNG (22 kb  1024x1293 pixels) PNG (53 kb  2048x2585 pixels) EPS (132 kb) 

Fig. 7 Fractional error in hybrid beam solid angle as a function of hybridization threshold, for the V2 DA. As the hybridization threshold is raised, noisy Jupiter data are excluded, so that the Jupiter data uncertainty (red) falls. At the same time, the model uncertainty (blue), estimated as a scaling error of 100%, increases because more of the twodimensional model beam is used. These contrary slopes produce a welldefined minimum in the total error (green). Hybridization threshold values for the fiveyear analysis are chosen near the location of this minimum, as shown by the vertical line. 
PNG (13 kb  731x1020 pixels) PNG (30 kb  1463x2040 pixels) EPS (42 kb) 

Fig. 8 (Top) Symmetrized radial profiles of hybridized, binned flight beams for the V2 DA. The central, high S/N part of the beam is taken directly from flight data of the planet Jupiter, whereas the part of the beam below a given gain cutoff is taken from beam models. The set of beam models shown comprises several normalizations of the extrapolated primarymirror distortions; the normalization favored in the analysis is zero, meaning that the extrapolated distortions are omitted. The noisefree lines at radii 3°  4° are portions of the lowernormalization profiles that include model points only. (Middle) Same profiles as in the top panel, after subtraction of the one with normalization 1.0. (Bottom) Beam transfer functions corresponding to the depicted beam profiles, relative to the one with normalization 1.0. Cf. Figure 9, especially the bottom panel. The beam transfer functions at l ≥ 100 are close to what is expected from the noisefree simulations, implying good solid angle recovery. 
PNG (30 kb  1024x1454 pixels) PNG (101 kb  2885x4095 pixels) EPS (164 kb) 

Fig. 9 (Top) Radially binned noisefree simulations of hybridized beams for the V2 DA. The central part of the beam is taken from the adopted DADRA model without augmented distortions. The part of the beam below a given gain cutoff is taken from augmented beam models with various normalizations of the extrapolated primarymirror distortions. The analysis of hybrid beams including real Jupiter data favors the zero normalization, meaning that the extrapolated distortions are omitted. (Middle) Same profiles as in the top panel, after subtraction of the one with normalization 1.0. (Bottom) Beam transfer functions corresponding to the depicted beam profiles, relative to the one with normalization 1.0. Fits using flight hybrid beams should approximate the curves shown here. 
PNG (19 kb  1024x1420 pixels) PNG (50 kb  2048x2840 pixels) EPS (160 kb) 

Fig. 10 Consistency of CMB power spectra across frequency bands, for window functions derived from various normalizations of the extrapolated primary mirror distortions. Complete omission of such extrapolated distortions from the main beam model is justified by this criterion. The CMB spectra and unresolved pointsource coefficients (A_{ps}) in this plot are from a preliminary stage of analysis and are not the final fiveyear WMAP results. (a)(e): Mean of yeartoyear cross spectra in each DA, relative to the final combined power spectrum from the threeyear analysis. The applied w_{l} are derived from hybridized beams in which the tail is from a beam model with extrapolated primary mirror distortions; hybridization thresholds in each DA optimize solid angle error for the nominal amplitude of these added distortions (Figure 5). Spectrum is binned in l with a bin size of Δl = 35. The panels differ in the scaling of the extrapolated distortion amplitude on the mirror: (a), 2; (b), 1; (c), 0.5; (d), 0.1; (e) no extrapolated distortions. (f) Scatter among the DAs in each l bin for the five normalizations of the extrapolated mirror distortions. Omitting the extrapolated distortions (Norm 0, black) minimizes the scatter in the CMB power spectrum over most of this l range, which includes the first peak near l ~ 200. 
PNG (30 kb  1028x1305 pixels) PNG (69 kb  2048x2600 pixels) EPS (137 kb) 

Fig. 11 Relative error in beam transfer functions (Δb_{l}/b_{l}) for the fiveyear beams (black) vs. the threeyear errors (red) (Jarosik et al. 2007). The fiveyear uncertainties are typically a factor of ~ 2 better than threeyear uncertainties. 
PNG (26 kb  1024x1390 pixels) PNG (62 kb  2048x2780 pixels) EPS (163 kb) 

Fig. 12 Relative error component that is due to estimated beam errors in the final WMAP TT power spectrum, which is combined from V and W band data. Cyan: eight independent instances of the square root of the diagonal of the covariance matrix for the coadded VW C_{l}; each instance is based on 5000 Monte Carlo realizations of V and W beam errors. Black: the instance that has been chosen for the C_{l} error bar, because it is approximately the upper envelope. 
PNG (8 kb  714x1024 pixels) PNG (19 kb  1427x2048 pixels) EPS (33 kb) 

Fig. 13 Consistency within ~ 1σ of fiveyear beam transfer functions with threeyear beam transfer functions. Black: difference of threeyear minus fiveyear b_{l}, relative to the fiveyear b_{l}; the lowl rise or fall for several DAs reflects solid angle changes detailed in Figure 14. Red: threeyear 1σ errors. For this plot, the beam profiles used to compute b_{l} are extended by including the profile of the inner portion of each sidelobe response pattern, and the resulting composite profiles are integrated out to the fiveyear transition radii. This removes the effect of the larger fiveyear transition radius. 
PNG (31 kb  1024x1407 pixels) PNG (71 kb  2048x2815 pixels) EPS (165 kb) 

Fig. 14 Much of the solid angle change between the threeyear and fiveyear beams arises inside the threeyear mainbeam boundaries. In this figure, the beam profiles are extended to a radius of 10° using the threeyear or fiveyear sidelobe response pattern, respectively, and the beams are normalized to give the same b_{200} for both three years and five years. Solid angle changes by ΔΩ_{B}/Ω_{B} < 0.5% for the K1, Ka1, Q1, Q2, and V1 DAs, and by 0.8% ≤ ΔΩ_{B}/Ω_{B} ≤ 1.5% for the V2 and W1–W4 DAs. Left : Fiveyear symmetrized hybrid beam profiles (red) and threeyear Hermitefitted beam profiles (black) for selected DAs. The fiveyear profiles include Jupiter data and so are noisy, whereas the threeyear profiles are the functional fit only. Dashed line: threeyear transition radius (Table 3). Dotted line: radius where fiveyear hybrid beams consist of 50% data and 50% model. Right : Change in beam solid angle from the three to the fiveyear analysis, as a function of radius, in annuli of 1°, expressed as a percentage of the fiveyear Ω_{B}. Dashed lines: Transition radii for three years and five years, respectively. Dotted lines: 50% data radius of hybrids, as in left column. 
PNG (28 kb  1024x1215 pixels) PNG (68 kb  2048x2429 pixels) EPS (224 kb) 

Fig. 15 Comparison of WMAP observations (Table 7) to the Mars brightness model of Wright (1976, 2007), evaluated at a wavelength of 3.2 mm (W band). Mean WMAP meaurements are shown for each observing season (diamonds), with error bars indicating the scatter among WMAP DAs W1–W4. Model values (red) are rescaled by 0.9 to bring them into overall agreement with the measurements; thick portions of the line indicate observing seasons. WMAP data are referenced to a fiducial distance of 1.5 AU and a solid angle of Ω^{ref}_{Mars} = 7.156×1010 sr (Hildebrand et al. 1985). 
PNG (11 kb  723x1024 pixels) PNG (25 kb  1446x2048 pixels) EPS (41 kb) 

Fig. 16 Seasonbyseason WMAP radiometry of Saturn in the W band (Table 7). Diamonds: Mean WMAP measurements for each of eight observing seasons, with error bars indicating the scatter among WMAP DAs W1–W4. WMAP data are referenced to a fiducial distance of 9.5 AU, corresponding to a Saturn solid angle of Ω^{ref} _{Sat} = 5.101 × 109 sr (Hildebrand et al. 1985). Red line: simple fitting model of the form T_{Sat} = T_{0} + α sin i, where i is the inclination of the ring plane from our line of sight. Thick portions of the line indicate observing seasons. Fitted parameters are α = 132 ± 16 and T_{0} = 102 ± 7. 
PNG (11 kb  723x1024 pixels) PNG (26 kb  1446x2048 pixels) EPS (37 kb) 