Optimal limits on f_{NL}^{local} from WMAP 5-year data

TL;DR

This work constrains local-type primordial non-Gaussianity by applying an optimal, foreground-marginalizing estimator to the WMAP5 data, achieving $f_{NL}^{\rm local}=38\pm21$ ($-4< f_{NL}^{\rm local} < 80$ at 95% CL). The method, which downweights foregrounds and avoids post-hoc choices, yields error bars about 40% smaller than earlier suboptimal analyses and shows no residual foreground contamination. Robustness tests across data releases, masks, and foreground treatments indicate no significant non-Gaussian signal, strengthening Gaussian initial-condition constraints and informing inflationary model space.

Abstract

We have applied the optimal estimator for f_{NL}^{local} to the 5 year WMAP data. Marginalizing over the amplitude of foreground templates we get -4 < f_{NL}^{local} < 80 at 95% CL. Error bars of previous (sub-optimal) analyses are roughly 40% larger than these. The probability that a Gaussian simulation, analyzed using our estimator, gives a result larger in magnitude than the one we find is 7%. Our pipeline gives consistent results when applied to the three and five year WMAP data releases and agrees well with the results from our own sub-optimal pipeline. We find no evidence of any residual foreground contamination.

Figures (7)

• Figure 1: Current constraints on $f_{NL}^{\rm local}$. Errors in this figure and throughout the paper are 2-$\sigma$. Panel (a) best results from WMAP 5 years from the WMAP team Komatsu:2008hk and WMAP 3 years from Yadav & Wandelt Yadav:2007yy together with the large scale structure results from Slosar et al Slosar:2008hx and the results from this paper using our optimal method (OPT). Panel (b) comparison of Komatsu:2008hk and Yadav:2007yy for the same choice of analysis parameters ($l_{max}=500$, raw maps and the Kp0 mask). Panels (c) and (d) show the effect of the mask for cleaned and raw maps respectively (from Komatsu:2008hk).
• Figure 2: Constraints on $f_{NL}^{\rm local}$ using 5-year data and KQ75 mask, using both the optimal estimator (squares) and the old estimator applied to clean maps (triangles). The top panel shows cumulative results (constraints using all the information up to a given $\ell$) while the bottom one shows contributions from separate $\ell$ bins. Our overall $f_{NL}^{\rm local}$ estimate, taking $\ell_{\rm max}=750$, is $(38\pm 21)$ for the optimal estimator and $(55\pm 33)$ for suboptimal.
• Figure 3: Estimates of $f_{NL}^{\rm local}$ using the optimal foreground-marginalized estimator with 3-year data and mask (squares), 5-year data with 3-year mask (triangles), and 5-year data and mask (circles). We have shown the contributions from separate $\ell$ bins; the overall estimates of $f_{NL}^{\rm local}$ obtained by summing all bins are $(58\pm 23)$, $(37\pm 21)$ and $(38\pm 21)$ respectively.
• Figure 4: Top panel: Comparison between 3-year results reported in Yadav:2007yy and results obtained from our pipeline, using either the optimal or suboptimal estimator. We apply the suboptimal estimator to 3-year raw maps for consistency with Yadav:2007yy. Bottom panel: Comparison between 5-year results (optimal estimator, raw maps) reported in Komatsu:2008hk and results obtained from our pipeline using the optimal or suboptimal estimator. We apply the suboptimal estimator to 5-year clean maps for consistency with Komatsu:2008hk.
• Figure 5: Comparison between the foreground-marginalized optimal estimator (${\widehat{f}_{NL}}$), and the optimal estimator without foreground marginalization (${\widehat{f}_{NL}}^0$) applied to either raw or clean 5-year maps with KQ75 mask.