The Clustering of Galaxies in SDSS-III DR9 Baryon Oscillation Spectroscopic Survey: Constraints on Primordial Non-Gaussianity

TL;DR

The study analyzes the SDSS-III DR9 BOSS CMASS galaxy clustering to constrain local primordial non-Gaussianity, f_{0L}^{local}, by modeling a scale-dependent bias on large scales. It introduces a robust framework that accounts for Galactic foreground systematics, particularly stellar density, through a parameterized correction and window-convolved power-spectrum modeling, validated with 600 mock catalogs. The results show no convincing evidence for non-zero f_{0L}^{local}; 95% CL constraints depend on the treatment of systematics, with a fiducial result of -45 < f_{0L}^{local} < 195 and P(f_{0L}^{local}>0) ≈ 91%, and more conservative marginalization broadening the interval to -82 < f_{0L}^{local} < 178. The work highlights the critical role of foreground systematics in large-scale structure f_{0L}^{local} analyses and provides a methodology for robust constraints in future surveys.

Abstract

We analyze the density field of 264,283 galaxies observed by the Sloan Digital Sky Survey (SDSS)-III Baryon Oscillation Spectroscopic Survey (BOSS) and included in the SDSS data release nine (DR9). In total, the SDSS DR9 BOSS data includes spectroscopic redshifts for over 400,000 galaxies spread over a footprint of more than 3,000 deg^2. We measure the power spectrum of these galaxies with redshifts 0.43 < z < 0.7 in order to constrain the amount of local non-Gaussianity, f_NL,local, in the primordial density field, paying particular attention to the impact of systematic uncertainties. The BOSS galaxy density field is systematically affected by the local stellar density and this influences the ability to accurately measure f_NL,local. In the absence of any correction, we find (erroneously) that the probability that f_NL,local is greater than zero, P(f_NL,local >0), is 99.5%. After quantifying and correcting for the systematic bias and including the added uncertainty, we find -45 < f_NL,local < 195 at 95% confidence, and P(f_NL,local >0) = 91.0%. A more conservative approach assumes that we have only learned the k-dependence of the systematic bias and allows any amplitude for the systematic correction; we find that the systematic effect is not fully degenerate with that of f_NL,local, and we determine that -82 < f_NL,local < 178 (at 95% confidence) and P(f_NL,local >0) = 68%. This analysis demonstrates the importance of accounting for the impact of Galactic foregrounds on f_NL,local measurements. We outline the methods that account for these systematic biases and uncertainties. We expect our methods to yield robust constraints on f_NL,local for both our own and future large-scale-structure investigations.

Figures (7)

• Figure 1: The mean power spectrum recovered from the mocks, $P_m(k)$, divided by the input power spectrum used to generate the mock galaxy catalogs that has been first translated to redshift space at $z=0.55$ assuming a real-space bias of 1.889 (the best-fit value) and then convolved with the window applied to the mock footprint. The error-bars are the 1$\sigma$ uncertainties on the mean of the 600 mocks and the dashed lines reflect the standard deviation of the 600 mocks.
• Figure 2: Top panel: The input redshift space power spectrum at $z=0.55$ assuming a real-space bias of 1.889 for the six labeled $f_{\mathrm{NL}}^{\mathrm{local}}$ values, divided by $P^o_g(k)$ for $f_{\mathrm{NL}}^{\mathrm{local}} = 0$. Bottom panel: The same information as the top panel, except the models have now been convolved with the window function and the range of the $P^o_g(k)/P^o_g(k,f_{\mathrm{NL}}^{\mathrm{local}}=0)$ axis has been decreased by more than an order of magnitude.
• Figure 3: The normalized (so that they integrate to 1) probability distributions for the local non-Gaussianity parameter $f_{\mathrm{NL}}^{\mathrm{local}}$, for our four treatment of systematics, applied to the DR9 CMASS sample. The blue curve shows the result using our fiducial treatment which uses the power spectrum, $P_{m,star}$ determined using the stellar density weights; the green curve shows the result when we use the power spectrum measurement, $P_{m,nw}$, that does not include the stellar density weights; the light blue curve shows the result when we marginalize over an additional term $S(P_{m,nw}-P_{m,star})$ in the model and allow it vary within a Gaussian prior of $0\pm0.1$; and the red curve shows the results when we marginalize over $S$ and allow it vary freely.
• Figure 4: Top panel: The measured DR9 CMASS $P(k)$ for the labeled treatment of systematics (points with error-bars) and the associated best-fit model (solid lines). The model with $f_{\mathrm{NL}}^{\mathrm{local}}=0$ and $S=0$ is displayed with a dotted line. We subtract $S(P_{m,nw}-P_{m,star})$ from the measured power spectrum, rather than add it to the theoretical model; in this way the points show the measurement assuming the given systematic treatment is the correct one. Bottom panel: The difference between the logarithms of a given power spectrum and the overall best-fit model power spectrum (which is for $S=0.45$, $f_{\mathrm{NL}}^{\mathrm{local}} =$-48), where the black circles represent the measured power spectrum (using weights for stellar density, $P_{m,star}$) and the lines represent the same four models and use the same scheme as in the top panel.
• Figure 5: The skewness, $G$, of the distributions of the 600 mock power spectra (red) and of the logarithm of the 600 mock power spectra (black). The dashed lines display the expectation of the standard deviation of the skewness of 600 values drawn from a Gaussian distribution (9.975$\times 10^{-2}$).