Neglecting primordial non-Gaussianity threatens future cosmological experiment accuracy

Future galaxy redshift surveys aim at probing the clustering of the cosmic large-scale structure with unprecedented accuracy, thus complementing cosmic microwave background experiments in the quest to deliver the most precise and accurate picture ever of our Universe. Analyses of such measurements are usually performed within the context of the so-called vanilla $\mathrm{\Lambda }\mathrm{CDM}$ model—the six-parameter phenomenological model which, for instance, emerges from best fits against the recent data obtained by the Planck satellite. Here, we show that such an approach is prone to subtle systematics when the Gaussianity of primordial fluctuations is concerned. In particular, we demonstrate that, if we neglect even a tiny amount of primordial non-Gaussianity—fully consistent with current limits—we shall introduce spurious biases in the reconstruction of cosmological parameters. This is a serious issue that must be properly accounted for in view of accurate (as well as precise) cosmology.

Figure 1

Galaxy (red) and galaxy cluster (blue) power spectra and their cross-spectrum (magenta) at $z=1$ with $f_{\mathrm{NL}}^{\text{loc}}=-2.17$, 3.51 and 9.3 (solid, short- and long-dashed curves, respectively).

Figure 2

Ratio of the non-Gaussian Schechter mass function to the Gaussian one as a function of mass at $z=1$ with $f_{\mathrm{NL}}^{\text{loc}}=-2.17$, 3.51, and 9.3 (solid, short-, and long-dashed curves, respectively).

Figure 3

Bias on cosmological parameters (in units of the error on the corresponding parameter) induced by neglecting PNG versus $f_{\mathrm{NL}}^{\text{loc}}$ when Fisher matrices for all probes are considered.

Figure 4

Forecast $1\sigma$ joint marginal contours in the $[{\mathrm{\Omega }}_m,\ln ({10}^{10}{A}_s)]$ and $(w_0,w_a)$ planes (left and right panels, respectively) for local- and orthogonal-type PNG with $f_{\mathrm{NL}}^{\text{loc}}=9.3$ and $f_{\mathrm{NL}}^{\text{ort}}=-65$. Solid ellipses are the true error contours, while dashed ellipses come from neglecting $f_{\mathrm{NL}}$ in the analysis. Black and red colors refer to results without and with Planck priors on $\mathrm{\Lambda }\mathrm{CDM}$ parameters, respectively. Top (bottom) panels are for TOT (PSTOT).