[Paper Review] Improving constraints on the neutrino mass using sufficient statistics

M. Wolk, István Szapudi+3 more|arXiv (Cornell University)|Apr 1, 2015

Astrophysics and Cosmic Phenomena1 references3 citations

TL;DR

This paper proposes using sufficient statistics—specifically the logarithmic transform $ A = \ln(1 + \delta) $—to extract maximal cosmological information from large-scale structure data, outperforming the standard matter power spectrum in constraining the sum of neutrino masses. Using $ N $-body simulations and Fisher information analysis, it demonstrates up to an 8-fold increase in information on $ M_\nu $ at $ z=0 $, translating to a nearly threefold tightening of constraints compared to the power spectrum alone.

ABSTRACT

We use the "Dark Energy and Massive Neutrino Universe" (DEMNUni) simulations to compare the constraining power of "sufficient statistics" with the standard matter power spectrum on the sum of neutrino masses, $M_ν\equiv \sum m_ν$. In general, the power spectrum, even supplemented with higher moments of the distribution, captures only a fraction of the available cosmological information due to correlations between the Fourier modes. In contrast, the non-linear transform of sufficient statistics, approximated by a logarithmic mapping A=ln(1+δ), was designed to capture all the available cosmological information contained in the matter clustering; in this sense it is an optimal observable. Our analysis takes advantage of the recent analytical model developed by Carron et al. 2014 to estimate both the matter power spectrum and the A-power spectrum covariance matrices. Using a Fisher information approach, we find that using sufficient statistics increases up to 8 times the available information on the total neutrino mass at z=0, thus tightening the constraints by almost a factor of 3 compared to the matter power spectrum.

Motivation & Objective

• To evaluate whether sufficient statistics—specifically the logarithmic transform $ A = \ln(1 + \delta) $—can extract more cosmological information than the standard matter power spectrum for constraining neutrino mass.
• To quantify the information gain in constraining $ M_\nu $ using Fisher matrix forecasts based on the DEMNUni $ N $-body simulations with massive neutrinos.
• To assess the impact of non-linear gravitational evolution and mode coupling on information loss in standard power spectrum analyses.
• To investigate how observational effects such as shot noise and galaxy bias affect the performance of sufficient statistics in real surveys.
• To compare the performance of sufficient statistics in local versus global density fluctuation definitions, particularly in weak lensing vs. galaxy redshift surveys.

Proposed method

• Utilizes the DEMNUni $ N $-body simulations, which include massive neutrinos with degenerate masses ($ M_\nu = 0.17, 0.3, 0.53 $ eV), to model non-linear structure formation across redshifts.
• Applies the non-linear transform $ A = \ln(1 + \delta) $, which is analytically shown to approximate sufficient statistics and thus capture all available information in the matter field.
• Employs an analytical model by Carron et al. (2014) to compute the covariance matrices for both the matter power spectrum $ P(k) $ and the $ A $-power spectrum $ P_A(k) $.
• Uses a Fisher information matrix approach to forecast the precision of $ M_\nu $ constraints, comparing the information content of $ P(k) $ and $ P_A(k) $ at $ z=0 $ and $ z=2 $.
• Adjusts for observational effects such as shot noise and galaxy bias by modifying the power spectrum estimator to $ P_g(k) = b^2 P(k) $ and incorporating a noise-modified window function $ w(k) $.
• Considers two cases: local density fluctuations (typical for galaxy surveys) and global density fluctuations (relevant for weak lensing), to assess the impact of super-survey mode contributions.

Experimental results

Research questions

• RQ1 How much more information on the total neutrino mass $ M_\nu $ can be extracted using the sufficient statistics transform $ A = \ln(1 + \delta) $ compared to the standard matter power spectrum?
• RQ2 What is the impact of non-linear gravitational evolution on the information content of the matter power spectrum, and can $ A $-statistics recover the lost information?
• RQ3 How does the performance of $ A $-statistics vary between local and global density fluctuation definitions, especially in the context of super-survey mode contributions?
• RQ4 What is the expected information gain from $ A $-statistics when accounting for observational effects like shot noise and galaxy bias in surveys such as SDSS?
• RQ5 Can the $ A $-statistics method be extended to 3D fields and applied to future surveys like LSST or Euclid with realistic observational systematics?

Key findings

• At $ z=0 $, the $ A $-power spectrum provides up to 8 times more Fisher information on $ M_\nu $ than the standard matter power spectrum, significantly improving constraint precision.
• The information gain from $ A $-statistics is reduced to a factor of ~2 at $ z=2 $, indicating that non-linear effects are less detrimental at higher redshifts.
• In the case of global density fluctuations (e.g., weak lensing surveys), the information gain from $ A $-statistics increases to a factor of ~25 at $ z=0 $ and ~8 at $ z=2 $, due to dominant super-survey mode contributions.
• For a realistic SDSS LRG-like survey with $ \bar{n} \sim 5 \times 10^{-4} \, h^3 \text{Mpc}^{-3} $ and bias $ b=2 $, the $ 1\sigma $ error on $ M_\nu $ is reduced from ~0.01 (using $ P(k) $) to ~0.005 (using $ P_{A^*}(k) $), indicating a ~4-fold information gain at $ z=0.3 $.
• The study confirms that the logarithmic transform $ A = \ln(1 + \delta) $ approximates sufficient statistics, capturing all available cosmological information in the matter field, especially in the non-linear regime.
• The results suggest that $ A $-statistics could be a powerful tool for future cosmological surveys, particularly in recovering information lost due to mode coupling in the non-linear regime.