Morphological Measures of nonGaussianity in CMB Maps

S. F. Shandarin, Hume A. Feldman, Yongzhong Xu, Max Tegmark

Abstract

Testing the Gaussianity of Cosmic Microwave Background (CMB) anisotropy is a fundamental step in validating standard inflationary models. This paper discusses the application of morphological statistics, specifically Minkowski functionals (MFs), to degree-scale CMB maps. We analyze the QMASK map—a combination of QMAP and Saskatoon data—using seven morphological functions: three standard MFs, the number of isolated regions ($N_c$), and the MFs of the largest percolating region. A key contribution of this work is the introduction of a new parameterization, $M_i(A)$, which defines morphological functions in terms of the total area ($A$) of the excursion set rather than temperature thresholds. This method significantly decorrelates the statistics and enhances robustness against noise in the distribution tails. By comparing the QMASK data to 1,000 Gaussian Monte Carlo simulations, we find that the QMASK map is consistent with a Gaussian field, with approximately 20% of simulations showing greater deviation from the mean than the observed data.

Introduction

Most standard cosmological models, particularly those involving simple inflation, predict that initial density fluctuations in the early universe should follow a Gaussian distribution. Detecting non-Gaussianity would be a landmark discovery, potentially pointing toward exotic physics such as cosmic strings or topological defects. However, distinguishing true physical non-Gaussianity from instrumental noise or foreground contamination requires highly robust statistical tools.

Traditional methods, such as the "genus" statistic, often suffer from high degrees of correlation with other measures, limiting the amount of independent information they provide. Furthermore, these statistics can be overly sensitive to fluctuations at extreme temperature peaks and valleys. This paper proposes a morphological framework designed to be both statistically independent and resilient to these "tail" fluctuations.

Methodology: Minkowski Functionals and Area Parameterization

The study utilizes Minkowski functionals (MFs), a set of geometric measures that describe the topology of a pattern. For a two-dimensional CMB map, these include:

  1. Area ($A$): The total surface area of regions above a specific threshold.
  2. Contour Length ($C$): The total perimeter of those regions.
  3. Genus ($G$): The number of isolated "islands" minus the number of "lakes" (holes).

To these, the authors add the number of regions ($N_c$) and the MFs of the single largest "percolating" region, bringing the total to seven diagnostic functions.

The Area-Based Innovation

Traditionally, these functionals are plotted against temperature levels ($\Delta T$). The authors introduce a parameterization based on the total area ($A$) of the excursion set. Instead of measuring the genus at a specific micro-Kelvin threshold, the statistic is measured when a specific percentage of the map (e.g., the hottest 10%) is highlighted. This transformation weights the statistics toward the "bulk" of the data, effectively insulating the results from noisy pixels at extreme temperature values.

Data and Analysis

The primary dataset is the QMASK map, a degree-scale CMB map covering approximately 648 square degrees, created by merging data from the QMAP and Saskatoon experiments. To establish a baseline for Gaussianity, the authors generated 1,000 Gaussian Monte Carlo simulations. These mock maps replicate the sky coverage, noise properties, and power spectrum of the QMASK data, providing a rigorous "apples-to-apples" comparison.

Key Insights

Statistical Decorrelation

A major finding of this study is the power of area-based parameterization to "untangle" different statistical measures. Under traditional temperature-based methods, 16 out of 21 pairs of statistics were found to be almost perfectly correlated (coefficients > 0.9). By switching to area-based parameterization, most of these correlations dropped below 0.5, and 13 pairs showed correlations lower than 0.3. This ensures that each functional provides unique, independent information about the map's morphology.

Percolation as a Diagnostic

The study demonstrates that the largest region in a map undergoes a "phase transition" or percolation. This transition is detectable even in maps with irregular boundaries or holes, making it a particularly powerful diagnostic tool for real-world datasets that lack perfect spherical symmetry.

Results

The analysis concludes that the QMASK map is a typical realization of a Gaussian field. While it is not a "perfect" Gaussian, it is far from an outlier. Specifically, about 20% of the 1,000 Gaussian mock maps differed more from the theoretical mean than the real QMASK map did. This result places the QMASK data well within the expected variance for a Gaussian universe.

Conclusion and Future Outlook

While the QMASK map provided a successful proof-of-concept, its relatively small size limits the utility of "partial" Minkowski functionals (the shapes of individual small islands). However, the authors anticipate that these morphological tools will be significantly more powerful when applied to larger, higher-resolution datasets, such as those from the WMAP (formerly MAP) satellite. These techniques offer a robust path forward for setting quantitative limits on foreground contaminants and searching for the subtle signatures of topological defects in the cosmic microwave background.

Summary updated

Abstract

We discuss the tests of nonGaussianity in CMB maps using morphological statistics known as Minkowski functionals. As an example we test degree-scale cosmic microwave background (CMB) anisotropy for Gaussianity by studying the \qmask map that was obtained from combining the QMAP and Saskatoon data. We compute seven morphological functions $M_i(\dt)$, $i=1,...,7$: six \mf and the number of regions $N_c$ at a hundred $\dt$ levels. We also introduce a new parameterization of the morphological functions $M_i(A)$ in terms of the total area $A$ of the excursion set. We show that the latter considerably decorrelates the morphological statistics and makes them more robust because they are less sensitive to the measurements at extreme levels. We compare these results with those from 1000 Gaussian Monte Carlo maps with the same sky coverage, noise properties and power spectrum, and conclude that the \qmask map is neither a very typical nor a very exceptional realization of a Gaussian field. At least about 20% of the 1000 Gaussian Monte Carlo maps differ more than the \qmask map from the mean morphological parameters of the Gaussian fields.
BibTeX
@misc{shandarin2001morphological,
  author = {S. F. Shandarin and Hume A. Feldman and Yongzhong Xu and Max Tegmark},
  title = {{Morphological Measures of nonGaussianity in CMB Maps}},
  year = {2001},
  eprint = {astro-ph/0107136},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/astro-ph/0107136},
}
Figure