Causa prima: cosmology meets causal discovery for the first time

TL;DR

This paper pioneers the application of causal discovery methods to astrophysical data, revealing potential causal links between galaxy properties and supermassive black hole mass, thus opening new avenues for observational cosmology.

Contribution

It introduces the first use of causal discovery algorithms in astrophysics to analyze galaxy and black hole data, uncovering causal relationships not previously established.

Findings

Central density causes SMBH mass

Velocity dispersion causes SMBH mass

Results are robust under data sub-sampling

Abstract

In astrophysics, experiments are impossible. We thus must rely exclusively on observational data. Other observational sciences increasingly leverage causal inference methods, but this is not yet the case in astrophysics. Here we attempt causal discovery for the first time to address an important open problem in astrophysics: the (co)evolution of supermassive black holes (SMBHs) and their host galaxies. We apply the Peter-Clark (PC) algorithm to a comprehensive catalog of galaxy properties to obtain a completed partially directed acyclic graph (CPDAG), representing a Markov equivalence class over directed acyclic graphs (DAGs). Central density and velocity dispersion are found to cause SMBH mass. We test the robustness of our analysis by random sub-sampling, recovering similar results. We also apply the Fast Causal Inference (FCI) algorithm to our dataset to relax the hypothesis of causal sufficiency, admitting unobserved confounds. Hierarchical SMBH assembly may provide a physical explanation for our findings.