From Cosmic Morphology to Survey-Ready Inference: Simulation-Based, Non-Gaussian Probes of Large-Scale Structure

by Wei Liu (University of Science and Technology of China)

Thursday 5 Feb 2026, 21:00 → 21:20 Asia/Shanghai

Tsung-Dao Lee Institute

Host: Dong Lai
Location: Online
Join Tencent Meeting：https://meeting.tencent.com/dm/iMmsWlsVRwn1
Meeting ID: 613535699 (no password）

Abstract:
Upcoming Stage-IV surveys will deliver exquisite statistical precision, but fully exploiting their information content requires going beyond traditional two-point statistics into the nonlinear, non-Gaussian regime—while still controlling modeling uncertainty and real-world observational systematics. In this talk, I will present my program to turn field morphology into a practical cosmological probe using Minkowski Functionals and Minkowski Tensors, which capture information from all orders of correlation and provide sensitivity to both massive neutrinos and modified gravity. I will highlight recent simulation-based analyses demonstrating how morphology and anisotropy in redshift space help disentangle new-physics signatures, and how emulator-driven forward models enable robust inference from realistic mock catalogs and survey data. I will then describe my contributions within Dark Energy Spectroscopic Instrument to build and validate a unified emulator pipeline for alternative clustering statistics (including wavelets, density-split statistics, voids, and morphology), with an emphasis on end-to-end validation and understanding implicit priors. Finally, I will outline near-term directions: portable simulation+ML pipelines for joint weak-lensing and galaxy-clustering map statistics for Chinese Space Station Telescope, and reconstruction-enhanced Minkowski Functionals for complementary dark-energy constraints with forthcoming DESI data releases.

Bio:

Wei Liu is a Ph.D. candidate in Astronomy at University of Science and Technology of China, advised by Wenjuan Fang, and a visiting graduate student at University of Michigan with Dragan Huterer. His research develops simulation-based and data-driven inference methods for large-scale structure, emphasizing nonlinear and non-Gaussian observables beyond two-point statistics—especially Minkowski Functionals/Tensors—to constrain massive neutrinos and test modified gravity. He contributes to emulator development and validation efforts in DESI and works on tests on gravity in the Effective Field Theory of Dark Energy framework.