Cosmological simulations of large scale structure and neutrinos

In this talk I will summarize the works I have been doing at KIAA, especially the the topics related to high performance computing of cosmological N-body simulations and their science outcomes. (1) In the world’s largest N-body simulation TianNu we discover the differential neutrino condensation effect, by which neutrino masses could be inferred independently by studying galaxy properties across different regions of the universe. (2) From the experience of application of TianNu, I wrote the next generation of cosmological N-body code “CUBEP3M”, named “CUBE”, which lowers the memory footprint of particle-mesh N-body algorithms down to 6 byte-per-particle, more than an order of magnitude lower than traditional N-body codes. This breaks the imbalance of the memory-computation ratio of N-body algorithms and this memory-lite N-body code can be more easily applied on Xeon-PHI or GPUs. (3) I will also mention recent works on Isobaric Reconstruction method on LSS and BAO, and (4) estimations of 21-cm absorption system surveys.

Haoran Yu
KIAA first floor meeting room
Fri, 2017-04-28 12:00 to 13:00