Parallelization strategies for QTT-based two-particle computations

Develop and evaluate effective parallelization strategies for QTT-based solvers of the parquet and Bethe–Salpeter equations, including distributing MPO–MPO contractions and tensor cross interpolation across computing resources to scale to combined frequency, momentum, and orbital dependencies.

Background

Scaling QTT-based two-particle calculations to models with large orbital and momentum spaces will require distributed-memory and multi-core strategies. Efficient parallelization is nontrivial due to data dependencies within MPO/MPS contractions and adaptive TCI sampling.

Identifying parallelization schemes that preserve numerical stability and compression efficiency would enable large-scale applications on modern HPC systems.