Can I read Using Matrix-Free Tensor-Network Optimizations to Construct a Reduced-Scaling and Robust Second-Order M{\o}ller-Plesset Theory on EtoBox?

Using Matrix-Free Tensor-Network Optimizations to Construct a Reduced-Scaling and Robust Second-Order M{\o}ller-Plesset Theory by Pierce, Karl; Morales, Miguel is a scholarly article available to read on EtoBox.

What is Using Matrix-Free Tensor-Network Optimizations to Construct a Reduced-Scaling and Robust Second-Order M{\o}ller-Plesset Theory about?

We investigate the application of the canonical polyadic decomposition (CPD) to the tensor hypercontraction (THC) and Laplace transform (LT) approximated second-order M{\o}ller-Plesset (MP2) method. By introducing these decompositions we formally reduce the scaling of the canonical MP2 method from $\mathcal{O}(N^5)$ to $\mathcal{O}(N^3)$ and the storage complexity from $\mathcal{O}(N^4)$ to $\mathcal{O}(N^2)$. We are able to construct the THC representation in $\mathcal{O}(N^3)$ time by employing the interpolative separable density fitting decomposition strategy. Furthermore, we introduce a CPD optimization strategy that takes advantage of the THC representation to decompose the order-four two-electron integral tensor with a computational scaling of $\mathcal{O}(N^3)$. Finally, we show that the rank of the CPD in the approximation of MP2 scales linearly with system size and that this CPD-ISDF-LT MP2 strategy realizes a performance advantage over canonical LT MP2 in both computational wall-times and memory resource requirements.

Author

Pierce, Karl; Morales, Miguel

Published

2025

Language

EN