Ab Initio Many Body Quantum Embedding and Local Correlation in Crystalline Materials using Interpolative Separable Density Fitting

Junjie Yang, Ning Zhang, Shunyue Yuan, Jincheng Yu, Hong-Zhou Ye, Garnet Chan · 2026

Read Ab Initio Many Body Quantum Embedding and Local Correlation in Crystalline Materials using Interpolative Separable Density Fitting by Junjie Yang, Ning Zhang, Shunyue Yuan, Jincheng Yu, Hong-Zhou Ye, Garnet Chan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-16379.txt.

Introduction

Junjie Yang and coauthors speed quantum sums for crystals with a fit trick. Their 2026 preprint deals with sums that join embed steps and local links. Crystals need sums that use the repeat form of the build. The fit trick splits charge clouds in a way that keeps that form. It is built from points that span the cloud with few terms.

Few terms cut the sum cost to a straight growth in size. Straight growth lets teams treat big crystals with high care. High care sums test new stuff for chips and power gear. The work joins two sum paths in one fast frame. It aims at real stuff, not toy sets.

Catalog record

CREATORJunjie Yang, Ning Zhang, Shunyue Yuan, Jincheng Yu, Hong-Zhou Ye, Garnet Chan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-16379

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Ab Initio Many Body Quantum Embedding and Local Correlation in Crystalline Materials using Interpolative Separable Density Fitting Junjie Yang, Ning Zhang, Shunyue Yuan, Jincheng Yu, Hong-Zhou Ye, Garnet Chan Abstract We present an efficient implementation of ab initio many-body quantum embedding and local correlation methods for infinite periodic systems through translational symmetry adapted interpolative separable density fitting, an approach which reduces the scaling of the calculations to only linear with the number of k-points. Employing this methodology, we compute correlated ground-state coupled cluster energies within density matrix embedding and local natural orbital correlation frameworks for both weakly and strongly correlated solids, using up to 1000 k-points. By extrapolating the local correlation domains and k-point sampling we further obtain estimates of the full coupled cluster with singles, doubles, and perturbative triples ground-state energies in the thermodynamic limit. I Introduction Correlated wavefunction methods have attracted increasing attention in computational materials science, because they provide a pathway to controllable accuracy via systematically improvable many-body expansions 58 ; 64 ; 6 ; 104 ; 43 ; 61 ; 26 ; 111 ; 83 ; 18 ; 89 . However, despite their theoretical advantages, these approaches naively suffer from steep computational scaling with system size, which presents an obstacle to obtaining thermodynamic limit (TDL) results 75 ; 10 ; 4 ; 26 . Consequently, several strategies are commonly used to reduce the cost of such calculations. The first exploits the translational symmetry inherent in periodic systems, by utilizing Bloch-type single-particle orbitals labelled by the crystal momentum k. When a uniform k-point mesh of size N_{\text{k}} (containing the \Gamma point) is employed to sample the Brillouin zone, the calculation is equivalent to one from a periodic supercell containing N_{\text{k}} unit cell images, but with a N_{\text{k}}^{-1} 63 to N_{\text{k}}^{-2} 61 reduction in cost. The second strategy exploits the locality and decay of correlations 71 ; 41 ; 53 . This “near-sightedness” is embodied in different quantum embedding and local correlation frameworks. While there are many variations 48 ; 49 ; 80 ; 81 ; 50 ; 29 ; 32 ; 79 ; 84 ; 27 ; 114 ; 42 ; 2 ; 86 ; 100 ; 25 ; 35 ; 52 ; 107 ; 66 ; 88 ; 76 ; 98 ; 47 ; 113 ; 99 ; 9 ; 62 ; 51 ; 106 ; 11 ; 14 ; 13 ; 12 ; 110 ; 95 , a common feature is that the system is divided into multiple fragments (or clusters), with each fragment subsequently treated independently or with a simplified level of coupling, for example, via the self-consistency of a single-particle quantity in quantum embedding methods 95 , or through the simplified treatment of weak pairs in certain local correlation methods. In this work, for concreteness we will focus on density matrix embedding 40 ; 102 as an example of a quantum embedding method, and the local natural orbital (LNO) method 81 ; 106 as an example of a local correlation framework. Both have been applied with success to crystalline materials in a diverse range of applications 106 ; 28 ; 44 ; 70 ; 45 ; 14 ; 13 ; 12 . The third strategy uses low-rank representations. Low-rank approximations of the Coulomb tensor have been extensively used, such as Cholesky decomposition and resolution-of-the-identity techniques 3 ; 101 ; 38 ; 5 ; 82 ; 77 , pseudospectral approaches 21 ; 65 ; 37 , as well as tensor hypercontraction (THC) 33 ; 68 ; 91 ; 60 . Interpolative separable density fitting (ISDF) has attracted attention as a protocol for generating THC-like factorizations in systems under periodic boundary conditions, where it has demonstrated the potential to accelerate multiple electronic structure methods 34 ; 17 ; 74 ; 73 ; 87 ; 90 ; 112 ; 54 ; 19 ; 112 ; 22 ; 55 ; 20 ; 57 . In this work, we describe an implementation of quantum embedding and local correlation for infinite periodic systems that combines the above reduced-scaling strategies. In particular, our methodology exploits translational symmetry through a tensor hypercontraction representation of the Coulomb interaction, implemented via a symmetry adapted ISDF 108 ; 109 ; 103 ; 72 ; 78 based on crystalline orbitals and utilizing the fast Fourier transform (FFTISDF). The primary algorithmic outcome is the ability to perform many-body quantum embedding and local correlation calculations in materials with only linear scaling with the number of k-points. This enables the practical extraction of thermodynamic limit results. We organize the rest of the work as follows. Section II recalls the FFTISDF algorithm for periodic solids utilizing k-point symmetry, and then describes the efficient implementation of the various components of the density matrix embedding and local natural orbital correlation frameworks in this representation and summarizes the parameters used for the benchmark calculations. Section III provides a set of benchmark studies using coupled cluster theory solvers on representative crystalline systems, including diamond, carbon dioxide, nickel monoxide, and the infinite layer cuprate CaCuO 2 . These calculations employ Gaussian polarized double-zeta basis sets with up to 1000 k-points to demonstrate the method’s scalability, and we extract the thermodynamic limit of the coupled cluster energies for several of the systems. Finally, Section IV summarizes our main findings and discusses potential directions for future work. II Theory Our periodic calculations are performed using a crystalline Gaussian-type orbital(GTO) basis, \phi_{\mu}^{\bm{k}}(\bm{r})=\sum_{\bm{T}}\text{e}^{\text{i}\bm{k}\cdot\bm{T}}\;\tilde{\phi}_{\mu}(\bm{r}-\bm{T}) (1) where \tilde{\phi}_{\mu}(\bm{r}) is a normal GTO centered on an atom in the reference unit cell; \bm{T} is a lattice translation vector; and \bm{k} is a momentum vector sampled in the first Brillouin zone. Our implementation builds upon the Gaussian-Plane-Wave infrastructure provided by the PySCF package 94 ; 93 ; 96 , where the basis functions are evaluated on a uniform grid in the unit cell at a set of points \{\bm{r}\} . (This infrastructure is termed FFTDF in PySCF ). II.1 Periodic k-point FFTISDF representation of integrals The core idea of the ISDF approach is to approximate products of basis functions on the grid points \bm{r} by using values from a select set of interpolating points \bm{r}_{I} (IPs). This reduces the computational complexity of evaluating two-electron integrals while maintaining systematic improvability of the approximation. In this work, the interpolation points are selected using a pivoted Cholesky decomposition of \phi_{\mu}^{\bm{k}_{\mu}}(\bm{r})\phi_{\nu}^{\bm{k}_{\nu}}(\bm{r})^{*} (where \bm{r} is the column index) closely following the procedure outlined in Ref. 59 . Only the largest N_{\text{IP}}=c_{\text{IP}}\times N_{\text{AO}} columns are used, where the pre-defined constant c_{\text{IP}} controls the accuracy 46 . This method is suitable for systems with moderate unit cell sizes. Although alternative approaches, such as the centroidal Voronoi tessellation k-means algorithm 17 , may be more efficient for large unit cells, they are not the focus of this study. The interpolation points are used to approximate the product of basis functions on the uniform grid \{\bm{r}\} : \displaystyle\rho_{\mu\nu}^{\bm{k}_{\mu}\bm{k}_{\nu}}(\bm{r}) \displaystyle=\phi_{\mu}^{\bm{k}_{\mu}}(\bm{r})\;\phi_{\nu}^{\bm{k}_{\nu}}(\bm{r})^{*}=\phi_{\mu}^{\bm{k}_{\mu}}(\bm{r})\;\phi_{\nu}^{-\bm{k}_{\nu}}(\bm{r}) (2) \displaystyle\approx\sum_{I}\xi_{I}^{\bm{q}}(\bm{r})\;X_{I\mu}^{\bm{k}_{\mu}}\;X_{I\nu}^{-\bm{k}_{\nu}} where in the above equation, X_{I\mu}^{\bm{k}_{\mu}}=\phi_{\mu}^{\bm{k}_{\mu}}(\bm{r}_{I}) is the interpolating vector (IV); \bm{q}=\bm{k}_{\mu}-\bm{k}_{\nu}+\bm{B}_{n} due to momentum conservation and is shifted to be within the first Brillouin zone ( \bm{B}_{n} is an integer multiple of the reciprocal lattice vectors); and \xi_{I}^{\bm{q}}(\bm{r}) is the interpolation function. The interpolation function is obtained from solving the least-squares problem corresponding to Eq. 2 , which leads

Read the complete text