Abstract:We present a method for total energy minimizations and molecular dynamics simulations based either on tight-binding or on Kohn-Sham hamiltonians. The method leads to an algorithm whose computational cost scales linearly with the system size. The key features of our approach are (i) an orbital formulation with single particle wavefunctions constrained to be localized in given regions of space, and (ii) an energy functional which does not require either explicit orthogonalization of the electronic orbitals, or inversion of an overlap matrix. The foundations and accuracy of the approach and the performances of the algorithm are discussed, and illustrated with several numerical examples including Kohn-Sham hamiltonians. In particular we present calculations with tight-binding hamiltonians for diamond, graphite, a carbon linear chain and liquid carbon at low pressure. Even for a complex case such as liquid carbon -- a disordered metallic system with differently coordinated atoms -- the agreement between standard diagonalization schemes and our approach is very good. Our results establish the accuracy and reliability of the method for a wide class of systems and show that tight binding molecular dynamics simulations with a few thousand atoms are feasible on small workstations.

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is to develop an algorithm in which the computational cost has a linear relationship with the system size for total energy minimization and molecular dynamics simulations. Specifically, the author proposes a method based on tight - binding or Kohn - Sham Hamiltonian, aiming to overcome the problem that in traditional methods the computational complexity grows cubically with the system size. ### Main problems and solutions 1. **Computational complexity problem**: - In traditional diagonalization methods, when dealing with large - scale systems, the computational complexity is usually on the order of the cube of the system (\(O(N^3)\)), which makes calculations for systems containing hundreds to thousands of atoms very time - consuming. - The paper proposes a new method. By constraining single - particle wave functions to be localized within specific spatial regions and using an energy functional that does not require explicit orthogonalization or inversion of the overlap matrix, the computational complexity is reduced to the linear level (\(O(N)\)). 2. **Improvement of the energy functional**: - A new energy functional \(E[Q]\) is introduced, where \(Q\) is the truncated series expansion of the inverse of the overlap matrix \(S\). This functional not only avoids explicit orthogonalization constraints but also automatically ensures the orthogonality of wave functions during the minimization process. - It is proved that when the wave functions are localized, the absolute minimum of the new functional corresponds to the Kohn - Sham ground - state energy \(E_0\). 3. **Numerical verification**: - The effectiveness and accuracy of this method are verified through a series of numerical experiments, including calculations of systems such as diamond, graphite, carbon chains, and low - pressure liquid carbon. The results show that even for complex disordered metallic systems (such as liquid carbon), this method can achieve results consistent with the standard diagonalization scheme. ### Key technical points - **Wave function localization**: Each single - particle wave function is restricted to a specific spatial region (called the localization region, LR), which greatly reduces the degrees of freedom required for calculations. - **Energy functional design**: An energy functional \(E[Q]\) with implicit orthogonalization constraints is introduced, and by choosing appropriate parameters \(\eta\) and the order of the series expansion \(N\), it is ensured that it has the correct ground - state energy. - **Linear - scaling algorithm**: Through localizing wave functions and sparse matrix operations, a linear relationship between the computational complexity and the system size is achieved, making it possible to simulate large - scale systems. ### Application prospects This method makes it possible to perform tight - binding molecular dynamics simulations of thousands of atoms on a small workstation, greatly expanding the scope and efficiency of materials science research.

Electronic structure calculations and molecular dynamics simulations with linear system-size scaling

Self-consistent field theory based molecular dynamics with linear system-size scaling

Towards Very Large-Scale Electronic-Structure Calculations

Subsystem Density Functional Theory Augmented by a Delta Learning Approach to Achieve Kohn–Sham Accuracy

Linear Scaling Electronic Structure Calculations With Numerical Atomic Basis Set

Gaussian Statistics of the Cosmic Microwave Background: Correlation of Temperature Extrema In the COBE DMR Two-Year Sky Maps

Scalable fine-grained parallelization of plane-wave-based ab initio molecular dynamics for large supercomputers

Linear-Scaling Subspace-Iteration Algorithm with Optimally Localized Nonorthogonal Wave Functions for Kohn-Sham Density Functional Theory

Linear-scaling ab-initio calculations for large and complex systems

A Sub-Linear Scaling Algorithm for Computing the Electronic Structure of Materials

Linear scaling DFT calculations for large Tungsten systems using an optimized local basis

O(N) methods in electronic structure calculations

Molecular hyperdynamics coupled with the nonorthogonal tight-binding approach: Implementation and validation

Born-Oppenheimer Molecular Dynamics with a Linear Combination of Atomic Orbitals and Hybrid Functionals for Condensed Matter Simulations Made Possible. Theory and Performance for the Microcanonical and Canonical Ensembles

2.5 Million-Atom Ab Initio Electronic-Structure Simulation of Complex Metallic Heterostructures with DGDFT

Two-Level Chebyshev Filter Based Complementary Subspace Method: Pushing the Envelope of Large-Scale Electronic Structure Calculations.

Linear-scaling local natural orbital CCSD(T) approach for open-shell systems: algorithm, benchmarks, and large-scale applications

Large-scale efficient Langevin dynamics, and why it works

Enabling Large-Scale Condensed-Phase Hybrid Density Functional Theory Based $Ab$ $Initio$ Molecular Dynamics I: Theory, Algorithm, and Performance

Efficient All-electron Hybrid Density Functionals for Atomistic Simulations Beyond 10,000 Atoms

Parallel Implementation of Large-Scale Linear Scaling Density Functional Theory Calculations with Numerical Atomic Orbitals in HONPAS.