Abstract:While most work on the quantum simulation of chemistry has focused on computing energy surfaces, a similarly important application requiring subtly different algorithms is the computation of energy derivatives. Almost all molecular properties can be expressed an energy derivative, including molecular forces, which are essential for applications such as molecular dynamics simulations. Here, we introduce new quantum algorithms for computing molecular energy derivatives with significantly lower complexity than prior methods. Under cost models appropriate for noisy-intermediate scale quantum devices, we demonstrate how low-rank factorization and other tomography schemes can be optimized for energy derivative calculations. We numerically demonstrate that our techniques reduce the number of circuit repetitions required by many orders of magnitude for even modest systems, and that the cost of estimating an entire force vector may in some systems be lower than the cost of estimating the energy. In the context of fault-tolerant algorithms, we develop new methods of estimating energy derivatives with Heisenberg limited scaling, incorporating state-of-the-art techniques for block encoding fermionic operators. In contrast to our near-term results, we find that the cost of estimating forces with any of our Heisenberg-limited methods is bounded by the cost of estimating energies, due to inner loops requiring either energy estimation or reflections around the ground state. This implies that applications such as geometry optimization, coupling parameter estimation, and spectral prediction may be practical on fault-tolerant quantum devices, but tractable molecular dynamics simulations of large-scale systems requires further algorithmic advances. https://doi.org/10.1103/PhysRevResearch.4.043210 Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI. Published by the American Physical Society

Exploring parameter dependence of atomic minima with implicit differentiation

Refining Potential Energy Surface through Dynamical Properties via Differentiable Molecular Simulation

Force field optimization by end-to-end differentiable atomistic simulation

Interpolation and differentiation of alchemical degrees of freedom in machine learning interatomic potentials

Ab Initio Validation on the Connection Between Atomistic and Hydrodynamic Description to Unravel the Ion Dynamics of Warm Dense Matter

Inverse molecular design and parameter optimization with Hückel theory using automatic differentiation

Stress and heat flux via automatic differentiation

Automated optimization and uncertainty quantification of convergence parameters in plane wave density functional theory calculations

Optimal design of experiments in the context of machine-learning inter-atomic potentials: improving the efficiency and transferability of kernel based methods

Uncertainty Quantification in Atomistic Simulations of Silicon using Interatomic Potentials

Learning Pair Potentials using Differentiable Simulations

Projected gradient descent algorithm for $\textit{ab initio}$ crystal structure relaxation under a fixed unit cell volume

Learning Interatomic Potentials at Multiple Scales

Differentiable molecular simulation can learn all the parameters in a coarse-grained force field for proteins

De novo exploration and self-guided learning of potential-energy surfaces

In operando active learning of interatomic interaction during large-scale simulations

Specialising neural network potentials for accurate properties and application to the mechanical response of titanium

Functional uncertainty quantification for isobaric molecular dynamics simulations and defect formation energies

Efficient quantum computation of molecular forces and other energy gradients

Diagonalization without Diagonalization: A Direct Optimization Approach for Solid-State Density Functional Theory

A Theoretical Case Study of the Generalisation of Machine-learned Potentials