Parallelizing the Variational Quantum Eigensolver: From JIT Compilation to Multi-GPU Scaling
R. Malarchick, A. Steed, [0.5em] Department of Physical Sciences, Embry-Riddle Aeronautical UniversityDaytona Beach, FL 32114, USA[0.5em] [email protected], [email protected] · 2026
Read Parallelizing the Variational Quantum Eigensolver: From JIT Compilation to Multi-GPU Scaling by R. Malarchick, A. Steed, [0.5em] Department of Physical Sciences, Embry-Riddle Aeronautical UniversityDaytona Beach, FL 32114, USA[0.5em] [email protected], [email protected] (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-2601-09951.txt.
Introduction
R. Malarchick and A. Steed run a key quantum task on a fast chip set. The task finds the low point of the energy of a molecule. It mixes quantum steps with classic steps in a loop. The team used the PennyLane set of tools for the run.
They ran the task for the hydrogen molecule. They swept one bond length at a time, for a full range of lengths. The run used a high end cluster with four large GPUs. The paper is about speed tricks. Just in time code builds and work split across GPUs let the loop scale up.
Catalog record
| CREATOR | R. Malarchick, A. Steed, [0.5em] Department of Physical Sciences, Embry-Riddle Aeronautical UniversityDaytona Beach, FL 32114, USA[0.5em] [email protected], [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-09951
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Parallelizing the Variational Quantum Eigensolver: From JIT Compilation to Multi-GPU Scaling R. Malarchick, A. Steed, [0.5em] Department of Physical Sciences, Embry-Riddle Aeronautical UniversityDaytona Beach, FL 32114, USA[0.5em] [email protected], [email protected] Abstract The Variational Quantum Eigensolver (VQE) is a hybrid quantum-classical algorithm for computing ground state energies of molecular systems. We implement VQE to calculate the potential energy surface of the hydrogen molecule (H 2 ) across 100 bond lengths using the PennyLane quantum computing framework on an HPC cluster featuring 4 × \times NVIDIA H100 GPUs (80GB each). We present a comprehensive parallelization study with four phases: (1) Optimizer + JIT compilation achieving 4.13 × \times speedup, (2) GPU device acceleration achieving 3.60 × \times speedup at 4 qubits scaling to 80.5 × \times at 26 qubits, (3) MPI parallelization achieving 28.5 × \times speedup, and (4) Multi-GPU scaling achieving 3.98 × \times speedup with 99.4% parallel efficiency across 4 H100 GPUs. The combined effect yields 117 × \times total speedup for the H 2 potential energy surface (593.95s → \rightarrow 5.04s). We conduct a CPU vs GPU scaling study from 4–26 qubits, finding GPU advantage at all scales with speedups ranging from 10.5 × \times to 80.5 × \times . Multi-GPU benchmarks demonstrate near-perfect scaling with 99.4% efficiency and establish that a single H100 can simulate up to 29 qubits before hitting memory limits. The optimized implementation reduces runtime from nearly 10 minutes to 5 seconds, enabling interactive quantum chemistry exploration. 1 Introduction 1.1 Background Quantum chemistry calculations help us understand how molecules are structured, how chemical reactions occur, and what properties materials have. A central problem in computational chemistry is finding the ground state energy (lowest energy configuration) and wavefunction (quantum state description) of a molecule. However, exact quantum mechanical calculations become exponentially harder as molecules get larger, making traditional computer methods impractical for large molecules. The Variational Quantum Eigensolver (VQE) is a promising quantum algorithm that combines both quantum and classical computing [ 1 ] . Unlike purely quantum algorithms that need perfect quantum computers, VQE works on today’s noisy quantum computers. The algorithm uses a quantum circuit (called an ansatz) with adjustable parameters to create trial wavefunctions on a quantum processor, while a classical computer adjusts these parameters to find the lowest energy. We focus on the hydrogen molecule (H 2 ), the simplest neutral molecule, which serves as a benchmark system for quantum chemistry methods. Despite its simplicity, H 2 exhibits key features of chemical bonding including equilibrium bond length, dissociation energy, and potential energy surface structure. 1.2 Issues and Questions to be Addressed This work addresses two primary questions: 1. Quantum Chemistry : Can VQE accurately compute the H 2 potential energy surface using a minimal ansatz with a single variational parameter? 2. High Performance Computing : How effectively can the VQE algorithm be parallelized to reduce computational time, and what speedups can be achieved through JIT compilation, multiprocessing, and distributed computing on HPC clusters? The serial implementation provides a baseline, and the parallel versions show good scaling up to 32 processes. 1.3 Related Work Peruzzo et al. [ 1 ] first demonstrated VQE experimentally on a photonic quantum processor. Since then, ansatz design has been a major focus: the Unitary Coupled Cluster (UCC) ansatz [ 2 ] is now standard for molecular simulations, while hardware-efficient ansatzes [ 3 ] reduce circuit depth on noisy devices. On the classical simulation side, several tools have emerged for accelerating variational quantum algorithms. PennyLane [ 4 ] provides automatic differentiation for quantum circuits, and its Lightning backend [ 5 ] offers optimized state-vector simulation. NVIDIA’s cuQuantum [ 6 ] targets multi-GPU simulation, and JAX [ 7 ] enables JIT compilation for numerical code. MPI parallelization of quantum chemistry is well-established [ 8 ] . We apply similar ideas to VQE, benchmarking JIT compilation, GPU acceleration, and MPI distribution on modern hardware. Our contribution is a systematic comparison of these techniques on a single HPC node with multiple GPUs. 2 Problem Description 2.1 The Molecular Hamiltonian Problem The goal is to compute the ground state energy E_{0} of the H 2 molecule as a function of internuclear distance d . The electronic Hamiltonian in the Born-Oppenheimer approximation is: H=-\frac{1}{2}\sum_{i=1}^{2}\nabla_{i}^{2}-\sum_{i=1}^{2}\left(\frac{1}{|\mathbf{r}_{i}-\mathbf{R}_{A}|}+\frac{1}{|\mathbf{r}_{i}-\mathbf{R}_{B}|}\right)+\frac{1}{|\mathbf{r}_{1}-\mathbf{r}_{2}|}+\frac{1}{d} (1) where \mathbf{r}_{i} are electron positions, \mathbf{R}_{A} and \mathbf{R}_{B} are nuclear positions separated by distance d , and atomic units are used. This continuous-space Hamiltonian must be converted to a finite basis set (we use STO-3G, a minimal basis set) and then transformed into qubit operators that quantum computers can work with using a method called the Jordan-Wigner transformation. 2.2 Computational Task The specific computational problem is: • Input : Set of bond lengths \{d_{1},\ldots,d_{100}\} uniformly spaced from 0.1 to 3.0 Å • Output : Ground state energies \{E_{1},\ldots,E_{100}\} at each bond length • Constraint : Each energy must converge to sufficient accuracy (200 VQE iterations) • Objective : Minimize total wall-clock time while maintaining accuracy The key computational challenge is that each bond length requires: • Hartree-Fock calculation to generate molecular Hamiltonian • 200 quantum circuit evaluations with gradient computation • Parameter updates via Adam optimizer This results in 8,000 total circuit evaluations taking approximately 50 seconds in the serial implementation. 3 Model Formulation 3.1 The Variational Principle VQE uses the variational principle from quantum mechanics: for any trial wavefunction |\psi(\theta)\rangle with adjustable parameters \theta , the energy we calculate will always be greater than or equal to the true ground state energy: E(\theta)=\langle\psi(\theta)|H|\psi(\theta)\rangle\geq E_{0} (2) where E_{0} is the true ground state energy and H is the molecular Hamiltonian. By finding the parameters \theta that give the lowest energy E(\theta) , we get a good approximation to the true ground state. 3.2 Molecular Hamiltonian in Second Quantization For the H 2 molecule, the electronic Hamiltonian in second quantization is: H=\sum_{i,j}h_{ij}a_{i}^{\dagger}a_{j}+\frac{1}{2}\sum_{i,j,k,\ell}h_{ijk\ell}a_{i}^{\dagger}a_{j}^{\dagger}a_{k}a_{\ell} (3) where: • h_{ij} are one-electron integrals (kinetic energy and nuclear attraction) • h_{ijk\ell} are two-electron integrals (electron-electron repulsion) • a_{i}^{\dagger},a_{i} are fermionic creation and annihilation operators These integrals are computed using the Hartree-Fock method with the STO-3G basis set, then mapped to Pauli operators on 4 qubits via the Jordan-Wigner transformation. 3.3 Quantum Circuit Ansatz The trial wavefunction is prepared using a parameterized quantum circuit: |\psi(\theta)\rangle=U(\theta)|\text{HF}\rangle (4) where: • |\text{HF}\rangle=|1100\rangle is the Hartree-Fock reference state (both electrons in lowest spatial orbital with opposite spins) • U(\theta) is a unitary operator implemented as a double excitation gate The double excitation gate is: U(\theta)=\exp\left(-i\frac{\theta}{2}(a_{0}^{\dagger}a_{1}^{\dagger}a_{2}a_{3}-a_{3}^{\dagger}a_{2}^{\dagger}a_{1}a_{0})\right) (5) This ansatz captures the most important electron correlation effects in H 2 (both electrons moving together from bonding to antibonding orbitals) while only needing a single adjustable parameter \theta . From a computational perspective, this gate is a parameterized unitary matrix applied to the 2^{N} -dimensional state vector. The optimization problem reduces to finding the parameter \theta that minimizes a matrix expectation value, which maps naturally to gradient-based optimization. 3.4 Optimization Problem The VQE algorithm finds the parameter value that gives the lowest energy: \theta^{*}=\arg\min_{\theta}E(\theta)=\arg\min_{\theta}\langle\psi(\theta)|H|\psi(\theta)\rangle (6) We use the Adam optimizer with: • Learning rate: \alpha=0.01 • Iterations per bond configuration: N_{\text{iter}}=200 • Initial parameter: \theta_{0}=0 (starts at Hartree-Fock state) 4 Methods 4.1 Problem Structure and Parallelization Opportunities The computational task consists of computing the potential energy surface by evaluating E(\theta^{*}) for N_{b}=100 bond