Toward chemically accurate quantum simulations on quantum hardware: A quantum-HPC hybrid approach
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Toward chemically accurate quantum simulations on quantum hardware".
Paper summary: Kai: To get into the details, the paper describes how they use SQD to decouple the quality of the energy from how well each individual circuit runs. They confine the energetic resolution to an exact diagonalization step over a specific subspace sampled by the hardware.
Mira: They applied this SQD framework to five specific molecules, and they compared two different ansatz designs: Sample-based Quantum Diagonalization with Local Unitary Cluster Jastrow, or SQD LUCJ, versus a Linear-CNOT Unitary Coupled-Cluster Singles and Doubles ansatz, or LCNot UCCSD.
Lev: The paper chose SQD LUCJ as the main thing to test because they found it was more efficient for the hardware and gave them better accuracy when they had to subsample their batches.
Conclusion: Kai: So, looking at this title and what we saw in this paper, it really boils down to finding a robust way to get reliable molecular energy calculations right now on these intermediate-scale quantum processors.
Mira: What the authors demonstrate is that even when you reduce the amount of sampling they do—for example, when subsampling is reduced to p times the size of the symmetry space—SQD LUCJ keeps outperforming LCNot UCCSD because it produces cleaner measurement distributions.
Lev: The implication for us as researchers building on this is that while enough sampling gets you close to the full configuration interaction accuracy, reduced sampling introduces noise that isn't systematic bias; it’s just random geometry-uncorrelated noise in the energy landscape.
Kai: And they showed this can be extended beyond small molecules, using Density Matrix Embedding Theory with SQD LUCJ to get accurate four electron in four orbital active space energies for eight ligand-like molecules and even for amantadine.
Mira: That extension means this approach isn't just a small molecule trick; it shows the practicality of this hybrid quantum classical method when you move to larger systems like those relevant to drug discovery.
Lev: The final point is that they show SQD LUCJ is a robust and hardware-efficient ansatz for near-term quantum simulation, but they are clear that future progress depends on better ways to generate the subspace or maybe moving beyond single reference coupled cluster initialization.
Anurag K. S. V., Ashish Kumar Patra, Manas Mukherjee, Alok Shukla, Sai Shankar P., Ruchika Bhat, Radhika T. S. L., §Jaiganesh G.
Qclairvoyance Quantum Labs · Centre for Quantum Technologies, National University of Singapore Centre for Quantum Technologies
quant-ph, cs.ET, physics.chem-ph, physics.comp-ph
Submitted: 2026-04-02
Updated: 2026-11-18
Comments: 90 pages, 42 figures
Journal ref: J. Chem. Phys. 165, 184111 (2026)
DOI: 10.1063/5.0354467
Code: https://github.com/qiskit-community/ffsim
Project page: https://iqm-finland.github.io/iqm-client/api/iqm
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 100/100
The gist: The work presents a comprehensive experimental demonstration of quantum-computing-based molecular simulation on IQM’s Sirius 24-qubit superconducting quantum processor…
Key concepts
- Sample-based Quantum Diagonalization (SQD)
- This method decouples the quality of the computed energy from circuit execution fidelity. It works by performing an exact diagonalization step only on a small, sampled subspace of the full problem space, making it resilient to noise inherent in current quantum hardware.
- SQD(LUCJ)
- This is the primary ansatz used in the study. It combines SQD with a Local Unitary Cluster Jastrow form. It was chosen because it offered superior hardware efficiency and robust accuracy, especially when sampling was limited, producing cleaner results than other tested ansätze.
- Potential Energy Surface (PES) Mapping
- This involves scanning molecular properties like bond length or angle to map out the energy landscape of a molecule. The study showed that SQD(LUCJ) could generate smooth, well-resolved PES maps for molecules like H2 and H2O, indicating its utility for studying chemical reactions.
- Density Matrix Embedding Theory (DMET)
- This is a technique used to obtain chemically accurate energies for larger systems by embedding the active space of interest into a larger computational framework. Integrating DMET with SQD allowed researchers to accurately simulate complex, ligand-like molecules on the hardware.
Terminology
Summary
The work presents a comprehensive experimental demonstration of quantum-computing-based molecular simulation on IQM’s Sirius 24-qubit superconducting quantum processor, utilizing up to 16 operational qubits across various quantum chemistry tasks. p. 0 The central methodology employed is Sample-based Quantum Diagonalization (SQD), which is designed to decouple the quality of the computed energy from the fidelity of individual circuit executions by confining energetic resolution to an exact diagonalization step over a hardware-sampled subspace, rendering it inherently robust against NISQ hardware noise. p. 4
The SQD framework was applied to five benchmark molecular systems: H 2, LiH, BeH 2, H 2 O, and NH 3. p. 0 The workflow involves classical pre-processing, quantum computation, and classical post-processing. p. 0 The SQD process is detailed in Algorithm 1, which executes the ansatz on the QPU to obtain measurement counts, performs configuration recovery if necessary (Algorithm 1 steps 5-8), selects and subsamples batches based on a parameter epsilon s (Algorithm 1 step 10), and finally solves the fermion problem via Davidson diagonalization to yield the ground-state energy (ESQD). p. 3
Two ansätze were compared: Sample-based Quantum Diagonalization with the Local Unitary Cluster Jastrow (SQD(LUCJ)) and a Linear-CNOT Unitary Coupled-Cluster Singles and Doubles (LCNot-UCCSD) ansatz within the SQD framework. p. 0 While LCNot-UCCSD offers reduced classical pre-computation overhead, SQD(LUCJ) was selected as the primary ansatz due to its superior hardware efficiency and robust accuracy under constrained subsampling conditions. p. 0 The comparison showed that for epsilon s = 108 (where the diagonalization subspace dimension S sub spans the entire symmetry space S, resulting in eta sub = 1. p. 30), both ansätze achieved essentially identical accuracy, with energy deviations from Full Configuration Interaction (FCI) falling below the sub-nanohartree threshold across all molecules. p. 0 When subsampling was reduced to epsilon s = pS, SQD(LUCJ) consistently outperformed SQD(LCNot-UCCSD), as the shallower LUCJ circuits produced cleaner bitstring distributions with higher eta post-cr, enabling a more representative sampled subspace even under reduced budgets, whereas deeper LCNot-UCCSD circuits suffered from greater hardware noise. p. 4
The results for potential energy surface (PES) mapping are particularly significant. p. 0 The 1D-PES scans for H 2 in the STO-3G and 6-31G bases showed that SQD(LUCJ) produced a smooth, well-resolved PES across the entire dissociation pathway under sufficient subsampling (epsilon s = 108). p. 3 For H 2 O in the STO-3G basis, the 2D-PES map (32x32 grid in bond length and bond angle) was numerically indistinguishable from the exact FCI surface when epsilon s = 108. p. 0 Under reduced subsampling (epsilon s = 21), deviations from FCI were governed by the incompleteness of S sub, manifesting as stochastic, geometry-uncorrelated noise in the energy landscape, rather than systematic bias.
Beyond small molecules, the methodology was extended using Density Matrix Embedding Theory (DMET) integrated with SQD(LUCJ). p. 0 This approach obtained chemically accurate 4-electron-in-4-orbital active-space energies for eight ligand-like molecules and for the pharmacologically relevant amantadine molecule (C 10 H 17 N), representing the first experimental demonstration of DMET-SQD at this molecular scale on IQM hardware. p. 0 The final DMET energies obtained using SQD converged reliably within a small number of self-consistency iterations, achieving energies close to the micro-Hartree regime relative to DMET-CASCI, well within the chemical accuracy threshold. p. 0
In summary, the study establishes SQD(LUCJ) as a robust and hardware-efficient ansatz for near-term quantum simulation on current devices. p. 3 The framework demonstrates that while sufficient sampling recovers FCI accuracy, reduced sampling introduces stochastic noise governed by subspace incompleteness, suggesting that future progress requires improved subspace generation methods and potentially extensions beyond single-reference CCSD initialization. p. 4 The integration with DMET successfully extends this capability to larger systems relevant to drug discovery, proving the practicality of this hybrid quantum-classical approach for simulating industrially relevant molecular systems. p. 0
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Adaptive ansatz selection can be used to choose between Sample-based Quantum Diagonalization (SQD) ansätze, specifically comparing Sample-based Quantum Diagonalization with Local Unitary Cluster Jastrow (LUCJ) and Linear CNOT Unitary Coupled-Cluster Singles and Doubles (LCNot-UCCSD). This allows for a direct hardware comparison of
ansatz expressivity and noise sensitivity
to revealcritical trade-offs.
-
Hybrid embedding strategies can be integrated with SQD, such as Density Matrix Embedding Theory (DMET), to obtain
chemically accurate 4-electron-in-4-orbital active-space energies for eight ligand-like molecules, as well as for the pharmacologically relevant amantadine molecular system.
This extends the reach of quantum solversbeyond small molecules to macroscopic systems.
-
The SQD framework can be used to map a two dimensional potential energy surface (2D-PES) for water by varying both bond length and bond angle simultaneously, providing a
considerably more demanding application of the method
relevant to characterizing vibrational frequencies and molecular force fields. -
The system can perform one dimensional potential energy surface (1D-PES) scans for molecules like H2 and HeH+ using STO-3G and 6-31G basis sets, with SQD(LUCJ) reproducing the FCI PES
with perfect visual agreement across the full dissociation pathway under both subsampling conditions.
-
The system can handle complex electronic structures by employing a fragmentation and embedding strategy where SQD is coupled with DMET, which enforces
self-consistency between a fragment and its bath via the one-particle reduced density matrix,
enabling simulations of systems like iron-sulfur clusters.
Sources
- Quantum measurements and the Abelian Stabilizer Problem
- Assessing requirements to scale to practical quantum advantage
- The vast world of quantum advantage
- Demonstrating an unconditional separation between quantum and classical information resources
- Mind the gaps: The fraught road to quantum advantage
- A Perspective on Quantum Computing Applications in Quantum Chemistry using 25--100 Logical Qubits
- Quantum-Selected Configuration Interaction: classical diagonalization of Hamiltonians in subspaces selected by quantum computers
- Closed-loop calculations of electronic structure on a quantum processor and a classical supercomputer at full scale
- Quantum-Centric Algorithm for Sample-Based Krylov Diagonalization
- Quantum chemistry with provable convergence via randomized sample-based Krylov quantum diagonalization
- Exact and Tunable Quantum Krylov Subspaces via Unitary Decomposition
- Auxiliary-field quantum Monte Carlo method with quantum selected configuration interaction
- Enhancing the accuracy and efficiency of sample-based quantum diagonalization with phaseless auxiliary-field quantum Monte Carlo
- Scaling Sample-Based Quantum Diagonalization on GPU-Accelerated Systems using OpenMP Offload
- GPU-Accelerated Selected Basis Diagonalization with Thrust for SQD-based Algorithms
- From Promise to Practice: Benchmarking Quantum Chemistry on Quantum Hardware
- Towards Compact Wavefunctions from Quantum-Selected Configuration Interaction
- HIVQE: Handover Iterative Variational Quantum Eigensolver for Efficient Quantum Chemistry Calculations
- Hybrid Quantum Algorithms for Computational Chemistry: Application to the Pyridine-Li ion Complex
- Quantum-centric simulation of hydrogen abstraction by sample-based quantum diagonalization and entanglement forging
Related papers
- Computational regimes in matrix-product-state-based quantum trajectory simulations
- Transversal surface-code game with reconfigurable qubits
- Fulqrum: a generalized framework for quantum subspace diagonalization
- Flux-Activated Resonant Control of a Bosonic Quantum Memory
- Machine learning the effects of many quantum measurements
- Barriers to dequantization of near-term molecular electronic structure methods