Fragmenting quantum problems cuts simulation time on IBM chips

Researchers at Qclairvoyance Quantum Labs and The University of Arizona have developed FMO-QVM-HCI-HSQD, a new framework designed to overcome limitations in simulating molecules with quantum computers. The work addresses the substantial classical computing demands and lengthy processing times that currently hinder quantum simulations of realistically sized molecules by fragmenting large problems and running those fragments in parallel using virtual regions of the quantum processing unit.

Experiments on IBM’s ibm_marrakesh and ibm_boston processors demonstrate a reduction in two-qubit depth of 39% for the water trimer compared to existing methods, alongside an 82% decrease in QPU time. These results show that combining fragmentation, virtual QPU partitioning and half-qubit sampling makes sample-based quantum simulation of chemically relevant systems feasible on current hardware.

FMO-QVM-HCI-HSQD Framework for Quantum Simulation

The FMO-QVM-HCI-HSQD framework uses fragment molecular orbital decomposition to reduce complex molecular simulations into smaller, independently solvable problems embedded within their electrostatic environment. This approach allows researchers to tackle large molecules by breaking them down into monomers and dimers, significantly reducing the computational burden of each individual calculation. The framework’s design prioritizes efficient resource allocation for sample-based quantum simulations, addressing a key limitation in current quantum computing workflows.

A core innovation within FMO-QVM-HCI-HSQD is the implementation of a half-qubit sampling ansatz, which effectively halves the number of qubits required for a given calculation and reduces the two-qubit depth of the circuits. This reduction in circuit complexity is critical for execution on near-term quantum hardware, enabling the generation of relevant bitstrings that are then expanded using a hardware-efficient configuration interaction (HCI) method. By keeping the variational determinant space compact, the framework minimizes the computational cost associated with post-processing quantum samples.

Further enhancing efficiency, the framework uses quantum virtual machines to concurrently execute independent fragment circuits on disjoint, quality-weighted regions of a single quantum processing unit (QPU). This virtual partitioning, extending DynQ’s region allocation with temporal reuse through a BRB sequence, allows all fragment circuits to run within a single job, streamlining the simulation process. Experiments on IBM Heron processors with fragments of up to 52 qubits demonstrate the framework’s capabilities when applied to the water trimer and the carmofur-M pro complex, comparing results against established classical benchmarks like FMO-FCI, FMO-CASCI and FMO-HCI.

“We presented FMO-QVM-HCI-HSQD, a resource-efficient framework for sample-based quantum simulation of molecules that combines three components,” stated the authors in their report. The team’s work represents a step towards more scalable and accurate quantum simulations of complex chemical systems.

Fragment Molecular Orbital Approach to Qubit Reduction

This efficiency arises from a combined approach using fragment molecular orbital decomposition, a quality-aware quantum virtual machine, and half-qubit sample-based diagonalization, all integrated into a single workflow for molecular simulation. The methodology combines fragment molecular orbital decomposition with sample-based quantum diagonalization and its half-qubit variant to determine the correlated energy of each fragment on quantum hardware, with the overall system energy assembled classically through the FMO expansion.

The fragment molecular orbital method itself approximates the electronic structure of large systems by dividing them into smaller, non-overlapping fragments, solving each individually and accounting for interactions between them. This decomposition is a well-established technique, with earlier approaches including DMET, divide and conquer and localized orbital embedding, all aimed at tackling the complexity of quantum simulations. The authors emphasize that the innovation lies in how these established methods are combined and executed on quantum hardware.

Virtual QPU Partitioning for Parallel Fragment Execution

The fmo-qvm-hci-hsqd framework achieves efficient quantum simulation by strategically dividing computational tasks and allocating virtualized portions of quantum hardware. This approach moves beyond simply sharing a device, instead treating qubit resources as an abstraction to enable parallel processing of fragmented molecular calculations. Specifically, the framework uses a technique called fragment molecular orbital (FMO) decomposition to break down complex systems into smaller, more manageable pieces, aligning with available qubit budgets.

This partitioning is not merely a scheduling problem. Six execution batches were created. This strategy allows researchers to address simulations with 52 spin orbitals for monomers and 104 for dimers, achieving energy deviations from FMO-HCI of only 1.22 mHa, a level considered within chemical accuracy.

Half-Qubit Ansatz Reduces Circuit Depth in Simulations

Selected-CI diagonalization, coupled with a heat-bath CI expansion, allows a limited sample set to generate a subspace sufficient for achieving chemical accuracy. The framework’s gains extend beyond circuit simplification, impacting execution time through parallel processing. This enhanced bitstring quality is a direct result of the reduced circuit depth. The impact of HCI-HSQD is particularly evident when comparing it to other methods like SQD, with the framework achieving a 39% reduction in two-qubit depth under QVM conditions.

“Together these give the HCI-HSQD solver used at every fragment,” the paper states, highlighting the integrated nature of the approach and its potential for broader application in quantum chemistry. The combination of fragmentation, virtual partitioning, and this half-qubit ansatz represents a step toward simulating increasingly complex molecules with limited quantum resources.

Water Trimer Results: Depth and QPU Time Improvements

Simultaneously, the two-qubit depth, a measure of circuit complexity, was lowered by 39% relative to FMO-QVM-SQD, indicating a streamlined computational pathway. The efficiency gains extend to the scale of the computational search, with the framework accessing only 3% of the determinant subspace to achieve accurate energies for the water trimer, a significant decrease from the 40% required by FMO-SQD.

Carmofur-Mpro Complex Achieves Chemical Accuracy

Researchers assessed accuracy against FMO-CASCI, using CASCI energies for monomers and the respective solvers for dimers in both methods. Despite the comparable energy deviations, the framework drastically altered the scale of computational exploration. FMO-QVM-HCI-HSQD accessed only 0.007% of the determinant subspace for the carmofur-Mpro complex, a figure dwarfed by the 40% accessed by FMO-QVM-SQD.

This minimized exploration of the determinant space is intentional, as the team prioritized achieving chemical accuracy with minimal resources, a goal they state is central to their work. “Since our goal is to reach chemical accuracy with minimal resources, we report results at these settings,” the researchers write, emphasizing the efficiency gains.

Determinant Subspace Access and Accuracy Trade-offs

Despite this nearly four orders of magnitude reduction, the framework maintains chemical accuracy, demonstrating its ability to extract relevant information from a significantly smaller computational space. These combined optimizations demonstrate that substantial reductions in computational effort do not necessarily compromise the quality of the results, offering a pathway towards simulating larger and more complex molecular systems.

NISQ Hardware Limitations and Sample-Based Solutions

Virtual partitioning of quantum processing units (QPUs) allows for concurrent execution of fragment circuits, significantly improving hardware use. The FMO-QVM-HCI-HSQD framework uses this capability by assigning disjoint regions of the QPU to individual molecular fragments, enabling parallel computation that reduces overall simulation time. This approach addresses a key limitation of near-term quantum devices, where individual workloads often underutilize available qubits, leaving substantial computational resources idle.

Beyond parallelization, the framework’s efficiency stems from a reduction in circuit complexity. The adoption of a half-qubit ansatz, coupled with the HCI-HSQD solver, minimizes the required circuit depth for generating samples. This is critical given that the concentration of gradients and hardware noise can create barren plateaus, hindering optimization in variational quantum algorithms. By shifting away from a variational loop altogether, sample-based methods circumvent this issue, but often at the cost of increased computational demands.

More like this

Quantum Research NewsA 33% saving in quantum solves with Bayesian weight selection

Quantum AlgorithmsNew feature map boosts quantum neural network accuracy by up to 44%

Quantum AlgorithmsHybrid quantum network flags suspicious accounts with fewer false alarms

Quantum AlgorithmsResearchers Map How Dipoles Tune Polaron Binding Energy