Quantum Computers Tackle Radical Molecules for the First Time

Last Updated on June 13, 2025 by Sushanta Barman

In a landmark collaboration, IBM Quantum and Lockheed Martin researchers have demonstrated for the first time that near-term quantum hardware can accurately model the electronic structure of an open-shell molecule—methylene (CH₂)—paving the way for improved simulations of highly reactive radical species in chemistry and materials science.

The research, published in the Journal of Chemical Theory and Computation on May 13, 2025, was conducted by teams from IBM Quantum and Lockheed Martin.

What was Discovered and Why It Matters?

Radical molecules—those with one or more unpaired electrons—present a notorious challenge for classical computational chemistry due to the exponential growth of electron-correlation complexity as system size increases.

In their new paper, researchers applied the sample-based quantum diagonalization (SQD) algorithm to methylene (CH₂), an archetypal open-shell system, using 52 qubits on an IBM quantum processor. “To our knowledge, this is the first study of an open-shell system (the CH₂ triplet) using SQD,” the authors note, underscoring the novelty of their work.

By simulating the dissociation of both the triplet ground state and the singlet excited state of CH₂, the team calculated key properties—including bond dissociation energies and the crucial singlet-triplet energy gap—with remarkable accuracy when benchmarked against high-fidelity classical methods and experimental data.

How was it done?

• Sample-Based Quantum Diagonalization (SQD): SQD is a variational algorithm that uses quantum circuits to sample relevant electronic configurations—or bit-strings—according to their contribution in the target wave function. These samples define a reduced Hamiltonian, which is then diagonalized classically to yield approximate eigenstates. Self-consistent “configuration recovery” restores physical symmetries lost to noise, iteratively refining both the sampled subspace and the predicted energies.
• Variational Ansatz & Orbital Optimization: To prepare initial guesses, the team employed a truncated Local Unitary Cluster Jastrow (LUCJ) ansatz, whose parameters stem from coupled-cluster calculations. Post-SQD orbital optimization using gradient-descent (ADAM) further improved energy estimates, with “warm-starting” ensuring stability along the dissociation pathway.
• Quantum-Centric Supercomputing: Experiments ran on IBM’s Nazca processor, leveraging 52 physical qubits (46 for spin-orbitals plus six ancillae) and executing circuits up to ~3,000 two-qubit gates. Error-mitigation techniques—dynamical decoupling and gate twirling—helped suppress hardware noise, while classical HPC resources managed the heavy lifting of subspace diagonalizations.

Key Results

• Singlet Dissociation Energies: SQD energies deviated by only 1–4 milli-Hartrees from classical references across bond lengths, demonstrating exceptional consistency for a near-term algorithm.
• Triplet Energies at Equilibrium: Near the equilibrium geometry, triplet energies agreed within ~7 milli-Hartrees of classical benchmarks, though discrepancies grew in regions of stronger static correlation.
• Singlet-Triplet Gap: At equilibrium, SQD predicted a 19 milli-Hartree gap, matching classical results and approaching the 14 milli-Hartree experimental value, thanks to beneficial error cancellation.

Implications and Future Directions

Accurate modeling of radical intermediates like CH₂ is crucial for understanding combustion mechanisms, atmospheric chemistry, and interstellar processes, as well as for designing advanced materials and sensors. By establishing SQD as a viable tool for open-shell systems, this work lays a foundation for near-term quantum advantage in chemical simulations.

As quantum hardware gets better—with less noise and some error correction—SQD and similar hybrid methods will be able to study more complex molecules, large catalysts, and reactive intermediates, giving accurate predictions that classical computers cannot easily achieve.

Read the full research paper: I. Liepuoniute et al., “Quantum-Centric Computational Study of Methylene Singlet and Triplet States,” J. Chem. Theory Comput., (2025). DOI: 10.1021/acs.jctc.5c00075