Los Alamos and QC Ware use sample-based diagonalization on IBM hardware to model iron oxides
A LANL–QC Ware preprint recovers the ground states of FeO and Fe2O2 models with sample-based quantum diagonalization on IBM's ibm_miami, where plain VQE fell short.
Researchers at Los Alamos National Laboratory and the quantum software company QC Ware have shown how a hybrid quantum-classical method can handle small models of iron oxides, a class of materials that conventional simulation methods struggle with. In a preprint titled Towards Quantum Computation of Bulk FeO, posted on October 26, the six authors study molecular FeO and the rhombic dimer Fe₂O₂ using classical exact diagonalization, the Variational Quantum Eigensolver (VQE) and Sample-based Quantum Diagonalization (SQD).
What was announced
The paper's author list includes five Los Alamos researchers and one from QC Ware. According to the abstract, plain VQE did not reach chemical accuracy for the ground state of either molecule. SQD with two "connectivity shells," however, recovered the exact ground-state energy for both while using, in the authors' words, remarkably few quantum computational resources. The method reproduced the experimentally known quintet ground state of FeO and found a nonet ground state for Fe₂O₂.
The technical details
Quantum Computing Report describes the workflow. The team built multi-orbital d–p tight-binding Hamiltonians from localized Wannier functions taken from density functional theory calculations. Circuits used a one-layer Local Unitary Cluster Jastrow (LUCJ) ansatz with nearest-neighbour connectivity, avoiding the overhead of non-local SWAP gates. To sidestep barren plateaus, the circuit parameters were optimized in classical simulation, and the quantum processor was then used only to sample electron configurations.
The experiments ran on IBM Quantum's ibm_miami system. QCR reports that FeO was modelled with 12 electrons in 16 spin orbitals (16 qubits), with six instances run in parallel across 96 qubits, and Fe₂O₂ with 24 electrons in 32 spin orbitals (32 qubits), with three parallel instances. The sampled bitstrings were filtered for the correct particle number, expanded through the nonzero elements of the Hamiltonian, and the resulting smaller problem was diagonalized classically.
According to QCR, raw VQE on FeO left a residual error of about 173 meV, above the 43 meV threshold for chemical accuracy. With two connectivity shells, SQD recovered FeO's experimentally observed quintet state within a compressed subspace of 184 determinants and gave an excitation energy of about 2,167 cm⁻¹.
Why it matters
Bulk FeO is a textbook strongly correlated material: QCR notes that standard density functional theory wrongly predicts it to be metallic, although it is an insulator. The abstract points out that iron-based materials matter for geology and astrophysics and are notoriously hard to calculate classically. The study fits a broader pattern of using today's noisy processors as samplers inside a mostly classical pipeline, rather than asking them to carry out the whole calculation.
What to watch
The work is a preprint and has not been peer reviewed. The molecules studied are small stand-ins for the real target. QCR says the authors outline a route to bulk, periodic FeO by distributing SQD calculations across points in the crystal's momentum space, which would be the real test of whether the approach scales to solids.
Sources
Facts on this page come from the sources below, retrieved by DesignQuantum.