Chemically Accurate Molecular Simulations Demonstrated on IQM Sirius Hardware

Why It Matters

Demonstrating chemical accuracy on near‑term quantum hardware proves that superconducting QPUs can reliably accelerate computational chemistry, opening pathways for faster drug discovery and materials design.

Key Takeaways

• SQD uses quantum hardware as configuration sampler, not energy estimator
• Local Unitary Cluster Jastrow yields shallow circuits, achieving chemical accuracy
• LCNot-UCCSD deeper circuits hinder larger molecule simulations on Sirius
• Full 2D PES for water matches exact classical FCI results
• Hybrid SQD+DMET enables drug‑like molecule simulations on 24‑qubit QPU

Pulse Analysis

Quantum chemistry has long been a benchmark for the practical utility of quantum computers, yet achieving "chemical accuracy"—within 1 kcal/mol of exact energies—remains elusive on noisy intermediate‑scale quantum (NISQ) devices. The recent demonstration on IQM's Sirius processor leverages Sample‑based Quantum Diagonalization, a paradigm that treats the quantum processor as a high‑throughput sampler of electron configurations. By offloading the final diagonalization to classical hardware, SQD sidesteps gate‑level errors that traditionally limit energy estimations, positioning superconducting qubits as specialized accelerators rather than standalone solvers.

Algorithmic choice proved decisive. The Local Unitary Cluster Jastrow (LUCJ) ansatz delivered shallow circuit depths, enabling the team to map a 32 × 32 potential energy surface for water with precision indistinguishable from full configuration interaction calculations. In contrast, the Linear‑CNOT Unitary Coupled‑Cluster (LCNot‑UCCSD) required deeper circuits, which degraded performance on larger systems such as ammonia. This comparative analysis offers a clear roadmap for researchers: prioritize low‑depth, hardware‑friendly ansätze when targeting chemically accurate results on current superconducting platforms.

Beyond isolated molecules, the integration of SQD with Density Matrix Embedding Theory (DMET) marks a pivotal step toward scalable quantum‑centric supercomputing. By fragmenting complex molecules and delegating sub‑problems to the quantum processor, the hybrid workflow successfully simulated eight ligand‑like compounds and the antiviral drug amantadine. As quantum hardware matures, such co‑design of algorithms and classical embedding techniques will accelerate drug discovery pipelines and materials innovation, cementing quantum processors as indispensable components of next‑generation high‑performance computing ecosystems.