News

Qubit Pharmaceuticals and Sorbonne Université Develop Polynomial-Time Lie Algebra Generator Pool Construction for Quantum Chemistry

A multi-panel convergence plot comparing standard ADAPT-VQE (with an QEB pool) and MB-ADAPT-VQE (with an Minimal Complete Pool) across molecular systems including LiH, H₆, H₈, and H₂O.

Researchers from Qubit Pharmaceuticals and Sorbonne Université (CNRS Laboratoire de Chimie Théorique) have published a peer-reviewed paper in Communications Physics introducing a polynomial-scaling algebraic framework to construct and verify Minimal Complete Pools (MCPs) for variational quantum eigensolvers (VQE). By mapping sets of Pauli operators to binary anti-commutation adjacency matrices over 𝔽2, the team proved that Lie-algebraic completeness can be certified directly via matrix rank in O(N3) time, eliminating the exponential overhead of explicitly computing nested commutators or product-group dimensions.

Operator Pool Dimensionality Reduction Across Molecular Benchmarks
Molecular SystemQubit Count (N)Standard QEB Pool SizeConstructed MCP / Augmented Pool (MCP + Starters)
LiH (R = 1.3 Å, STO-3G)12 qubits561 operators20 (MCP) / 34 (Augmented)
H6 Chain (R = 3.0 Å, STO-3G)12 qubits561 operators20 (MCP) / 59 (Augmented)
H8 Chain (R = 0.8 Å, STO-3G)16 qubits1,940 operators28 (MCP) / 174 (Augmented)
H2O (6-31G basis)26 qubits15,275 operators48 (MCP) / 667 (Augmented)

For symmetry-preserving molecular Hamiltonians, the authors showed that generating pools attain a target rank of 2N − 4 while scaling linearly in qubit count rather than the O(N4) scaling of standard unitary coupled-cluster or qubit-excitation-based (QEB) pools. Because a strictly minimal pool can limit optimization directions in practice, the authors augmented the mathematically verified MCP core with physics-informed starter excitations and paired it with a batched adaptive algorithm, MB-ADAPT-VQE, as well as the fixed-ansatz non-iterative NI-DUCC-VQE method.

In GPU-accelerated statevector benchmarks on the Jean Zay supercomputer up to 26 qubits, MB-ADAPT-VQE reduced energy-evaluation overhead by up to two orders of magnitude on strongly correlated and multi-orbital molecules compared with standard QEB-based ADAPT-VQE, while NI-DUCC-VQE scaled beyond its prior 14-qubit ceiling to reach chemical accuracy on 26-qubit H2O. The authors have open-sourced the C++ and CUDA routines for rank-based MCP verification on GitHub.

Read the open-access paper in Communications Physics here, and access the open-source MCP verification code on GitHub here.

In GQI Portal

The players behind the news

The team that writes QCR tracks every company, deal and technology in the GQI Factory, GQI's verified database of the quantum industry. Next up: every story linked to its players, coming to QCR's paid plans.

Leave a comment

All fields are required. Your email address will not be published.