Arclight Quantum Cuts Quantum Simulation Cost with New Method

Introduction

Arclight Quantum Computing Inc., in collaboration with researchers from the Chinese Academy of Sciences, has introduced Fast Imaginary‑Time Evolution (FITE), a new probabilistic scheme for quantum simulation. By performing an extensive offline basis transformation that diagonalises the target Hamiltonian, FITE keeps the online computational effort constant regardless of simulation length. Early numerical studies show dramatically lower gate counts compared with traditional Pauli‑based imaginary‑time evolution (PITE) methods while preserving comparable accuracy.

How FITE Works

1. Offline Basis Optimisation – Before any quantum circuit runs, the Hamiltonian is transformed into a diagonal representation using a customised basis. This step is computationally intensive but performed only once per problem class.
2. Probabilistic Imaginary‑Time Evolution – The simulation proceeds with a series of probabilistic operations that effectively evolve the system in imaginary time, steering it toward its ground state or thermal equilibrium.
3. Constant‑Cost Online Phase – Because the diagonal representation is reused, each additional time step requires the same quantum circuit depth, eliminating the linear scaling of gate resources typical of conventional methods.

Reported Benefits

• Gate‑Count Reduction – Benchmarks on Hubbard, XY and XYZ models report online gate‑cost reductions on the order of 10⁴‑10⁵ times relative to first‑order Pauli‑PITE.
• Higher Success Probabilities – Reported state‑acceptance rates reach 0.317 for a two‑qubit Hubbard model, 0.064 for an XY chain, and 4.09×10⁻⁴ for an XYZ chain, outperforming grouped‑filtering approaches on the same models.
• Accuracy Retention – Infidelity remains below 0.005 across tested systems, indicating that the resource savings do not come at the expense of solution quality.

Caveats and Open Questions

While the online savings are striking, the offline preparation incurs a non‑trivial cost. The current benchmarks do not fully account for the time and classical resources needed to compute the diagonal basis or estimate coefficients for larger, more complex Hamiltonians. Consequently, the net advantage of FITE will depend on the balance between one‑time preparation overhead and the number of simulation steps required for a given application. Additionally, success probabilities for some models (e.g., the XYZ chain) remain low, meaning repeated attempts could erode the overall efficiency gain.

Outlook

FITE represents a promising direction for quantum simulation, especially for problems where long‑time evolution or repeated queries are essential—such as material‑science calculations, quantum chemistry, and thermal‑state sampling. Future work will need to:

• Quantify offline preparation costs for realistic, high‑dimensional systems.
• Explore hybrid strategies that combine FITE with error‑mitigation or variational techniques to boost success rates.
• Extend the method to broader classes of Hamiltonians beyond the benchmark models.

If these challenges can be addressed, the constant‑cost online phase of FITE could become a valuable tool for scaling quantum simulations toward the fault‑tolerant era.

--- This article summarises the key ideas and early results of the FITE method without introducing new data or quotations.