Verifying Quantum Simulations
How to trust the output of a quantum simulator when its regime is, by design, beyond classical checking.
The verification problem
The point of quantum simulation is to reach regimes classical computers cannot. But that raises a hard question: if you cannot compute the answer classically, how do you know the quantum simulator is right? Verification is a genuine scientific concern, especially for analog devices and for claims of quantum advantage.
Cross-validation in tractable regimes
The first line of defense is testing the simulator where classical answers exist, small system sizes, short times, exactly solvable models like the 1D Heisenberg chain or free-fermion systems, then extrapolating trust into the hard regime. Agreement on all reachable checks builds confidence but is not proof for the untested regime.
Physical consistency checks
- Conservation laws: energy, particle number, and symmetries should be preserved.
- Sum rules and known asymptotic behavior of correlation functions.
- Consistency across independent methods (digital versus analog, different encodings).
- Convergence: results should stabilize as Trotter steps or shots increase.
Interactive and cryptographic verification
Theoretical work has produced protocols by which a classical verifier can, under cryptographic assumptions, check a quantum computation through interaction, even one it cannot simulate. Simpler tools include randomized benchmarking and cross-entropy benchmarking to characterize hardware fidelity, though the latter has itself been the subject of classical-spoofing debates in advantage claims.
Error bars, not just answers
A trustworthy simulation reports uncertainty: statistical error from finite sampling, algorithmic error from Trotter truncation, and hardware error from noise. Rigorous digital algorithms have provable error bounds, which is a major advantage over analog simulation, where the error budget is set by uncharacterized control imperfections. Verification is thus tied to methodology: the more an algorithm comes with proven bounds and passes physical consistency checks, the more its extrapolation into the classically hard regime can be trusted. For any engineering use, quantum-simulation results would need this full verification pedigree before informing design, the same standard applied to validated classical codes.