arXiv:2610.24160
Trajectree - A quantum trajectory-based quantum circuit simulator
Ansh Singal, Allen Zang, Jordan Thomas, Alexander Kolar, Anirudh Ramesh, Joaquin Chung, Rajkumar Kettimuthu, Prem Kumar
Abstract
We introduce Trajectree, a quantum trajectory-based quantum circuit simulator that uses a novel, state-saving partial execution of the quantum circuits to simulate broad and deep noisy quantum circuits efficiently. Trajectree can efficiently generate a large number of Monte Carlo samples by saving partially executed trajectories. We conduct extensive performance analysis of the Trajectree approach, proving the utility of the trajectory saving scheme for simulations of open quantum systems. Using this architecture, we show that Trajectree can simulate standard benchmark noisy circuits over five times faster than Google's QSim, a part of the Cirq package. We also analyze Trajectree's performance for simulating quantum optical systems relevant to communication and computing applications. We demonstrate applications of the optical backend to simulate three optical experiments: CHSH Bell test, entanglement swapping, and optical CNOT gate. This shows Trajectree's potential for simulating noisy qudit and optical systems efficiently.
AI Summary
Trajectree addresses the repeated computation required to generate Monte Carlo trajectories for noisy quantum circuits. It caches intermediate states and Kraus-branch probabilities in a tree, allowing subsequent independent samples to reuse previously computed circuit prefixes. A frequency-based cache policy limits stored trajectories, while a matrix-product-state/operator backend built on Quimb supports qubits and finite-dimensional optical modes. Performance is evaluated numerically using optical entanglement swapping and 25-qubit W-state, quantum-neural-network, and variational-eigensolver circuits. For entanglement swapping, the reported average sample time falls from approximately 55 seconds for the first sample to 0.1 seconds over 1,000 samples; the estimated uncached total is 55,000 seconds versus approximately 100 seconds with caching. Saving about 30 nodes provides similar timing to retaining all approximately 226 explored nodes in that example. Against Google's QSim, the paper reports speedups up to approximately $5\times$ on benchmarks with amplitude damping of 0.1% after each gate on each involved qubit. The optical backend reproduces a CHSH maximum just above 2.30 near mean photon number 0.66, entanglement-swapping visibility of $77\%\pm2\%$ from 100,000 samples versus a theoretical 77.7%, and an optical CNOT coincidence pattern. The stated tradeoffs include cache memory versus recomputation and stronger caching benefits when probability concentrates on few trajectories. Optical simulations require photon-number truncation; increasing mean photon number raises tensor-network bond dimensions and reduces their performance advantage. The implementation evaluated is single-core CPU, with parallelization left for future work.
Error found: The abstract reports a QSim speedup of 'over five times' while the introduction reports 'up to 5×', a minor inconsistency that the supplied text cannot resolve. The broad performance claim is supported by QSim comparisons only for three 25-qubit circuit families with 0.1% amplitude damping, leaving its applicability beyond favorable trajectory distributions unestablished. The approximately 550-fold optical speedup uses an extrapolated uncached baseline from the first sample rather than a measured uncached simulation. The main-text thermal-loss Kraus operators incorrectly incorporate the thermal probability f(n) instead of its square root, disagreeing with the channel weighting in the appendix. The appendix's beamsplitter expansion omits bosonic factorial normalization factors, giving incorrect multiphoton amplitudes and undermining the printed optical model.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 15 pages, 22 figures