The Signal
Researchers Model Cavity-Loss Solver for MaxCut and MIS
A paper proposes a cavity-based quantum solver that uses photon loss to help escape local minima in MaxCut and Maximum Independent Set. Numerical studies identify favorable parameter regions and estimate time complexities for the two proble
Quantum.world Team · Written with AI · fact-checked ·
Photon loss usually drains energy from a quantum system; five researchers propose turning that loss into a route through difficult optimization problems. In a paper submitted to arXiv on October 29, Zhi-Hao Huang, Tian-Mu Hu, Shuaifan Cao, Ye Lu and Xiaopeng Li model a cavity-based solver for MaxCut and Maximum Independent Set.
A cavity offers a way out of local minima
The proposal encodes an optimization problem in an atomic-spin Hamiltonian, whose lowest-energy states represent optimal solutions. The spins couple to a single-mode optical cavity, whose frequency can be tuned to connect spin configurations with particular energy differences. Coherent transitions help the system move between configurations; photons leaking from the cavity then carry energy away.
That combination is meant to help the system leave a metastable local minimum—a configuration that cannot improve through a nearby single-spin flip—and continue toward a lower-energy state. The authors call the mechanism dissipative false-vacuum decay, borrowing a term for escape from a long-lived, higher-energy state.
These results highlight cavity dissipation as a controllable resource for quantum optimization.
Zhi-Hao Huang, Tian-Mu Hu, Shuaifan Cao, Ye Lu and Xiaopeng Li, authors of the paper, write in arXiv:2610.39266.
Simulations find parameter windows for two problems
The paper tests the method on MaxCut and Maximum Independent Set using random graphs with an average degree of three. In parameter scans for 10-vertex, 3-regular graphs, the researchers average results across 10 graph instances. They find higher ground-state preparation probabilities for MaxCut around cavity detunings of 3.0–3.5 and 6.0–6.5, with coupling near 0.25–0.3 in the paper’s energy units; for Maximum Independent Set, the favorable region lies near detuning 2.0 and coupling 0.15.
The best settings do not all match a single spin-flip energy exactly. The authors suggest that several resonant and near-resonant pathways contribute, while intermediate coupling balances movement between states against energy selectivity. Photon leakage makes the energy transfer irreversible, helping reduce the chance that population simply returns to its starting configuration.
Finite-size results use two different models
For its finite-size analysis of the full cavity solver, the team examines graphs from 5 to 12 vertices and takes median mixing times across 100 instances per problem. MaxCut’s mixing time grows roughly linearly with graph size in these simulations; Maximum Independent Set mixing times fluctuate around 200 time units without a clear systematic increase. Combining those results with estimated ground-state preparation probabilities, the authors derive time-complexity estimates of O(N·20.16N) for MaxCut and O(20.25N) for Maximum Independent Set.
The paper compares those estimates with Grover search’s O(20.5N) oracle-query scaling and reports more favorable estimated time complexities for its solver. The supplementary material also reports a separate effective-model dataset for graphs up to 18 vertices; the full, finite-coupling solver dataset covers graphs up to 12 vertices.
The next test is whether the cavity can scale
The authors describe possible extensions, including adding a cavity-cooling stage after quantum annealing and using multiple cavity modes to address different spin-flip energies. Both ideas remain proposals in the paper. Its concrete result is a numerical study of small instances showing that, under selected cavity settings, photon loss can assist rather than simply disrupt the search for low-energy solutions.