The Signal
Alice & Bob Team Maps a Dissipative Phase Diagram with 7x Fewer Solves
A variational method cut the number of exact steady-state calculations needed to map a simulated dissipative phase diagram by a factor of seven. The paper also reports a further threefold reduction using a multipoint version and tests the m
Quantum.world Team · Written with AI · fact-checked ·
Seven times fewer exact steady-state calculations: that is how many fewer anchor points variational perturbation theory needed than standard perturbation theory to map a simulated dissipative phase diagram. The result comes from researchers at Alice & Bob, whose paper, published in Quantum on October 22, 2026, targets a practical computational burden: repeating a costly calculation every time a system parameter changes.
Reusing solutions across parameter space
Open quantum systems exchange energy or information with their surroundings. Their long-time behavior is often described by a steady state, which can reveal how a device responds under sustained driving and dissipation. But researchers may need to calculate that state across many combinations of settings—for example, to trace a phase diagram or fit a model to measurements.
For each parameter combination, a direct numerical method can solve the system’s Lindblad master equation. Standard perturbation theory offers a shortcut by expanding the steady state around a reference point, then using that expansion for nearby settings. Its reach is limited, however, and conventional calculations rely on a numerically costly matrix operation called the Moore–Penrose pseudoinverse.
André Melo, Gaspard Beugnot and Fabrizio Minganti replace the usual fixed power-series weights with coefficients chosen through a low-dimensional variational calculation. Their method, which they call variational perturbation theory, reuses perturbative corrections from one reference point to approximate the steady state nearby. The authors also present a multipoint version that combines information from several reference points.
A wider reach, with extra care near transitions
In a driven, dissipative Kerr-resonator model, the team varied detuning and drive strength to map a region that includes a transition between high- and low-photon states. Standard perturbation theory required dense sampling near the transition, where the system changes non-analytically and the Liouvillian gap closes. Variational perturbation theory covered the tested phase diagram with seven times fewer points where the researchers had to calculate the steady state directly.
Using information from neighboring reference points in the multipoint version reduced the number of those direct calculations by a further factor of three compared with the single-point variational method in the same mapping exercise. The benefit has a boundary: the paper says variational perturbation theory performs less well near critical points, and that accurately resolving the transition requires a tighter error tolerance. Increasing the expansion order can extend coverage, but the authors also identify a trade-off: the resulting representation becomes less compact.
“We propose two model-independent methods based on VPT that avoid computing the pseudo-inverse of the Liouvillian, both also enabling the differentiation of the steady state.”
André Melo, Gaspard Beugnot and Fabrizio Minganti, authors of the paper
From resonators to cat-state parameter fitting
The paper tests the approach on more than one model. In a simulated superconducting memory-and-buffer system designed to host Schrödinger cat states, the authors used variational estimates of steady-state gradients to fit three unknown parameters: a coupling strength, a drive amplitude and a Kerr nonlinearity. Starting from a random guess, their optimizer recovered the parameters within 15 iterations on numerically generated, noisy spectroscopy data.
That test is a numerical demonstration, not a report of a new experiment: the team generated the data themselves and added noise to imitate measurement imperfections. The distinction matters for practical use. The method could reduce the repeated calculations involved in parameter fitting, but the paper does not establish that it will recover unknown parameters from every real device or measurement set.
Krylov recycling reaches a 3-by-3 spin model
For larger systems, the researchers also recast the method as a preconditioned Krylov-space calculation, which uses iterative steps rather than a full matrix factorization. They apply it to a dissipative XYZ spin model on a 3-by-3 lattice with periodic boundaries. Symmetries reduce the relevant Liouvillian block from a dimension of about 260,000 to about 15,000; the authors describe full repeated LU factorizations across a phase diagram as prohibitively expensive at the larger dimension.
Across their examples, the authors report computational-cost reductions of up to a factor of 100 compared with direct calculations. That is a maximum across the tested models, not a guaranteed speedup for every system. The paper’s results point to a useful computational strategy for scanning parameter regions, while leaving open how performance will vary with model size, solver choices and the accuracy required near a phase transition.