The Signal
Hartree Theory Fails for Some Non-Hermitian Boson Dynamics
A paper in Quantum finds that the Hartree approximation can fail to describe bosonic many-body dynamics governed by non-Hermitian Hamiltonians. The authors demonstrate the problem in solvable and numerical examples, including a model whose
Quantum.world Team · Written with AI · fact-checked ·
“Our findings challenge the validity of the mean-field Hartree approximation for non-Hermitian Hamiltonians, and call for additional conditions for the validity of the mean-field regime to model the dynamics of particle gain and loss and the open-system dynamics in bosonic many-body systems.”
Matias Ginzburg, Simone Rademacher and Giacomo De Palma, the paper’s authors
Where the Hartree approximation breaks
A paper published October 22 in Quantum finds that a standard shortcut for describing large bosonic systems can fail when their dynamics are governed by a non-Hermitian Hamiltonian. Hartree theory replaces the many-particle problem with an effective equation for one particle, treated as moving in the average field produced by the others. The approach is widely used in ordinary Hermitian settings, but the authors show that extending it to non-Hermitian dynamics does not always preserve the correct one-particle behavior.
Non-Hermitian Hamiltonians can describe particle gain or loss, and the evolution of an open quantum system conditioned on no quantum jump occurring. The paper’s result matters for those models because a large particle count alone does not guarantee that the single-particle description works. The authors also connect the issue to a proposed quantum algorithm that uses many-copy dynamics to reproduce nonlinear differential equations: that method relies on the Hartree approximation holding for the chosen non-Hermitian dynamics.
An overlooked contribution from other particles
For the general class of bosonic mean-field Hamiltonians with two-body interactions, the authors compare the one-particle state’s early-time evolution with the Hartree equation. They prove that the approximation gives the correct first time derivative at the initial moment, but can give the wrong second derivative even in the limit of infinitely many particles. That discrepancy means the two descriptions can begin to separate immediately, rather than only after a long evolution.
The paper traces the mismatch to anti-Hermitian terms acting on particles other than the one being tracked. In Hermitian dynamics, terms acting only on traced-out particles cancel when calculating the reduced state; the corresponding anti-Hermitian terms do not. Their contribution can scale with the particle number and amplify departures from the factorized state that Hartree theory assumes.
A bosonic-qubit model changes character at t = 1/2
Ginzburg, Rademacher and De Palma demonstrate the problem in an exactly solvable model of N bosonic qubits with a purely anti-Hermitian Hamiltonian. For almost every initial condition they analyze, the one-particle state approaches a pure large-N limit, but that limit follows a different evolution from the non-Hermitian Hartree equation. A pure state is therefore not, by itself, evidence that the Hartree description is correct.
One balanced initial condition produces a sharper change: when the two initial basis-state probabilities are each one-half, the limiting one-particle state follows Hartree theory only until the model’s critical time, t = 1/2. After that point it becomes mixed, while the Hartree solution remains constant. The time is specific to the model studied, not a universal threshold for bosonic systems.
Real eigenvalues do not remove the caveat
The authors also numerically study a non-Hermitian Hamiltonian with a real spectrum. They find initial product states for which the one-particle marginal still fails to converge to the Hartree solution as the particle number grows. That example shows the issue is not confined to the paper’s purely anti-Hermitian model.
The results do not establish that every non-Hermitian Hartree model fails. Instead, the authors argue that particle-loss and open-system calculations need additional conditions or case-by-case checks before relying on the approximation. For quantum algorithms built on many-copy mean-field embeddings, the same test applies: the nonlinear evolution is reproduced only if the specific non-Hermitian system actually has the required Hartree limit.