Failure of the mean-field Hartree approximation for a bosonic many-body system with non-Hermitian Hamiltonian

This paper demonstrates that the mean-field Hartree approximation fails for bosonic many-body systems governed by non-Hermitian Hamiltonians, as evidenced by an analytically solvable model where the exact large-$N$ limit diverges from the Hartree prediction and exhibits finite-time transitions to mixed states absent in Hermitian systems.

The Gist

The Big Idea: When the "Average" Lie

Imagine you are trying to predict the weather. Instead of tracking every single air molecule (which is impossible), you look at the "average" temperature and pressure. This is called Mean-Field Theory. It's a shortcut that assumes every particle in a system behaves like the "average" particle, ignoring the messy, complex ways they might be secretly connected to each other.

For a long time, physicists believed this shortcut worked perfectly for almost everything, even when the system was losing energy or particles (like a leaky bucket). They thought they could just tweak the math slightly to handle these "non-Hermitian" (lossy) systems.

This paper says: "Stop. That shortcut is broken."

The authors proved that for certain systems where particles are gained or lost, the "average" behavior does not actually describe what is happening to a single particle. The shortcut fails, sometimes spectacularly.

---

The Analogy: The Dance Floor

To understand why, let's imagine a massive dance floor with $N$ dancers (particles).

1. The Hermitian Case (The Normal Party)

In a normal, closed party (where no one leaves or enters), everyone is dancing to the same beat. If you look at one dancer, they are just doing the "average" dance move of the crowd. Even though they are all connected, the connection is weak enough that the "average" description works perfectly. This is the standard Hartree approximation.

2. The Non-Hermitian Case (The Leaky Party)

Now, imagine a party where people are constantly leaving the room (particle loss) or new people are being teleported in (particle gain). The rules of the game change. The music is different. The authors of this paper set up a specific, simplified version of this "leaky party" using quantum bits (qubits).

They asked: If we watch just one dancer, can we predict their moves by just looking at the "average" dance of the whole crowd?

The Answer: No.

The Two Big Surprises

The paper found two shocking things happen when you try to use the "average" shortcut in this leaky system:

Surprise #1: The "Average" is Wrong (Even when it looks right)

For most starting positions, the single dancer does look like they are in a pure state (they are dancing clearly, not confused). However, their specific dance moves do not match the moves predicted by the "average" equation.

• The Metaphor: Imagine a crowd of people walking in a straight line. The "average" equation predicts they are all walking at 5 mph. But in reality, because of the "leak" (the non-Hermitian part), the crowd is actually walking at 5.1 mph. The difference seems small, but in the quantum world, this tiny error means the prediction is completely wrong. The "average" equation is blind to a hidden time-dependent force that only appears when particles are lost.

Surprise #2: The "Sudden Chaos" (The Mixed State)

This is the most dramatic finding. The authors found a specific starting condition where everything looks fine for a while. The single dancer is dancing clearly (a "pure" state).

But then, at a specific critical time (like a clock striking 12:00), something snaps.

• The Metaphor: Imagine a perfectly synchronized flash mob. Suddenly, at a specific moment, the synchronization breaks. The single dancer you are watching stops being a clear, single entity and becomes a "blur" of possibilities. They become a mixed state.
• In the "average" equation, the dancer should stay clear and synchronized forever. But in the real system, the loss of particles causes the dancer to become confused and mixed up with the rest of the crowd instantly. This phenomenon never happens in normal, non-leaky systems.

Why Should You Care?

This isn't just about abstract math; it has real-world consequences for two big fields:

1. Quantum Computing (The "Nonlinear Solver"):
   There is a proposed quantum algorithm that tries to solve complex, non-linear math problems (like predicting chaotic weather or fluid dynamics) by simulating a huge number of copies of a system. The algorithm relies on the assumption that the "average" behavior of these copies will give the correct answer.

   • The Problem: This paper proves that assumption is false for systems with gain/loss. If you use this algorithm without checking, you might get the wrong answer. It's like trying to bake a cake by averaging the ingredients of 100 other cakes, but forgetting that one ingredient evaporates during the process.

2. Modeling Real Systems (Lasers, Biology, etc.):
   Scientists use these "leaky" equations to model lasers, biological systems, and open quantum systems.

   • The Problem: If you use the standard "average" equation to predict how a laser loses energy, you might be wrong. The system might suddenly become "mixed" or behave unpredictably in a way the standard math doesn't see.

The Takeaway

The authors didn't just say "it's broken"; they built a new, correct map for this specific type of broken system. They showed that to understand these systems, you can't just look at the "average." You have to account for the fact that the "average" changes over time in a way that depends on how many particles are left.

In short: When particles are gained or lost, the crowd doesn't just move as a simple average. They develop hidden correlations that make the "average" lie. If you want to predict the future of these systems, you need a new set of rules.

Technical Summary

1. Problem Statement

The paper addresses the validity of the mean-field Hartree approximation when applied to non-Hermitian many-body systems.

• Context: Mean-field theory (leading to the Hartree or Gross–Pitaevskii equations) is a standard tool for reducing interacting $N$-body bosonic dynamics to a single nonlinear one-particle evolution equation. This relies on the assumption that as $N \to \infty$, entanglement between particles is suppressed, and the one-particle marginal state remains approximately pure (factorized).
• The Gap: While rigorously established for Hermitian Hamiltonians, the validity of this approximation for non-Hermitian Hamiltonians (which model particle gain/loss, open quantum systems, and are used in quantum algorithms for nonlinear differential equations) has not been rigorously justified.
• Core Question: Does the Hartree approximation correctly describe the limit of the one-particle marginal state for generic non-Hermitian bosonic Hamiltonians?

2. Methodology

The authors employ a combination of analytical exact solutions and numerical simulations to construct a counterexample.

• Model System: They analyze a system of $N$ bosonic qubits (dimension $d=2$) with a purely anti-Hermitian two-body interaction Hamiltonian:
  $$H_{eff} = i \sum_{1 \le i < j \le N} Z_i Z_j$$
  where $Z$ is the Pauli-Z matrix. The initial state is a fully factorized product state $|\phi\rangle^{\otimes N}$.
• Exact Solution:
  • They diagonalize the $N$-body evolution operator in the symmetric subspace using a basis of Dicke states $|N, n\rangle$ (states with $n$ excitations).
  • They derive an exact analytical expression for the unnormalized $N$-body wavefunction $|\Psi(N)(t)\rangle$ and the unnormalized one-particle marginal state $\rho^{(1)}(t)$.
• Large-$N$ Limit Analysis:
  • They compute the limit $N \to \infty$ of the normalized one-particle marginal state $\hat{\rho}^{(1)}(t)$ using Laplace's method (saddle-point approximation) on the sum over the Dicke states.
  • They analyze the function $f_t(x)$ governing the exponent of the sum to determine the global maximizers, which dictate the behavior of the limit state.
• Comparison:
  • They derive the non-Hermitian Hartree equation (a nonlinear Schrödinger equation for a single wavefunction $|\varphi\rangle$) based on the standard factorization ansatz $\hat{\rho}^{(2)} \approx \hat{\rho}^{(1)} \otimes \hat{\rho}^{(1)}$.
  • They compare the exact large-$N$ limit of the marginal state against the solution of this Hartree equation using two metrics:
    1. Linear Entropy ($S_L$): To check if the limit state is pure.
    2. Infidelity ($I$): To measure the distance between the exact limit and the Hartree solution.

3. Key Contributions and Results

The paper demonstrates that the Hartree approximation fails for non-Hermitian systems, even in this simple, exactly solvable model. The results are categorized by initial conditions:

A. Failure for Generic Initial Conditions ($|\phi_0|^2 \neq |\phi_1|^2$)

• Result: For almost all initial conditions, the limit $N \to \infty$ of the one-particle marginal state remains pure.
• Discrepancy: However, this pure limit state does not satisfy the non-Hermitian Hartree equation.
• Mechanism: The exact limit state evolves according to a different effective equation that includes an explicit time-dependent term not present in the standard Hartree derivation. The Hartree equation only matches the exact dynamics at $t=0$ or at fixed points.
• Implication: The failure is not due to the emergence of mixedness (entanglement surviving the limit) but due to a structural mismatch in the evolution equations.

B. Finite-Time Mixedness Transition (Critical Initial Condition)

• Result: For the specific initial condition $|\phi_0|^2 = |\phi_1|^2 = 1/2$ (unstable fixed point of the Hartree equation), a phase transition occurs at a critical time $t_c = 1/2$.
  • For $t \le 1/2$: The limit state is pure and matches the Hartree solution.
  • For $t > 1/2$: The limit state undergoes a sharp transition to a mixed state (linear entropy $S_L > 0$).
• Significance: This phenomenon is completely absent in Hermitian systems. It indicates that non-unitarity can amplify correlations such that they survive the mean-field limit, causing the factorization ansatz to break down completely after a finite time. The Hartree equation, which preserves purity, fails to capture this transition.

C. Numerical Verification

• Numerical simulations confirm the analytical predictions.
• The infidelity between the exact finite-$N$ state and the Hartree solution converges to a non-zero constant for generic initial conditions as $N \to \infty$.
• For the critical case, the infidelity remains zero only up to $t=1/2$ and then diverges.
• The convergence speed of the exact state to its large-$N$ limit scales as $O(N^{-1})$ (similar to Hermitian cases) for most times, but changes to $O(N^{-2})$ in the mixed regime of the critical case.

4. Significance and Implications

1. Quantum Algorithms for Nonlinear Equations:

   • The paper challenges the validity of a specific quantum algorithm proposed by Lloyd et al. [30], which uses mean-field embeddings of linear dynamics on multiple copies to simulate nonlinear differential equations.
   • The algorithm relies on the assumption that the Hartree reduction is valid for non-Hermitian Hamiltonians. The authors show this assumption is false; the effective dynamics of the marginal state may differ from the target nonlinear equation, potentially leading to incorrect algorithmic outputs.

2. Modeling Open Quantum Systems:

   • Non-Hermitian Hamiltonians are widely used to model particle loss (e.g., in cold atoms or quantum optics) via the "no-jump" trajectory of Lindblad dynamics.
   • The results imply that using a non-Hermitian Hartree equation to predict one-body observables in lossy systems is not generically justified. The effective dynamics may require additional time-dependent terms or may fail to capture the emergence of mixed states.

3. Theoretical Framework:

   • The work highlights that the standard "product-state ansatz" is insufficient for non-Hermitian mean-field limits.
   • It calls for a new theoretical framework to derive correct effective evolution equations for non-Hermitian systems, potentially involving explicit time dependence or higher-order correlation information that survives the $N \to \infty$ limit.

Conclusion

The authors provide a rigorous counterexample proving that the mean-field Hartree approximation is not universally valid for non-Hermitian bosonic systems. They demonstrate that non-unitarity can lead to two distinct failure modes: (1) a structural mismatch in the evolution equation even when the state remains pure, and (2) a finite-time transition to a mixed state where the mean-field ansatz completely collapses. These findings necessitate caution when applying mean-field theories to open quantum systems or quantum algorithms involving non-Hermitian dynamics.