The Quantum Adiabatic Theorem for Non-Hermitian Dynamics

Technical Summary: The Quantum Adiabatic Theorem for Non-Hermitian Dynamics

Problem Statement
The quantum adiabatic theorem is a cornerstone of quantum mechanics, asserting that a system governed by a slowly varying Hamiltonian remains in its instantaneous eigenspace. However, the standard formulation relies heavily on the Hermiticity of the Hamiltonian, which ensures unitary evolution and the existence of orthogonal spectral projections. For non-Hermitian Hamiltonians, which describe open quantum systems and PT-symmetric systems, the evolution is not unitary with respect to the standard inner product, and the spectral structure is not preserved in a straightforward manner. Consequently, standard resolvent-projection techniques and adiabatic estimates cannot be directly applied. While previous works have addressed adiabaticity in open systems or specific non-Hermitian cases using biorthogonal methods or complex scaling, a general operator-theoretic framework for finite-dimensional non-Hermitian Schrödinger dynamics with real spectra remains to be fully established.

Methodology
The authors develop a framework that bridges non-Hermitian dynamics with the well-understood Hermitian adiabatic theory through a time-dependent Dyson map. The methodology proceeds in three main stages:

1. Construction of a Dynamically Compatible Metric:
   Assuming the non-Hermitian Hamiltonian $\hat{H}_T(s)$ is diagonalizable with a real, non-crossing spectrum, the authors construct a time-dependent metric operator $\hat{\Theta}_T(s)$. This operator is defined via the evolution operator $\hat{U}_T(s)$ and a strictly positive initial metric $\hat{\Theta}_T(0)$, satisfying the relation $\hat{H}_T^*(s)\hat{\Theta}_T(s) - \hat{\Theta}_T(s)\hat{H}_T(s) = \frac{i}{T}\frac{d}{ds}\hat{\Theta}_T(s)$. The positive square root of this metric, $\hat{\eta}_T(s) = \hat{\Theta}_T(s)^{1/2}$, serves as the time-dependent Dyson map.

2. Dyson Transformation to a Hermitian Counterpart:
   The Dyson map transforms the original non-Hermitian state $|\psi(s)\rangle$ into a transformed state $|\phi(s)\rangle = \hat{\eta}_T(s)|\psi(s)\rangle$. This transformation yields a Dyson-transformed Hamiltonian $\bar{H}_T(s)$, defined as:
   $$\bar{H}_T(s) = \hat{\eta}_T(s)\hat{H}_T(s)\hat{\eta}_T^{-1}(s) + \frac{i}{T}\dot{\hat{\eta}}_T(s)\hat{\eta}_T^{-1}(s)$$
   The authors prove that $\bar{H}_T(s)$ is Hermitian. If this transformed Hamiltonian satisfies standard adiabatic assumptions (smoothness, spectral gaps, non-crossing eigenvalues), the quantitative Hermitian adiabatic theorem can be applied to the transformed system.

3. Pullback and Projection Comparison:
   The adiabatic estimate derived for the Hermitian system is "pulled back" to the original non-Hermitian representation. This involves comparing two types of projections:

   • The Dyson-pulled-back projection $\Pi_j(s)$, which corresponds to the instantaneous eigenspace of the Hermitian counterpart in the original representation.
   • The original spectral projection $\hat{P}_j(s)$ of the non-Hermitian Hamiltonian.

   The authors treat the Hermitian counterpart as a perturbation of a similarity transform of the original operator. Using a contour-resolvent estimate, they derive an explicit bound on the difference between $\Pi_j(s)$ and $\hat{P}_j(s)$. This allows the final error estimate to be expressed directly in terms of the spectral subspaces of the original non-Hermitian Hamiltonian.

Key Contributions and Results
The paper establishes a quantitative adiabatic estimate for a class of finite-dimensional non-Hermitian Schrödinger dynamics. The main result (Theorem 12) provides an upper bound on the deviation between the exact evolution and the adiabatic evolution for an initial state in a spectral subspace of the non-Hermitian Hamiltonian.

The final error bound consists of two distinct contributions:

1. The Pulled-Back Hermitian Adiabatic Error: This term arises from the standard adiabatic theorem applied to the Hermitian counterpart $\bar{H}_T(s)$. It depends on the spectral gaps and the derivatives of the transformed Hamiltonian, scaled by the inverse of the total evolution time $T$.
2. The Projection-Comparison Error: This term quantifies the discrepancy between the Dyson-pulled-back spectral projections and the true spectral projections of the non-Hermitian Hamiltonian. It is derived using a contour-resolvent estimate and depends on the norm of the Dyson correction term $B_T(s)$ and the resolvent bounds of the similarity-transformed operator.

The paper demonstrates that under specific conditions (diagonalizability, real spectrum, non-crossing eigenvalues, and a sufficiently large $T$ to satisfy the contour-resolvent condition), the system exhibits adiabatic behavior with an error of order $O(1/T)$.

Illustrative Example
To validate the hypotheses and the structure of the error bound, the authors present a two-level non-Hermitian model. In this example:

• The original Hamiltonian is non-Hermitian with time-dependent coefficients.
• The Dyson map is explicitly constructed and is time-dependent.
• The Hermitian counterpart is non-constant.
• Both the pulled-back Hermitian adiabatic error and the projection-comparison error are shown to be non-zero and explicitly calculable, confirming that the final estimate captures both sources of deviation.

Significance and Scope
The paper claims to provide a rigorous operator-theoretic framework that extends the quantitative quantum adiabatic theorem to non-Hermitian systems with real spectra. Its significance lies in:

• Reformulation: It reformulates the adiabatic mechanism for non-unitary dynamics by mapping the problem to a Hermitian representation via a dynamically compatible metric.
• Explicit Bounds: It provides explicit error bounds that are directly applicable to the spectral subspaces of the original non-Hermitian Hamiltonian, rather than just the transformed representation.
• Decomposition of Error: It clearly separates the adiabatic error into the intrinsic Hermitian adiabatic error and the error introduced by the non-unitary nature of the original system (projection mismatch).

The authors note limitations: the analysis is restricted to finite-dimensional systems, and the construction of the metric operator relies on the exact evolution operator, meaning the framework does not yet provide an independent constructive procedure for determining the Dyson map in arbitrary models without prior knowledge of the dynamics. The work sets the stage for future extensions to infinite-dimensional settings and the development of constructive criteria for Dyson maps.