Adiabatic Wave Operators in Quantum Dynamics

• Adiabatic wave operators are a formalism that extends conventional adiabatic theory to describe quantum dynamics when systems deviate slightly from the strict adiabatic limit.
• The framework uses a generalized time-dependent wave operator to map between the evolving active subspace and the instantaneous adiabatic subspace, incorporating non-linear Bloch-type equations.
• It provides practical correction techniques for non-Hermitian and dissipative systems, notably around exceptional points, making it invaluable for precise quantum dynamic simulations.

Adiabatic wave operators generalize conventional time-dependent wave-operator techniques to quantum systems evolving near, but not exactly within, the adiabatic limit. They play a central role in accurately describing quantum dynamics when the system’s true evolution departs non-negligibly from the idealized adiabatic subspace, including cases where the strict adiabatic approximation fails, yet the system trajectory remains within a “limbo” neighborhood of the adiabatic projector. This formalism enables systematic corrections to the adiabatic approximation and provides a rigorous framework for analyzing geometric phases and non-Hermitian Hamiltonians, especially around exceptional points (Viennot, 2013).

1. Standard Time-Dependent Wave Operators and Active Subspaces

Let $\mathcal{H}$ denote a separable Hilbert space and $P_0$ an orthogonal projector of rank $m$ onto a fixed “active subspace” $\operatorname{Ran} P_0 \subset \mathcal{H}$. The quantum evolution is generated by a self-adjoint $H(t)$ via

$$\dot U(t,0) = -\frac{i}{\hbar} H(t) U(t,0),\quad U(0,0) = \mathbb{1}.$$

The instantaneous projector

$P(t) = U(t,0) P_0 U(t,0)^\dagger$

obeys the von Neumann equation: $$\dot P(t) = -\frac{i}{\hbar} [H(t), P(t)],\quad P(0) = P_0.$$ If the trajectory satisfies $\operatorname{dist}(P(t), P_0) < \pi/2$ (measured in Fubini–Study distance), one defines the standard time-dependent wave operator: $\Omega(t) = P(t) [P_0 P(t) P_0]^{-1},$ where the inverse acts on $\operatorname{Ran} P_0$. This operator lifts the effective dynamics residing in the active subspace back to the full Hilbert space.

2. Adiabatic Deformation and the “Almost Adiabatic” Regime

Consider a continuous family of adiabatic projectors $P_0(t)$, each of rank $m$, satisfying

$[H(t), P_0(t)] = 0,\quad P_0(0) = P_0.$

In the strict adiabatic limit, $P(t) \approx P_0(t)$ up to $O(1/T)$ for large slowness parameter $T$. The almost-adiabatic regime relaxes this to

$$\exists~r < \frac{\pi}{2}:\,\forall t,\,\operatorname{dist}(P(t), P_0(t)) \leq r,$$

requiring only that the true subspace $\operatorname{Ran} P(t)$ stays within a bounded (not infinitesimal) neighborhood of the adiabatic one $\operatorname{Ran} P_0(t)$. This accounts for situations where adiabaticity is imperfect but the Hilbert-space trajectory does not stray arbitrarily far.

3. Generalized Time-Dependent Wave Operator: Construction and Dynamics

The core tool is the generalized time-dependent wave operator

$\Omega(t) = P(t) [P_0(t) P(t) P_0(t)]^{-1}$

(with inverse on $\operatorname{Ran} P_0(t)$) mapping between the dynamical and instantaneous adiabatic subspaces. Equivalently, $\Omega(t)$ satisfies a non-linear Bloch-type equation: $$\dot\Omega(t) = -\frac{i}{\hbar} H(t) \Omega(t) + \frac{i}{\hbar} \Omega(t) H(t) \Omega(t) + \Omega(t) \dot\Omega(t), \quad\Omega(0) = P_0(0).$$ Given $\Omega(t)^2 = \Omega(t)$, this can be rewritten in Floquet form: $$[H(t) - i\hbar \partial_t]\,\Omega(t) = \Omega(t)\,[H(t) - i\hbar \partial_t]\,\Omega(t).$$ Effective dynamics within the adiabatic subspace are then governed by the operator

$$H^{\rm eff}(t) = [P_0(t) H(t) - i\hbar\,\dot P_0(t)] \Omega(t),$$

with evolution propagator $U^{\rm eff}(t,0)$ solving

$$\dot U^{\rm eff}(t,0) = -\frac{i}{\hbar} H^{\rm eff}(t) U^{\rm eff}(t,0), \quad U^{\rm eff}(0,0) = P_0(0).$$

The full evolution connects to this effective motion via

$U(t,0) P_0(0) = \Omega(t) U^{\rm eff}(t,0).$

4. Geometric and Dynamical Phases

For the case $\operatorname{Ran} P_0(t)$ one-dimensional, let $|\phi_0(t)\rangle$ be the normalized instantaneous eigenvector, so $P_0(t) = |\phi_0(t)\rangle\langle\phi_0(t)|$. A general solution with initial state $|\psi(0)\rangle = |\phi_0(0)\rangle$ is

$$\psi(t) = c(t) \Omega(t) \phi_0(t),\quad c(t) \in \mathbb{C}^*.$$

Projecting Schrödinger’s equation as detailed in (Viennot, 2013), the transport law becomes

$$\psi(t) = \exp\left\{ -\frac{i}{\hbar} \int_0^t \lambda^{\rm eff}(t') dt' - \int_0^t A(t') dt' - \int_0^t \hat\eta(t') dt' \right\} \Omega(t) \phi_0(t),$$

where the phase terms are

$$\lambda^{\rm eff}(t) = \langle\phi_0(t)|\Omega^{-1} H \Omega|\phi_0(t)\rangle, \qquad A(t) = \langle\phi_0(t)|\dot\phi_0(t)\rangle, \qquad \hat\eta(t) = \langle\phi_0(t)|\Omega^{-1} \dot\Omega|\phi_0(t)\rangle.$$

Compared to the standard adiabatic phases, the dynamical phase is replaced by the effective phase, and there is an additional “wave-operator” geometric phase.

5. Practical Computation of $\Omega(t)$ and Adiabatic Correction

Direct integration of the non-linear Bloch equation for $\Omega(t)$ is computationally intensive in large spaces. To alleviate this, the decomposition

$$\Omega(t) = P_0(t) + X(t), \qquad X(t) = Q_0(t) \Omega(t) P_0(t),\quad Q_0(t) = \mathbb{1} - P_0(t)$$

is used. Define an intertwiner $V(t)$ satisfying

$$\dot V(t) = [\dot P_0(t) P_0(t) + \dot Q_0(t) Q_0(t)] V(t),\quad V(0) = \mathbb{1},$$

so that $V(t) P_0(0) = P_0(t) V(t)$. Then, the reduced variable

$Y(t) = V^{-1}(t) X(t) V(t)$

obeys the standard time-dependent wave-operator equation (for fixed initial active subspace): $$\dot Y(t) = [Q_0(0) - Y(t)]\,\widetilde{H}_{\rm ad}(t)\,[P_0(0) + Y(t)]$$ for

$$\widetilde{H}_{\rm ad}(t) = V^{-1}(t) [H(t) - (\dot P_0 P_0 + \dot Q_0 Q_0)] V(t).$$

Recursive Dimensional Wave-Operator Algorithms (RDWA) or split-operator techniques in Floquet space permit efficient integration on reduced dimensionality, after which the full $\Omega(t)$ and corrected wavefunctions are reconstructed.

6. Application: Non-Hermitian Two-Level Model and Exceptional Point

Consider the non-Hermitian two-level Hamiltonian in basis $|0\rangle, |1\rangle$: $$H(t) = \frac{\hbar}{2} \begin{pmatrix} 0 & W(t)\ W(t) & 2\Delta(t) - i\frac{\Gamma}{2} \end{pmatrix},$$ exhibiting an exceptional point at $(W, \Delta) = (\Gamma/4, 0)$. Let the system’s parameters describe a loop encircling the exceptional point twice in time $T$. The instantaneous eigenprojector for the chosen branch is $P_0(t) = |\phi_0^+(t)\rangle\langle\phi_0^+(t)^*|$, with complementary state $\phi_1^+(t)$. The reduced wave operator reads

$$\Omega(t) = |\phi_0^+(t)\rangle\langle\phi_0^+(t)^*| + x(t) |\phi_1^+(t)\rangle\langle\phi_0^+(t)^*|.$$

The Riccati-type ODE

$$\dot x(t) = \mathcal{A}_{10}^+(t) x(t)^2 - \frac{i}{\hbar} [\lambda_1(t) - \lambda_0(t)] x(t) + \mathcal{A}_{10}^+(t),\quad\mathcal{A}_{10}^+=\langle\phi_1^+*|\dot\phi_0^+\rangle,$$

is directly solvable by first-order schemes for each $t$. The almost-adiabatic wave operator accurately recovers the system’s population transfer across moderate speeds, where the strict adiabatic approximation fails by large margins (Viennot, 2013).

7. Context and Implications

The almost-adiabatic wave-operator formalism by Viennot provides a robust mechanism to systematically correct the breakdowns of strict adiabatic theory without leaving the adiabatic regime’s local topological structure entirely. This allows for controlled descriptions of quantum evolution around singularities such as exceptional points in non-Hermitian systems. A plausible implication is that the approach is particularly well-suited to non-Hermitian or dissipative quantum dynamics where standard adiabatic corrections are insufficient, and that it may generalize to other nearly adiabatic quantum processes in complex parameter regimes (Viennot, 2013).