Quantum Internal Model Principle

• Quantum Internal Model Principle is a theoretical framework that enables perfect decoherence control by embedding a model of environmental disturbance into the control algebra.
• It leverages geometric control techniques and ancillary qubits to extend the control family, ensuring robust disturbance decoupling in open quantum systems.
• Numerical and theoretical analyses confirm that incorporating an internal model of environmental interactions can scale to multi-qubit registers for effective decoherence suppression.

The Quantum Internal Model Principle (QIMP) is a theoretical framework for perfect decoherence control in open quantum systems. It establishes that effective disturbance rejection—specifically, rendering quantum-system observables invariant to environmental decoherence—is possible if and only if the controller embeds a model of the environmental interaction within its control algebra. This provides a unifying and foundational perspective on quantum disturbance decoupling, integrating core ideas from geometric control theory, classical disturbance rejection, and internal model principles from control engineering (Ganesan et al., 2010).

1. Decoherence Control as a Disturbance Rejection Problem

Consider an open quantum system $S$ coupled to an environment $E$, with joint Hilbert space $\mathcal{H}_s \otimes \mathcal{H}_e$. The system evolution is generated by the free Hamiltonians $H_0$ (system), $H_e$ (environment), the system-environment interaction Hamiltonian $H_{SE}$, and system control Hamiltonians $\{H_i\}_{i=1}^r$.

The Schrödinger equation governing the joint state $\xi(t) \in \mathcal{H}_s \otimes \mathcal{H}_e$ is: $$\frac{d}{dt}\xi(t) = \left( H_0 \otimes I_e + I_s \otimes H_e + H_{SE} + \sum_{i=1}^r u_i(t) H_i \otimes I_e \right) \xi(t)$$ where $u_i(t)$ are admissible (generally time-dependent) scalar control inputs.

The output to be protected from decoherence is typically a system operator $C$ (such as a coherence matrix element or nondemolition observable), with output

$y(t) = \langle \xi(t) | C | \xi(t) \rangle.$

Decoherence control is thereby formulated as a disturbance rejection problem: $H_{SE}$ is viewed as an external disturbance and the goal is to render $y(t)$ invariant under the disturbance vector field $K_I(\xi) = H_{SE}\,\xi$. Decoupling is achieved if

$y(\xi(t); H_{SE}) = y(\xi(t); H_{SE}=0)$

for all admissible controls and initial conditions.

2. The Quantum Internal Model Principle: Statement and Conditions

In classical control, disturbance decoupling and the Internal Model Principle are distinct: the former is achievable without explicit knowledge of the disturbance generator, while the latter requires a controller to embed the exosystem. The quantum setting differs crucially:

Quantum Internal Model Principle: Perfect decoupling of a quantum disturbance $H_{SE}$ is achievable if and only if the controller incorporates an internal model of $H_{SE}$ in its control algebra.

This is formalized by requiring the existence of an extended system (system plus ancilla) and analytic, state-dependent control laws $\alpha(\xi)$, $\beta(\xi)$, yielding the affine control law

$$\frac{d\xi}{dt} = (H_0 + H_e + \sum_i \alpha_i(\xi) H_i + H_{SE})\xi + \sum_{i, j} \beta_{ij}(\xi) v_j H_i \xi,$$

such that one can find an involutive distribution $\Delta$ of vector fields on the state manifold satisfying:

• $K_I \in \Delta \subset \ker(dy)$,
• $[\Delta, \tilde K_0] \subset \Delta$, $[\Delta, \tilde K_j] \subset \Delta$ for the controlled drifts $\tilde K_0, \tilde K_j$.

These necessary and sufficient conditions (Theorem 2) can alternatively be stated as: $$[\Delta, K_0] \subset \Delta + G, \quad [\Delta, K_i] \subset \Delta + G, \quad K_I \in \Delta \subset \ker(dy)$$ where $G = \operatorname{span}\{K_i\}$ and $K_i$ are the control vector fields.

3. Geometric Control Approach and Ancillary Quantum Controller

The geometric control-theoretic analysis starts by examining the open-loop scenario (no active controller), introducing the Lie-algebraic span: $$\widetilde{\mathcal{C}} = \operatorname{span}\left\{ ( \mathrm{ad}_{H_i} )^j \{ ( \mathrm{ad}_{H_0} + \frac{\partial}{\partial t} )^k C \} \right\}_{i=1,\ldots, r;\, j, k \ge 0}.$$ Open-loop invariance (Theorem 1) holds only if $[\widetilde{\mathcal{C}}, H_{SE}] = 0$, which for most physically relevant configurations fails.

When open-loop or simple feedback fails, active controllers utilizing state-dependent feedback ($u = \alpha(\xi) + \beta(\xi) v$) are considered, but analytic control-theoretic conditions on $\alpha$ and $\beta$ are unsatisfiable unless control vector fields act on $\mathcal{H}_e$ in nontrivial fashion.

The solution provided by the QIMP is to introduce a single ancillary (“internal model”) qubit $b$ with its own Hamiltonian $H_b$, environment coupling $H_{bE}$, and $\sigma_{z}$-type couplings $H_{Sb}$ to the principal system. These, combined with rapid commutator sequences, synthesize a larger algebra of effective control vector fields (e.g., $\sigma_{x, y}^{(j)} \otimes (b + b^\dagger)^k$) that explicitly act on both system and environment. Only this extended controller algebra supports the invariance and integrability conditions required by the QIMP.

4. Two-Qubit Decoherence Suppression: Construction and Analysis

For the two-qubit system coupled collectively to a bosonic bath: $$H_0 = \sum_{j=1}^2 \frac{\omega_0}{2}\,\sigma_z^{(j)}, \quad H_e = \sum_k \omega_k b_k^\dagger b_k,$$

$$H_{SE} = \sum_k \left( \sigma_z^{(1)} + \sigma_z^{(2)} \right) (g_k b_k^\dagger + g_k^* b_k)$$

with control on each qubit through $$H_1 = \sigma_x^{(1)}, H_2 = \sigma_y^{(1)}, H_3 = \sigma_x^{(2)}, H_4 = \sigma_y^{(2)}$$, and observable $C = |01\rangle\langle 10|$, open-loop methods fail: $[\widetilde{\mathcal{C}}, H_{SE}] \neq 0$ and no suitable $\Delta$ exists.

The ancillary qubit $b$ is introduced: $$H_b = \frac{\omega_0}{2}\sigma_z^{(b)}, \quad H_{bE} = \sum_k \sigma_z^{(b)} (w_k b_k^\dagger + w_k^* b_k),$$ with Ising couplings $H_7 = J_1 \sigma_z^{(1)}\sigma_z^{(b)}$, $H_8 = J_2 \sigma_z^{(2)}\sigma_z^{(b)}$ and control fields $H_5 = \sigma_x^{(b)}, H_6 = \sigma_y^{(b)}$. Effective Hamiltonians of the form $\sigma_x^{(j)} (b + b^\dagger)^k$ for $j = 1,2$ and $k = 0,1,2$ are constructed via high-order commutators.

The total restructured system then has a 24-parameter control family: $$\dot{\xi} = (H_0 + H_e + H_{SE})\xi + \sum_{i=1}^8 \sum_{k=0}^2 u_{i, k}\left( \sigma_{x / y}^{(j)} \otimes (b + b^\dagger)^k \right)\xi,$$ whose control algebra fails to commute with $H_{SE}$. This allows an involutive $\Delta$ spanned by five vector fields to be constructed, providing both the inclusion $K_I \in \Delta$ and the required invariance. Numerical results confirm that $y(t)$ behaves as if $H_{SE} \equiv 0$ for any analytic controls.

5. Comparison to Classical Internal Model and Disturbance Decoupling

Classical disturbance decoupling depends on finding a subspace $\Delta$ incorporating the disturbance vector $E_d$ and respecting algebraic closure under system evolution, without explicit use of the disturbance generator. In contrast, the classical Internal Model Principle (IMP) for robust output regulation entails embedding the exosystem's dynamics ($\dot d = A_d d$) directly in the controller, ensuring perfect rejection of initial unknown disturbance components.

The QIMP merges these concepts: perfect quantum disturbance decoupling requires that the controller be capable of generating control vector fields acting on the environment in precisely the “modes” as the disturbance Hamiltonian $H_{SE}$. Thus, unlike classical disturbance rejection, explicit modeling and embedding of the disturbance are necessary. Disturbance rejection and tracking coincide in the quantum regime, resulting in a hybridization of these two classical principles.

6. Theorems, Scalability, and Outlook

The main theorems are:

Theorem 1 (Open-Loop Invariance):

$$y = \langle C \rangle \text{ is invariant under } H_{SE} \iff [\widetilde{\mathcal{C}}, H_{SE}] = 0.$$

Theorem 2 (Active Decoupling Criterion):

There exist analytic feedback laws $(\alpha, \beta)$ that render $y$ invariant if and only if there is an involutive $\Delta$ with $K_I \in \Delta \subset \ker(dy)$ and $[\Delta, K_i] \subset \Delta + G$.

Corollary (Scalability):

For any finite $N$-qubit register, exact protection of coherence can be attained by the addition of a single ancillary qubit whose controller algebra incorporates the complete environmental interaction model.

The QIMP framework is potentially extensible to higher-dimensional systems, continuous-mode environments, and integration with quantum error correction. The central insight is the necessity of an internal environmental interaction model within the controller for perfect decoherence suppression—a point where disturbance rejection and reference tracking fundamentally merge within quantum dynamics (Ganesan et al., 2010).