arXiv:2610.40072
Dissipative controller as a hidden source of errors in quantum gates
Serhat C. Kadıoğlu, Saulo V. Moreira, Mark T. Mitchison, Anthony Kiely
Abstract
Quantum systems are typically controlled by classical devices that necessarily dissipate energy to the surrounding environment. By the fluctuation-dissipation theorem, this implies noise in the control field produced. We investigate the relationship between the resulting errors and the entropy produced by the macroscopic controller, working in the context of a concrete minimal model. In the most experimentally relevant regime of fast control and high fidelity, we find a tradeoff between dissipation and control errors. These findings are exemplified by perturbative calculations of the gate fidelity for simple gate implementations and of the fidelity of state transfer via shortcuts to adiabaticity in the Landau-Zener model. Our results quantitatively demonstrate that suppressing errors in quantum control typically comes at the cost of macroscopic dissipation, which is needed to tame fluctuations in the classical controller.
AI Summary
The paper asks how dissipation in a macroscopic classical controller relates to errors in the quantum operation it drives. It models a magnetic-field controller as a resistive–inductive circuit with Johnson–Nyquist noise. The resulting control fluctuations have correlation $C(t,s)=(\gamma/\tau_c)e^{-|t-s|/\tau_c}$, while the controller’s entropy production is $\Sigma/k_B=\gamma^{-1}\int_0^\tau g(t)^2dt$, neglecting a positive transient contribution. Errors are calculated through second-order, noise-averaged state and gate fidelities, expressed as integrals of the noise correlation against ideal-evolution operator correlations. For commuting control dynamics, the susceptibility is $\chi=2\gamma K[\tau+\tau_c(e^{-\tau/\tau_c}-1)]$, independent of pulse shape. Combining this with the pulse-area cost bound gives $\chi\Sigma/k_B\geq K\theta^2F(\tau/\tau_c)$, where $\theta=\int g\,dt$ and $F(x)=2[1+(e^{-x}-1)/x]$. Shorter operations therefore reduce error while increasing dissipation. Examples cover single-qubit gates and an effective CNOT generator. For counterdiabatic Landau–Zener state transfer, analytical propagators and numerical evaluation with independent noisy control channels show the same tradeoff at short durations. Fitted error–cost exponents range approximately from $-1.92$ to $-1.15$. At long durations, both error and dissipation increase. A general state-fidelity product bound is also derived, but depends on the protocol and initial state and can vanish. The analysis assumes perfectly prescribed mean voltage, stationary controller noise, and perturbatively small fluctuations, using $\tau^2\gamma/\tau_c\ll1$ as its validity condition. It excludes other noise sources; realistic multi-qubit dissipation prefactors require hardware-specific controller models.
Error found: The gate-fidelity kernel omits its trace-subtraction term, making the stated CNOT susceptibility incorrect: its leading infidelity is 25% below the reported value and does not equal that of the listed single-qubit gates at fixed duration. The abstract's characterization of fast control as the most experimentally relevant regime is not supported by a quantitative connection between the controller model and experimental operating parameters.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- Comments are welcome