Quantum Control and Feedback

Quantum control asks how deliberately chosen drives, pulses, measurements, couplings, and reservoirs can steer a quantum system toward a desired behavior. In open systems, this is not only a unitary-control problem. Decoherence, relaxation, measurement backaction, finite bandwidth, leakage, calibration drift, and environmental memory are part of the control design.

This chapter is the control map for the open-system volume. It connects coherent drives, pulse-based noise filtering, optimal-control ideas, measurement-based feedback, coherent feedback, reservoir engineering, dissipative state preparation, Zeno constraints, and practical limits from noise.

Core Equation

A common Markovian model for controlled open dynamics is

\dot\rho = - \frac{i}{\hbar} \left[ H_0+\sum_j u_j(t)H_j,\rho \right] + \sum_\alpha \mathcal D[L_\alpha(t)]\rho,

with

\mathcal D[L]\rho = L\rho L^\dagger - \frac12 \{L^\dagger L,\rho\}.

Here $u_j(t)$ are externally chosen control waveforms, $H_j$ are control Hamiltonians, and the jump operators $L_\alpha(t)$ may represent natural noise, measurement channels, or engineered dissipation. This equation is a useful organizing form, not a universal law. Strong coupling, finite delay, non-Markovian environments, leakage outside the model space, and measurement conditioning can require a different description.

The basic design question is:

given a model, a target, and constraints, choose allowed interventions
so that the target is reached with verified robustness

The constraints are as important as the target. A pulse that works only with infinite amplitude, zero timing error, no bandwidth limit, and no dissipation is not a physical control solution.

Control Strategies

Strategy	Typical control object	What it is good for
Coherent open-loop control	predetermined $u_j(t)$ in the Hamiltonian	rotations, gates, state transfer, spectroscopy
Dynamical decoupling	sign changes or frame changes by pulses	suppressing slow dephasing and shaping noise filters
Optimal control	numerically optimized waveforms	constrained high-fidelity operations and robustness tradeoffs
Measurement-based feedback	controls depending on a measurement record	stabilization, cooling, tracking, conditional correction
Coherent feedback	routed quantum fields or auxiliary systems	feedback without an intermediate classical record
Reservoir engineering	designed dissipators or lossy auxiliaries	autonomous stabilization and dissipative state preparation
Zeno control	frequent measurement or strong dissipation	confinement to subspaces and inhibited transitions

These strategies overlap in real experiments. A cavity can be a measurement port, an engineered reservoir, a coherent feedback element, or an unwanted loss channel depending on what is retained, monitored, and traced out.

Targets

Control targets should be stated operationally. Common targets include:

preparing a state or subspace;
implementing a unitary gate or quantum channel;
maximizing a transition probability;
cooling a mode or suppressing heating;
stabilizing a nonequilibrium steady state;
tracking a continuously measured state estimate;
rejecting noise in a selected frequency band;
estimating a signal with a chosen sensing protocol;
making a target dark state or unique attractor of an engineered Liouvillian.

For closed, finite-dimensional systems, controllability can sometimes be analyzed through the Lie algebra generated by the available Hamiltonians. In open systems, controllability is more delicate because dissipative channels can both help and hurt. A lossy auxiliary can remove entropy and prepare a pure state, while an uncontrolled relaxation channel can erase the same coherence a pulse sequence is trying to protect.

Open-Loop Control

Open-loop control means the waveform is chosen without using a contemporaneous measurement record. Examples include resonant Rabi driving, Ramsey interferometry, spin echo, shaped pulses, composite pulses, and Floquet modulation.

For a two-level system, the rotating-frame Hamiltonian often has the schematic form

H_{\mathrm{rot}} = \frac{\hbar}{2} \left( \Delta Z+\Omega_x(t)X+\Omega_y(t)Y \right),

where $\Delta$ is detuning and $\Omega_x,\Omega_y$ are controlled quadratures. This is the baseline model behind many pulse and calibration protocols. The closed-system starting point is Rabi Oscillations: First Encounter. With decay and dephasing included, the natural open-system entry is Optical Bloch Equations.

Open-loop control is powerful, but it cannot respond to information learned during the run unless the protocol is updated between shots. It also cannot refocus noise that has no memory on the relevant time scale.

Noise Filtering

Dynamical Decoupling is the canonical page for pulse-based noise filtering. In a simple dephasing model,

H_{\mathrm{noise}}(t) = \frac{\hbar}{2} y(t)\xi(t)Z,

where $y(t)$ is the toggling-frame modulation produced by pulses. The accumulated phase samples the noise through a filter,

Y(\omega,T) = \int_0^T dt\, y(t)e^{i\omega t}.

Changing the pulse sequence changes the frequency band in which the system is sensitive. This is why Ramsey, echo, CPMG, and more elaborate sequences can produce different coherence times in the same physical device.

The limit is just as important: if the damaging noise is effectively white over the relevant band, ideal sign flips do not remove its total effect. If the dominant error is energy relaxation through a Markovian jump operator, dephasing refocusing pulses may not help.

Feedback

Measurement-based feedback uses an observed record to choose later controls. In continuous measurement, the record is noisy and the conditioned state obeys a stochastic equation. A schematic feedback law is

u_j(t) = f_j\!\left[I(s):0\le s\le t\right],

where $I(s)$ is the measurement record up to time $t$ . This makes the dynamics conditional, record-dependent, and sensitive to delays, estimator bandwidth, detector inefficiency, and backaction. The local control-design entry point is Measurement-Based Feedback, while Feedback from Measurement Records and Stochastic Master Equations provide the record and state-update language.

Coherent Feedback routes quantum fields or auxiliary systems without first converting the signal into a classical record. It is natural in quantum optics, circuit QED, and network models, where Input–Output Theory supplies the port language. The conceptual boundary is operational: ask whether the field is measured, coherently routed, traced out, or retained as part of an enlarged system.

Engineered Dissipation

Reservoir Engineering turns dissipation into a design resource. Instead of fighting every environmental channel, one chooses lossy auxiliaries, drives, or bath spectra so that useful states become steady, attractive, or protected. For the focused preparation workflow, use Dissipative State Preparation.

In Liouvillian language, a target state $\rho_\star$ must satisfy

\mathcal L(\rho_\star)=0.

For preparation, stationarity is not enough. The target must be attractive on a useful time scale and robust against unwanted channels. A dark state in a large dark subspace is protected, but not uniquely prepared. For the fixed-point and relaxation-mode language behind this check, see Steady States and Relaxation.

Planned Reading Path

The complete chapter is organized around these article-level topics:

Topic	Core question
Driven Open Systems	How do controlled Hamiltonians coexist with dissipators?
Rabi and Ramsey Control	What do familiar two-level protocols teach in the presence of decoherence?
Dynamical Decoupling	How do pulses filter noise spectra?
Pulse Sequences	What are the standard pulse patterns and convention pitfalls?
Optimal Control	How are pulses designed by optimizing an objective under constraints?
Measurement-Based Feedback	How does a noisy record become a control signal?
Coherent Feedback	What changes when the feedback signal remains quantum?
Reservoir Engineering	How can dissipation prepare or stabilize useful states?
Dissipative State Preparation	When is a target state a unique attractive steady state?
Quantum Zeno Dynamics	How do frequent measurements or strong dissipation constrain motion?
Control Limits from Noise	What prevents control from being arbitrarily accurate?

For optimized pulse design, Optimal Control gives the conceptual guide, while Optimal Control Toy Problems gives a reproducible computational entry point.

What This Chapter Owns

This chapter owns the open-system control viewpoint: how drives, pulses, records, feedback laws, and engineered reservoirs modify reduced dynamics. It does not own:

the basic density-operator and channel formalism, which belongs to Core Formalism and Quantum Channels;
the closed-system theory of time evolution and its control bridge, which belong to Dynamics and Formulations;
the derivation of Lindblad generators, which belongs to Markovian Master Equations;
the full theory of noise spectra, which belongs to Quantum Noise, Dissipation, and Baths;
hardware-specific gate calibration, which belongs to platform or quantum-information volumes;
general classical control engineering, which should be cited rather than recreated.

When a control page needs a master equation, link to the canonical generator page. When it needs a spectrum, link to the canonical noise page. When it needs a platform-specific implementation, link outward to the relevant AMO, quantum matter, or quantum information chapter.

Common Mistakes

Treating controllability in an ideal closed model as evidence of high-fidelity control in an open experiment.
Reporting a pulse sequence without specifying the rotating frame, phase convention, and detuning convention.
Comparing Ramsey, echo, and dynamically decoupled coherence times as if they measured the same noise band.
Using feedback formulas without accounting for measurement inefficiency, delay, and estimator noise.
Calling engineered dissipation “error correction” when no protected logical information or recovery condition has been specified.
Assuming stronger drive always improves control; it can increase leakage, heating, off-resonant excitation, and calibration sensitivity.
Forgetting that a dark state need not be the unique attractor.
Optimizing a waveform against a model that omits the dominant experimental constraint.

Exercises

Open-loop or feedback?

A qubit is driven by a pulse $u(t)$ chosen before the experiment begins. In a second protocol, the drive amplitude at time $t$ depends on the homodyne current measured during the same run. Which protocol is open-loop, and which is feedback?

Solution

The predetermined pulse is open-loop control: $u(t)$ is fixed before the run and does not depend on the observed record. The homodyne-dependent drive is measurement-based feedback because the control law uses information obtained during the run. The second protocol must model measurement backaction, detector noise, latency, and the conditioned state estimate.

Dark does not mean prepared

Suppose a dissipator has two orthogonal dark states $\lvert d_1\rangle$ and $\lvert d_2\rangle$ . Why is $\lvert d_1\rangle$ not automatically prepared from arbitrary initial states?

Solution

If both states are dark, the dissipator does not remove population from either one. The steady-state space is at least two-dimensional, so the final state can depend on the initial overlap with the dark subspace. To prepare $\lvert d_1\rangle$ uniquely, additional Hamiltonian or dissipative terms must drain $\lvert d_2\rangle$ or otherwise make $\lvert d_1\rangle$ the unique attractive steady state.

Pulse filtering and white noise

Why can spin echo refocus quasi-static dephasing but fail to improve coherence under ideal white dephasing noise?

Solution

Spin echo reverses the sign of the accumulated phase, so noise that is nearly constant during the sequence cancels to leading order. White noise has no such memory: each time interval contributes independent fluctuations. In the filter-function language, a flat spectrum makes the total dephasing depend on the integrated exposure $\int_0^Tdt\,|y(t)|^2$ , which is unchanged by ideal sign flips with $|y(t)|=1$ .

Cross-Links

References

D. D’Alessandro, Introduction to Quantum Control and Dynamics, Chapman and Hall/CRC (2007).
H. M. Wiseman and G. J. Milburn, Quantum Measurement and Control, Cambridge University Press (2010).
K. Jacobs, Quantum Measurement Theory and its Applications, Cambridge University Press (2014).
C. Brif, R. Chakrabarti, and H. Rabitz, “Control of quantum phenomena: past, present and future,” New Journal of Physics 12, 075008 (2010).
D. Dong and I. R. Petersen, “Quantum control theory and applications: a survey,” IET Control Theory & Applications 4, 2651–2671 (2010).
N. Khaneja, T. Reiss, C. Kehlet, T. Schulte-Herbrüggen, and S. J. Glaser, “Optimal control of coupled spin dynamics: design of NMR pulse sequences by gradient ascent algorithms,” Journal of Magnetic Resonance 172, 296–305 (2005).
L. Viola and S. Lloyd, “Dynamical suppression of decoherence in two-state quantum systems,” Physical Review A 58, 2733–2744 (1998).
J. F. Poyatos, J. I. Cirac, and P. Zoller, “Quantum reservoir engineering with laser cooled trapped ions,” Physical Review Letters 77, 4728–4731 (1996).