Quantum Noise, Open Systems & Control
An independent twelve-unit research-level course on open quantum dynamics, coherent errors and leakage, microscopic system-bath models, non-Markovian processes, Ramsey and noise spectroscopy, dynamical decoupling, shaped and optimized pulses, randomized benchmarking, readout calibration, feedback and engineered reservoirs. Quantum error correction has moved into its own full specialist course.
Published syllabus
Course Units
Open-system reduction, dynamical maps and operational CPTP tests
Derive a physical noise channel from joint unitary system-environment evolution and a reduced partial trace. A fixed product environment state yields a linear completely positive trace-preserving map. Initial correlations break the premise of a single universal reduced channel across arbitrary hypothetical preparations, so record the preparation intervention before fitting a process matrix. The distinction between a phenomenological fit and a microscopic model is central to quantitative control.
Coherent overrotations, stochastic noise and leakage
Calibration errors can be coherent rotations rather than stochastic jumps. Pauli twirling removes off-diagonal transfer-matrix components and often yields an operationally convenient surrogate, but it cannot prove that real errors accumulate stochastically. A small coherent angle may have quadratic average infidelity and linear worst-case diamond distance, producing different failure trends across repeated gates.
Microscopic bath coupling, Redfield models and Lindblad generators
Microscopic reduced dynamics begins with a Hamiltonian containing system, bath and interaction terms. Weak coupling, decay of bath correlations, rotating-wave or secular approximations and time-scale separation support Markovian master-equation approximations. Different approximation orders lead to Redfield or Gorini–Kossakowski–Sudarshan–Lindblad generators, which can predict different behavior near spectral degeneracies.
Non-Markovian memory, process tensors and CP divisibility
A process with correlated environment memory cannot in general be characterized by fitting one CPTP channel at each time and extrapolating it to arbitrary control sequences. The operational process tensor represents multi-time conditional probabilities given specified interventions. It separates a bath's memory from apparent drift created by measurement instability or by uncontrolled preparation correlations.
Noise power spectra, Ramsey dephasing and coherence spectroscopy
Dephasing spectroscopy relates a fluctuating qubit frequency to measured phase variance. For Gaussian stationary noise, a filter function describes how a pulse sequence weights the environment's spectral density. For non-Gaussian telegraph fluctuations or drifting calibration parameters, a second-order power spectrum can be insufficient to predict coherence.
Dynamical decoupling, finite pulses and filter-function optimization
Dynamical decoupling suppresses selected environmental couplings by driving the toggling-frame interaction through sign changes. Echo, Carr–Purcell and Uhrig-style timings have different filter structures and assumptions. Ideal delta-function pi pulses eliminate certain low-frequency terms, but finite-width pulses suffer control errors, leakage and pulse-induced heating.
Rotating-frame Hamiltonians, pulse envelopes and Bloch–Siegert corrections
A microwave control pulse couples to a qubit through an oscillating drive whose amplitude, frequency and phase determine a desired rotation. A rotating frame exposes the detuning and transverse drive; the rotating-wave approximation neglects counterrotating motion only when drive strengths and detunings remain small compared with the carrier frequency. Outside that hierarchy the Bloch–Siegert shift and leakage can create coherent calibration errors.
Composite pulses, robust optimal control and GRAPE gradients
Composite rotations exploit noncommuting axes so that selected calibration errors cancel across a designed sequence. GRAPE and other optimal-control methods instead optimize a discretized pulse waveform against a target unitary or channel, often through forward and backward propagations. Control constraints are part of the physics: amplitude bounds, minimum rise time, filtering, drift and hardware sampling all limit feasible solutions.
Randomized benchmarking, gate-set tomography and coherent-error detection
Randomized benchmarking estimates average error parameters from decay curves under assumptions about Markovian gate noise, the randomizing group and state-preparation and measurement behavior. The fitted parameter is not a direct image of every physical error channel. Gate-dependent noise, coherent drift and leakage can distort an exponential model and invalidate naive conversion to per-gate infidelity.
Measurement backaction, dispersive readout and calibration drift
Qubit readout is an interaction with a physical meter, not a preexisting bit copied without disturbance. An ideal POVM provides outcome probabilities, while an instrument specifies conditional state transformations. In dispersive superconducting readout, resonator frequency shifts encode qubit state into an emitted microwave field; integrating and classifying that field introduces assignment error, measurement-induced dephasing and possible state transitions.
Closed-loop stabilization, Kalman filtering and quantum trajectory control
Real-time quantum control uses a measurement record to adapt drive frequency, pulse phase or dissipative feedback. A classical filter may estimate hidden detuning from noisy Ramsey shots, while a quantum filter propagates a conditional state under measurement backaction. The two are related only when the measurement and hidden-state assumptions are carefully matched.
Decoherence-free subspaces, reservoir engineering and control research
A decoherence-free subspace exploits symmetry of environment coupling: selected code states experience the same noise operator action, so their relative coherence is preserved under that interaction. This is different from generic error correction because no repeated syndrome measurement is required in the ideal symmetry limit. Real devices have imperfect collective coupling and symmetry-breaking residual noise.