| quantum tomography | |
|---|---|
| Name | Quantum tomography |
| Field | Quantum information science |
| Invented by | Mauro D'Ariano? |
| Year | 1990s |
| Related | Quantum state estimation, Quantum process tomography |
quantum tomography
Quantum tomography is a collection of methods for reconstructing the quantum state, process, or measurement operators of a quantum system from experimental data. It provides operationally meaningful representations (for example, a density matrix or a quantum channel) that are essential for validation, benchmarking, and development in Quantum information science and experimental physics. Quantum tomography matters because it links abstract mathematical descriptions like density matrixs and quantum channels to measurable statistics and enables certification of devices such as quantum computers and quantum sensors.
Quantum tomography encompasses procedures to estimate quantum states (state tomography), quantum dynamics (process tomography or quantum process tomography), and measurement devices (detector tomography). Typical targets include finite-dimensional systems such as qubits in superconducting qubit circuits (e.g., work at IBM, Google and Rigetti Computing) and trapped-ion platforms (ion traps developed at institutions like IQOQI and NIST), as well as continuous-variable systems in quantum optics using homodyne detection pioneered in experiments by groups like those of Serge Haroche and Raymond Laflamme. Tomography is applied in laboratory verification, quantum error characterization, and compliance with theoretical models such as those used in quantum state estimation benchmarks and quantum certification.
The mathematical basis of quantum tomography rests on statistical estimation theory and linear algebra. The quantum state of a finite system is represented by a density matrix ρ, a positive semidefinite, unit-trace operator on a Hilbert space. Measurement outcomes are modeled by Positive operator-valued measures (POVMs). Tomographic reconstruction solves an inverse problem: given measurement statistics for a set of known POVMs, infer ρ or a quantum map represented by a completely positive trace-preserving (CPTP) superoperator. Fundamental results draw on the Born rule, spectral theorem, and concepts from estimation theory such as unbiasedness and the Cramér–Rao bound. Connections to convex optimization and maximum-likelihood estimation formulate physically constrained estimators, while information-theoretic criteria like the Fisher information and quantum Fisher information quantify precision limits.
Common algorithms include linear inversion, maximum-likelihood (MLE), Bayesian estimation, and convex optimization approaches such as semidefinite programming. Linear inversion provides direct reconstruction but can produce nonphysical (non-positive) density matrices; MLE enforces physicality by optimizing the likelihood under positivity constraints. Bayesian tomography incorporates priors and yields credible regions via techniques like Markov chain Monte Carlo (MCMC). Recent advances integrate compressed sensing methods leveraging low-rank matrix recovery for sparse states and use randomized measurement schemes inspired by unitary t-designs and randomized benchmarking protocols. Algorithmic implementations rely on numerical libraries and solvers used in quantum computing research at institutions like MIT, Caltech, and industrial labs.
Experimental tomography has been realized across platforms: quantum optics experiments use homodyne tomography and balanced homodyne detectors to reconstruct continuous-variable states and Wigner functions; trapped ion systems perform state and process tomography via laser-driven gates and fluorescence detection (groups at University of Innsbruck and NIST); superconducting qubit platforms use microwave control and dispersive readout (research at Yale University and Google Quantum AI). Detector tomography characterizes single-photon detectors and superconducting nanowire single-photon detectors used in telecom experiments. Implementation details include calibration of POVMs, selection of informationally complete measurement sets (e.g., SIC-POVMs or mutually unbiased bases), and mitigation of state-preparation and measurement (SPAM) errors using techniques like gate set tomography developed by researchers at Sandia National Laboratories and University of Colorado Boulder.
Quantum tomography underpins tasks in quantum computation and quantum communication: benchmarking quantum gates, verifying entanglement (via reconstructed density matrices and entanglement measures such as negativity), characterizing quantum channels for quantum error correction research, and certifying quantum random number generators. Tomographic reconstructions provide data for testing foundational phenomena—Bell inequality violations and decoherence models—and are used in quantum metrology to characterize probe states for enhanced sensitivity. Industry applications include device certification for nascent quantum processors by companies like Xanadu and academic–industrial collaborations.
Scalability is a primary limitation: the number of parameters grows exponentially with system size (e.g., 4^n − 1 real parameters for n qubits), making full tomography infeasible for large registers. Statistical noise, SPAM errors, and model mismatch complicate inference; regularization and physical constraints mitigate but do not eliminate these issues. Error analysis uses bootstrap methods, Bayesian credible intervals, and asymptotic bounds such as the Cramér–Rao limit. Practical workarounds include partial tomography, tomography of reduced density matrices, and cross-entropy benchmarking; theoretical work on complexity establishes hardness results connecting tomography to problems in computational complexity theory.
Extensions adapt tomography to realistic resource limits. Compressed sensing tomography assumes low-rank states and uses convex optimization to reconstruct with far fewer measurements, demonstrated in nuclear magnetic resonance (NMR) and photonic systems. Bayesian tomography yields full posterior distributions, enabling principled uncertainty quantification and adaptive experiment design; implementations use MCMC and variational Bayesian methods. Continuous-variable tomography reconstructs Wigner functions and Gaussian states via homodyne detection and techniques like inverse Radon transforms, important in quantum optics and continuous-variable quantum information protocols. Hybrid methods combine ideas from machine learning (neural-network quantum states), compressed sensing, and Bayesian inference to scale tomography for medium-size quantum devices.