Quantum state tomography

Quantum state tomography (QST) is the original quantum characterization protocol: reconstruct the complete density matrix of a prepared state from measurement statistics. It is a characterization tool rather than a benchmark (it returns a matrix, not a score), but most reported state fidelities trace back to it, and it is the readout subroutine inside quantum process tomography. Attribution is conventional rather than crisp: inferring a state from measurements goes back to the Pauli problem (1933) and Fano (1957), the modern origin is Vogel & Risken’s 1989 homodyne-reconstruction proposal, the word “tomography” arrived with Smithey et al.’s 1993 experiment, and James, Kwiat, Munro & White (2001) fixed the standard qubit recipe.

How it works

Prepare the same state many times and measure the copies in an informationally complete set of bases: for n qubits, typically all 3^n combinations of Pauli measurement settings; in optics, rotated field quadratures via homodyne detection. From the outcome frequencies, estimate the density matrix by linear inversion or, more robustly, by maximum-likelihood or Bayesian estimation that enforces physicality. An n-qubit density matrix has 4^n − 1 real parameters, so measurement and post-processing costs grow exponentially with n.

Strengths and limitations

QST is maximally informative (the reconstructed density matrix predicts every observable), and tooling is ubiquitous (Qiskit Experiments ships a StateTomography experiment). The costs are equally well known. Exponential scaling confines full QST to a handful of qubits; beyond that, classical shadows, matrix-product-state tomography, and direct fidelity estimation take over. Estimator choice matters: naive linear inversion can return unphysical (non-positive) matrices, while maximum-likelihood estimation is biased. And QST assumes calibrated measurements, so SPAM errors bias the reconstruction, the self-consistency gap that gate set tomography was invented to close. Nearly four decades on it remains in active use and development, with recent work on selective multi-qubit and continuous-variable variants.