Quantum State Tomography
Reconstructing a state from measurement statistics.
A density matrix is not an observable: no single measurement reveals it. Tomography reconstructs from the statistics of many measurements on many copies. How many copies it takes is one of the sharpest quantitative results in quantum information, with an exponential price for many qubits and some remarkable ways around it.
What you’ll learn
- Define informational completeness and reconstruct a qubit from Pauli or SIC-POVM statistics by linear inversion.
- Quantify shot noise and state the optimal sample complexities: copies with collective and with single-copy measurements.
- Turn raw estimates into physical states (projection, maximum likelihood) and recognize their pitfalls.
- Use classical shadows to predict many properties from few randomized measurements, with explicit sample bounds.
Informationally complete measurements
A density matrix on has real parameters. A measurement yields only the numbers , so it can pin down only if these numbers determine it uniquely.
For qubits (), the Pauli strings form an orthogonal operator basis with , so every state is fixed by its Pauli expectation values:
For one qubit this is : three bases suffice. Reconstruct a mystery state yourself.
Think first Write down your estimate . What should it be close to, and how close?
The prepared state has , , so . Each component should typically land within about one standard error () and almost always within . Because this state is pure, will often come out slightly larger than 1: an unphysical estimate, a problem fixed in step 3.
How many copies?
Each run yields one outcome, so every estimate carries shot noise. Start with a single Pauli observable , estimated from runs:
The outcomes are i.i.d. with mean . Use their sample mean:
Since , each outcome has variance :
Hoeffding's inequality for variables in gives an exponential tail:
A union bound over observables: precision costs , while the number of observables costs only . With this many shots per observable, all estimates are within with probability at least :
You want the standard error of your estimate of to be . How many shots do you need?
Show the answer
10000 shots
The standard error is at , so . Halving the error costs four times as many shots.
Full tomography must estimate all parameters at once, and the measurement model changes the answer qualitatively:
| Measurements allowed | Copies for trace-distance error , full rank | What it takes |
|---|---|---|
| Collective | a quantum memory holding all copies; Schur–Weyl-based measurements | |
| Single-copy, adaptive or not | one copy at a time, e.g. random-basis measurements | |
| Local Pauli bases on qubits | exponential in | settings just to be informationally complete |
From numbers to a physical state
Linear inversion is unbiased, but with finite data it can return an operator with negative eigenvalues: for a nearly pure qubit, pokes outside the Bloch ball about half the time. Two standard repairs:
Linear inversion on a qubit returns . Which state does the Frobenius-norm projection return?
- itself: linear inversion is unbiased, so no correction is needed
- , obtained by shrinking the largest component
- , the radial projection onto the sphere
- , because the estimate is unphysical
Show the answer
C. , the radial projection onto the sphere
For qubits , so the closest physical state is the closest point of the ball: the radial projection, since . In eigenvalue language, becomes : the estimate is projected to a pure state.
Classical shadows: predicting without reconstructing
Often you do not need at all, only numbers : energies, fidelities, correlators. Huang, Kueng & Preskill (2020) showed that a few randomized measurements suffice, at a cost logarithmic in .
In each round, measure every qubit in a uniformly random basis , or ; equivalently, apply a random single-qubit Clifford and measure in the computational basis. Record the bases and the outcome bits .
Averaged over bases and outcomes, one qubit's snapshot is a depolarized copy of :
Invert this channel on the classical record, qubit by qubit. The resulting snapshot is an unbiased estimator, , even though itself is not a state:
Estimate each by a median of means of the numbers , computed classically from the stored snapshots.
Two features stand out. grows only logarithmically with the number of observables, because every snapshot is reused for all of them; and for local observables it does not depend on the number of qubits at all. The Clifford version makes fidelities with pure targets cheap, since for . Huang, Kueng and Preskill also proved that these scalings are close to optimal.
Key takeaways
- Tomography needs an informationally complete measurement, whose effects span all Hermitian directions; for qubits, Pauli settings suffice.
- Each expectation value costs about shots. Full-rank tomography costs copies with collective and with single-copy measurements: exponential in either way.
- Linear inversion can be unphysical; projection or maximum likelihood repairs it, and honest error bars need care and a trusted measurement model.
- Classical shadows predict local properties from randomized snapshots, independently of system size.
Check yourself
In the interactive lesson these are scored with your confidence, and what you miss comes back for review.
How many measurement settings does standard Pauli tomography of qubits use, and why not ?
- , one per Pauli string
- : each setting fixes , or on every qubit, and strings containing are read off by ignoring bits
- , one per computational basis state
- , one per free parameter
Show the answer
B. : each setting fixes , or on every qubit, and strings containing are read off by ignoring bits
Measuring every qubit in a chosen Pauli basis gives the joint distribution of all outcomes. Marginalizing gives every string obtained by replacing some letters with , so settings cover all strings.
Why must an informationally complete POVM on have at least outcomes?
- Its effects must span the -dimensional real vector space of Hermitian operators
- Each outcome reveals one amplitude, and a -dimensional state has amplitudes
- Each outcome needs its own copy of the state, and tomography needs copies
- It need not: a projective measurement in a well-chosen basis, with outcomes, suffices
Show the answer
A. Its effects must span the -dimensional real vector space of Hermitian operators
If the span of the missed a Hermitian direction , which is automatically traceless since is in the span, then and would give identical statistics. Spanning a -dimensional space takes at least vectors.
Using random-Pauli classical shadows, you want to estimate different 2-local observables, each with , to accuracy . How does the required number of snapshots scale?
- , one batch of shots per observable
- , because the state must effectively be reconstructed
- , independent of the number of qubits
Show the answer
C. , independent of the number of qubits
The shadow norm of a -local observable is at most for , and median-of-means gives , with no dependence on . Every snapshot is reused for all observables.
Which statement about tomography of a full-rank -dimensional state to trace-distance error is correct?
- Collective and single-copy measurements both need copies
- Collective measurements need copies and single-copy ones
- Adaptive single-copy measurements match collective ones at
- Collective measurements need copies; single-copy ones need , even when chosen adaptively
Show the answer
D. Collective measurements need copies; single-copy ones need , even when chosen adaptively
Measurements entangled across copies (built on Schur–Weyl duality) reach . Without a quantum memory each copy is measured on its own, the optimal rate is , and adaptivity does not close the gap.
Further reading
- Z. Hradil, "Quantum-state estimation", Physical Review A 55, R1561 (1997).
- D. Gross, Y.-K. Liu, S. T. Flammia, S. Becker & J. Eisert, "Quantum state tomography via compressed sensing", Physical Review Letters 105, 150401 (2010).
- J. Haah, A. W. Harrow, Z. Ji, X. Wu & N. Yu, "Sample-optimal tomography of quantum states", IEEE Transactions on Information Theory (2017); R. O'Donnell & J. Wright, "Efficient quantum tomography", STOC (2016).
- H.-Y. Huang, R. Kueng & J. Preskill, "Predicting many properties of a quantum system from very few measurements", Nature Physics 16, 1050 (2020).
- J. M. Renes, R. Blume-Kohout, A. J. Scott & C. M. Caves, "Symmetric informationally complete quantum measurements", Journal of Mathematical Physics 45, 2171 (2004).