Can measuring two quantum copies together beat measuring them one by one?
Measuring two copies of a noisy qubit jointly with an entangling circuit estimated two rotation angles more precisely than any one-at-a-time measurement can, on real quantum computers.
Source
Approaching optimal entangling collective measurements on quantum computing platforms
Study at a glance
- Design
- Other — Theory-designed measurement circuits run on several quantum processors (IBM, Rigetti, trapped-ion, photonic) to estimate two small qubit rotations under controlled decoherence
- N
- No participant sample; each unknown angle was estimated 400 times, each estimate averaging 512 circuit repetitions (341 for three-copy circuits)
- Population
- Qubits on the Fraunhofer IBM Q System One, 11 cloud IBM Q processors, Rigetti Aspen-9, the AQTION trapped-ion processor and the JenQuant photonic processor
- Outcome
- Mean squared error of simultaneous estimates of two rotation angles, compared with Nagaoka and Holevo bounds
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors considered a qubit rotated by small unknown angles about the x and y axes and then partly decohered, and derived optimal measurements that act on one, two or three copies at once. They compiled these into quantum circuits and ran them on superconducting processors (a dedicated IBM system, cloud IBM machines, Rigetti), a trapped-ion processor and, for single-copy measurements, a photonic processor. A simple calibration-based error mitigation removed constant bias, and the mean squared error of the angle estimates was compared with theoretical bounds.
What they found
On the dedicated IBM device, two-copy measurements gave errors 19 ± 4% below the theoretical single-copy limit when averaged over a range of angles, only 6 ± 4% above the ultimate Holevo bound, whereas single-copy measurements must be at least 33% above it. The maximum two-copy advantage across decoherence levels was 21 ± 4%. The trapped-ion and photonic devices reached single-copy limits without mitigation, but Rigetti reached none of the limits and three-copy circuits failed on all devices due to gate errors. Two-copy measurements also violated the Lu–Wang uncertainty relation by more than three standard deviations.
The limits
What it doesn't show
The advantage is shown for a specific toy estimation problem (small rotations of one qubit with a particular noise model), not a practical sensor. Three-copy measurements, which should get closer to the Holevo bound, performed worse in practice because deeper circuits accumulate more errors, so scaling to many copies is unproven. Error mitigation worked best where the team had unrestricted device access, and the photonic device could only do single-copy measurements.
Key terms
- Collective measurement
- A measurement that acts jointly on several copies of a quantum state, entangling them during readout, instead of measuring each copy separately.
- Holevo Cramér–Rao bound
- The ultimate precision limit for estimating several parameters from a quantum state, reachable only with collective measurements on infinitely many copies.
- Nagaoka bound
- The best precision achievable for two-parameter qubit estimation with measurements restricted to a given number of copies; here the single- and two-copy versions are the targets.
- Mean squared error
- The average of the squared differences between estimated and true parameter values, used as the figure of merit for precision.
- Error mitigation
- Post-processing that corrects noisy processor outputs using calibration data, here a constant offset fitted from known angles so the estimator stays unbiased.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What makes a measurement 'collective' in this paper's terminology?
Common questions
Why doesn't an entangled input state suffice here, as in single-parameter sensing?
When two parameters are estimated at once, the optimal observables may be incompatible, so entanglement is also needed at the measurement stage to approach the ultimate limit.
Does violating the Lu–Wang relation mean the uncertainty principle is wrong?
No. That relation assumes single-copy measurements; the violation shows it is not truly universal and that tighter relations are needed when collective measurements are allowed.
Why did three-copy measurements not help?
The theoretical gain over two copies is marginal for this problem, and the larger circuits suffer more gate and readout errors on current hardware.
More on Entanglement and quantum information