Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2609.39941

Fast and Sure-ious Quantum Process Tomography

A. Afham, Sayantan Sen, Marco Tomamichel

quant-phSubmitted Sep 30, 2026• Announced Oct 1, 2026• 0 citations• Revised Oct 2, 2026 (v2)
v1 · Sep 30, 2026v2 · Oct 2, 2026
arXivHTMLPDF

Abstract

Quantum process tomography (QPT) protocols are often designed by lifting quantum state tomography (QST) protocols through the Choi-Jamiolkowski isomorphism, at the cost of an extra step enforcing the trace-preserving (TP) constraint. Existing approaches fall into two classes. The folklore partial normalization, which sandwiches the density estimate between inverse square roots of its input marginal, has a closed form but is used as a heuristic, its only analyses being asymptotic and losing a factor polynomial in the dimension. Projection-based methods are principled but have no closed forms and scale poorly with dimension. We resolve this dichotomy by showing that the partial normalization is the exact projection onto the Choi states of quantum channels in fidelity (equivalently in Bures or purified distance). Consequently, any QST algorithm with guarantees in fidelity lifts to a QPT algorithm that inherits its sample complexity, through a single closed-form congruence that preserves the Choi rank and is computed in Kraus form without any eigendecomposition of Choi dimension. We lift two single-copy protocols and prove that the non-adaptive one (Fidelity Projected Least Squares) recovers the Choi rank and reaches infidelity $\varepsilon$ from a number of channel uses proportional to $1/\varepsilon$. Numerically, for channels of low Choi rank, it improves on the state-of-the-art projected least squares in accuracy, and by orders of magnitude in the time and memory of the TP-regularization step and of the returned estimate.

AI Summary

Ancilla-assisted quantum process tomography reduces channel estimation to state tomography of a Choi state, but the estimate must additionally satisfy the trace-preserving marginal constraint. This paper studies whether the usual partial normalization can enforce that constraint while retaining statistical guarantees and low-rank structure. Using a fixed-marginal fidelity-projection result established in a companion paper, it identifies $\mathrm{N}_A(\hat\rho)=d_A^{-1}(\hat\rho_A^{-1/2}\otimes I)\hat\rho(\hat\rho_A^{-1/2}\otimes I)$ as the unique closest channel Choi state in Bures or purified distance, assuming $\hat\rho_A>0$. Triangle inequality bounds the increase in Bures error by $2$, and in root-infidelity by $4$, yielding a generic lift of fidelity-guaranteed state-tomography protocols. The correction preserves rank and costs $O(rd_A^2d_B+d_A^3)$ operations in Kraus form. For non-adaptive local-Pauli tomography, Fidelity Projected Least Squares thresholds the least-squares spectrum at a matrix-Bernstein concentration radius. Concentration, rank recovery and a quadratic operator-error-to-infidelity bound give channel-use complexity $O(r^2D^{1.6}\log(D/\delta)/\varepsilon^2)$ generally, improving to $O(r\lambda_r^{-1}D^{1.6}\log(D/\delta)/\varepsilon)$ when $\varepsilon\le r\lambda_r/3$, where $D=d_Ad_B$ and $\lambda_r$ is the smallest nonzero Choi eigenvalue. Simulations through $D=1024$ compare against projected least squares, reporting low-rank infidelity scaling as $N^{-1}$ rather than $N^{-1/2}$. Limitations include spectrum-dependent onset, poorer pre-onset and nearly-full-rank performance, and dimension-dependent conversion to diamond error. Numerical experiments generally halve the proven threshold heuristically; reported computational savings concern TP regularization and output storage, not the full reconstruction pipeline.

Likely largely AI-generatedAI use disclosed
the LLM was asked to formalize this idea and provided the proofs of \cref{App:ProjectionStep} to \cref{App:QPTProof}.

Error found: The abstract's unqualified rank-preservation statement requires an invertible input marginal, while completing the correction for singular marginals can increase rank. The headline caption reports an empirical rank-recovery onset of 10^6 shots without specifying the threshold, whereas the accuracy section reports 3×10^5 for the halved threshold normally used in the numerics. Part (3) of the universal lifting theorem does not ensure that the QST estimate has an invertible input marginal, so its displayed correction can be undefined even on the promised accuracy event.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Comments
39 pages and 9 figures. v1 to v2 change: numerical experiments modified, including one extra plot

Related papers

Optimal learning of quantum channels in diamond distance

Antonio Anna Mele, Lennart Bittel

Quantum process tomography, the task of estimating an unknown quantum channel, is a central problem in quantum information theory. A long-standing open question is how many uses of an unknown channel are required to learn it in diamond distance, the standard metric for distinguishing quantum processes. While quantum state tomography is well understood, for general channels the problem remained open beyond the unitary case. Here we establish the query complexity of channel tomography with optimal dependence on the dimension parameters, at any fixed constant accuracy. We design an algorithm showing that any channel with input/output dimensions $d_{\mathrm{in}},d_{\mathrm{out}}$ and Kraus rank at most $k$ can be learned to accuracy $\varepsilon$ using $O(d_{\mathrm{in}}d_{\mathrm{out}}k/\varepsilon^{2})$ channel uses. Conversely, we prove that $Ω(d_{\mathrm{in}}d_{\mathrm{out}}k)$ uses are necessary at constant accuracy and that, for non-minimal Kraus rank, a separate $Ω(1/\varepsilon^{2})$ contribution is unavoidable. Since channels subsume states, unitaries, isometries, and measurements as special cases, our protocol provides a unified framework for these tomography tasks, yielding new guarantees for isometry and measurement tomography while recovering known optimal scalings for state and unitary tomography. Our algorithm follows the natural strategy of performing optimal tomography on the Choi state. The main technical contribution is to show that this suffices to control the induced diamond-distance error, avoiding the dimension loss incurred by a naive conversion from Choi-state trace distance to channel diamond distance. The protocol uses the channel non-adaptively to prepare Choi-state copies, purifies them in parallel, and performs optimal pure-state tomography on the resulting purifications. Hence, we reduce channel tomography to pure-state tomography.

quant-pharXiv:2512.10214•Nature Physics•Dec 12, 2025•36 citations

Projected Least-Squares Quantum Process Tomography

Trystan Surawy-Stepney, Jonas Kahn, Richard Kueng +1 more

We propose and investigate a new method of quantum process tomography (QPT) which we call projected least squares (PLS). In short, PLS consists of first computing the least-squares estimator of the Choi matrix of an unknown channel, and subsequently projecting it onto the convex set of Choi matrices. We consider four experimental setups including direct QPT with Pauli eigenvectors as input and Pauli measurements, and ancilla-assisted QPT with mutually unbiased bases (MUB) measurements. In each case, we provide a closed form solution for the least-squares estimator of the Choi matrix. We propose a novel, two-step method for projecting these estimators onto the set of matrices representing physical quantum channels, and a fast numerical implementation in the form of the hyperplane intersection projection algorithm. We provide rigorous, non-asymptotic concentration bounds, sampling complexities and confidence regions for the Frobenius and trace-norm error of the estimators. For the Frobenius error, the bounds are linear in the rank of the Choi matrix, and for low ranks, they improve the error rates of the least squares estimator by a factor d2 , where d is the system dimension. We illustrate the method with numerical experiments involving channels on systems with up to 7 qubits, and find that PLS has highly competitive accuracy and computational tractability.

arXiv:2107.01060•Quantum•Oct 20, 2022•48 citations

Quantum tomography for non-iid sources

Leonardo Zambrano

Quantum state and process tomography are typically analyzed under the assumption that devices emit independent and identically distributed (i.i.d.) states or channels. In realistic experiments, however, noise, drift, feedback, or adaptive behavior can violate this assumption. We show that projected least-squares tomography remains statistically optimal even under fully adaptive state and channel preparation. Specifically, we prove that the sample complexity for reconstructing the time-averaged state or channel matches the optimal i.i.d. scaling up to logarithmic factors for non-adaptive, single-copy measurements. For average states of rank $r$, the sample complexity is $\tilde{\mathcal{O}}(d r^2/ε^2)$ to achieve trace-norm accuracy $ε$, while for process tomography it is $\tilde{\mathcal{O}}(d^6/ε^2)$ to achieve diamond-norm accuracy $ε$. Thus, dropping the i.i.d.\ assumption does not increase the fundamental sample complexity of quantum tomography, but only changes the interpretation of the reconstructed object. Finally, by tracking the drift of the source using only past outcomes, we certify high average preparation purity and entanglement even when the time-averaged state is mixed or separable.

quant-pharXiv:2602.22057•Feb 26, 2026•5 citations

Quantum channel tomography: optimal bounds in the near-boundary regime

Kean Chen, Aadil Oufkir

Quantum channel tomography (also known as quantum process tomography) is a fundamental task in quantum information and quantum computing, aiming to learn the full description of an unknown quantum channel from black-box queries. For channels with input dimension $d_1$, output dimension $d_2$, and Kraus rank at most $r$, [CGO+26] identifies the dilation rate $τ=rd_2/d_1\geq 1$ as a key parameter governing a Heisenberg-to-classical transition in query complexity. While the optimal bounds in the boundary regime ($τ=1$) and the away-from-boundary regime ($τ\geq 1+Ω(1)$) are now known, the precise behavior in the near-boundary regime ($1<τ<1+o(1)$) has remained unresolved. In this paper, we close this gap, thereby fully settling the query complexity of quantum channel tomography. Specifically, we show that the query complexity is - $Ω(rd_1d_2/\varepsilon^2)$ for diamond norm error $\varepsilon$, matching the upper bound by [MB26, CGO+26], - $Θ(d_1^2/\varepsilon+\min\{(τ-1)d_1^2,(τ-1)^2d_1^2d_2^2\}/\varepsilon^2)$ for Choi trace norm error $\varepsilon$. Our results reveal two qualitatively different forms of the Heisenberg-to-classical transition. For diamond norm tomography, the Heisenberg scaling disappears suddenly as soon as $τ$ leaves the boundary regime. In contrast, Choi trace norm tomography exhibits a gradual transition from Heisenberg to classical scaling as $τ$ moves from the boundary regime to the away-from-boundary regime.

quant-pharXiv:2611.03639•Nov 3, 2026•0 citations

Gradient-Descent Quantum Process Tomography by Learning Kraus Operators

Shahnawaz Ahmed, Fernando Quijandría, Anton Frisk Kockum

We perform quantum process tomography (QPT) for both discrete- and continuous-variable quantum systems by learning a process representation using Kraus operators. The Kraus form ensures that the reconstructed process is completely positive. To make the process trace preserving, we use a constrained gradient-descent (GD) approach on the so-called Stiefel manifold during optimization to obtain the Kraus operators. Our ansatz uses a few Kraus operators to avoid direct estimation of large process matrices, e.g., the Choi matrix, for low-rank quantum processes. The GD-QPT matches the performance of both compressed-sensing (CS) and projected least-squares (PLS) QPT in benchmarks with two-qubit random processes, but shines by combining the best features of these two methods. Similar to CS (but unlike PLS), GD-QPT can reconstruct a process from just a small number of random measurements, and similar to PLS (but unlike CS) it also works for larger system sizes, up to at least five qubits. We envisage that the data-driven approach of GD-QPT can become a practical tool that greatly reduces the cost and computational effort for QPT in intermediate-scale quantum systems.

arXiv:2208.00812•Physical Review Letters•Apr 14, 2023•35 citations

Fast quantum measurement tomography with optimal error bounds

Leonardo Zambrano, Sergi Ramos-Calderer, Richard Kueng

We present a two-step protocol for quantum measurement tomography that is light on classical co-processing cost and still achieves optimal sample complexity. Given measurement data from a known probe state ensemble, we first apply least-squares estimation to produce an unconstrained approximation of the POVM, and then project this estimate onto the set of valid quantum measurements. For a POVM with L outcomes acting on a d -dimensional system, we show that the protocol requires O ( ( d 3 + d 2 L ) / ϵ 2 ) samples to achieve error ϵ in worst-case distance, and O ( d 2 L / ϵ 2 ) samples in average-case distance. We further establish two matching sample complexity lower bounds of Ω ( ( d 3 + d 2 L ) / ϵ 2 ) and Ω ( d 2 L / ϵ 2 ) for any non-adaptive, single-copy POVM tomography protocol. Hence, our projected least squares POVM tomography is sample-optimal in both the dimension and the number of outcomes for both distances. Our method admits an analytic form when using global or local 2-designs as probe ensembles and enables rigorous non-asymptotic error guarantees. Finally, we also complement our findings with empirical performance studies carried out on a noisy superconducting quantum computer with flux-tunable transmon qubits.

arXiv:2507.04500•Quantum•Jul 15, 2026•18 citations
Quantum process tomography (QPT) protocols are often designed by lifting quantum state tomography (QST) protocols through the Choi-Jamiolkowski isomorphism, at the cost of an extra step enforcing the trace-preserving (TP) constraint. Existing approaches fall into two classes. The folklore partial normalization, which sandwiches the density estimate between inverse square roots of its input marginal, has a closed form but is used as a heuristic, its only analyses being asymptotic and losing a factor polynomial in the dimension. Projection-based methods are principled but have no closed forms and scale poorly with dimension. We resolve this dichotomy by showing that the partial normalization is the exact projection onto the Choi states of quantum channels in fidelity (equivalently in Bures or purified distance). Consequently, any QST algorithm with guarantees in fidelity lifts to a QPT algorithm that inherits its sample complexity, through a single closed-form congruence that preserves the Choi rank and is computed in Kraus form without any eigendecomposition of Choi dimension. We lift two single-copy protocols and prove that the non-adaptive one (Fidelity Projected Least Squares) recovers the Choi rank and reaches infidelity ε\varepsilonε from a number of channel uses proportional to 1/ε1/\varepsilon1/ε. Numerically, for channels of low Choi rank, it improves on the state-of-the-art projected least squares in accuracy, and by orders of magnitude in the time and memory of the TP-regularization step and of the returned estimate.
Ancilla-assisted quantum process tomography reduces channel estimation to state tomography of a Choi state, but the estimate must additionally satisfy the trace-preserving marginal constraint. This paper studies whether the usual partial normalization can enforce that constraint while retaining statistical guarantees and low-rank structure.
Using a fixed-marginal fidelity-projection result established in a companion paper, it identifies NA(ρ^)=dA−1(ρ^A−1/2⊗I)ρ^(ρ^A−1/2⊗I)\mathrm{N}_A(\hat\rho)=d_A^{-1}(\hat\rho_A^{-1/2}\otimes I)\hat\rho(\hat\rho_A^{-1/2}\otimes I)NA​(ρ^​)=dA−1​(ρ^​A−1/2​⊗I)ρ^​(ρ^​A−1/2​⊗I) as the unique closest channel Choi state in Bures or purified distance, assuming ρ^A>0\hat\rho_A>0ρ^​A​>0. Triangle inequality bounds the increase in Bures error by 222, and in root-infidelity by 444, yielding a generic lift of fidelity-guaranteed state-tomography protocols. The correction preserves rank and costs O(rdA2dB+dA3)O(rd_A^2d_B+d_A^3)O(rdA2​dB​+dA3​) operations in Kraus form.
For non-adaptive local-Pauli tomography, Fidelity Projected Least Squares thresholds the least-squares spectrum at a matrix-Bernstein concentration radius. Concentration, rank recovery and a quadratic operator-error-to-infidelity bound give channel-use complexity O(r2D1.6log⁡(D/δ)/ε2)O(r^2D^{1.6}\log(D/\delta)/\varepsilon^2)O(r2D1.6log(D/δ)/ε2) generally, improving to O(rλr−1D1.6log⁡(D/δ)/ε)O(r\lambda_r^{-1}D^{1.6}\log(D/\delta)/\varepsilon)O(rλr−1​D1.6log(D/δ)/ε) when ε≤rλr/3\varepsilon\le r\lambda_r/3ε≤rλr​/3, where D=dAdBD=d_Ad_BD=dA​dB​ and λr\lambda_rλr​ is the smallest nonzero Choi eigenvalue.
Simulations through D=1024D=1024D=1024 compare against projected least squares, reporting low-rank infidelity scaling as N−1N^{-1}N−1 rather than N−1/2N^{-1/2}N−1/2. Limitations include spectrum-dependent onset, poorer pre-onset and nearly-full-rank performance, and dimension-dependent conversion to diamond error. Numerical experiments generally halve the proven threshold heuristically; reported computational savings concern TP regularization and output storage, not the full reconstruction pipeline.
Quantum process tomography, the task of estimating an unknown quantum channel, is a central problem in quantum information theory. A long-standing open question is how many uses of an unknown channel are required to learn it in diamond distance, the standard metric for distinguishing quantum processes. While quantum state tomography is well understood, for general channels the problem remained open beyond the unitary case. Here we establish the query complexity of channel tomography with optimal dependence on the dimension parameters, at any fixed constant accuracy. We design an algorithm showing that any channel with input/output dimensions din,doutd_{\mathrm{in}},d_{\mathrm{out}}din​,dout​ and Kraus rank at most kkk can be learned to accuracy ε\varepsilonε using O(dindoutk/ε2)O(d_{\mathrm{in}}d_{\mathrm{out}}k/\varepsilon^{2})O(din​dout​k/ε2) channel uses. Conversely, we prove that Ω(dindoutk)Ω(d_{\mathrm{in}}d_{\mathrm{out}}k)Ω(din​dout​k) uses are necessary at constant accuracy and that, for non-minimal Kraus rank, a separate Ω(1/ε2)Ω(1/\varepsilon^{2})Ω(1/ε2) contribution is unavoidable. Since channels subsume states, unitaries, isometries, and measurements as special cases, our protocol provides a unified framework for these tomography tasks, yielding new guarantees for isometry and measurement tomography while recovering known optimal scalings for state and unitary tomography. Our algorithm follows the natural strategy of performing optimal tomography on the Choi state. The main technical contribution is to show that this suffices to control the induced diamond-distance error, avoiding the dimension loss incurred by a naive conversion from Choi-state trace distance to channel diamond distance. The protocol uses the channel non-adaptively to prepare Choi-state copies, purifies them in parallel, and performs optimal pure-state tomography on the resulting purifications. Hence, we reduce channel tomography to pure-state tomography.
We propose and investigate a new method of quantum process tomography (QPT) which we call projected least squares (PLS). In short, PLS consists of first computing the least-squares estimator of the Choi matrix of an unknown channel, and subsequently projecting it onto the convex set of Choi matrices. We consider four experimental setups including direct QPT with Pauli eigenvectors as input and Pauli measurements, and ancilla-assisted QPT with mutually unbiased bases (MUB) measurements. In each case, we provide a closed form solution for the least-squares estimator of the Choi matrix. We propose a novel, two-step method for projecting these estimators onto the set of matrices representing physical quantum channels, and a fast numerical implementation in the form of the hyperplane intersection projection algorithm. We provide rigorous, non-asymptotic concentration bounds, sampling complexities and confidence regions for the Frobenius and trace-norm error of the estimators. For the Frobenius error, the bounds are linear in the rank of the Choi matrix, and for low ranks, they improve the error rates of the least squares estimator by a factor d2 , where d is the system dimension. We illustrate the method with numerical experiments involving channels on systems with up to 7 qubits, and find that PLS has highly competitive accuracy and computational tractability.
Quantum state and process tomography are typically analyzed under the assumption that devices emit independent and identically distributed (i.i.d.) states or channels. In realistic experiments, however, noise, drift, feedback, or adaptive behavior can violate this assumption. We show that projected least-squares tomography remains statistically optimal even under fully adaptive state and channel preparation. Specifically, we prove that the sample complexity for reconstructing the time-averaged state or channel matches the optimal i.i.d. scaling up to logarithmic factors for non-adaptive, single-copy measurements. For average states of rank rrr, the sample complexity is O~(dr2/ε2)\tilde{\mathcal{O}}(d r^2/ε^2)O~(dr2/ε2) to achieve trace-norm accuracy εεε, while for process tomography it is O~(d6/ε2)\tilde{\mathcal{O}}(d^6/ε^2)O~(d6/ε2) to achieve diamond-norm accuracy εεε. Thus, dropping the i.i.d.\ assumption does not increase the fundamental sample complexity of quantum tomography, but only changes the interpretation of the reconstructed object. Finally, by tracking the drift of the source using only past outcomes, we certify high average preparation purity and entanglement even when the time-averaged state is mixed or separable.
Quantum channel tomography (also known as quantum process tomography) is a fundamental task in quantum information and quantum computing, aiming to learn the full description of an unknown quantum channel from black-box queries. For channels with input dimension d1d_1d1​, output dimension d2d_2d2​, and Kraus rank at most rrr, [CGO+26] identifies the dilation rate τ=rd2/d1≥1τ=rd_2/d_1\geq 1τ=rd2​/d1​≥1 as a key parameter governing a Heisenberg-to-classical transition in query complexity. While the optimal bounds in the boundary regime (τ=1τ=1τ=1) and the away-from-boundary regime (τ≥1+Ω(1)τ\geq 1+Ω(1)τ≥1+Ω(1)) are now known, the precise behavior in the near-boundary regime (1<τ<1+o(1)1<τ<1+o(1)1<τ<1+o(1)) has remained unresolved. In this paper, we close this gap, thereby fully settling the query complexity of quantum channel tomography. Specifically, we show that the query complexity is - Ω(rd1d2/ε2)Ω(rd_1d_2/\varepsilon^2)Ω(rd1​d2​/ε2) for diamond norm error ε\varepsilonε, matching the upper bound by [MB26, CGO+26], - Θ(d12/ε+min⁡{(τ−1)d12,(τ−1)2d12d22}/ε2)Θ(d_1^2/\varepsilon+\min\{(τ-1)d_1^2,(τ-1)^2d_1^2d_2^2\}/\varepsilon^2)Θ(d12​/ε+min{(τ−1)d12​,(τ−1)2d12​d22​}/ε2) for Choi trace norm error ε\varepsilonε. Our results reveal two qualitatively different forms of the Heisenberg-to-classical transition. For diamond norm tomography, the Heisenberg scaling disappears suddenly as soon as τττ leaves the boundary regime. In contrast, Choi trace norm tomography exhibits a gradual transition from Heisenberg to classical scaling as τττ moves from the boundary regime to the away-from-boundary regime.
We perform quantum process tomography (QPT) for both discrete- and continuous-variable quantum systems by learning a process representation using Kraus operators. The Kraus form ensures that the reconstructed process is completely positive. To make the process trace preserving, we use a constrained gradient-descent (GD) approach on the so-called Stiefel manifold during optimization to obtain the Kraus operators. Our ansatz uses a few Kraus operators to avoid direct estimation of large process matrices, e.g., the Choi matrix, for low-rank quantum processes. The GD-QPT matches the performance of both compressed-sensing (CS) and projected least-squares (PLS) QPT in benchmarks with two-qubit random processes, but shines by combining the best features of these two methods. Similar to CS (but unlike PLS), GD-QPT can reconstruct a process from just a small number of random measurements, and similar to PLS (but unlike CS) it also works for larger system sizes, up to at least five qubits. We envisage that the data-driven approach of GD-QPT can become a practical tool that greatly reduces the cost and computational effort for QPT in intermediate-scale quantum systems.
We present a two-step protocol for quantum measurement tomography that is light on classical co-processing cost and still achieves optimal sample complexity. Given measurement data from a known probe state ensemble, we first apply least-squares estimation to produce an unconstrained approximation of the POVM, and then project this estimate onto the set of valid quantum measurements. For a POVM with L outcomes acting on a d -dimensional system, we show that the protocol requires O ( ( d 3 + d 2 L ) / ϵ 2 ) samples to achieve error ϵ in worst-case distance, and O ( d 2 L / ϵ 2 ) samples in average-case distance. We further establish two matching sample complexity lower bounds of Ω ( ( d 3 + d 2 L ) / ϵ 2 ) and Ω ( d 2 L / ϵ 2 ) for any non-adaptive, single-copy POVM tomography protocol. Hence, our projected least squares POVM tomography is sample-optimal in both the dimension and the number of outcomes for both distances. Our method admits an analytic form when using global or local 2-designs as probe ensembles and enables rigorous non-asymptotic error guarantees. Finally, we also complement our findings with empirical performance studies carried out on a noisy superconducting quantum computer with flux-tunable transmon qubits.