pith. sign in

arxiv: 2606.27254 · v1 · pith:Q35CYAVXnew · submitted 2026-06-25 · 🪐 quant-ph

Particle-preserving fermionic shadows with mode-independent sample complexity

Maxwell West , M. Cerezo , Martin Larocca This is my paper

Pith reviewed 2026-06-26 04:07 UTC · model grok-4.3

classification 🪐 quant-ph
keywords fermionic shadowsclassical shadowsparticle-preservingSlater determinantssample complexityAIII symmetric spaceJacobi ensembles
0
0 comments X

The pith

Classical shadows learn overlaps with any Slater determinant using O(η log η) samples independent of total modes n.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a classical shadow protocol restricted to particle-preserving fermionic unitaries. It proves that the worst-case number of samples needed to estimate the overlap between an unknown η-particle state and an arbitrary Slater determinant drops to O(η log η). The same framework yields an O(η ||h0||_2^2) bound for any particle-preserving quadratic observable. These bounds matter because they remove dependence on the orbital count n, so the cost stays manageable when particle number is small even if the mode space is large. The post-processing stays polynomial and the randomization can be realized with shallow circuits in a first-quantized encoding.

Core claim

The central claim is that particle-preserving fermionic shadows achieve a mode-independent sample complexity of O(η log η) for worst-case estimation of overlaps with Slater determinants. This is obtained by showing that the extremal shadow variance reduces to an integral over the AIII symmetric space U(n)/(U(η) imes U(n-η)) that evaluates, via Jacobi-ensemble and orthogonal-polynomial techniques, to a quantity scaling only with η. The same reduction supplies the O(η ||h0||_2^2) bound for general traceless particle-preserving quadratics, with classical post-processing costing O(n η^2) or O(n^2 η) respectively.

What carries the argument

The reduction of extremal shadow variance to an integral over the AIII symmetric space U(n)/(U(η) imes U(n-η)), evaluated with Jacobi-ensemble and orthogonal-polynomial methods.

If this is right

  • Slater-determinant overlaps require only O(η log η) samples in the worst case.
  • General particle-preserving quadratic observables are estimated with O(η ||h0||_2^2) samples.
  • Classical post-processing runs in O(n η^2) time for a dense orbital.
  • Approximate unitary designs realize the required randomization with polylog(n) depth in first-quantized encoding.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The symmetric-space analysis may extend to other conserved-quantity shadow protocols beyond fermions.
  • Resource estimates for variational algorithms on fixed-particle fermionic systems could drop by a √ n factor.
  • The Jacobi-ensemble integral technique might tighten bounds in related random-measurement schemes for symmetry-constrained states.

Load-bearing premise

The extremal shadow variance for particle-preserving measurements reduces to an integral over the AIII symmetric space that can be evaluated exactly with Jacobi-ensemble techniques.

What would settle it

Numerical Monte-Carlo sampling of the shadow variance for fixed η and increasing n that shows growth faster than O(log η) would falsify the claimed mode independence.

read the original abstract

We consider the problem of learning expectation values of particle-preserving operators with respect to an unknown $\eta$-particle $n$-mode fermionic state via classical shadows. Our main application is to estimating overlaps with arbitrary Slater determinant states: While it is known that such overlaps can, in the average case, be learnt to a fixed additive precision with a constant number of samples, the best-known worst case bound is $\mathcal{O}(\sqrt n \log n)$; here we improve this to $\mathcal{O}(\eta\log\eta)$, achieving a mode-independent sample cost. Our procedure is also computationally efficient, requiring only classical post-processing which for a generic dense orbital runs in time $\mathcal{O}(n\eta^2)$. For the task of estimating the expectation value of a general particle-preserving quadratic fermionic observable $h$, we prove a sample complexity bound of $\mathcal{O}(\eta \|h_0\|_2^2)$, where $h_0$ is the traceless component of $h$; the associated classical post-processing scales as $\mathcal{O}(n^2\eta )$. Finally, we discuss implementation of the required randomization: in a first-quantized encoding, approximate unitary designs give circuit depths polylogarithmic in the number of modes, contrasting with linear-depth requirements for nearest-neighbor second-quantized matchgate implementations. On the technical side, our proof reduces the extremal shadow variance to harmonic analysis on the AIII symmetric space $U(n)/(U(\eta)\times U(n-\eta))$ and evaluates the resulting integral using techniques from the theories of Jacobi ensembles and orthogonal polynomials, in a calculation which may be of independent interest.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 3 minor

Summary. The manuscript develops particle-preserving fermionic classical shadows for estimating expectation values of number-conserving observables on an unknown η-particle state in n modes. The central results are (i) a worst-case sample complexity of O(η log η) for additive-error estimation of overlaps with arbitrary Slater determinants (improving on the prior O(√n log n) bound), (ii) an O(η ‖h₀‖₂²) bound for general particle-preserving quadratic observables, and (iii) polylog-depth implementations via approximate unitary designs in the first-quantized encoding. The proofs reduce the extremal shadow variance to an integral over the AIII symmetric space U(n)/(U(η)×U(n-η)) and evaluate it via Jacobi-ensemble techniques and orthogonal polynomials.

Significance. If the integral evaluation is correct, the work establishes the first mode-independent sample bound for this task and supplies an explicit, computationally efficient protocol. The reduction to harmonic analysis on the symmetric space and the closed-form evaluation via orthogonal polynomials constitute a technical contribution that may be reusable beyond shadows. No machine-checked proofs or code are provided, but the derivation is parameter-free once the representation-theoretic identification is accepted.

major comments (1)
  1. [Symmetric-space integral evaluation (the section following the representation-theoretic reduction)] The O(η log η) claim (Abstract and the section deriving the sample bound) is load-bearing on the evaluation of the extremal variance integral over U(n)/(U(η)×U(n-η)). The change of variables to principal angles, the identification of the Jacobi weight, and the large-n/η asymptotic analysis must be verified to ensure no residual n-dependent factor appears; an error here would invalidate mode independence.
minor comments (3)
  1. [Theorem statements] Clarify whether the O(η log η) bound is achieved exactly or up to constants, and state the precise dependence on the target precision ε.
  2. [Computational complexity paragraph] The post-processing complexity O(n η²) for dense orbitals is stated; confirm whether this remains sub-quadratic when the orbital is sparse.
  3. [Implementation section] The discussion of approximate unitary designs should quantify the approximation error and its propagation into the sample-complexity bound.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading, accurate summary of our results, and constructive feedback. We address the single major comment below.

read point-by-point responses

  1. Referee: [Symmetric-space integral evaluation (the section following the representation-theoretic reduction)] The O(η log η) claim (Abstract and the section deriving the sample bound) is load-bearing on the evaluation of the extremal variance integral over U(n)/(U(η)×U(n-η)). The change of variables to principal angles, the identification of the Jacobi weight, and the large-n/η asymptotic analysis must be verified to ensure no residual n-dependent factor appears; an error here would invalidate mode independence.

    Authors: We agree that the mode-independent bound rests on this integral evaluation. After the representation-theoretic reduction to the AIII symmetric space, the change of variables to the principal angles φ_i maps the problem to the Jacobi ensemble with weight proportional to ∏_i (sin 2φ_i) and the appropriate Vandermonde factor arising from the root system. Standard results on the asymptotics of Jacobi orthogonal polynomials (via the Christoffel–Darboux kernel and the known large-n limit of the ensemble with η fixed or η = o(n)) then yield an explicit leading term for the extremal variance that is O(η log η) with all n-dependent prefactors canceling exactly against the normalization of the invariant measure. The calculation is parameter-free once the identification is made and contains no hidden n factors. We are prepared to expand the relevant section or add an appendix with the intermediate steps of the orthogonal-polynomial evaluation if the referee finds any step insufficiently detailed. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external RMT and representation theory

full rationale

The central O(η log η) bound is obtained by mapping the extremal shadow variance to an integral over the AIII symmetric space and evaluating that integral via Jacobi ensembles and orthogonal polynomials. This step invokes standard external techniques from harmonic analysis and random matrix theory rather than any self-definitional reduction, fitted-input-as-prediction, or load-bearing self-citation chain. The paper explicitly flags the integral calculation as potentially of independent interest, confirming it is not tautological with the input assumptions. No equations or claims reduce the result to its own fitted parameters or prior self-citations by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of mapping the shadow variance to an integral over the AIII symmetric space and its evaluation via Jacobi ensembles, which are standard mathematical tools applied here.

axioms (1)
  • standard math The extremal shadow variance reduces to an integral over the AIII symmetric space U(n)/(U(η)×U(n-η)) evaluable using Jacobi ensembles and orthogonal polynomials.
    Described as the technical core of the proof in the abstract.

pith-pipeline@v0.9.1-grok · 5825 in / 1127 out tokens · 59427 ms · 2026-06-26T04:07:23.516389+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 18 canonical work pages · 2 internal anchors

  1. [1]

    Worst This work O(ηlogη) O(η∥h 0∥2

  2. [2]

    relative unitary

    Worst TABLE I. Comparison of our sample complexity results with those of previously existing fermionic shadow schemes for our two families of observables, along with whether an average-case or worst-case bound is considered. The most direct comparison is with Ref. [6], over which we obtain clear improvements in the caseη≪ √n. h, we show that the required ...

  3. [3]

    Predicting many properties of a quantum system from very few measurements,

    Hsin-Yuan Huang, Richard Kueng, and John Preskill, “Predicting many properties of a quantum system from very few measurements,” Nature Physics16, 1050–1057 (2020)

    2020

  4. [4]

    Shadow tomography of quantum states,

    Scott Aaronson, “Shadow tomography of quantum states,” inProceedings of the 50th annual ACM SIGACT sympo- sium on theory of computing(2018) pp. 325–338

    2018

  5. [5]

    Sample-optimal tomography of quantum states,

    Jeongwan Haah, Aram W Harrow, Zhengfeng Ji, Xiaodi Wu, and Nengkun Yu, “Sample-optimal tomography of quantum states,” IEEE Transactions on Information The- ory63, 5628–5641 (2017)

    2017

  6. [6]

    Optimal lower bounds for quantum state tomog- raphy,

    Thilo Scharnhorst, Jack Spilecki, and John Wright, “Optimal lower bounds for quantum state tomog- raphy,” arXiv preprint arXiv:2510.07699 (2025), 10.48550/arXiv.2510.07699

    work page doi:10.48550/arxiv.2510.07699 2025

  7. [7]

    Classical shadows of fermions with par- ticle number symmetry,

    Guang Hao Low, “Classical shadows of fermions with par- ticle number symmetry,” arXiv preprint arXiv:2208.08964 (2022), 10.48550/arXiv.2208.08964

    work page doi:10.48550/arxiv.2208.08964 2022

  8. [8]

    Matchgate shadows for fermionic quantum sim- ulation,

    Kianna Wan, William J Huggins, Joonho Lee, and Ryan Babbush, “Matchgate shadows for fermionic quantum sim- ulation,” Communications in Mathematical Physics404, 629–700 (2023)

    2023

  9. [9]

    Fermionic partial tomography via classical shadows,

    Andrew Zhao, Nicholas C Rubin, and Akimasa Miyake, “Fermionic partial tomography via classical shadows,” Physical Review Letters127, 110504 (2021)

    2021

  10. [10]

    Classical shad- ows with arbitrary group representations,

    Maxwell West, Frederic Sauvage, Aniruddha Sen, Roy Forestano, David Wierichs, Nathan Killoran, Dmitry Grinko, M Cerezo, and Martin Larocca, “Classical shad- ows with arbitrary group representations,” arXiv preprint arXiv:2604.01429 (2026), 10.48550/arXiv.2604.01429

    work page doi:10.48550/arxiv.2604.01429 2026

  11. [12]

    Hardware-efficient learning of quan- tum many-body states,

    Katherine Van Kirk, Jordan Cotler, Hsin-Yuan Huang, and Mikhail D Lukin, “Hardware-efficient learning of quan- tum many-body states,” arXiv preprint arXiv:2212.06084 (2022), 10.48550/arXiv.2212.06084

    work page doi:10.48550/arxiv.2212.06084 2022

  12. [13]

    Classical shadows based on locally- entangled measurements,

    Matteo Ippoliti, “Classical shadows based on locally- entangled measurements,” Quantum8, 1293 (2024)

    2024

  13. [14]

    Group-theoretic error mitigation enabled by classical shadows and symmetries,

    Andrew Zhao and Akimasa Miyake, “Group-theoretic error mitigation enabled by classical shadows and symmetries,” npj Quantum Information10, 57 (2024)

    2024

  14. [15]

    Triply efficient shadow tomography,

    Robbie King, David Gosset, Robin Kothari, and Ryan Babbush, “Triply efficient shadow tomography,” PRX Quantum6, 010336 (2025)

    2025

  15. [16]

    Shadows of quantum ma- chine learning,

    Sofiene Jerbi, Casper Gyurik, Simon C Marshall, Riccardo Molteni, and Vedran Dunjko, “Shadows of quantum ma- chine learning,” Nature Communications15, 5676 (2024)

    2024

  16. [17]

    Classical shadows with noise,

    Dax Enshan Koh and Sabee Grewal, “Classical shadows with noise,” Quantum6, 776 (2022)

    2022

  17. [18]

    Robust shadow estimation,

    Senrui Chen, Wenjun Yu, Pei Zeng, and Steven T Flammia, “Robust shadow estimation,” PRX Quantum2, 030348 (2021)

    2021

  18. [19]

    Efficient local classical shadow tomography with number conservation,

    Sumner N Hearth, Michael O Flynn, Anushya Chandran, and Chris R Laumann, “Efficient local classical shadow tomography with number conservation,” Physical Review Letters133, 060802 (2024)

    2024

  19. [20]

    Algorithmic shadow spectroscopy,

    Hans Hon Sang Chan, Richard Meister, Matthew L Goh, and Bálint Koczor, “Algorithmic shadow spectroscopy,” PRX Quantum6, 010352 (2025)

    2025

  20. [21]

    Classical shadows withsymmetries,

    Frederic Sauvage and Martin Larocca, “Classical shadows withsymmetries,” arXivpreprintarXiv:2408.05279 (2024), 10.48550/arXiv.2408.05279

    work page doi:10.48550/arxiv.2408.05279 2024

  21. [22]

    Del Grosso, G

    Jonas Helsen and Michael Walter, “Thrifty shadow estimation: Reusing quantum circuits and bounding tails,” Physical Review Letters131(2023), 10.1103/phys- revlett.131.240602

    work page doi:10.1103/phys- 2023

  22. [23]

    Shadow process tomography of quantum channels,

    Jonathan Kunjummen, Minh C Tran, Daniel Carney, and Jacob M Taylor, “Shadow process tomography of quantum channels,” Physical Review A107, 042403 (2023)

    2023

  23. [24]

    Learning fermionic correlationsbyevolvingwithrandomtranslationallyinvari- ant hamiltonians,

    Janek Denzler, Antonio Anna Mele, Ellen Derbyshire, Tommaso Guaita, and Jens Eisert, “Learning fermionic correlationsbyevolvingwithrandomtranslationallyinvari- ant hamiltonians,” (2023), arXiv:2309.12933 [quant-ph]

    arXiv 2023

  24. [25]

    Random ensembles of symplectic and unitary states are indistinguishable,

    Maxwell West, Antonio Anna Mele, Martin Larocca, and M Cerezo, “Random ensembles of symplectic and unitary states are indistinguishable,” arXiv preprint 7 arXiv:2409.16500 (2024), 10.48550/arXiv.2409.16500

    work page doi:10.48550/arxiv.2409.16500 2024

  25. [26]

    Sample-optimal classical shadows for pure states,

    Daniel Grier, Hakop Pashayan, and Luke Schaeffer, “Sample-optimal classical shadows for pure states,” Quan- tum8, 1373 (2024)

    2024

  26. [27]

    Fast and robust quan- tum state tomography from few basis measure- ments,

    Fernando GSL Brandão, Richard Kueng, and Daniel Stilck França, “Fast and robust quan- tum state tomography from few basis measure- ments,” arXiv preprint arXiv:2009.08216 (2020), https://doi.org/10.4230/LIPIcs.TQC.2021.7

    work page doi:10.4230/lipics.tqc.2021.7 2009

  27. [28]

    Shal- low shadows: Expectation estimation using low-depth ran- domCliffordcircuits,

    Christian Bertoni, Jonas Haferkamp, Marcel Hinsche, Mar- ios Ioannou, Jens Eisert, and Hakop Pashayan, “Shal- low shadows: Expectation estimation using low-depth ran- domCliffordcircuits,” PhysicalReviewLetters133,020602 (2024)

    2024

  28. [29]

    Estimation of the quantum fisher information on a quantum proces- sor,

    Vittorio Vitale, Aniket Rath, Petar Jurcevic, Andreas El- ben, Cyril Branciard, and Benoît Vermersch, “Estimation of the quantum fisher information on a quantum proces- sor,” PRX Quantum5, 030338 (2024)

    2024

  29. [30]

    Shadow hamiltonian simulation,

    Rolando D Somma, Robbie King, Robin Kothari, Thomas O’Brien, and Ryan Babbush, “Shadow hamiltonian simulation,” arXiv preprint arXiv:2407.21775 (2024), 10.48550/arXiv.2407.21775

    work page doi:10.48550/arxiv.2407.21775 2024

  30. [31]

    Real classical shadows,

    Maxwell West, Antonio Anna Mele, Martín Larocca, and M Cerezo, “Real classical shadows,” Journal of Physics A: Mathematical and Theoretical58, 245304 (2025)

    2025

  31. [32]

    Classicalshadowsforsample-efficientmea- surements of gauge-invariant observables,

    Jacob Bringewatt, Henry Froland, Andreas Elben, and NiklasMueller,“Classicalshadowsforsample-efficientmea- surements of gauge-invariant observables,” arXiv preprint arXiv:2511.02904 (2025)

    Pith/arXiv arXiv 2025

  32. [33]

    Random unitaries in extremely low depth,

    Thomas Schuster, Jonas Haferkamp, and Hsin-Yuan Huang, “Random unitaries in extremely low depth,” Sci- ence389, 92–96 (2025)

    2025

  33. [34]

    Unitary designs in nearly opti- mal depth,

    Laura Cui, Thomas Schuster, Fernando Brandao, and Hsin-Yuan Huang, “Unitary designs in nearly opti- mal depth,” arXiv preprint arXiv:2507.06216 (2025), 10.48550/arXiv.2507.06216

    work page doi:10.48550/arxiv.2507.06216 2025

  34. [35]

    No-go theorems for sublinear-depth group designs,

    Maxwell West, Diego García-Martín, NL Diaz, M Cerezo, and Martin Larocca, “No-go theorems for sublinear-depth group designs,” arXiv preprint arXiv:2506.16005 (2025), 10.48550/arXiv.2506.16005

    work page doi:10.48550/arxiv.2506.16005 2025

  35. [36]

    Will it glue? on short-depth designs be- yond the unitary group,

    Lorenzo Grevink, Jonas Haferkamp, Markus Heinrich, Jonas Helsen, Marcel Hinsche, Thomas Schuster, and Zoltan Zimboras, “Will it glue? on short-depth designs be- yond the unitary group,” arXiv preprint arXiv:2506.23925 (2025), 10.48550/arXiv.2506.23925

    work page doi:10.48550/arxiv.2506.23925 2025

  36. [37]

    William Fulton and Joe Harris,Representation Theory: A First Course(Springer, 1991)

    1991

  37. [38]

    Introduction to Haar measure tools in quantum information: A beginner’s tutorial,

    Antonio Anna Mele, “Introduction to Haar measure tools in quantum information: A beginner’s tutorial,” Quantum 8, 1340 (2024)

    2024

  38. [39]

    Explicit construction of exact unitary designs,

    Eiichi Bannai, Yoshifumi Nakata, Takayuki Okuda, and Da Zhao, “Explicit construction of exact unitary designs,” Advances in Mathematics405, 108457 (2022)

    2022

  39. [40]

    Recursive cartan decompositions for unitary synthesis,

    David Wierichs, Maxwell West, Roy T Forestano, M Cerezo, and Nathan Killoran, “Recursive cartan decompositions for unitary synthesis,” arXiv preprint arXiv:2503.19014 (2025), 10.48550/arXiv.2503.19014

    work page doi:10.48550/arxiv.2503.19014 2025

  40. [41]

    A general framework for recursive decompo- sitions of unitary quantum evolutions,

    Mehmet Dagli, Domenico D’Alessandro, and Jonathan D H Smith, “A general framework for recursive decompo- sitions of unitary quantum evolutions,” Journal of Physics A: Mathematical and Theoretical41, 155302 (2008)

    2008

  41. [42]

    A tight lower bound for non-coherent index erasure,

    Nathan Lindzey and Ansis Rosmanis, “A tight lower bound for non-coherent index erasure,” arXiv preprint arXiv:1902.07336 (2019), 10.48550/arXiv.1902.07336

    work page doi:10.48550/arxiv.1902.07336 1902

  42. [43]

    Operator growth in random quantum circuits with symmetry

    Nicholas Hunter-Jones, “Operator growth in random quantum circuits with symmetry,” arXiv preprint arXiv:1812.08219 (2018), 10.48550/arXiv.1812.08219

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1812.08219 2018

  43. [44]

    On the average-case complexity of learning states from the circular and gaussian en- sembles,

    Maxwell West, “On the average-case complexity of learning states from the circular and gaussian en- sembles,” arXiv preprint arXiv:2601.10197 (2026), 10.48550/arXiv.2601.10197

    work page doi:10.48550/arxiv.2601.10197 2026

  44. [45]

    Classical shadows over symmetric spaces

    Rebecca Chang, Maureen Krumtünger, Martin Larocca, and Maxwell West, “Classical shadows over symmet- ric spaces,” arXiv preprint arXiv:2605.05518 (2026), 10.48550/arXiv.2605.05518

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.05518 2026

  45. [46]

    Optimal haar random fermionic linear optics circuits,

    Paolo Braccia, NL Diaz, Martin Larocca, M Cerezo, and Diego García-Martín, “Optimal haar random fermionic linear optics circuits,” arXiv preprint arXiv:2505.24212 (2025), 10.48550/arXiv.2505.24212

    work page doi:10.48550/arxiv.2505.24212 2025

  46. [47]

    Herbert Wilf,generatingfunctionology(CRC Press, 2005)

    2005

  47. [48]

    Singularvaluestatisticsofmatrixproductswithtruncated unitarymatrices,

    Mario Kieburg, Arno B.J. Kuijlaars, and Dries Stivigny, “Singularvaluestatisticsofmatrixproductswithtruncated unitarymatrices,” InternationalMathematicsResearchNo- tices2016, 3392–3424 (2016)

    2016

  48. [49]

    Chap- ter 5: Matrix-variate gamma and beta distributions,

    Arak Mathai, Serge Provost, and Hans Haubold, “Chap- ter 5: Matrix-variate gamma and beta distributions,” in Multivariate Statistical Analysis in the Real and Complex Domains(Springer International Publishing, Cham, 2022) pp. 289–393

    2022

  49. [50]

    Orthogonal polynomials,

    Steven Lalley, “Orthogonal polynomials,” (2013)

    2013

  50. [51]

    Convolution algebras for Heckman–Opdam polynomials derived from compact Grassmannians,

    Heiko Remling and Margit Rösler, “Convolution algebras for Heckman–Opdam polynomials derived from compact Grassmannians,” Journal of Approximation Theory197, 30–48 (2015). 8 Appendix A: Estimating Slater determinant overlaps In this Appendix we prove Theorems 1, 2 and 3. We begin by recalling some notation. We letZη ={Z⊂[n] :|Z|=η}, withD=|Z η|= n η . To ...

    2015