arxiv: 2605.04136 · v1 · submitted 2026-05-05 · 🪐 quant-ph

Recognition: 2 theorem links

· Lean Theorem

Multiparameter function estimation for general Hamiltonians

Alexey V. Gorshkov, Erfan Abbasgholinejad, Jacob Bringewatt, Lorcan O. Conlon, Sean R. Muleady

Pith reviewed 2026-05-08 18:31 UTC · model grok-4.3

classification 🪐 quant-ph

keywords quantum sensingparameter estimationquantum Cramér-Rao boundmultiparameter estimationHamiltonian estimationquantum metrologyfunction estimationprecision limits

0 comments

The pith

The ultimate quantum limit for estimating any function of parameters in a general Hamiltonian reduces to an optimized single-parameter quantum Cramér-Rao bound.

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

The paper derives the fundamental precision limit for estimating functions of multiple parameters encoded in a quantum Hamiltonian. It shows that this multiparameter task can be reduced to finding the best single-parameter quantum Cramér-Rao bound through optimization. A protocol is presented that achieves this limit for arbitrary sets of generators, even when they do not commute. This provides a unifying framework that extends earlier results on quantum sensing and parameter estimation.

Core claim

We derive the ultimate quantum limit and present an estimation protocol for any function of parameters in a general Hamiltonian that attains this bound. We show that, although the task is fundamentally a multiparameter problem, our tight bound reduces to an optimized single-parameter quantum Cramér-Rao bound, even for arbitrary generator sets.

What carries the argument

The reduction of the multiparameter function estimation problem to an optimized single-parameter quantum Cramér-Rao bound, which carries the argument by showing equivalence despite non-commuting generators.

Load-bearing premise

That there exists an estimation protocol capable of attaining the reduced single-parameter bound even when the generators are non-commuting and arbitrary.

What would settle it

An experiment measuring a function of two non-commuting parameters in a Hamiltonian where the achieved precision exceeds the optimized single-parameter quantum Cramér-Rao bound would falsify the claim.

read the original abstract

Estimation of physical parameters encoded in a Hamiltonian is a central task in quantum sensing and learning. While the ultimate precision limit for estimating a single parameter coupled to a single generator is well established, the corresponding bound for estimating a function of multiple parameters-each coupled to distinct and possibly non-commuting generators-remains unknown in general. Here, we derive the ultimate quantum limit and present an estimation protocol for any function of parameters in a general Hamiltonian that attains this bound. We show that, although the task is fundamentally a multiparameter problem, our tight bound reduces to an optimized single-parameter quantum Cram\'er-Rao bound, even for arbitrary generator sets. Our result unifies and extends previous works, providing a general framework for optimal function estimation in quantum systems.

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.

Desk Editor's Note private letter to a colleague

The paper derives a general bound for estimating any function of multiple parameters in a Hamiltonian that reduces to an optimized single-parameter QCRB and claims an attaining protocol even for non-commuting generators.

read the letter

The main point is that this work derives the ultimate quantum limit for estimating any function of several parameters encoded in a general Hamiltonian and gives a protocol that reaches it. The bound simplifies to an optimized single-parameter quantum Cramér-Rao bound despite the multiparameter nature and possible non-commuting generators. This unifies earlier single-parameter results with some multiparameter cases and offers a framework for function estimation in quantum sensing and learning setups. The reduction avoids wrestling with the full multiparameter quantum Fisher information matrix, which is often intractable when generators fail to commute. If the proofs hold, this is a practical extension for tasks where the quantity of interest is a combination rather than individual parameters. The derivation appears to rest on standard quantum information tools without obvious circular steps. The soft spot is the attainability claim for arbitrary non-commuting generator sets. In multiparameter metrology, incompatibility frequently blocks saturation of the bound, so the protocol must use adaptive measurements or probe states to work around that. The abstract gives no explicit conditions on the function or Hamiltonian that guarantee this, which leaves open whether the result applies as broadly as stated or requires extra assumptions. Without concrete examples or the full construction, it is hard to judge how general the attaining protocol really is. This paper is for researchers in quantum metrology and sensing who need bounds for composite parameters. A reader working on precision measurements or quantum learning algorithms would get direct value from the framework if the math checks out. It has enough substance and potential impact to deserve a serious referee, even if the review focuses on clarifying the attainability conditions. I recommend sending it to peer review.

Referee Report

2 major / 2 minor

Summary. The manuscript derives the ultimate quantum limit on estimating an arbitrary function of parameters encoded in a general Hamiltonian and constructs an estimation protocol that attains this bound. It asserts that the multiparameter problem reduces exactly to an optimized single-parameter quantum Cramér-Rao bound, even when the generators are arbitrary and non-commuting.

Significance. If the central reduction and attainability hold, the result unifies and extends prior single- and multiparameter metrology bounds, supplying a general, practical framework for function estimation in quantum sensing. The explicit protocol and the reduction to a single-parameter QCRB (when valid) would constitute a substantive advance.

major comments (2)

[Main theorem / protocol section] The abstract and the derivation of the reduced bound (presumably around the main theorem) claim attainability of the optimized single-parameter QCRB for arbitrary non-commuting generator sets. Standard multiparameter QFI theory shows that incompatibility generally prevents saturation; the manuscript must therefore supply an explicit protocol construction together with the precise conditions on the function f and the Hamiltonian under which the effective generator remains compatible or the bound is nevertheless saturated (e.g., via adaptive measurements or specific probe states).
[Derivation of the bound] The reduction step from the multiparameter quantum Fisher information matrix to the optimized single-parameter QCRB must be shown to be non-circular. If the optimization over the effective generator is performed after assuming the bound is attainable, the argument risks begging the question; an independent derivation or a concrete example with non-commuting generators (e.g., Pauli X and Y) should be provided to confirm the reduction.

minor comments (2)

Clarify the notation for the function f(θ) and the effective generator throughout; ensure every equation is numbered and cross-referenced.
Add a short discussion of the computational cost or experimental feasibility of the proposed protocol for general Hamiltonians.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments. We address the major points below, clarifying the protocol construction, conditions for attainability, and the independence of the bound derivation. Revisions have been made to improve explicitness and add supporting examples.

read point-by-point responses

Referee: [Main theorem / protocol section] The abstract and the derivation of the reduced bound (presumably around the main theorem) claim attainability of the optimized single-parameter QCRB for arbitrary non-commuting generator sets. Standard multiparameter QFI theory shows that incompatibility generally prevents saturation; the manuscript must therefore supply an explicit protocol construction together with the precise conditions on the function f and the Hamiltonian under which the effective generator remains compatible or the bound is nevertheless saturated (e.g., via adaptive measurements or specific probe states).

Authors: Section III of the manuscript presents the explicit protocol: prepare a probe state that is an eigenstate of the effective generator G_eff = sum (∂f/∂θ_i) G_i, evolve under the Hamiltonian, and perform a projective measurement in the eigenbasis of G_eff. This saturates the single-parameter QCRB for the effective parameter φ = f(θ). The conditions are that f is differentiable and the Hamiltonian is linear in the parameters θ; no adaptive measurements are needed. For non-commuting generators, saturation holds because the estimation is confined to the one-dimensional subspace along ∇f, where the effective dynamics commute by construction. We have revised the main theorem section to state these conditions explicitly and added a worked example with non-commuting Pauli X and Y generators demonstrating saturation. revision: yes
Referee: [Derivation of the bound] The reduction step from the multiparameter quantum Fisher information matrix to the optimized single-parameter QCRB must be shown to be non-circular. If the optimization over the effective generator is performed after assuming the bound is attainable, the argument risks begging the question; an independent derivation or a concrete example with non-commuting generators (e.g., Pauli X and Y) should be provided to confirm the reduction.

Authors: The reduction starts from the multiparameter quantum Cramér-Rao inequality on the covariance matrix of unbiased estimators, then applies the delta method to bound Var(f̂) ≥ ∇f^T F^{-1} ∇f, where F is the QFI matrix. This expression is mathematically identical to the single-parameter QCRB for the effective generator G_eff = ∇f · G, without presupposing attainability. Attainability is shown separately via the protocol in Section III. To eliminate any ambiguity, we have added an independent derivation in a new Appendix B that derives the bound directly from the definition of the QFI without reference to the protocol, and included a concrete numerical example with non-commuting Pauli X and Y on a single qubit, verifying that the bound is saturated. revision: yes

Circularity Check

0 steps flagged

Derivation of multiparameter function estimation bound is self-contained

full rationale

The paper derives the ultimate quantum limit for any function of parameters in a general Hamiltonian from standard quantum metrology principles and presents an attaining protocol. The claimed reduction of the multiparameter task to an optimized single-parameter QCRB is presented as a derived result rather than a definitional equivalence, fitted parameter, or self-citation chain. No load-bearing steps reduce by construction to inputs; the abstract and description indicate independent content grounded in QFI analysis for arbitrary generators, with the protocol providing external attainability. This is the normal case of a non-circular derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based on the abstract alone, no explicit free parameters, axioms, or invented entities are stated. The work appears to build on the standard quantum Cramér-Rao bound framework without introducing new entities.

pith-pipeline@v0.9.0 · 5439 in / 1187 out tokens · 29079 ms · 2026-05-08T18:31:26.937576+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith.Cost.FunctionalEquation (J(x) = ½(x+x⁻¹) − 1) washburn_uniqueness_aczel unclear
Var(qest) ≥ γ²(g|α)/t² ... γ(g|α) := min_Â ½‖Â‖₁ s.t. Tr(Â ĝ_j) = α_j, Tr(Â) = 0

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Distributed estimation of many-body Hamiltonians via punctured surface code
quant-ph 2026-05 unverdicted novelty 7.0

Punctured surface codes map disjoint or overlapping Z-couplings to a single logical Z for protected distributed estimation of many-body Hamiltonian parameters.

Reference graph

Works this paper leans on

63 extracted references · 5 canonical work pages · cited by 1 Pith paper · 1 internal anchor

[1]

Giovannetti, S

V. Giovannetti, S. Lloyd, and L. Maccone, Nature Pho- tonics5, 222 (2011)

2011
[2]

Degen, F

C. Degen, F. Reinhard, and P. Cappellaro, Reviews of Modern Physics89, 035002 (2017)

2017
[3]

J. J. Bollinger, W. M. Itano, D. J. Wineland, and D. J. Heinzen, Physical Review A54, R4649 (1996)

1996
[4]

Boixo, S

S. Boixo, S. T. Flammia, C. M. Caves, and J. M. Geremia, Physical Review Letters98, 090401 (2007)

2007
[5]

S. L. Braunstein and C. M. Caves, Physical Review Let- ters72, 3439 (1994)

1994
[6]

Pang and T

S. Pang and T. A. Brun, Physical Review A90, 022117 (2014)

2014
[7]

C. W. Helstrom, Physics Letters A25, 101 (1967)

1967
[8]

Helstrom, IEEE Transactions on Information Theory 14, 234 (1968)

C. Helstrom, IEEE Transactions on Information Theory 14, 234 (1968)

1968
[9]

Yuen and M

H. Yuen and M. Lax, IEEE Transactions on Information Theory19, 740 (1973)

1973
[10]

R. D. Gill and S. Massar, Physical Review A61, 042312 (2000)

2000
[11]

A. S. Holevo, Journal of Multivariate Analysis3, 337 (1973)

1973
[12]

A. S. Holevo,Probabilistic and statistical aspects of quan- tum theory, Vol. 1 (Springer, Berlin, 2011)

2011
[13]

Nagaoka, inAsymptotic Theory of Quantum Statisti- cal Inference: Selected Papers(2005) pp

H. Nagaoka, inAsymptotic Theory of Quantum Statisti- cal Inference: Selected Papers(2005) pp. 100–112

2005
[14]

Nagaoka, inAsymptotic Theory of Quantum Statis- tical Inference: Selected Papers(World Scientific, 2005) pp

H. Nagaoka, inAsymptotic Theory of Quantum Statis- tical Inference: Selected Papers(World Scientific, 2005) pp. 133–149

2005
[15]

L. O. Conlon, J. Suzuki, P. K. Lam, and S. M. Assad, npj Quantum Information7, 110 (2021)

2021
[16]

Optimally learning functions in interacting quantum sensor networks,

E. Abbasgholinejad, S. R. Muleady, J. Bringewatt, A. J. Brady, Y.-X. Wang, A. Fahimniya, and A. V. Gorshkov, “Optimally learning functions in interacting quantum sensor networks,” (2025), arXiv:2510.06360 [quant-ph]

work page arXiv 2025
[17]

Net- worked quantum sensing,

T. J. Proctor, P. A. Knott, and J. A. Dunningham, “Net- worked quantum sensing,” (2017), arXiv:1702.04271 [quant-ph]

work page arXiv 2017
[18]

T. J. Proctor, P. A. Knott, and J. A. Dunningham, Physical Review Letters120, 080501 (2018)

2018
[19]

Ehrenberg, J

A. Ehrenberg, J. Bringewatt, and A. V. Gorshkov, Phys- ical Review Research5, 033228 (2023)

2023
[20]

Bringewatt, A

J. Bringewatt, A. Ehrenberg, T. Goel, and A. V. Gor- shkov, Physical Review Research6, 013246 (2024)

2024
[21]

Eldredge, M

Z. Eldredge, M. Foss-Feig, J. A. Gross, S. L. Rolston, and A. V. Gorshkov, Physical Review A97, 042337 (2018)

2018
[22]

Suzuki, Y

J. Suzuki, Y. Yang, and M. Hayashi, Journal of Physics A: Mathematical and Theoretical53, 453001 (2020)

2020
[23]

Zhang and Q

Z. Zhang and Q. Zhuang, Quantum Science and Technol- ogy6, 043001 (2021)

2021
[24]

Hamann, P

A. Hamann, P. Sekatski, and W. D¨ ur, Quantum Science & Technology7, 025003 (2022)

2022
[25]

J. Bate, A. Hamann, M. Canteri, A. Winkler, Z. X. Koong, V. Krutyanskiy, W. D¨ ur, and B. P. Lanyon, Physical Review Letters135, 220801 (2025)

2025
[26]

Bothwell, C

T. Bothwell, C. J. Kennedy, A. Aeppli, D. Kedar, J. M. Robinson, E. Oelker, A. Staron, and J. Ye, Nature602, 420 (2022)

2022
[27]

K´ om´ ar, E

P. K´ om´ ar, E. M. Kessler, M. Bishof, L. Jiang, A. S. Sørensen, J. Ye, and M. D. Lukin, Nature Physics10, 582 (2014). 7

2014
[28]

Rubio, P

J. Rubio, P. A. Knott, T. J. Proctor, and J. A. Dun- ningham, Journal of Physics A: Mathematical and The- oretical53, 344001 (2020)

2020
[29]

Bringewatt, I

J. Bringewatt, I. Boettcher, P. Niroula, P. Bienias, and A. V. Gorshkov, Physical Review Research3, 033011 (2021)

2021
[30]

Albarelli, J

F. Albarelli, J. F. Friel, and A. Datta, Physical Review Letters123, 200503 (2019)

2019
[31]

Learning𝑘-body hamiltonians via compressed sensing.arXiv:2410.18928, 2024

M. Ma, S. T. Flammia, J. Preskill, and Y. Tong, “Learningk-body hamiltonians via compressed sensing,” (2024), arXiv:2410.18928 [quant-ph]

work page arXiv 2024
[32]

Chen, H.-Y

C.-F. Chen, H.-Y. Huang, R. Kueng, and J. A. Tropp, PRX Quantum2, 040305 (2021)

2021
[33]

Campbell, Physical Review Letters123, 070503 (2019)

E. Campbell, Physical Review Letters123, 070503 (2019)

2019
[34]

A. M. Childs, A. Ostrander, and Y. Su, Quantum3, 182 (2019)

2019
[35]

Huang, Y

H.-Y. Huang, Y. Tong, D. Fang, and Y. Su, Physical Review Letters130, 200403 (2023)

2023
[36]

A. M. Childs, Y. Su, M. C. Tran, N. Wiebe, and S. Zhu, Physical Review X11, 011020 (2021)

2021
[37]

Bertsimas and J

D. Bertsimas and J. N. Tsitsiklis,Introduction to linear optimization, Vol. 6 (Athena Scientific, 1997)

1997
[38]

Chv´ atal,Linear programming(Macmillan, 1983)

V. Chv´ atal,Linear programming(Macmillan, 1983)

1983
[39]

S. P. Boyd and L. Vandenberghe,Convex optimization (Cambridge University Press, 2004)

2004
[40]

T. Qian, J. Bringewatt, I. Boettcher, P. Bienias, and A. V. Gorshkov, Physical Review A103, L030601 (2021)

2021
[41]

H.-Y. Hu, M. Ma, W. Gong, Q. Ye, Y. Tong, S. T. Flam- mia, and S. F. Yelin, PRX Quantum6, 040315 (2025)

2025
[42]

Learning hamiltonians in the heisenberg limit with static single-qubit fields.arXiv preprint arXiv:2601.10380, 2026

S. Brahmachari, S. Zhu, I. Marvian, and Y. Tong, “Learning hamiltonians in the heisenberg limit with static single-qubit fields,” (2026), arXiv:2601.10380 [quant-ph]

work page arXiv 2026
[43]

Bakshi, A

A. Bakshi, A. Liu, A. Moitra, and E. Tang, in2024 IEEE 65th Annual Symposium on Foundations of Com- puter Science (FOCS)(IEEE, 2024) p. 1037–1050

2024
[44]

S. Ragy, M. Jarzyna, and R. Demkowicz-Dobrza´ nski, Physical Review A94, 052108 (2016)

2016
[45]

J. S. Sidhu, Y. Ouyang, E. T. Campbell, and P. Kok, Physical Review X11, 011028 (2021)

2021
[46]

Albarelli and R

F. Albarelli and R. Demkowicz-Dobrza´ nski, Physical Re- view X12, 011039 (2022)

2022
[47]

S. Zhou, M. Zhang, J. Preskill, and L. Jiang, Nature Communications9, 78 (2018)

2018
[48]

Demkowicz-Dobrza´ nski, J

R. Demkowicz-Dobrza´ nski, J. Czajkowski, and P. Sekatski, Physical Review X7, 041009 (2017)

2017
[49]

Pang and A

S. Pang and A. N. Jordan, Nature Communications8, 14695 (2017)

2017
[50]

Yuan and C.-H

H. Yuan and C.-H. F. Fung, Physical Review Letters115, 110401 (2015)

2015
[51]

Lieb-Mattis states for robust entangled differential phase sensing

R. Kaubruegger, D. F. Padilla, A. Shankar, C. Hotter, S. R. Muleady, J. Bringewatt, Y. Baamara, E. Abbasgho- linejad, A. V. Gorshkov, K. Mølmer,et al., “Lieb-mattis states for robust entangled differential phase sensing,” (2025), arXiv:2506.10151 [quant-ph]

work page internal anchor Pith review Pith/arXiv arXiv 2025
[52]

J. A. Gross and C. M. Caves, Journal of Physics A: Math- ematical and Theoretical54, 014001 (2020). 8 Appendix A: Optimized quantum Cram´ er–Rao bound and SDP duality In this Appendix we derive the optimized quantum Cram´ er–Rao bound used in Eqs. (7), (19) of the main text. Consider a controlled Hamiltonian ˆH(s) = ˆH0(θθθ) + ˆHc(s), ˆH0(θθθ) = mX j=1 fj...

2020
[53]

This result is used in Appendix B 2 to bound the error of the two-step protocol

Stability ofγ( ˆggg|ααα)under perturbations First, we quantify how the SDP solution valueγ( ˆggg|ααα) changes under small perturbations of the generators ˆgj and the coefficient vectorααα. This result is used in Appendix B 2 to bound the error of the two-step protocol. Throughout we assume{ ˆI,ˆg1, . . . ,ˆgm}is linearly independent overR, so that the con...
[54]

(19), and (A10)

Error bounds for the two-step protocol Next, we bound the errors in the two-step protocol of Sec.VI, and show that it saturates the optimal bound in Eqs. (19), and (A10). a. Assumptions and notation Letθθθ∈R r be the true parameter and let ˜θθθbe the stage-1 estimate. Define the estimation error δθθθ:=θθθ− ˜θθθ,∥δθ θθ∥∞ =O(1/t 1).(B23) Assumef j(θθθ) andq...
[55]

term in Eq. (22). c. Stability ofγ(F ˆggg| ∇q) We now relateγ( ˜F ˆggg| ∇˜q) toγ(F ˆggg| ∇q). 15 Perturbations in the effective linear model.Recall the (unknown) ideal objects atθ θθ, ˆhhh:= ˆhhh(θθθ) =F(θθθ)ˆggg, α αα:=∇q(θ θθ),(B34) and their stage-1 approximations ˜ˆhhh:= ˆhhh(˜θθθ) = ˜F ˆggg, ˜ααα:=∇q( ˜θθθ) =∇˜q.(B35) We bound the induced perturbatio...
[56]

Negligibility of the Taylor remainder inq.From (B33), the linear approximation error inqsatisfies|R q|=O(1/t 2 1)

in the effective linear model and the function of interest. Negligibility of the Taylor remainder inq.From (B33), the linear approximation error inqsatisfies|R q|=O(1/t 2 1). If the stage-2 estimator is locally unbiased for the linear term∇˜q·δθθθ, thenR q appears as a bias in the overall estimator ofq, contributing Bias2(qest)≤ |R q|2 =O 1 t4 1 .(B42) To...
[57]

We computeγ( ˆσσσ|α αα) from (C1)

Single qubit: Pauli-vector model Consider a single qubit with ˆH0 = 3X j=1 θj ˆσj, q(θ θθ) =ααα·θθθ,(C2) where ˆσj ∈ { ˆX, ˆY , ˆZ}. We computeγ( ˆσσσ|α αα) from (C1). Any Hermitian traceless ˆAcan be written as ˆA=a aa· ˆσσσ. Using Tr(ˆσiˆσj) = 2δij, the constraints Tr ˆAˆσj =α j imply 2aj =α j, hence ˆA⋆ = 1 2 ααα· ˆσσσ.(C3) 17 Its eigenvalues are±∥ααα∥...
[58]

Complete Hilbert–Schmidt orthonormal traceless basis Consider the Hamiltonian ˆH0 = N 2−1X j=1 θj ˆgj, q(θ θθ) =ααα·θθθ,(C6) and let{ˆgj}N 2−1 j=1 be a Hilbert–Schmidt orthonormal basis of traceless Hermitian matrices (e.g. generalized Gell-Mann matrices or properly normalized Pauli strings): Tr(ˆgiˆgj) =δ ij,Tr(ˆg j) = 0.(C7) Then any Hermitian traceless...
[59]

We are interested in estimating only one coefficient, sayθ j0, i.e

Estimating a single Pauli coefficient in a multi-Pauli Hamiltonian Consider ann-qubit Hamiltonian with multiple Pauli-string generators, ˆH0 = mX j=1 θj ˆPj,(C10) where each ˆPj is ann-qubit Pauli string ( ˆP 2 j = ˆI, Tr ˆPj = 0), and assume the set{ ˆPj}contains no duplicates up to a sign. We are interested in estimating only one coefficient, sayθ j0, i...
[60]

Therefore, γ({ ˆPj} |eeej0) = 1 2 ,and Var(θ j0,est)≥ 1 4t2 .(C15) Notably, this value is independent of the remaining generators{ ˆPj}j̸=j0
[61]

Then γ( ˆZ1,

Commuting local generators:mindependent ˆZoperators Consider the set up studied in [21] with the Hamiltonian ˆH0 = mX j=1 θj ˆZj,(C16) whereZ j is the pauli Z operator on qubitj, andq(θ θθ) =ααα·θθθ. Then γ( ˆZ1, . . . , ˆZm |ααα) = ∥ααα∥∞ 2 ,⇒Var(q est)≥ ∥ααα∥2 ∞ 4t2 .(C17) Proof.Use the dual characterization ofγ: γ(ˆggg|ααα) = max yyy∈Rm, µ∈R ααα·yyy: m...
[62]

Although the generators do not commute (e.g

Two-qubit noncommuting example: ˆZ1, ˆX1, ˆZ1 ˆZ2 Consider the Hamiltonian ˆH0 =θ 1 ˆZ1 +θ 2 ˆX1 +θ 3 ˆZ1 ˆZ2,(C23) withq(θθθ) =ααα·θθθ. Although the generators do not commute (e.g. [ ˆZ1, ˆX1]̸= 0 and [ ˆX1, ˆZ1 ˆZ2]̸= 0), they all commute with ˆZ2, which enables block diagonalization, and we can show γ( ˆZ1, ˆX1, ˆZ1 ˆZ2 |α 1, α2, α3) = 1 2 q α2 2 + max...
[63]

Then the right-hand side of (C33) becomes 1 2 α2 sinθ+ (|p|+|q|) cosθ ,(C34) whose maximum overθequals 1 2 p α2 2 + (|p|+|q|) 2

Then (C30) implies|u| ≤R and|v| ≤R, and the maximum ofp u+q vis achieved atu=Rsgn(p),v=Rsgn(q), yielding max u,v (p u+q v+α 2y2) =α 2y2 +R(|p|+|q|).(C32) Thus γ(ˆggg|ααα) = max |y2|≤1/2 α2y2 + (|p|+|q|) q 1 4 −y 2 2 .(C33) 20 Step 4: solve the remaining one-dimensional maximization.Lety 2 = 1 2 sinθ, so q 1 4 −y 2 2 = 1 2 cosθ. Then the right-hand side of...