Genuine Multipartite Nonlocality Is Fermionic Magic

Jamal Slim

REVIEW

Genuine Multipartite Nonlocality Is Fermionic Magic

Jamal Slim

Quantum Physics · Submitted 30 Sep 2026

Pith's verdict: Proven part is solid and answers a real open question; the abstract slightly overstates scope by dropping the measurement hypothesis, but the authors flag exactly where the proof ends.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 1, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

T0 AI review An AI read the paper and answered a fixed list of questions.
T1 Audit Automatic checks that the references and files the paper relies on exist and match.
T2 Compute The paper's code or calculations re-run with the same results.
T3 Formal statement The claims written out in a formal math language a computer can read.
T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 3 major objections · 4 minor · Ask about this paper · The compiled record · Read the full review →

Challenge this review Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2609.39411 (version 2)

Pith IDC4MRWEWC, Pith's permanent record of this paper and its review

CitationJamal Slim. "Genuine Multipartite Nonlocality Is Fermionic Magic." arXiv:2609.39411 (2026). https://pith.science/paper/C4MRWEWC

Cite this reviewBibTeX for Pith's review of this paper

Claim paper

Why Pith says this

Highlight any text to comment on it.

Editor's note

read the letter

The headline: this paper proves that free fermions, under local bilinear readout in which each party's two planes share a direction or commute, carry at most the Bell nonlocality of a single Tsirelson pair—Mermin capped at 2√2 for three parties, 2^(n/2) for n. Read with the cited result that one non-Gaussian state makes matchgates universal, a violation becomes a device-independent certificate of fermionic magic. That is a genuine answer to an open question, and it is neatly opposite to the stabilizer case, where GHZ states reach maximal genuine multipartite nonlocality while remaining simulable. Fermionic parity is the obstruction, which is a satisfying mechanism.

The proven part is solid. The proof is first-principles: Pfaffian reduction to two modes per party, the vacuum-full occupancy inequality P(n)P(n̄) ≤ 2^-M, Lemma 4, and a clean Cauchy-Binet argument for the n-party law, with named numerical cross-checks at 10^-15 to 10^-9. The experiment is about as careful as hardware papers get: raw values only, a correctly derived threshold showing post-selection is not free for this bound, all twelve runs disclosed including superseded ones, an explicit retraction of an earlier optimizer artifact, and margins of 0.48–0.71 over the bound at 16 to 32 sigma across three devices and two architectures.

The soft spot is the one both the reader and the stress-test flag, and it is real but contained. The abstract's unrestricted "never genuinely multipartite nonlocal" holds under the direction-sharing/commuting hypothesis. For parties reading general bilinears over four or more Majoranas, only 3.26 is certified by interval branch-and-bound, and 3.26√2 ≈ 4.61 > 4, so a Svetlichny violation is not excluded there; the sharp 2√2 for general planes is Conjecture 1. That said, the authors are honest about this—b156 and b179 state the limitation plainly, and the Outlook says the Svetlichny corollary is proven only under the hypothesis. Fourteen multi-start optimizations all return 2.828427, so the conjecture is plausible, but it is unproven. The abstract should say "under local bilinear readout" rather than "never". The proven result for the physically relevant class—dual-rail and tetron qubits, where the compiled measurements satisfy the hypothesis exactly—stands. Reproducibility is capped until code and data are deposited (promised, not released), which is minor but worth noting.

Bottom line: careful, honest, and a real result. Send it to a serious referee; the general-plane conjecture and the abstract's scope wording are the two things to press on. I would cite it.

Referee report

Significance

Under its measurement hypothesis the mathematics is sound and derived, not fitted: the vacuum-full inequality P(n)P(nbar) <= 2^-M (with tightening for odd M), Lemma 4's inequality on V(epsilon), and the Cauchy-Binet argument for Theorem 3 all check out, and the constraints 2*sqrt(2), sqrt(2), and beta are derived. The manuscript ships named numerical scripts with cross-checks at 10^-6 to 10^-16 and discloses superseded runs and an optimizer-budget artifact, which supports trust in the numerics. The experiment is careful: raw values only, Eq. (10) gives a corrected post-selection threshold that the flagged data do not pass, and the monogamy statistic uses configuration-level standard errors. The identification of GMNL with fermionic magic, if the unrestricted version holds, would be a notable unification of the nonlocality and simulability boundaries. The principal weakness is scope: the abstract-level claim that free fermions 'are never genuinely multipartite nonlocal' is proven only for the direction-sharing/commuting measurement class, and for the general-bilinear class the certified constant is too weak to exclude Svetlichny violations.

Major objections (3)

[Abstract; b3; Theorem 1 (b26)] The abstract states unconditionally that free fermions 'are never genuinely multipartite nonlocal' and that GMNL 'cannot arise from any mean-field state'. Theorem 1 proves |<QAQBQC>| <= 2*sqrt(2) only under the hypothesis that each party's two planes share a direction or commute; the class of general bilinears over four or more Majoranas is explicitly outside the hypothesis (b35). For that class SM Sec. 6 certifies only 3.26 by interval branch-and-bound, with numerical relaxation 3.12. Since |S3| = sqrt(2)|<QAQBQC>|, the certified 3.26 permits |S3| ~ 4.61, above Svetlichny's hybrid-local value 4, so a genuinely tripartite nonlocal Gaussian correlation is not excluded for general bilinear readout. The authors state this honestly (b156, b179, Outlook) and make the sharp general-plane value Conjecture 1, supported only by direct searches (Table S1; every run returns 2.828427). Because this is the paper's headline claim, the abstract and b3 must be restricted to the hypothesis class (or to the dual-rail/encoded setting where the hypothesis is automatic, b34) unless Conjecture 1 is proved at least for n = 3. The authors acknowledge this: “The sharp constant, and hence the Svetlichny corollary, is proven only under the hypothesis of Theorem 1 of the main text.”
[Theorem 3; b45] The n-party statement that free fermions 'are therefore never genuinely multipartite nonlocal for any number of parties' inherits the same restriction. The proof requires the factorization Qj = i(uj*c)(lj*c), which exists only when the two settings share a direction (the symmetric parts of the Wick matrix vanish only then); for arbitrary planes the Pfaffian identity fails (b179) and Conjecture 1 is verified only numerically to n = 5 (SM Sec. 6, to 10^-6). The n-party Svetlichny bound |Sn| <= sqrt(2)*2^(n/2) <= 2^(n-1) therefore also holds only for direction-sharing settings. Please attach the hypothesis to the n-party claims in the main text in the same way Theorem 3's own statement does, and mark the 'never genuinely multipartite nonlocal for any number of parties' sentence as conditional on it.
[Setting paragraph; Table S1 footnote; b350] The main text states that letting each party read all its occupation numbers with any of the 16 Boolean post-processings per setting 'gives no improvement (SM)'. The Supplement concedes these E1 searches are local (Table S1 footnote: 'The E1 searches are local'; b350: 'They are not a global search over class E1, which has 84 continuous and 24 Boolean parameters at k=2'). As written in the main text the sentence reads as an exhaustive claim. Please flag the locality of the E1 evidence at the point of citation, since a reader relying on the main text would not know the search is not global. The authors acknowledge this: “They are not a global search over class E1, which has 84 continuous and 24 Boolean parameters atk= 2.”

Minor objections (4)

[SM Sec. 11.1 (ibm berlin)] The ibm berlin scan fit has chi^2/dof = 3.1 with a monotone residual trend that a constant offset of -0.04 largely absorbs; consider showing this trend in a figure or noting it in the main text next to Fig. 2, which currently displays only the ibm fez residuals.
[Main text, 'Against the setting-agnostic bound'] For ibm berlin the eight-cell joint value 0.457 falls 0.005 short of the certified setting-agnostic bound beta'_8 = 0.462 while clearing the numerical 0.394 (SM Sec. 11.1). The main text quotes 0.457 only against beta^8 = 0.282; please state in the main text that the certified setting-agnostic monogamy bound is not cleared on berlin.
[Main text proof sketch of Theorem 1] The notation V(1) and V(2) for the two largest classes of the sign vector epsilon is easy to misread as numeric evaluation of V; consider writing V^(1), V^(2) to match the definition V(epsilon) with epsilon in {+-,}^3.
[SM Sec. 7 (extremal sets)] The statement that for obtuse angles relabeling conjugates Qp and the bound is unchanged, with attainment verified by direct maximization, is found only in the SM; a one-sentence note in the main text after Theorem 2 would help, since Theorem 2 as stated covers only phi_p in [0,pi/2].

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

Free-fermion states are never genuinely multipartite nonlocal under parity-preserving local bilinear readout, and any violation certifies the non-Gaussian 'magic' that makes fermionic computation universal.

The reading

Free-fermion systems are efficiently classically simulable, but a single non-Gaussian state turns them into universal quantum computers. This paper shows the same dividing line appears in correlations: under local bilinear measurements whose two settings share a direction or commute, every fermionic Gaussian state, however many parties share the fermions, carries no more Bell nonlocality than one entangled pair and is never genuinely multipartite nonlocal. The three-party Mermin correlator is at most $2\sqrt{2}$, the Svetlichny polynomial at most $4$, and the $n$-party product correlator at most $2^{n/2}$. Since every pure non-Gaussian state is known to make matchgate circuits universal, exceeding those bounds is a statistics-only certificate of fermionic magic. The paper also reports hardware certification, including eight disjoint six-qubit cells on 48 qubits that all certify and jointly win with probability $0.503$ against a Gaussian ceiling of $0.282$.

What carries the argument

The carrying object is the covariance matrix $\Gamma$ of a fermionic Gaussian state together with the Pfaffian, or Wick, formula for correlators of Majorana linear forms. The proof compresses the state to two modes per party, introduces the vacuum-full inequality $P(n)P(\bar n)\le 2^{-M}$, which follows from the doubly degenerate spectrum of $S=\Gamma\Gamma_n$ and tightens when $M$ is odd, and decomposes each bilinear setting into raising and lowering operators inside fixed fermion-parity sectors; a short linear program over pair weights then yields $2\sqrt{2}$. The $n$-party law is carried by the Pfaffian identity $\left|\operatorname{Pf}(W^\top\Gamma W)\right|^2\le\det(W^\dagger W)=\prod_j\|\ell_j\|^2=2^n$, so each complex measurement form contributes at most norm $\sqrt{2}$. The named objects are the Mermin polynomial $M_n=\operatorname{Re}\prod_j Q_j$, whose violation witnesses genuine multipartite nonlocality, and the Svetlichny polynomial $S_3=\sqrt{2}\operatorname{Re}(e^{-i\pi/4}Q_AQ_BQ_C)$, whose hybrid-local bound is $4$.

Extended reading notes

Core claim

The paper proves Theorem 1: for any fermionic Gaussian state, a free-fermion state fixed by a covariance matrix, and any local bilinear settings in which each party's two planes share a direction or commute, $\left|\left\langle Q_AQ_BQ_C\right\rangle\right|\le 2\sqrt{2}$, with the bound attained; the three-party Mermin polynomial is therefore capped at its biseparable value and the Svetlichny polynomial at its hybrid-local value. Theorem 3 extends the cap to every number of parties: $\left|\left\langle\prod_{j=1}^n Q_j\right\rangle\right|\le 2^{n/2}$, also attained, so free fermions violate the stronger of the two Mermin-family inequalities by exactly $\sqrt{2}$, the violation of one Tsirelson pair, and never the other. The mechanism is fermion parity: Gaussian correlators are Pfaffians, and the inequality $P(n)P(\bar n)\le 2^{-M}$ between occupation and complementary-occupation probabilities blocks the three-party paradox. Combined with the cited result that every pure non-Gaussian state makes matchgate circuits universal, the paper concludes that genuine multipartite nonlocality is a device-independent signature of fermionic magic, accessible from the outcome table alone under the covered readouts.

Load-bearing premise

The main bound stands on the measurement hypothesis that each party's two settings are single Majorana bilinears whose planes share a direction or commute, which is automatic for dual-rail and tetron qubit encodings; if parties instead measure general bilinears over four or more Majoranas, the sharp $2\sqrt{2}$ bound and the Svetlichny consequence are only Conjecture 1, with a certified relaxation $3.26$ that is too weak to exclude a Svetlichny violation.

Editorial extensions

If this is right

Any Mermin value above $2\sqrt{2}$ under the covered readouts rules out every Gaussian state, pure or mixed, on any number of modes, from the outcome statistics alone, which makes fermionic magic a device-independently certifiable resource.
Mean-field and free-fermion matter, including BCS superconductors, Kitaev chains, Slater determinants, and Hartree-Fock states, can never produce genuine multipartite nonlocality under such readouts, so observing it is evidence of interactions beyond mean field.
The chaining theorem gives a free-fermion counterpart of the Clifford-magic shallow-circuit separation: $m$ disjoint cells win jointly with probability at most $\beta^m$ for Gaussian states, with $\beta=(2+\sqrt{2})/4$, while adding one non-Gaussian gate angle solves the corresponding relational problem with certainty.
The multipartite correlation power of free fermions is exactly one Tsirelson pair: the violation of the stronger Mermin-family inequality is $\sqrt{2}$ for every $n$, and the other member of the pair is never violated.
The protocol certifies on current hardware: eight disjoint six-qubit cells on 48 qubits all certify individually and jointly win with probability about $0.503$, against $0.282$ allowed to Gaussian strategies.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the open general-plane conjecture, sharp $2\sqrt{2}$ for arbitrary bilinear settings, is true, then the Svetlichny conclusion would hold without the shared-direction hypothesis, making genuine multipartite nonlocality an unconditional witness for all parity-preserving local bilinears; closing the gap between the certified $3.26$ and the numerical $2.828$ is the clear next test.
The same parity logic might extend to the full fermionic readout class, including readout of all occupation numbers with Boolean post-processing; the paper's evidence there is numerical and local, so a proof in that setting would make the resource claim a theorem for the entire class.
The monogamous cell construction reads like a distributed hardware benchmark: per-cell acceptance flags tracked degraded chip regions with correlations above $+0.9$ on two processors, suggesting that correlation statistics alone could serve as chip-quality metrology.
More broadly, if nonlocality and universal fermionic computation share one resource, the violation size could become a quantitative input for verifying fermionic simulators, in the spirit of using magic-state measures to quantify overhead in other simulation frameworks.

Original abstract

Free fermion dynamics is classically simulable, and a single non-Gaussian state makes it universal. We show that the same boundary governs quantum nonlocality. Measured as fermionic qubits, fermionic Gaussian states reach at most the biseparable bound of the Mermin inequality for three parties and, for any number of parties, violate the Mermin family by exactly the factor of a single entangled pair. They are never genuinely multipartite nonlocal. Genuine multipartite nonlocality and computational universality are thus one resource for fermions, the opposite of the stabilizer case. The mechanism is fermion parity. The bound is monogamous and certifies non-Gaussianity from measured correlations alone. On superconducting processors we watch nonlocality beyond free fermions switch on under an interaction between densities, confirm the monogamy law on 96 qubits, and measure the ground state of twelve interacting fermions on a ring cross the bound as the interaction grows, detecting the charge density wave transition.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-01: 1 request, 4,221 input tokens.

What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.23 · human participants 0.02 · controlled comparison 0.11 · randomized assignment 0.02 · states quantitative result 0.96 · compares with prior results 0.43 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

A system of free fermions can be simulated on a classical computer, and a single non-Gaussian state makes it a universal quantum computer. We show that the same line bounds quantum nonlocality. Measured as fermionic qubits, free fermions carry the Bell nonlocality of one entangled pair, no more, however many parties share them, and are never genuinely multipartite nonlocal. Fermion parity is the reason. Genuine multipartite nonlocality is thus the resource of universal fermionic computation, cannot arise from any mean-field state, and we certify it on $48$ qubits of IBM processors.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	A system of free fermions can be simulated on a classical computer, and a single non-Gaussian state makes it a universal quantum computer.	background 1.00	0.11	0.08	asserted 1.00
s1	We show that the same line bounds quantum nonlocality.	result 0.53	0.82	0.19	asserted 0.99
s2	Measured as fermionic qubits, free fermions carry the Bell nonlocality of one entangled pair, no more, however many parties share them, and are never genuinely multipartite nonlocal.	result 1.00	0.84	0.54	asserted 1.00
s3	Fermion parity is the reason.	interpretation 0.96	0.49	0.07	asserted 1.00
s4	Genuine multipartite nonlocality is thus the resource of universal fermionic computation, cannot arise from any mean-field state, and we certify it on $48$ qubits of IBM processors.	result 0.46	0.92	0.77	asserted 1.00

What kind of paper the classifier reports this is: theoretical result 0.99 · evidence from proof or derivation 0.58.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures

**Figure 1.** FIG. 1. The cell, three dual-rail logical qubits on six physical

Figures reproduced from arXiv:2609.39411 by the authors.

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.