{"id":"20776ed0-d25f-445c-9bfc-7214d2aafb2d","arxiv_id":"2610.38716","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A collective bootstrap built from virial constraints, Schur complements, and loop product lifting produces SDP lower bounds and close numerical estimates for planar ground states of multi-matrix quantum mechanics.","lead":"Large N matrix quantum mechanics is reformulated as a semidefinite program in the space of coordinate loop expectation values, giving rigorous energy lower bounds and numerical ground state estimates for one, two, and three matrix models. The paper's value is a new numerical framework that may extend to the bosonic sectors of BFSS and BMN matrix models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The three-matrix numerical claim rests on an unverified S=3 support cage: without it the search leaves the physical region, so the reported 0.022% agreement is not a test of the collective bootstrap.","rationale":"The reader's verdict already identifies the S=3 cage as the weakest assumption, and this pass confirms it as the load-bearing point while sharpening why it matters. The central lower-bound statement (3.23) is well protected: the exact solution is feasible at every truncation, the Schur-complement step replaces an inverse by a positive semidefinite block without shrinking the physical feasible set, and dropping the high sector only removes a nonnegative correction. No internal inconsistency was found there. The one-matrix and decoupled two-matrix benchmarks are credible tests of the machinery. The problem is the three-matrix estimate: the final fit is obtained under an artificial cage, the support assumption behind it is unchecked, and the uncaged search demonstrates that the imposed constraints allow configurations with ϕ2≈31 and ϕ3≈30. Stability over S=3,4,8 does not imply correctness, since S=16 changes the result. The omitted higher dynamical SD equations, acknowledged in the Discussion, are the natural place from which support control should come. The proposed check—either the independent spectral edge of R or the addition of the first non-virial SD equations—would settle whether the cage is benign or load-bearing. The reader's CONDITIONAL verdict remains appropriate; no change is required.","tokens_in":32253,"tokens_out":13676,"duration_ms":109192,"concrete_test":"Compute the spectral edge of R = Σ_a M_a^2 at the master-field ground state of Ref. [19], either from that paper's finite-matrix ansatz or by rerunning its master-field optimization, and compare it with S=3: if the edge is at or above 3, the S=3 cage excludes the physical solution and the Table 5 fit is not a controlled estimate. As a cross-check, rerun the three-matrix search at ℓmax=7 without cages but with the first non-virial dynamical SD equations (the multi-matrix analog of §B.2) added; if uncaged rank-one points with ϕ2≈31 and ϕ3≈30 survive, the cages are supplying dynamical information that the truncated bootstrap itself lacks.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The inequality chain (3.23) is internally sound: the exact planar solution satisfies every truncated constraint, so the rank relaxation and the rank-one constraint place E0 above both ER and EP. The load-bearing weakness is in the numerical estimate presented as the three-matrix test. Section 7 imposes cages ⟨ϕ(R^{n+1})⟩ < S⟨ϕ(R^n)⟩, and the final fits use S=3. As the text admits ([b269]-[b270]), this is equivalent to assuming the spectral edge of R lies below S, and that support statement is not derived anywhere. This is not merely a missing rigor certificate: Table 5 shows that without cages the ℓmax=7 search exits with ϕ2≈31 and ϕ3≈30, so the reported agreement Eg=1.3672 and θ≈0.7895 is produced by the cage, not by the collective bootstrap constraints. The S=16 row shows the result is cage-sensitive once the cage is loose; stability among S=3,4,8 does not by itself certify S=3. Moreover, the method imposes only the virial equation among the dynamical SD equations (Discussion, [b288]); the uncaged direction (large ϕ2, ϕ3 with ϕ2−ϕ3 finite) is a concrete nonphysical configuration allowed by the truncated constraints. The missing physical input is exactly the support bound behind the cage.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper develops a collective bootstrap for large-N matrix quantum mechanics, formulated directly in the space of single-trace coordinate moments. The loop joining and splitting operations organize the kinetic Schwinger-Dyson constraints, the virial theorem together with a Schur complement eliminates the inverse joining matrix, and a lifted matrix Y relaxes quadratic products of loops. The resulting semidefinite program yields a lower bound on the planar ground-state energy at every truncation; a sequential log-det rank-minimization heuristic searches for factorized rank-one points. The framework is benchmarked on the one-matrix model, the decoupled and coupled two-matrix quartic models, and two- and three-matrix Yang-Mills models. The one-matrix and decoupled two-matrix comparisons are excellent, and the two-matrix Yang-Mills results agree well with master-field optimization. The three-matrix comparison, however, is obtained with additional 'cage' constraints whose spectral-support assumption is not independently established.","tokens_in":32657,"tokens_out":10258,"duration_ms":73967,"significance":"If the numerical results hold, the method is a valuable complement to phase-space matrix bootstrap and master-field optimization: it avoids momentum-containing words, permits longer word truncations, and gives a transparent relaxation hierarchy with a rigorous lower-bound statement. The formal core of the paper, inequality (3.23), is carefully argued and the paper is commendably explicit in Section 3.2.4 about what is rigorous and what is heuristic. The one-matrix benchmark (six-digit energy agreement), the free-independence structure of the decoupled two-matrix model, the exact rescaling predictions for massless Yang-Mills exponents, and the Hellmann-Feynman consistency check are genuine positive results. The main limitation is the three-matrix application, where the reported 0.022% agreement is not a test of the collective-bootstrap constraints alone but of those constraints supplemented by an unverified cage with S=3.","major_comments":[{"comment":"The headline three-matrix result is not produced by the collective-bootstrap constraints alone. The final estimates (7.8)-(7.12) use the cages \\langle\\phi(R^{n+1})\\rangle < S\\langle\\phi(R^n)\\rangle with S=3, and the paper itself states ([b269]-[b270]) that the cage is equivalent to assuming the spectral edge of R lies below S, an assumption that is not derived. The uncaged search (\\langle\\phi_2\\rangle\\approx 31, \\langle\\phi_3\\rangle\\approx 30) and the S=16 row (E_g=1.34970241, with loop values changed by 10-20%) show that the truncated constraints alone do not confine the search to the physical region and that the result is cage-sensitive once the cage is loose. The stability among S=3,4,8 does not by itself certify that S=3 exceeds the true spectral edge, and the manuscript does not explain how S=3 was chosen. Because this is the only three-matrix benchmark, the paper should either provide an independent argument for the support bound or explicitly reclassify the three-matrix numbers as conditional estimates under an unverified spectral-support assumption.","section":"Section 7, Table 5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The unqualified claim that the framework \"exhibits close agreement with analytic solutions and master field optimization\" is too strong for the three-matrix model. For the one-matrix and decoupled two-matrix models the agreement is a genuine test of the constraints, but for the three-matrix model it tests the combination of constraints plus the artificial cage. The abstract and introduction should carry a qualifier such as \"under an assumed spectral-edge bound\" for the three-matrix case, and the 0.022% agreement should not be presented as a direct validation of the collective bootstrap until the cage assumption is justified.","section":"Abstract and Section 1","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The bisection tolerance \\varepsilon_E is not reported for the three-matrix runs, and the rank tolerance is only stated generically in Section 6.3. Without these values, the eight-decimal entries in Table 5 overstate the numerical precision and hinder reproducibility.","section":"Algorithm 2 and Table 5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The horizontal axis of Figure 12 lists loop indices 1, 3, 6, 13, 26, 61, 146 without explaining the enumeration or ordering; a short note or legend mapping these indices to representative words would make the scaling-exponent plot much easier to interpret.","section":"Figure 12","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The two-matrix massless angle \\theta\\approx 0.671 in (6.22) is reported without comparison to a previous value, although the analogous three-matrix angle is compared with the master-field result; adding the available comparison value, if one exists, would strengthen the benchmark.","section":"Section 6.3.1","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision","confidential_remarks":"The central formal claim (3.23) appears sound and is presented with unusual candor about its heuristic parts; I would not reject the paper. The main issue is that the three-matrix application, which is a headline result in the abstract, rests on the unverified spectral-support cage. This is fixable either by supplying an independent support bound or by reframing the three-matrix section as a conditional demonstration. The citation practice looks appropriate and I have no novelty concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Colleague,\n\nYou should know two things before reading. The central inequality chain (3.23) is sound, and the paper is unusually explicit about which of its numbers are rigorous and which are heuristic. But the headline three-matrix result, the 0.022% agreement with master field optimization, is not produced by the collective bootstrap constraints; it is produced by a cage constraint whose physical justification is assumed, not established.\n\nWhat is actually new is the assembly: coordinate-only loop variables that eliminate momentum words, the virial theorem plus a Schur complement that trades the inverse joining matrix for a PSD block, Kazakov-Zheng lifting for the quadratic splitting, and log-det rank minimization to seek factorized points. Each ingredient exists in the literature, but the combination does not, and the lower-bound logic is internally consistent: the exact planar solution satisfies every truncated constraint, each relaxation enlarges the feasible set, and Section 3.2.4 correctly states that ER, EP, and Eg are not all ordered. The paper earns real credit there.\n\nThe benchmarks are genuine. One-matrix energies agree with the analytic solution to six digits across truncations and couplings, including the approach to the critical coupling. The decoupled two-matrix model reproduces free-independence factorization. The massless two-matrix scaling exponents are fixed by exact rescaling rather than fitted, and the massive-coefficient check against Hellmann-Feynman is a legitimate consistency test.\n\nThe soft spot is Section 7, and it is substantial. The paper does acknowledge that the cages are artificial and equivalent to assuming the spectral edge of R lies below S, but that acknowledgment does not neutralize the consequence. Table 5 shows the uncaged search exiting at phi2 around 31 and phi3 around 30, an order of magnitude above the master field values. So the reported Eg = 1.3672 and theta = 0.7895 are produced by the cage, not by the bootstrap constraints, and the S = 16 row shows the result is cage-sensitive once the cage is loose. The paper also admits that only the virial equation among the dynamical SD equations is imposed; that is exactly why the large-phi directions remain allowed. These are acknowledged limitations, but they mean the three-matrix section reads as an exploratory case study, not a validation. Minor point: no code or data is shipped, so the tables are not reproducible from the text.\n\nNone of this breaks the lower-bound construction, and the one- and two-matrix parts stand on their own. This paper deserves a serious referee; the method is a plausible route to longer-word truncations and eventually to BFSS/BMN bosonic sectors. The referee should ask for a derivation or independent certification of the support bound behind the cage, or a three-matrix presentation that reports cage dependence openly rather than as a fitted number. I would cite it for the SDP formulation and the inequality chain.","headline":"The lower-bound construction is sound and honestly labeled, but the three-matrix 'agreement' is produced by a cage constraint whose physical input is assumed, not derived.","tokens_in":33126,"tokens_out":3258,"would_cite":true,"duration_ms":22068,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["11.15.Pg","02.60.Pn"],"model":"deepseek/deepseek-v4-flash","headline":"Loop-space bootstrap certifies a lower bound on the large-N ground state at every truncation, matching one-matrix energies to six digits.","keywords":["matrix quantum mechanics","large N limit","collective bootstrap","semidefinite programming","loop space","master field","Yang-Mills matrix models","rank minimization"],"falsifier":"Evaluate the spectral edge of $R=\\sum_a M_a^2$ at the master-field ground state of the three-matrix model, or at bootstrap solutions with larger $S$ where results still shift, and compare it with $S=3$; if the edge exceeds 3, the cage used for the headline fits excludes the physical solution. A complementary check is the uncaged collective bootstrap at truncations beyond $\\ell_{\\max}=7$: if the runaway directions with $\\langle\\phi^2\\rangle\\approx 31$ and $\\langle\\phi^3\\rangle\\approx 30$ persist while only cage-dependent fits converge, the three-matrix estimates are cage artifacts.","tokens_in":33737,"feed_emoji":"🧮","tokens_out":17379,"duration_ms":110082,"temperature":0.7,"pith_summary":"This paper aims to make the planar (large-$N$) ground state of matrix quantum mechanics computable with certified error control. It formulates the problem directly in the space of single-trace coordinate moments and recasts the collective energy minimization as a semidefinite program (a convex optimization whose constraints are matrix-positivity conditions), claiming that every truncation yields a rigorous lower bound on the exact ground state energy divided by $N^2$. A sequential rank-minimization search then estimates the energy and loop expectation values by driving the lifted variables toward planar factorization, the property $\\langle\\phi(C)\\phi(C')\\rangle = \\langle\\phi(C)\\rangle\\langle\\phi(C')\\rangle$ that characterizes a classical large-$N$ configuration. The reported energies agree with the exact free-fermion solution to six digits in the one-matrix model and with master-field optimization to a few parts in a thousand in two- and three-matrix Yang–Mills models. Because momentum-containing words never appear as variables, longer word-length truncations become affordable, which the paper argues opens a route to the bosonic sectors of the BFSS and BMN models.","feed_headline":"Matrix ground states pinned down with certified lower bounds","feed_subtitle":"A loop-space SDP matches exact one-matrix energies and master-field Yang–Mills results.","key_machinery":"Four objects carry the argument. The loop space of single-trace coordinate moments, $\\phi(C) = N^{-\\ell(C)/2-1}\\,\\mathrm{tr}(C)$, replaces the phase-space operator basis so that momentum words never appear as independent variables. Loop joining, $\\Omega$, and loop splitting, $\\omega$, from the collective field formulation organize the kinetic constraints. The virial theorem together with a Schur complement replaces the inverse joining matrix $\\Omega^{-1}$ by the positive-semidefinite block condition $S = \\begin{pmatrix} \\Omega & \\omega^\\dagger \\\\ \\omega^T & \\mu \\end{pmatrix} \\succeq 0$, making the energy linear in the loops. Quadratic products of loops in the splitting term are lifted to auxiliary variables $Y(C,C') = \\phi(C)^\\dagger\\phi(C')$ with $Y \\succeq 0$, and a physical configuration is recovered when $\\mathrm{rank}(Y)=1$, which is the planar factorization condition. The relaxed program with the rank condition dropped produces the certified lower bound $E_R(\\ell_{\\max})$, while the log-det heuristic and energy bisection search for the rank-one point that yields the estimate $E_g$.","core_discovery":"The central claim is inequality (3.23): at every truncation $\\ell_{\\max}$, the optimum $E_R(\\ell_{\\max})$ of the relaxed semidefinite program and the optimum $E_P(\\ell_{\\max})$ of the rank-constrained problem satisfy $E_R(\\ell_{\\max}) \\le E_P(\\ell_{\\max}) \\le E_0$, where $E_0$ is the exact large-$N$ ground state energy divided by $N^2$. Each truncated computation is therefore a genuine lower bound, and both optima are nondecreasing in $\\ell_{\\max}$, so the bound tightens monotonically toward $E_0$. The numerical estimate $E_g$ is produced by bisecting in the energy and using a log-det rank-minimization heuristic to test whether the feasible region contains a point whose lifted matrix $Y$ has rank one, which is exactly the planar factorization condition; the paper acknowledges that this search is a local heuristic, although every accepted energy level is certified by an approximate rank-one point. On this basis the reported relative energy deviations run from below $10^{-6}$ in the one-matrix model to a few times $10^{-3}$ in the Yang–Mills models, with the three-matrix commutator angle $\\theta \\approx 0.7895$ against a master-field value of $0.791(2)$. The paper emphasizes that the rigorous bound relies only on constraints satisfied by the exact solution, while the three-matrix estimates further depend on artificial cage constraints stated in Section 7.","pith_inferences":["The three-matrix numbers are only as strong as the cage assumption: without $\\langle\\phi(R^{n+1})\\rangle < 3\\langle\\phi(R^n)\\rangle$ the search yields $\\langle\\phi^2\\rangle \\approx 31$ and $\\langle\\phi^3\\rangle \\approx 30$, an order of magnitude above master-field values, so the 0.022% energy agreement partly measures the cage's influence; computing the spectral edge of $R$ at the master-field sol","The accuracy asymmetry visible in the tables, energies at $10^{-6}$ to $10^{-3}$ relative error while long loops sit at the percent level, suggests that adding the higher dynamical equations would improve loop values before the energy; this is the paper's own suggested next step, and my reading is that the energy accuracy is carried almost entirely by the virial constraint.","The certified lower bound $E_R$ and the heuristic estimate $E_g$ are logically independent outputs: a failure of the rank-one search raises $E_g$ but never breaks the bound, so the two numbers should carry different trust levels in future applications.","The quadratic growth of lifted variables with the number of short loops is the practical bottleneck toward nine-matrix models; the symmetry reductions used here suggest that flavor and rotation symmetry, not loop counting alone, will decide whether BFSS is reachable."],"forward_implications":["Every truncated computation returns a certified lower bound on the planar ground state energy, and the bound is nondecreasing in the truncation length, so systematic improvement is guaranteed rather than assumed.","Because only coordinate moments appear, the single-trace variable count grows like $d^\\ell$ instead of $(2d)^\\ell$, making longer word lengths and more matrices affordable than in phase-space bootstraps.","The decoupled two-matrix benchmark reproduces the free-independence factorization rules of the large-$N$ limit, confirming that the rank-one search finds the correct factorization structure.","At a factorized solution the collective Hamiltonian determines the $O(1)$ fluctuation spectrum around the large-$N$ background, so the same computation can yield excitation energies alongside the ground state.","The framework is aimed at the bosonic sectors of the BFSS and BMN nine-matrix models, where the coordinate-only formulation is expected to control the growth of variables."],"supporting_citations":[{"why":"Supplies the collective field Hamiltonian, the loop variables, and the collective potential whose inverse joining matrix the bootstrap removes.","marker":"[21, 22]"},{"why":"The phase-space matrix bootstrap whose stationarity, positivity, and symmetry constraints are the starting point being reorganized.","marker":"[27]"},{"why":"Introduces the semidefinite relaxation that lifts quadratic products of loops to a positive-semidefinite matrix, the core relaxation step.","marker":"[26]"},{"why":"Provides the log-det heuristic that the sequential SDP uses to drive the lifted matrix toward rank one.","marker":"[35]"},{"why":"Gives the exact free-fermion solution of the one-matrix model used as the six-digit benchmark.","marker":"[2]"},{"why":"Master-field optimization for multi-matrix systems, the benchmark for the coupled two-matrix model and the source of the finite-matrix parametrization.","marker":"[17]"},{"why":"Two-matrix Yang–Mills master-field results, including the energy coefficient 0.8890(2), that the bootstrap matches.","marker":"[18]"},{"why":"Three-matrix Yang–Mills master-field results, including the angle 0.791(2), that the bootstrap reproduces.","marker":"[19]"}],"fun_headline_variants":["Collective bootstrap bounds matrix ground states from below","Loop-space SDP certifies planar ground state energy bounds","Matrix quantum mechanics ground states via loop-space semidefinite bounds","Bootstrap lower bounds tighten toward exact matrix ground state","Collective bootstrap in loop space pins matrix energies from below"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The load-bearing premise behind the three-matrix estimates is that the true eigenvalue distribution of $R=\\sum_a M_a^2$ is supported below the artificial cage bound $S=3$; the paper acknowledges the cages are artificial and that the bound (3.23) remains valid only if $S$ exceeds the true spectral edge, a condition the paper does not independently verify, and if the edge lies above $S$ the cage excludes the physical solution and the reported agreement with master-field values does not test the method.","fun_headline_variants_meta":{"raw":{"variants":["Collective bootstrap bounds matrix ground states from below","Loop-space SDP certifies planar ground state energy bounds","Matrix quantum mechanics ground states via loop-space semidefinite bounds","Bootstrap lower bounds tighten toward exact matrix ground state","Collective bootstrap in loop space pins matrix energies from below"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":8.3e-05,"raw_usage":{"cost":8.34652e-05,"is_byok":false,"cost_details":{"upstream_inference_cost":8.34652e-05,"upstream_inference_prompt_cost":2.77732e-05,"upstream_inference_completions_cost":5.5692e-05},"total_tokens":1396,"prompt_tokens":1090,"completion_tokens":306,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":981,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":240}},"tokens_in":1090,"tokens_out":306,"cost_source":"openrouter_usage","duration_ms":5241,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":240,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-31T04:11:04.720580+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"falsifier":"Evaluate the spectral edge of $R=\\sum_a M_a^2$ at the master-field ground state of the three-matrix model, or at bootstrap solutions with larger $S$ where results still shift, and compare it with $S=3$; if the edge exceeds 3, the cage used for the headline fits excludes the physical solution. A complementary check is the uncaged collective bootstrap at truncations beyond $\\ell_{\\max}=7$: if the runaway directions with $\\langle\\phi^2\\rangle\\approx 31$ and $\\langle\\phi^3\\rangle\\approx 30$ persist while only cage-dependent fits converge, the three-matrix estimates are cage artifacts.","supporting_citations":[{"cited_title":"Bootstrapping Matrix Quantum Mechanics","cited_arxiv_id":"2004.10212","evidence_quote":"The phase-space matrix bootstrap whose stationarity, positivity, and symmetry constraints are the starting point being reorganized."},{"cited_title":"Analytic and Numerical Bootstrap for One-Matrix Model and \"Unsolvable\" Two-Matrix Model","cited_arxiv_id":"2108.04830","evidence_quote":"Introduces the semidefinite relaxation that lifts quadratic products of loops to a positive-semidefinite matrix, the core relaxation step."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the log-det heuristic that the sequential SDP uses to drive the lifted matrix toward rank one."},{"cited_title":"Brezin, C","cited_arxiv_id":null,"evidence_quote":"Gives the exact free-fermion solution of the one-matrix model used as the six-digit benchmark."},{"cited_title":"Large N Optimization for multi-matrix systems","cited_arxiv_id":"2108.08803","evidence_quote":"Master-field optimization for multi-matrix systems, the benchmark for the coupled two-matrix model and the source of the finite-matrix parametrization."},{"cited_title":"Large N Master Field Optimization: the Quantum Mechanics of two Yang-Mills coupled Matrices","cited_arxiv_id":"2306.00935","evidence_quote":"Two-matrix Yang–Mills master-field results, including the energy coefficient 0.8890(2), that the bootstrap matches."},{"cited_title":"\"Glueballs\" in the Quantum Mechanics of three large massless Yang-Mills coupled Matrices","cited_arxiv_id":"2503.05304","evidence_quote":"Three-matrix Yang–Mills master-field results, including the angle 0.791(2), that the bootstrap reproduces."}],"review_version":1}