{"id":"420ed5f9-988c-4406-b7f3-ba396904da5e","arxiv_id":"1908.03574","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The velocity-dependent Lyapunov exponent inside the butterfly cone satisfies λ(v) ≤ 2πT(1-|v|/v_B), a generalization of the chaos bound, saturated in SYK chains, holographic theories, and large N CFTs.","lead":"This paper proves a bound on how fast chaos can grow along each direction of a spreading 'butterfly cone' in a quantum system: the growth rate at velocity v cannot exceed 2πT(1-|v|/vB). It generalizes the Maldacena-Shenker-Stanford chaos bound to moving frames and is saturated in several strongly coupled models.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The VDLE bound is conditional on the ray ansatz (1.5), which the paper does not derive from microscopic dynamics; without it, (1.8) has no well-defined object.","rationale":"I read the paper as aiming to establish a universal upper bound on the velocity-dependent Lyapunov exponent inside the butterfly cone. The derivation is mathematically clean once the ansatz (1.5) is granted: the fixed-x MSS bound gives (1.7), and the integration with λ(v_B)=0 gives (1.8). The examples in Section 3 and the Regge analysis in Section 4 are consistent with the bound and illustrate saturation above a critical velocity. The load-bearing weakness is that (1.5) is an assumption rather than a consequence of the MSS axioms; the paper is transparent about this, and the reader's conditional verdict already reflects it. I do not see an internal inconsistency in the derivation itself, nor a reason to reject the bound for systems in which a well-defined, differentiable VDLE exists. My independent read agrees with the reader's weakest assumption, and the conditional verdict should stand unchanged.","tokens_in":28066,"tokens_out":17065,"duration_ms":192690,"concrete_test":"Independently re-derive the bound (1.8) starting from the fixed-x MSS bound (1.6) and the definition of λ(v) in (1.4) alone, without postulating the exponential ray form (1.5). Concretely, attempt to prove |λ(v)-v·∇λ(v)|≤2π/β using only the existence of the ray limit lim_{t→∞}(1/t) log(1/(1-f(t,vt))). If the proof requires additional control of ∂_x f beyond what (1.4) provides, then the derivation of (1.8) is established only under (1.5), and the concern that the VDLE ansatz is load-bearing is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central result, Eq. (1.8), is derived in Section 2.2 by substituting the VDLE ansatz (1.5), f(t,x)=1-ε e^{λ(x/t)t}, into the fixed-x MSS bound (1.6). This substitution is what converts the bound into |λ(v)-v·∇λ(v)|≤2π/β and, with λ(v_B)=0, into (1.8). The ansatz is introduced in the Introduction as 'it seems natural to propose' and is not derived from microscopic dynamics; the paper explicitly relies on it as the definition of λ(v) for every velocity inside the butterfly cone. If a local system's OTOC does not organize into smooth exponential rays—for example, because the front is broadened or diffusive, because λ(v) is not differentiable, or because multiple saddles contribute—then ∂_t f/(1-f) is not equal to λ(v)-v·∇λ(v), and the bound (1.8) does not follow. The Regge-theory discussion in Section 4.2 supplies a microscopic justification only for large-N CFTs, and that justification itself assumes the leading Regge pole dominates and, for the lower bound on v*, the convexity of the large-N Regge trajectory, an assumption the paper explicitly states is not proven (Section 4.2, near Eq. (4.22)). Thus the central universal claim is conditional on an unproven structural assumption, not a direct consequence of the MSS bound alone.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives an upper bound on the velocity-dependent Lyapunov exponent lambda(v) inside the butterfly cone, lambda(v) <= (2*pi/beta)(1 - |v|/v_B), by applying the Maldacena-Shenker-Stanford chaos bound to a ray-ansatz form for the out-of-time-ordered correlator. The derivation starts from a proposed form f(t,x) = 1 - epsilon e^{lambda(x/t)t}, converts the MSS bound into a bound on the Legendre transform of lambda(v), and then integrates the resulting differential inequality subject to lambda(v_B)=0. The paper also generalizes the bound to anisotropic butterfly cones, analyzes SYK chains, MSW models, chiral SYK models, stringy corrections in holography, and conformal Regge theory, and re-examines chaos bounds for boosted and rotating ensembles in Section 5.","tokens_in":28339,"tokens_out":4426,"duration_ms":46722,"significance":"If the result holds in the stated generality, it is a valuable refinement of the MSS chaos bound: it upgrades a single Lyapunov exponent to a velocity-resolved bound that is saturated in a broad class of strongly coupled systems, and it gives a concrete meaning to 'maximal chaos along rays' with a sharp critical velocity v*. The derivation in Section 2 is clean, parameter-free, and directly follows from the MSS bound combined with the ray ansatz; the examples are taken from the literature and the saturation of the bound is demonstrated by independent saddle/pole calculations. The connection to conformal Regge theory (Section 4) is a useful bridge between chaos and the analytic structure of large-N CFT correlators, and the rotating-ensemble analysis in Section 5 clarifies previous conflicting claims about rotating BTZ black holes. The paper is transparent about several assumptions, but the central claim is conditional on the unproven ray-ansatz structure, which limits the universality of the abstract's statement.","major_comments":[{"comment":"The bound (1.8) is obtained by substituting the ray ansatz f(t,x) = 1 - epsilon e^{lambda(x/t)t} into the fixed-x MSS bound (1.6), which yields a bound on the Legendre transform lambda(v) - v·grad lambda(v). The ansatz is proposed in the Introduction as 'it seems natural to propose' and is not derived from microscopic dynamics. If a local system's OTOC does not organize into smooth exponential rays—for example, if the front is broadened or diffusive, lambda(v) is not differentiable, or multiple saddles contribute—then partial_t f/(1-f) is not equal to lambda(v) - v·grad lambda(v), and Eq. (1.8) does not follow. Since this is exactly the central claim advertised in the abstract, the paper should either derive the ansatz under stated conditions or explicitly frame the result as conditional on the ray structure and adjust the abstract and introduction accordingly.","section":"§2.2, Eq. (1.5) and Eq. (1.8)"},{"comment":"The Regge-theoretic justification of the ansatz for large-N CFTs is itself conditional: it assumes the leading Regge pole dominates the integral (4.14) and explicitly assumes the convexity of the large-N Regge trajectory j(-ir), a property that the paper states has not been proven for the large-N trajectory. The text near Eq. (4.22) acknowledges this, but the Introduction and Abstract present the result as a universal bound for local systems. The dependence of the Section 4 analysis on the unproven convexity assumption should be flagged in the summary of results, not only in the technical section.","section":"§4.2, near Eq. (4.22)"},{"comment":"The anisotropic generalization in Eq. (2.12) is presented as an 'ideal bound' because the extremal function lambda_max(v) defined in Eq. (2.13) saturates the Legendre-transform bound (1.7) for all v. This is verified locally, but the definition of v±_B(v) via intersection of a half-line with the butterfly cone assumes that the zero set of lambda is a well-defined closed cone with no reentrant structure. The paper does not discuss what happens if the butterfly cone is not star-shaped or if lambda(v) has non-monotonic behavior along a ray. This is a minor gap for the generality of the anisotropic statement, though it does not affect the isotropic bound.","section":"§2.3, Eq. (2.12)"}],"minor_comments":[{"comment":"The abstract states the bound as 'lambda(v) <= 2*pi*T(1-|v|/v_B)' without mentioning that this presumes the ray-ansatz definition of lambda(v). A one-sentence caveat in the abstract would make the scope of the result clear to a reader who does not go through the derivation.","section":"General / Abstract"},{"comment":"The string 'lambda'’t Hooft' appears with an apostrophe that is likely a LaTeX artifact; it should be written as lambda_{'t Hooft} or similar. This is a presentational issue that should be cleaned up.","section":"§4.2, Eq. (4.26) and surrounding text"},{"comment":"In the chiral SYK discussion, the text says 'there is both a lower and an upper critical velocity v±_* such that for v < v_-_* and v > v_+_* , the VDLE is ballistic.' Given the formulas v±_* = (2-2J^2)/(2∓J) and v±_B = 1±J, it would be helpful to state explicitly which of the two critical velocities is negative or positive for J in [0,1], as this affects the reader's picture of the asymmetric cone.","section":"§3.4"},{"comment":"The statement that 'on average, the Lyapunov exponent is lambda_L = 2*pi/beta' would benefit from a precise definition of the averaging procedure; the period of the modulation is stated but the average over one period is not explicitly computed in the text.","section":"§5.2, after Eq. (5.18)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid synthesis of previously known results with a new bound, and the examples are analyzed correctly. The main concern is that the central claim is presented in the abstract and introduction as a general result, while the derivation requires the ray ansatz (1.5) and, in the CFT case, the unproven convexity of the large-N Regge trajectory. This is a fixable framing issue rather than a fatal error, but it needs to be addressed honestly. The paper also leans heavily on prior work by Gu-Qi-Stanford, Murugan-Stanford-Witten, Lian-Sondhi-Yang, and Shenker-Stanford; the novelty is the bound itself and the unified interpretation of the pole-dominance mechanism. The treatment of rotating BTZ in Section 5 is a useful correction to earlier claims, although the regime of validity for the averaged Lyapunov exponent deserves a bit more care."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe one-line take: Eq. (1.8) is genuinely new and the derivation is clean, but its scope is narrower than the abstract suggests. It is a theorem about the function defined by the ray ansatz (1.5), not a direct consequence of the MSS bound alone.\n\nThe paper is at its best in Section 2. There is no heavy machinery: you substitute f = 1 - epsilon e^{lambda(x/t)t} into the MSS inequality, get the Legendre-transform bound, integrate, and find the linear envelope. That is a real step beyond the old pointwise bound, and it matches the intuition from holography. The examples are also handled honestly. The SYK chain, MSW models, chiral SYK, and stringy holographic computations are taken from the literature, and the paper shows how each saturates the bound through saddle-pole exchange. The Regge-theory section adds a nice interpretation of v* as the saddle-to-pole transition and of lambda(v) as the Legendre transform of the leading Regge trajectory. The rotating BTZ analysis also earns its keep: the OTOC is not purely exponential; it has a periodic modulation; the average exponent is 2 pi / beta, and the stronger boosted bound holds in the decompactified limit. That corrects an earlier misreading in refs. [20,21].\n\nThe soft spot is exactly where the stress-test pointed. Eq. (1.5) is an ansatz, introduced as 'it seems natural to propose.' Nothing in the paper shows that local OTOCs actually organize into smooth exponential rays with a well-defined lambda(v) throughout the butterfly cone. If the front is broadened, diffusive, or multiple saddles contribute, then partial_t f / (1-f) is not lambda(v) - v dot grad lambda(v), and the bound (1.8) does not follow. The paper knows this and says so, but the abstract and the claim about 'any chaotic large N 2d CFT' run ahead of the proof. The Regge argument supplies a microscopic justification only for large-N CFTs, and even there the lower bound on v* depends on the unproven convexity of the large-N Regge trajectory, which the paper admits. These are not fatal to the central result if it is restated as a conditional bound, but they need to be flagged clearly in a revision.\n\nWho should read it: people working on operator growth, SYK chains, and conformal Regge theory. It deserves a serious referee. My own recommendation is to send it out and ask for revision, not for the bound but for a scope statement.\n\nBest,","headline":"A genuinely new velocity-dependent refinement of the MSS bound, but its universality is conditional on a structural ansatz the paper does not derive; worth refereeing and citing with care.","tokens_in":28913,"tokens_out":2621,"would_cite":true,"duration_ms":25903,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves a universal bound on the velocity-dependent Lyapunov exponent inside the butterfly cone.","keywords":["velocity-dependent Lyapunov exponent","butterfly cone","chaos bound","out-of-time-order correlator","SYK chain","conformal Regge theory","rotating black holes","quantum chaos"],"falsifier":"Numerically compute the OTOC for a local, translation-invariant spin chain with a clear separation of time scales, extract $\\lambda(v)$ by fitting $\\log(1-f(t,vt))$ against $t$ at fixed rays inside the cone, and look for any ray with $0<|v|<v_B$ where $\\lambda(v)>\\frac{2\\pi}{\\beta}(1-|v|/v_B)$. One such data point, with error bars below the gap to the bound, would falsify the central claim.","tokens_in":27857,"feed_emoji":"🦋","tokens_out":10855,"duration_ms":100394,"temperature":0.7,"pith_summary":"The paper claims that in any local quantum system whose out-of-time-order correlator organizes along rays with a velocity-dependent Lyapunov exponent $\\lambda(v)$, the growth rate inside the butterfly cone is bounded by $\\lambda(v) \\leq \\frac{2\\pi}{\\beta}(1-|v|/v_B)$, where $\\beta$ is the inverse temperature and $v_B$ is the butterfly speed. This generalizes the single-rate chaos bound of [2], which is recovered at $v=0$. The derivation applies the known analyticity-based chaos bound to the exponential ray ansatz, turns the result into a bound on the Legendre transform of $\\lambda$, and solves the resulting differential inequality with the boundary condition that $\\lambda$ vanishes at the cone edge. If the bound is right, strongly coupled systems saturate it beyond a critical velocity $v_*<v_B$, so that a boosted probe scrambles at the fastest allowed rate. The paper also uses the bound to sharpen the notion of scrambling time and to clarify that rotating black hole computations, once periodicity is included, do not violate the chaos bound.","feed_headline":"Butterfly-cone chaos growth obeys a universal velocity bound","feed_subtitle":"Generalizing the fast-scrambling bound, systems saturate it beyond a critical velocity, boosting chaos.","key_machinery":"The central object is the velocity-dependent Lyapunov exponent $\\lambda(v)$, defined by the ray ansatz $f(t,x)=1-\\epsilon e^{\\lambda(x/t)t}$ for the normalized OTOC inside the butterfly cone. The argument is carried by three pieces: the known chaos bound [2], $|\\partial_t f|/(1-f) \\leq 2\\pi/\\beta$, which rests on analyticity in a strip and factorization of correlators; the conversion of this bound into $|\\lambda(v)-v\\cdot\\nabla\\lambda(v)| \\leq 2\\pi/\\beta$, a bound on the Legendre transform of $\\lambda$; and the boundary condition $\\lambda(v_B)=0$ at the cone edge, which turns the differential inequality into the integrated linear bound. The proof technique writes $\\lambda(v)-v\\lambda'(v)=a(v)$ with $|a(v)|\\leq 2\\pi/\\beta$, solves it as $\\lambda(v)=v\\int_v^{v_B} du\\, a(u)/u^2$, and bounds the integral by replacing $a(u)$ with its maximum. Saturation above $v_*$ is mediated by a pole in the mode integral, interpreted through the ladder identity of [9] or through the stress tensor pole in conformal Regge theory.","core_discovery":"Starting from the normalized out-of-time-order four-point function $f(t,x)$ and the ray ansatz $f(t,x)=1-\\epsilon e^{\\lambda(x/t)t}$ inside the butterfly cone, the paper derives the universal inequality $\\lambda(v) \\leq \\frac{2\\pi}{\\beta}(1-|v|/v_B)$. The known bound $|\\partial_t f|/(1-f) \\leq 2\\pi/\\beta$ becomes the Legendre-transform bound $|\\lambda(v)-v\\cdot\\nabla\\lambda(v)| \\leq 2\\pi/\\beta$; in the isotropic case this differential inequality, together with $\\lambda(v_B)=0$, integrates to the stated linear bound. The paper identifies a critical velocity $v_* < v_B$ above which the bound is saturated in SYK chains, two-dimensional SYK-like CFTs, the chiral SYK model, and holographic gauge theories with stringy corrections, and explains saturation as an exchange of dominance between a saddle point and a pole in the integral defining the OTOC. In conformal Regge theory the critical velocity is the inverse slope of the leading large-$N$ Regge trajectory at the stress tensor, and the velocity-dependent Lyapunov exponent interpolates between the Regge and light-cone limits. For rotating black holes in three-dimensional anti-de Sitter space, the growing part of the OTOC is a periodic modulation on top of an exponential with average Lyapunov exponent $2\\pi/\\beta$, so the previously claimed violation of the bound disappears once the periodicity of the shockwave profile is taken into account.","pith_inferences":["Beyond the paper, the same Legendre-transform structure suggests a universal statement about front shapes: if a system's $\\lambda(v)$ is smooth and concave, the butterfly front should be linearly sharp where the bound is saturated and rounded below $v_*$, a signature directly measurable in cold-atom or trapped-ion simulations of operator spreading.","Beyond the paper, because the argument uses only the ray ansatz and the analyticity bound, it should also constrain classical and semiclassical chaotic systems with a local growth rate; testing the inequality with classical spin-chain numerics would show whether the bound is specifically quantum or a general feature of local chaos.","Beyond the paper, the saddle-pole exchange suggests that the subexponential prefactor $\\epsilon(t,x)$ carries the signature of saturation: near $v_*$ the prefactor should show crossover behavior, which an exponential-only analysis cannot resolve.","Beyond the paper, a local system with a strongly broadened or non-differentiable front would escape this bound, marking the boundary of the ray-ansatz regime and motivating a refined bound involving the front width."],"forward_implications":["At every velocity inside the butterfly cone, the growth rate is no larger than $\\frac{2\\pi}{\\beta}(1-|v|/v_B)$, so the bound and the definition of the cone edge are mutually consistent.","Boosting a probe, or working in a boosted thermal ensemble, cannot exceed the faster of the two chiral rates: $\\lambda_L \\leq \\min\\{2\\pi/\\beta_+, 2\\pi/\\beta_-\\}$, with saturation when the critical velocity is crossed.","The local scrambling time satisfies $t_{\\rm scr}(x) \\geq \\frac{\\beta}{2\\pi}\\log(1/\\epsilon)+|x|/v_B$, so the boundary of the scrambling region is a cone only when chaos is maximal at all velocities; otherwise the tip is smoothed out.","In large-$N$ CFTs, the critical velocity is the inverse slope of the leading Regge trajectory at the stress tensor; in planar $N=4$ SYM it equals $1/\\alpha_1(\\lambda)$ for the planar coupling, and the ballistic maximal-chaos front appears for coupling above about 37.74.","For rotating black holes, the average Lyapunov exponent is exactly $2\\pi/\\beta$ and the instantaneous version of the bound can be violated only after times that scale with the system size; in the decompactified limit the VDLE obeys the boosted bound."],"supporting_citations":[{"why":"Supplies the analyticity bound $|\\partial_t f|/(1-f)\\leq 2\\pi/\\beta$ that the paper applies to the ray ansatz.","marker":"[2]"},{"why":"Introduced the velocity-dependent Lyapunov exponent and ray-organized growth that the paper formalizes.","marker":"[4]"},{"why":"Provides the SYK-chain OTOC whose saddle-pole exchange gives the first worked example of VDLE saturation.","marker":"[5]"},{"why":"Supplies the two-dimensional SYK-like models whose Regge trajectory and pole structure determine $v_*$.","marker":"[6]"},{"why":"Chiral SYK model with an asymmetric butterfly cone used to test the anisotropic bound.","marker":"[8]"},{"why":"Ladder identity producing the pole at $\\kappa=1$, the mechanism behind saturation above $v_*$.","marker":"[9]"},{"why":"Stringy corrections to holographic OTOCs, giving the same saddle-pole mechanism and the VDLE form.","marker":"[11]"},{"why":"Rotating black hole OTOC calculation that the paper reinterprets with the periodic shockwave profile.","marker":"[20]"},{"why":"Parallel rotating black hole calculation whose claimed violation of the chaos bound is clarified by the periodicity analysis.","marker":"[21]"},{"why":"Conformal Regge resummation and Regge trajectory formalism used to connect $\\lambda(v)$ to the leading trajectory.","marker":"[36]"}],"fun_headline_variants":["Universal velocity bound for chaos in butterfly cone","Chaos bound generalizes to velocity-dependent Lyapunov exponent","Boosting quantum systems enhances chaos, bound saturates at v*","Butterfly-cone chaos: new velocity bound, critical boost saturation","Saturated chaos growth beyond critical velocity in butterfly cone"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that, inside the butterfly cone, the normalized OTOC actually takes the ray form $f=1-\\epsilon e^{\\lambda(x/t)t}$ with only subexponential corrections throughout the Lyapunov regime; if a local system's growing region broadens rather than organizing into smooth rays, the derived bound does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Universal velocity bound for chaos in butterfly cone","Chaos bound generalizes to velocity-dependent Lyapunov exponent","Boosting quantum systems enhances chaos, bound saturates at v*","Butterfly-cone chaos: new velocity bound, critical boost saturation","Saturated chaos growth beyond critical velocity in butterfly cone"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000662,"raw_usage":{"total_tokens":3115,"prompt_tokens":1124,"completion_tokens":1991,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":740,"completion_tokens_details":{"reasoning_tokens":1907}},"tokens_in":740,"tokens_out":1991,"duration_ms":14743,"temperature":1.0,"reasoning_tokens":1907,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:10:10.235263+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically compute the OTOC for a local, translation-invariant spin chain with a clear separation of time scales, extract $\\lambda(v)$ by fitting $\\log(1-f(t,vt))$ against $t$ at fixed rays inside the cone, and look for any ray with $0<|v|<v_B$ where $\\lambda(v)>\\frac{2\\pi}{\\beta}(1-|v|/v_B)$. One such data point, with error bars below the gap to the bound, would falsify the central claim.","supporting_citations":[{"cited_title":"On the relation between the magnitude and exponent of OTOCs","cited_arxiv_id":"1812.00120","evidence_quote":"Ladder identity producing the pole at $\\kappa=1$, the mechanism behind saturation above $v_*$."}],"review_version":1}