{"id":"4e5693ee-ed07-4c07-a0ea-8f2281c625c9","arxiv_id":"1908.05025","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a two-species weak-competition system, the exact invasion speeds are c1=2√(dr), c2=max{c_LLW, c_nlp}, c3=-c̃_LLW, with c_nlp a nonlocally determined speed, resolving a 1997 conjecture.","lead":"This paper computes the exact spreading speeds of two competing species invading an empty habitat, showing that the slower species is chased by the faster one at a speed that depends nonlocally on the habitat ahead. The result answers a 1997 conjecture by Shigesada and Kawasaki and introduces a geometric optics method for problems with several moving fronts.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Freidlin condition in Proposition 3.5(b) is asserted rather than verified; the explicit formula for c_nlp collapses without it.","rationale":"The reader identified the compact-support hypothesis (H∞) as the weakest assumption. That is a legitimate scope note, but it is not a load-bearing correctness concern: Theorem 1.3 is explicitly stated under (H∞), and the whole proof is conditional on it. The genuinely load-bearing step is the Freidlin condition in Proposition 3.5(b). Without it, the Hamilton-Jacobi solution w1 cannot be identified with max{J1,0}, and therefore the explicit formula for c_nlp does not follow from the variational calculation. The proof of Proposition 3.5(b) is only two observations and omits the argument for boundary points whose unconstrained optimal path lies on the boundary of P. This is exactly the situation in case (b) at x/t = 2 sqrt(1-a). My proposed test supplies the missing argument: a family of paths that remain in the open set P with cost converging to J1 = 0. If the test is carried out and succeeds, the proof is complete. Otherwise, the claimed value of c2 is not established. I found no other internal inconsistency: the half-relaxed limit framework, the comparison arguments in Proposition 2.1, the traveling-wave comparison in Lemma 2.4, and the algebraic reduction in Proposition 4.2 are coherent, and the numerical experiment corroborates the formula. The verdict should therefore be conditional on completing the verification of Freidlin's condition rather than a flat acceptance.","tokens_in":32616,"tokens_out":39955,"duration_ms":378163,"concrete_test":"Verify Freidlin's condition (41) at the boundary point (t,x) = (1, 2 sqrt(1-a)) in the case c1/2 > sqrt(a)+sqrt(1-a). Let c0 = 2 sqrt(1-a), and for epsilon > 0 define gamma_epsilon(s) = c0(1-s) + epsilon s(1-s), so that gamma_epsilon(0) = x, gamma_epsilon(1) = 0, and gamma_epsilon(s) > c0(1-s) for s in (0,1). Compute I_epsilon = integral_0^1 [ |gamma_epsilon'(s)|^2 / 4 - 1 + a ] ds, noting the indicator in L1 equals 1 for small epsilon because c0 < c1, and show I_epsilon -> 0 = J1(1,c0) as epsilon -> 0. If this limit holds, condition (41) is satisfied for the delicate boundary case; if it fails, the explicit formula for c_nlp in Theorem 1.3 is unsupported. A supplementary check should repeat the construction for boundary points in case (a), where the piecewise-linear optimal path already lies in P except at the endpoint.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The most load-bearing step is the identification w1 = max{J1,0} in Lemma 3.7, which converts the variational formula into the explicit zero set {x <= c_nlp t}. This identification is exactly Freidlin's condition (41), whose verification in Proposition 3.5(b) is reduced to two observations: (i) P = {J1>0} is a cone, and (ii) optimal paths are piecewise linear. The verification is incomplete in the case c1/2 > sqrt(a)+sqrt(1-a), where the boundary of P is x = 2 sqrt(1-a) t. For a boundary point with x/t = 2 sqrt(1-a), the unconstrained minimizer in Proposition 3.5(a) is the straight line from (0,0) to (t,x), which lies exactly on the boundary for every s in (0,t). Condition (41) requires the infimum over paths that remain in the open set P to equal J1(t,x)=0, and the paper does not demonstrate this. If (41) failed, w1 would not be max{J1,0}, the zero set of w1 would not be x <= c_nlp t, and both Proposition 4.1 (lower bound c2 >= c_nlp) and Corollary 2 (the exponential rate used in Proposition 4.2) would lose their foundation. This is an internal correctness risk, not merely a scope limitation, because it is needed exactly under the stated hypotheses.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the invasion of two competing species modeled by the Lotka–Volterra competition-diffusion system (1) under the initial condition (H∞): species u initially occupies the left half-line, species v is compactly supported, and the right habitat is empty. The main result (Theorem 1.3) states that for dr>1 the solution develops three fronts with speeds c1=2√(dr), c2=max{c_LLW,c_nlp}, and c3=−\\tilde c_LLW, where c_LLW and \\tilde c_LLW are the single-front spreading speeds of Lewis et al. and c_nlp is given explicitly by (4); between the fronts the solution converges to (0,0), (0,1), (k1,k2), and (1,0), respectively. The proof combines comparison arguments for c1 and c3 and rough bounds on c2, a WKB transform w_ε=−ε log u_ε with half-relaxed limits leading to the Hamilton–Jacobi equations (33)–(34), a variational solution w1 of the limiting problem whose zero set is {x≤c_nlp t}, and a large-deviation estimate together with wave comparisons (Lemma 2.4) to obtain the upper bound on c2. Theorem 1.4 treats the case dr=1 by a perturbation argument using the dr>1 and dr<1 results.","tokens_in":32837,"tokens_out":29815,"duration_ms":286066,"significance":"The result is significant: it provides the first exact determination of the second spreading speed in the two-species competition invasion problem with separated fronts, resolving a question raised by Shigesada and Kawasaki, and it exhibits a speed c_nlp that is nonlocally determined and can strictly exceed the minimal traveling wave speed. The geometric-optics method, adapted from Freidlin and Evans–Souganidis, is a promising tool for multi-front problems, and the paper contains detailed appendices for the variational calculus and comparison arguments, together with numerical simulations matching the predicted speeds. The main formula is explicit and falsifiable, and the overall derivation is coherent. The only substantive weakness is that one load-bearing verification, Freidlin’s condition in Proposition 3.5(b), is too terse as written; however, the missing argument is local and straightforward, so I do not view it as a threat to the main theorem.","major_comments":[],"minor_comments":[{"comment":"The verification of Freidlin’s condition is too terse. In the case c1/2>√a+√(1−a), the boundary is ∂P={x=2√(1−a)t} and the unconstrained minimizer for J1=0 is the straight line lying on ∂P, so (41) is not immediate. Please add the explicit perturbation argument, for example by considering paths γ_ε(s)=2√(1−a)s+ε s(t−s)/t, whose action tends to 0 because the boundary lies strictly below x=c1s and L1 is unchanged there.","section":"Proposition 3.5(b), Eq. (41)"},{"comment":"The sentence “It remains to check that w∗(0,x)=∞ for x>0” appears to refer to the lower half-relaxed limit w_* rather than the upper one, since the proof that follows uses a limit inferior; please correct the symbol to avoid confusion.","section":"Lemma 3.4"},{"comment":"The notation for maximal and minimal spreading speeds loses the overline/underline distinction in several places, in particular in Proposition 2.1(ii) and in Step 7 of its proof; please disambiguate which of \\underline c2 and \\bar c2 is used in (11b) and (11c).","section":"Section 2 and Proposition 2.1"},{"comment":"The line “Furthermore,” immediately before the Proof is an incomplete sentence and should be removed or completed.","section":"Lemma 3.2"},{"comment":"The construction of the compactly supported function \\tilde u0 with 0≤\\tilde u0≤u0 should be stated explicitly, for instance by taking a cut-off of u0 on a bounded interval inside (−∞,0], rather than merely asserted.","section":"Proposition 2.1, Step 4"},{"comment":"The existence of a traveling wave for the perturbed system (71) at any speed above c^δ_LLW is invoked without proof or reference; please add a citation or a short justification.","section":"Lemma 2.4, Appendix A"}],"recommendation":"minor_revision","confidential_remarks":"The stress-test concern about Freidlin’s condition does not lead me to doubt the main theorem: at the boundary x=2√(1−a)t the Lagrangian is unchanged under small perturbations into P, so condition (41) holds. I recommend asking the authors to expand Proposition 3.5(b) explicitly, since the current proof is too compressed for a step on which Lemma 3.7 and Propositions 4.1–4.2 depend."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is the headline: for a two-species weak-competition system spreading into empty habitat, it gives the exact rightward speed of the slower species as c2 = max{c_LLW, c_nlp}, with c_nlp a genuinely nonlocal quantity. That resolves a 1997 question of Shigesada and Kawasaki and goes beyond earlier work of Girardin and Lam, which covered a different parameter regime. The geometric optics machinery — passing through the WKB transform, half-relaxed limits, and a Hamilton-Jacobi equation with coefficients depending on the faster front — is new for this three-front problem and is a good fit. The paper is clearly written, the structure is sensible, and the numerical check matches the formulas. The proofs are detailed enough that I could follow the main thread, and I do not see circular argumentation: c_nlp is computed independently in Proposition 3.5 and only later compared with c_LLW, which is an external quantity from the single-front theory.\n\nThe soft spot is the verification of Freidlin's condition in Proposition 3.5(b). The stress-test concern is fair: in the case c1/2 > sqrt(a)+sqrt(1-a), the boundary of the positivity set P is x = 2 sqrt(1-a) t, and the unconstrained minimizer for J1 at a boundary point is the straight line lying exactly on that boundary. The paper says the condition holds because P is a cone and the optimal paths are piecewise linear, but it does not explicitly show that the infimum over paths staying in the open set P equals J1(t,x)=0. This is a genuine gap in exposition, not a fatal error. For a boundary point, you can take the straight-line path and perturb it slightly upward so it stays strictly above the boundary until the final instant; the cost tends to zero as the perturbation size tends to zero. I checked this for the case at hand. So the concern is fillable, but the authors should write out that approximation argument. As it stands, the assertion is too terse for a result that the later propositions lean on.\n\nOther cautions are minor: the compact-support hypothesis (H∞) is sharp and the explicit c_nlp will not survive with merely exponential decay; the proof of the dr=1 case is a perturbation argument that relies on continuous dependence of c_LLW on parameters, which is cited rather than proved. Neither of these undercuts the main result.\n\nThis is a solid paper worth serious refereeing. I would send it out for review, with a request to expand the proof of the Freidlin condition and to make the boundary-approximation step explicit. The main result is likely correct, and the methods will be useful to people working on nonlocal pulled fronts and multi-species spreading.","headline":"Exact spreading speeds for weak competition with a nonlocally pulled front; the main theorem is new and the Hamilton-Jacobi framework works, but the verification of Freidlin's condition needs a fill-in.","tokens_in":33454,"tokens_out":5232,"would_cite":true,"duration_ms":47439,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K58","35B40","35D40"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that a slower competing species invading empty habitat can spread at a nonlocally pulled speed, given explicitly in terms of the model parameters, and that the solution develops exactly three invasion fronts with these…","keywords":["Lotka-Volterra competition-diffusion","spreading speeds","geometric optics","Hamilton-Jacobi equation","nonlocally pulled front","multiple invasion fronts","viscosity solution","WKB transformation"],"falsifier":"Run a high-resolution numerical simulation of system (1) with a=0.6, b=0.5, d=1.5, r=1 and initial data u0=χ_{[-1000,0]}, v0=χ_{[-20,0]}, and measure sup{x:u(t,x)>0.4}/t at t=200 and t=400; Theorem 1.3 predicts the ratios converge to c2 = max{c_LLW, c_nlp} ≈ 1.3387, so a limit below that, such as 1.265, would falsify the claimed formula.","tokens_in":32356,"feed_emoji":"🌱","tokens_out":5526,"duration_ms":54255,"temperature":0.7,"pith_summary":"The paper establishes exact spreading speeds for the Lotka-Volterra two-species competition-diffusion system when the right half-line is initially empty. For dr>1, the solution develops three fronts: the faster species v invades empty habitat at speed 2√dr, the coexistence state invades the v-only region at speed max{c_LLW, c_nlp}, and the u-only state invades the coexistence region at speed -c̃_LLW. The novel quantity c_nlp is a nonlocally pulled speed that can strictly exceed the minimal traveling wave speed c_LLW. This resolves a question raised by Shigesada and Kawasaki about multiple invasion fronts during biological invasions, and it shows that the slower competitor's speed is determined by habitat quality far ahead of the front, not just by local conditions.","feed_headline":"Three-front invasion speeds pinned down exactly","feed_subtitle":"The slower competitor can spread at a nonlocally pulled speed exceeding the classical traveling wave speed.","key_machinery":"The load-bearing object is the WKB transformation w_ε(t,x) = -ε log u(t/ε, x/ε) and its half-relaxed limits w_* and w^*, which satisfy viscosity sub- and super-solutions of a Hamilton-Jacobi equation whose coefficient switches at the known faster front x=2√dr t. The limits are pinned down by comparison with the explicit solution w1 = max{J1,0}, where J1 is the infimum of the path integral ∫ [|γ̇|²/4 - 1 + a χ_{γ≤c1 s}] ds over curves ending at (t,x). Solving this variational problem gives the front x=c_nlp t, and an exponential decay estimate for u along a line x=ĉt then controls the spreading speed from above via comparison with traveling waves. The machinery is the Evans-Souganidis PDE approach to geometric optics, including half-relaxed limits and Freidlin's condition, which handles coefficients depending on several moving frames.","core_discovery":"The central discovery is that the spreading speed c2 of the slower species u equals max{c_LLW, c_nlp}, where c_LLW is the classical spreading speed of the coexistence equilibrium into the v-only state, and c_nlp has the explicit formula c_nlp = √dr - √a + (1-a)/(√dr-√a) when √dr ≤ √a + √(1-a), and c_nlp = 2√(1-a) otherwise. The speed c_nlp is characterized by the zero level set of the viscosity solution w1 of the Hamilton-Jacobi equation min{∂t w + |∂x w|^2 + 1 - a χ_{x≤2√dr t}, w} = 0 with initial data 0 on the left half-line and ∞ on the right. In the variational representation of w1, minimizing paths spend time ahead of the moving front, so c_nlp depends on the environment in front of the invasion, making it nonlocally pulled. The paper proves convergence to the four homogeneous equilibria in the regions separated by the three fronts, and obtains the borderline case dr=1 as a limit.","pith_inferences":["If the compact-support hypothesis (H∞) is relaxed to merely exponential decay of u0 on the right, the WKB initial data would be finite rather than infinite, so the limiting Hamilton-Jacobi problem and the formula for c_nlp should change; a testable prediction is that the second front speed becomes a function of the decay rate and approaches c_nlp as the decay rate tends to infinity.","The algebraic coincidence of c_nlp with the speed found in a related monostable competition case suggests that nonlocally pulled fronts are a general mechanism in reaction-diffusion systems where a fast front modifies the effective growth rate ahead of a slower front, not a peculiarity of this parameter regime.","A direct numerical test could look for the sharp transition in (4): plotting the measured second-front speed against √dr should show a kink exactly at √dr = √a + √(1-a), with the speed formula switching branches there."],"forward_implications":["If Theorem 1.3 is correct, the slower species u can spread strictly faster than any traveling wave of the homogeneous system would permit, with c_nlp > c_LLW whenever the faster species is not too fast relative to the competition parameters.","The asymptotic state of the system is fully classified by three speeds: ahead of c1 t the habitat is empty, between c2 t and c1 t only v persists at density 1, between c3 t and c2 t both species coexist at (k1,k2), and left of c3 t only u persists at density 1.","When dr=1, the two rightward fronts merge into one front at speed 2, so the invasion goes (1,0) ← (k1,k2) → (0,0) with the coexistence region filling the interval between -c̃_LLW t and 2t.","The same geometric optics framework is claimed to extend to systems with three or more species and to higher dimensions, because the limiting Hamilton-Jacobi problem accommodates coefficients depending on several distinct front speeds."],"supporting_citations":[{"why":"Supplies the definition of c_LLW as the spreading speed of the coexistence state into (0,1) and the linear-determinacy criterion used as a baseline.","marker":"[27]"},{"why":"Raises the original biological question of two invasion fronts chasing each other during tree-species invasions, which this paper settles.","marker":"[36]"},{"why":"Introduces the large-deviation geometric optics method for the Fisher-KPP equation and the Freidlin condition used to identify w1.","marker":"[17]"},{"why":"Provides the PDE approach to geometric optics, including half-relaxed limits, comparison principles, and representation formulas that form the core machinery.","marker":"[14]"},{"why":"Gives the monostable-case spreading speed whose formula coincides with c_nlp and motivates the super/sub-solution comparison used here.","marker":"[18]"},{"why":"Establishes prior two-species spreading estimates for compactly supported initial data, including the bounds c2 ∈ [2√(1-a), 2] improved in this paper.","marker":"[31]"},{"why":"Supplies the half-relaxed limit method used to define the upper and lower limits w* and w^* of the WKB transform.","marker":"[5]"}],"fun_headline_variants":["Three-front spreading speeds solved exactly via geometric optics","Nonlocally pulled fronts: exact speeds for three-species invasion","Exact three-front speeds settle 1997 open question","Geometric optics reveals nonlocally pulled invasion speed","Three fronts, exact speeds: nonlocally pulled spreading"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on the slower species being exactly absent on a right half-line initially, so that the WKB variable starts as infinity there; if u0 only decays exponentially, the explicit formula for c_nlp no longer follows from the argument.","fun_headline_variants_meta":{"raw":{"variants":["Three-front spreading speeds solved exactly via geometric optics","Nonlocally pulled fronts: exact speeds for three-species invasion","Exact three-front speeds settle 1997 open question","Geometric optics reveals nonlocally pulled invasion speed","Three fronts, exact speeds: nonlocally pulled spreading"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000986,"raw_usage":{"total_tokens":4139,"prompt_tokens":860,"completion_tokens":3279,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":3199}},"tokens_in":476,"tokens_out":3279,"duration_ms":22673,"temperature":1.0,"reasoning_tokens":3199,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:26:00.597735+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a high-resolution numerical simulation of system (1) with a=0.6, b=0.5, d=1.5, r=1 and initial data u0=χ_{[-1000,0]}, v0=χ_{[-20,0]}, and measure sup{x:u(t,x)>0.4}/t at t=200 and t=400; Theorem 1.3 predicts the ratios converge to c2 = max{c_LLW, c_nlp} ≈ 1.3387, so a limit below that, such as 1.265, would falsify the claimed formula.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the definition of c_LLW as the spreading speed of the coexistence state into (0,1) and the linear-determinacy criterion used as a baseline."},{"cited_title":"Shigesada and K","cited_arxiv_id":null,"evidence_quote":"Raises the original biological question of two invasion fronts chasing each other during tree-species invasions, which this paper settles."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the large-deviation geometric optics method for the Fisher-KPP equation and the Freidlin condition used to identify w1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the PDE approach to geometric optics, including half-relaxed limits, comparison principles, and representation formulas that form the core machinery."},{"cited_title":"Girardin and K","cited_arxiv_id":null,"evidence_quote":"Gives the monostable-case spreading speed whose formula coincides with c_nlp and motivates the super/sub-solution comparison used here."},{"cited_title":"Lin and W","cited_arxiv_id":null,"evidence_quote":"Establishes prior two-species spreading estimates for compactly supported initial data, including the bounds c2 ∈ [2√(1-a), 2] improved in this paper."},{"cited_title":"Barles and B","cited_arxiv_id":null,"evidence_quote":"Supplies the half-relaxed limit method used to define the upper and lower limits w* and w^* of the WKB transform."}],"review_version":1}