{"id":"bed3b6c9-a143-43f8-bb95-aa3df68ce77f","arxiv_id":"1908.01331","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The 3D critical Brezis-Nirenberg energy drops quadratically, S(a+εV)-S = -c sup Q_V^2/|a| ε^2 + o(ε^2), and minimizers concentrate at the maximizer.","lead":"This paper computes the exact energy change in a famous 3D optimization problem from partial differential equations when a small adjustment is made to the setup. The change is quadratic in the adjustment, and the solutions cluster at a precise point.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Dependence on Druet's non-attainment theorem is the load-bearing point; no internal gap found, so verdict stands.","rationale":"The reader identified Druet's non-attainment theorem as the weakest assumption, and I agree. The central claim, Theorem 1.3, requires that almost minimizers concentrate; the proof of that concentration is Proposition 3.1, whose first step rules out a nonzero weak limit precisely by citing [12, Step 1]. Everything after Proposition 3.1 is a self-contained bootstrap of expansions and coercivity estimates, so the external theorem is genuinely load-bearing. However, this is a standard published result, and the paper's use of it is explicit and covers exactly the required regularity class. I found no internal inconsistency, omitted step that affects the main line (Lemma 2.6 is routine), or hidden assumption beyond those stated in Assumption 1.2. The dependence on Druet's theorem is a risk only if that theorem is false, which is not suggested by the current literature. The verdict of ACCEPT with moderate confidence remains appropriate; the proposed check would independently confirm the theorem's applicability and thereby complete the weakest link.","tokens_in":40580,"tokens_out":24902,"duration_ms":238779,"concrete_test":"Settle the application by verifying the specific non-attainment statement for the class used here: prove that a minimizer u0 for S(a) with a critical would imply S(a-tχ)<S(a) for some small t>0 and a suitable nonnegative cutoff χ supported where u0>0, contradicting criticality. If this implication is reproduced from [12, Step 1] or an independent proof is supplied, the load-bearing step is sound; if a counterexample in this class exists, Proposition 3.1 fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The lower-bound bootstrap in Sections 4-6 presupposes Proposition 3.1: almost minimizers decompose as u=α(PU_{x,λ}+w) with ||∇w||→0. Appendix B proves this by first showing the weak limit u0 must vanish. The only mechanism excluding u0≠0 is Druet's theorem that S(a) is not attained for critical a, used at the Brézis-Lieb comparison around (B.1). If that external theorem failed, M=0 and u0 would be an S(a)-minimizer; the concentration decomposition would not follow, and the matching lower bound in Sections 4-6 would have no starting point. The paper does not reprove this theorem, but it explicitly identifies the dependence and the cited theorem is standard in the field. This is a reliance on a deep literature result rather than an internal inconsistency; the rest of the argument is carefully structured and no circularity or parameter fitting is evident.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the three-dimensional Brezis–Nirenberg problem for the functional S(a+εV) with a critical function a in the sense of Hebey and Vaugon. The main result, Theorem 1.3, gives a two-term asymptotic expansion of S(a+εV)-S with leading order -(3/S)^{1/2}(8π²)^{-1} sup_{x∈N_a(V)} Q_V(x)²/|a(x)| ε², where Q_V is defined via the Green's function G_a and N_a(V) is the subset of the Robin-function zero set where Q_V<0. Theorem 1.4 treats the complementary case N_a(V)=∅, showing S(a+εV)=S+o(ε²) and equality for small ε when Q_V>0 everywhere on N_a. Theorem 1.7 characterizes almost minimizers: they concentrate at maximizers of Q_V²/|a| in N_a(V), with ελ → 4π²|a(x_0)|/|Q_V(x_0)|. The upper bound is obtained by explicit trial functions ψ_{x,λ}; the lower bound is obtained through a three-stage bootstrap (Sections 4–6) using coercivity of a quadratic form on the orthogonal complement of the approximate kernel, together with a concentration-compactness argument for almost minimizers in Appendix B.","tokens_in":40709,"tokens_out":12913,"duration_ms":121451,"significance":"Assuming the main results are correct, this is a significant contribution: it computes a higher-order (ε²) asymptotic in the critical case where the first-order term cancels, and it shows that the coefficient is non-local (through Q_V) rather than pointwise in V. The proof is careful: the upper bound is parameter-free and the lower bound tracks error terms explicitly to o(ε²). The paper also gives precise concentration-point selection for almost minimizers, going beyond earlier works that identify only N_a. Strengths include the explicit two-sided bounds in Theorem 2.1, the transparent three-step refinement of the decomposition of almost minimizers, and the honest identification of the external input from Druet's non-attainment theorem in Proposition 3.1. I find no circularity or parameter fitting in the argument, provided the cited external theorem is sound.","major_comments":[],"minor_comments":[{"comment":"The proof of Lemma 2.6 is omitted: the text says the proof is similar but simpler than that of Lemma 2.5. Since this lemma is used in the denominator expansion (2.3) and in the estimate leading to Lemma 6.6, the authors should supply the short proof or give a precise reference. In its present form this is a missing support for a step in the main argument; I do not regard it as fatal, but it should be fixed before publication.","section":"Section 2.2, Lemma 2.6"},{"comment":"The symbols 'opλ2q' and 'opǫλ´1q' in the display (6.18) and in the surrounding text should presumably read o(λ^{-2}) and o(ελ^{-1}), respectively. Please correct these typographical errors so the order estimates are unambiguous.","section":"Section 6.4, around (6.18)"},{"comment":"The proof that u_ε converges weakly to zero imports Druet's non-attainment theorem [12, Step 1]. The authors note this dependence and that the relevant part of [12] requires only a∈L^{3/2}. For completeness, please state the imported theorem precisely and confirm that the hypotheses of the present setting (in particular, criticality in the sense of Definition 1.1 and the regularity of a) match those of [12, Step 1]. This is a clarification request about a load-bearing literature result, not an assertion of circularity.","section":"Appendix B, Step 1"}],"recommendation":"minor_revision","confidential_remarks":"This is a strong, careful paper in the scope of math.AP. The external dependence on Druet's theorem is legitimate and explicitly flagged; I see no reason to doubt the central claim. The minor revisions requested (proof of Lemma 2.6, a typo fix, and a precise statement of the imported theorem) are local and do not affect the validity of the argument."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is the right paper for the 3D case, and the proof holds up. The main theorem is the two-term expansion (1.6) with the ε^2 coefficient given by Q_V^2/|a|, plus the sharp selection rule for concentration points in Theorem 1.7. That is genuinely new: in N≥4 the first-order term survives, while here criticality kills it, so you need the quadratic correction. The paper works.\n\nWhere it earns credit: the upper bound is a clean trial-function argument; the lower bound is a three-stage bootstrap that refines the decomposition of almost minimizers, using coercivity twice more than previous arguments. The cancellation in Lemma 6.3, where the tangential components of the correction drop out, is the subtle step and it is handled carefully. The asymptotics for Theorem 1.4 are also a nice touch, showing the N_a(V)≠∅ condition is almost necessary.\n\nSoft spots, in proportion: Lemma 2.6 is stated without proof; it is a routine analogue of Lemma 2.5, so this is minor, but a referee should ask for the proof. More importantly, the lower bound depends on Proposition 3.1, which in turn depends on Druet's non-attainment theorem and on Bahri–Coron. The stress-test note is right: if Druet's theorem were false, the decomposition would fail. But that theorem is a cornerstone of the field, and the paper identifies the dependence explicitly. This is a reliance on literature, not a hidden gap. The other heavy inputs—Esposito's coercivity and Rey's estimates—are standard and partly reproved.\n\nCaveats: the estimates are numerous and not machine-checked, so confidence should be moderate, but I see no circular or fitted parameter anywhere. The constant in Lemma 2.7 has a typo ('As x→∞' should read 'As λ→∞'), trivial to fix.\n\nFor whom: anyone working on critical Sobolev problems, Brezis–Nirenberg asymptotics, or concentration of almost minimizers. It deserves a serious referee and, after minor revisions, publication.","headline":"Solid, careful analysis that nails the 3D Brezis–Nirenberg energy gap at order ε^2; the reliance on Druet's theorem is real but standard.","tokens_in":41333,"tokens_out":1923,"would_cite":true,"duration_ms":20622,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35J20","35J60","35B40","46E35"],"pacs":[],"model":"deepseek-v4-flash","headline":"For critical potentials, small perturbations lower the sharp Sobolev energy quadratically, with the coefficient set by a nonlocal Green-function average.","keywords":["critical Sobolev exponent","Robin function zero set","Green's function regular part","energy asymptotics","almost minimizer concentration","nonlocal perturbation functional","three-dimensional Sobolev inequality"],"falsifier":"Work in the unit ball with the constant potential $a=-\\pi^2/4$, which is critical and has $N_a=\\{0\\}$. For any bounded $V$ with $q_V=\\int_B V(y)\\cos^2(\\pi|y|/2)/|y|^2\\,dy<0$, the theorem predicts $S(a+\\epsilon V)-S=-(3/S)^{1/2}(2\\pi^4)^{-1}q_V^2\\epsilon^2+o(\\epsilon^2)$. A high-precision numerical minimization of the quotient over $H_0^1(B)$ for several small $\\epsilon$ should reproduce this exact quadratic slope; a systematically different slope would falsify the theorem.","tokens_in":40338,"feed_emoji":"📉","tokens_out":10708,"duration_ms":100595,"temperature":0.7,"pith_summary":"This paper studies the sharp Sobolev minimization problem in a bounded three-dimensional domain when the potential is critical—meaning its infimum already equals the sharp Sobolev constant and any pointwise smaller potential lowers it—and is then perturbed by a small multiple of an arbitrary bounded function. It proves that the first-order energy change vanishes, so the leading effect is a quadratic drop in the perturbation parameter, whose coefficient is a nonlocal average of the perturbation against the square of the Dirichlet Green function, optimized over the zero set of the Robin function. It also proves that almost minimizers concentrate at the optimizing point, with a blow-up rate proportional to the inverse perturbation size. The result matters because it identifies the precise order at which a critical threshold turns into strict subcritical behavior and gives a nearly sharp criterion for when a perturbation lowers the energy.","feed_headline":"Perturbing a critical potential lowers energy quadratically","feed_subtitle":"The drop's size is fixed by a Green-weighted average of the perturbation over the Robin-function zero set.","key_machinery":"The driving object is the zero set of the Robin function $\\varphi_a(x)=H_a(x,x)$, the diagonal of the regular part of the Green function of $-\\Delta+a$ with Dirichlet boundary conditions. Criticality forces $\\varphi_a\\ge0$ with a nonempty zero set, and the paper shows concentration must occur there. The new mechanism is the nonlocal functional $Q_V(x)=\\int_\\Omega V(y)G_a(x,y)^2\\,dy$: negative values of $Q_V$ on $N_a$ select the concentration point and set the energy coefficient, so the perturbation acts through a Green-weighted average rather than pointwise. Technically, the argument is carried by projected bubbles $\\psi_{x,\\lambda}=PU_{x,\\lambda}-\\lambda^{-1/2}(H_a(x,\\cdot)-H_0(x,\\cdot))$ and by the coercivity of the quadratic form $\\int_\\Omega(|\\nabla v|^2+av^2-15U_{x,\\lambda}^4v^2)\\,dy$ on the orthogonal complement of the finite-dimensional tangent space of the bubble ansatz; three successive applications of this coercivity extract the remainder structure to order $\\epsilon^2$.","core_discovery":"The paper's central claim is a second-order energy expansion for the critical Sobolev quotient under a small potential perturbation. Let $G_a$ be the Dirichlet Green function of $-\\Delta+a$, let $H_a(x,y)=|x-y|^{-1}-G_a(x,y)$ be its regular part, and set $\\varphi_a(x)=H_a(x,x)$. For a critical $a$, one has $\\varphi_a\\ge0$ and the zero set $N_a=\\{\\varphi_a=0\\}$ is nonempty. The paper proves that if $Q_V(x)=\\int_\\Omega V(y)G_a(x,y)^2\\,dy$ is negative somewhere on $N_a$, then $S(a+\\epsilon V)<S$ for all $\\epsilon>0$ and $$S(a+\\epsilon V)-S=-\\left(\\frac{3}{S}\\right)^{1/2}\\frac{1}{8\\$pi^{2}$}\\sup_{x\\in N_a(V)}\\frac{Q_V(x)^2}{|a(x)|}\\$epsilon^{2}$+o(\\$epsilon^{2}$),$$ with $N_a(V)=\\{x\\in N_a: Q_V(x)<0\\}$. Moreover, almost minimizers concentrate at a point attaining the supremum and satisfy $\\epsilon\\lambda\\to4\\pi^2|a(x_0)|/|Q_V(x_0)|$. In the complementary case the drop is only $o(\\epsilon^2)$, and if $Q_V$ is strictly positive on $N_a$ the energy stays exactly at $S$ for all sufficiently small $\\epsilon$.","pith_inferences":["Because the coefficient is a Green-weighted average rather than a pointwise value, a sign-changing perturbation $V$ can lower the energy even where $V$ is locally positive, so the zero set of the Robin function acts as a nonlocal 'sensor' for the perturbation.","The iterative coercivity scheme is not obviously limited to two orders; pushing it further should produce the $\\epsilon^3$ term and a systematic asymptotic expansion of $S(a+\\epsilon V)$, likely with the same maximization principle applied to higher-order Green-function functionals.","The leftover borderline case $\\min_{N_a}Q_V=0$ is a natural next target: there the $\\epsilon^2$ coefficient vanishes, and one expects the next nonzero order to be governed by higher-order terms in the regular-part expansion around the zero set.","The same machinery could transfer to compact manifolds or to perturbations that depend on $\\epsilon$ themselves, wherever a Dirichlet Green function and its Robin function are available."],"forward_implications":["If $N_a(V)\\neq\\emptyset$, the strict inequality $S(a+\\epsilon V)<S$ holds for every $\\epsilon>0$, and the energy drop is quadratic with the explicit leading coefficient given by the supremum of $Q_V^2/|a|$.","Any family of almost minimizers concentrates, along a subsequence, at a point in $N_a(V)$ maximizing $Q_V(x)^2/|a(x)|$, with the blow-up scale $\\epsilon\\lambda\\to4\\pi^2|a(x_0)|/|Q_V(x_0)|$.","If $N_a(V)=\\emptyset$, the energy drop is only $o(\\epsilon^2)$; if additionally $Q_V>0$ on $N_a$, the energy remains exactly at the Sobolev constant for all sufficiently small $\\epsilon$.","The condition $N_a(V)\\neq\\emptyset$ is therefore almost necessary and sufficient for a small perturbation to lower the critical energy, with the only unresolved borderline being $\\min_{N_a}Q_V=0$.","In the unit-ball example the whole formula becomes explicit: the coefficient is a universal constant times $q_V^2$, where $q_V=\\int_B V(y)\\cos^2(\\pi|y|/2)/|y|^2\\,dy$."],"supporting_citations":[{"why":"introduces the minimization problem and the critical-exponent setting.","marker":"[8]"},{"why":"supplies the non-attainment theorem for critical potentials, used to rule out a nonzero weak limit of almost minimizers.","marker":"[12]"},{"why":"provides the coercivity bound on the orthogonal complement and initial a priori estimates, adapted here to almost minimizers.","marker":"[15]"},{"why":"gives asymptotic expansions for projected Sobolev bubbles and establishes the role of the Robin function.","marker":"[23]"},{"why":"classifies optimizers of the Sobolev inequality, used in the concentration decomposition.","marker":"[21]"},{"why":"supplies the global compactness decomposition for concentration near the boundary.","marker":"[5]"},{"why":"links negativity of the Robin function to strict subcriticality, implying criticality forces a nonempty zero set.","marker":"[6]"},{"why":"provides the pointwise-convergence lemma used to separate the weak limit from the remainder.","marker":"[7]"}],"fun_headline_variants":["Quadratic energy drop for critical Sobolev perturbations","Robin-function zeros dictate Sobolev energy shift","Negative Green-average triggers sharp energy drop","Negative Green average yields quadratic drop"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on a previously established result that, for a potential sitting exactly at the critical threshold, the minimization problem has no minimizer at all; without that result, the argument that almost minimizers must have zero weak limit, and hence the entire concentration decomposition, would collapse.","fun_headline_variants_meta":{"raw":{"variants":["Quadratic energy drop for critical Sobolev perturbations","Robin-function zeros dictate Sobolev energy shift","Negative Green-average triggers sharp energy drop","Negative Green average yields quadratic drop"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00067,"raw_usage":{"total_tokens":3113,"prompt_tokens":1062,"completion_tokens":2051,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":678,"completion_tokens_details":{"reasoning_tokens":1996}},"tokens_in":678,"tokens_out":2051,"duration_ms":16009,"temperature":1.0,"reasoning_tokens":1996,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:16:31.488855+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Work in the unit ball with the constant potential $a=-\\pi^2/4$, which is critical and has $N_a=\\{0\\}$. For any bounded $V$ with $q_V=\\int_B V(y)\\cos^2(\\pi|y|/2)/|y|^2\\,dy<0$, the theorem predicts $S(a+\\epsilon V)-S=-(3/S)^{1/2}(2\\pi^4)^{-1}q_V^2\\epsilon^2+o(\\epsilon^2)$. A high-precision numerical minimization of the quotient over $H_0^1(B)$ for several small $\\epsilon$ should reproduce this exact quadratic slope; a systematically different slope would falsify the theorem.","supporting_citations":[{"cited_title":"Brézis, L","cited_arxiv_id":null,"evidence_quote":"introduces the minimization problem and the critical-exponent setting."},{"cited_title":"Druet, Elliptic equations with critical Sobolev exponents in dime nsion 3","cited_arxiv_id":null,"evidence_quote":"supplies the non-attainment theorem for critical potentials, used to rule out a nonzero weak limit of almost minimizers."},{"cited_title":"Esposito, On some conjectures proposed by Haim Brezis","cited_arxiv_id":null,"evidence_quote":"provides the coercivity bound on the orthogonal complement and initial a priori estimates, adapted here to almost minimizers."},{"cited_title":"Rey, The role of the Green ’s function in a non-linear elliptic equ ation involving the critical Sobolev exponent","cited_arxiv_id":null,"evidence_quote":"gives asymptotic expansions for projected Sobolev bubbles and establishes the role of the Robin function."},{"cited_title":"Lions, The concentration-compactness principle in the calculus o f variations","cited_arxiv_id":null,"evidence_quote":"classifies optimizers of the Sobolev inequality, used in the concentration decomposition."},{"cited_title":"Bahri, J.-M","cited_arxiv_id":null,"evidence_quote":"supplies the global compactness decomposition for concentration near the boundary."},{"cited_title":"Brézis, Elliptic equations with limiting Sobolev exponents—the im pact of topology","cited_arxiv_id":null,"evidence_quote":"links negativity of the Robin function to strict subcriticality, implying criticality forces a nonempty zero set."},{"cited_title":"Brézis, E","cited_arxiv_id":null,"evidence_quote":"provides the pointwise-convergence lemma used to separate the weak limit from the remainder."}],"review_version":1}