{"id":"d801741b-936f-4313-ad32-c1fd14ab6730","arxiv_id":"2506.06736","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives fractional Euler-Lagrange equations and claims to prove the Legendre condition by the classical method, but the fixed-endpoint Legendre proof contains an erroneous estimate.","lead":"This math paper proves a generalized Du Bois-Reymond lemma for functionals built from Caputo fractional derivatives and Riemann-Liouville integrals, then uses it to derive Euler-Lagrange equations. It claims to prove the Legendre condition by the classical second-variation method, but the fixed-endpoint proof has a serious gap.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.2's first-term estimate uses |f−f(c)|≤2M to claim a negative O(ε) upper bound; this is a direction error, and no lower bound on ∫(f−f(c))² is supplied, so the Legendre proof is invalid as written.","rationale":"The reader's weakest assumption identifies the same step: the proof needs a lower bound on ∫(f−f(c))², while the text only supplies an upper bound through M. My independent check of the estimate (19) confirms it has the wrong direction: with ⟨P r,r⟩≤−γ, bounding (f−f(c))² above bounds the negative quantity from below, not above. Moreover the standard repair is not present: one would need to choose f with bounded sup norm and a quantitative lower bound on the squared deviation over a positive fraction of the interval, e.g. f(t)=sin(π(t−(σ−ε))/(2ε)); the proof as written does neither. The other components (Du Bois-Reymond lemma, Euler-Lagrange integral equations) appear mostly sound, with only the minor k=0 technicality in Lemma 3.1. Since the central advertised result, Theorem 4.2, depends on this invalid estimate, the reader's REJECT verdict is appropriate; I am not recommending a change in verdict.","tokens_in":23406,"tokens_out":12600,"duration_ms":127009,"concrete_test":"Compute the left-hand side of (19) in Theorem 4.2 with α=1/2, β=1, t0=0, t1=2, σ=1, r=1, P≡−1, and f(t)=t on [1−ε,1+ε]. Then M≈1 and the term equals −∫_{1−ε}^{1+ε}(2−t)(t−c)²dt ~ −Cε³, where c is the weighted mean of t. The claimed upper bound is ≲ −8ε. For sufficiently small ε, −Cε³ > −8ε, so inequality (19) is false. This directly tests whether the central estimate of the Legendre proof is valid.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the estimate of the first term in the second variation (12) in the proof of Theorem 4.2. Since ⟨P(t)r,r⟩ ≤ −γ on [σ−ε,σ+ε], the integrand satisfies (t1−t)^{β−1}⟨P(t)r,r⟩(f(t)−f(c))² ≤ −γ(t1−t)^{β−1}(f(t)−f(c))². The paper replaces (f(t)−f(c))² by its upper bound 4M² and concludes the integral is ≤ −8M²γ(t1−t0)^{β−1}ε (inequality (19)). This is a sign/direction error: from (f−f(c))²≤4M² one obtains −γ(t1−t)^{β−1}(f−f(c))² ≥ −4γM²(t1−t)^{β−1}, i.e. a lower bound, not an upper bound. To get a negative O(ε) upper bound one needs ∫_{σ−ε}^{σ+ε}(f−f(c))²dt ≥ cε for some c>0. The proof only states that f is continuous and non-constant; such an f may have ∫(f−f(c))² arbitrarily small relative to ε (a narrow spike), or of order ε³ when f(t)=t on the interval. In that case the negative term is O(ε³), while the Q-term in (12) is O(ε^{1+α}); for 0<α<1, ε^{1+α}≫ε³ and the claimed domination fails. No construction of f with the required lower bound appears. Because Theorem 4.2 is the paper's advertised refutation of the obstruction raised in [9], the central claim is not supported as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a fractional variational problem whose functional is a Riemann-Liouville fractional integral of order β of a Lagrangian depending on the Caputo fractional derivative of order α. The main results are a fractional Du Bois-Reymond lemma, Euler-Lagrange equations in integral form for four endpoint regimes, and — as the advertised novelty — a Legendre necessary condition for the fixed-endpoint and free-initial/fixed-final cases. The authors claim this Legendre proof adapts the classical needle variation and thereby overcomes the obstruction described in [9]. The Euler-Lagrange part is developed through a direct, non-circular proof of the Du Bois-Reymond lemma, and several illustrative examples are included.","tokens_in":23768,"tokens_out":11553,"duration_ms":124175,"significance":"If correct, the explicit dependence of the necessary conditions on the relation between α and β would be a useful contribution, and the Legendre condition for fixed final constraints would settle a question raised in [9]. The Du Bois-Reymond lemma and the Euler-Lagrange derivations are largely coherent and are supported by explicit constructions. However, the central Legendre proof contains a load-bearing estimate with the wrong direction, and the theorem is not established as written. The examples and the comparison with [9] are helpful, but they do not compensate for the gap in the main proof.","major_comments":[{"comment":"The estimate of the first term in the second variation is not an upper bound. From (15), that term equals ∫(t1−t)^{β−1}(f(t)−f(c))²⟨P(t)r,r⟩dt ≤ −γ∫(t1−t)^{β−1}(f(t)−f(c))²dt. Replacing (f(t)−f(c))² by its maximum 4M² gives a lower bound, not an upper bound: the integral is ≥ −4M²γ∫(t1−t)^{β−1}dt. To conclude that the first term is at most −M₁γε, the proof would need a lower bound of the form ∫_{σ−ε}^{σ+ε}(t1−t)^{β−1}(f(t)−f(c))²dt ≥ cε for some c>0. No such bound is stated or proved for the arbitrary continuous nonconstant f. For instance, f(t)=t gives an integral of order ε³, while the Q-term in (12) is O(ε^{1+α}); for 0<α<1 the latter dominates, so the claimed contradiction δ²J<0 does not follow. Thus Theorem 4.2 is unsupported as written.","section":"Section 4.1, proof of Theorem 4.2, inequality (19)"},{"comment":"In the case 0<β≤α≤1, the statements require a constant k≠0 (or k∈R^n\\{0}) in conditions 2). This makes the stated 'necessary and sufficient' claims false: f≡0 satisfies identity (1), and the corresponding Euler-Lagrange condition (10) may hold with k=0, but these are excluded by the nonzero-constant requirement. The proof itself determines k from a boundary integral and never uses k≠0. The conditions should allow arbitrary k∈R (or k∈R^n), including zero.","section":"Lemmas 3.1 and 3.3; Theorems 4.1, 4.3, 4.5"},{"comment":"The weak local minimum in problem (P) is defined in the space C^α([t0,t1],R^n), whose Caputo derivative is continuous. The Legendre proof, however, works with x₀∈PC^α and variations h∈PC^α₀, whose Caputo derivatives may be piecewise continuous with jump discontinuities. Then x₀+λh need not belong to the admissible class C^α, so the second-variation inequality (12) is not justified for the variations used in the proof. The admissible class or the notion of weak local minimum must be extended to the piecewise-smooth setting before Theorem 4.2 can be applied.","section":"Section 4.1, setup before Theorem 4.2"}],"minor_comments":[{"comment":"In the definition of PC^α₀([t0,t1],R^n), the right-hand side should refer to the set PC^α([t0,t1],R^n), not PC^α₀, otherwise the definition is circular.","section":"Section 2"},{"comment":"The notation cAC^{α,∞}_{0+} is invoked without definition or reference; please state or cite the space explicitly.","section":"Remark 4.1"},{"comment":"The manuscript contains several LaTeX artifacts and inconsistent notations, such as a manuscript header 'JOT A manuscript No.', unclear spacing in author names, and the repeated use of '0 ⁄= k' where the intended meaning is 'k ≠ 0'. A careful editorial pass is needed.","section":"Throughout"}],"recommendation":"reject","confidential_remarks":"The Euler-Lagrange part of the paper has merit and may be worth publishing separately if the k=0 issue is corrected. The Legendre result, however, is the paper's central advertised contribution, and its proof rests on a sign error in the key estimate. Since the required lower bound on ∫(f−f(c))² is absent and is not derivable from the hypotheses, the main theorem is not proved. I do not see a simple local correction within the current proof framework, so I recommend rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the generalized Du Bois-Reymond lemma (Lemma 3.1) with the beta>alpha vs beta<=alpha split is genuinely new, and the integral-form Euler-Lagrange equations that display the alpha-beta coupling are not in [9] or [13]. Those parts are largely coherent, direct, and not circular. Second, the paper's headline claim — that the Legendre condition for fixed-final-constraint problems can be proved by the standard classical method — fails at the key estimate in Theorem 4.2. The stress-test note is right: the proof replaces (f(t)-f(c))^2 by its upper bound 4M^2 inside an integral with a negative coefficient, which gives a lower bound, not the claimed negative O(epsilon) upper bound. To make the argument work you need a positive lower bound on the integral of (f-f(c))^2 over the needle interval, and the paper neither states nor constructs an f with that property. For f(t)=t, the integral is O(epsilon^3), while the Q-term is O(epsilon^{1+alpha}); for 0<alpha<1 the claimed domination fails. This is a load-bearing flaw, not a minor gap.\n\nCredit where it is due. The proof of Lemma 3.1 is direct and the special variation with parameter k is a real idea. Theorem 4.4, for the free-initial/fixed-final problem, avoids the problematic f because the variation's fractional derivative is constant on the needle, so its first-term estimate has the correct sign and order. The examples illustrating the alpha-beta distinction are useful. The k!=0 restriction in Lemmas 3.1/3.3 and Theorem 4.1 is a smaller technical issue: for f=0 the equality holds but the stated condition requires k!=0, making the \"if and only if\" literally false. That is fixable by allowing k=0 or adding a separate zero case.\n\nThe reader's skeptic is mostly on target. I would add that Theorem 4.4's proof appears sound in the part that matters, which the reader's report does not emphasize. But the central assertion of the paper, Theorem 4.2, is not supported as written.\n\nWho is this for? People working on fractional variational necessary conditions. The lemma and EL equations deserve a serious referee, and the Legendre question is important enough that a referee could push the authors to either fix the proof or substantially qualify the claim. I would not desk-reject; I would send it out with a clear request to repair the k issue and rewrite the Legendre proof or remove the unsupported claim.","headline":"The new Du Bois-Reymond lemma and integral-form Euler-Lagrange equations are worth attention, but the advertised proof of the Legendre condition for fixed endpoints has a sign/direction error in its key estimate and is not valid as written.","tokens_in":24331,"tokens_out":4433,"would_cite":false,"duration_ms":46843,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["26A33","49K99","49K05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the classical proof of the Legendre necessary condition can be adapted to fractional calculus of variations with fixed endpoints, contrary to a claim in the existing literature.","keywords":["fractional calculus of variations","Caputo derivative","Riemann-Liouville integral","Euler-Lagrange equation","Legendre condition","Du Bois-Reymond lemma","weak local minimum","necessary conditions"],"falsifier":"For the needle variation in Theorem 4.2, take $f(t)=1+\\delta\\cos(\\pi(t-\\sigma)/\\varepsilon)$ on $[\\sigma-\\varepsilon,\\sigma+\\varepsilon]$; then the integral of $(f(t)-f(c))^2$ over the needle is $O(\\delta^2\\varepsilon)$ rather than the $O(\\varepsilon)$ needed for estimate (19), so for small $\\delta$ the negative term no longer dominates and the contradiction argument fails, which would refute the paper's claim that the standard classical proof goes through without additional hypotheses.","tokens_in":23166,"feed_emoji":"📐","tokens_out":19644,"duration_ms":158426,"temperature":0.7,"pith_summary":"The paper studies minimization problems in which the cost is a Riemann-Liouville fractional integral of order $\\beta>0$ of a Lagrangian that depends on a Caputo fractional derivative of order $0<\\alpha\\le 1$, under fixed, free, or mixed endpoint conditions. Its central claim is that, for the problems with a fixed final endpoint, a weak local minimum must satisfy the classical Legendre condition: the second derivative of the Lagrangian with respect to the derivative variable is positive semidefinite along the extremal, at every point where the Caputo derivative is continuous. This is proved by the standard classical method, using a fractional analogue of the Du Bois-Reymond lemma and a needle variation whose fractional derivative has support concentrated on a small interval. The result directly contradicts a well-known assertion in the literature that the classical fixed-endpoint proof cannot be adapted to fractional problems. If the claim is correct, fractional variational problems inherit the same second-order test as classical ones, and the relation between $\\alpha$ and $\\beta$ enters the Euler-Lagrange equations explicitly.","feed_headline":"Legendre's condition survives fractional derivatives","feed_subtitle":"A needle-variation proof extends the fixed-endpoint second-order test to Caputo-derivative variational problems.","key_machinery":"The main tool is the generalized Du Bois-Reymond lemma (Lemma 3.1), which characterizes when $\\int_{t_0}^{t_1}(t_1-t)^{\\beta-1}f(t)(cD_{t_0+}^\\alpha h)(t)\\,dt=0$ for every zero-endpoint $h$: if $\\beta>\\alpha$ the function $(t_1-t)^{\\beta-\\alpha}f(t)$ must vanish, and if $0<\\beta\\le\\alpha$ then $f$ must be a constant multiple of $(t_1-t)^{\\alpha-\\beta}$. This lemma converts the first variation into the Euler-Lagrange equations in integral form. The second mechanism is the needle variation (16), $h(t)=\\frac{1}{\\Gamma(\\alpha)}\\int_{t_0}^t (t-\\tau)^{\\alpha-1}g(\\tau)\\,d\\tau$ with $g=(f(t)-k)r$ on the needle and zero elsewhere; because the Caputo derivative of $h$ equals $g$, the variation has a compactly supported fractional derivative, which is exactly the feature the classical proof needs and which the earlier literature said could not be arranged under final constraints.","core_discovery":"The central claim is Theorem 4.2: for the fixed-endpoint problem $(P)$, if $x_0$ is a weak local minimum and $L,L_x,L_y,L_{xx},L_{xy},L_{yy}$ are continuous near its graph, then $\\langle P(t)r,r\\rangle\\ge 0$ for every $t$ at which the Caputo derivative of $x_0$ is continuous and every $r\\in\\mathbb{R}^n$, where $P(t)=L_{yy}(t,x_0(t),(cD_{t_0+}^\\alpha x_0)(t))$. The proof constructs a variation $h$ whose Caputo derivative equals $(f(t)-k)r$ on a small interval $[\\sigma-\\varepsilon,\\sigma+\\varepsilon]$ and vanishes outside; the constant $k$ is fixed by $h(t_1)=0$. Substituting this $h$ into the second variation, the paper shows that a negative value of $\\langle P(\\sigma)r,r\\rangle$ would make the leading negative term dominate the remaining terms for sufficiently small $\\varepsilon$, contradicting minimality. The same argument is repeated for the free-initial and fixed-final problem in Theorem 4.4.","pith_inferences":["If the Legendre condition does hold for all problems of this type, it yields a quick nonexistence check: any candidate extremal along which the second derivative of the Lagrangian in the derivative variable has a negative eigenvalue at a continuity point cannot be a weak local minimum, just as in the classical theory.","The power-law family appearing in the fractional Du Bois-Reymond lemma suggests that fractional first integrals may take a similar power-law form, which could be used to construct conserved quantities for fractional Euler-Lagrange equations.","Because the paper's needle variation has nonzero Caputo derivative only on a small interval, similar variations may be used to prove higher-order necessary conditions or to treat inequality constraints in fractional variational problems."],"forward_implications":["Fixed-endpoint fractional variational problems now have a second-order necessary condition of the same form as in the classical calculus of variations.","The Euler-Lagrange equations come in two regimes depending on the ordering of the two fractional orders, and ignoring this relation can produce systems with no admissible solution, as Example 4.1 shows.","For problems with a free right endpoint in the regime where the integral order exceeds the derivative order, a necessary condition for existence is that the partial derivative of the terminal cost with respect to the final value vanishes along the extremal.","The fractional Du Bois-Reymond lemma reduces to the classical lemma when both orders equal one and to an earlier fractional lemma when the two orders are equal, so the new proof contains those earlier results as special cases.","The paper's variation method is presented as a template that can be reused for other fractional variational problems under constraints."],"supporting_citations":[{"why":"The prior work whose claim that the standard proof cannot be adapted to final constraints is directly contradicted; the paper takes its problem class from here.","marker":"[9]"},{"why":"The doctoral dissertation cited as arguing that the natural fractional problem takes the two fractional orders equal; the paper's parameter analysis addresses this.","marker":"[10]"},{"why":"The earlier paper establishing that a Riemann-Liouville integral functional is needed for existence and proving the lemma for equal orders, which the present lemma generalizes.","marker":"[13]"},{"why":"The paper that first formulated the fractional Euler-Lagrange equation, which is the starting point for the first-order conditions derived here.","marker":"[27]"},{"why":"An early derivation of Euler-Lagrange equations for fractional variational problems, which the present integral-form equations extend to the full parameter range.","marker":"[2]"}],"fun_headline_variants":["Legendre condition holds for fractional derivatives","Fractional Euler-Lagrange gets Legendre condition","Classical proof adapts to Caputo fractional variations","No counterexample: Legendre condition survives fractional"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the Legendre condition rests on the unstated assumption that the needle test function $f$ deviates from its midpoint value by at least a fixed fraction of its maximum amplitude over a non-negligible portion of the tiny interval, whereas the paper only assumes $f$ is continuous and non-constant.","fun_headline_variants_meta":{"raw":{"variants":["Legendre condition holds for fractional derivatives","Fractional Euler-Lagrange gets Legendre condition","Classical proof adapts to Caputo fractional variations","No counterexample: Legendre condition survives fractional"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000207,"raw_usage":{"total_tokens":1389,"prompt_tokens":920,"completion_tokens":469,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":536,"completion_tokens_details":{"reasoning_tokens":410}},"tokens_in":536,"tokens_out":469,"duration_ms":5154,"temperature":1.0,"reasoning_tokens":410,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:52:48.185335+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the needle variation in Theorem 4.2, take $f(t)=1+\\delta\\cos(\\pi(t-\\sigma)/\\varepsilon)$ on $[\\sigma-\\varepsilon,\\sigma+\\varepsilon]$; then the integral of $(f(t)-f(c))^2$ over the needle is $O(\\delta^2\\varepsilon)$ rather than the $O(\\varepsilon)$ needed for estimate (19), so for small $\\delta$ the negative term no longer dominates and the contradiction argument fails, which would refute the paper's claim that the standard classical proof goes through without additional hypotheses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The prior work whose claim that the standard proof cannot be adapted to final constraints is directly contradicted; the paper takes its problem class from here."},{"cited_title":"1-183 (2020)","cited_arxiv_id":null,"evidence_quote":"The doctoral dissertation cited as arguing that the natural fractional problem takes the two fractional orders equal; the paper's parameter analysis addresses this."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The earlier paper establishing that a Riemann-Liouville integral functional is needed for existence and proving the lemma for equal orders, which the present lemma generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The paper that first formulated the fractional Euler-Lagrange equation, which is the starting point for the first-order conditions derived here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"An early derivation of Euler-Lagrange equations for fractional variational problems, which the present integral-form equations extend to the full parameter range."}],"review_version":1}