{"id":"91217aa7-3f67-4523-aac0-2f3c96fdca31","arxiv_id":"1908.03027","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Positive linear maps that converge on a small test set of continuous functions converge on the whole space, even when the limit is a nonlinear isometry.","lead":"The authors prove new Korovkin-type theorems: a sequence of positive linear operators that converges on a small test set of continuous functions converges on the whole space, even when the limiting map is a nonlinear isometry. This extends a classical benchmark result in approximation theory and gives a unified route to pointwise and uniform convergence criteria.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Step 2 invokes [9, Thm 2.3] without verifying its hypotheses; if that theorem requires T∞(M) to separate points of Y, the proof of linearity of T∞ is incomplete and Theorem 3.1 lacks the needed assumption.","rationale":"The reader's weakest assumption is the same one I would flag: the imported classification theorem. The novelty of the paper is precisely that T∞ is not assumed linear; Step 2 is the bridge from pointwise convergence on S to linearity of T∞ on all M. A missing hypothesis in the cited theorem is not a cosmetic issue. However, I do not see an internal contradiction in the paper's own arguments assuming [9] applies; Step 1 is plausible from Lemma 2.2, and the boundary argument in Step 3 is standard once linearity is available. The non-separating range issue is the sharpest form of the concern: the paper's definition of function space requires separation, and T∞(M) is never shown to satisfy it. A quotient fix is likely, which is why conditional acceptance (not rejection) is appropriate. If the authors verify the hypotheses or add a sentence about quotienting, the paper can be accepted.","tokens_in":8823,"tokens_out":34099,"duration_ms":375892,"concrete_test":"Obtain the statement of [9, Theorem 2.3] and check two items: (1) Does it require the codomain F to be a function space on Y, i.e. to separate points of Y? (2) Does Step 1's triple-separation property for Ch(M) match a hypothesis of that theorem? Then run the following analytical check: let X=[0,1], Y=X×{0,1}, M=C(X), and T∞f(x,i)=f(x). This T∞ is an isometry from M onto a subspace that does not separate Y, and S=M satisfies all other assumptions of Theorem 3.1. If [9, Theorem 2.3] excludes this example, Step 2 needs an additional quotient argument; verify that the quotient argument can reproduce the representation on Ch(T∞(S)) and the conclusion T∞1=1. If it cannot, Theorem 3.1 as stated needs an extra hypothesis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Step 2 is the only place where the paper proves the decisive fact that T∞ is complex-linear. The proof is: Mazur–Ulam gives real-linearity, then 'taking into account Step 1, from [9, Theorem 2.3] it follows that T∞1=1 and ...' The hypotheses of [9, Theorem 2.3] are never stated. The paper defines 'function space' to mean a subspace separating points of the underlying compact space. M is a function space on X because it contains the test space S. But T∞(M) is only assumed to be a subspace of C(Y); nothing in Theorem 3.1 guarantees that T∞(M) separates points of Y. If [9, Theorem 2.3] requires its codomain to be a function space on Y, it cannot be applied to T∞:M→T∞(M). A concrete failure mode is any quotient-type isometry T∞f=f∘π with π:Y→X continuous surjective and non-injective: the range is isometric to M but does not separate Y. The same step also leaves unstated how the theorem yields T∞1=1 from positivity and why i∈S is used. Since Step 3 invokes [7, Lemma 2.5] and part (b) uses the quotient construction, both depend on Step 2. Thus the central claim is not fully justified as written unless [9]'s hypotheses are verified or the proof is repaired by quotienting T∞(M).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Korovkin-type convergence for sequences of positive linear maps T_n from a self-conjugate subspace M of C(X) into C(Y). The limit map T∞ is assumed only to be an isometry from M onto its image, not a linear operator. Theorem 3.1(a) claims pointwise convergence of T_n f to T∞f on the Choquet boundary of T∞(S) whenever this convergence holds on the test space S; part (b) uses a boundary inclusion plus boundedness of {T_n 1} to extend convergence to all of Y. Theorem 3.2 gives the uniform version, Corollary 3.4 recovers earlier results of Hachiro and Okayasu, and the examples include classical Korovkin-type statements.","tokens_in":9148,"tokens_out":24983,"duration_ms":284499,"significance":"If correct, the paper broadens the usual Korovkin framework in two useful ways: the approximating maps need only be positive rather than unital contractions, and the limit map need not be assumed linear. The use of the quotient by N and of extreme-boundary measures in Theorem 3.1(b) is a clever device, and the examples illustrate the intended scope. The contribution is potentially interesting, but the manuscript as written cannot be certified: the proof of the decisive linearity step depends on an unverified application of an external classification theorem, and one later passage in Step 3 uses the Choquet boundary in a way that is not justified.","major_comments":[{"comment":"The proof of the linearity of T∞ is incomplete as written. After Mazur-Ulam gives real-linearity, the manuscript invokes [9, Theorem 2.3] without stating its hypotheses and without verifying that the pair (M, T∞(M)) satisfies them. In particular, if that theorem requires the codomain to be a function space on Y, or a strongly separating subspace, then the conclusion cannot be applied to T∞:M→T∞(M): an isometry of the form T∞f = f∘π with π:Y→X continuous, surjective and non-injective has range isometric to M but this range does not separate the fibres of π. Since Step 2 is the only place where T∞1 = 1 and complex-linearity of T∞ are established, and since Step 3 and part (b) both use that linearity, the central claim of Theorem 3.1 is not justified unless the hypotheses of [9, Theorem 2.3] are stated and checked. Theorem 3.2 inherits the same dependence.","section":"Theorem 3.1, Step 2"},{"comment":"The passage 'it is observed that the above relation holds for all z, z′ ∈ Y' is not justified. The displayed inequality before this sentence is established only for z′ ∈ Ch(T∞(M)); the Choquet boundary property of T∞(M) does not by itself imply that a real-valued function in T∞(M) is nonnegative on all of Y whenever it is nonnegative on Ch(T∞(M)). Moreover, the final estimate is later applied at y′ ∈ Ch(T∞(S)), which need not lie in Ch(T∞(M)): for a subspace inclusion the Choquet boundary of the smaller space can be strictly larger. Thus the proof needs an explicit argument for this step, or the final estimate needs to be derived directly with z′ = y′.","section":"Theorem 3.1, Step 3"}],"minor_comments":[{"comment":"The displayed representation of T∞ writes f∘φ in both alternatives on K and on its complement; one of the two cases should be the conjugate function −̅{f∘φ} or an equivalent expression.","section":"Theorem 3.1, Step 2"},{"comment":"In the displayed equicontinuity estimate, the term η T∞ f_{y′}(y) appears where a real part such as η Re T∞ f_{y′}(y) is expected, and the symbol η is defined only after it is used; please reorder and correct the displayed inequality.","section":"Theorem 3.2, Proof of Claim"},{"comment":"The notation R̂{Ch}(T∞(S)) is used without definition; please define it as the image of Ch(T∞(S)) under the quotient map and explain the inclusion Ch(R̂N) ⊆ R̂{Ch}(T∞(S)).","section":"Theorem 3.1(b)"},{"comment":"The peak function for z0 ∈ T is not h(z) = (z + z0)/2, since h(z0) = z0; it should be (1 + ̅z0 z)/2, or an equivalent rotation, in order to satisfy h(z0) = 1 and |h(z)| < 1 for z ≠ z0.","section":"Example 4.3"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the main issue is verification. The authors should be asked to state the precise form of [9, Theorem 2.3] and to verify its hypotheses for T∞(M), or to replace Step 2 with a self-contained argument. If the theorem requires the range to separate points, the statement of Theorem 3.1 may need an additional hypothesis or a quotient-based modification. The Step 3 boundary passage also needs a rigorous justification."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real result here is Theorem 3.1 and its uniform companion Theorem 3.2: they move Korovkin convergence from unital contractions to arbitrary positive linear maps, and allow the limit to be a non-linear isometry. That is a legitimate extension of [7], not a repackaging. The proof strategy is the classical Korovkin barrier argument dressed up with Choquet boundary techniques, and for real-valued M it is mostly self-contained and convincing. The examples are standard illustrations, but they do show the intended scope.\n\nThe main soft spot is exactly where the stress-test note lands: Step 2 of Theorem 3.1, the complex case. The paper invokes [9, Theorem 2.3] without stating its hypotheses. The reader is asked to accept that a real-linear isometry between function spaces has the form f∘φ (or its conjugate), and that T∞1=1 follows from positivity. But T∞(M) is never shown to separate the points of Y, and the classification theorem for real-linear isometries typically assumes both domain and codomain are function spaces on their respective compacta. A quotient isometry f↦f∘π with π:Y→X continuous, surjective, but not injective would satisfy the hypotheses of Theorem 3.1 yet fail to separate points on Y, and [9] may not apply. That is a load-bearing gap, because everything downstream—the representation used in Step 3, the positivity argument, and the conclusion that T∞ is complex-linear—depends on Step 2. It may be repairable by passing to the quotient Y/∼ as in part (b), or by adding an explicit hypothesis that T∞(M) separates Y, but as written the complex-case proof is incomplete.\n\nSmaller issues: Example 4.2 has a malformed formula (the indices on b are off), and there are a few notational slips in the inequalities of the proofs. None of that is fatal, but it is the kind of thing that should be cleaned before publication. The citation pattern is fine; the authors cite the relevant Korovkin literature and the isometry classification papers.\n\nWho gets value from this paper: approximation theorists and anyone using Korovkin-type convergence on C(X). The real-scalar case is solid enough to stand alone; the complex case needs the gap closed or at least a clear statement of [9, Thm 2.3] and verification of its hypotheses.\n\nI would send it to a serious referee. The paper deserves referee time, but I would expect the referee to push on Step 2 and probably require a repair before acceptance.","headline":"A genuine extension of Korovkin-type results, with a solid real-scalar proof and a genuine gap in the complex case: the proof calls on an unstated classification theorem whose hypotheses are not checked.","tokens_in":9670,"tokens_out":2371,"would_cite":false,"duration_ms":27165,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["41A36","46E15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves generalized Korovkin-type theorems: pointwise or uniform convergence of positive linear maps on a test function space forces convergence on the whole space, even when the limit is only assumed to be an isometry.","keywords":["Korovkin theorem","positive linear maps","Choquet boundary","function space","isometry","pointwise convergence","uniform convergence","approximation theory"],"falsifier":"One concrete test: construct a real-linear isometry $T_\\infty$ between subspaces of $C(X)$ with $T_\\infty 1 \\ge 0$ but $T_\\infty i \\ne iT_\\infty 1$, together with positive linear maps $T_n$ that converge to it on the test space $S$. If the hypotheses of Theorem 3.1 hold yet convergence fails outside $S$, the linearity step—and with it the appeal to the classification of real-linear isometries—is the point to inspect. A simpler check of that step is to verify whether the cited classification theorem actually applies to arbitrary self-conjugate unital subspaces $M$ used here.","tokens_in":8630,"feed_emoji":"📈","tokens_out":10115,"duration_ms":94757,"temperature":0.7,"pith_summary":"Approximation theory's classical Korovkin theorem says that positive linear operators that converge on the handful of test functions $1,x,x^2$ converge on every continuous function. This paper proves a broad extension: a sequence of positive linear maps from a function space $M$ to $C(Y)$ that converges pointwise on a smaller test space $S$ to an isometry $T_\\infty$—which need not be assumed linear—also converges pointwise on the Choquet boundary of $T_\\infty(S)$ for every $f\\in M$. With two mild extra conditions, $\\operatorname{Ch}(N)\\subseteq\\operatorname{Ch}(T_\\infty(S))$ and boundedness of $\\{T_n1\\}$, the convergence extends to all of $Y$; the uniform version holds on compact subsets of the boundary, and globally when the boundary is compact. The authors establish this by first showing the limit isometry is actually linear and has the form $T_\\infty f=f\\circ\\varphi$, then using positivity and peak test functions to push convergence from $S$ to all of $M$.","feed_headline":"Few test functions force convergence to non-linear isometries","feed_subtitle":"A generalized Korovkin theorem: checking a small space spreads pointwise and uniform convergence to all continuous functions.","key_machinery":"The engine is the Choquet boundary $\\operatorname{Ch}(S)$ together with the peak-function criterion of Lemma 2.2: a point $x_0$ lies in $\\operatorname{Ch}(S)$ exactly when one can find functions in $S$ that are small at $x_0$ and large outside any neighbourhood of it. The proof uses this to build the test functions $f_{y'}$ that sandwich $f$ through the tensor-product positivity inequality $|(T_n\\otimes_{T_n1}T_\\infty)(F)| \\le (T_n\\otimes_{T_n1}T_\\infty)(1\\otimes\\epsilon+\\|F\\|\\,\\operatorname{Re} F_{y'})$. Linearity of $T_\\infty$ is forced via the Mazur–Ulam theorem and a classification of real-linear isometries; a boundary-measure representation then carries convergence from $\\operatorname{Ch}(T_\\infty(S))$ to all of $Y$, and the Arzelà–Ascoli compactness criterion upgrades pointwise to uniform convergence on compact subsets.","core_discovery":"The central discovery is Theorem 3.1: under the stated hypotheses, the limit isometry $T_\\infty$ is forced to be a linear isometry of the form $T_\\infty f = f\\circ\\varphi$ on the Choquet boundary of $T_\\infty(M)$, and once that representation is available, a positivity sandwich forces convergence from $S$ to all of $M$. Part (a) states $T_nf\\to T_\\infty f$ on $\\operatorname{Ch}(T_\\infty(S))$ for every $f\\in M$; part (b) upgrades this to all of $Y$ when $\\operatorname{Ch}(N)\\subseteq\\operatorname{Ch}(T_\\infty(S))$ and $\\{T_n1\\}$ is bounded. Theorem 3.2 is the uniform analogue: uniform convergence on $S$ gives uniform convergence on compact subsets of $\\operatorname{Ch}(T_\\infty(S))$, and global uniform convergence when $\\operatorname{Ch}(T_\\infty(S))$ or $\\operatorname{Ch}(N)$ is compact. The examples show the classical Korovkin theorem, its complex variants on the torus and disk, and a smooth multivariate version all follow from the same mechanism.","pith_inferences":["The paper does not pursue rates, but the proof's explicit $\\epsilon$ control suggests that a quantitative version—uniform modulus of convergence on compact subsets in terms of the test-space convergence—can be extracted from the inequalities in Theorem 3.2.","Because Step 2 is the only place the isometry is shown to be linear, a promising stress test is to search for real-linear isometries of subspaces that satisfy $T_\\infty 1\\ge0$ but are not complex-linear; if one is also a pointwise limit of positive linear maps, it would mark exactly where the classification hypothesis bites.","The transfer mechanism via tensor-product positivity is not tied to $C(X)$ in any essential way, so analogues should hold for positive maps on other Banach function spaces that admit a Choquet-type boundary, such as spaces of differentiable functions or Lipschitz spaces."],"forward_implications":["The classical theorem for $C[0,1]$ follows as the case $M=S=\\operatorname{Span}\\{1,x,x^2\\}$ with $T_n$ converging to the identity, so the quadratic test functions alone force convergence for every continuous function.","Complex Korovkin-type statements on the torus and disk are recovered: convergence of $T_n1\\to1$ and $T_nz\\to z$ forces convergence on $C(\\mathbb{T})$, and adding $T_n|z|^2\\to|z|^2$ does the same on $C(\\mathbb{D})$.","For smooth functions $D_K$ on a compact subset of $\\mathbb{R}^p$, convergence on $1$, the coordinate projections $P_k$, and $\\sum P_k^2$ forces convergence for every $C^\\infty$ function on $K$.","Pointwise convergence on $S$ propagates to all of $Y$ when $\\operatorname{Ch}(N)\\subseteq\\operatorname{Ch}(T_\\infty(S))$ and $\\{T_n1\\}$ is bounded; uniform convergence on $S$ propagates uniformly on compact subsets of $\\operatorname{Ch}(T_\\infty(S))$, and globally when the relevant Choquet boundary is compact.","The proofs work for nets as well as sequences, so the results are sequential only in presentation."],"supporting_citations":[{"why":"Supplies the earlier Korovkin-type theorem for unital linear contractions and the proof skeleton (boundary measures, quotient by N) used in part (b).","marker":"[7]"},{"why":"Provides the classification of real-linear isometries between subspaces of continuous functions used to prove T∞ is linear and has the form f ↦ f∘φ.","marker":"[9]"},{"why":"Mazur–Ulam theorem, used to conclude T∞ is a real-linear isometry from the fact that it is an isometry with T∞0 = 0.","marker":"[10]"},{"why":"Browder's Lemma 2.2, the peak-function criterion for Choquet boundary points used to construct the test functions fy'.","marker":"[5]"},{"why":"Gives the continuous surjection φ with T∞f = f∘φ on the Choquet boundary, used in Step 3 and in Corollary 3.4.","marker":"[3]"},{"why":"Korovkin's original theorem, the classical statement this paper generalizes and the source of the test-space idea.","marker":"[8]"},{"why":"Bishop–de Leeuw representation of linear functionals by measures on extreme points, used to carry convergence from the boundary to all of Y.","marker":"[4]"},{"why":"Phelps' lectures, supplying the representing measure on the extreme-point set used in the dominated-convergence step.","marker":"[12]"}],"fun_headline_variants":["Korovkin with non-linear isometry limits: linearity emerges","Small test functions force linear isometry in Korovkin","Choquet boundary key for Korovkin convergence to isometries","Pointwise and uniform Korovkin results via Choquet boundary"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the external classification of real-linear isometries between subspaces of continuous functions applies to the self-conjugate unital subspaces M allowed in the theorem; the paper invokes it without restating its hypotheses, and if it does not apply, T∞ need not be linear and the representation T∞f = f∘φ—the engine of Steps 2 and 3—fails.","fun_headline_variants_meta":{"raw":{"variants":["Korovkin with non-linear isometry limits: linearity emerges","Small test functions force linear isometry in Korovkin","Choquet boundary key for Korovkin convergence to isometries","Pointwise and uniform Korovkin results via Choquet boundary"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000372,"raw_usage":{"total_tokens":1923,"prompt_tokens":809,"completion_tokens":1114,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":425,"completion_tokens_details":{"reasoning_tokens":1041}},"tokens_in":425,"tokens_out":1114,"duration_ms":12296,"temperature":1.0,"reasoning_tokens":1041,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:26:41.157366+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"One concrete test: construct a real-linear isometry $T_\\infty$ between subspaces of $C(X)$ with $T_\\infty 1 \\ge 0$ but $T_\\infty i \\ne iT_\\infty 1$, together with positive linear maps $T_n$ that converge to it on the test space $S$. If the hypotheses of Theorem 3.1 hold yet convergence fails outside $S$, the linearity step—and with it the appeal to the classification of real-linear isometries—is the point to inspect. A simpler check of that step is to verify whether the cited classification theorem actually applies to arbitrary self-conjugate unital subspaces $M$ used here.","supporting_citations":[{"cited_title":"Hachiro and T","cited_arxiv_id":null,"evidence_quote":"Supplies the earlier Korovkin-type theorem for unital linear contractions and the proof skeleton (boundary measures, quotient by N) used in part (b)."},{"cited_title":"Koshimizu, T","cited_arxiv_id":null,"evidence_quote":"Provides the classification of real-linear isometries between subspaces of continuous functions used to prove T∞ is linear and has the form f ↦ f∘φ."},{"cited_title":"Mazur and S","cited_arxiv_id":null,"evidence_quote":"Mazur–Ulam theorem, used to conclude T∞ is a real-linear isometry from the fact that it is an isometry with T∞0 = 0."},{"cited_title":"Browder, Introduction to Function Algebras , W","cited_arxiv_id":null,"evidence_quote":"Browder's Lemma 2.2, the peak-function criterion for Choquet boundary points used to construct the test functions fy'."},{"cited_title":"Araujo and J","cited_arxiv_id":null,"evidence_quote":"Gives the continuous surjection φ with T∞f = f∘φ on the Choquet boundary, used in Step 3 and in Corollary 3.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Korovkin's original theorem, the classical statement this paper generalizes and the source of the test-space idea."},{"cited_title":"Bishop and K","cited_arxiv_id":null,"evidence_quote":"Bishop–de Leeuw representation of linear functionals by measures on extreme points, used to carry convergence from the boundary to all of Y."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Phelps' lectures, supplying the representing measure on the extreme-point set used in the dominated-convergence step."}],"review_version":1}