{"id":"dee750cb-c396-4d5b-a2c7-457f4c571cc6","arxiv_id":"2607.19641","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Relation semigroups whose domain is orthogonal to their fixed multivalued part reduce to classical operator semigroups; their generators are maximal anti-accumulative (contraction case) or anti-selfadjoint (unitary case).","lead":"The paper builds a semigroup theory for 'domain-multivalued orthogonal' linear relations, showing that such relation semigroups are just classical semigroups of bounded operators on a subspace plus a fixed multivalued part. It then repackages the Hille-Yosida and Lumer-Phillips theorems for these relations in terms of maximal anti-accumulative relations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Definition 2.3 equates 'densely d.m.o.' and 'maximal d.m.o.' (same equality), and the proof of Theorem 5.12 asserts domA=K^⊥; taken literally this forces every C0 generator to be bounded, so the Hille-Yosida/Lumer-Phillips theorems rest on an unstated density-vs-equality fix.","rationale":"The paper's main goal is to characterize generators of relation semigroups. The dictionary in Theorem 5.3 is valid for the class defined in Definition 5.1, and the restriction to a fixed multivalued part K is a deliberate modeling choice rather than an internal inconsistency. The weakest point is instead the formal handling of 'densely d.m.o.', because Theorems 5.12 and 5.17 are the paper's headline results and they rely on a condition whose two readings lead to different mathematical content. Definition 2.3 gives 'densely' and 'maximal' the same equality; if read literally, every C0 generator would have bounded operator part, contradicting the acknowledged unboundedness of C0 generators. The proofs repeat the equality domA=(mulT0)^⊥, so this is not a one-line typo but a systematic gap in the central argument. The proposed test with the Dirichlet Laplacian discriminates sharply between the two readings and checks whether the main theorems survive for unbounded generators. The reader already set a conditional verdict and flagged the dense/maximal ambiguity; this stress test sharpens that ambiguity into a concrete, load-bearing correctness concern without changing the verdict. The false Lemma 2.2(i) is a real flaw but is not the single most load-bearing issue for the central claim, since the Section 5 applications assume a common multivalued part where the product formula can be justified directly.","tokens_in":20866,"tokens_out":18224,"duration_ms":183397,"concrete_test":"Take K=0 and A=-Δ on L^2(0,1) with Dirichlet boundary conditions, a standard maximal anti-accumulative operator generating a contractive C0-semigroup. Check whether Theorem 5.12 and Theorem 5.17 accept A: under the literal Definition 2.3, domA≠(mulA)^⊥=H, so A is not 'densely d.m.o.' and the theorem excludes a textbook generator; under the density reading, the proof must replace the equality domA=K^⊥ by density. Rerunning the proof of Theorem 5.12 on this example with the corrected definition will settle which reading was intended and whether the stated theorems remain valid for genuinely unbounded generators.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central reduction (Theorem 5.3) is internally sound for the class it defines, but the main generator characterizations (Theorems 5.8, 5.12, 5.17) all depend on the notion of a 'densely d.m.o.' relation, and that notion is not coherently defined. In Definition 2.3, 'densely d.m.o.' and 'maximal d.m.o.' are given by the same displayed condition, domT=(mulT)^⊥. For a C0-generator A=A^⊙⊕K, however, domA=domA^⊙ is only dense in K^⊥, not equal to it. The proof of Theorem 5.11(i) and the proof of Theorem 5.12 both assert domA=domA^⊙=(mulT0)^⊥, directly contradicting Remark 5.10(ii), which correctly states that A^⊙ is a closed densely defined, generally unbounded, operator. If the equality is taken literally, every A satisfying the stated hypothesis has everywhere-defined closed operator part, hence is bounded by the closed graph theorem, and the theory excludes standard unbounded generators such as the Dirichlet Laplacian. If 'densely' is meant to allow closure, then the displayed equality and the repeated proof equality must be changed to \\(\\overline{\\mathrm{dom}\\,T}=(\\mathrm{mul}\\,T)^\\perp\\) and density statements, respectively. Because Theorem 5.17 and Corollaries 5.18–5.19 inherit this hypothesis, the headline Hille-Yosida/Lumer-Phillips characterization is not acceptable as written until this is resolved. A separate false Lemma 2.2(i) exists, but it is not the main blockage for the common-multivalued-part setting used in Section 5; it should still be corrected.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a theory of one-parameter semigroups for linear relations in Hilbert space, introducing the class of domain-multivalued orthogonal (d.m.o.) relations. It claims characterizations for generators of uniformly and strongly continuous semigroups: maximal d.m.o. relations generate uniformly continuous semigroups (Theorem 5.8), densely d.m.o. relations with Hille–Yosida resolvent bounds generate C0-semigroups (Theorem 5.12), and maximal anti-accumulative relations generate C0-semigroups with contractive operator part (Theorem 5.17), with unitary analogs (Corollary 5.18). The central reduction, Theorem 5.3, asserts that a family of closed maximal d.m.o. relations with common multivalued part is a semigroup iff its operator parts form a classical operator semigroup on the orthogonal complement of the multivalued part.","tokens_in":21326,"tokens_out":10847,"duration_ms":84942,"significance":"If the central theorems were correct, the d.m.o. framework would provide a clean reduction of semigroups of linear relations to classical operator semigroups on the orthogonal complement of the multivalued part, and the Lumer–Phillips analog (Theorem 5.17) would be a useful result. The paper uses standard tools from [5, 17, 19] and contains explicit statements of resolvent estimates and generator formulas. However, the current manuscript contains a definitional collapse (Section 2, Definition 2.3) and a false lemma (Lemma 2.2(i)) that directly affect the main theorems. The central claims are defensible in spirit, but the presentation is not acceptable as written and requires substantive repair.","major_comments":[{"comment":"The definitions of 'densely d.m.o.' and 'maximal d.m.o.' are identical as stated: both are given by domT=(mulT)^⊥. This conflation is not a typo: the proofs of Theorem 5.11(i) and Theorem 5.12 use the equality to conclude domA=domA⊙=(mulT0)^⊥, which would force the generator's operator part to be everywhere defined and hence bounded by the closed graph theorem. That contradicts Remark 5.10(ii), where A⊙ is correctly stated to be a closed densely defined, generally unbounded operator, and excludes standard examples such as the Schrödinger operator in §6.2. The intended notion for C0-generators is evidently density: 'densely d.m.o.' should mean domT is dense in (mulT)^⊥, while 'maximal d.m.o.' means equality. All subsequent statements and proofs involving this definition must be re-evaluated and corrected accordingly.","section":"Section 2, Definition 2.3"},{"comment":"Lemma 2.2(i) is false as stated. The proof claims that from (h,k)∈TS and the hypothesis domT∩mulS={0} it follows that the intermediate vector f lies in the domain of S⊙. This is not valid: membership in S⊙ requires f⊥mulS, while domT∩mulS={0} only excludes f∈mulS; f may have a nonzero component in mulS. A concrete counterexample: in H=C^2 take S=span{(e2,e2),(0,e1)} and T=span{((e2+e1),e2)}. Then S is closed, domT∩mulS={0}, but (e2,e2)∈TS while (e2,e2)∉TS⊙. Since Lemma 2.2(i) is used in Theorem 2.4(ii) and hence in the proof of Theorem 5.3, the reduction to the operator part needs a corrected hypothesis (e.g., ranS∩domT⊂(mulS)^⊥) and a revised proof.","section":"Lemma 2.2(i)"},{"comment":"The proof of Theorem 5.12 contains the line 'domA=domA⊙=(mulT0)^⊥=(mulA)^⊥'. For a C0-semigroup whose generator has unbounded operator part, this equality is false; in fact domA⊙ is only dense in (mulT0)^⊥. Consequently the forward direction of the Hille–Yosida proof is invalid, and the theorem's statement 'A is densely d.m.o.' cannot be interpreted under the current Definition 2.3 without forcing A to be bounded. Since Theorem 5.17 and Corollaries 5.18–5.19 inherit this hypothesis through Theorem 5.12, the headline Lumer–Phillips and unitary characterizations are not acceptable as written. The statement may be salvageable by changing 'densely d.m.o.' to mean density and by using the resolvent bounds of the operator part, but the proofs must be rewritten.","section":"Theorem 5.12 and Corollaries 5.17–5.19"},{"comment":"Definition 5.1 restricts the semigroup to closed maximal d.m.o. relations with a common multivalued subspace K for all t. By Remark 5.2, this forces each Tt⊙ to be a bounded operator on K^⊥. Thus the theory covers semigroups whose elements are bounded-operator lifts with a fixed multivalued part, but not genuinely unbounded or time-varying multivalued dynamics. The paper does not compare this class with the m-dissipative relation semigroups of Arendt–Chalendar–Moletsane [1], where the multivalued part of the generator plays a more essential role. The claimed extension of the Hille–Yosida and Lumer–Phillips theorems should be framed as applying to this restricted class; as written, the scope is overstated.","section":"Definition 5.1 and scope"}],"minor_comments":[{"comment":"The notation 'TT', 'LL', 'AA', 'SS', 'BB' is used inconsistently; it appears to denote the operator part (T⊙, L⊙, etc.) but is never defined. This makes statements such as Theorem 3.5 ('with LL a maximal dissipative operator') and Theorem 5.12 ('(ω,∞)⊂A A') difficult to parse. Please define and use a consistent notation for the operator part.","section":"Throughout §§3–5"},{"comment":"Typo: 'if if there exists' should be 'if there exists'.","section":"Proof of Theorem 5.3"},{"comment":"Typo: 'a a semigroup' should be 'a semigroup'.","section":"Corollary 5.4"},{"comment":"The word 'exits' should be 'exists'.","section":"Equation (5.2)"},{"comment":"The proof lists items (i), (ii), (iii), then repeats '(ii):' for the final item; it should be '(iv):'.","section":"Theorem 5.11 proof"},{"comment":"'Lema 3.4' should be 'Lemma 3.4'.","section":"Theorem 4.5 proof"},{"comment":"The statement '(ω,∞)⊂A A' in the proof of Theorem 5.12 should almost certainly be '(ω,∞)⊂ρ(AA)' or '(ω,∞)⊂ρ(A⊙)'; the typo makes the argument hard to follow.","section":"Remark 5.10 and Theorem 5.12"}],"recommendation":"major_revision","confidential_remarks":"The paper has a promising central idea—reducing semigroups of d.m.o. relations to operator semigroups—and the final examples are relevant. However, the definitional collapse in Definition 2.3 and the false Lemma 2.2(i) are load-bearing and currently invalidate the main theorems as stated. The authors should also be asked to clarify how their d.m.o. class relates to the existing m-dissipative relation semigroup theory of [1]. Given that the core reduction is likely repairable, I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: there is a load-bearing bug in the definition of “densely d.m.o.” that makes the main Hille–Yosida/Lumer–Phillips claims false as written. The paper is otherwise a careful dictionary between a restricted class of relation semigroups and classical operator semigroups, and the cleanest part—Theorem 5.3—is worth preserving.\n\nWhat is actually good: the d.m.o. notion and the reduction Theorem 5.3 are neat; once you assume a constant multivalued subspace K and each T_t maximal d.m.o., the operator parts T_t^⊙ form a semigroup in B(K^⊥), and conversely any such operator semigroup lifts. That makes Sections 5.1 and the uniform case (Theorem 5.8) essentially sound. The characterizations of dissipative/accumulative and symmetric/selfadjoint relations in terms of operator parts (Theorems 3.5, 3.11, 4.5, 4.12) are useful packaging, and the one-dimensional perturbation examples in Section 6 give concrete generators. The authors cite [1] and don’t oversell the history.\n\nSoft spots, in order of size. First and decisive: Definition 2.3 gives the same equality, dom T = (mul T)^⊥, for “densely d.m.o.” and “maximal d.m.o.” That can’t be right. For a C0-semigroup generator A = A^⊙ ⊕ K, you only have dom A dense in K^⊥, not equal; Remark 5.10(ii) even says A^⊙ is generally unbounded. But the proofs of Theorems 5.11(i) and 5.12 assert dom A = (mul A)^⊥. Taken literally, that forces A^⊙ to be everywhere defined and hence bounded, which excludes elementary generators like the Dirichlet Laplacian and makes the Hille–Yosida statements vacuous outside the bounded case. The fix is to define “densely d.m.o.” as \\overline{dom T} = (mul T)^⊥ and replace those equalities by density statements. Until that is done, Theorems 5.12, 5.17, and Corollaries 5.18–5.19 are not acceptable.\n\nSecond, Lemma 2.2(i) is false as stated: from (h,f)∈S and f∈domT you can’t conclude (h,f)∈S^⊙, because f need not lie in (mulS)^⊥. It needs a stronger hypothesis or a counterexample. It doesn’t block the main section, but it should be fixed.\n\nThird, the paper never positions itself against [1], which already extends Hille–Yosida/Lumer–Phillips to m-dissipative linear relations. Given Theorem 5.3, the main characterizations are mostly Pazy’s theorems restated for the operator part on K^⊥. That’s fine as a packaging exercise, but the novelty claim needs to be moderated, and the authors should say exactly which cases [1] does not cover (constant multivalued part? contractive operator part?).\n\nBottom line: worth a serious referee, not for the current version. The reduction theorem and examples deserve to be in the literature; the generator theorems need a definitional fix and a more careful comparison with prior work. I’d send it to review conditionally, and I wouldn’t cite the HY/LP claims until the dense/maximal issue is resolved.","headline":"Clean dictionary between a restricted class of relation semigroups and operator semigroups, but a definitional bug in 'densely d.m.o.' makes the main Hille–Yosida/Lumer–Phillips claims false as written.","tokens_in":21830,"tokens_out":6218,"would_cite":false,"duration_ms":55244,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47A06","47D06","47D60"],"pacs":[],"model":"deepseek-v4-flash","headline":"A family of closed maximal d.m.o. linear relations is a semigroup if and only if its operator parts form a classical operator semigroup, reducing relation semigroup theory to the classical case.","keywords":["linear relations","domain-multivalued orthogonal relations","C0-semigroups","infinitesimal generators","anti-accumulative relations","Hille-Yosida theorem","Lumer-Phillips theorem","multivalued operators"],"falsifier":"Check the resolvent equality ||R_A(λ)|| = ||R_{A_⊙}(λ)|| for a concrete relation A = A_⊙ ⊕ K with K ≠ {0}; for example, take H = C^2, K = span{(0,1)}, and A_⊙ = [[0,1],[-1,0]] on the first coordinate, computing both norms at λ=1. If they differ—say by using a non-d.m.o. relation where mul A is not orthogonal to dom A—the reduction in Theorem 2.7 breaks, demonstrating that the d.m.o. condition is essential and that such a relation cannot be part of the semigroup theory developed here.","tokens_in":20737,"feed_emoji":"🔄","tokens_out":11376,"duration_ms":92533,"temperature":0.7,"pith_summary":"Linear relations—multivalued linear operators—arise naturally in singular differential equations, boundary value problems, and adjoints of non-densely defined operators. This paper develops a semigroup theory for them by restricting to domain-multivalued orthogonal (d.m.o.) relations, whose domains are orthogonal to their multivalued parts. The central result is a reduction: a family of closed maximal d.m.o. relations with a common multivalued subspace K is a semigroup exactly when the operator parts form a semigroup of bounded operators on K^⊥. This reduction lets the authors translate the classical generation theorems: generators of uniformly continuous semigroups are maximal d.m.o.; generators of C0-semigroups are densely d.m.o. with a positive resolvent ray; and generators of C0-semigroups of contractions are precisely maximal anti-accumulative relations. The paper thus gives a complete characterization of relation semigroups, with the multivalued part appearing as a fixed vertical component.","feed_headline":"Relation semigroups are just operator semigroups in disguise","feed_subtitle":"Fix the multivalued subspace, and Hille-Yosida and Lumer-Phillips carry over unchanged.","key_machinery":"The central object is the d.m.o. relation and its canonical decomposition T = T_⊙ ⊕ T_∞ into an operator part (a closed linear operator on (mul T)^⊥) and a purely multivalued part T_∞ = {0}⊕mul T. For closed maximal d.m.o. relations, dom T = (mul T)^⊥, so T_⊙ is everywhere defined on that subspace and, by the closed graph theorem, bounded. Theorem 2.4 shows that products of such relations decompose in the same way, which is the mechanism: the semigroup law T_{t+s} = T_t T_s becomes the operator semigroup law T_{t+s}^⊙ = T_t^⊙ T_s^⊙, reducing the entire theory to classical bounded-operator semigroups.","core_discovery":"On a Hilbert space, a linear relation is a subspace of H⊕H; a relation is d.m.o. when its domain is orthogonal to its multivalued part. The paper proves that every semigroup {T_t} of closed maximal d.m.o. relations with common multivalued part K decomposes as T_t = T_t^⊙ ⊕ T_0∞, where {T_t^⊙} is a C0-semigroup (or uniformly continuous semigroup, as appropriate) of bounded operators on K^⊥. Consequently the infinitesimal generator of the relation semigroup is A = A_⊙ ⊕ T_0∞, the sum of the generator of the operator semigroup and the constant multivalued part. The main generation theorems state that a closed relation generates a uniformly continuous semigroup iff it is maximal d.m.o.; a strong","pith_inferences":["The constant-multivalued-subspace assumption is more than a convenience: it forces the operator parts to be bounded, so the whole theory lives on K^⊥. Relaxing it to a time-varying multivalued subspace would break the reduction, and the paper's examples suggest that such generality may be needed for systems with changing constraints.","Because generators are characterized geometrically as maximal anti-accumulative relations, the framework opens a path to perturbation and approximation results for relation semigroups, mirroring the operator case; the paper does not pursue these.","The one-dimensional perturbation construction—where a multivalued relation emerges only at the parameter value τ=∞—could be transferred to concrete quantum-mechanical models, suggesting a physical interpretation of the multivalued part as an ideal boundary or zero-capacity degree of freedom."],"forward_implications":["The Hille-Yosida theorem holds verbatim for relations: a closed relation generates a C0-semigroup of relations with growth bound Me^{ωt} iff it is densely d.m.o., (ω,∞) ⊂ ρ(A), and ||R_A(λ)^n|| ≤ M/(λ−ω)^n for all n.","Lumer-Phillips for relations: a closed relation generates a C0-semigroup whose operator parts are contractions iff it is maximal anti-accumulative; this is a purely dissipative-geometric generator condition.","Unitary analogue of Stone's theorem: generators of C0-semigroups with unitary operator parts are exactly anti-selfadjoint relations; equivalently T_t = e^{itL} for a maximal dissipative or selfadjoint relation L, computed via Yosida approximation.","The generator of any d.m.o. relation semigroup is unique and has the form A_⊙ ⊕ T_0∞, so the multivalued part is a constant, inactive component that does not affect evolution on the domain.","The one-dimensional perturbation constructions show how to build explicit contractive and unitary relation semigroups from differential operators with boundary conditions, and classify all maximal anti-dissipative and anti-accumulative extensions."],"fun_headline_variants":["Relation semigroups decompose into operator semigroups","Semigroups on d.m.o. relations: operators plus a constant","Hille-Yosida and Lumer-Phillips extend to relation semigroups","Relation semigroup generators split into operator and constant parts"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is Definition 5.1, which fixes the multivalued subspace K for all times and requires each relation to be closed and maximal d.m.o.; this is what makes the operator part everywhere defined and bounded, collapsing the semigroup theory to the classical bounded case.","fun_headline_variants_meta":{"raw":{"variants":["Relation semigroups decompose into operator semigroups","Semigroups on d.m.o. relations: operators plus a constant","Hille-Yosida and Lumer-Phillips extend to relation semigroups","Relation semigroup generators split into operator and constant parts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0014,"raw_usage":{"total_tokens":5465,"prompt_tokens":678,"completion_tokens":4787,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":422,"completion_tokens_details":{"reasoning_tokens":4713}},"tokens_in":422,"tokens_out":4787,"duration_ms":28985,"temperature":1.0,"reasoning_tokens":4713,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T12:11:01.057954+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the resolvent equality ||R_A(λ)|| = ||R_{A_⊙}(λ)|| for a concrete relation A = A_⊙ ⊕ K with K ≠ {0}; for example, take H = C^2, K = span{(0,1)}, and A_⊙ = [[0,1],[-1,0]] on the first coordinate, computing both norms at λ=1. If they differ—say by using a non-d.m.o. relation where mul A is not orthogonal to dom A—the reduction in Theorem 2.7 breaks, demonstrating that the d.m.o. condition is essential and that such a relation cannot be part of the semigroup theory developed here.","supporting_citations":[],"review_version":1}