{"id":"a195fc01-cdba-46f5-9399-c85759d7287e","arxiv_id":"2411.09976","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Joint moments of occupation times on the legs of a diffusion spider are computed recursively, with an explicit closed form for Bessel and Brownian spiders.","lead":"This paper gives recursive formulas and, for Bessel and Brownian spiders, explicit closed forms for the joint moments of time spent on each leg of a diffusion spider. The results make occupation-time statistics on star-shaped random processes directly computable.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed generality of The recursion in Theorem 3 and the closed form in Theorem 5 is not established: both proofs are written only for r=2, with higher r deferred to 'the same method can be repeated', leaving equation (32) for r≥3 as an unproved assertion.","rationale":"The reader's verdict is already CONDITIONAL and their rationale flags the same r=2-only derivational gap. I make that gap the single load-bearing concern because, unlike the cited Green-kernel and D_k formulas, it is an internal gap in the argument: if the missing induction cannot be filled, the central explicit formula (32) is unproved for exactly the range of r where it is claimed. The concern is not that the formulas look wrong: the r=2 Brownian cases I checked numerically reproduce known arcsine moments, and the self-similar recurrence is consistent for small exponents. The issue is epistemic: the paper's own proof stops at r=2, so the general theorem is presently an assertion rather than a demonstrated result. A symbolic check of (32) against (16) for r=3,4 would settle whether the generalization is true; if it is, the conditional can be upgraded. No verdict change is needed because the reader already conditioned on filling this gap.","tokens_in":19153,"tokens_out":14321,"duration_ms":146522,"concrete_test":"Use a computer algebra system to verify that the explicit RHS of (32) satisfies the recurrence (16) for r=3 and r=4, using D_k from (30), for all exponent vectors with n_i ≤ 4. Any mismatch refutes Theorem 5 as stated; a full match would close the r≥3 gap and justify upgrading the verdict. As a second confirmatory step, independently derive (30) from (13) for k=1,2 using the Bessel Green kernel and E_y(H0^{k-1}e^{-λH0}).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that joint occupation-time moments on any number r of spider legs are computable by recursion (Theorem 3) or by the explicit Stirling-number formula (32) for Bessel spiders (Theorem 5). In both cases the proof stops at r=2. The Theorem 3 proof says 'the method is the same for any higher values of r', and the Appendix A proof of Theorem 5 says the procedure 'can then readily be repeated for a larger value of r'. This is load-bearing because the r=2 proof is not a trivial induction: it relies on two nontrivial combinatorial identities (bottom of p.18) to collapse sums involving Stirling numbers. For r≥3 the analogous sums contain products of three Stirling factors, and no displayed identity or argument shows they collapse to the claimed form (32). Thus a reader cannot currently verify whether Theorem 5 is true for r≥3 or only for r=2. The formula also depends on the imported D_k expression (30) from the same authors' earlier paper [20] with a different normalization; that citation is not re-derived, so a normalization error there would also collapse Theorem 5. The special cases r=1,2 and the Brownian case check out, so the formulas are credible, but the full claimed generality is currently unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the joint moments of occupation times on the legs of a homogeneous diffusion spider. It presents four main results: Theorem 3, a recursive formula for the Laplace transform of the joint occupation-time moments for any number r of legs; Theorem 4, a formula for the moment generating function of the occupation times at an independent exponential time; Theorem 5, an explicit Stirling-number closed form for the joint moments in the Bessel spider case; and Theorem 6, a specialized closed form for the Brownian spider. The proofs of Theorems 3 and 5 are written out only for r = 2, with the extension to larger r asserted by repetition of the same method; Theorem 4 is proved for general R. The paper also connects the results to Yano's double Laplace transform formula and to Walsh Brownian motion on sectors.","tokens_in":19351,"tokens_out":3959,"duration_ms":42082,"significance":"If the stated formulas are correct, the paper offers a useful extension of the single-leg occupation-time results in [20] to joint occupation times on spider legs, and the explicit formula (32) for Bessel spiders is elegant and likely to be of interest to researchers working on diffusions on graphs. The proof strategy is sound for the r = 2 case: the recursion is derived from Kac's moment formula and the explicit Green kernel (7), without circularity, and Theorem 4 is proved for general R in a clean manner. The special cases r = 1 and r = 2, and the Brownian specialization in Theorem 6, check out. However, the claimed generality of the central results is not currently supported: both Theorem 3 and Theorem 5 are proved only for two legs, and the text explicitly leaves the generalization to the reader. Since the r = 2 proof of Theorem 5 relies on nontrivial combinatorial identities involving Stirling numbers, the status of (32) for r ≥ 3 is an open question within the manuscript; similarly, the dependence of (16) on the r = 2 proof means the foundation for higher r is not fully documented.","major_comments":[{"comment":"Theorem 5 is stated for any r ∈ {1,...,R}, but the proof in Appendix A is carried out only for r = 2, ending with the statement that the procedure 'can then readily be repeated for a larger value of r'. This is a load-bearing gap: the r = 2 induction collapses the sums using two nontrivial combinatorial identities involving products of Stirling numbers of the first and second kind, and for r ≥ 3 the analogous expressions would contain products of three or more such Stirling factors. No displayed identity or argument shows that these higher-r expressions collapse to the claimed form (32). The formula for r ≥ 3 is therefore currently an unproved assertion. Please either provide a full proof for general r, or state Theorem 5 for r = 2 (with a clearly labeled conjecture for r ≥ 3).","section":"Appendix A / Theorem 5, Eq. (32)"},{"comment":"The proof of Theorem 3 stops at r = 2 and says 'Proving the result for r = 2 should be sufficient, since the method is the same for any higher values of r'. The structure of (15) is additive, so the extension to r > 2 is plausible, but the proof is not written out, and Remark 3 points out that the single-leg case r = 1 is not a special case of the theorem. Because Theorem 5's recursive derivation uses (16), the generality of Theorem 3 has direct consequences for the main closed-form result. I ask that a complete induction argument for general r be included, or that the theorem be stated with its proof for r = 2 and the extension as a remark.","section":"Section 4.2 / Theorem 3"},{"comment":"The formula (30) for D_k^{(i)}(λ) is imported from the authors' previous paper [20] with the note that a different normalization of m and S is used there, and Remark 2.1 additionally flags a sign change relative to [20]. Since Eq. (30) is an essential input to Theorem 5, and since a normalization or sign mismatch in this imported result would propagate into the main formula, the manuscript should include a self-contained derivation of (30) under the present normalization, or at least a detailed and explicit statement of the conversion between the normalizations. As it stands, a reader cannot verify the correctness of (30) without consulting [20] and reconstructing the change of variables.","section":"Section 6.1, Eq. (30)"}],"minor_comments":[{"comment":"In the sentence 'Restating the equation above with N = 2', the displayed formula contains a missing parenthesis: 'Ex (At(V1))n1At(V2))n2 )' should be 'E_x((A_t(V_1))^{n_1}(A_t(V_2))^{n_2})'.","section":"Section 3, proof of Proposition 1"},{"comment":"The phrase 'see also Theorem 4 in Barlow, Pitman and Yor [3], where where the formula is presented' contains a duplicated 'where'.","section":"Section 5, before Corollary 2"},{"comment":"The notation switches between R (number of legs of the spider) and r (number of legs with positive z_i) without explicit definition of how legs with z_i = 0 are handled; a brief clarification would avoid ambiguity, especially since (27) is then claimed for any r ≤ R.","section":"Corollary 2 and Eq. (27)"},{"comment":"The remark states that D_k^{(i)}(λ) cannot depend on λ for self-similar spiders, and this is used repeatedly; making the short induction explicit in the remark would improve readability.","section":"Remark 2.2"}],"recommendation":"major_revision","confidential_remarks":"The central formulas are credible and the r = 2 proofs are convincing, but the paper's advertised generality for r ≥ 3 is not established. The reliance on [20], including a normalization change and a sign change in the imported D_k^{(i)} formula, warrants a careful self-contained verification. The proofs for general r may be completed within the manuscript's scope, so I do not recommend rejection, but a major revision is needed to close the proof gaps."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Salminen and Stenlund extend their earlier one-dimensional occupation-time results to diffusion spiders. The genuinely new pieces are the recursive Laplace-transform formula for joint moments (Theorem 3), the exponential-killing MGF (Theorem 4), and the explicit Stirling-number formulas for Bessel spiders (Theorem 5) and Brownian spiders (Theorem 6). The r=2 derivations are carried out in real detail: the Kac-moment step, the strong-Markov decomposition at the first hit of zero, and the combinatorial collapse in Appendix A are all checkable. Theorem 6 is a nice explicit payoff, and Corollary 3 gives a simple sanity check. The paper also connects honestly to Yano's and Barlow-Pitman-Yor's earlier characterizations; it is positioned as an extension, not an overwrite.\n\nThe soft spot is exactly the one the authors flag: both Theorem 3 and Theorem 5 are proved only for r=2, with higher r deferred to 'the method is the same' or 'can readily be repeated.' For the recursion in Theorem 3 this is fairly benign, because adding a leg just adds terms of the same shape. For Theorem 5 it is a real gap. The r=2 proof depends on two nontrivial combinatorial identities to collapse sums over Stirling numbers; for r>=3 the analogous sums involve products of three Stirling factors, and no identity is displayed that reduces them to (32). The formula may well be true—the r=1 and r=2 cases, plus the Brownian specialization, all check out—but as written the claimed generality for arbitrary r is not demonstrated. A referee should ask for the general induction or for a statement that (32) is proved for r=2 and conjectured beyond.\n\nThe paper imports D_k from the authors' prior [20] without re-derivation. That is normal citation practice, and the sign/normalization change is explicitly noted. There is no data or code; this is pure analysis.\n\nBottom line: the machinery is sound, the r=2 results are solid, and the paper deserves a serious referee. The main requested revision is to tighten the r>=3 argument. I would cite it for the recursive scheme and for Theorem 6; I would bring it to reading group only if someone in the group is working on spider occupation times.","headline":"Useful and mostly careful extension of the authors' one-dimensional occupation-time results to spiders; the r=2 proofs are convincing, but Theorem 5's full claimed generality for r>=3 rests on 'same method can be repeated' rather than a displayed induction.","tokens_in":19958,"tokens_out":2851,"would_cite":true,"duration_ms":28720,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J60","60J55","60J65","05A10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes that the Laplace transform of the joint moments of occupation times on the legs of a homogeneous diffusion spider obeys an explicit recursion in terms of the spider's Green kernel, and that for Bessel…","keywords":["diffusion spider","Walsh Brownian motion","occupation times","Green's function","Kac's moment formula","Bessel process","Stirling numbers","self-similar diffusion"],"falsifier":"Compute the two-leg Brownian spider moment $E_0(A_1^{(1)}A_1^{(2)})$ with $\\beta_1=\\beta_2=1/2$ by direct Monte Carlo simulation of Walsh Brownian motion; the explicit formula (Corollary 3) gives $1/8$, so a result that differs from $1/8$ by more than sampling error would refute the paper's closed-form theorem.","tokens_in":18882,"feed_emoji":"🕷️","tokens_out":13338,"duration_ms":117794,"temperature":0.7,"pith_summary":"This paper aims to make the joint law of occupation times on the legs of a diffusion spider tractable. It proves a recursion that expresses the Laplace transform of any joint moment of the occupation times in terms of lower-order moments, using only the spider's Green kernel and the Laplace transforms of its hitting times of the origin. For a Bessel spider—where each leg has the radial law of a Bessel process with parameter $\\nu\\in(-1,0)$—the recursion is solved in closed form: every joint moment is a finite sum over Stirling numbers. The Brownian spider is the special case $\\nu=-1/2$, and the same results cover sector occupation times for Walsh Brownian motion. If the paper is right, occupation-time moments on these multiray graphs are no longer an iterative or simulation-only problem.","feed_headline":"Recursion gives joint occupation moments on diffusion spiders","feed_subtitle":"For Bessel spiders every moment is a finite sum of Stirling numbers; Brownian spiders are a special case.","key_machinery":"A diffusion spider is a star graph of $R$ legs carrying the same one-dimensional diffusion on each leg, with the process choosing leg $i$ with probability $\\beta_i$ at the vertex. The machinery is the spider's Green kernel $g_\\lambda$ (Theorem 1) together with the coefficients $D_k^{(i)}(\\lambda)=\\frac{\\lambda^k}{(k-1)!}\\int_0^\\infty g_\\lambda(0,(y,i))\\,E_{(y,i)}(H_0^{k-1}e^{-\\lambda H_0})\\,\\beta_i m(dy)$ from Eq. (13); these coefficients carry all the model data. The recursive step is the generalized Kac moment formula (Proposition 1), which breaks a product of occupation-time powers into a sum over legs and lower-order moments, with the integrals over the transition density collapsing onto the $D$ coefficients. For a self-similar spider, the scaling identity $A_t^{(i)}\\stackrel{d}{=}t A_1^{(i)}$ replaces Laplace transforms by direct recursions for moments at time 1. In the Bessel case the $D$ coefficients become explicit, $D_k^{(i)}=-\\beta_i\\binom{\\nu+k-1}{k}$, and the recursion solves to a finite Stirling-number sum.","core_discovery":"The paper's central claim is Theorem 3: for a homogeneous diffusion spider with $R$ legs and occupation times $A_t^{(i)}$, the Laplace transform of any joint moment satisfies\n$$L_t\\{E_0(\\prod_{i=1}^r ($A_t^{{(i)}}$)^{n_i})\\}(\\$\\lambda$)=\\sum_{i=1}^r\\sum_{k=1}^{n_i}\\frac{n_i!\\,$D_k^{{(i)}}$(\\$\\lambda$)}{(n_i-k)!\\,\\$\\lambda$^k} L_t\\{E_0(\\prod_{j=1}^r ($A_t^{{(j)}}$)^{n_j}/($A_t^{{(i)}}$)^k)\\}(\\$\\lambda$),$$\nwith $D_k^{(i)}(\\lambda)$ defined by Eq. (13) through the Green kernel. For self-similar spiders this becomes the direct recursion (16) for $E_0(\\prod_i (A_1^{(i)})^{n_i})$. The paper then evaluates the $D$ coefficients for Bessel spiders, obtaining $D_k^{(i)}=-\\beta_i\\binom{\\nu+k-1}{k}$, and solves the recursion to arrive at Theorem 5: the explicit finite sum of Eq. (32) over Stirling numbers of the first and second kind. As a corollary, the joint first moment on $r$ legs is $E_0(A_1^{(1)}\\cdots A_1^{(r)})=(-\\nu)^{r-1}\\beta_1\\cdots\\beta_r$, and the Brownian case $\\nu=-1/2$ simplifies further to Eq. (34).","pith_inferences":["The paper leaves implicit that Theorem 3 is an algorithm: for any homogeneous spider with an explicit Green kernel, the $D_k^{(i)}$ integrals can be evaluated once, and the recursion then yields moments of arbitrary order without further simulation.","A natural testable extension is the inhomogeneous spider, where each leg has its own speed measure; the Kac-moment structure is leg-agnostic, so the recursion likely survives with leg-dependent coefficients, while the closed Bessel form would not.","The exponential-time moment generating function (21) could be inverted numerically in $\\lambda$ to obtain densities of the joint occupation law, connecting the moment formulas to the double-Laplace-transform characterization of earlier work."],"forward_implications":["Joint occupation moments on the legs of any homogeneous diffusion spider become recursively computable once the Green kernel and the Laplace transforms of first-hitting times of 0 are known.","For Bessel spiders the recursion closes into a finite Stirling-number sum, so moments at all orders are available in closed form rather than by iterative integration.","For Brownian spiders the closed form simplifies via Bessel numbers and gives concrete formulas for Walsh Brownian motion on finitely many rays.","The moment generating function at an independent exponential time has a rational closed form, linking the moment recursion to the known double-Laplace-transform description of the joint law.","For Walsh Brownian motion with a continuous angular distribution, sector occupation times follow the same formulas with $\\beta_i$ equal to sector probabilities."],"supporting_citations":[{"why":"Supplies the explicit Green kernel formula for the diffusion spider (Theorem 1), the starting point of all recursions.","marker":"[17]"},{"why":"Provides the one-dimensional occupation-time recursion and the Bessel $D_k$ formula that this paper extends to joint moments.","marker":"[20]"},{"why":"Gives Kac's moment formula, whose generalized product version (Proposition 1) is the engine of the recursion.","marker":"[15]"},{"why":"Supplies the Stirling-number identities used to solve the recursion and prove Theorem 5.","marker":"[13]"},{"why":"Characterizes the joint law of occupation times via a double Laplace transform, the earlier result this paper's moments complement.","marker":"[26]"},{"why":"Provides the signed Bessel-number identity used to simplify the Brownian spider moments in Theorem 6.","marker":"[21]"}],"fun_headline_variants":["Bessel spider occupation moments as Stirling sums","Explicit moments for Bessel diffusion spiders","Recursion yields explicit spider occupation moments","Stirling numbers for occupation moments on spider legs","Diffusion spider occupation moments: recursion to Stirling sums"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the quoted formulas—the explicit Green kernel of the spider and the explicit Bessel coefficients $D_k^{(i)}$—are correct; the paper uses them without reproving them, so an error in either would invalidate every recursion and closed-form moment that follows.","fun_headline_variants_meta":{"raw":{"variants":["Bessel spider occupation moments as Stirling sums","Explicit moments for Bessel diffusion spiders","Recursion yields explicit spider occupation moments","Stirling numbers for occupation moments on spider legs","Diffusion spider occupation moments: recursion to Stirling sums"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000978,"raw_usage":{"total_tokens":4134,"prompt_tokens":909,"completion_tokens":3225,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":525,"completion_tokens_details":{"reasoning_tokens":3157}},"tokens_in":525,"tokens_out":3225,"duration_ms":21499,"temperature":1.0,"reasoning_tokens":3157,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:06:17.438593+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the two-leg Brownian spider moment $E_0(A_1^{(1)}A_1^{(2)})$ with $\\beta_1=\\beta_2=1/2$ by direct Monte Carlo simulation of Walsh Brownian motion; the explicit formula (Corollary 3) gives $1/8$, so a result that differs from $1/8$ by more than sampling error would refute the paper's closed-form theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the explicit Green kernel formula for the diffusion spider (Theorem 1), the starting point of all recursions."},{"cited_title":"and Stenlund, D","cited_arxiv_id":null,"evidence_quote":"Provides the one-dimensional occupation-time recursion and the Bessel $D_k$ formula that this paper extends to joint moments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Kac's moment formula, whose generalized product version (Proposition 1) is the engine of the recursion."},{"cited_title":"L., Knuth, D","cited_arxiv_id":null,"evidence_quote":"Supplies the Stirling-number identities used to solve the recursion and prove Theorem 5."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Characterizes the joint law of occupation times via a double Laplace transform, the earlier result this paper's moments complement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the signed Bessel-number identity used to simplify the Brownian spider moments in Theorem 6."}],"review_version":1}