{"id":"f14250b2-2bbe-44a1-b4c3-0ed1fc8db699","arxiv_id":"2501.04593","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A weighted Besov calculus on Heisenberg groups gives pathwise existence and uniqueness for the parabolic Anderson model under the condition (n+1)/2-(1-zeta) < alpha < (n+1)/2.","lead":"This paper builds a new function-space calculus on Heisenberg groups and uses it to solve a stochastic heat equation with multiplicative noise in a pathwise sense. The advertised exponent window, (n+1)/2 - (1-zeta) < alpha < (n+1)/2, says the spatial noise must be smooth enough relative to how rough it is in time.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 3.25 transfers the radial-only frequency localization (Prop. 3.23) to arbitrary Besov blocks without proof; the paraproduct and fixed-point estimates depend on this transfer.","rationale":"Strengths: the paper builds a detailed weighted-Besov calculus on Hn (Gevrey bump functions, Bernstein lemma, heat-flow smoothing), and the proofs of Propositions 2.24, 3.6, 3.8 and Lemma 4.15 are fairly complete and internally consistent. The main theorem is likely true, and the exponent condition (1.15) matches the heuristic comparison to Rd with Q=2n+2. The risk is concentrated in the one imported-and-transferred paraproduct step.\n\nThe dependency chain for the central claim is: Theorem 4.16 -> Theorem 4.7 -> Lemma 4.6 -> Proposition 3.11 -> Lemma 3.29 -> Lemmas 3.27/3.28 -> Corollary 3.25 -> Proposition 3.23. The last step is the least secure because Proposition 3.23 is explicitly radial, while Corollary 3.25 needs non-radial blocks. The non-diagonal entries of the projective Fourier transform are essential: for radial functions \\hat f(m,ell,lambda)=0 for m != ell, which trivializes the matrix product (2.17) and makes [4]'s Laguerre calculus applicable. For a general block, the sum over j in (2.17) can mix different frequencies, and the annulus/ball support of the product is not a formal consequence of the support of each factor unless one proves a twisted-convolution localization estimate.\n\nThe reader's CONDITIONAL verdict is therefore appropriate. The sign error in Theorem 4.16's proof is real but does not affect the result; it is exactly the kind of typo that a careful revision should fix. I do not see circularity: the paper uses [3,4,6] as external input, not as a consequence of the present results, and no parameters are fitted. The verdict should remain CONDITIONAL, contingent on closing the radial-to-block gap or supplying a direct proof of Corollary 3.25.","tokens_in":49970,"tokens_out":8905,"duration_ms":76043,"concrete_test":"Prove Corollary 3.25 from the matrix product formula (2.17)-(2.18) for arbitrary (non-radial) f,g. Concretely, for h_k^S = S_{k-1}f sigma_k g write the Fourier transform as \\hat h_k^S(m,ell,lambda) = sum_j (sum_{i' is required; the final condition (1.15) is nevertheless the correct one.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops weighted Besov spaces on Heisenberg groups H_n using the projective Fourier transform of Bahouri--Chemin--Danchin, proves Bernstein-type inequalities, heat-flow smoothing estimates, and a paraproduct calculus in these spaces, and then applies the framework to the pathwise (Stratonovich--Young) parabolic Anderson model driven by a Gaussian noise with time covariance |t-s|^{-\\zeta} and spatial covariance given by the kernel G_{2\\alpha}. The main existence-uniqueness result is Theorem 4.7 together with Theorem 4.16, which identifies the admissible range (n+1)/2-(1-\\zeta)<\\alpha<(n+1)/2 for the noise parameters. The authors emphasize that the weighted Besov space construction is new and of independent interest.","tokens_in":50278,"tokens_out":15207,"duration_ms":149277,"significance":"If the technical gaps are repaired, the paper makes a substantial contribution: it gives a genuinely new weighted Besov-space framework on Heisenberg groups that is compatible with heat-flow estimates, and it provides a pathwise well-posedness result for the parabolic Anderson model whose exponent condition matches the Hausdorff dimension Q=2n+2, in line with the authors' earlier It\\^o-setting results. The paper contains many explicit and checkable estimates: the Gevrey-class construction of Littlewood--Paley blocks, the scaling arguments in Section 2.4, the heat-flow smoothing in Proposition 3.8, and the variance computation in Lemma 4.15 are all concrete and mostly coherent. The main unresolved point is a single but load-bearing localization step in the paraproduct proof, and the paper would be acceptable after that step is supplied.","major_comments":[{"comment":"Corollary 3.25 is asserted as a \"direct application\" of Proposition 3.23, but Proposition 3.23 is proved only for radial functions f,g; its proof uses the diagonalization (3.55) and imports [4, Proposition 4.2]. The blocks S_{k-1}f and \\sigma_k g are not radial for general f, and the projective convolution identity (2.17)--(2.18) couples all indices j\\in\\mathbb{N}^n, so the claimed support localization of \\widehat{h^S_k}(m,\\ell,\\lambda) and \\widehat{h^{\\sigma,\\varepsilon}_k}(m,\\ell,\\lambda) does not follow from Proposition 3.23 without an additional argument. This step is load-bearing: Lemma 3.29 invokes Corollary 3.25 to apply Lemmas 3.27--3.28, Proposition 3.11 then uses Lemma 3.29, and Lemma 4.6 and Theorem 4.7 rely on Proposition 3.11. Please provide a complete proof of the non-radial frequency localization, or state and prove the needed non-radial version of [4, Proposition 4.2] in the projective Fourier setting.","section":"§3.3.3, Corollary 3.25"}],"minor_comments":[{"comment":"The displayed compatibility inequality \"2-\\zeta<1+(n+1)/2-\\alpha\" has the wrong direction: the argument requires 1+(n+1)/2-\\alpha<2-\\zeta, which is equivalent to the stated condition \\alpha>(n+1)/2-(1-\\zeta). The final condition (1.15)/(4.66) is correct, but the displayed line should be corrected.","section":"§4.3, proof of Theorem 4.16"},{"comment":"The abstract and introduction say the spatial covariance is generated by negative powers (-\\Delta)^{-\\alpha}, while Definition 4.8 and equation (1.14) use G_{2\\alpha}, i.e. (-\\Delta)^{-2\\alpha}. Please harmonize the wording so that the exponent convention is unambiguous.","section":"§1.1 and §1.2, versus §4.3"},{"comment":"The heat semigroup P_t is defined in (1.11) as the semigroup generated by \\Delta, but Proposition 3.6 calls it e^{t\\Delta/2}. Since the main SPDE (1.13) contains the factor 1/2, this normalization should be fixed consistently (for instance by defining p_t for the operator \\Delta/2 or by writing the mild solution with P_{(t-s)/2}). This is a notational issue and does not change the estimates, because the constants can absorb a factor 2.","section":"§1.1, (1.11), and §3.2, Proposition 3.6"},{"comment":"The weight \\rho_b is defined inconsistently: Definition 3.9 sets \\rho_b(q)=|q|_*^{-b} with b\\in\\mathbb{R}, while Definition 4.2 and Theorem 4.16 use \\rho_b(q)=c(1+|q|_*^b) with b>0. The integrability condition in Lemma 4.15 depends on which convention is used. Please align the notation and state the exact class of polynomial weights used in Hypothesis 4.5.","section":"§3.3, Definition 3.9, and §4.1, Definition 4.2"},{"comment":"Theorem 4.7 states existence and uniqueness in D^{\\theta,\\kappa,\\nu,b}_{\\alpha,\\beta} for arbitrary exponents satisfying \\theta+\\vartheta>1 and \\gamma<\\kappa<1, but Lemma 4.6 proves contraction only for the particular pair constructed with \\theta=1-\\vartheta+\\varepsilon and \\kappa=\\gamma+2\\varepsilon. Please either prove the more general statement or state Theorem 4.7 with the specific pair produced in Lemma 4.6, since the application in Theorem 4.16 only needs the existence of some such pair.","section":"§4.2, Theorem 4.7"}],"recommendation":"major_revision","confidential_remarks":"The paper relies heavily on the authors' previous work [6] and on [3,4]; this is acceptable in context, but the referee's main concern is the unproved transfer from radial to non-radial functions in Corollary 3.25. If the authors can provide a complete proof of that localization step, or a precise adaptation of [4, Proposition 4.2] to the projective non-radial setting, the central theorem is likely sound and the contribution would be publishable in a good probability journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my take on Baudoin-Chen-Huang-Ouyang-Tindel-Wang, arXiv:2501.04593. The paper builds a weighted Besov calculus on the Heisenberg group via the projective Fourier transform and uses it to solve a pathwise (Stratonovich-Young) parabolic Anderson model under the exponent condition α > (n+1)/2 − (1−ζ). The Besov machinery is a real contribution: new weighted spaces, heat-flow smoothing, Bernstein inequalities, and a noise embedding that uses no fitted parameters. The exponent threshold matching the flat-space Bessel-kernel result with Hausdorff dimension Q=2n+2 is a nice sanity check, and the comparison with the earlier Itô paper is honest.\n\nI agree with the reader's conditional verdict, but I want to put the stress-test concern right at the center. Corollary 3.25 is not justified. Proposition 3.23 proves product localization only for radial functions, using the radial diagonalization of the projective Fourier transform. Corollary 3.25 asserts the same for arbitrary Besov blocks by \"direct application.\" That is a non-sequitur: the blocks are not radial, their projective Fourier transforms are off-diagonal, and the pointwise product on H_n is not the matrix product in (2.17)-(2.18). The off-diagonal terms are exactly what [4] controls only for radial inputs. This step feeds Lemma 3.29, Proposition 3.11, and the fixed-point argument in Lemma 4.6, so it is load-bearing. A referee should demand a proof of Corollary 3.25 for non-radial blocks, or a precise restatement of the support conditions that avoids the radial assumption.\n\nThere is also a minor sign error in the proof of Theorem 4.16: the compatibility inequality is displayed with the wrong direction, although the final condition (1.15) is correct. That is easy to fix. Some estimates in the Young-integration section are left to the reader, but the structure is standard.\n\nThe paper is serious and not circular: no fitted parameters, no conclusion-as-input. Self-citation is contextual. If the paraproduct gap is repaired, the theorem is likely to stand.\n\nBottom line: this deserves a serious referee, but with a specific mandate to verify Corollary 3.25. I would not cite the paraproduct estimates in their current form.","headline":"Main theorem likely true, but Corollary 3.25 is a genuine missing proof: product localization is radial-only and the transfer to Besov blocks is not shown.","tokens_in":50840,"tokens_out":14556,"would_cite":false,"duration_ms":134847,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","35R60","43A80","46E35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Unique solution found for parabolic Anderson on Heisenberg groups","keywords":["parabolic Anderson model","Heisenberg group","weighted Besov spaces","projective Fourier transform","paraproduct","Stratonovich/Young integration","sub-Laplacian","Gaussian noise"],"falsifier":"Take a non-radial Schwartz pair on $\\mathbf{H}_n$, for instance $f(q)=x_1 e^{-|q|_h^2}$ and $g(q)=e^{-|q|_h^2}$, and check whether the projective Fourier coefficients of $h=S_{k-1}f\\cdot \\sigma_k g$ vanish outside $(2|m|+n)^{-1}2^k$ times the annulus $C_0'$ asserted in Corollary 3.25; a single pair with support leaking outside that annulus would invalidate Lemma 3.29 and the fixed-point contraction.","tokens_in":49770,"feed_emoji":"🧮","tokens_out":7462,"duration_ms":71071,"temperature":0.7,"pith_summary":"The paper's aim is to give a rigorous pathwise (Stratonovich) meaning to the parabolic Anderson model on the Heisenberg group $\\mathbf{H}_n$ and to determine exactly for which noises a unique solution exists. The noise is smoother than white noise in time, with spatial covariance given by a negative power of the sub-Laplacian, and the solution lives in a new family of weighted Besov spaces tailored to the group. These spaces, defined through a projective Fourier transform whose frequency variable is the pair $(m,\\ell,\\lambda)$ of Hermite-mode indices and a real weight, are developed in detail and are claimed to be new and of independent interest. The payoff is an explicit condition on the noise parameters under which existence and uniqueness hold.","feed_headline":"Unique solution found for parabolic Anderson on Heisenberg groups","feed_subtitle":"Exact noise window pinned down by new weighted Besov spaces from a projective Fourier transform.","key_machinery":"The load-bearing object is the weighted Besov scale $B^{\\gamma,\\nu}_{\\alpha,\\beta}(\\mathbf{H}_n)$, defined by frequency blocks $\\sigma_k f$ whose projective Fourier transform is supported on the diagonal $\\{m=\\ell\\}$ with frequency $|\\lambda|(2|m|+n)$ comparable to $2^k$; the exponential weight $e^{-\\nu|q|_*^\\eta}$ is what lets the method handle spatially unbounded noises. The projective Fourier transform itself — Fourier modes indexed by Hermite indices $(m,\\ell)$ and $\\lambda\\in\\mathbb{R}^*$, with multiplication turned into a discrete convolution — supplies the analogue of Euclidean frequency localization. The argument then runs on two rails: Bernstein-type estimates and heat-semigroup smoothing in these weighted spaces, and a paraproduct continuity theorem whose key input is a frequency-localization statement for products of two localized functions, imported from an existing paraproduct on $\\mathbf{H}_n$ and transferred through radial test functions.","core_discovery":"The central discovery is that the parabolic Anderson model $\\partial_t u_t = \\tfrac12 \\Delta u_t + u_t \\dot W_{\\zeta,\\alpha}$ on $\\mathbf{H}_n$ admits a unique mild solution, in the Young/Stratonovich sense, whenever the noise has temporal exponent $\\zeta\\in(0,1)$ and spatial regularity $\\alpha$ satisfying $\\frac{n+1}{2}-(1-\\zeta)<\\alpha<\\frac{n+1}{2}$. The noise $\\dot W_{\\zeta,\\alpha}$ is the centered Gaussian field whose time covariance is $|t-s|^{-\\zeta}$ and whose space covariance is the resolvent kernel $G_{2\\alpha}$ of $(-\\Delta)^{-2\\alpha}$; the paper proves that such a noise falls into the Hölder-in-time distribution class $\\mathcal{C}^{\\vartheta,-\\gamma,\\rho_b}_{\\infty,\\infty}$ required by its general fixed-point theorem. Along the way it constructs weighted Besov spaces $B^{\\gamma,\\nu}_{\\alpha,\\beta}$ on $\\mathbf{H}_n$ via Littlewood–Paley blocks built from Gevrey partitions of unity in the projective Fourier picture, and proves the Bernstein lemma, heat-flow smoothing, and paraproduct bounds needed for the contraction argument.","pith_inferences":["If the block-localization step of Corollary 3.25 is genuinely valid for arbitrary blocks, the same paraproduct machinery should transfer to any homogeneous Lie group with a projective Fourier calculus, giving analogous exponent windows with the group's homogeneous dimension in place of $Q=2n+2$.","The paper stops at existence and uniqueness; a natural next test would be whether the boundary $\\alpha=(n+1)/2-(1-\\zeta)$ also governs moment growth and intermittency exponents.","Below the Young threshold one would expect regularity structures or paracontrolled calculus on $\\mathbf{H}_n$ to extend the result; the weighted Besov scale constructed here would be the natural base scale.","The radial-to-arbitrary transfer in Corollary 3.25 could be checked on $n=1$ by explicit Hermite-coefficient computation; a counterexample would require a different localization argument for the paraproduct."],"forward_implications":["In the white-noise-in-time limit $\\zeta\\to 1$, the interval in condition (1.15) collapses; the model is solvable only if the spatial noise is a function rather than a distribution, matching known Stratonovich phenomena.","With $Q=2n+2$, condition (1.15) takes the shape $\\alpha>Q/4-(1-\\zeta)$, the analogue of the Euclidean Bessel-kernel condition $d/4-(1-\\zeta)$.","The weighted Besov spaces and their paraproduct give a setting for other semilinear SPDEs on noncompact sub-Riemannian manifolds with spatially unbounded distributions.","Because the time integral is pathwise Young/Stratonovich, the solution map supports the polymer-type measures that motivate the paper."],"supporting_citations":[{"why":"Supplies the radial-function/Laguerre identities used to connect the imported paraproduct to the projective Fourier setting.","marker":"[1]"},{"why":"Provides the Euclidean partition-of-unity and Littlewood–Paley framework that the construction adapts.","marker":"[2]"},{"why":"Supplies the projective Fourier transform, Schwartz-space isomorphism, and differential operators on $\\tilde{\\mathbf{H}}_n$ that define the frequency side.","marker":"[3]"},{"why":"Supplies the Heisenberg paraproduct and product-frequency-localization estimates that Proposition 3.23 imports.","marker":"[4]"},{"why":"Defines the Gaussian noise family and Riesz kernel $G_{2\\alpha}$, and gives the Itô-setting exponent window being extended.","marker":"[6]"},{"why":"Gives the explicit heat kernel of the sub-Laplacian used throughout the smoothing estimates.","marker":"[9]"},{"why":"Provides the Euclidean Stratonovich threshold $d/4-(1-\\zeta)$ against which the Heisenberg condition is compared.","marker":"[10]"},{"why":"Supplies the weighted Bernstein-lemma and Besov-space template that is transferred to $\\mathbf{H}_n$.","marker":"[12]"}],"fun_headline_variants":["Parabolic Anderson on Heisenberg groups: unique solution","Unique solution for Heisenberg parabolic Anderson","Parabolic Anderson solved uniquely on Heisenberg groups","Noise window pinned for parabolic Anderson on Heisenberg","Weighted Besov spaces crack Heisenberg Anderson"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is the assertion, via Corollary 3.25, that the product of two frequency-localized functions—not merely two radial ones—remains supported in the expected projective frequency annulus or ball; the paper imports this from the radial case without giving a fully detailed proof for arbitrary Littlewood–Paley blocks.","fun_headline_variants_meta":{"raw":{"variants":["Parabolic Anderson on Heisenberg groups: unique solution","Unique solution for Heisenberg parabolic Anderson","Parabolic Anderson solved uniquely on Heisenberg groups","Noise window pinned for parabolic Anderson on Heisenberg","Weighted Besov spaces crack Heisenberg Anderson"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00048,"raw_usage":{"total_tokens":2378,"prompt_tokens":952,"completion_tokens":1426,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":1355}},"tokens_in":568,"tokens_out":1426,"duration_ms":10076,"temperature":1.0,"reasoning_tokens":1355,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:29:50.482569+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a non-radial Schwartz pair on $\\mathbf{H}_n$, for instance $f(q)=x_1 e^{-|q|_h^2}$ and $g(q)=e^{-|q|_h^2}$, and check whether the projective Fourier coefficients of $h=S_{k-1}f\\cdot \\sigma_k g$ vanish outside $(2|m|+n)^{-1}2^k$ times the annulus $C_0'$ asserted in Corollary 3.25; a single pair with support leaking outside that annulus would invalidate Lemma 3.29 and the fixed-point contraction.","supporting_citations":[{"cited_title":"Bahouri, Hajer, D","cited_arxiv_id":null,"evidence_quote":"Supplies the radial-function/Laguerre identities used to connect the imported paraproduct to the projective Fourier setting."},{"cited_title":"Bahouri, J.-Y","cited_arxiv_id":null,"evidence_quote":"Provides the Euclidean partition-of-unity and Littlewood–Paley framework that the construction adapts."},{"cited_title":"Bahouri, J.-Y","cited_arxiv_id":null,"evidence_quote":"Supplies the projective Fourier transform, Schwartz-space isomorphism, and differential operators on $\\tilde{\\mathbf{H}}_n$ that define the frequency side."},{"cited_title":"Bahouri, and I","cited_arxiv_id":null,"evidence_quote":"Supplies the Heisenberg paraproduct and product-frequency-localization estimates that Proposition 3.23 imports."},{"cited_title":"Baudoin, C","cited_arxiv_id":null,"evidence_quote":"Defines the Gaussian noise family and Riesz kernel $G_{2\\alpha}$, and gives the Itô-setting exponent window being extended."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the explicit heat kernel of the sub-Laplacian used throughout the smoothing estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Euclidean Stratonovich threshold $d/4-(1-\\zeta)$ against which the Heisenberg condition is compared."},{"cited_title":"Mourrat and H","cited_arxiv_id":null,"evidence_quote":"Supplies the weighted Bernstein-lemma and Besov-space template that is transferred to $\\mathbf{H}_n$."}],"review_version":1}