{"id":"ce19ea28-77c0-45ef-8b73-46ac0f93cc1c","arxiv_id":"2506.16203","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The authors extend their unified gas-kinetic wave-particle method to frequency-dependent phonon transport with dispersion and polarization, adding an implicit steady-state variant and adaptive frequency sampling that cuts memory and runtime dramatically.","lead":"New computational methods solve the phonon Boltzmann equation, which describes heat flow in tiny devices, while including real material details like phonon dispersion and polarization. The methods are claimed to run large 3D heat simulations on a laptop in about one to two hours, which could speed up nanoscale thermal design.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"IUGKP validation rests on a variable-τ formulation that is not derived or cited in the main text; the unstated dependence of the macroscopic prediction on τg needs a concrete check.","rationale":"The Reader's weakest assumption is exactly the load-bearing concern I identify: the algorithm as derived in the manuscript only covers constant τg, but all validation tests use the variable-τ Terris model, so the IUGKP results are produced by an algorithm the paper does not specify. I agree with that identification and with the CONDITIONAL verdict. The paper's central claim—that IUGKP achieves rapid convergence at both large and small Knudsen numbers through the macroscopic prediction equation (Eq. 76)—is tested only in the variable-τ setting, so the derivation-to-validation gap is not a peripheral formatting issue. I do not see a more severe internal inconsistency that would justify REJECT: the UGKWP part, the frequency-adaptive sampling, and the 1D/3D accuracy comparisons are plausibly correct, and the reference to previous work [44] means the missing formulation likely exists. But the paper as written does not make the IUGKP validation reproducible or the claim auditable, which is a genuine correctness risk. The concrete test I propose—re-deriving the variable-τ compatibility and running the constant-τ version as controls—would settle whether the gap is merely expository or whether the variable-τ coupling changes the algorithm's behavior. I set verdict_should_be to CONDITIONAL to match the Reader; the concern strengthens the case for the requested revisions but does not move the verdict to REJECT or UNVERDICTED, because the accuracy plots against reference solutions provide independent support for the overall method family even if the IUGKP derivation is incomplete.","tokens_in":18756,"tokens_out":1905,"duration_ms":19077,"concrete_test":"Independent re-derivation and implementation check: (1) Take the 1D cross-plane case L = 100 nm with Terris relaxation times and the variable-τ IUGKP from ref. [44]. Verify that the steady-state limit of the particle evolution and the macro-prediction operator (76) satisfy the correct equilibrium temperature determined by the full compatibility condition (17) when τg varies in space. (2) Run the same case with τg held constant per group at the value used in the paper's setup and compare the converged temperature to the reference; if IUGKP with constant τg does not reproduce the reference profiles in Figure 3, then the paper's derivation is not what produced the validation data. (3) Reproduce the 1D convergence table (Table 3) using the unmodified code path described in Algorithm 2 with the variable-τ equations from ref.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The IUGKP method's derivation in Section 4 applies only to constant τg. The paper explicitly states in Section 4.2: 'we only present the algorithm formulation for the case where τg is constant. For the case where τg varies with space, please refer to our previous work [44].' Yet all numerical tests use the Terris relaxation rule (Table 2), where τ depends on temperature and frequency, so τg varies spatially with T in every run. The steady-state integral solution (52) and the CDF-based free-streaming length (57) are used with a spatially varying τ, and the macroscopic prediction equation (76) includes κeff = Σ C_g τ_g |V_g|^2 / 3 dω, which is the diffusion-limit conductivity only if the full variable-τ compatibility condition is enforced. Because the manuscript neither derives the variable-τ particle resampling (sampling rule Pg,f, compatibility weights) nor reproduces the equations from [44], the numerical validations in Sections 5.1-5.5 do not directly test the derivation given in this paper. This is not a claim of mathematical error, but of a gap: the central efficiency claim for IUGKP (fast convergence at small Kn in Eq. (76)) is validated only by an algorithm whose specification is absent from the manuscript. The concern is load-bearing because the speedup claims are the paper's most distinctive contribution, and they depend on the coupling between the particle solver and the macroscopic prediction operator operating with the variable-τ used in the tests.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops two particle-based multiscale methods for the frequency-dependent phonon Boltzmann transport equation with dispersion and polarization: the UGKWP method for unsteady problems and the IUGKP method for steady-state problems. Both methods solve multiple groups of BGK equations at discrete frequency points, construct multiscale fluxes from integral solutions of the BGK equation, and use an adaptive frequency-space sampling technique to control particle counts. The IUGKP method adds a macroscopic diffusion-synthetic prediction equation to accelerate convergence at small Knudsen numbers. Numerical validation covers 1D, 2D, and 3D silicon heat conduction cases against deterministic reference solutions, with reported speedups of one to two orders of magnitude and 3D simulations running in 40--100 minutes on a single-core laptop.","tokens_in":19007,"tokens_out":6542,"duration_ms":72147,"significance":"If the methods perform as claimed, they offer a practical route to large-scale phonon transport simulations that are currently intractable for deterministic DOM-type solvers. The constant-relaxation-time derivations are self-contained, the algorithms are clearly structured, and the efficiency results in Sections 5.4 and 5.5 are impressive if reproducible. However, the validation and the headline efficiency claims rest on an unstated variable-relaxation-time formulation and an unspecified tuning parameter, so the significance cannot be fully assessed from the manuscript in its present form. The paper does not ship code, but the algorithmic descriptions are sufficiently detailed for the constant-τ case to be reimplemented.","major_comments":[{"comment":"The IUGKP algorithm is derived only for constant relaxation time per group, with the text explicitly stating 'we only present the algorithm formulation for the case where τg is constant' and referring to [44] for the spatially varying case. However, all numerical tests use Terris' relaxation rule (Table 2), in which τ depends on temperature and frequency; because temperature varies in every test, τg varies spatially in every run. The steady-state integral solution in Eq. (52) and the CDF in Eq. (53) are written with spatially varying τ, but the particle evolution step and the free-streaming length sampling in Eq. (57) use a constant τg. The manuscript does not provide the variable-τ particle resampling, compatibility conditions, or the corresponding form of the macroscopic prediction equation. As a result, the numerical validations in Sections 5.1--5.5 exercise an algorithm whose specification is absent from this paper, and the central convergence claim for IUGKP at small Knudsen numbers is not directly supported by the derivation presented here. Please include the variable-τ formulation or restrict the tests and claims to the constant-τ case.","section":"Section 4.2, Eq. (57)"},{"comment":"The same constant-τ limitation affects the UGKWP derivation: the integral solution (20), the expansion (21)--(23), and the flux coefficients (25) assume a constant τg along the characteristic and across the time step. With the temperature-dependent Terris rule used in Sections 5.1 and 5.2, τg varies with x and t, and the manuscript does not state how τg is evaluated in the integral solution, in the coefficients, or in the compatibility condition (11). This is load-bearing for the UGKWP validation as well, since the transient tests use the same relaxation model. Please clarify the treatment of spatially varying τ in the UGKWP flux construction, or restrict the claims accordingly.","section":"Section 3, Eqs. (20)--(25)"},{"comment":"The amplification factor γ in the macroscopic prediction equation is a free parameter, but its value or selection criterion is never reported in the numerical tests or in Algorithm 2. Because the convergence speed of IUGKP in the diffusive regime depends directly on this factor (as explained in Remark 3), the iteration counts in Table 3 and the CPU times in Tables 4 and 6 cannot be reproduced without this information. Please state the γ values used for each test and the criterion for choosing them.","section":"Section 4.4, Eq. (76) and Algorithm 2"},{"comment":"The transition from the integral solution to the particle-based representation uses the identity P_{g,c} = dP_{g,f}/dx, which the text states holds for constant τg. For variable τg, this identity does not hold in the form needed for the cumulative weighting in Eq. (56). Since the tests use variable τg, the derivation of Eq. (56) is not applicable to the tested settings. This is a specific instance of the gap identified in the first major comment and should be addressed in the revision.","section":"Section 4.1, Eqs. (53)--(56)"}],"minor_comments":[{"comment":"In the second term of F^{fr,wave}_{g,ij}, 'c2' appears to be a typo and should read 'V_g^2'.","section":"Eq. (45)"},{"comment":"The iteration count for UGKWP at L = 100 μm is reported as '12,0000' (120,000) and the ratio as 38; please check the typography and arithmetic consistency.","section":"Table 3"},{"comment":"The captions and text do not consistently distinguish the UGKWP variant without frequency adaptive sampling (Table 4) from the variant with it (Table 6); please label the variants identically in the text and in the table captions.","section":"Section 5.3, Tables 4 and 6"},{"comment":"The text states that the IUGKP results for the 3D cubic domain are 'shown in Fig .11', but the contour plot for that test is Fig. 8; the figure cross-references should be corrected.","section":"Section 5.4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on reference [44], a parallel preprint by the same group, for the variable-τ formulation of IUGKP that is actually used in all numerical tests. If [44] is not yet published or readily available, the present manuscript is not self-contained with respect to its own validation. This is the main reason for the major-revision recommendation rather than acceptance. The paper is otherwise a plausible contribution to the multiscale phonon transport literature, and the reported efficiency gains are striking; the authors should be encouraged to close the variable-τ gap and report γ values."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. This paper extends the authors' UGKWP/IUGKP framework to frequency-resolved phonon transport, and that's a real step: the multi-group compatibility condition and the adaptive frequency-space sampling are genuinely useful ideas, and the 1D and 3D benchmarks against the SIS reference are plausible. The derivations in Sections 3-4 follow the standard UGKS integral-solution route and appear algebraically consistent; I only spotted a typo in Eq. (45), where the second term has 'c2' that should be V_g^2.\n\nThe soft spots are concentrated in the IUGKP part. Section 4.2 states plainly that the algorithm is formulated only for constant relaxation time per group, and refers to the companion paper [44] for the space-varying τ case. But every numerical test uses the Terris relaxation rule from Table 2, which depends on temperature and frequency, so τ is spatially varying in every run. That means the convergence speedups in Tables 3, 4, and 6 are produced by an algorithm that is not specified in this manuscript. That's a load-bearing gap, because the 'rapid convergence at small Knudsen numbers' is the paper's most distinctive claim. It's not a mathematical error — I see no reason to doubt the extension works — but as it stands, the validation does not test the derivation that's actually included.\n\nOther issues are smaller: γ, the amplification factor in the macro prediction equation, is never given a value or selection rule; the Monte Carlo results have no error bars or variance estimates; the 2D square-domain test compares only UGKWP against IUGKP, with no independent reference; and the reference solutions come from a paper sharing a co-author, so independent validation is thin. No code or data is released, which matters for a paper whose headline is efficiency.\n\nOn balance, I think the central idea holds. The UGKWP extension is clearly explained, and the adaptive frequency sampling is a nice contribution in itself. The IUGKP gap is presentation-level but must be fixed before I'd trust the efficiency claims. I'd send it to peer review with a request that the authors either derive the variable-τ formulation or reproduce the relevant equations from [44], report γ, include variance estimates, and ideally release code. It deserves serious referee time, but not acceptance in its current form.","headline":"A useful extension of UGKWP to multi-frequency phonon BTE with a real efficiency claim, but the IUGKP validation relies on a variable-relaxation-time algorithm that the paper only references, not derives.","tokens_in":19576,"tokens_out":3221,"would_cite":false,"duration_ms":35584,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M08","80A19","82C70"],"pacs":["44.10.+i","63.20.-e"],"model":"deepseek-v4-flash","headline":"The paper claims that two wave-particle schemes, UGKWP and IUGKP, solve the frequency-dependent phonon Boltzmann equation across all Knudsen regimes, with IUGKP converging in 4 to 320 iterations and running 3D multiscale heat conduction…","keywords":["unified gas-kinetic wave particle method","phonon Boltzmann transport equation","multiscale heat conduction","dispersion","polarization","implicit unified gas kinetic particle method","frequency-space adaptive sampling","Knudsen number"],"falsifier":"Run the one-dimensional cross-plane silicon film tests at L = 10 nm, 100 nm, 1 $\\mu$m, and 100 $\\mu$m with Terris relaxation times using only the constant-tau IUGKP derivation given in Section 4; if the results fail to match the DUGKS reference solutions or lose the reported convergence ratios of 400, 180, 20, and 38, the IUGKP validation collapses.","tokens_in":18485,"feed_emoji":"🔥","tokens_out":3835,"duration_ms":39017,"temperature":0.7,"pith_summary":"This paper claims that the frequency-dependent phonon Boltzmann equation, with full dispersion and polarization effects, can be solved efficiently across every transport regime by two wave-particle schemes. The explicit UGKWP method adapts automatically: in the diffusive limit its flux collapses to Fourier's law, while in the ballistic limit free-streaming particles carry the non-equilibrium flux. The steady-state IUGKP method adds a macroscopic prediction equation so that convergence is fast at both small and large Knudsen numbers, and adaptive sampling in frequency space keeps the particle count per cell at the level of a single-frequency gray method. If these claims hold, large three-dimensional multiscale heat-conduction simulations become laptop-scale rather than supercomputer-scale problems.","feed_headline":"Multiscale phonon transport solved on a laptop in under two hours","feed_subtitle":"Wave-particle schemes with adaptive frequency sampling cover ballistic to diffusive heat flow without a supercomputer.","key_machinery":"The carrier of the argument is the wave-particle decomposition of each frequency group's distribution into an equilibrium 'wave' flux, computed analytically from the integral solution of the unsteady BGK equation, and a free-transport 'particle' flux. For IUGKP, the load-bearing new object is the macroscopic prediction equation (Eq. 76), $-\\nabla\\cdot(\\gamma\\kappa_{\\mathrm{eff}}\\nabla\\delta T)=\\nabla\\cdot q$, an inexact-Newton acceleration that dominates convergence at small Knudsen numbers while automatically shutting off in the ballistic limit. Adaptive sampling maps particles to frequency intervals by cumulative energy fraction, reducing the total particle count from order $N_B\\times N_{\\mathrm{ref}}$ to order $N_{\\mathrm{ref}}$.","core_discovery":"The central discovery is that treating each discrete phonon frequency as its own gray BGK group, joined only by a frequency-integrated collision compatibility condition, lets a unified gas-kinetic wave-particle solver handle dispersion and polarization without extra memory growth. The IUGKP variant replaces explicit time-marching with a steady-state iteration in which particles free-stream a distance drawn from the physical mean free path, and a newly constructed macroscopic prediction equation of the form $-\\nabla\\cdot(\\gamma\\kappa_{\\mathrm{eff}}\\nabla\\delta T)=\\nabla\\cdot q$ supplies the missing small-Knudsen acceleration. The paper reports that IUGKP converges in 4 to 320 iterations on one-dimensional films from 10 nm to 100 $\\mu$m, up to two orders of magnitude faster than UGKWP, and completes three-dimensional cubic and heat-dissipation simulations in 40 and 100 minutes respectively while using about 2 GB of memory.","pith_inferences":["The constant-tau derivation gap implies that the paper's validation rests on a variable-relaxation-time formulation drawn from the companion gray IUGKP work; a reader applying the method to real silicon must obtain that formulation.","Because adaptive sampling allocates particles by equilibrium energy fraction, the same idea could reduce memory in deterministic discrete-ordinate or unified gas-kinetic solvers by weighting angular quadrature per frequency.","The macroscopic prediction equation is effectively a Poisson solve for the temperature correction, so its amplification factor could be viewed as a tunable preconditioner that may extend to nonlinear or strongly temperature-dependent materials.","If the 40-to-100-minute three-dimensional runs hold, routine device-scale thermal simulation with resolved dispersion becomes a desktop task, enabling fast optimization of micro- and nano-scale electronics cooling."],"forward_implications":["If correct, steady-state multiscale heat conduction is no longer memory-bound: full three-dimensional runs fit in about 2 GB and finish on one core.","UGKWP retains accuracy for unsteady problems, so the method pair covers transient and steady simulations with the same machinery.","Frequency-adaptive sampling should make the cost of adding dispersion and polarization nearly independent of the number of frequency groups used.","The reported IUGKP iteration counts, from 4 in the diffusive limit to 320 in the ballistic regime, suggest the method needs no regime-specific algorithmic switching.","The automatic collapse to Fourier's law in the diffusive limit means no separate diffusion solver is needed for low-Knudsen regimes."],"supporting_citations":[{"why":"Supplies the variable-relaxation-time formulation that the IUGKP derivation here explicitly defers to.","marker":"[44]"},{"why":"Provides the DUGKS reference solutions used to validate the one-, two-, and three-dimensional test cases.","marker":"[48]"},{"why":"Supplies Terris' frequency- and temperature-dependent relaxation-time rule used in all numerical tests.","marker":"[47]"},{"why":"Introduces the unified implicit kinetic scheme and inexact Newton iteration that motivate the IUGKP macroscopic prediction.","marker":"[23]"},{"why":"Supplies the quadratic dispersion coefficients and the implicit kinetic framework for frequency-dependent phonon transport.","marker":"[46]"},{"why":"The authors' earlier gray-model UGKWP for phonons, which this paper extends to dispersion and polarization.","marker":"[41]"},{"why":"Supplies the reference-number particle sampling strategy used to control particle counts.","marker":"[35]"}],"fun_headline_variants":["New solver runs multiscale heat flow on a laptop","Phonon transport: 2 hours on a laptop, no cluster","Wave-particle scheme speeds up phonon simulations 100x","From ballistic to diffusive: fast phonon solver"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The IUGKP algorithm is derived only for a constant relaxation time per frequency group, but every numerical test uses a relaxation time that depends on temperature and frequency, so the validation leans on a variable-tau formulation the paper does not present.","fun_headline_variants_meta":{"raw":{"variants":["New solver runs multiscale heat flow on a laptop","Phonon transport: 2 hours on a laptop, no cluster","Wave-particle scheme speeds up phonon simulations 100x","From ballistic to diffusive: fast phonon solver"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000232,"raw_usage":{"total_tokens":1505,"prompt_tokens":976,"completion_tokens":529,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":592,"completion_tokens_details":{"reasoning_tokens":460}},"tokens_in":592,"tokens_out":529,"duration_ms":6483,"temperature":1.0,"reasoning_tokens":460,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T23:45:36.361528+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the one-dimensional cross-plane silicon film tests at L = 10 nm, 100 nm, 1 $\\mu$m, and 100 $\\mu$m with Terris relaxation times using only the constant-tau IUGKP derivation given in Section 4; if the results fail to match the DUGKS reference solutions or lose the reported convergence ratios of 400, 180, 20, and 38, the IUGKP validation collapses.","supporting_citations":[{"cited_title":"Implicit unified gas kinetic particle method for steady-state solution of multiscale phonon transport","cited_arxiv_id":"2506.09465","evidence_quote":"Supplies the variable-relaxation-time formulation that the IUGKP derivation here explicitly defers to."},{"cited_title":"A fast synthetic iterative scheme for the stationary phonon boltzmann transport equation","cited_arxiv_id":null,"evidence_quote":"Provides the DUGKS reference solutions used to validate the one-, two-, and three-dimensional test cases."},{"cited_title":"Modeling semiconductor nanos- 32 tructures thermal properties: The dispersion role","cited_arxiv_id":null,"evidence_quote":"Supplies Terris' frequency- and temperature-dependent relaxation-time rule used in all numerical tests."},{"cited_title":"Unified implicit kinetic scheme for steady multiscale heat transfer based on the phonon Boltzmann transport equation","cited_arxiv_id":null,"evidence_quote":"Introduces the unified implicit kinetic scheme and inexact Newton iteration that motivate the IUGKP macroscopic prediction."},{"cited_title":"An implicit kinetic scheme for multiscale heat transfer problem accounting for phonon dispersion and polarization","cited_arxiv_id":null,"evidence_quote":"Supplies the quadratic dispersion coefficients and the implicit kinetic framework for frequency-dependent phonon transport."},{"cited_title":"Unified gas-kinetic wave-particle methods","cited_arxiv_id":null,"evidence_quote":"Supplies the reference-number particle sampling strategy used to control particle counts."}],"review_version":1}