{"id":"c95ae6c9-1a1d-47a9-a3d4-4f630377b0ab","arxiv_id":"2411.18106","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In 2D simulations of hydrogen/air head-on quenching, thermodiffusive instabilities raise mean wall heat flux by up to about threefold and reduce quenching distance, while hydrodynamic instabilities alone have little effect.","lead":"This paper uses computer simulations of lean hydrogen flames hitting a wall to show that flame instabilities, not smooth flame fronts, drive most of the extra heat delivered to the wall. The result gives engineers a way to estimate wall heat loads from simpler one-dimensional flame calculations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-realization HOQ statistics: the spatial average over one periodic domain per condition is treated as a converged mean, so the linear I0 fit in Fig. 12a may shift.","rationale":"The paper's central claim has two levels. Qualitatively, thermodiffusive instabilities substantially alter head-on quenching while hydrodynamic instabilities do not; this is supported by consistent visual and statistical contrasts across the parameter space and by the p10/Tu700 cases, so I do not see a reason to reject that claim. Quantitatively, the paper proposes a linear model for Phi_q,2D/Phi_q,1D as a function of I0 and a second linear model for I0 as a function of Ze/Pecd. The load-bearing step is the empirical mean ratio in Fig. 11, because both fits inherit every error in those means. The manuscript reports only one 2D realization per operating condition, and the standard deviation shown is the across-wall spatial variation, not the sampling uncertainty of the mean. That matters more here than in many DNS studies because the quenching process is transient and intermittent: the mean is controlled by a small number of flame fingers and local side-wall quenching events, so a single periodic box of 100 delta_T0 may not be a converged sample of the cell statistics. Part I performed a domain-size study, but only for one condition; the parametric variation here changes the instability cell size and intensity strongly, so the earlier convergence evidence cannot be transferred without an additional check. The reader's conditional verdict is appropriate: the qualitative conclusions are credible, but the quantitative model should be treated as provisional until sampling convergence is demonstrated or the model is tested on independent conditions. I therefore recommend keeping the verdict unchanged rather than strengthening or weakening it.","tokens_in":18280,"tokens_out":4328,"duration_ms":44336,"concrete_test":"For the p10 case (highest I0, strongest effect) and the Ref case, generate at least five statistically independent initial conditions by taking snapshots separated by O(100 tau_T0) from the already statistically stationary freely propagating flame simulation, and run one HOQ per snapshot; compute the mean and standard error of Phi_q,2D/Phi_q,1D. Separately, repeat one high-pressure case with Ly = 200 delta_T0 to test domain-size sensitivity. If the realization-to-realization or domain-size shift in the mean ratio exceeds about 10% of the value reported in Fig. 11, or if the refit slope b in Eq. 9 changes by more than the point scatter, the single-realization statistics are the limiting uncertainty and the quantitative model should be re-fit or qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative core of the paper is the mean quenching wall-heat-flux ratio in Fig. 11 and its linear correlation with I0 in Fig. 12a, used to build the two-step model (Eqs. 9-11). For each operating condition, this mean is computed from a single 2D HOQ realization on Ly = Lx = 100 delta_T0 (Sec. 2.1.2). The +/-1 sigma band in Fig. 11 is the spatial scatter along the wall, not an uncertainty of the mean; it cannot tell whether the periodic box contains enough instability cells to make the mean representative. Part I's domain-size study is cited for the link between mean heat flux and I0, but it was performed at one operating condition, so it does not certify convergence across phi, Tu, and p variations, especially where cell sizes and intensities change strongly (e.g., p10, phi0.4). If the spatial average is under-sampled, the positions of the data points in Fig. 12a change, the constants a and b in Eq. 9 and hence e and f in Eq. 11 are refit, and the claimed strong correlation across all operating conditions is not established. The qualitative finding that thermodiffusive instabilities increase wall heat flux may survive; the quantitative predictive model is the part at risk.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends Part I's single-condition study of flame-wall interaction of intrinsically unstable hydrogen/air flames to a parametric space of equivalence ratio (0.4–1.0), unburnt temperature (298–700 K), and pressure (1.01325–20 bar). One-dimensional head-on quenching (HOQ) simulations provide stable-flame baselines, and two-dimensional HOQ simulations capture the effects of thermodiffusive and hydrodynamic instabilities. The central claims are that thermodiffusive instabilities significantly increase the mean wall heat flux and reduce the mean quenching distance relative to one-dimensional stable flames, that hydrodynamic instabilities have negligible influence on quenching quantities, and that the relative increase in wall heat flux correlates linearly with the reactivity factor I0. A joint model (Eqs. 9–11) is proposed to estimate the heat-flux increase from one-dimensional flame quantities, namely the Zeldovich number and the convection-diffusion Péclet number.","tokens_in":18576,"tokens_out":2791,"duration_ms":27426,"significance":"If the quantitative model holds beyond the fitted dataset, it would be a practically valuable tool for estimating wall thermal loads in hydrogen combustion systems from inexpensive one-dimensional simulations. The paper contributes a broad, detailed DNS dataset that is otherwise scarce in the literature, and the qualitative picture—thermodiffusive instability intensity controls the heat-flux enhancement while hydrodynamic instability does not—is convincingly supported by the PDFs, spatial profiles, and the contrast between the Tu700 and p10 cases. The explicit use of one-dimensional reference flames for normalization is a strength. However, the quantitative model currently rests on in-sample fits with no uncertainty quantification or out-of-sample validation, so the predictive claim is not yet established to the standard required for a modeling paper.","major_comments":[{"comment":"The model coefficients a, b, c, d (and hence e, f) are fitted to the same dataset used to demonstrate agreement. The I0 values come from the freely propagating simulations and the heat-flux ratios come from the HOQ simulations, but both are generated at the same operating conditions and no independent validation is performed. Fig. 13 shows in-sample comparisons only, and no confidence intervals or goodness-of-fit metrics (e.g., R² or RMS error) are reported. This is load-bearing because the abstract and Section 4.2 claim the model can 'estimate' the heat-flux increase. Please add an out-of-sample test (e.g., leave-one-condition-out cross-validation on the four parametric series) or otherwise demonstrate that the linear relationships are not artifacts of the particular set of operating conditions.","section":"§4.2, Eqs. (9)–(11)"},{"comment":"The mean and standard deviation of Φq,2D/Φq,1D are computed from a single 2D realization per operating condition on a periodic domain of Ly = Lx = 100δ_T. The ±1σ band in Fig. 11 is the spatial scatter along the wall, not an uncertainty of the mean; it does not quantify whether the periodic box contains enough instability cells for the spatial average to be statistically representative. The authors cite the domain-size study in Part I, but that was performed at a single operating condition, and the present paper does not provide a convergence check for the conditions where cell sizes and intensities change strongly (e.g., p10, phi0.4). If the spatial average is under-sampled, the data points in Fig. 12a shift, the fitted coefficients in Eqs. (9)–(11) change, and the claimed universal linear correlation is not established. Please include at least a representative convergence study (multiple domain sizes or realizations) for two or three conditions spanning the instability-intensity range, and report the resulting sensitivity of the fit coefficients.","section":"§2.1.2, Fig. 11"},{"comment":"The claim that 'a linear relationship also holds across all operating conditions' is based on a visual fit with no reported uncertainty. The data set contains only ten operating conditions spanning three distinct parametric series, and the text does not report residuals or per-series slopes. Given that the two-step model in Eq. (11) is derived by composing two linear fits, any deviation from linearity in Fig. 12a or 12b propagates directly into the final model. Please quantify the fit quality (e.g., R², residual plots) and discuss whether the linear form is adequate across the full range of Ze/Pecld, rather than merely being the simplest two-parameter choice.","section":"§4.2, Fig. 12a"}],"minor_comments":[{"comment":"The abstract says 'a novel model fit is proposed' and the novelty statement repeats this; given that the fit is a linear correlation with no independent validation, the wording 'model fit' is appropriate but 'estimate' should be tempered to 'correlate' or 'suggest a predictive relationship' until out-of-sample performance is demonstrated.","section":"Abstract / Section 1"},{"comment":"In the bullet on the pressure variation at Tu = 700 K, the text says 'the trends for the laminar flame speed and thermal flame thickness are similar to those at Tu = 700 K', which appears to be a typo; it should likely read 'similar to those at Tu = 298 K'.","section":"§3.1"},{"comment":"The acronym 'HQO' appears in the caption and in the text where 'HOQ' is intended; please make the terminology consistent throughout.","section":"§4.2 (Figure 9 caption)"},{"comment":"The table lists 'Pecld' with an accented 'e' in the header, but the text uses 'Pécd' and 'Peclet' in several places; please standardize the notation and define all symbols in the table caption.","section":"§2.2, Table 1"},{"comment":"The paragraph introducing Fig. 12 states that the relationship between the relative heat flux and I0 was shown in Part I 'through varying the numerical domain sizes', but Part I is cited as 'in review'; please ensure the citation includes the arXiv identifier and that the claim is verifiable from the cited document.","section":"§4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid parametric extension of the authors' Part I, but the central quantitative claim—the predictive linear model for the relative wall-heat-flux increase—is currently an in-sample correlation without uncertainty quantification or cross-validation. This is the main barrier to acceptance. The qualitative findings are convincing and likely correct, but the modeling claim needs either additional validation or a more circumscribed statement. The fit to the editor is good for a combustion/fluids journal, though the novelty relative to Part I and the existing scaling literature (Berger, Rieth) is incremental rather than transformative."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a straightforward, honest parametric extension of the authors' Part I. The new deliverable is the systematic map of how thermodiffusive instabilities alter head-on quenching across phi (0.4-1), Tu (298-700 K), p (1-20 bar), plus a correlation that ties the wall-heat-flux enhancement to the reactivity factor I0, and a two-step model that estimates I0 from 1D flame quantities (Ze/Pec_d). That is genuinely useful for people estimating thermal loads in lean hydrogen systems.\n\nThe paper earns its keep on the qualitative side. The 1D vs 2D comparisons, PDFs, and the contrasting cases (e.g., Tu700 where only hydrodynamic instabilities are active) convincingly show that the heat-flux enhancement tracks thermodiffusive intensity, and that hydrodynamic wrinkling alone does not change the mean quenching flux. The setup follows Berger et al. and Howarth et al., so the operating-condition trends are consistent with the literature, which adds credibility.\n\nWhere it gets soft is the quantitative layer. Each operating condition rests on one 2D realization; the ±1σ band in Fig. 11 is the spatial scatter along the wall, not the uncertainty of the mean, and there is no run-to-run or domain-size convergence check for the heat-flux statistics. The cell count in a 100δ_T box is probably adequate for the most intense cases, but for mild instability (e.g., p at 1 bar, phi 0.65) the number of cells is smaller, so the mean ratio could shift. Second, Eqs. 9-11 are linear fits evaluated on the same data used for the fit: no confidence intervals, no leave-one-out, no independent test. The stress-test note is right that the coefficients would be refit if new realizations moved the points. That said, the authors only claim a \"model fit\" and an \"estimate,\" not a validated prediction, so this is a limitation to fix in revision, not a fatal flaw. The strongest version of the paper would add one ensemble case (or a smaller-domain repeat) and an out-of-sample check, even a crude one.\n\nI'd send it to review. The combustion/FWI community will want this parametric data on record, and the concerns are addressable with additional analysis rather than a redo. I'd cite the paper for the parametric trends, less for the model constants.","headline":"A solid, useful parametric extension of Part I; the qualitative conclusions hold, but the quantitative correlation is a same-data linear fit atop single-realization statistics and needs a convergence/out-of-sample check before it is used as a predictive model.","tokens_in":19101,"tokens_out":3472,"would_cite":true,"duration_ms":29526,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Thermodiffusive instabilities, not hydrodynamic ones, drive the enhanced wall heat flux and shorter quenching distance in hydrogen/air flames, and the enhancement scales linearly with the reactivity factor I0 across all tested operating…","keywords":["flame-wall interaction","head-on quenching","thermodiffusive instability","hydrogen/air flames","wall heat flux","quenching distance","reactivity factor","direct numerical simulation"],"falsifier":"Run several independent two-dimensional head-on quenching simulations for one high-pressure case (for example $p=10$ bar, $\\phi=0.5$, $T_u=298$ K) with different initial perturbation seeds and lateral domain sizes, and check whether the ensemble mean of $\\Phi_{q,2D}/\\Phi_{q,1D}$ still lies on the line $-1.16 + 2.07 I_0$; a deviation larger than the reported one-standard-deviation spread would undermine the proportionality claim.","tokens_in":18084,"feed_emoji":"🔥","tokens_out":7155,"duration_ms":58804,"temperature":0.7,"pith_summary":"The paper asks whether intrinsic flame instabilities change how hydrogen/air flames quench against a solid wall, and how that change depends on operating conditions. It runs detailed two-dimensional simulations of head-on quenching across a wide range of equivalence ratios, unburnt temperatures, and pressures, and compares them with one-dimensional stable-flame references. The central finding is that thermodiffusive instabilities, not hydrodynamic ones, substantially increase the mean wall heat flux and reduce the mean quenching distance, with the effect growing at lower equivalence ratio, lower unburnt temperature, and higher pressure. The relative increase in wall heat flux is linearly proportional to the reactivity factor $I_0$ across every condition tested. If true, this means one-dimensional quenching simulations systematically underestimate the thermal load on combustor walls for the lean hydrogen conditions of practical interest.","feed_headline":"Thermodiffusive instability can triple wall heat flux at quenching","feed_subtitle":"Lean, low-temperature, high-pressure H2/air flames push up to 3x more heat into the wall than 1D models predict.","key_machinery":"The load-bearing object is the reactivity factor $I_0 = (s_c/s_l)/(A/L_y)$, the ratio of the normalized consumption speed to the normalized flame-surface length. It isolates the local enhancement of flame reactivity caused by thermodiffusive instabilities, separating it from the purely geometric surface-area increase that hydrodynamic instabilities also produce. The argument's other pillar is the ratio $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$ formed from the Zeldovich number (flame sensitivity to temperature) and the convection-diffusion Péclet number (balance of convective and diffusive transport), taken from one-dimensional freely propagating flames. The two linear fits—$I_0$ against $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$, and $\\Phi_{q,2D}/\\Phi_{q,1D}$ against $I_0$—are chained through a composition of linear functions, yielding a single estimate of wall-heat-flux enhancement from one-dimensional flame quantities. The simulations use a periodic two-dimensional domain of $100\\delta_T^0 \\times 100\\delta_T^0$ with detailed hydrogen chemistry, and one-dimensional head-on quenching under identical conditions supplies the normalization.","core_discovery":"On the paper's own terms, the discovery is that when a hydrogen/air flame is intrinsically unstable and quenches head-on against a wall, the thermodiffusive part of the instability—not the hydrodynamic part—controls the thermal impact. For a given operating condition, the mean quenching wall heat flux of the two-dimensional unstable flame is enhanced relative to a one-dimensional stable flame by an amount that depends only on the reactivity factor $I_0$, which measures local enhancement of flame reactivity beyond what flame-surface wrinkling would produce. The data collapse onto the linear relation $\\Phi_{q,2D}/\\Phi_{q,1D} = -1.16 + 2.07 I_0$ over the whole parameter space, and $I_0$ itself is fit by $I_0 = 0.978 + 0.0115\\,\\mathrm{Ze}/\\mathrm{Pe}_{cd}$ from one-dimensional freely propagating flame quantities. Combining the two gives a model $\\Phi_{q,2D}/\\Phi_{q,1D} = 0.864 + 0.0238\\,\\mathrm{Ze}/\\mathrm{Pe}_{cd}$, meaning the peak wall heat load of an unstable flame can be estimated from one-dimensional flame data alone. The paper concludes that hydrodynamic instabilities leave the quenching characteristics essentially unchanged, while thermodiffusive instabilities can raise the mean wall heat flux by up to a factor of about three in the high-pressure lean cases.","pith_inferences":["A natural testable extension is to repeat the same quenching statistics in three dimensions; if 3D cellular dynamics redistribute wall heat flux differently, the 2D-derived slope and offset would need recalibration.","The same $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$ pathway may transfer to other low-Lewis-number fuels or hydrogen blends, but the fitted constants should be re-evaluated per fuel since they are empirical.","The proportionality could be tested locally, not just on the mean, by conditioning local quenching wall heat flux on local flame curvature or local equivalence ratio; the paper presents PDF evidence but does not build a local model.","The proposed model offers a cheap design-rule route: a one-dimensional flame calculation plus the fits gives an early estimate of whether a given operating point carries a thermal-load penalty before expensive multi-dimensional simulations are run."],"forward_implications":["For lean hydrogen/air flames at low unburnt temperature and high pressure, one-dimensional head-on quenching simulations underestimate the mean wall heat flux; in the 10 bar, $\\phi=0.5$, 298 K case the enhancement reaches about a factor of three.","Hydrodynamic instabilities can be neglected when predicting mean quenching heat flux and quenching distance; the governing quantity is the thermodiffusive reactivity factor $I_0$.","The relative wall-heat-flux increase follows $\\Phi_{q,2D}/\\Phi_{q,1D} = -1.16 + 2.07 I_0$ across all operating conditions examined.","Because $I_0$ is itself modeled from one-dimensional flame quantities through $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$, the combined model estimates combustor wall heat load without running a two-dimensional unstable-flame simulation.","Lower equivalence ratio, lower unburnt temperature, and higher pressure each intensify thermodiffusive instabilities and therefore increase the heat-flux enhancement relative to the one-dimensional baseline."],"supporting_citations":[{"why":"Supplies the two-dimensional head-on quenching configuration, initialization procedure, and the single-condition finding that this parametric study extends.","marker":"[1]"},{"why":"Defines the operating-condition matrix and establishes how instability intensity varies with equivalence ratio, temperature, and pressure, motivating the parameter space.","marker":"[12, 13]"},{"why":"Introduces the Zeldovich-number-to-Péclet-number ratio $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$ and the reactivity-factor scaling that the model fit is based on.","marker":"[21]"},{"why":"Justifies the choice of $L_y = 100\\delta_T^0$ so that no intrinsic instability length scales are suppressed, and provides the parameter-space visualization.","marker":"[16]"},{"why":"Provides the flame-finger phenomenology and the reactivity factor $I_0$ used to separate thermodiffusive from hydrodynamic effects.","marker":"[30]"},{"why":"Supplies the flame-surface definition and empirical scaling approach used to compute $A/L_y$ and consumption-speed enhancement.","marker":"[19]"},{"why":"Provides the detailed hydrogen reaction mechanism used in all one- and two-dimensional simulations.","marker":"[23]"}],"fun_headline_variants":["Thermodiffusive instability triples H2 flame wall heat flux","Unstable H2/air flames push wall heat up to 3x","Wall heat in H2 flames: thermodiffusive instability matters most","Quenching heat flux predicted from 1D flame data alone","H2 flame quenching: thermodiffusive instability rules, hydrodynamic doesn't"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative conclusions rest on a single two-dimensional realization per operating condition; if the periodic domain under-samples the cellular structures of the instabilities, the reported mean ratios and linear fits could shift.","fun_headline_variants_meta":{"raw":{"variants":["Thermodiffusive instability triples H2 flame wall heat flux","Unstable H2/air flames push wall heat up to 3x","Wall heat in H2 flames: thermodiffusive instability matters most","Quenching heat flux predicted from 1D flame data alone","H2 flame quenching: thermodiffusive instability rules, hydrodynamic doesn't"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000572,"raw_usage":{"total_tokens":2786,"prompt_tokens":1111,"completion_tokens":1675,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":727,"completion_tokens_details":{"reasoning_tokens":1580}},"tokens_in":727,"tokens_out":1675,"duration_ms":10306,"temperature":1.0,"reasoning_tokens":1580,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:30:43.607998+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run several independent two-dimensional head-on quenching simulations for one high-pressure case (for example $p=10$ bar, $\\phi=0.5$, $T_u=298$ K) with different initial perturbation seeds and lateral domain sizes, and check whether the ensemble mean of $\\Phi_{q,2D}/\\Phi_{q,1D}$ still lies on the line $-1.16 + 2.07 I_0$; a deviation larger than the reported one-standard-deviation spread would undermine the proportionality claim.","supporting_citations":[{"cited_title":"Flame-wall interaction of thermodiffusively unstable hydrogen/air flames -- Part I: Characterization of governing physical phenomena","cited_arxiv_id":"2411.17590","evidence_quote":"Supplies the two-dimensional head-on quenching configuration, initialization procedure, and the single-condition finding that this parametric study extends."},{"cited_title":"Rieth, A","cited_arxiv_id":null,"evidence_quote":"Introduces the Zeldovich-number-to-Péclet-number ratio $\\mathrm{Ze}/\\mathrm{Pe}_{cd}$ and the reactivity-factor scaling that the model fit is based on."},{"cited_title":"Berger, K","cited_arxiv_id":null,"evidence_quote":"Justifies the choice of $L_y = 100\\delta_T^0$ so that no intrinsic instability length scales are suppressed, and provides the parameter-space visualization."},{"cited_title":"Berger, M","cited_arxiv_id":null,"evidence_quote":"Provides the flame-finger phenomenology and the reactivity factor $I_0$ used to separate thermodiffusive from hydrodynamic effects."},{"cited_title":"Howarth and A","cited_arxiv_id":null,"evidence_quote":"Supplies the flame-surface definition and empirical scaling approach used to compute $A/L_y$ and consumption-speed enhancement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the detailed hydrogen reaction mechanism used in all one- and two-dimensional simulations."}],"review_version":1}