{"id":"a90fa3a6-e3f8-444a-a281-c7153572c71a","arxiv_id":"2412.10167","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Madwave3 is a parallelized quantum wave packet code for state-to-state triatomic reaction and photodissociation dynamics, validated against the ABC code on H+DH to H2+D.","lead":"MADWAVE3 is an open-source Fortran code that simulates quantum collisions and photodissociation in triatomic molecules, including reactions that jump between electronic states. It is validated on the benchmark reaction H + DH to H2 + D, where its results match the established ABC code.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The advertised nonadiabatic and photodissociation capabilities are not exercised by any benchmark; the single-state J=0 ABC comparison cannot validate those code paths, leaving the paper's broader central claim unsupported.","rationale":"The reader's verdict of CONDITIONAL is appropriate. The central validated result—single-state J=0 reactive/inelastic probabilities matching ABC—is well supported by the comparison in Figure 2, assuming the in-house PES is close enough to BKMP2 for a meaningful code-to-code comparison. However, the paper advertises features (multiple coupled diabatic states, photodissociation) that are not benchmarked, and the reader correctly identifies this as the weakest load-bearing assumption. A bug in those code paths would not affect the ABC comparison but would invalidate the advertised scope. I agree with the reader's recommendation to either add benchmarks for the multi-state and photodissociation paths or narrow the stated scope, and to pin the repository version and fix textual inconsistencies. I do not think the evidence warrants rejection: the code clearly works for the single-state collision case, and the unvalidated features are plausibly correct given the authors' prior work with related methods (e.g., Refs. [30, 36, 51, 53]). Thus the verdict should remain CONDITIONAL, pending the additional benchmarks or a revised scope. The concrete test proposed here—a two-state charge-transfer benchmark against published time-independent results—would directly settle whether the diabatic path is trustworthy and would be a decisive addition to the paper.","tokens_in":22748,"tokens_out":4431,"duration_ms":48066,"concrete_test":"Run MADWAVE3 for the J=0 nonadiabatic charge-transfer reaction H+ H2+ -> H2 + H+ using the coupled two-state diabatic H3+ surfaces employed in Ref. [36] (Sanz-Sanz et al., J. Chem. Phys. 154, 104104 (2021)), and compare the resulting state-to-state probabilities against the time-independent hyperspherical results reported there. If the agreement is not comparable to the H+DH benchmark shown in Figure 2, the diabatic machinery is not validated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim, as stated in the abstract and introduction, is that MADWAVE3 computes state-to-state probabilities for inelastic, reactive, and photodissociation processes over one or multiple coupled diabatic electronic states. The only quantitative validation is a single-electronic-state J=0 collision benchmark (H+DH -> H2+D and HD+H) compared to the time-independent ABC code. This benchmark exercises only the single-state, iphoto=0 code path. The diabatic-state machinery (Section 2.1, the electronic Hamiltonian in Eq. (5) with off-diagonal potential couplings), the photodissociation initial-state construction (Section 2.3.2, Eq. (13), including the dipole operator and bound-state Lanczos), and the product analysis in a coupled electronic-state manifold (Section 2.4, Eq. (18)) are not tested. If any of these untested paths contains an implementation error, the excellent agreement shown in Figure 2 would not reveal it, yet the advertised capabilities would fail. Furthermore, the in-house H3 potential used in the example is described only as reproducing BKMP2 'very well', with no quantitative comparison; if the potential differs subtly from BKMP2, the ABC comparison is not a strict code-to-code validation. The combination of an untested feature set and an unvalidated potential means the strongest evidence in the paper supports only the narrow single-state collision claim, not the full advertised scope.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents MADWAVE3, a Fortran90 time-dependent wave packet code for triatomic quantum dynamics. The theoretical framework uses body-fixed Jacobi coordinates, a real Chebyshev propagator with complex absorbing potentials, DVR/FBR transformations for the angular kinetic terms, and reactant- or product-coordinate-based analysis for state-to-state probabilities. The paper documents the modular structure, installation, input namelists, and auxiliary programs, and gives a case study of H+DH(v=0,j=0) at J=0: total and state-to-state probabilities are compared with the time-independent ABC code, partial-wave cross sections are shown, and MPI/OpenMP speedups are reported. The abstract advertises inelastic, reactive, and photodissociation processes over one or multiple coupled diabatic electronic states.","tokens_in":23025,"tokens_out":10915,"duration_ms":116650,"significance":"If the advertised capabilities are correct, MADWAVE3 is a useful open-source alternative to time-independent codes for triatomic state-to-state dynamics, and the repository with example inputs, the potential, and auxiliary analysis tools is a practical strength. The J=0 comparison against the independent ABC code is a genuinely non-circular check of the single-surface reactive/inelastic machinery, and the parallel-scaling study is informative. However, the evidence in the manuscript covers only a single-electronic-state, iphoto=0 collision benchmark; the nonadiabatic and photodissociation paths named in the abstract are not exercised. The paper would be considerably strengthened by quantitative convergence data and by at least one benchmark of the coupled-state or photodissociation code paths.","major_comments":[{"comment":"The only quantitative validation in the paper is for nelec=1 and iphoto=0. The off-diagonal electronic couplings in the Hamiltonian of Eq. (5), the coupled-electronic-state product basis of Eq. (18), and the photodissociation initial-state construction of Eqs. (13)-(16) are never exercised by any reported test. Since the abstract states that the code computes photodissociation processes and works over multiple coupled diabatic electronic states, the present evidence supports only the single-surface collision claim. Please add a validation case for at least one nonadiabatic or photodissociation process, or explicitly restrict the claims made in the abstract and conclusions to single-surface collisions.","section":"§6.2 and Figure 2; §2.1 Eq. (5); §2.3.2 Eq. (13); §2.4 Eq. (18)"},{"comment":"The agreement with ABC is described only as 'excellent', with no numerical deviations, no maximum or mean absolute error, and no systematic convergence study over the parameters listed in Table 1 (grid sizes, absorption parameters, number of Chebyshev iterations). As written, the benchmark is visual and cannot be independently verified quantitatively. Please report numerical differences from ABC and a convergence test for the key propagation and grid parameters.","section":"§6.2 and Figure 2"},{"comment":"Eq. (22) defines the cumulative probability with a free index Ω' on the S-matrix element but no sum over Ω'. For a state-to-state cross section to a final v',j', all final helicity projections must be summed. As written, the formula would omit the final-helicity degeneracy contribution and is incomplete. Please clarify whether the sum over Ω' is implicit in the code’s definition of the S-matrix files, and correct Eq. (22) accordingly.","section":"§2.5, Eq. (22)"}],"minor_comments":[{"comment":"The divisibility condition in Eq. (27) appears to be reversed: if nγ angular points are distributed over nγ_proc processors, the requirement should be that nγ is divisible by nγ_proc (mod(nγ,nγ_proc)=0), not that nγ_proc is divisible by nγ. The nproc values used in Figure 5 are inconsistent with the printed condition.","section":"§7, Eq. (27)"},{"comment":"The text says the inelastic probabilities are in the top panel and reactive probabilities in the bottom panel, while the caption says the reverse. The panel labels in the figure itself also disagree with the caption. Please make the text, caption, and panels consistent.","section":"Figure 2 caption and §6.2 text"},{"comment":"The in-house H3 potential is described only as reproducing BKMP2 'very well', with no quantitative comparison or fitting details. Since the ABC benchmark and the presented cross sections both use this potential, please document its deviations from BKMP2 or use the published BKMP2 surface so that the physical results are reproducible and independently checkable.","section":"§6.1 and §8"},{"comment":"The conclusions state that 'the D + H2+ → DH+ + H is presented as an use example', but the example in Section 6 is H + DH → H2 + D. Please correct this sentence.","section":"§8, Conclusions"},{"comment":"The state-to-state cross sections in Figure 4 are shown without any independent comparison or convergence test. Even a J>0 ABC comparison for one partial wave would help; alternatively, state explicitly that these cross sections are illustrative. Also, the notation 'nOM P proc' in Figure 5 and the statement that J=9 has Ω=0,...,10 (with nΩ=10) should be clarified, since Ω usually ranges only up to J.","section":"§6.3 and Figure 5"}],"recommendation":"major_revision","confidential_remarks":"The core J=0 single-surface validation against ABC and the availability of the repository are strong points, and I do not doubt the code’s usefulness. The main issue is that the abstract overclaims relative to the validation: the nonadiabatic and photodissociation paths are not tested. The missing Ω' sum in Eq. (22) should be checked carefully; if the code does sum over Ω', the equation is simply a typo, but it must be fixed because cross sections are a central advertised output."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a software release paper for MADWAVE3, a FORTRAN90 wave packet code for triatomic collisions and photodissociation. The genuinely new contribution is the combination in one MPI/OpenMP package: state-to-state reactive and inelastic scattering plus photodissociation, with an optional manifold of coupled diabatic electronic states. The single-surface J=0 benchmark against ABC for H+DH is a legitimate check, and the agreement shown in Fig. 2 is the right kind of evidence that the basic propagation machinery works.\n\nThe paper does several things well. The formalism section is reasonably complete. The code is open source with a worked example, and the parallelization strategy—over helicity and then over angular grid points—is sensible and clearly described. The authors are also honest about known issues like FFTW thread-safety and the difficulty of resonant states in RPD.\n\nThe soft spots are real but fixable. The headline capabilities—multiple coupled diabatic states and photodissociation—are never exercised. Every quantitative test is single-state, iphoto=0, so an implementation error in the diabatic couplings or the photoinitiation code would go completely undetected. The in-house H3 potential is only described as reproducing BKMP2 'very well', with no numbers; that weakens the ABC comparison as a strict code-to-code test. There are also several internal inconsistencies: the conclusion names D+H2+ instead of H+DH, figure captions swap 'inelastic' and 'reactive' panels, and Fig. 5's caption says J=9 with nΩ=1 while the text says J=10 with nΩ=10. The placeholder CPC and Code Ocean links and the lack of a pinned commit hash are minor but relevant in a reproducibility-focused software paper.\n\nNone of this sinks the paper. The narrow claim—that MADWAVE3 reproduces ABC for the J=0 H+DH case—appears well supported. The broader claim in the abstract is not yet supported, but it is testable. The authors should either add at least one benchmark that exercises the multi-state and photodissociation paths, or narrow the advertised scope. They should also quantify the difference between their H3 PES and BKMP2 and fix the textual inconsistencies.\n\nI would send this to peer review. It is a usable tool with a credible single-surface validation, and a good referee can push the authors to close the gap between what is advertised and what is demonstrated. I would not desk-reject it.","headline":"Solid single-surface wave packet code with a credible J=0 ABC check; the advertised diabatic and photodissociation paths are unbenchmarked and need either demonstrations or a narrowed scope.","tokens_in":23632,"tokens_out":3079,"would_cite":false,"duration_ms":30851,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper presents MADWAVE3, a parallel quantum wave-packet code that computes state-to-state probabilities for triatomic inelastic, reactive, and photodissociation processes over one or multiple coupled diabatic electronic states, and…","keywords":["quantum wave packet","reactive scattering","photodissociation","diabatic representation","state-to-state probabilities","Chebyshev propagator","triatomic systems","parallel computing"],"falsifier":"Run MADWAVE3 on a photodissociation case with published cross sections (for example the LiHF or HCN systems cited in the paper) and compare the computed absolute absorption cross section; or run a two-state nonadiabatic benchmark against MCTDH or known results. If the photodissociation autocorrelation (Eq. 24) or the state-to-state products from the diabatic coupling do not match, the central claim of multi-process capability would be contradicted.","tokens_in":22480,"feed_emoji":"⚛️","tokens_out":5722,"duration_ms":53462,"temperature":0.7,"pith_summary":"MADWAVE3 is a parallel Fortran 90 code for propagating quantum wave packets in triatomic systems, and the paper's claim is that it delivers converged state-to-state probabilities and cross sections for inelastic collisions, reactive collisions, and photodissociation, on one or several coupled diabatic electronic states. The authors validate this by benchmarking the H+DH(v=0,j=0) system at total angular momentum J=0: their state-to-state inelastic and reactive probabilities agree closely with results from the time-independent ABC code on the same potential. If the claim holds, the code offers a wave-packet alternative to ABC that scales more gently with channel number, supports nonadiabatic and photoinitiated processes that ABC does not implement, and parallelizes efficiently over helicity and angular grid points. The paper provides the potential surface, input files, and installation instructions so the benchmark is reproducible.","feed_headline":"MADWAVE3 matches reference reaction probabilities","feed_subtitle":"Open-source Fortran code also targets photodissociation and coupled diabatic states.","key_machinery":"The load-bearing machinery is the modified Chebyshev propagator of Mandelshtam and Taylor: the wave packet is expanded as real Chebyshev components Ψ(k) that stay real when the initial packet is real, and the Hamiltonian action is split into radial kinetic terms evaluated with FFTW sine transforms, angular kinetic terms applied via DVR-to-FBR transformations, and a potentially L-shaped grid for the potential. To extract state-to-state probabilities the code uses either reactant-coordinate-based (RCB) or product-coordinate-based (PCB) sequential coordinate transformations, and the analysis is done through flux and projection coefficients at a fixed product distance. Parallelization is done by distributing helicity Ω components (coupled only to Ω±1) and, for low J, the γ angle grid.","core_discovery":"The central claim is that MADWAVE3 computes correct state-to-state S-matrix elements, and hence probabilities and cross sections, for triatomic A+BC → AB+C reactions and ABC+hν → AB+C photodissociation, including multiple coupled diabatic electronic states. The machinery works in body-fixed Jacobi coordinates (r, R, γ) with the wave packet expanded in helicity Ω and electronic components; propagation uses a modified Chebyshev scheme that keeps the wave packet real, and state-to-state analysis is done by projecting onto product or reactant Jacobi coordinates via sequential transformations. The key demonstration is the H+DH(v=0,j=0) example: for J=0, the MADWAVE3 state-to-state probabilities for HD and H2 product channels are reported to be in excellent agreement with the ABC hypherspherical close-coupling code, which the authors take as evidence of convergence and accuracy of the grids, basis, and propagation parameters.","pith_inferences":["The paper's advertised nonadiabatic and photodissociation capabilities rest on untested code paths: the only benchmark is single-electronic-state J=0 reactive/inelastic scattering, so a bug in the diabatic coupling or in the Eq. (13) initial-state construction would not be caught by the ABC comparison. A natural next test is to reproduce a published photodissociation cross section (e.g., HCN or Li","Because the code reads user-supplied potential routines and transition dipoles, its practical accuracy for a new molecular system will depend on the quality of the diabatization, which the code itself does not build; users must supply a diabatic representation that is diagonal in the reactant asymptotic channel.","The J-shifting interpolation results shown for higher J indicate that approximate cross sections can be obtained from a few partial waves, so the code's advertised scaling advantage matters most when many J or many channels are needed."],"forward_implications":["For systems where the ABC code becomes expensive—high total angular momentum, deep wells, or many channels—MADWAVE3 is claimed to be a better-scaled alternative because it computes one column of the S-matrix and parallelizes over helicity and angular grid points.","The code extends state-to-state wave-packet dynamics to photodissociation, using a first-order perturbation-theory initial packet built from a bound state and transition dipole moments, which ABC does not provide.","The diabatic multi-state treatment allows nonadiabatic reactive and photoinitiated processes to be studied in a single framework, with products' rovibrational states computed in the full electronic manifold when needed.","The accompanying potential grid, input files, and analysis programs (for cross sections via J-shifting interpolation) make the benchmark fully reproducible, so users can check convergence parameters before running new systems."],"supporting_citations":[{"why":"Supplies the time-independent hyperspherical close-coupling results used as the benchmark for the J=0 state-to-state probabilities.","marker":"[3]"},{"why":"Provides the sequential reactant/product Jacobi coordinate transformation method used to extract state-to-state probabilities (RCB/PCB).","marker":"[2]"},{"why":"Introduces the modified Chebyshev propagator with damping that the code uses to keep Chebyshev components real.","marker":"[22]"},{"why":"Original Chebyshev polynomial propagation scheme that the modified propagator builds on.","marker":"[23]"},{"why":"Defines the real-wavepacket initial condition and the analysis of outgoing flux/projection coefficients for S-matrix elements.","marker":"[28]"},{"why":"Describes the DVR-to-FBR angular kinetic treatment and the helicity-index parallelization strategy.","marker":"[1]"},{"why":"Gives the first-order perturbation theory construction of the photodissociation initial wave packet used in Eq. (13).","marker":"[30]"}],"fun_headline_variants":["MADWAVE3: quantum wave packet code for triatomic reactions","New Fortran code simulates triatomic reactions and photodissociation","MADWAVE3 parallelized for state-to-state quantum dynamics","Open-source code for nonadiabatic triatomic dynamics","MADWAVE3 achieves ABC-level accuracy in H+DH test"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the code's photodissociation and multi-diabatic-state machinery is as correct as the single-electronic-state reactive machinery, because the ABC benchmark exercises only the latter and would not expose an error in the former.","fun_headline_variants_meta":{"raw":{"variants":["MADWAVE3: quantum wave packet code for triatomic reactions","New Fortran code simulates triatomic reactions and photodissociation","MADWAVE3 parallelized for state-to-state quantum dynamics","Open-source code for nonadiabatic triatomic dynamics","MADWAVE3 achieves ABC-level accuracy in H+DH test"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000595,"raw_usage":{"total_tokens":2781,"prompt_tokens":938,"completion_tokens":1843,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":554,"completion_tokens_details":{"reasoning_tokens":1751}},"tokens_in":554,"tokens_out":1843,"duration_ms":12565,"temperature":1.0,"reasoning_tokens":1751,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T16:16:27.635520+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run MADWAVE3 on a photodissociation case with published cross sections (for example the LiHF or HCN systems cited in the paper) and compare the computed absolute absorption cross section; or run a two-state nonadiabatic benchmark against MCTDH or known results. If the photodissociation autocorrelation (Eq. 24) or the state-to-state products from the diabatic coupling do not match, the central claim of multi-process capability would be contradicted.","supporting_citations":[{"cited_title":"Skouteris, J","cited_arxiv_id":null,"evidence_quote":"Supplies the time-independent hyperspherical close-coupling results used as the benchmark for the J=0 state-to-state probabilities."},{"cited_title":"G´ omez-Carrasco, O","cited_arxiv_id":null,"evidence_quote":"Provides the sequential reactant/product Jacobi coordinate transformation method used to extract state-to-state probabilities (RCB/PCB)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the modified Chebyshev propagator with damping that the code uses to keep Chebyshev components real."},{"cited_title":"Tal-Ezer, R","cited_arxiv_id":null,"evidence_quote":"Original Chebyshev polynomial propagation scheme that the modified propagator builds on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the real-wavepacket initial condition and the analysis of outgoing flux/projection coefficients for S-matrix elements."},{"cited_title":"Roncero, D","cited_arxiv_id":null,"evidence_quote":"Describes the DVR-to-FBR angular kinetic treatment and the helicity-index parallelization strategy."},{"cited_title":"Paniagua, A","cited_arxiv_id":null,"evidence_quote":"Gives the first-order perturbation theory construction of the photodissociation initial wave packet used in Eq. (13)."}],"review_version":1}