{"id":"ff3b54f8-194a-4dc5-8eee-f542396a1258","arxiv_id":"2505.09418","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"DIALECT implements FMO-LC-TDDFTB to compute excited-state spectra, nonadiabatic dynamics, and polaritonic couplings of large molecular aggregates.","lead":"DIALECT is a new open-source software package for computing how large molecular assemblies absorb light, and for simulating energy and charge transfer between molecules, including inside optical cavities. It brings together fragment-based tight-binding quantum chemistry with mixed quantum-classical dynamics, making studies of systems with thousands of atoms feasible for a wider community.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline capacity claim leans on FMO-LC-TDDFTB's fragment-excited Hamiltonian and on Eq. 31's zero inter-fragment nonadiabatic coupling; neither is validated here, so the paper supports implementation but not yet quantitative capability.","rationale":"The paper is best read as a software-release paper: it demonstrates that DIALECT runs, gives plausible examples, and provides open-source code and data. That is real evidence. However, the central claim in Section 5 says the package is capable of simulating ground and excited state properties and nonadiabatic dynamics for large assemblies. That capability is only meaningful if the FMO-LC-TDDFTB Hamiltonian and the Ehrenfest equations are faithful. The least secured piece is the statement, immediately after Eq. 31, that nonadiabatic couplings vanish between quasi-diabatic states on different fragments. The LE and CT states are obtained as TDA solutions of independent fragment calculations; they are not constructed through an explicit diabatization procedure that would make inter-fragment derivative couplings small. The paper gives no numerical estimate of the omitted Dnm terms and no dynamics comparison with and without them. Since Section 3.3 uses population transfer rates as proof of principle, a systematic error in Dnm would directly compromise the headline dynamics capability. I therefore agree with the reader's weakest-assumption identification and with CONDITIONAL: the software may be correct, but the dynamical accuracy claim is currently unvalidated. I did not find an internal mathematical inconsistency that would force rejection; the concern is an uncontrolled approximation, not a contradiction. A focused numerical test of the omitted couplings can settle it.","tokens_in":20089,"tokens_out":6828,"duration_ms":80406,"concrete_test":"Run DIALECT's Ehrenfest dynamics on a small anthracene chain, e.g., 8 monomers, on the same set of trajectories in two ways: (i) as implemented, with inter-fragment Dnm set to zero, and (ii) with Dnm computed by finite-difference overlaps of the quasi-diabatic states along those trajectories, implemented in a test branch if needed. If the maximal omitted |Dnm| exceeds 10 percent of the corresponding HExc-FMO matrix element, or if the monomer-population transfer curves differ by more than about 20 percent, then the zero-coupling assumption is not negligible and the headline dynamics claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing assumption is not merely that the code runs, but that the FMO-LC-TDDFTB excited-state Hamiltonian built from monomer and dimer fragments (Eqs. 12, 16–22) and the Ehrenfest propagation of Eq. 31 reproduce the dynamics of the full assembly. The text after Eq. 31 states that nonadiabatic coupling is limited to the LE and CT states on the same fragments and is zero between all other quasi-diabatic states. These states are never explicitly diabatized across fragments; they are just fragment-local TDA solutions. For close-packed anthracene or naphthalene aggregates, nuclear motion modulates inter-fragment overlaps and orbital energies, so derivative couplings between LE states on different fragments need not vanish. If they are comparable to the excitonic couplings that drive energy transfer, the exciton population dynamics in Section 3.3 is biased in an uncontrolled way. Similarly, the entire spectrum rests on the unbenchmarked monomer/dimer reconstruction. No comparison to full-system LC-TDA-DFTB or to high-level reference data is provided in this paper. Thus the central claim of quantitative simulation capability is plausible but unverified; the weakest link is the zero inter-fragment nonadiabatic coupling in Eq. 31.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript introduces DIALECT, an open-source Rust package implementing DFTB2, DFTB3, LC-DFTB, FMO-LC-TDDFTB, surface-hopping and decoherence-corrected Ehrenfest nonadiabatic dynamics, and a Tavis-Cummings-based polaritonic coupling scheme. The theoretical part recapitulates the authors' earlier method papers (Refs. 58–60). Three example calculations are presented: LC-TDDFTB surface-hopping photodynamics of cis-stilbene with product branching ratios, polariton dispersion of a 203-molecule naphthalene cluster, and FMO-LC-TDDFTB Ehrenfest exciton transfer in a 30-molecule anthracene chain. Timings for water and tetracene clusters with up to about 9000 atoms are reported. The stated goal is to demonstrate that DIALECT can simulate excited-state spectra and dynamics in large molecular assemblies from weak to strong light-matter coupling regimes.","tokens_in":20327,"tokens_out":9045,"duration_ms":86838,"significance":"The value of the paper, if the implementation is sound, is as a publicly available tool that brings FMO-LC-TDDFTB and polaritonic excited-state calculations to a community that currently lacks a single package for these tasks. The authors deserve credit for open-sourcing the code, depositing example data, and reporting wall-clock timings; the examples run and the equations are consistent with the authors' prior work. However, the quantitative claims are not yet backed by adequate validation: the cis-stilbene branching ratios differ substantially from reference methods, the anthracene dynamics rests on 10 trajectories, and the central FMO-LC-TDDFTB reconstruction is not benchmarked against full-system or high-level references. The paper is therefore best read as an implementation report, and the stronger capability claims in Section 5 should be softened.","major_comments":[{"comment":"The excited-state Hamiltonian of a full assembly is reconstructed from monomer and dimer fragment TDA calculations. This reconstruction is the central approximation of FMO-LC-TDDFTB, but the manuscript provides no validation: it does not compare a computed spectrum with a full-system LC-TDA-DFTB calculation on a small aggregate, nor with high-level reference data, and it does not test how the results depend on the number of included LE and CT states. Because the polariton dispersion (Section 3.2) and the exciton dynamics (Section 3.3) both inherit this Hamiltonian, the quantitative capability claim in Section 5 is not supported by the evidence presented here. The authors should either include such a benchmark or explicitly state that the examples are algorithmic demonstrations, not accuracy benchmarks.","section":"2.3, Eqs. (12)-(22)"},{"comment":"The text following Eq. (31) states that the nonadiabatic coupling is limited to LE and CT states on the same fragment and is zero between all other quasi-diabatic states. These states are fragment-local TDA solutions; they are not obtained by an explicit diabatization across fragments. For a close-packed aggregate, nuclear motion changes inter-fragment overlaps and fragment orbital energies, so derivative couplings between LE states on different fragments need not vanish. If these couplings are comparable to the excitonic couplings that drive energy transfer, the population dynamics in Figures 4 and 5 would be biased in an uncontrolled way. Please provide a numerical estimate of the neglected inter-fragment derivative couplings for a representative geometry, or benchmark the Ehrenfest dynamics against a method that includes them.","section":"2.5, Eq. (31)"},{"comment":"The anthracene example is the only FMO-LC-TDDFTB nonadiabatic dynamics demonstration in the paper, and it is based on 10 trajectories per method; the authors themselves call it a proof of concept. That is acceptable for an implementation paper, but the concluding statement in Section 5 that DIALECT can be used to simulate localized exciton transport goes beyond what the data show. In addition, two tunable parameters enter this example—the atom-specific TAB decoherence parameter alpha_R in Eq. (35) and the harmonic restraint force constant—without any sensitivity analysis. Please add a sensitivity study, increase the number of trajectories, or explicitly restrict the conclusion to algorithmic demonstration.","section":"3.3 and 4.2"},{"comment":"The number of LE states per monomer and CT states per pair is chosen as an input (three LE and one CT per pair in anthracene; four LE and one CT per pair in tetracene) with no convergence test with respect to this basis truncation. The excitonic Hamiltonian is only as complete as the selected diabatic basis, and omitted charge-transfer or higher-lying LE states can change both spectra and dynamics. A convergence study for a small aggregate, or a physical justification for the truncation, is needed before the spectra and dynamics can be considered converged.","section":"2.3, 3.3, 4.2"}],"minor_comments":[{"comment":"In Eq. (4), the long-range gamma-matrix is written as erf(C_lr_AB) R_AB / R_AB; the argument of the error function appears to be missing a factor C_lr_AB R_AB. Please compare with Eq. (2) and correct.","section":"2.1, Eq. (4)"},{"comment":"In Eq. (24), E_n(R(t)) c_n(t) is missing the closing parenthesis on R(t); the equation should read E_n(R(t)) c_n(t).","section":"2.4, Eq. (24)"},{"comment":"The acronym 'FMO-LC-TDDTB' appears in the first paragraph of Section 2.6; it should be 'FMO-LC-TDDFTB'.","section":"2.6"},{"comment":"The parameter set is called 'ob263' in Sections 3.1 and 3.2 but 'ob2' in Section 3.3; please clarify whether these are the same parameter set.","section":"3.1 and 3.3"},{"comment":"The paper's title promises 'weak to strong light-matter coupling regimes', but the polariton example only exercises the strong-coupling regime; a sentence explaining what the weak-coupling capability corresponds to in the implemented formalism would help.","section":"3.2"},{"comment":"The monomer labels 'M. 1-3' etc. in Figure 4 are not defined in the text; please state how the monomers are ordered along the anthracene chain.","section":"Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"This is essentially a software paper whose methodological core was published in the authors' earlier papers (Refs. 58–60). The novel contribution is the open-source package and the demonstrations. The main risk is overclaiming quantitative capability; the revision should focus on benchmarks and softened claims rather than on new methodology. The paper fits the journal's scope, and I do not see a circularity problem: the examples do not fit parameters to the target results, though alpha_R and the harmonic restraint are user-chosen."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should read this as a software announcement, not as a validation paper. The genuinely new thing is DIALECT itself: an open-source Rust code that packages the group's previously published FMO-LC-TDDFTB method with surface hopping, decoherence-corrected Ehrenfest dynamics, and cavity coupling. The examples show the code runs and produces plausible results, and the timings are meaningful: ground-state SCC scales nearly linearly, and an excited-state calculation on 8850 atoms takes about an hour. The repo and supporting data are on GitHub, which makes the package immediately usable. I also give the authors credit for reporting the cis-stilbene branching ratios honestly against CASPT2 and OM3 results, even though the agreement is poor, and for explicitly labeling the anthracene dynamics a proof of concept with only 10 trajectories.\n\nThe soft spots are real but proportionate. The largest one is what the stress-test note flagged: after Eq. 31, the paper states that nonadiabatic coupling between quasi-diabatic states on different fragments is zero. Those states are just fragment-local TDA solutions, not states that were diabatized across fragments, so the assumption is not self-evident for close-packed aggregates. The paper provides no benchmark against full-system TDA-DFTB or high-level reference data, so the exciton dynamics in Section 3.3 rest on an untested approximation. The same is true for the monomer/dimer reconstruction of the excited-state Hamiltonian that underlies the spectra and polariton dispersion. In addition, the claim that systems over 10000 atoms are feasible is an extrapolation, since the largest benchmark is 8850 atoms. Finally, the reproducibility artifacts would benefit from a pinned version of the code. None of these are fatal for a software paper, but they do bound what the paper can claim.\n\nWho should read it: anyone planning to run semiempirical excited-state dynamics on large molecular aggregates or polaritonic systems. The package fills a practical gap, and the paper tells you what it does and roughly how fast it is. It does not tell you how accurate the underlying fragment approximation is in a new regime.\n\nMy recommendation: send it to peer review. It deserves referee time, but the referee should push for a small full-system benchmark (e.g., a dimer or trimer where FMO-LC-TDDFTB can be compared to the unfragmented calculation) and for a versioned code release. With those, the paper would be solid.","headline":"A genuinely useful open-source implementation of the group's FMO-LC-TDDFTB methods, but the paper validates the software's operation rather than the underlying fragment reconstruction, so the quantitative capability claims should be read as plausible but not proven.","tokens_in":20885,"tokens_out":1793,"would_cite":true,"duration_ms":20676,"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":"DIALECT is an open-source package that computes excited-state spectra and nonadiabatic dynamics for molecular assemblies of thousands of atoms, from weak to strong light-matter coupling.","keywords":["DIALECT","FMO-LC-TDDFTB","exciton dynamics","nonadiabatic molecular dynamics","polariton dispersion","density-functional tight-binding","fragment molecular orbital","charge transfer"],"falsifier":"Benchmark the method against a full-system TD-DFTB or high-level ab initio calculation on a small aggregate, for example a chain of five to ten anthracene or naphthalene molecules, and compare excited-state energies, oscillator strengths, and exciton-transfer dynamics; significant deviations beyond roughly 0.1 eV or qualitatively different population dynamics would disprove the fragment-pair assumption and the neglect of inter-fragment nonadiabatic couplings.","tokens_in":19877,"feed_emoji":"⚛️","tokens_out":6619,"duration_ms":56919,"temperature":0.7,"pith_summary":"The paper introduces DIALECT, a Rust-based simulation package that treats excited states and dynamics of large molecular assemblies by dividing the system into fragments, computing locally excited and charge-transfer states for each monomer and pair, and diagonalizing the resulting excitonic Hamiltonian. The central claim is that this FMO-LC-TDDFTB approach is efficient enough to handle assemblies with thousands of atoms, while still describing nonadiabatic dynamics, exciton transport, and polaritonic coupling to microcavities. If correct, DIALECT would let researchers model energy and charge transfer in molecular materials, biological aggregates, and optoelectronic devices at a scale previously out of reach for first-principles methods.","feed_headline":"Software simulates exciton dynamics in 10,000-atom assemblies","feed_subtitle":"DIALECT couples tight-binding fragments with cavity modes to track energy and charge flow.","key_machinery":"The central object is the quasi-diabatic excitonic Hamiltonian, whose basis comprises locally excited states on each molecular fragment and charge-transfer states between pairs of fragments, with matrix elements obtained from long-range-corrected Tamm-Dancoff density-functional tight-binding calculations on monomers and dimers. Exciton couplings are expressed through transition charges and the long-range-corrected $\\gamma$ matrix, and nonadiabatic dynamics are propagated either by Tully surface hopping or by Ehrenfest dynamics with a collapse-to-block decoherence correction. For strong light-matter coupling, photonic basis states are appended to the Hamiltonian with light-matter matrix elements proportional to the transition dipole moments of the LE states.","core_discovery":"The discovery is a unified computational framework in which a quasi-diabatic excitonic Hamiltonian containing locally excited (LE) and charge-transfer (CT) states is built from monomer and dimer fragment calculations, then coupled to cavity photon modes. The paper demonstrates this machinery by computing the polariton dispersion of a 203-molecule naphthalene aggregate for different polarization directions, tracking exciton transfer along an anthracene chain with both standard and decoherence-corrected Ehrenfest dynamics, and showing that excited-state calculations on tetracene aggregates up to 8850 atoms are feasible in about an hour. The authors assert that the package enables first-principles atomistic simulations of exciton and charge transport in large biomolecular systems and realistic optoelectronic models.","pith_inferences":["If the fragment-pair assumption is robust, DIALECT could compute exciton diffusion lengths from first principles by averaging many decoherence-corrected Ehrenfest trajectories, which the paper leaves as future work.","The same excitonic Hamiltonian could be extended beyond single cavity modes to include multiple photon modes or vibronic coupling, giving a path to polariton chemistry and vibrational spectra that the authors do not pursue here.","The accuracy of the method for systems with strong inter-fragment polarization remains untested; a direct comparison with full-system TD-DFT on small clusters would quantify the approximation's error, a benchmark the paper does not provide."],"forward_implications":["Exciton and charge-transfer dynamics in aggregates of thousands of atoms can be simulated directly with surface hopping or decoherence-corrected Ehrenfest dynamics.","Polariton dispersions of realistic molecular aggregates, including intermolecular excitonic couplings, can be obtained for arbitrary cavity polarization and mode energy.","The near-linear scaling of the ground-state SCC step and roughly one-hour timings for 200 excited states of an 8850-atom system make the method practical for organic semiconductor and biomolecular models.","The photoisomerization branching ratios obtained for cis-stilbene, though not quantitatively on par with high-level methods, provide a fast exploratory route to photodynamics."],"supporting_citations":[{"why":"Supplies the FMO-LC-TDDFTB method and the excitonic Hamiltonian built from LE and CT states.","marker":"58"},{"why":"Provides the Ehrenfest dynamics formulation and energy gradients in the quasi-diabatic basis.","marker":"59"},{"why":"Introduces the coupling of FMO-LC-TDDFTB to microcavity modes via the generalized Tavis-Cummings Hamiltonian.","marker":"60"},{"why":"Supplies the FMO-DFTB method for ground-state fragment calculations.","marker":"49"},{"why":"Provides the long-range corrected FMO-DFTB formalism used for charge-transfer states.","marker":"53"},{"why":"The long-range corrected TD-DFTB formalism that enables charge-transfer excitations.","marker":"26"},{"why":"The SCC-DFTB base method underlying the fragment calculations.","marker":"23"},{"why":"Tully's fewest-switches surface hopping algorithm used for nonadiabatic dynamics.","marker":"65"},{"why":"Granucci decoherence correction used in the surface hopping simulations.","marker":"71"},{"why":"State-pairwise decoherence times and the TAB collapse procedure used in the Ehrenfest dynamics.","marker":"73"}],"fun_headline_variants":["DIALECT simulates exciton dynamics in 8,850-atom aggregates","Excitons plus cavity photons: DIALECT scales to 9,000 atoms","Large-scale exciton and polariton dynamics with DIALECT","DIALECT: from weak to strong light-matter coupling at scale"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire method assumes that the excited states of a large assembly are fully determined by electronic structure calculations on single molecules and pairs of molecules, and that nonadiabatic couplings between different fragments can be neglected; if many-chromophore polarization or inter-fragment couplings are significant, the computed spectra and dynamics will be biased.","fun_headline_variants_meta":{"raw":{"variants":["DIALECT simulates exciton dynamics in 8,850-atom aggregates","Excitons plus cavity photons: DIALECT scales to 9,000 atoms","Large-scale exciton and polariton dynamics with DIALECT","DIALECT: from weak to strong light-matter coupling at scale"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000807,"raw_usage":{"total_tokens":3520,"prompt_tokens":898,"completion_tokens":2622,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":2540}},"tokens_in":514,"tokens_out":2622,"duration_ms":16692,"temperature":1.0,"reasoning_tokens":2540,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:31:56.033425+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Benchmark the method against a full-system TD-DFTB or high-level ab initio calculation on a small aggregate, for example a chain of five to ten anthracene or naphthalene molecules, and compare excited-state energies, oscillator strengths, and exciton-transfer dynamics; significant deviations beyond roughly 0.1 eV or qualitatively different population dynamics would disprove the fragment-pair assumption and the neglect of inter-fragment nonadiabatic couplings.","supporting_citations":[{"cited_title":"P.; Levine, B","cited_arxiv_id":null,"evidence_quote":"State-pairwise decoherence times and the TAB collapse procedure used in the Ehrenfest dynamics."},{"cited_title":"Long-range corrected fragment molecular orbital density functional tight-binding method for excited states in large molecular systems","cited_arxiv_id":null,"evidence_quote":"Supplies the FMO-LC-TDDFTB method and the excitonic Hamiltonian built from LE and CT states."},{"cited_title":"Nonadiabatic Exciton Dynamics and Energy Gradients in the Framework of FMO - LC - TDDFTB","cited_arxiv_id":null,"evidence_quote":"Provides the Ehrenfest dynamics formulation and energy gradients in the quasi-diabatic basis."},{"cited_title":"N.; Mitrić, R","cited_arxiv_id":null,"evidence_quote":"Introduces the coupling of FMO-LC-TDDFTB to microcavity modes via the generalized Tavis-Cummings Hamiltonian."},{"cited_title":"G.; Irle, S","cited_arxiv_id":null,"evidence_quote":"Supplies the FMO-DFTB method for ground-state fragment calculations."},{"cited_title":"Q.; Nishimoto, Y.; Fedorov, D","cited_arxiv_id":null,"evidence_quote":"Provides the long-range corrected FMO-DFTB formalism used for charge-transfer states."},{"cited_title":"Long-range correction for tight-binding TD - DFT","cited_arxiv_id":null,"evidence_quote":"The long-range corrected TD-DFTB formalism that enables charge-transfer excitations."},{"cited_title":"Including quantum decoherence in surface hopping","cited_arxiv_id":null,"evidence_quote":"Granucci decoherence correction used in the surface hopping simulations."}],"review_version":1}