{"id":"cc0c9ea2-7386-428a-912d-aa4ef017c046","arxiv_id":"1908.03772","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The authors derive mean-field and mapping-variable ring polymer instantons for multistate systems and show that mean-field rates match Fermi's Golden Rule and adiabatic instanton benchmarks across coupling regimes.","lead":"Two new versions of a multistate ring polymer instanton method are derived and tested on model electron-transfer reactions. The mapping-variable version reports how electron populations change along the tunneling path, which earlier instanton methods did not show.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The two fixed beads and N1/N2 scan in Sec. IV.A likely yield a constrained stationary point, not a true saddle of the full ring-polymer potential; Eq. 27's rate prefactor is then unjustified.","rationale":"The reader identified the constrained saddle search in Sec. IV.A as the weakest assumption; I agree that this is the single most load-bearing concern. The paper's abstract and conclusion claim that the MF-RPI rate accurately describes both adiabatic and nonadiabatic regimes, and that the solutions are true first-order saddles. Both claims depend on the optimized configuration being a stationary point of the full ring-polymer potential. The manuscript explicitly states that unconstrained initial guesses converge to a minimum or to the acceptor well, which is why two beads were fixed at the diabatic crossing and N1/N2 was scanned. This protocol produces a stationary point of a constrained problem, not necessarily of the full potential. A true ring-polymer instanton requires the gradient to vanish in all N directions; the paper does not report a gradient check, and the symmetry argument for the fixed-bead stationarity applies only to the symmetric model I, not to the driven models II and III. Without this check, the derivation of the rate prefactor in Eq. 27, which assumes harmonic fluctuations about a full-space saddle, is not justified. The secondary concerns raised by the reader, such as missing bead-convergence data and the overstatement in the intermediate-coupling claim, are real but addressable and less central. The intermediate-coupling claim relies on the MF rate being accurate where no exact benchmark exists; however, the observed deviations from FGR at Delta=5.0 (about 1.7 orders of magnitude) are expected because FGR is inapplicable there, so this alone would not invalidate the method. The decisive issue is the saddle point itself. I do not recommend changing the reader's CONDITIONAL verdict: the concern is concrete, falsifiable, and can be resolved by the proposed unconstrained gradient test. If the test shows nonzero gradient and a material rate change, the verdict should move toward REJECT; if the test confirms stationarity and rate insensitivity, the paper's central claim would be substantially strengthened. Until then, CONDITIONAL is the appropriate verdict.","tokens_in":13932,"tokens_out":8025,"duration_ms":89579,"concrete_test":"Take the converged 256-bead MF-RPI (and MV-RPI) configuration from Sec. V and compute the full gradient of the ring-polymer potential V_MF (and V_MV) with respect to all N bead positions, including the two beads fixed at the crossing in Sec. IV.A. If the gradient norm is not essentially zero, release the two fixed-bead constraints and re-optimize using an unconstrained saddle-finding method (e.g., eigenvector-following or a dimer method) initialized from the constrained solution. Then recompute the MF-RPI rate at the resulting stationary point using Eq. 27 and compare with Table III. If the configuration shifts significantly or the rate changes by more than a factor of 2, the constrained search biases the instanton path and the central claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that MF-RPI rates are accurate across three orders of magnitude in coupling rests on the assertion that the optimized ring-polymer configuration is a true first-order saddle of the full multistate ring-polymer potential. The protocol of Sec. IV.A, however, fixes two beads exactly at the diabatic crossing and scans the discrete integer ratio N1/N2, maximizing the effective potential over that ratio. These two fixed beads are never optimized, and N1/N2 is not a degree of freedom of the potential; it is merely a property of the bead labeling. Consequently, the final configuration generally has nonzero gradient components in the directions of the fixed beads, so it is not a stationary point of the full potential V_MF(R_1,...,R_N). The paper claims numerical verification of a single unstable mode, but a Hessian with one negative eigenvalue at a point with nonzero gradient does not establish a first-order saddle. The rate expression in Eq. 27 is derived from a Gaussian fluctuation integral about a true saddle of the full potential, including the zero-mode normalization z_N and the product over the N-1 stable modes. If the configuration is only a constrained stationary point, the fluctuation prefactor is missing contributions from the constrained variables and the semiclassical derivation does not apply. Notably, the crossing point is a stationary point by symmetry only for the symmetric model; for asymmetric models II and III, the force at the crossing is generally nonzero, making the bias more severe. No unconstrained or differently constrained calculation is reported to validate that the constrained solution is a true instanton or that rates are insensitive to the constraint. This is the most load-bearing weakness because it directly undermines the validity of the rate constants that constitute the paper's central quantitative contribution.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents two multistate ring polymer instanton (RPI) formulations, a Mean-Field (MF)-RPI and a Mapping-Variable (MV)-RPI, both derived from exact path integral representations of the canonical partition function. The MF-RPI rate expression, Eq. (27), is used to compute rate constants for model two-state systems with coupling strengths spanning three orders of magnitude, and is compared to Fermi's Golden Rule rates in the nonadiabatic limit and to single-surface RPI rates in the adiabatic limit. The MV-RPI is shown to provide electronic state populations along the instanton path, which the MF-RPI populations obtained from Eq. (34) do not. The authors also claim analytic zero modes and numerical verification of a single unstable mode for both formulations. The paper concludes that the MF-RPI accurately calculates adiabatic and nonadiabatic reaction rates and that the MV-RPI gives mechanistic population information.","tokens_in":14112,"tokens_out":7277,"duration_ms":81436,"significance":"If fully validated, the paper would be a valuable contribution: it offers a practical instanton-based rate theory for nonadiabatic processes across coupling regimes, and the MV-RPI population analysis is a novel mechanistic diagnostic. The limit checks are meaningful: Table IV shows agreement with FGR to three decimal places, and in the adiabatic regime the MF-RPI is within a factor of about 2 of single-surface RPI. The analytic zero-mode derivations in the appendices are useful. However, the central rate-accuracy claim rests on the identification of the computed configurational as a true first-order saddle of the full ring-polymer potential, and the intermediate-coupling accuracy claim currently lacks a benchmark. These issues are load-bearing for the abstract's claim of accurate rates over three orders of magnitude in coupling.","major_comments":[{"comment":"The optimization protocol fixes two ring-polymer beads exactly at the diabatic crossing and scans the discrete integer ratio N1/N2, selecting the ratio that maximizes the effective potential. These fixed beads are not optimized, and N1/N2 is not a degree of freedom of the ring-polymer potential; it is a property of the bead labeling. Consequently, the final configuration generally has nonzero gradient components in the directions of the fixed beads, especially for the asymmetric models II and III, and is not a stationary point of the full potential V_MF(R_1,...,R_N). A Hessian with one negative eigenvalue at a point with nonzero gradient does not establish that the point is a first-order saddle. The rate expression in Eq. (27) is derived from a Gaussian fluctuation integral about a true saddle of the full potential, including the zero-mode normalization z_N and the product over the N-1 stable modes (Eqs. 24-26). If the configuration is only a constrained stationary point, the fluctuation prefactor omits contributions from the constrained variables and the semiclassical derivation does not apply. The paper should demonstrate that the constrained solution is the true unconstrained saddle, for example by reporting the gradient norm after relaxation of the constraints or by locating the saddle with an algorithm that does not fix beads.","section":"Sec. IV.A and Eq. (27)"},{"comment":"The claim that the MF-RPI rate expression 'interpolates smoothly and accurately' between the nonadiabatic and adiabatic regimes is not supported by the data. In Table III at Δ=5.0 a.u., the MF-RPI rate is about 50 times the FGR rate (log10 k = -30.269 vs -31.997), and while the adiabatic single-surface RPI is closer, the intermediate coupling regime has no exact benchmark. Agreement with FGR in the weak-coupling limit and with single-surface RPI in the strong-coupling limit brackets the two ends but does not establish accuracy in between. The abstract's claim of 'accurately calculate rate constants ... with the coupling strength varying over three orders of magnitude' is therefore an overstatement. The authors should either provide a benchmark in the intermediate regime (e.g., a numerically exact quantum rate or a converged golden-rule instanton result where valid) or soften the accuracy claim to 'smoothly interpolates between the two limiting rate expressions'.","section":"Table III and Sec. V"}],"minor_comments":[{"comment":"The phrase 'we numerically and that these solutions are true instantons' is missing a verb; it should read 'we numerically find that these solutions are true instantons'.","section":"Abstract"},{"comment":"In the definition of M_nn, the argument 'V_nn(R_α)+V_nn(R_α)' appears twice; based on the structure of M_nm it likely should be V_nn(R_α)+V_nn(R_{α+1}). Please correct the typo.","section":"Eq. (9)"},{"comment":"The caption reports 'Δ = 0.077 a.u.' for the nonadiabatic case, but the text and Table I give Δ = 0.0077 a.u. at β = 3.25 a.u. The factor-of-10 discrepancy is misleading.","section":"Fig. 5 caption"},{"comment":"The zero-mode proof for the MV instanton is formal because the electronic variables enter the action only through a total derivative, so the equation of motion for x is first-order and the second variation in x has no differential operator. The claim that the collective velocity mode in both R and x is a zero mode of the stability matrix in Eq. (41) should be justified more carefully, or the scope of the analytic zero-mode statement should be limited to the MF case.","section":"Appendix B"}],"recommendation":"major_revision","confidential_remarks":"The constrained-saddle issue in Sec. IV.A is the central technical concern. I would ask the authors to provide direct numerical evidence that the reported configurations are stationary points of the full ring-polymer potential, or to re-derive the rate expression for the constrained optimization. The intermediate-coupling accuracy claim also needs a benchmark or a more modest wording. These are fixable with additional numerical work and do not require rejecting the underlying idea, but the manuscript in its current form overstates the reliability of the rates."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a real methods paper, not a dressed-up null result. The new thing is the mapping-variable RPI: explicit continuous electronic variables that give bead-resolved populations along the instanton path, which golden-rule instantons don't provide. The mean-field rate expression is also new in this form, and the benchmarks in the nonadiabatic and adiabatic limits look right: MF-RPI tracks FGR to four decimal places at small Delta, and single-surface RPI within a factor of about two at large Delta. The zero-mode analytic argument is standard but correct, and the MV populations genuinely distinguish adiabatic from nonadiabatic behavior in a way the ad hoc MF populations don't.\n\nSoft spots. The big one is the constrained saddle search in Sec. IV.A. Two beads are fixed at the diabatic crossing and N1/N2 is scanned as a discrete integer. That is not a stationary point of the full ring-polymer potential unless the force at the crossing happens to vanish; for asymmetric models II and III it generally won't. A Hessian with one negative eigenvalue at a point with nonzero gradient is not a first-order saddle, so Eq. 27's fluctuation prefactor is not rigorously justified for the constrained configuration. The paper claims 'true instantons' but never reports the gradient norm or an unconstrained search. This matters most for the intermediate-coupling rates, where the 'interpolates smoothly and accurately' claim is also overstated: at Delta=5, MF is about 1.7 orders above FGR, and the single-surface RPI is not exact there either. So the central quantitative claim rests on a saddle that hasn't been shown to exist. That's addressable—report the gradient, try a nudged-elastic-band or a different constraint, test sensitivity—but it is load-bearing.\n\nMinor: no error bars or bead-convergence data on the rates; the discrete N1/N2 scan may miss the true optimum. These are minor.\n\nWho's it for: anyone working on nonadiabatic rates or instanton methods; it's a useful stepping stone. I'd send it to a serious referee—the derivation and the MV populations are worth the referee's time even if the constraint issue needs work. My own verdict would be conditional, not reject.","headline":"A genuinely new mapping-variable instanton with useful bead-resolved populations, but the constrained saddle search undercuts the intermediate-coupling rate claims.","tokens_in":14785,"tokens_out":3820,"would_cite":true,"duration_ms":36581,"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":"Two path-integral instanton formulations give multistate reaction rates across three orders of magnitude in electronic coupling.","keywords":["nonadiabatic instanton","ring polymer instanton","mapping variables","mean-field","two-state system","reaction rates","path integral","golden-rule rate"],"falsifier":"Take the symmetric two-state model at inverse temperature $\\beta=3.25$ a.u. with the weak coupling $\\Delta=0.0077$ a.u. and search for a first-order saddle of the full multistate ring polymer potential without fixing any beads at the crossing. An unconstrained saddle with a different action, bead distribution, or stability spectrum than the constrained $N_1/N_2$-scanned solution would falsify the paper's central claim.","tokens_in":13614,"feed_emoji":"⚛️","tokens_out":12356,"duration_ms":117997,"temperature":0.7,"pith_summary":"The paper sets out to show that two path-integral instanton constructions—a mean-field version that averages over electronic states and a mapping-variable version with explicit Cartesian electronic variables—are both legitimate multistate ring polymer instantons. It argues that each is a true first-order saddle with a single unstable mode and a zero mode, and that the two nuclear instanton paths essentially coincide. The distinctive payoff is that the mapping-variable version resolves how electronic state populations change bead by bead along the instanton path, which the mean-field version cannot. If the claims are right, the mean-field instanton gives a single rate formula that matches golden-rule rates at weak coupling and single-surface ring polymer instanton rates at strong coupling, covering three orders of magnitude in coupling strength.","feed_headline":"Multistate instantons give rates across three coupling regimes","feed_subtitle":"A mean-field instanton matches known rate limits; a mapping-variable twin tracks electronic state changes along the path.","key_machinery":"The load-bearing object is the multistate ring polymer potential, a discretized imaginary-time action whose saddle points are the instantons. In the mean-field representation the effective potential is $V_{\\mathrm{MF}}(\\{R_\\alpha\\}) = U(\\{R_\\alpha\\}) - \\frac{1}{\\beta}\\ln|\\mathrm{Re}\\,\\Gamma_{\\mathrm{MF}}|$, where $U$ is the harmonic bead coupling and $\\Gamma_{\\mathrm{MF}}$ is the transfer matrix obtained by tracing over electronic states. In the mapping-variable representation it is $V_{\\mathrm{MV}}(\\{R_\\alpha\\},\\{x_\\alpha\\}) = U(\\{R_\\alpha\\}) + \\frac{1}{\\beta}\\sum_\\alpha x_\\alpha^T x_\\alpha - \\frac{1}{\\beta}\\ln|\\mathrm{Re}\\,\\Gamma_{\\mathrm{MV}}|$, with $x_\\alpha$ continuous Cartesian variables for the electronic states. The argument works by solving the stationarity conditions on all coordinates and then classifying the solution through the stability matrix: exactly one negative eigenvalue certifies a first-order saddle, and the zero eigenvalue is identified analytically with the collective velocity mode $\\dot{X}(\\tau)$ in all degrees of freedom. The rate follows from Gaussian integration around the saddle after analytically continuing the unstable mode.","core_discovery":"The central claim is that multistate ring polymer instantons can be computed from the exact path-integral partition function in two ways, and both are genuine instantons: analytically they possess the zero mode associated with imaginary-time translation, and numerically their stability matrices show exactly one unstable mode. The mean-field and mapping-variable nuclear paths nearly coincide for the two-state single-mode models studied, so the nuclear tunneling mechanism is robust to the electronic representation. The mapping-variable version goes beyond the mean-field one by assigning every bead a fractional donor/acceptor population; in the nonadiabatic limit these populations change smoothly along the path, while in the adiabatic limit they change sharply and with oscillatory structure near the crossing. The paper also supplies an MF-RPI rate expression and shows it reproduces golden-rule results at weak coupling and single-surface ring polymer instanton results at strong coupling, interpolating accurately across the intermediate regime.","pith_inferences":["The two-bead pinning protocol is a natural stress point: repeating the calculation with an unconstrained saddle search on an asymmetric model would show whether the scanned ratio $N_1/N_2$ is a genuine saddle coordinate or an artifact of the optimizer's basin.","The MV-RPI population profiles suggest a measurable nonadiabatic reaction coordinate—for instance, the bead index at which donor population crosses 1/2—that could be plugged into a transition state theory calculation, something the paper does not do.","All reported models lie in the normal electron-transfer regime; applying the method in the inverted regime or near the crossover temperature would test whether the single-unstable-mode instanton picture holds outside the demonstrated conditions.","Because the MV-RPI carries explicit electronic coordinates, it is a plausible seed for mapping-variable ring polymer molecular dynamics trajectories, an extension the authors flag but do not execute."],"forward_implications":["The MF-RPI rate formula can serve as a single expression for reaction rates in both the nonadiabatic and adiabatic limits, matching golden-rule results at weak coupling and single-surface ring polymer instanton results at strong coupling.","The MV-RPI provides per-bead electronic state populations along the instanton path, giving a mechanism-level picture of where the donor-to-acceptor transition occurs without assuming the nuclear geometry of the transition.","Both formulations are first-order saddles with one unstable mode and a zero mode, so they fit into the standard ring polymer instanton framework, including its rate expression.","Because the mean-field and mapping-variable path integrals are equivalent representations of the same partition function, their nuclear instantons coincide; no separate MV rate formula is needed."],"supporting_citations":[{"why":"Supplies the ring-polymer instanton saddle-point formalism and the single-surface rate expression that the mean-field rate generalizes.","marker":"15"},{"why":"Provides the mean-field path-integral representation and the earlier nonadiabatic instanton rate expression whose form the mean-field rate follows.","marker":"18"},{"why":"Gives the golden-rule instanton used for comparison, for the two-bead crossing-pinning initial guess, and for the contrast with beads that sit wholly on one surface.","marker":"26"},{"why":"Introduces the mapping of electronic states to singly excited oscillators that underlies the mapping-variable representation.","marker":"27"},{"why":"Supplies the mapping-variable path-integral partition function from which the mapping-variable effective potential is derived.","marker":"29"},{"why":"Provides the box-constrained quasi-Newton optimization routine used to locate the reported instanton saddles.","marker":"35"}],"fun_headline_variants":["Two multistate instanton paths, one accurate rate formula","Mapping-variable instantons expose electronic transitions along the path","Multistate instantons: exact partition function, two representations","Nonadiabatic rates from mean-field and mapping-variable instantons","Multistate instantons bridge weak to strong coupling seamlessly"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that pinning two ring-polymer beads exactly at the crossing point of the two electronic-state curves and scanning how many remaining beads sit on each side locates the true first-order saddle of the unconstrained multistate ring polymer potential; if that constraint biases the saddle, the instanton paths and rates inherit the bias.","fun_headline_variants_meta":{"raw":{"variants":["Two multistate instanton paths, one accurate rate formula","Mapping-variable instantons expose electronic transitions along the path","Multistate instantons: exact partition function, two representations","Nonadiabatic rates from mean-field and mapping-variable instantons","Multistate instantons bridge weak to strong coupling seamlessly"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00033,"raw_usage":{"total_tokens":1848,"prompt_tokens":966,"completion_tokens":882,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":582,"completion_tokens_details":{"reasoning_tokens":796}},"tokens_in":582,"tokens_out":882,"duration_ms":10107,"temperature":1.0,"reasoning_tokens":796,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:03:17.885678+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the symmetric two-state model at inverse temperature $\\beta=3.25$ a.u. with the weak coupling $\\Delta=0.0077$ a.u. and search for a first-order saddle of the full multistate ring polymer potential without fixing any beads at the crossing. An unconstrained saddle with a different action, bead distribution, or stability spectrum than the constrained $N_1/N_2$-scanned solution would falsify the paper's central claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the mean-field path-integral representation and the earlier nonadiabatic instanton rate expression whose form the mean-field rate follows."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the golden-rule instanton used for comparison, for the two-bead crossing-pinning initial guess, and for the contrast with beads that sit wholly on one surface."},{"cited_title":"Ananth \\ and\\ author T","cited_arxiv_id":null,"evidence_quote":"Supplies the mapping-variable path-integral partition function from which the mapping-variable effective potential is derived."},{"cited_title":"Zhu , author R","cited_arxiv_id":null,"evidence_quote":"Provides the box-constrained quasi-Newton optimization routine used to locate the reported instanton saddles."}],"review_version":1}