{"id":"431eab0e-1f0d-46d9-9017-fd131261957d","arxiv_id":"2607.06122","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":5,"one_line_summary":"Analytic expressions for hydrogen diffusion and retention in multi-occupancy traps (monovacancies and voids) in tungsten and vanadium are validated against molecular dynamics simulations with no fitted parameters, showing superiority over single-occupancy trap models.","lead":"This paper validates a parameter-free analytic model for how hydrogen diffuses and gets trapped in radiation-damaged metals, by comparing it directly to molecular dynamics simulations. A smart generalist might read it because accurate prediction of tritium retention in fusion reactor walls is a safety and economic bottleneck for fusion energy.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"Void diffusivity validation is sensitive to mobile-gas-fraction noise and untested surface-model assumptions, but monovacancy validation is clean and the central claim holds.","rationale":"The reader identified the void surface model assumptions and the mobile-gas-fraction sensitivity as the weakest point, which I agree is the correct target. The paper's central claim is primarily validated by the monovacancy results (Figures 5, 6), where full binding-energy data is available and the agreement with MD is tight. The void model (Figures 12-14) is a reasonable extension with honestly acknowledged limitations. The 'no parameters fitted to diffusion data' claim is technically correct—all parameters come from static calculations—but the quartic binding form and equal-frequency assumption are ansätze whose impact on the void diffusivity prediction is amplified by the extreme sensitivity to mobile gas fraction. This is a real soft spot in the void-model validation, but it does not undermine the monovacancy validation or the superiority claim over single-occupancy models. The paper appropriately positions the void results as approximate and transferable rather than precision-fitted. An ACCEPT with MODERATE confidence is the right call; the moderate rather than high confidence is justified by the void-model issues, and the reader's rationale captures this correctly. The one thing I would add to the reader's assessment is that the finite-size statistics in the void MD (1-2 mobile atoms) are a distinct concern from the model assumptions themselves—both contribute to the 800K discrepancy, and disentangling them would require the larger-cell test I propose.","tokens_in":25540,"tokens_out":2710,"duration_ms":158180,"concrete_test":"Run void MD simulations at 800K with a simulation cell 4× larger in each dimension (e.g., 96³ instead of 48³ conventional cells) to increase the mobile gas atom count from ~1-2 to ~10-20, reducing the finite-size noise in the measured Deff by a factor of ~3-4. Simultaneously, compute ω_s for H on void surfaces (at least for the dominant surface site type) using DFT phonon calculations or frozen-phonon estimates with the EAM potential, and recompute the analytic diffusivity with the corrected ω_s. If the corrected analytic curve shifts by more than the MD error bars at 800K for the 60-vacancy void, the ω_s = ω_L assumption is load-bearing for the void model and should be replaced; if it shifts negligibly, the discrepancy is dominated by finite-size statistics and the current model is adequate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly identifies the void retention model as the soft spot. The central claim has two pillars: (1) the monovacancy model agrees with MD, and (2) multi-occupancy is superior to single-occupancy. Pillar (1) is well-supported for monovacancies—Figures 5 and 6 show tight agreement using full binding-energy data with the configurational-entropy prefactor. Pillar (2) is also well-established across both materials. The concern lands specifically on the void extension (Section IVB, eqs 16-17). The diffusivity in the void case is computed via eq 6/7, which depends on var(θ)/x. The paper itself notes 'there may be only one or two mobile gas atoms in the MD simulation box,' meaning the MD-measured Deff has large finite-size noise, and the analytic Deff is exponentially sensitive to the predicted mobile fraction. A small overestimate of surface retention (which the paper acknowledges at 800K, Figure 14) propagates into a large underestimate of diffusivity. Two specific untested assumptions drive this: (a) the quartic surface binding form (eq 37) with a single parameter α fitted to the monovacancy limit, which may not hold for larger voids where surface site heterogeneity matters (Hou et al [39] show site-dependent binding); and (b) ω_s = ω_L (Table II), which sets the vibrational entropy of surface atoms equal to lattice atoms—this directly affects the surface-to-lattice chemical potential balance and hence the mobile fraction. At 800K, kBT ≈ 0.069 eV, so a modest error in the effective surface free energy (say 0.05-0.1 eV from the ω_s assumption alone) could shift the mobile fraction by a factor of 2-4, which is amplified in Deff. The paper is transparent about these limitations, which is why this weakens confidence in the void-model validation specifically but does not overturn the central claim, which rests primarily on the monovacancy results.","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"This manuscript presents a multiscale modelling study of hydrogen diffusion and retention in irradiated bcc metals (tungsten and vanadium), focusing on multi-occupancy traps (monovacancies and nanovoids). The authors validate an analytic framework for effective diffusivity (derived in prior work, ref [14]) against molecular dynamics (MD) simulations and lattice kinetic Monte Carlo (kMC). The central claims are: (1) the analytic multi-occupancy trap model, parameterized from static calculations with no fitted diffusion parameters, agrees with MD for hydrogen in W and V containing monovacancies; (2) the model is quantitatively superior to the single-occupancy McNabb-Foster formalism across concentration ranges; and (3) the framework extends to voids, with reasonable but imperfect MD agreement. The paper also includes kMC simulations of void nucleation and growth to establish plausible void sizes, and concludes with first-principles predictions for retention and diffusivity in post-irradiation-annealed tungsten.","tokens_in":25806,"tokens_out":1533,"duration_ms":249156,"significance":"The manuscript makes a valuable contribution to hydrogen transport modelling in fusion-relevant materials. The parameter-free derivation of effective diffusivity for multi-occupancy traps, validated against both kMC (steady-state convergence, Appendix VI, Fig. 16) and full MD (monovacancies in W and V, Figs. 5–6), is a significant strength. The demonstration that multi-occupancy treatment is necessary and superior to single-occupancy McNabb-Foster models across materials and concentration ranges is well-supported and practically important for fusion reactor modelling. The void growth kMC study (Section IVA) and the forward prediction of retention/diffusivity in annealed microstructures (Fig. 15) add practical value. The authors are commendably transparent about the limitations of the void model and the sensitivity of diffusivity to mobile-gas fraction.","major_comments":[{"comment":" ","section":null},{"comment":"Section IVB, Eqs. (16–17) and Table II: The void retention model assumes equal vibrational frequencies for surface and lattice hydrogen atoms (ω_s = ω_L = 0.254 PHz, Table II). This assumption directly affects the surface-to-lattice chemical potential balance and hence the predicted mobile fraction, which the paper itself notes is extremely sensitive (Section IVC: 'there may be only one or two mobile gas atoms in the MD simulation box'). At 800 K, the model overestimates retention and underestimates diffusivity (Fig. 14, low-occupancy voids). The authors acknowledge this discrepancy but do not quantify how much of it is attributable to the ω_s = ω_L assumption versus the quartic surface binding form (Eq. 37). A brief sensitivity analysis or discussion of the expected sign and magnitude of the error from this assumption would strengthen the void model validation. This is load-bearing for,","section":null},{"comment":"Section IVB, Eq. (37): The quartic surface binding energy model uses a single parameter α fitted to the monovacancy limit and treated as constant across void sizes. Hou et al. [39] (cited by the authors) show site-dependent binding on void surfaces. The authors note that 'gas atoms on one surface site may have a different binding energy to those on a different site' (Section IVB) but then proceed with the single-α model without testing whether surface site heterogeneity matters for the larger voids (n_v = 15, 60) used in the MD validation. Since the void diffusivity prediction depends exponentially on the mobile fraction, which depends on surface retention, this is a correctness-risk for the void extension. The authors should discuss whether the reasonable MD agreement at 1200–1600 K (Fig. 14) provides sufficient evidence that the single-α approximation is adequate, or whether the 800 K","section":null},{"comment":"Section III, discussion of g'_i: The authors state 'it was difficult to recommend one model which would be best for both tungsten and vanadium' and recommend the configurational-entropy prefactor (Eq. 14) largely for computational convenience. Figure 4 shows that different choices of g'_i shift the predicted diffusivity substantially. Since the choice of g'_i is load-bearing for the monovacancy validation (Figs. 5–6) and for the central claim of parameter-free prediction, the authors should clarify whether the recommended g'_i (therm+conf) is the one used in all subsequent figures (Figs. 5, 6, 10, 12, 14), and whether the MD agreement would degrade significantly with the thermodynamic-only prefactor. The current text is ambiguous about which prefactor is used where.","section":null}],"minor_comments":[{"comment":"Section I, paragraph 3: 'microstructre' should be 'microstructure'.","section":null},{"comment":"Table I caption: the footnote markers (a, b, c) are placed after the values but the caption text lists them as 'a Ref [47], b Ref [45], c Ref [48]' — the mapping is clear but the superscript placement in the table body could be more explicit.","section":null},{"comment":"Figure 2 caption: 'PALIOXIS' is written in uppercase in the caption but as 'Palioxis' in the text (Section IIA). Consistent capitalization would help.","section":null},{"comment":"Section IIA: the statement 'Changing to g'_i = g has no effect on the conclusion' could briefly note what conclusion is referred to (presumably the kMC–analytic agreement).","section":null},{"comment":"Reference [58] is listed as 'in preparation, 2025' — this should be updated to a published reference or preprint if available at revision stage.","section":null},{"comment":"Figure 14: the shaded regions indicating expected instantaneous diffusivity range are described in the text but the figure caption could state more explicitly that these correspond to one standard deviation of the mobile gas fraction.","section":null},{"comment":"Section IVD, Fig. 15a: the experimental comparison to ref [74] is for a 'similar but not identical scenario' — a brief sentence clarifying the key differences (irradiation temperature vs. post-irradiation annealing) would help the reader assess the comparison.","section":null},{"comment":"The data availability statement says code and data 'will be made available on acceptance' — standard practice would be to provide a repository link at revision stage for reproducibility.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The reader's report and stress-test note correctly identify the void model (Section IVB) as the soft spot. The monovacancy validation (Figs. 5–6) is clean and the central claim of multi-occupancy superiority is well-established. The void model issues are real but localised: they affect the void extension, not the core framework. The authors are transparent about the 800 K discrepancy and the mobile-gas-fraction sensitivity. I judge these as fixable with additional discussion and possibly a brief sensitivity check, hence minor revision rather than major. The overlapping authorship with refs [14], [46], and [72] is disclosed in the text and does not appear to create circularity — the analytic framework is independently verified by kMC, and the MD simulations use the same potential but represent ground-truth dynamics for that potential. The paper is within scope for a materials modelling journal."},"author_rebuttal":{"model":"glm-5.2","summary":"We thank the referee for their careful reading and constructive comments. The referee correctly identifies the key strengths of the manuscript and raises three substantive points concerning (1) the assumption of equal vibrational frequencies for surface and lattice hydrogen in the void model, (2) the single-parameter quartic surface binding model and its neglect of surface site heterogeneity, and (3) ambiguity about which detrapping prefactor g'_i is used in subsequent figures. We address each point below. In brief: we will add a sensitivity analysis for the ω_s = ω_L assumption, add discussion of surface site heterogeneity and its likely impact, and clarify explicitly which prefactor is used in each figure. We agree with all three comments and will revise the manuscript accordingly.","responses":[{"response":"The referee is correct that the ω_s = ω_L assumption is load-bearing for the surface-to-lattice chemical potential balance and that we did not quantify its impact. We will add a brief sensitivity analysis in the revised manuscript. To anticipate the result: the vibrational frequency enters the surface free energy (Eq. 39) through the term n_s * F_vib_s, where F_vib_s ≈ 3 k_B T log(ℏω_s / k_B T) in the classical limit. The difference between surface and lattice vibrational free energies is ΔF_vib = 3 k_B T log(ω_s / ω_L). If ω_s < ω_L (as is physically expected, since surface H atoms are more weakly bound and thus have softer vibrational modes), then ΔF_vib < 0, meaning the surface free energy is lowered, retention is increased, and the mobile fraction is further reduced. This would worsen the discrepancy at 800 K, not improve it. Conversely, if ω_s > ω_L, the error would partially offset the overestimation. We will include this analysis explicitly, showing the effect of varying ω_s over a plausible range (e.g., 0.15–0.35 PHz) on the predicted retention and diffusivity at 800 K and 1200 K. This will allow the reader to assess whether the ω_s = ω_L assumption is a major contributor to the 800 K discrepancy or whether the quartic binding form (Eq. 37) and the neglect of surface site heterogeneity are the dominant sources of error. Based on our preliminary assessment, the quartic binding form and site heterogeneity are likely the larger contributors, but we will present the sensitivity analysis to let the reader judge.","revision_made":"yes","referee_comment":"Section IVB, Eqs. (16-17) and Table II: The void retention model assumes equal vibrational frequencies for surface and lattice hydrogen atoms (ω_s = ω_L = 0.254 PHz). This assumption directly affects the surface-to-lattice chemical potential balance and hence the predicted mobile fraction, which the paper itself notes is extremely sensitive. At 800 K, the model overestimates retention and underestimates diffusivity. The authors acknowledge this discrepancy but do not quantify how much of it is attributable to the ω_s = ω_L assumption versus the quartic surface binding form. A brief sensitivity analysis or discussion of the expected sign and magnitude of the error would strengthen the void model validation."},{"response":"The referee raises a valid concern. We will add a discussion of the expected impact of surface site heterogeneity and why the single-α model may be adequate at high temperature but not at low temperature. The key physical argument is as follows: at high temperatures (1200–1600 K), the thermal energy k_B T is comparable to or larger than the spread in site-dependent binding energies reported by Hou et al. [39] (which are on the order of 0.1–0.3 eV). In this regime, hydrogen atoms sample multiple surface sites rapidly, and the effective binding is well-approximated by an average value—hence the single-α model works well and the MD agreement at 1200–1600 K is genuine evidence of adequacy in that range. At 800 K, however, k_B T ≈ 0.069 eV is smaller than the site-to-site variation, so hydrogen atoms preferentially occupy the most strongly bound sites. The single-α model, which uses an average binding energy, underestimates the retention on the strongest sites and overestimates it on the weakest, but because the mobile fraction depends exponentially on the strongest binding sites (the last atoms to desorb), the net effect is that the model underestimates retention at low temperature—wait, this is the opposite of what we observe. In fact, our model overestimates retention at 800 K. This suggests that the discrepancy at 800 K is more likely due to the ω_s = ω_L assumption or to the quartic form's behavior at low occupancy rather than to the neglect of site heterogeneity per se. We will discuss this reasoning explicitly in the revised manuscript, acknowledging that we cannot fully disentangle the contributions without explicit site-resolved calculations, and noting that the 800 K discrepancy is a known limitation of the void model that we have been transparent about. We agree a","revision_made":"yes","referee_comment":"Section IVB, Eq. (37): The quartic surface binding energy model uses a single parameter α fitted to the monovacancy limit and treated as constant across void sizes. Hou et al. show site-dependent binding on void surfaces. The authors note that gas atoms on different surface sites may have different binding energies but proceed with the single-α model without testing whether surface site heterogeneity matters for the larger voids (n_v = 15, 60). Since the void diffusivity prediction depends exponentially on the mobile fraction, which depends on surface retention, this is a correctness-risk for the void extension. The authors should discuss whether the reasonable MD agreement at 1200–1600 K provides sufficient evidence that the single-α approximation is adequate, or whether the 800 K discrepancy suggests it is not."},{"response":"The referee is correct that the text is ambiguous about which prefactor is used in which figure. We will clarify this explicitly. The therm+conf prefactor (Eq. 14, g'_i^{therm+conf}) is used in all subsequent figures: Figs. 5, 6, 10, 12, and 14. We will state this clearly in the revised manuscript at the point where the recommendation is made (end of Section III, before the MD results). Regarding whether the MD agreement would degrade with the thermodynamic-only prefactor (g'_i = γg/i): Figure 4 shows that for tungsten at 1200 K, the thermodynamic-only prefactor gives a diffusivity that is noticeably lower than the therm+conf curve, and further from the MD data points. For vanadium at 500 K, the difference between the two prefactors is smaller but still present. We will add a sentence noting this, so the reader understands that the choice of prefactor does affect the quality of agreement, and that the therm+conf choice is not arbitrary but gives demonstrably better agreement with MD in both materials. We agree that this clarification is important for the reader to assess the robustness of the parameter-free claim.","revision_made":"yes","referee_comment":"Section III, discussion of g'_i: The authors state 'it was difficult to recommend one model which would be best for both tungsten and vanadium' and recommend the configurational-entropy prefactor (Eq. 14) largely for computational convenience. Figure 4 shows that different choices of g'_i shift the predicted diffusivity substantially. Since the choice of g'_i is load-bearing for the monovacancy validation (Figs. 5–6) and for the central claim of parameter-free prediction, the authors should clarify whether the recommended g'_i (therm+conf) is the one used in all subsequent figures (Figs. 5, 6, 10, 12, 14), and whether the MD agreement would degrade significantly with the thermodynamic-only prefactor. The current text is ambiguous about which prefactor is used where."}],"tokens_in":25553,"tokens_out":1734,"duration_ms":278035,"standing_objections":[]},"desk_editor":{"model":"glm-5.2","letter":"This paper validates a previously derived analytic model for hydrogen diffusion and retention in multi-occupancy traps (Kaur et al 2025, ref [14], overlapping authors) against full molecular dynamics simulations in tungsten and vanadium. The central result is that the parameter-free analytic model agrees with MD for monovacancies in both materials and is quantitatively superior to single-occupancy McNabb-Foster across the concentration ranges tested. That claim holds up well on reading the paper. The monovacancy validation (Figures 5–6) is the strongest part: tight agreement using the configurational-entropy detrapping prefactor, no parameters fitted to diffusion data, and the MD and analytic curves track each other across varying hydrogen and vacancy concentrations in both materials. The kMC verification of steady-state convergence (Appendix VI, Figure 16) is also clean — the Oriani approximation is explicitly tested and holds. The kMC void growth study (Section IVA) is a useful side result, showing spontaneous void nucleation in tungsten without impurities, consistent with experimental void sizes after annealing. The forward prediction in Figure 15 for post-irradiation annealed microstructures is a genuine prediction using independently determined parameters, and the comparison to experimental retention data is reasonable given the simplifications. The void retention model (eqs 16–17) is where the soft spots are, and the reader and stress-test correctly identify this. The diffusivity in the void case is exponentially sensitive to the mobile gas fraction, and the paper itself notes there may be only one or two mobile gas atoms in the MD box. The quartic surface binding form (eq 37) with a single parameter α fitted to the monovacancy limit, and the assumption ω_s = ω_L, are untested for larger voids where surface site heterogeneity matters. The discrepancy at 800K for low-occupancy voids (Figure 14) is visible and acknowledged. But these limitations are honestly stated and do not undermine the central claim, which rests primarily on the monovacancy results. The unresolved detrapping prefactor g'_i — no single form works best for both W and V — is a real open question but not a flaw in this paper per se; they recommend the configurational entropy form on practical grounds. The shared interatomic potential between MD and analytic model means the validation tests the analytic approximations, not the potential accuracy. The paper says this explicitly. One minor point: the circularity burden from overlapping authorship on the analytic framework, the potential, and the void model parameters is real but manageable — the MD serves as an independent check on the analytic approximations, not on the potential. This paper is for researchers doing hydrogen transport modeling in fusion structural materials, particularly those building finite-element codes who need parameter-free trap models. It deserves a serious referee. The monovacancy validation and the multi-occupancy vs single-occupancy comparison are the load-bearing results and they are solid. The void extension is a reasonable first step with clearly stated limitations. I would accept for peer review.","headline":"Solid validation of multi-occupancy trap model against MD; monovacancy results are clean, void model has honest but proportionate limitations.","tokens_in":26449,"tokens_out":1176,"would_cite":true,"duration_ms":55013,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"Multi-occupancy trap model beats single-occupancy for hydrogen in irradiated metals","keywords":["hydrogen diffusion","multi-occupancy traps","irradiated tungsten","vanadium","nanovoids","McNabb-Foster","molecular dynamics","effective diffusivity"],"falsifier":"Run molecular dynamics or experiment at a temperature and concentration where the model predicts a specific mobile fraction (e.g., 800 K, low-occupancy voids in tungsten, where the paper already notes discrepancy) and show that the multi-occupancy model's diffusivity prediction deviates from measurement by more than the single-occupancy model fitted at that point — which would undermine the claim of quantitative superiority across concentration ranges.","tokens_in":25775,"feed_emoji":"","tokens_out":1486,"duration_ms":318431,"temperature":0.7,"pith_summary":"This paper sets out to prove that an analytic model for hydrogen diffusion and retention in irradiated metals — one that treats defect traps as sites capable of holding multiple hydrogen atoms simultaneously — can reproduce full molecular dynamics simulations without any parameters fitted to diffusion data, and is quantitatively superior to the widely used single-occupancy McNabb-Foster model across realistic concentration ranges. The central object is the multi-occupancy trap: a monovacancy, vacancy cluster, or nanoscale void that can bind several hydrogen atoms at once, with each successive atom experiencing a different binding energy. The authors derive effective diffusivity from the steady-state occupation statistics of these traps, parameterized entirely from static atomistic or density-functional-theory calculations (binding energies, migration barriers, attempt frequencies). They then validate this analytic theory against three independent simulation methods: lattice kinetic Monte Carlo confirms the steady-state equations are correctly implemented; molecular dynamics simulations of hydrogen in tungsten and vanadium containing monovacancies confirm the physical approximations (negligible H-H interaction in the lattice, rapid equilibration between trapped and mobile populations) hold; and molecular dynamics of hydrogen in tungsten nanovoids confirms a simplified free-energy model for surface-bound and molecular hydrogen inside voids. The paper also demonstrates that voids form spontaneously in tungsten without impurity stabilisation, and makes a first-principles prediction for how hydrogen retention and diffusivity change as irradiation-induced monovacancies coalesce into voids during post-irradiation annealing. The key finding is that single-occupancy trap models cannot simultaneously match both the retention and the concentration-dependent diffusivity that molecular dynamics produces, while the multi-occupancy model does so across two materials and multiple trap geometries with no fitted parameters.","feed_headline":"Multi-occupancy trap model beats single-occupancy for hydrogen in irradiated metals","feed_subtitle":"Analytic equations parameterised from first principles reproduce molecular dynamics for hydrogen in tungsten and vanadium with no fitted","key_machinery":"The central mechanism is the steady-state probability distribution y_eq_i (equation 10) for finding i hydrogen atoms in a single trap, computed from the product of successive trapping-to-detrapping rate ratios. From this distribution, the mean occupancy ⟨θ⟩ gives the fraction of gas immobilised and hence the effective diffusivity via the Oriani approximation (equation 3), while the variance var(θ) gives the effective diffusivity for finite-element transport modelling (equation 6). For voids, the model extends to two coupled populations — surface-bound atoms and diatomic molecules in the interior — whose steady-state balance is governed by matching chemical potentials across lattice, surface,","core_discovery":"The multi-occupancy trap model — where a single defect site binds multiple hydrogen atoms with occupancy-dependent energies — reproduces molecular dynamics diffusivity and retention data for hydrogen in tungsten and vanadium with no fitted parameters, while the standard single-occupancy McNabb-Foster model fails at concentrations comparable to the trap density. The steady-state occupation distribution of a multi-occupancy trap (equation 10) depends on the ratio of trapping to detrapping rates at each occupancy level, and from this distribution both the mean retention (which sets effective diffusivity measured in simulation) and the variance (which sets the diffusivity needed in finiteelement","pith_inferences":["The unresolved choice of detrapping geometric/entropic prefactor g'_i — where no single form works best for both tungsten and vanadium — suggests the model's transferability across materials depends on a physical effect (likely anharmonic phonon contributions or site-specific elastic relaxation) that the current approximations do not fully capture. A systematic study across additional bcc metals c","The prediction that diffusivity jumps sharply at the monovacancy-to-void transition temperature (~750 K for tungsten) is testable by ion-irradiation experiments with controlled post-annealing followed by deuterium depth profiling: a step-change in penetration depth at this temperature would confirm the mechanism.","The quartic surface binding energy model (equation 37) with a single parameter α fitted to monovacancy data may break down for very large voids where facet-dependent surface site energies become significant; extending the model to anisotropic surface binding could be necessary for voids larger than a few nanometres.","The convergence scaling of ~4/ρ_v hops per hydrogen atom to reach steady state implies that at the very low mobile fractions expected in reactor conditions (x ~ 10^-8), the equilibration time could become long relative to transient loading events, potentially violating the Oriani steady-state assumption that underpins the analytic diffusivity."],"forward_implications":["Fusion reactor designers modelling tritium retention in irradiated tungsten divertor components can use these analytic equations, parameterised from static calculations, instead of fitting trap parameters to a limited set of thermal desorption experiments, improving predictive reliability under plasma-loading conditions where trap saturation is expected.","The finding that void coalescence during annealing above 750 K sharply reduces hydrogen retention and increases diffusivity provides a concrete design guideline: post-irradiation annealing protocols could be optimised to reduce tritium inventory in structural materials.","The demonstration that voids nucleate and grow in tungsten without carbon or other impurity stabilisation simplifies the defect landscape that future retention models need to consider, ruling out the need to track impurity-hydrogen coupling for this class of damage.","The framework extends directly to deuterium and tritium by substituting isotope masses and zero-point energies, making it immediately applicable to the tritium fuel-cycle problem that motivated the work.","The sensitivity of predicted diffusivity to the mobile gas fraction — sometimes only one or two atoms in the simulation box — implies that experimental validation will require carefully controlled low-concentration measurements rather than saturation-loading studies."],"fun_headline_variants":["First-principles multi-occupancy trap model matches MD for hydrogen in irradiated metals","Analytic multi-occupancy trap model reproduces hydrogen diffusivity in tungsten and vanadi","No-fit multi-occupancy trap theory predicts hydrogen retention in irradiated bcc metals","Multi-occupancy trap model outperforms single-occupancy formalism for hydrogen in tungsten","First-principles hydrogen trapping model validated against atomistic simulation in bcc met"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The void retention model assumes hydrogen atoms on void surfaces are in dynamic steady state with mobile interstitial hydrogen, and uses a simplified quartic surface binding energy with a single fitted parameter and equal vibrational frequencies for surface and lattice atoms. Because the predicted diffusivity is extremely sensitive to the mobile gas fraction — sometimes only one or two mobile atoms in the simulation box — small errors in the surface binding model propagate","fun_headline_variants_meta":{"raw":{"variants":["First-principles multi-occupancy trap model matches MD for hydrogen in irradiated metals","Analytic multi-occupancy trap model reproduces hydrogen diffusivity in tungsten and vanadium","No-fit multi-occupancy trap theory predicts hydrogen retention in irradiated bcc metals","Multi-occupancy trap model outperforms single-occupancy formalism for hydrogen in tungsten","First-principles hydrogen trapping model validated against atomistic simulation in bcc metals"]},"model":"glm-5.2","effort":"high","cost_usd":0.0,"raw_usage":{"total_tokens":733,"prompt_tokens":618,"completion_tokens":115,"prompt_tokens_details":null},"tokens_in":618,"tokens_out":115,"duration_ms":56055,"temperature":1.0,"reasoning_tokens":null,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-08T15:49:38.883763+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"Run molecular dynamics or experiment at a temperature and concentration where the model predicts a specific mobile fraction (e.g., 800 K, low-occupancy voids in tungsten, where the paper already notes discrepancy) and show that the multi-occupancy model's diffusivity prediction deviates from measurement by more than the single-occupancy model fitted at that point — which would undermine the claim of quantitative superiority across concentration ranges.","supporting_citations":[],"review_version":1}