{"id":"e517e290-0c3c-4244-a895-26196d4ffc3a","arxiv_id":"2607.15387","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Simulations show that isolated 'dead' sodium formed at grain-boundary junctions during stripping persists and accelerates dendrite penetration across cycles in solid-state sodium batteries.","lead":"This paper uses computer simulations to show that in solid-state sodium batteries, tiny pieces of sodium metal can become stranded inside the ceramic electrolyte during cycling. These leftover pieces then help new dendrites grow deeper on the next charge, making the battery fail faster.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation 10's voltage waveform cannot alternate plating/stripping as printed, so the cycling evidence for the isolated-Na memory claim is unsupported.","rationale":"The reader's weakest assumption—mapping DFT slab surface electron densities to grain-boundary/internal-interface densities—is a legitimate physical concern that the authors explicitly acknowledge in Section 2.1. However, the boundary-condition problem is more load-bearing because it undermines the demonstration of the central mechanism before any physics approximation is invoked. The central claim is specifically about what happens between plating and stripping cycles; if Eq. 10 does not alternate sign, no stripping occurs, so isolated Na cannot be created by the stripping process as described. The stated parameters make Eq. 10 monotone in the negative direction and dimensionally suspect; the only way around this is to assume a typographical error and/or normalized time, but without code/data that rescue is unverifiable. This concern is independent of the DFT-to-GB mapping: even an exact GB electron density would not help if the simulation never strips. I credit the parametric study isolating c_surf, the parallel-GB control (SI S7), and the uniform-conductivity control as good experimental design within the simulation, but the printed model cannot distinguish the memory mechanism from a boundary-condition artifact. The reader's CONDITIONAL verdict already flags the boundary condition in its rationale, so my recommendation is UNCHANGED; if a corrected Eq. 10 and the simulation code/data were supplied, the central claim could be re-evaluated fairly.","tokens_in":12954,"tokens_out":9306,"duration_ms":86542,"concrete_test":"Analytically evaluate Eq. 10 over t∈[0,0.01] s with t_total=0.01 s, t_cycle=0.002 s and count the zero crossings of φ(t); if there are none, the boundary condition contains no stripping segments. Then re-run the Figure 3 cycling simulation with a corrected boundary condition, e.g. φ(0,t) = −0.2 V·tanh(10 sin(2πt/0.002 s)), and check whether isolated Na still forms at grain-boundary triple junctions. If no isolated Na forms, the memory effect is an artifact of an unstated or corrected cycling signal; if it persists, the central mechanism survives the boundary-condition fix.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing issue is the applied-voltage boundary condition in Eq. 10. With t_total=0.01 s and t_cycle=0.002 s, the sine argument is 2π(t_total/t_cycle)t = 10πt. Over the stated simulation interval t∈[0,0.01] s, this argument spans [0,0.1π], so sin is non-negative and φ(t) = −0.2 V·tanh(10 sin(...)) never changes sign: the cell is only polarized negatively and never strips. The formula is also dimensionally suspect unless t is implicitly normalized, which is not stated. Five plating–stripping cycles cannot be produced by this boundary condition, so the simulated 'isolated Na persists between cycles' memory effect—the paper's central claim—is not supported by the simulation as described. This is a more immediate threat than the acknowledged DFT surface-to-GB electron-density approximation, because stripping is a prerequisite for forming the isolated remnants. With no code or data deposited, one cannot tell whether the actual COMSOL implementation used a corrected waveform, making the central claim unverifiable as printed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a two-dimensional phase-field model, parameterized by DFT-computed surface electron densities, to simulate Na dendrite plating and stripping in a polycrystalline Na3SbS4 solid electrolyte over five consecutive cycles. The central claim is that dendrite stripping is intrinsically asymmetric with respect to plating because grain-boundary triple junctions cause premature electronic disconnection, leaving isolated metallic Na that is kinetically trapped, persists between cycles, and serves as a reactivation seed for deeper dendrite penetration in subsequent plating steps. The authors support this with simulations under different voltages, microstructures, and a Na-Na3Sb composite anode, and propose design principles for suppressing the 'dendrite memory' effect.","tokens_in":13343,"tokens_out":4713,"duration_ms":53297,"significance":"If the central claim is correct, the paper identifies a concrete, microstructurally rooted mechanism for cycle-by-cycle dendrite penetration in Na solid-state batteries, which is important for battery safety and lifetime. The strengths of the paper are its explicit DFT-to-phase-field parameter derivation (c_surf is computed, not fitted), the careful documentation of parameters in Table 1, and the sensitivity analyses for surface electron density and its anisotropy. These features make the model internally transparent and partially falsifiable. However, the paper's main predictive claim depends on the simulation actually switching between plating and stripping, and the published voltage waveform appears unable to do so; this is a load-bearing issue that must be resolved. The acknowledged DFT slab-to-GB electron-density mapping is another major source of uncertainty, as is the lack of any direct experimental benchmark. The paper is likely of interest to the solid-state-battery modeling community, but the core evidence as printed does not yet support the 'isolated Na memory' conclusion.","major_comments":[{"comment":"The voltage boundary condition as printed cannot produce the claimed five plating-stripping cycles. With t_total=0.01 s and t_cycle=0.002 s, the sine argument is 2π(t_total/t_cycle)t = 10πt. Over the simulated interval t∈[0,0.01] s, this argument spans [0, 0.1π], so sin(10πt) is nonnegative and φ(t) = −0.2 V·tanh(10 sin(...)) never changes sign. The cell is only polarized negatively; no stripping occurs. The formula is also dimensionally inconsistent unless t is implicitly normalized, which is not stated. Since the formation of isolated Na requires stripping, the central claim of the paper is not supported by the simulation as described. The authors must either supply the correct waveform and show that it indeed alternates, or deposit the code/data from the actual COMSOL implementation.","section":"§2.2, Eq. (10)"},{"comment":"There is a direct inconsistency in the quoted time scales. Table 1 gives a physical time step Δt = 0.2 s (from Δt0 = 4×10^3 s and reduced Δt/Δt0 = 5×10^-5), while t_total = 0.01 s and t_cycle = 0.002 s in Eq. (10). A 0.01 s total simulation with a 0.2 s time step would contain less than one time step, making the reported 'five consecutive cycles' impossible. This suggests that either the normalization of Eq. (10) is different from what is printed, or Table 1 is in error. Because reproducibility and internal consistency of the cycling protocol are prerequisites for the paper's core claim, this needs to be corrected and clarified before the results can be evaluated.","section":"Table 1 vs. Eq. (10)"},{"comment":"The mapping of DFT-derived surface electron densities of free (100), (110), and (111) slabs to grain-boundary and internal-interface electron densities is acknowledged to be an approximation, but it is load-bearing: the phase-field model localizes excess electrons at GBs through c_surf, and the parametric study in §3.2.2 shows c_surf is a critical parameter controlling penetration depth and isolated Na size. If the DFT slab charges are not representative of GB excess electrons, the entire GB-facilitated dendrite-penetration and isolated-Na-memory mechanism could be an artifact. The authors should test this mapping, for example by computing excess charges at representative GB structures via DFT or by comparing the predicted penetration depths with available experimental dead-Na observations. This is a scientific concern, not a formatting issue.","section":"§2.1, §3.2.1"},{"comment":"The central mechanistic claim—that GB triple-junction geometry induces asymmetric stripping—is demonstrated only in a two-dimensional simulation. In 2D, a triple junction is a point, whereas in 3D it is a line; the necking and disconnection dynamics are qualitatively different. The paper also draws strong conclusions about 'intrinsically asymmetric' stripping and design principles without any experimental validation. The authors should at least discuss whether the 2D point-junction mechanism survives in 3D, and ideally perform a 3D simulation on a small domain to confirm the isolated-Na formation. Without this, the generality of the central claim remains speculative.","section":"§3.2, Figure 3"}],"minor_comments":[{"comment":"State the units of t and t_total/t_cycle explicitly. The sine argument should be dimensionless; as written it has units of time. A corrected form such as sin(2πt/t_cycle) would make the intended five cycles visible.","section":"Eq. (10)"},{"comment":"Typo: 'where also examined' should be 'were also examined'.","section":"§3.2.1"},{"comment":"The sentence 'the calculated anisotropy has a negligible impact on the overall results' is supported by the sensitivity study, but the criterion for 'negligible' (penetration depth agreement? morphology?) is not quantified. Please provide a quantitative threshold or comparison.","section":"§3.1"},{"comment":"The phrase 'As a result, dead Na already forms at the grain triple junction' uses 'dead Na' interchangeably with 'isolated Na.' Define the term once and use it consistently, especially because 'dead' can imply electrochemical inactivity whereas the paper argues the isolated Na is reactivatable.","section":"§3.2"},{"comment":"Reference [45] (Windl, diffusion in silicon) is cited for the statement that dangling bonds lead to deep gap states. This is a nonstandard citation; a textbook or a reference on surface states in chalcogenides would be more appropriate.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The Eq. (10) and time-step issues are serious and directly undermine the central claim. If the authors cannot supply a corrected waveform and confirm that the simulations actually alternate plating/stripping, the manuscript should be rejected. The DFT-slab-to-GB mapping is a further weakness but is at least acknowledged; however, the paper's value hinges on this mapping, so it needs more than an acknowledgment. The paper would also benefit from even one experimental comparison (e.g., a published dead-Na or dendrite-depth measurement) to anchor the predictions. I recommend major revision, not acceptance, because the core mechanism may be salvageable with corrected simulations but is not verifiable as printed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The paper is a phase-field + DFT study of Na dendrite growth in Na3SbS4 that claims dendrite stripping is intrinsically asymmetric at grain-boundary triple junctions, leaving isolated Na that persists between cycles and acts as a 'memory' driving deeper penetration. That's a concrete, mechanism-level idea that goes beyond prior dead-Li phase-field work. But the printed voltage boundary condition in Eq. 10 cannot produce the five plating-stripping cycles the paper says it runs. The sine argument is 2π(t_total/t_cycle)t = 10πt; over t in [0,0.01] s that spans 0 to 0.1π, so sin stays positive and φ(t) never changes sign. The cell is only polarized one way, no stripping occurs, and the memory mechanism is unsupported as printed.\n\nWhat's actually good: the DFT inputs are used in good faith — c_surf is computed from slabs, not fitted to the dendrite results, so the main mechanism isn't circular. The authors run useful controls: parallel GBs dissolve fully instead of isolating Na, and uniform conductivity also avoids isolation, which supports the geometric/electronic origin they claim. They also check sensitivity to surface-electron-density anisotropy and find it negligible, justifying their isotropic approximation. The model and parameters are tabulated clearly, and the acknowledged surface-to-GB mapping is an honest limitation.\n\nSoft spots: the Eq. 10 problem is load-bearing, not cosmetic. It might be a typesetting slip — the natural intended waveform would be sin(2π t/t_cycle) with t_cycle=0.002 s giving five full cycles over 0.01 s. But as printed, the formula reads otherwise, and without code or data we can't tell what COMSOL actually used. Second, there is no experimental validation; the central claim is a simulation prediction. That's acceptable for this type of paper if the simulation is internally consistent, but the boundary condition issue makes that uncertain. Third, D_Na = D_SE and the 2D geometry are assumptions that could affect quantitative penetration depths, though they likely don't kill the qualitative mechanism.\n\nWho it's for: people working on Na or Li solid-state batteries, phase-field modelers, and anyone interested in 'dead metal' memory effects. The paper deserves a serious referee — the mechanism is plausible and the framework is established — but the referee must demand a corrected Eq. 10 and preferably code or data sharing. If the authors fix the waveform and confirm the cycling results, this could be a useful contribution. I'd hold off citing it until then.","headline":"Plausible mechanism for dendrite 'memory' in Na SSBs, but the printed voltage waveform can't produce the claimed cycling, so the central result is unverifiable as written.","tokens_in":13740,"tokens_out":3955,"would_cite":false,"duration_ms":35109,"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":"Dendrite stripping leaves isolated sodium metal at grain-boundary junctions, and that residual sodium reconnects on the next plating cycle to drive deeper dendrite penetration.","keywords":["sodium solid-state battery","dendrite growth","phase-field modeling","grain boundary","isolated dead sodium","Na3SbS4 electrolyte","DFT surface electron density","battery cycling"],"falsifier":"Cycle a Na | Na3SbS4 | Na cell, strip completely, then section and inspect with cryo-FIB/SEM or EELS: if no isolated Na remains at grain-boundary triple junctions, the claimed memory mechanism is falsified. Alternatively, compute the excess electron density at an explicit grain boundary rather than a free surface and rerun the phase-field simulation; if the isolated Na disappears, the mechanism is an artifact of the surface-electron approximation.","tokens_in":12908,"feed_emoji":"🔋","tokens_out":4291,"duration_ms":48658,"temperature":0.7,"pith_summary":"This paper tries to establish why dendrites in all-solid-state sodium batteries get worse cycle after cycle rather than reversing during stripping. Using a phase-field model fed with density-functional-theory values for excess electrons at the solid-electrolyte surface, the authors find that stripping is intrinsically asymmetric: sodium in a dendrite dissolves first at the stem, cutting off the portion of the dendrite that sits beyond a grain-boundary triple junction. That cut-off sodium is thermodynamically unstable but kinetically trapped, so it survives the stripping half-cycle. On the next plating half-cycle it acts as a preferential seed, reconnects electronically, and lets the dendrite advance deeper. The paper argues that this \"dead\" sodium is a structural memory that accelerates penetration, and that voltage, grain-boundary density, and anode composition can control it.","feed_headline":"Stripped sodium fragments reseed dendrites every cycle","feed_subtitle":"Phase-field simulations show grain-boundary junctions trap residual Na metal that reconnects and drives deeper penetration in solid-state so","key_machinery":"The argument runs on a nonlinear phase-field electrodeposition model coupled to an electrochemical reaction-rate law (Butler-Volmer kinetics) and reaction-diffusion equations for Na+ and electric potential. Its load-bearing input is the excess electron concentration at the electrolyte, computed by DFT as surface electron densities of NAS slabs and inserted into the electron concentration field as c_e-/c0 = xi + (c_surf/c0)(1-xi)(1-phi_g), so grain boundaries are the only places with extra electrons available to seed deposition. The geometric key is the grain-boundary triple junction: the narrow dendrite stem acts as an electron bottleneck during stripping, causing premature disconnection and","core_discovery":"The central claim is that dendrite stripping is intrinsically asymmetric with respect to plating because of grain-boundary geometry: when a dendrite occupying a grain-boundary channel is stripped, dissolution proceeds fastest along sidewalls and through the constricted stem, so the portion of metal beyond a triple junction loses electronic contact before it can dissolve. That isolated Na metal, rather than disappearing, becomes kinetically stabilized—a negative-potential basin, local Na+ shielding, and loss of electron supply all suppress further stripping. It is therefore present at the start of the next plating cycle, reconnects with the advancing dendrite, and reactivates as a site for fu","pith_inferences":["Beyond the paper's own claims: if the same geometric asymmetry applies to other polycrystalline sulfide electrolytes, cycling protocols that interrupt the electron path at the dendrite stem—short pulses or intermittent rest—could be tested for their ability to dissolve isolated metal before it reconnects.","The DFT slab approximation could be checked directly: calculating excess electron density at an explicit Na3SbS4 grain boundary and rerunning the phase-field simulation would show whether the trapping effect survives a more physical interfacial electron distribution.","A quantitative experimental signature follows from the model: after stripping, residual sodium imaged at grain-boundary triple junctions should correlate with the penetration depth observed in the next cycle; this could be tested with cross-sectional microscopy on cycled cells.","Because the model treats the conductivity mismatch as essential, a materials-level prediction is that electrolytes with electronic conductivity closer to ionic conductivity should show little or no isolated Na, since the authors note a uniform-conductivity control dissolves fully."],"forward_implications":["If isolated Na is a memory for dendrite penetration, then suppressing its formation should slow cycle-over-cycle degradation, not just the initial plating depth.","Dense microstructures with many fine grain boundaries dissolve Na more completely, so fine-grained electrolytes should outperform coarse-grained or void-containing ones.","Low-diffusivity Na3Sb alloy anodes reduce penetration (about 8% in the simulations) by slowing deposition and improving dissolution, pointing to anode alloying as a mitigation lever.","Applied voltage sets a threshold: in the model, penetration stays shallow with little isolated Na near 0.15 V, while 0.18 V and 0.20 V produce deep penetration and trapped Na.","Intergranular voids are worse than grain boundaries because they allow fast propagation and early disconnection, making void-free processing a concrete design target."],"fun_headline_variants":["Stripped sodium bits regrow dendrites each cycle","Residual Na metal at grain junctions reignites dendrites","Dendrite remnants trapped in grain boundaries reseed growth","Asymmetric stripping leaves Na pockets that reactivate next plating","Grain-boundary Na leftovers make dendrites worse every cycle"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that excess electrons at grain boundaries and internal interfaces inside the polycrystalline electrolyte can be represented by the DFT-calculated surface electron densities of free NAS slabs; the authors themselves state that this mapping is an approximation. If real grain boundaries do not carry comparable electron densities, the grain-boundary plating and the whole isolated-Na memory mechanism may not occur.","fun_headline_variants_meta":{"raw":{"variants":["Stripped sodium bits regrow dendrites each cycle","Residual Na metal at grain junctions reignites dendrites","Dendrite remnants trapped in grain boundaries reseed growth","Asymmetric stripping leaves Na pockets that reactivate next plating","Grain-boundary Na leftovers make dendrites worse every cycle"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00016,"raw_usage":{"total_tokens":1063,"prompt_tokens":732,"completion_tokens":331,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":263}},"tokens_in":476,"tokens_out":331,"duration_ms":4145,"temperature":1.0,"reasoning_tokens":263,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T23:30:31.215557+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Cycle a Na | Na3SbS4 | Na cell, strip completely, then section and inspect with cryo-FIB/SEM or EELS: if no isolated Na remains at grain-boundary triple junctions, the claimed memory mechanism is falsified. Alternatively, compute the excess electron density at an explicit grain boundary rather than a free surface and rerun the phase-field simulation; if the isolated Na disappears, the mechanism is an artifact of the surface-electron approximation.","supporting_citations":[],"review_version":1}