{"id":"e6da4e61-08f6-497f-8b7d-ddca60f005a2","arxiv_id":"1908.06523","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Alternating the shear plane during cyclic loading lowers the potential energy of a model glass below the single-plane plateau, with three-axis alternation approaching the slow-cooled reference energy.","lead":"Molecular dynamics simulations of a model glass show that cycling shear in alternating directions drives the material to lower energy states than cycling in a single plane. The result points to a simple protocol change that could make mechanical annealing of metallic glasses faster.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The protocol-ordering claim rests on one MD trajectory per protocol with no error bars; energy differences of ~0.01–0.02 epsilon are not statistically established, as the paper itself notes in Sec. II.","rationale":"The reader's weakest_assumption identifies the same single-sample limitation, and this is indeed the most load-bearing concern. The paper's own Sec. II flags it, and the central claim is a comparative statement about protocols that requires representative sampling. The energy differences (0.01-0.02 epsilon) are small relative to what could plausibly vary between independent glass preparations, so without replicates the headline ordering is not established. A conditional verdict is appropriate rather than rejection: the reported trends are internally consistent, the protocol design is reasonable, and the single-trajectory limitation is an acknowledged gap rather than a demonstrated error. The proposed multi-sample test would settle whether the ordering is robust. I find no independent basis to strengthen the concern beyond conditional, and no reason to accept the claim as fully verified.","tokens_in":9798,"tokens_out":3389,"duration_ms":41629,"concrete_test":"Prepare 4-5 independent Kob-Andersen glasses with the same density, cooling rate, and final temperature but different random initial configurations. Run 1400 shear cycles for each protocol (xz; alternating xz,yz; alternating xz,yz,xy) at gamma0 = 0.03 and 0.06, and record the mean and standard error of the final potential energy. If the ordering U_xz > U_xz,yz > U_xz,yz,xy is not preserved with separations exceeding roughly two standard errors, the central claim is not statistically established. A cheaper supplementary check is to rerun the existing protocols with different thermostat seeds on the same initial configuration to gauge trajectory-level stochasticity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II explicitly states: 'Due to computational limitations, the simulations of periodic shear deformation were performed only for one sample.' The central claim—that each additional shear orientation lowers the final potential energy—is read from Figs. 2 and 3, where at gamma0 = 0.06 the final energies are roughly U ≈ -8.26 (xz), -8.275 (xz,yz), and -8.29 (xz,yz,xy) epsilon. These differences are of order 0.01-0.02 epsilon per atom. With a single preparation, the monotonic ordering could reflect properties of one particular initial configuration or one stochastic trajectory rather than a general protocol effect; the absence of replicate samples means the reported separation cannot be distinguished from sample-to-sample fluctuations. The argument is internally consistent and the curves are suggestive, but the load-bearing premise 'one trajectory is representative' is explicitly unverified. This is the weakest point because if another realization reversed or flattened the ordering at gamma0 = 0.01 or 0.03, the headline claim would collapse. The mechanical-property data in Fig. 5 are also scattered, but the energy ordering is the primary claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses molecular dynamics simulations of a Kob-Andersen binary Lennard-Jones glass (60,000 atoms) to compare cyclic shear deformation protocols: single-plane shear (xz), alternating two-plane shear (xz, yz), and alternating three-plane shear (xz, yz, xy). For strain amplitudes in the claimed elastic range (γ0 ≤ 0.065), the potential energy decreases with increasing amplitude within each protocol. The central claim is that, for a fixed amplitude, each additional alternation of the shear orientation produces a lower final potential energy, with the three-plane protocol approaching the energy of a slowly cooled glass. The paper further reports, after 1400 cycles, an increase in yield stress and orientation-dependent shear moduli, along with qualitative observations of shear localization from nonaffine displacement fields in subsequent steady shear.","tokens_in":9956,"tokens_out":4126,"duration_ms":42384,"significance":"If the central ordering result holds, the paper identifies a simple and potentially useful protocol for mechanical annealing of glasses: multi-axis cyclic loading at sub-yield amplitudes drives the system to deeper energy minima than single-plane cycling alone. The study uses a standard, well-characterized model and a direct simulation approach with no fitted parameters; the γ0 = 0 baseline is stable, and the energy series are internally consistent. The main significance is therefore as a candidate protocol comparison. However, the strength of the claim is currently limited by the single-trajectory statistics, since the reported inter-protocol energy differences (≈0.01–0.02 ε per atom) are of the same order as typical sample-to-sample fluctuations in such systems.","major_comments":[{"comment":"The central claim that each additional shear orientation brings the glass to lower energy states rests on a single molecular dynamics trajectory per protocol, as the paper explicitly states in Section II: 'the simulations of periodic shear deformation were performed only for one sample.' At γ0 = 0.06 the inter-protocol differences in Fig. 3 are roughly 0.01–0.02 ε per atom (U ≈ −8.26, −8.275, −8.29 for xz, xz–yz, and xz–yz–xy, respectively), while at γ0 = 0.01 and 0.03 the separation is comparable to the cycle-to-cycle fluctuations visible in Fig. 1. Without independent initial configurations or error bars, the ordering in Fig. 3 cannot be distinguished from sample-to-sample fluctuations, so the headline conclusion is not statistically established. The authors should run replicate samples per protocol (at least at the smaller amplitudes where the separation is marginal) and report averages with error bars, or alternatively restrict the claim to the single sample studied.","section":"Section II; Figs. 1–3"},{"comment":"The mechanical-property trends — that the yield peak increases when an additional shear orientation is introduced and that the shear modulus is larger along directions not used during cyclic loading — are also drawn from one sample per protocol. The data in Fig. 5 are visibly scattered (e.g., the shear-modulus values for a given protocol vary by several units across directions and amplitudes), so without replicate samples these trends, like the energy ordering, are not statistically supported. This is a secondary but still load-bearing part of the abstract's claim of increased strength and modulus anisotropy.","section":"Section III; Fig. 5"},{"comment":"The statement that the three-plane protocol attains U ≈ −8.29 ε, 'approaching' the slowly cooled value of U ≈ −8.31 ε, is made without an uncertainty estimate. Given the single-sample limitation, the residual gap of 0.02 ε per atom may be statistically meaningful or may be within fluctuations; the paper should either provide error bars or soften this comparison.","section":"Section III, comparison with slowly cooled reference"}],"minor_comments":[{"comment":"The cooling rate is rendered as '10 −2ε/kBτ' with an awkward spacing; please format as 10^{-2} ε/(kB τ) for clarity.","section":"Section II"},{"comment":"The mention of test simulations at γ0 = 0.07 that show flow localization within the first 300 cycles is not shown; consider reporting at least the final energy levels or the time of localization in a supplementary figure, since this is directly relevant to the claim that the studied amplitudes are in the elastic range.","section":"Section III (not shown result)"},{"comment":"The legend entry '10 xz, 10 yz' would benefit from an explicit statement that this denotes 10 consecutive cycles along xz followed by 10 consecutive cycles along yz; the caption currently describes it in the text but not in the legend itself.","section":"Fig. 2 caption"},{"comment":"Reference [40] is cited as an arXiv preprint (Das, Parmar, and Sastry, 'Annealing glasses by cyclic shear deformation'); if a peer-reviewed version has appeared, please cite the published version instead.","section":"References"},{"comment":"The title word 'Accelerated' implies a rate comparison, but the paper reports final energies after a fixed number of cycles rather than relaxation rates. Consider clarifying in the introduction whether 'accelerated' refers to reaching lower energies after the same number of cycles or to a faster relaxation rate.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The single-sample statistics are the key issue. The work is within the journal's scope and the simulation methodology is sound, but the central ordering claim needs either additional independent samples (which are expensive but feasible) or a carefully worded revision that confines the conclusions to the studied sample. I would not reject the paper on the current evidence; a major revision with added statistics or a scaled-back abstract would make it publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the central result—that alternating the shear plane every cycle in two or three dimensions puts a rapidly cooled Kob–Andersen glass into deeper potential-energy minima than single-plane cycling at the same amplitude—is plausible, internally consistent, and genuinely new. But it is built on one MD trajectory per protocol, and the paper says so in Section II. The energy gaps between protocols are ~0.01–0.02 epsilon per atom, so without replicated samples the ordering is not statistically established. I think the paper deserves a serious referee; the claim is important enough that the single-sample limitation should be addressed by the authors or, failing that, clearly flagged as provisional.\n\nWhat's actually new: prior single-plane cyclic annealing is well studied, and ref [40] reported no significant effect for alternating planes near the critical amplitude. Here, at amplitudes 0.01–0.065, each added orientation lowers the plateau energy, with the three-axis protocol reaching U ≈ -8.29 epsilon, close to the slow-cooled value -8.31. The authors also show an increase in the yielding peak and a shear-modulus anisotropy consistent with the energy ordering. The nonaffine displacement snapshots are a nice complement. The paper is honest: it reports the one-sample limitation, notes the yielding transition at 0.07, and doesn't overclaim.\n\nSoft spots, in proportion: the single-sample issue is the main one. The energy differences are small, and the Fig. 5 mechanical data are scattered. The ordering at gamma0 = 0.01 and 0.03 rests on even smaller separations, so for those amplitudes the conclusion is more like a hypothesis than a measurement. I'd also like to see at least one independent initial configuration or a second cooling history to check that the ordering is robust. That said, the simulations are long (1400 cycles, 60k atoms) and expensive; a full ensemble is a lot to ask, but even a single repeat for one amplitude (say 0.06) would materially change the confidence.\n\nVerdict: this is a useful exploratory result for the mechanical-annealing community. It deserves peer review rather than a desk rejection, but the referee should ask for replicate simulations or a much clearer statistical statement. The paper is worth citing as the first systematic comparison of multi-axis cyclic shear, with the caveat that the ordering is provisional.","headline":"The multi-axis mechanical annealing result is new and plausible, but single-sample statistics leave the protocol ordering provisional.","tokens_in":10505,"tokens_out":2484,"would_cite":true,"duration_ms":26750,"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":"Cyclic shear with an alternating shear plane—especially cycling through all three spatial directions—drives a model metallic glass to lower potential energies than single-plane cycling at the same sub-yield strain amplitude, approaching…","keywords":["metallic glasses","cyclic loading","periodic deformation","shear orientation","structural relaxation","potential energy landscape","shear modulus anisotropy","molecular dynamics"],"falsifier":"Repeat the three protocols (xz; xz+yz; xz+yz+xy) for at least five independently quenched samples at $\\gamma_0 = 0.01$, $0.03$, and $0.06$, and compare the end-of-cycle potential energies as an ensemble. If the single-plane sample is sometimes as deep as the three-plane sample, or if the ordering does not hold within the spread, the claim that each added orientation lowers the energy is not supported. A complementary check: run the same protocols under athermal quasistatic shear and see whether the limit-cycle energy still decreases with added orientations.","tokens_in":9548,"feed_emoji":"🔁","tokens_out":7875,"duration_ms":70159,"temperature":0.7,"pith_summary":"This paper tries to establish that the direction of cyclic mechanical loading can be used as a control knob for structural relaxation in glasses. In molecular dynamics simulations of a rapidly cooled binary glass, 1400 cycles of shear at strain amplitudes below yielding lower the potential energy, and the lowering is greater when the shear plane alternates between two or three perpendicular directions. The deepest energy, $U \\approx -8.29\\,\\varepsilon$ per atom at $\\gamma_0 = 0.06$, is reached by alternating among the xz, yz, and xy planes every cycle, close to the slowly cooled reference value of about $-8.31\\,\\varepsilon$. If correct, this makes multi-axis cyclic loading a more efficient mechanical annealing protocol than single-plane cycling, and it gives a practical route to stronger, more relaxed glasses without slow thermal annealing.","feed_headline":"Three-axis shear cycling relaxes glass near to slow-cooled state","feed_subtitle":"Simulations show each added shear direction in cyclic loading lowers the glass energy and raises yield strength.","key_machinery":"The object doing the work is the deformation protocol itself: periodic shear $\\gamma(t) = \\gamma_0\\sin(2\\pi t/T)$ with period $T=5000\\tau$, applied along one plane, alternating between two planes, or alternating among three mutually perpendicular planes. The quantity that carries the comparison is the potential energy at the end of each shear cycle, tracked over 1400 cycles, with the slowly cooled glass as the reference energy. Under cyclic strain the energy landscape is effectively tilted, so groups of atoms can undergo irreversible rearrangements into lower minima; changing the shear plane appears to open additional rearrangement pathways. The nonaffine displacement measure $D^2$ (squared deviation of atomic trajectories from the best affine fit) is used to show where these rearrangements localize, including the shear band that forms during subsequent monotonic loading.","core_discovery":"The central claim is that, within the elastic range ($\\gamma_0 \\le 0.065$), each additional alternation of the shear orientation in the deformation protocol relocates the glass to deeper energy minima. The ordering holds for every amplitude tested: single-plane shear along xz gives the highest energy, alternating xz/yz goes lower, and alternating among xz, yz, and xy goes lower still. At $\\gamma_0 = 0.06$, the last protocol reaches $U\\approx -8.29\\,\\varepsilon$, within about $0.02\\,\\varepsilon$ of the slowly cooled glass, and the energy curves nearly coincide at $\\gamma_0 = 0.06$ and $0.065$ for the multi-plane protocols, suggesting a slightly lower critical amplitude there. After cycling, steady-shear tests show a pronounced yielding peak, and the shear modulus is larger along planes that were not used during cyclic loading. The paper concludes that cyclic loading in the elastic regime increases strength and produces shear-modulus anisotropy, with more orientations giving stronger relaxation.","pith_inferences":["The same ordering might hold in athermal quasistatic cycling, where a fully reversible limit cycle bounds the attainable energy; alternating the shear direction could lower that bound. This is a direct, testable extension of the paper's mechanism.","Because the energy differences between protocols are only about $0.01$-$0.02\\,\\varepsilon$ per atom, the protocol ranking should be rechecked with several independently prepared samples; the paper itself notes only one sample was simulated per protocol.","A practical optimization question follows: how often the shear plane should be switched (every cycle vs every 10 cycles) and whether non-perpendicular plane sets work as well. The paper only tests every-cycle and every-10-cycle switching at $\\gamma_0=0.06$.","If multi-axis mechanical annealing works in bulk metallic glasses, it could complement thermal cycling treatments for tuning ductility and strength."],"forward_implications":["For a given sub-yield strain amplitude, adding shear orientations monotonically lowers the attainable potential energy; three-axis cycling is therefore a faster mechanical annealing route than single-plane cycling.","Cyclic loading raises the yield stress of the relaxed glass, and the yield peak grows with both strain amplitude and the number of shear orientations in the protocol.","Cyclic loading creates mechanical anisotropy: the shear modulus is larger along planes that were not deformed during cycling, which should be measurable in experiments on trained metallic glasses.","Deeper relaxation correlates with more pronounced strain localization: after multi-plane cycling, subsequent monotonic shear produces a clear shear band soon after the yielding peak.","At the higher sub-yield amplitudes, the two- and three-plane protocols nearly saturate in energy between $\\gamma_0 = 0.06$ and $0.065$, implying a slightly lower protocol-dependent critical strain amplitude."],"supporting_citations":[{"why":"Defines the strongly non-additive binary mixture and its parameters; the model glass used throughout the simulations.","marker":"[43]"},{"why":"The molecular dynamics code used to equilibrate, cyclically deform, and mechanically test the samples.","marker":"[44]"},{"why":"Established oscillatory athermal quasistatic deformation and the reversible limit cycle, the phenomenon this paper extends to alternating orientations.","marker":"[26]"},{"why":"Located the yielding transition and critical strain amplitude for the same mixture, fixing the sub-yield range used here.","marker":"[33]"},{"why":"Supplies the slowly cooled reference energy and earlier evidence of collective nonaffine displacements under oscillatory shear.","marker":"[34]"},{"why":"Earlier finite-temperature study of mechanical annealing in the same model, supporting the protocol's ability to lower energy and providing context on yield onset.","marker":"[37]"},{"why":"Compared single-plane and alternating-plane shear near the critical amplitude; the paper's results extend that comparison into the sub-yield regime.","marker":"[40]"},{"why":"Introduced the nonaffine displacement measure used to identify shear transformations and shear bands in the snapshots.","marker":"[47]"}],"fun_headline_variants":["More shear axes, deeper glass energy","Alternating shear planes lower glass energy","Each added shear direction deepens glass relaxation","Three-axis cyclic shear approaches slow-cooled glass"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ordering of energies across protocols is read from a single simulated sample per protocol; if sample-to-sample fluctuations exceed the roughly 0.01-0.02 energy-unit differences between protocols, the central claim about protocol efficiency collapses.","fun_headline_variants_meta":{"raw":{"variants":["More shear axes, deeper glass energy","Alternating shear planes lower glass energy","Each added shear direction deepens glass relaxation","Three-axis cyclic shear approaches slow-cooled glass"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001086,"raw_usage":{"total_tokens":4498,"prompt_tokens":862,"completion_tokens":3636,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":478,"completion_tokens_details":{"reasoning_tokens":3594}},"tokens_in":478,"tokens_out":3636,"duration_ms":28105,"temperature":1.0,"reasoning_tokens":3594,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:42:12.816122+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the three protocols (xz; xz+yz; xz+yz+xy) for at least five independently quenched samples at $\\gamma_0 = 0.01$, $0.03$, and $0.06$, and compare the end-of-cycle potential energies as an ensemble. If the single-plane sample is sometimes as deep as the three-plane sample, or if the ordering does not hold within the spread, the claim that each added orientation lowers the energy is not supported. A complementary check: run the same protocols under athermal quasistatic shear and see whether the limit-cycle energy still decreases with added orientations.","supporting_citations":[{"cited_title":"Kob and H","cited_arxiv_id":null,"evidence_quote":"Defines the strongly non-additive binary mixture and its parameters; the model glass used throughout the simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The molecular dynamics code used to equilibrate, cyclically deform, and mechanically test the samples."},{"cited_title":"Fiocco, G","cited_arxiv_id":null,"evidence_quote":"Established oscillatory athermal quasistatic deformation and the reversible limit cycle, the phenomenon this paper extends to alternating orientations."},{"cited_title":"Leishangthem, A","cited_arxiv_id":null,"evidence_quote":"Located the yielding transition and critical strain amplitude for the same mixture, fixing the sub-yield range used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the slowly cooled reference energy and earlier evidence of collective nonaffine displacements under oscillatory shear."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier finite-temperature study of mechanical annealing in the same model, supporting the protocol's ability to lower energy and providing context on yield onset."},{"cited_title":"Annealing glasses by cyclic shear deformation","cited_arxiv_id":"1805.12476","evidence_quote":"Compared single-plane and alternating-plane shear near the critical amplitude; the paper's results extend that comparison into the sub-yield regime."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the nonaffine displacement measure used to identify shear transformations and shear bands in the snapshots."}],"review_version":1}