{"id":"03b6da73-c8f9-409b-ae44-cdc76131185c","arxiv_id":"2501.16806","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Reversing the Ti/W stacking order inverts the electric current vorticity and flips the sign of the spin torque generated by a compositionally graded interface, supporting the spin-vorticity coupling mechanism.","lead":"Researchers find that flipping the deposition order of ultrathin titanium and tungsten layers reverses the direction of a spin current generated at a compositionally graded interface. The result points to a stacking-order knob for controlling spin-torque sign in spintronic devices made from abundant metals.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The sign-reversal claim rests on a cross-paper comparison between sample series grown under deliberately different Ti sputtering conditions; this deposition confound, rather than current-vorticity direction, could produce the opposite Δξ.","rationale":"Read in good faith, the paper has real strengths: within the current W/Ti series the t_i dependence is systematic, STEM-EDS confirms that the insertion modifies the interface (CGI width changes from 0.97 to 1.94 nm), and the reduced-Ti-rate control shows that the negative Δξ is tied to CGI formation. These internal checks are valuable. The load-bearing weakness is specifically the cross-series sign reversal used to claim that current-vorticity polarity controls the spin-current sign. The authors deliberately changed the Ti sputtering rate between Ref. [22] and this work, and their own control experiment shows that this parameter determines whether a CGI effect appears in the W/Ti geometry. Therefore the opposite Δξ cannot be uniquely attributed to the polarity of electric current vorticity unless the two stacking orders are compared under controlled deposition conditions. This concern is adjacent to the reader's weakest_assumption about t_i-independent bulk and interface properties, but it targets the comparison across two separate sample series, which is more directly tied to the sign-reversal claim. Given the thorough structural characterization and the internal consistency of the current series, a conditional verdict remains appropriate; the mechanistic conclusion should be explicitly conditioned on a matched side-by-side experiment. No change to the reader's CONDITIONAL verdict is needed, but the condition should include a matched deposition matrix rather than only additional t_i-independent characterization.","tokens_in":11494,"tokens_out":7692,"duration_ms":77993,"concrete_test":"Grow a matched matrix: for both Ti/W and W/Ti stacking orders, deposit samples at Ti sputtering rates of 0.022 and 0.039 nm/s (plus at least one intermediate rate) in the same session with W rate fixed; measure CGI width by STEM-EDS and Δξ(t_i) by ST-FMR for each condition. Then compare the sign of Δξ at a fixed deposition rate and, separately, at matched CGI widths. If opposite signs persist under matched parameters, the polarity claim survives; if the W/Ti negative Δξ appears only at the high Ti rate, the deposition-condition confound is decisive.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that inverting the Ti/W stacking order reverses the sign of the CGI spin torque. The evidence is a comparison of Δξ(t_i) measured here on W/Ti/Ni-Cu samples with data from Ref. [22] for Ti/W/Ni-Cu samples. These two series are not matched: the present samples use a Ti sputtering rate of 0.039 nm/s, whereas Ref. [22] used about 0.022 nm/s, and the authors explicitly show that reducing the Ti rate to 0.022 nm/s in the W/Ti geometry makes the CGI effect disappear (ξ stays near +0.031 to +0.034 over t_i). Thus the reversed sign is entangled with a deposition-condition change that the authors themselves demonstrate can turn the CGI contribution on or off. The two series also peak at different t_i values (1.5 nm vs about 0.5 nm) and have different magnitudes (0.008 vs 0.011), so the 'reversal' is inferred after separate interpretations of each series rather than from a controlled side-by-side comparison. If deposition conditions, not vorticity polarity, determine whether the CGI-type spin-current contribution appears, the strongest claim is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports spin-torque ferromagnetic resonance measurements on W/Ti/Ni95Cu5 microstrips with a compositionally graded interface formed by an ultrathin Ti/W insertion layer of thickness t_i. The authors find that the damping-like torque efficiency decreases monotonically with t_i up to 1.5 nm, opposite in sign to the increase reported earlier for the Ti/W/Ni-Cu stacking order, and they attribute the opposite sign to the opposite electric-current vorticity produced by the reversed conductivity gradient. The manuscript includes structural characterization (2D-XRD, AFM, STEM-EDS) showing a compositional gradient, and it evaluates alternative mechanisms (bulk spin Hall effect, orbital Hall effect, Rashba–Edelstein and orbital Rashba–Edelstein effects, a Ti-W alloy control) before concluding that spin-vorticity coupling with a polarity determined by the gradient direction is the most likely source.","tokens_in":11673,"tokens_out":4640,"duration_ms":41653,"significance":"If the sign reversal were established, the paper would identify a practical control parameter (stacking order of a compositionally graded interface) for the sign of current-induced spin torque without relying on strong spin-orbit materials. The work's strengths include the clean use of a standard ST-FMR analysis, a systematic t_i series with error bars, and direct structural evidence of the gradient and of a Ti-W alloy at larger t_i. The exclusion of the bulk Ti-W alloy contribution using a separate periodic stack is a useful falsifiable control. However, the central comparison between the two stacking orders is not yet controlled, because the W/Ti and Ti/W series were grown under different Ti sputtering rates, and the paper itself demonstrates that the W/Ti CGI effect disappears when the Ti rate is matched to the earlier work. This makes the headline claim of a vorticity-polarity reversal currently unproven, although the underlying SVC hypothesis remains plausible and testable.","major_comments":[{"comment":"The reported control experiment, in which reducing the Ti deposition rate from 0.039 nm/s to 0.022 nm/s makes the W/Ti CGI effect vanish (ξ stays near +0.031 to +0.034 over t_i), directly undermines the attribution of the sign reversal in Fig. 7b to vorticity polarity, because the Ti/W reference series (Ref. [22]) was grown at 0.022 nm/s for Ti; a controlled side-by-side comparison with identical deposition conditions is needed to separate the stacking-order effect from the growth-condition effect.","section":"Section 2.4, paragraph on sputtering condition dependence"},{"comment":"The extraction of ξ_DL assumes the bulk SHE/OHE, Oersted field, and interface torques are independent of t_i, but the paper does not report per-sample saturation magnetization μ0Ms (used in Eq. (3)) or current-shunting/resistivity data; if μ0Ms or the conductivity distribution changes with t_i, the monotonic trend in Fig. 7a could be an artifact of the extraction rather than CGI spin-current generation. Please provide the Kittel-fit values and layer-resistance estimates for each t_i.","section":"Section 2.3, Eq. (3) and Fig. 7"},{"comment":"The negative-spin-torque control using Sub./[Ti(0.5 nm)/W(0.5 nm)]10/Ni-Cu(10 nm) reports ξ=+0.026, but this is a periodically repeated stack rather than the actual CGI formed in the device series; the argument that bulk alloy effects are negligible would be stronger if the alloy composition and thickness of the CGI were directly correlated with the measured ξ_DL trend.","section":"Section 2.4, paragraph discussing the Ti-W nanoalloy"}],"minor_comments":[{"comment":"The thickness d_NM used in Eq. (3) is not defined; since the stack contains Ti(3 nm)/W(10 nm)/Ti(t_i/2)/W(t_i/2)/Ti(10 nm), it is unclear whether d_NM is a constant (23 nm) or includes t_i, and this choice affects the reported ξ_DL values.","section":"Equation (3) and Section 2.3"},{"comment":"The yellow triangle representing the Ti(10 nm)/Ni-Cu reference sample is not described in the caption; please add a sentence identifying it and explaining its significance.","section":"Figure 7 caption"},{"comment":"The abstract states that reversing the Ti/W stacking order 'suppresses positive spin torque from the orbital Hall effect in Ti'; given the deposition-rate confound discussed in Section 2.4, a more cautious formulation such as 'is accompanied by' would be more appropriate.","section":"Abstract"},{"comment":"Reference [3] is cited as a preprint with a submission date of Oct 2021; if it has been published or updated in the intervening time, please provide the final citation.","section":"Reference [3]"}],"recommendation":"major_revision","confidential_remarks":"The paper is a good candidate for this journal if the authors can close the deposition-rate gap, either by growing both stacking orders under identical conditions (and at multiple Ti rates) or by reformulating the claim as a material-specific observation rather than a general polarity law. The current evidence is not sufficient for acceptance as is, but the experiment is well-executed and the central question is important. I would not recommend rejection because the confound is identifiable and addressable within a normal revision: a side-by-side growth-series comparison would resolve it."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is a solid ST-FMR study showing that inserting a W/Ti compositionally graded interface produces a negative change in damping-like torque, opposite to the positive change reported earlier for the Ti/W gradient. The structural characterization is thorough—XRD, AFM, and STEM-EDS all support the gradient picture, and the control experiment ruling out the bulk Ti-W alloy contribution is convincing. The sign reversal is genuinely new and worth reporting.\n\nThe soft spot is the central mechanistic claim. The comparison between the two stacking orders is cross-paper, and the two series were grown under different Ti sputtering rates (0.039 nm/s here versus 0.022 nm/s in the earlier PRB). The authors themselves show that lowering the Ti rate in the W/Ti geometry makes the CGI effect disappear entirely. That means the sign difference could reflect deposition-condition sensitivity rather than vorticity polarity. They acknowledge the deposition sensitivity but interpret it as a need to re-optimize conditions; that does not remove the confound. A controlled side-by-side comparison under identical deposition conditions, or a deposition-condition matrix, is needed to support the sign-reversal claim.\n\nThere are other minor gaps. The t_i=0 cancellation of bulk W spin Hall and Ti orbital Hall effects is argued by comparison to a Ti/Ni-Cu bilayer, not demonstrated with shunting measurements. Per-sample saturation magnetization values are not reported, and the effect size (about 0.008) is modest relative to scatter at some thicknesses. No raw data or code are provided.\n\nThe paper is honest. It explicitly discusses the nano-alloy alternative, tests it, and reports the deposition-rate dependence even though it complicates the headline. The within-series trend is systematic, so there is a real effect under the stated conditions. The SVC interpretation is plausible and consistent with their prior model, but the sign-reversal claim should be treated as a hypothesis supported by a single non-controlled comparison.\n\nThis paper deserves a serious referee. It should not be desk-rejected. A good referee should ask for a controlled stacking-order comparison, per-sample magnetization and shunting data, and ideally raw spectra. For a reading group I would say maybe—useful for people in charge-to-spin conversion and SVC, but not essential outside that niche. I would cite it as a data point with a caveat.\n\nRecommendation: send to peer review, but expect major revision. The observation is publishable; the mechanism needs stronger support.","headline":"A clean, novel sign-reversal observation in the W/Ti CGI system, but the central vorticity claim is entangled with a deposition-condition change between the two sample series.","tokens_in":12303,"tokens_out":3530,"would_cite":true,"duration_ms":31982,"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":"This paper reports that reversing the Ti/W stacking order flips the sign of the spin torque produced by a compositionally graded interface, so the polarity of the compositional gradient controls the spin-current direction.","keywords":["spintronics","spin-vorticity coupling","spin torque","compositional gradient","spin-torque ferromagnetic resonance","Ti/W interface","orbital Hall effect","spin Hall effect"],"falsifier":"Measure the saturation magnetization and the sheet conductances of every $t_i$ sample and recompute $\\xi_{\\mathrm{DL}}$ with those per-sample values; if the $\\Delta\\xi_{\\mathrm{DL}}(t_i)$ curve keeps its sign and shape, the CGI-polarity claim stands, whereas a drifting magnetization or current split would show the assignment to SVC is contaminated.","tokens_in":11248,"feed_emoji":"🧲","tokens_out":7378,"duration_ms":63488,"temperature":0.7,"pith_summary":"This paper tries to establish that a nanoscale compositional gradient by itself can generate a spin current, and that the sign of the resulting spin torque is set by the polarity of the gradient rather than by the bulk spin Hall or orbital Hall effects of the constituent metals. The authors build W/Ti/Ni-Cu stacks in which an ultrathin Ti/W bilayer is inserted at the W/Ti interface to form a compositionally graded interface, and they measure the damping-like torque by spin-torque ferromagnetic resonance. They find that inserting the W/Ti CGI produces a negative change in torque efficiency (about -0.008 at $t_i = 1.5$ nm), opposite to the positive change (about +0.011) reported earlier for the Ti/W CGI. This matters because it would give device designers a simple structural knob, stacking order, for choosing the sign of spin torque without using rare or strongly spin-orbit-coupled metals.","feed_headline":"Inverting a Ti/W composition gradient flips the spin torque","feed_subtitle":"A W/Ti graded interlayer generates torque opposite to Ti/W, giving a stacking-order switch for spin current without heavy metals.","key_machinery":"The load-bearing object is spin-vorticity coupling (SVC), expressed by the Hamiltonian $H_{\\mathrm{SVC}} = -\\frac{1}{2}\\mathbf{S}\\cdot\\boldsymbol{\\omega}$, where $\\boldsymbol{\\omega} = \\nabla \\times \\mathbf{v}$ is the vorticity of the electron velocity field. In a film with a conductivity gradient along its thickness, the electron drift velocity varies across the film, creating a nonzero vorticity and an effective magnetic field aligned with it; the gradient of that field exerts a spin-dependent force that produces a spin current flowing out of the plane into the adjacent ferromagnet. The experiment realizes this by inserting an ultrathin Ti/W bilayer (thickness $t_i$ from 0 to 2.0 nm) at the W(10 nm)/Ti(10 nm) interface so that atomic diffusion forms a compositionally graded interface, and ST-FMR reads out the damping-like torque efficiency $\\xi_{\\mathrm{DL}}$ from the amplitude ratio of Lorentzian to anti-Lorentzian components. The same machinery explains why reversing the stacking order reverses the torque sign: the Ti–W resistivity changes monotonically with composition, so the vorticity direction is set by whether low-conductivity Ti or high-conductivity W sits next to the ferromagnet.","core_discovery":"The central claim is that the polarity of the spin current generated in a compositionally graded layer is dictated by the direction of the electric-current vorticity. In the W/Ti/Ni-Cu series, the W/Ti CGI creates a region where the electrical conductivity decreases toward the Ni-Cu layer, producing a vorticity whose sign is opposite to that in the previously studied Ti/W/Ni-Cu series, where the conductivity increases toward the Ni-Cu layer. The measured damping-like torque efficiency decreases with $t_i$ up to 1.5 nm ($\\Delta\\xi_{\\mathrm{DL}} \\approx -0.008$), while the earlier Ti/W series increased ($\\Delta\\xi_{\\mathrm{DL}} \\approx +0.011$ at 0.5 nm). The authors argue that the $t_i = 0$ value (+0.031) matches a Ti/Ni-Cu reference, so the bulk W spin Hall and Ti orbital Hall contributions approximately cancel, and that the sharpest W/Ti interface and the Ti/Ni-Cu interface produce no $t_i$-dependent torque, ruling out Rashba–Edelstein and orbital Rashba–Edelstein effects. They conclude that spin-vorticity coupling is the most likely source and that its sign follows the vorticity orientation.","pith_inferences":["The authors leave implicit that the same vorticity-based conversion should operate in any graded conductor carrying a steady current, so SVC could become a general route to spin-current generation in light-element and even non-metallic systems.","The two series are not directly matched: the W/Ti effect peaks at $t_i = 1.5$ nm while the Ti/W effect peaks at 0.5 nm, and the deposition rates differed, so a matched-gradient experiment is needed to test whether the sign reversal is quantitatively symmetric.","A decisive control would be to measure the per-sample saturation magnetization and the W/Ti current split and recompute the torque efficiency; only if the $\\Delta\\xi_{\\mathrm{DL}}(t_i)$ curve survives those corrections can the polarity claim be isolated from hidden $t_i$-dependent backgrounds."],"forward_implications":["If SVC is the source, then any material pair with a conductivity contrast can act as a spin-current source, with the sign chosen by stacking order rather than by the sign of bulk spin-orbit coupling.","The torque magnitude can be engineered by the width of the gradient: in the W/Ti series the effect grows to $t_i = 1.5$ nm and weakens at 2.0 nm, while in the Ti/W series it peaks at 0.5 nm.","Since the sharpest W/Ti interface and the Ti/Ni-Cu interface produce no $t_i$-dependent torque, the Rashba–Edelstein and orbital Rashba–Edelstein mechanisms are not the controlling ones in this system.","Materials with weak spin-orbit coupling could still give large torques if the CGI is built from a pair with a large conductivity contrast, as the Si/Al example with $\\xi_{\\mathrm{DL}} = 0.67$ indicates."],"supporting_citations":[{"why":"Prior Ti/W/Ni-Cu gradient series that supplies the positive $\\Delta\\xi_{\\mathrm{DL}}$ baseline and the monotonic resistivity-versus-composition relation used to fix the vorticity sign.","marker":"[22]"},{"why":"Introduces the SVC Hamiltonian $H = -\\frac{1}{2}\\mathbf{S}\\cdot\\boldsymbol{\\omega}$ from which the spin-current generation mechanism is taken.","marker":"[10]"},{"why":"Demonstrates SVC-mediated spin torque in surface-oxidized Cu, the precedent on which the CGI interpretation is modeled.","marker":"[2]"},{"why":"Reports the Si/Al compositionally graded interface with $\\xi_{\\mathrm{DL}} = 0.67$, the benchmark for large conductivity contrast and the claim that heavy metals are unnecessary.","marker":"[3]"},{"why":"Provides the ST-FMR protocol used to measure the damping-like torque efficiency.","marker":"[26]"},{"why":"Gives the efficiency formula (Eq. 3) relating the Lorentzian and anti-Lorentzian amplitudes to $\\xi_{\\mathrm{DL}}$.","marker":"[30]"},{"why":"Supplies the analysis that Rashba–Edelstein effects mainly produce field-like torque, used to exclude REE and OREE from the $t_i$-dependent signal.","marker":"[33]"}],"fun_headline_variants":["Spin-torque sign flips with inverted composition gradient","Inverting gradient polarity reverses spin-torque sign","Composition gradient direction controls spin torque","Vorticity orientation sets spin-torque polarity","Gradient polarity flips torque without heavy metals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole reading assumes that inserting the ultrathin Ti/W bilayer changes only the CGI spin current, leaving the bulk W spin Hall effect, bulk Ti orbital Hall effect, interface torques, current shunting, and the Ni-Cu magnetic properties unchanged with $t_i$.","fun_headline_variants_meta":{"raw":{"variants":["Spin-torque sign flips with inverted composition gradient","Inverting gradient polarity reverses spin-torque sign","Composition gradient direction controls spin torque","Vorticity orientation sets spin-torque polarity","Gradient polarity flips torque without heavy metals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000749,"raw_usage":{"total_tokens":3346,"prompt_tokens":969,"completion_tokens":2377,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":585,"completion_tokens_details":{"reasoning_tokens":2286}},"tokens_in":585,"tokens_out":2377,"duration_ms":14991,"temperature":1.0,"reasoning_tokens":2286,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T10:32:58.850695+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the saturation magnetization and the sheet conductances of every $t_i$ sample and recompute $\\xi_{\\mathrm{DL}}$ with those per-sample values; if the $\\Delta\\xi_{\\mathrm{DL}}(t_i)$ curve keeps its sign and shape, the CGI-polarity claim stands, whereas a drifting magnetization or current split would show the assignment to SVC is contaminated.","supporting_citations":[{"cited_title":"Nakayama, T","cited_arxiv_id":null,"evidence_quote":"Prior Ti/W/Ni-Cu gradient series that supplies the positive $\\Delta\\xi_{\\mathrm{DL}}$ baseline and the monotonic resistivity-versus-composition relation used to fix the vorticity sign."},{"cited_title":"Matsuo, J","cited_arxiv_id":null,"evidence_quote":"Introduces the SVC Hamiltonian $H = -\\frac{1}{2}\\mathbf{S}\\cdot\\boldsymbol{\\omega}$ from which the spin-current generation mechanism is taken."},{"cited_title":"Okano, M","cited_arxiv_id":null,"evidence_quote":"Demonstrates SVC-mediated spin torque in surface-oxidized Cu, the precedent on which the CGI interpretation is modeled."},{"cited_title":"Horaguchi, C","cited_arxiv_id":null,"evidence_quote":"Reports the Si/Al compositionally graded interface with $\\xi_{\\mathrm{DL}} = 0.67$, the benchmark for large conductivity contrast and the claim that heavy metals are unnecessary."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the ST-FMR protocol used to measure the damping-like torque efficiency."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the efficiency formula (Eq. 3) relating the Lorentzian and anti-Lorentzian amplitudes to $\\xi_{\\mathrm{DL}}$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analysis that Rashba–Edelstein effects mainly produce field-like torque, used to exclude REE and OREE from the $t_i$-dependent signal."}],"review_version":1}