{"id":"963f1f4e-4ee5-4e15-b920-e2c97b250b55","arxiv_id":"2501.04655","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Numerical simulations show two previously unreported vortex modes in the four-roll mill, with transitions to simple extensional flow controlled by Reynolds number, power-law index, and roller radius.","lead":"The authors simulate a classic four-roll mill device with a lattice Boltzmann method and find that, at low rotation speeds, the expected simple stretching flow is replaced by two new patterns of four symmetrical vortices. The result matters because four-roll mills are widely used to stretch droplets, cells, and polymer solutions, so knowing which flow pattern appears is important for experiments and device design.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The advertised 'supercritical bifurcation' is not established: the paper shows two steady modes separated by a Reynolds-number threshold, but never tests stability, hysteresis, or branch scaling. This makes the strongest claim a classification rather than a demonstrated result.","rationale":"This is a careful numerical study with validated benchmark cases, a grid-independence analysis, and a plausible new observation of vortex modes in the four-roll mill. The reader's verdict is CONDITIONAL, and my concern supports that conditionality rather than overturning it. The most load-bearing weakness is not primarily the boundary scheme or the convergence of velocity gradients, although resolution certainly matters near the critical point as the authors themselves show for n=0.7-0.9. The more fundamental gap is that the word 'supercritical' is a stability classification, and no stability information is produced anywhere in the paper. The measurements show that the steady flow changes character as Re varies, but a subcritical bifurcation would also produce different steady states on either side of a threshold, and it could even produce hysteresis between two stable branches. Since all solutions are obtained by forward time integration from some initial condition, the solver will converge to whichever attracting steady state lies in the basin of the initial condition; without a reverse sweep or a perturbation analysis, bistability would be missed entirely. The paper's own closing sentence defers stability analysis to future work, which is an explicit admission that the advertised bifurcation type is not actually demonstrated. A concrete continuation and scaling test would settle this. I do not think the paper should be rejected: the mode observations and phase diagrams are valuable and likely reproducible, but the headline claim should be conditional on either supplying the missing stability evidence or softening 'supercritical bifurcation' to 'mode transition.' Since the reader already assigned CONDITIONAL, no adjustment to the verdict is needed.","tokens_in":18871,"tokens_out":3059,"duration_ms":34538,"concrete_test":"Perform bidirectional continuation at fixed n=1.0 and n=1.3 with h=1, delta=2.56, r=9 on both 800x800 and 1600x1600 grids: start from a well-converged simple-extensional state at Re=50, decrease Re in steps of 0.1 using the previous solution as the initial condition, and compare the resulting dux/dx, duy/dy, lx, and ly against the published increasing-Re path. If the two paths coincide with no bistable interval, supercriticality is supported. Then fit each vortex length L near the critical point to A*(Re_c - Re)^beta; a supercritical pitchfork requires beta = 1/2 for the stable branch. If hysteresis appears, or if beta differs significantly from 1/2, the paper should soften its claim from 'supercritical bifurcation' to 'mode transition at a critical Reynolds number.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The title and abstract assert a supercritical bifurcation, but a supercritical bifurcation is a dynamical-systems statement: a stable branch emerges continuously at a critical parameter value, and it is the only stable branch locally. The paper reports forward-converged steady states only, with no reverse continuation, no perturbation decay or growth rates near the transition, and no scaling of any order parameter. Subcritical or hysteretic behavior would be completely invisible to the method. The authors appear to recognize this: the conclusion states that 'Future work should explore the stability characteristics of these newly discovered flow modes.' That admission directly targets the central claim. In addition, the two diagnostics used to locate the transition disagree in exactly the regimes where the bifurcation type matters: for n=1.3, lx vanishes at Re=38 while the other three criteria give Re=36, and the paper then invents the dumbbell-shaped intermediate mode to accommodate the discrepancy. Without a stability or continuation analysis, the correct claim is that the flow exhibits mode transitions at some critical Reynolds numbers, not that these transitions are supercritical bifurcations.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports two-dimensional lattice Boltzmann simulations of generalized Newtonian power-law fluids in a four-roll mill, spanning Re in [1,50], power-law index n in [0.7,1.3], and variations of roller radius r and roller-container gap delta. The TRT-RLB solver is validated against analytic solutions for force-driven power-law channel flow and Taylor-Couette flow, and grid convergence of velocity and shear rate is documented at Re=10. The authors discover two flow configurations in the central region in addition to the classical simple extensional flow: a quadrifoliate vortex mode (four symmetric counter-rotating vortices around the stagnation point, with horizontal extension and vertical compression at O) at low Re, and, for n greater than about 1, a dumbbell-shaped quad-vortex mode in which vortices detach laterally from the stagnation point. Vortex lengths lx and ly decrease monotonically with Re and vanish at critical Reynolds numbers Rec that roughly coincide with sign reversal of the velocity gradients dux/dx and duy/dy at the stagnation point. The paper labels these transitions 'supercritical bifurcations' in the title, abstract, and conclusion, presents phase diagrams in (n, Re), (delta, n), and (r, n) space, and finds that r controls the transition while delta has minimal influence.","tokens_in":19013,"tokens_out":23696,"duration_ms":219573,"significance":"The finding that the central region of the four-roll mill is not always simple extensional flow is genuinely interesting: the device is a standard tool for extensional rheometry and droplet and cell deformation studies, and the identification of a low-Re vortex regime in which the stagnation-point deformation axes are exchanged (du/dx and du/dy both reverse sign relative to the classical state) bears directly on the interpretation of low-speed experiments. The numerical work is careful in several respects that deserve explicit credit: validation against two analytic benchmarks (power-law channel flow for n=0.5, 1.0, 1.5 and Taylor-Couette flow over beta=0.1-0.8), a documented grid-convergence study, and honest reporting of resolution-dependent nonphysical states near the transition for n=0.7-0.9 together with their resolution by mesh refinement to 1600 and 3200 grid points. If the mode classification is accepted, the phase diagrams (Figs. 20, 23, 25) provide a useful engineering map of the flow regimes.","major_comments":[{"comment":"The term 'supercritical bifurcation' is used in the title, abstract, and conclusion, but the analysis establishes only that forward-converged steady states switch between two modes at a Reynolds-number threshold. A supercritical bifurcation in the dynamical-systems sense requires a stability statement: the emerging branch must be the locally unique stable branch, and the transition must be continuous in the order parameter. None of the required evidence is provided: (i) there is no reverse continuation (for example, initializing at Re=50 with the converged extensional solution and decreasing Re through 36 to test whether the quadrifoliate mode reappears, which would rule out hysteresis); (ii) there are no perturbation experiments near Rec measuring growth or decay rates; and (iii) there is no scaling analysis of any order parameter (a pitchfork would give lx, ly proportional to (Rec-Re)^(1/2), a transcritical bifurcation a linear dependence, and so on). A subcritical or transcritical structure would be completely invisible to this forward-only procedure. The manuscript itself concedes this in the closing paragraph of Section V: 'Future work should explore the stability characteristics of these newly discovered flow modes.' Either the missing stability, continuation, and scaling analysis must be added, or the claim must be re-scoped to 'mode transitions at critical Reynolds numbers' throughout, including the title and abstract.","section":"Abstract; Section IV A; Section IV B; Section V"},{"comment":"The critical-Reynolds-number detection is internally inconsistent with the claims built on it. At n=1.3 the reported values are Rec=36 from dux/dx, Rec=36 from duy/dy, Rec=38 from lx, and Rec=36 from ly; the 2-unit discrepancy is then absorbed by introducing the dumbbell-shaped quad-vortex mode as an intermediate state, even though at Re=36 and 37 the velocity gradients already have the simple-extensional signs. At n=0.7 the four criteria give 32, 31, 32, and 31. Yet conclusion item (c) states that in Transition I 'all characteristic parameters ... yield identical critical Reynolds numbers,' and Section IV C 1 repeats the coincidence claim for the delta-sweep. The paper should report a single grid-converged Rec per case and state which criterion defined each phase boundary; with the phase diagram of Fig. 20 claimed at Delta-Re=0.1 resolution near transitions, a 1-2-unit spread among the defining criteria needs a quantitative explanation, not a narrative one. I note that the grid-independence check of the intermediate state (Fig. 19) supports its physical reality; the open question is whether it is a distinct dynamical branch or a continuous deformation, which again requires the stability and continuation analysis of Major Comment 1. A related definitional coupling should also be acknowledged: the quadrifoliate mode is defined by the sign pair (dux/dx greater than 0, duy/dy less than 0), and the transition is then detected from the sign reversal of those same gradients; the vortex-length criterion is the independent evidence and should be given equal prominence in the mode definitions.","section":"Section IV B; Section V, item (c); Section IV C 1"},{"comment":"The grid-convergence study (Table I) is performed at Re=10, far from the transition region, and monitors only ux at a midpoint and |gamma-dot| at O. The transition diagnostic, namely the signs of dux/dx and duy/dy at the stagnation point, is exactly the quantity shown in Figs. 15-16 to be unconverged on the 800x800 grid: at n=0.7 and Re=31, the 800x800 solution exhibits the nonphysical state dux/dx<0 and duy/dy<0 (locally compressible flow), and dux/dx flips sign when the grid is refined to 1600x1600 and 3200x3200. The text states that this 'grid-dependent behavior occurs for 0.7 < n < 0.9,' which is precisely the range feeding into the phase diagrams, yet the paper never specifies which resolution was used for the points of Figs. 20, 23, and 25 or whether the reported Rec values are the resolution-converged ones. Without this policy, the transition points and phase boundaries for shear-thinning fluids are not reproducible. The authors should state the resolution used for each regime of the phase diagrams and demonstrate convergence of Rec itself (not just bulk velocity) at least along the phase boundary.","section":"Section III D; Section IV B; Figs. 15, 16, 20, 23, 25"}],"minor_comments":[{"comment":"The phrase 'supercritical bifurcation modes' in the abstract conflates steady flow modes with transitions between modes; after the re-scoping requested in Major Comment 1, the modes themselves should not be described as bifurcations.","section":"Abstract; Section V"},{"comment":"The TRT-RLB method is cited as Yu et al. (2023), arXiv:2312.10318; if an archival version exists, it should be cited, since the method description is delegated to this unpublished preprint.","section":"Section III A"},{"comment":"The priority claim that this is 'the first successful extension of the TRT-RLB methodology to power-law fluid dynamics' should be checked against the existing regularized-LBM literature for non-Newtonian fluids and softened if related work exists.","section":"Section I"},{"comment":"The construction of the phase diagrams should specify the operational criterion (gradient signs, vortex lengths, or streamline topology) used to assign each point to a mode, so that the boundaries are reproducible.","section":"Figs. 20, 23, 25; Section IV B"},{"comment":"The exclusion of the n=1.3, r=13 and 14 cases 'due to numerical convergence issues' appears only in the running text; it should also be stated in the caption of Fig. 25, together with the number of excluded points.","section":"Section IV C 2; Fig. 25"},{"comment":"The reference 'V ona, M. and Lauga, E.' should read 'Vonna, M. and Lauga, E.'; hyphenation of 'four-roll mill' is also inconsistent in a few places.","section":"References"},{"comment":"The eight vortices visible in Fig. 6a at Re=50 are likely the wall-attached vortices reported by Lagnado and Leal (1990); a sentence connecting the two observations would help situate the new central-region modes relative to the known high-Re behavior.","section":"Section IV A; Fig. 6"}],"recommendation":"major_revision","confidential_remarks":"The main scientific gap is precisely the one the authors acknowledge at the end of Section V: no stability analysis has been performed, yet the title and abstract assert a specific bifurcation type. I would urge the editors to insist either that a continuation- and perturbation-based stability argument (at minimum a reverse-continuation hysteresis test and an order-parameter scaling fit) be added, or that the 'supercritical' terminology be removed from the title. The paper also delegates its methodological novelty claim to a self-cited unpublished preprint (Yu et al. 2023, arXiv:2312.10318). Given that the benchmark validations are solid and the flow-mode observations are likely reproducible, I view the manuscript as salvageable in major revision rather than rejectable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The numerical work is solid: the TRT-RLB extension to power-law fluids with curved boundaries is a legitimate first, and the two benchmark validations against channel flow and Taylor-Couette give me confidence. Grid convergence is shown at 800^2 and finer for the tricky cases. The new observations are plausible and appear genuinely new: the quadrifoliate vortex mode at low Re, and the dumbbell-shaped quad-vortex mode as an intermediate state for shear-thickening fluids, with a phase diagram over Re, n, r, and delta. The roller radius matters a lot; the container gap barely matters. That is a useful map for anyone using four-roll mills or similar extensional devices.\n\nThe soft spot is the word 'supercritical' in the title and throughout. What they actually show is that forward-converged steady states change continuously with Re and vortex lengths shrink to zero at a threshold. They never do backward continuation, perturbation decay/growth, or hysteresis, so subcritical or hysteretic behavior would be invisible. The paper itself concedes this in the conclusion: 'Future work should explore the stability characteristics...' That admission targets the central claim. The mismatch between diagnostics at n=1.3 (lx vanishing at Re=38 while gradients and ly flip at 36) is papered over by naming the dumbbell state, which is a descriptive fix, not evidence for a bifurcation type. The grid-dependent nonphysical states near the critical Re for n=0.7-0.9 are acknowledged, and they refine to 1600x1600 or 3200x3200, which is fine, but it means some phase boundaries sit on less certain footing. A minor circularity: the same velocity gradients define the modes and identify transitions, though the vortex-length diagnostics are independent, so this is not fatal.\n\nWhat needs to happen for the claim to stand: either run a proper linear stability or hysteresis test, or soften the language to 'mode transitions at critical Reynolds numbers' and reserve bifurcation terminology for what is actually established. Reporting code and data would also help, since the phase boundaries are not accompanied by uncertainty estimates.\n\nWho is this for: anyone working on extensional flow devices, LBM for non-Newtonian fluids, or flow-mode classification. Deserves a serious referee. The revision should be conditional on fixing the bifurcation language or adding the missing analysis.","headline":"Solid numerical observations of new vortex modes in the four-roll mill, but the 'supercritical bifurcation' label is asserted, not demonstrated, and should be either backed by stability analysis or softened.","tokens_in":19599,"tokens_out":3350,"would_cite":true,"duration_ms":26515,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76A05","76M28"],"pacs":[],"model":"deepseek-v4-flash","headline":"Four-roll mill, long assumed to produce simple extensional flow, is shown to host two previously unreported vortex modes whose disappearance at a critical Reynolds number is a supercritical bifurcation strongly controlled by roller radius…","keywords":["four-roll mill","power-law fluid","lattice Boltzmann method","supercritical bifurcation","quadrifoliate vortex mode","dumbbell-shaped quad-vortex mode","extensional flow","stagnation point"],"falsifier":"A high-resolution experiment or simulation that tracks tracer particles in a Newtonian four-roll mill with $r=9$, $h=1$, $\\delta\\approx2.56$ should see the four central vortices shrink to zero and the stagnation-point velocity gradients reverse sign around $Re=36$; if the vortices persist at $Re=50$ or disappear below $Re=30$, the claimed bifurcation point is wrong. For $n=1.3$, a grid-converged simulation at $Re=36$ should show the dumbbell-shaped mode with $l_y=0$ but $l_x>0$; absence of that state under resolution refinement would refute Transition II.","tokens_in":18643,"feed_emoji":"🌀","tokens_out":10002,"duration_ms":82200,"temperature":0.7,"pith_summary":"At the center of a four-roll mill, the flow is usually described as pure hyperbolic extension: fluid is pulled in along one axis and pushed out along the other, with zero velocity at the stagnation point. This paper argues that this picture is incomplete at low roller speeds. Using lattice Boltzmann simulations of Newtonian and power-law fluids, the authors find a quadrifoliate vortex mode — four symmetric counter-rotating vortices clustered around the stagnation point — and, for shear-thickening fluids, an intermediate dumbbell-shaped quad-vortex state in which the vortices detach from the center and align along one axis. As the Reynolds number rises, the vortices shrink monotonically and disappear at a critical value, which the paper identifies as a supercritical bifurcation to simple extensional flow. The result matters because four-roll mills are standard tools for stretching droplets, cells, and polymers, and any experiment run below the critical speed is actually probing a different base flow than intended.","feed_headline":"Four-roll mill hides two vortex modes that vanish at a critical speed","feed_subtitle":"Simulations show a smooth supercritical transition to simple extensional flow, with roller radius as the key control.","key_machinery":"The analysis is carried by a two-relaxation-time regularized lattice Boltzmann (TRT-RLB) model augmented with a one-point second-order curved boundary scheme, which the paper presents as the first application of this TRT-RLB variant to power-law fluids in curved geometries. The quantitative backbone is a pair of equivalent bifurcation criteria: the stagnation-point gradients $\\partial_x u_x$ and $\\partial_y u_y$ computed by central differences, and the vortex lengths $l_x$ and $l_y$ measured along the symmetry axes from the point where the axial velocity first vanishes to the center $O$. Agreement between the two criteria marks a direct transition; persistence of $l_x$ after $l_y$ has vanished marks the dumbbell mode. The phase diagram in the $(n, Re)$ plane is built with a step of $\\Delta Re = 0.1$ near the critical lines, and the roller geometry is fixed by the geometric condition $(r+I)^2 = 2(r+h/2)^2$ to minimize container-wall influence.","core_discovery":"The central claim is that the four-roll mill's stagnation flow undergoes a supercritical bifurcation: as $Re$ increases, the vortex lengths $l_x$ and $l_y$ decrease continuously to zero at a critical Reynolds number $Re_c$, and the stagnation-point velocity gradients $\\partial_x u_x$ and $\\partial_y u_y$ reverse sign at the same point. For Newtonian fluids under the reference geometry ($h=1$, $r=9$, $\\delta\\approx 2.56$) that critical value is $Re_c = 36$. The paper further claims that the transition path is controlled by the power-law index $n$: for $0.7 \\le n < 1.0$ the quadrifoliate mode passes directly to simple extensional flow (Transition I), while for $1.0 \\le n \\le 1.3$ it passes through the dumbbell-shaped quad-vortex mode (Transition II), in which $l_y$ vanishes before $l_x$. The authors state that the roller radius $r$ is the dominant geometric parameter — larger $r$ enlarges the vortices, can shift $Re_c$ beyond the studied range, and for $r=4$ suppresses the vortex mode entirely — whereas the roller-container gap $\\delta$ has only a weak effect.","pith_inferences":["Because the same stagnation-point topology appears in cross-slot microchannels, the quadrifoliate and dumbbell modes may also exist there; if so, microfluidic extensional rheometry would need its own phase diagram in $(Re, n)$ before deformation measurements can be trusted.","The transition type could be used as a passive rheological indicator: measuring the critical Reynolds number and whether an intermediate dumbbell state appears gives a coarse read on the power-law index $n$ without any force measurement.","The mechanism behind Transition II may be the local viscosity increase in shear-thickening fluids near the stagnation point, which would alter the balance of extension and compression in the neighborhood of $O$; a targeted simulation with an artificially fixed local viscosity could isolate that effect.","If the bifurcation is robust in 2D, the 3D roller-length effect (short rollers) may break the assumed symmetry and shift $Re_c$, so extending the computation to finite-length rollers is a natural next test."],"forward_implications":["Four-roll mill experiments conducted below $Re_c$ were not in simple extensional flow, so published deformation and breakup data obtained in the quadrifoliate regime may need to be re-interpreted with the vortex circulation taken into account.","The critical Reynolds number can be determined from two independent checks — sign reversal of $\\partial_x u_x$ and $\\partial_y u_y$ at the center, and vanishing of $l_x$ and $l_y$ — giving experimentalists a direct way to verify which flow mode they are in.","Roller radius, not container gap, is the design lever: small radii (around $r=4$) eliminate the vortex mode, while large radii (around $r=14$) can keep the quadrifoliate or dumbbell modes present up to $Re=50$.","Shear-thickening fluids will pass through an extra intermediate state, so rheological measurements or particle trapping in such fluids should avoid the dumbbell-shaped regime or treat it as a distinct flow condition.","The weak dependence on $\\delta$ means compact four-roll mill configurations can be built without significantly shifting the flow-mode boundaries."],"supporting_citations":[{"why":"Introduces the four-roll mill as the device for generating simple extensional flow with a stagnation point, the baseline flow that the paper claims is only one of several modes.","marker":"Taylor (1934)"},{"why":"Establishes the computer-controlled four-roll mill as the standard tool for particle and drop studies; its assumption of stable simple extensional flow is the premise the paper questions at low $Re$.","marker":"Bentley and Leal (1986)"},{"why":"Supplies the geometric condition $(r+I)^2 = 2(r+h/2)^2$ used to set the roller positions and minimize container-gap disturbances to the central flow.","marker":"Andreotti, Douady, and Couder (2001)"},{"why":"Provides the kinematic analysis and optimal-geometry criteria for four-roll mill configurations that the paper's parameter choices build on.","marker":"Higdon (1993)"},{"why":"Reports vortex emergence at high $Re$ in a four-roll mill, the nearest prior observation of non-simple-extensional states that the new modes extend.","marker":"Lagnado and Leal (1990)"},{"why":"Introduces the TRT-RLB model; the paper extends it to power-law fluids with curved boundaries, so this reference supplies the numerical machinery.","marker":"Yu et al. (2023)"},{"why":"Provides the one-point second-order curved boundary scheme used to treat the rotating roller and container walls, which is essential for accurate near-wall flow.","marker":"Tao et al. (2018)"},{"why":"Earlier numerical simulation of four-roll mill flow used as a comparison point for the Newtonian central-flow behavior at $Re=50$.","marker":"Feng and Leal (1997)"}],"fun_headline_variants":["Four-roll mill flow bifurcates at a critical Reynolds number","Two hidden vortex modes vanish in four-roll mill","Roller radius controls four-roll mill vortex disappearance","Power-law index steers four-roll mill vortex vanishing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the TRT-RLB solver and curved-boundary scheme resolve the near-stagnation velocity gradients accurately enough that the sign of $\\partial_x u_x$ and $\\partial_y u_y$ at the center is trustworthy, since the authors show grid-dependent nonphysical states near the critical Reynolds number for $n = 0.7$–$0.9$ that disappear only with refinement to $1600\\times1600$ or $3200\\times3200$ grids.","fun_headline_variants_meta":{"raw":{"variants":["Four-roll mill flow bifurcates at a critical Reynolds number","Two hidden vortex modes vanish in four-roll mill","Roller radius controls four-roll mill vortex disappearance","Power-law index steers four-roll mill vortex vanishing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00082,"raw_usage":{"total_tokens":3675,"prompt_tokens":1120,"completion_tokens":2555,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":736,"completion_tokens_details":{"reasoning_tokens":2500}},"tokens_in":736,"tokens_out":2555,"duration_ms":23135,"temperature":1.0,"reasoning_tokens":2500,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:28:48.538838+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A high-resolution experiment or simulation that tracks tracer particles in a Newtonian four-roll mill with $r=9$, $h=1$, $\\delta\\approx2.56$ should see the four central vortices shrink to zero and the stagnation-point velocity gradients reverse sign around $Re=36$; if the vortices persist at $Re=50$ or disappear below $Re=30$, the claimed bifurcation point is wrong. For $n=1.3$, a grid-converged simulation at $Re=36$ should show the dumbbell-shaped mode with $l_y=0$ but $l_x>0$; absence of that state under resolution refinement would refute Transition II.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the four-roll mill as the device for generating simple extensional flow with a stagnation point, the baseline flow that the paper claims is only one of several modes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the kinematic analysis and optimal-geometry criteria for four-roll mill configurations that the paper's parameter choices build on."}],"review_version":1}