{"id":"27ee1c01-e167-473a-bacb-e59a35151c4d","arxiv_id":"1908.05942","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Randomly removing half-circle obstacles from a touching square lattice reverses the net current of active particles at low disorder, an effect attributed to an imbalance of trapping on curved versus flat obstacle sides.","lead":"Active particles swimming through a grid of half-circle barriers can spontaneously flow backwards when some barriers are randomly removed, even though the barriers are shaped to push them forward. The effect appears only when the half-circles exactly touch, forming traps that catch more forward-moving particles than backward-moving ones.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The current inversion rests on a single L=100 array; without error bars or an L-scaling of Jx, the negative minimum in Fig. 2 could be a finite-size artifact, so the main claim is not yet secure.","rationale":"The reader's weakest assumption is exactly the load-bearing point: the negative current is reported for a single system size with no finite-size scaling or statistical error bars. My reading of the manuscript confirms that Fig. 2 shows only smooth curves with no realization counts, and the only L-dependence shown anywhere is for isolated-obstacle probabilities, not for Jx. The control simulations with D=9, circles, and wedges are useful supporting evidence that the effect is tied to touching half-circle obstacles, but they do not rule out a size-dependent magnitude of the inversion. The analytic model is explicitly stated by the authors to fail below Jmin, so the explanation of the negative minimum is not independently verified. Thus the central claim remains plausible but not fully established. The reader's CONDITIONAL verdict is appropriate; I do not see grounds to reject or to accept unconditionally, so the verdict should remain unchanged pending the proposed finite-size and statistical check.","tokens_in":9856,"tokens_out":5168,"duration_ms":54186,"concrete_test":"Run the same athermal active-disk protocol for L=200 and L=400 with D=10 and φ=0.5, using at least 50 independent disorder realizations per f and time averages long enough to reach steady state, and plot Jx(f) for f=0.05–0.5 with standard errors. If the negative minimum at f≈0.15–0.25 persists with comparable magnitude and does not shift or shrink with L, the inversion is not a finite-size artifact; if it weakens or disappears, the Fig. 2 result is L-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's core observation is the negative Jx minimum at low f (Fig. 2), but the evidence for it is one box size, L=100 with D=10, i.e. a 10x10 array of obstacles, and periodic boundary conditions. The text states averages over time and distinct realizations, but no realization count, error bars, or convergence checks for Jx are reported. The only finite-size test (Fig. 5) concerns the isolated-obstacle probability, not the current itself. Since f is changed in steps of 0.05, the low-f region corresponds to removing only 5–20 out of 100 obstacles; percolation and current can vary strongly from realization to realization. For L=100, the persistence length vo/η=1000 exceeds the box size by a factor of 10, so periodic images may bias trapping and the sign of Jx. The analytic model (Eq. 4/A9) is an admitted fit that fails below Jmin, so it does not independently validate the inversion. Therefore the central claim that disorder reverses the current is load-bearing on an unquantified finite-size simulation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies transport of self-propelled disks in a square lattice of touching half-circle obstacles, with disorder introduced by randomly deleting a fraction f of obstacles. Using Brownian dynamics simulations at one box size (L=100, D=10, so a 10×10 array of obstacles), the authors find that for packing fractions φ=0.1–0.9 the x-component of the mean current is negative for low/intermediate f and positive for larger f, i.e., an inversion relative to the usual easy-flow direction; no inversion occurs for D=9. They propose a trap-imbalance mechanism—positive-moving particles are captured in the spaces between touching obstacles, while negative-moving particles are impeded at the flat sides—and derive Eq. (4) from a mean-field counting of traps and particle layers. The calculation is compared with two simulation curves above the negative minimum using per-curve fitted values of v′, c+, and c−. Additional simulations with circular obstacles and wedges are used as controls.","tokens_in":10136,"tokens_out":5134,"duration_ms":50467,"significance":"The claimed inversion is potentially significant: if confirmed, it shows that quenched disorder in an asymmetric obstacle lattice can reverse, rather than merely reduce, the rectified current of active particles, with implications for microfluidic sorting and active-matter control. The D=9 control and the wedge comparison are well-designed falsifiable checks that support the proposed mechanism. The manuscript is also honest about the model's limitations, explicitly stating that Eq. (4) fails below Jmin. However, the central numerical result currently rests on a single array size with no statistical or finite-size characterization, which is the main barrier to accepting the claim.","major_comments":[{"comment":"The central observation Jx<0 at low/intermediate f is reported for a single system size, L=100 with D=10 (a 10×10 array of obstacles), and no realization count, standard deviation, or finite-size analysis of Jx is provided. Since the low-f region corresponds to removing only 5–20 of 100 obstacles and the persistence length v0/η=1000 is ten times the box size, the negative minimum could in principle be a finite-size or periodic-image artifact; the finite-size check in Fig. 5 concerns the isolated-obstacle probability, not the current. Please report error bars and the evolution of Jx(f) for at least L=150, 200, and 300, showing that the negative minimum persists and converges.","section":"Sec. II and Fig. 2"},{"comment":"The analytical result is not an independent validation of the inversion because c+, c−, and v′ are fitted per curve to the very simulation data with which they are compared, and the text acknowledges that Eq. (4) fails below Jmin, which is precisely the low-f region containing the inversion. Please state the number of fitted parameters explicitly, restrict the claimed agreement to f above f(Jmin), and test the underlying assumptions (statistical independence of n± and v±, equality of the mean positive and negative velocities) by measuring these quantities separately in the simulations.","section":"Eq. (4) and Fig. 2"},{"comment":"The derivation of ⟨T+⟩ relies on the ad hoc counting rule that 'each removed obstacle corresponds to one removed trap,' which is not derived from the cluster statistics of the random lattice, and it is combined with an isolated-obstacle probability fitted in Fig. 5 without error bars or residuals. This makes the quantitative agreement above Jmin less compelling than stated; presenting the fit parameters and their uncertainties, or deriving ⟨T+⟩ from the actual cluster-size distribution, would place the model on firmer ground.","section":"Appendix, Eq. (A6)"}],"minor_comments":[{"comment":"The line 'Jx = ⟨J+⟩⟨J−⟩' appears to be a typo for 'Jx = ⟨J+⟩ − ⟨J−⟩'; please correct it and check the surrounding notation.","section":"Appendix, before Eq. (A2)"},{"comment":"The number of particles N used for each φ and the number of realizations included in the average in Eq. (2) are not stated; please provide these values together with the length of the time averaging window.","section":"Sec. II"},{"comment":"The notation is inconsistent between v′ and ⟨v⟩; please clarify that v′ is the f-independent prefactor used in the linear ansatz ⟨v⟩=v′f, and define the exact relation used to produce the curves in Fig. 2.","section":"Eqs. (4) and (A9)"},{"comment":"The fitting functions 1.0014f^2.002 and 1.04f^2.02 are plausible, but the text should report the fit range, the number of data points, and a goodness-of-fit measure for each curve.","section":"Fig. 5"},{"comment":"Minor typographical issues include 'disorderd' in the arXiv title and 'The arrange of the disks' in Sec. II; the color-coded dashed lines in Fig. 2 should also be distinguishable by line style or labels for accessibility.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The required revisions are technical and feasible: the central idea is worth another round, and the main gap is the lack of statistical and finite-size support for the headline current inversion. I saw no inappropriate citation behavior or novelty disclosure problem; the contrast with Refs. [13,14] and [22] is appropriately framed. The finite-size requirement should be enforced during revision, because the negative current may not survive larger systems."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper reports a new simulation result — randomly removing a fraction of half-circle obstacles from a touching square lattice reverses the direction of the mean active current at low disorder, with no external field. I think the effect is probably real, but the evidence doesn't yet rule out a finite-size artifact, so it needs another round.\n\nWhat's actually new: previous work has rectification in ordered lattices (Potiguar et al. 2014) and density-induced inversion in periodic substrates (McDermott et al. 2016), but not disorder-induced inversion by random obstacle removal in the touching half-circle geometry. The mechanism — more particles trapped in the contact regions between obstacles than on the flat sides — is physically sensible and is supported by control runs: D=9 half-circles show no inversion, circles show no net current, and wedges show the same inversion with reduced magnitude. That's a decent piece of evidence.\n\nThe soft spots are statistical. There are no error bars, no number of realizations, and no finite-size scaling of Jx — all of it is one box, L=100 (a 10x10 array). The bare persistence length is 1000, ten times the box size, so periodic images could bias the trapping statistics. That doesn't make the effect spurious, but it does mean the central negative minimum in Fig. 2 is not yet secure. The analytical model in Eq. (4) is an admitted fit: c+, c-, and v are chosen per curve, it reproduces only two curves, and it fails below Jmin — which is exactly the region where the new effect lives. The paper is upfront about this, and the qualitative mechanism is plausible, but the model doesn't independently corroborate the inversion.\n\nWho is this for? People who care about active matter rectification and transport in disordered environments. It deserves a serious referee, but the referee should ask for error bars, realization counts, and an L-scaling study of Jx, plus a genuinely predictive version of the trap-imbalance calculation (or at least a fit with shared parameters across curves). With those, this could be a solid finite-size-verified paper. As it stands, it's a reasonable preprint but not a definitive one.\n\nI'd send it to peer review with a request for major revision along those lines.","headline":"Plausible new simulation result — disorder-induced current inversion in a touching half-circle lattice — but the absence of error bars and finite-size checks means the central effect is not yet secure.","tokens_in":10649,"tokens_out":4470,"would_cite":true,"duration_ms":42254,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["87.80.Fe","47.63.Gd","87.15.hj","05.40.-a"],"model":"deepseek-v4-flash","headline":"This paper claims that randomly removing a fraction of touching half-circle obstacles from a square lattice makes active particles drift opposite the easy-flow direction at low disorder, with no external field.","keywords":["active matter","self-propelled particles","current inversion","disordered lattice","asymmetric obstacles","trapping","rectification","nonequilibrium transport"],"falsifier":"Repeat the simulations with the same obstacle diameter D=10 and periodic boundary conditions in larger boxes, say L=200 and L=400, at a mid-range area fraction such as φ=0.5 and fractions f=0.15, 0.25, 0.35; if the negative minimum of Jx shrinks toward zero, changes sign, or scales with 1/L, the inversion is a finite-size effect rather than a spontaneous reversal.","tokens_in":9627,"feed_emoji":"🧭","tokens_out":7570,"duration_ms":66553,"temperature":0.7,"pith_summary":"This paper asks whether translational disorder alone can change the direction in which self-propelled particles move through a lattice of asymmetric obstacles. It claims the answer is yes: when half-circle obstacles of diameter equal to the lattice spacing touch one another and a fraction f of them is randomly removed, the net particle current points opposite the easy-flow direction for low and intermediate f, then reverts to the easy direction at high f. The reason, the authors argue, is that spaces between touching obstacles trap more particles moving in the +x direction than the flat sides of obstacles trap particles moving in the -x direction, so the untrapped population has a net drift to the negative side. The paper backs this with simulations and a closed-form expression for the current that reproduces the numerical curves above the current minimum. If right, it gives a purely geometric control knob—obstacle removal fraction—for reversing active transport without external fields.","feed_headline":"Randomly removing obstacles flips active-particle current","feed_subtitle":"Deleting a fraction of touching half-circle obstacles makes particles drift opposite the easy-flow direction—no external field needed.","key_machinery":"The load-bearing object is the trap-imbalance formula for the x-current, $J_x = v'(L/D)^2(1-f)f[(\\pi D+d)/(2d)(f^2-1)\\langle c_+\\rangle + (D/d)\\langle c_-\\rangle]$, where $\\langle c_+\\rangle$ and $\\langle c_-\\rangle$ are the mean numbers of particle layers around the curved and flat sides of an obstacle. The traps are the two capture regions: for a particle moving in the +x direction, a trap is the space between two touching obstacles, reached by sliding along the curved side; for a particle moving in the -x direction, a trap is the flat side of an obstacle. The counting uses the probability that a remaining obstacle is isolated, which the simulations find scales as $f^2$, to estimate how many curved-side traps survive removal, while every remaining obstacle keeps its flat-side trap. The argument then assumes positive and negative travelers have equal mean speeds, so the sign of $J_x$ is decided purely by whether more particles are held in curved-side traps ($\\langle c_+\\rangle > \\langle c_-\\rangle$) than in flat-side traps, which is the condition for the inversion.","core_discovery":"On the paper's own terms, the central discovery is a spontaneous current inversion in a disordered lattice of asymmetric obstacles. For half-circles of diameter D equal to the unit-cell length of a square lattice, arranged so the easy-flow direction is +x, randomly removing a fraction f of the obstacles produces Jx < 0 for low and intermediate f, with the negative minimum deepening as the area fraction φ decreases; for sufficiently large f the current becomes positive and follows the easy-flow direction, with a crossover f*(φ) that decreases as φ increases. The inversion disappears when the obstacle diameter is smaller than the cell (D=9), and circular obstacles produce no current at all, while wedge-shaped obstacles produce a weaker inversion. The mechanism asserted is an imbalance in trapping: particles moving in +x are trapped between the curved sides of touching neighbors, whereas particles moving in -x are trapped at flat sides, and because more particles are held in the curved-side traps the free particles preferentially drift in the -x direction. The paper derives $J_x = v'(L/D)^2(1-f)f[(\\pi D+d)/(2d)(f^2-1)\\langle c_+\\rangle + (D/d)\\langle c_-\\rangle]$ from this trap-counting argument and shows it reproduces the simulated curves above the negative current minimum when $\\langle c_+\\rangle > \\langle c_-\\rangle$.","pith_inferences":["A testable extension the paper does not run: sweeping the obstacle diameter continuously between D=9 and D=10 should reveal a threshold diameter at which the negative current first appears, pinning the trap-size condition quantitatively.","The empirical $f^2$ scaling for isolated-obstacle probability could be derived from site-percolation statistics on the square lattice; if exact, it would turn Eq. (4) into a parameter-free prediction for the f-dependence of Jx.","Particles with different angular noise η should be trapped at different rates, so a binary mixture in the same disordered lattice is likely to develop opposite or unequal net currents for the two species, giving a noise-based sorting mechanism—the paper notes this possibility but does not test it.","If the negative-current regime is tied to percolating paths of touching obstacles, then the onset near f≈0.10 (one removed column) may be connected to a directed-percolation threshold; checking whether f*(φ) follows a percolation scaling would sharpen the mechanism."],"forward_implications":["At fixed area fraction, tuning the removal fraction f from low to high values sweeps the net current from negative through zero to positive, so a single quenched disorder parameter controls transport direction.","The crossover fraction f*(φ) moves to smaller f as area fraction increases, meaning denser suspensions reach the easy-flow regime with less disorder.","The sign and magnitude of the inverted current depend on obstacle shape: touching half-circles give the strongest negative minimum, wedges give a weaker one, circles give none, and half-circles that do not touch (D=9) give no inversion at all.","Because the negative current only appears when neighboring obstacles touch, the effect is tied to the existence of curved-side traps, allowing shape and spacing to serve as design parameters.","The derived expression for Jx, when its assumptions hold, predicts the current from obstacle geometry, the layer numbers $\\langle c_\\pm\\rangle$, and the self-propulsion speed, so it can be used to estimate transport in microfluidic obstacle arrays."],"supporting_citations":[{"why":"Supplies the active Langevin model (angular Brownian motion with self-propulsion and rotational noise) used in all simulations.","marker":"[8]"},{"why":"Establishes that particles slide along the curved side of half-circle obstacles and defines the easy-flow direction; the baseline result this paper's inversion is compared against.","marker":"[14]"},{"why":"The funnel-lattice rectification result whose positive easy-flow current the disorder-induced inversion is explicitly contrasted with.","marker":"[13]"},{"why":"Provides the prior observation of current inversion in periodic substrates, which this paper distinguishes from its own disorder-driven inversion.","marker":"[22]"},{"why":"Shows accumulation of active particles between two bodies, supporting the claim that the spaces between touching obstacles act as traps.","marker":"[24]"},{"why":"Reports related trapping of active matter against obstacles, reinforcing the flat-side and curved-side capture picture.","marker":"[25]"},{"why":"Demonstrates trapping of active particles by funnel-shaped obstacles as a function of aperture angle, used when interpreting the weaker inversion for 90-degree wedges.","marker":"[27]"},{"why":"The review documenting that active particles slide along surfaces, the microscopic ingredient behind the trap-capture argument.","marker":"[5]"}],"fun_headline_variants":["Random removal flips active particle current","Disordered lattice inverts active matter transport","Missing obstacles reverse active swimmer drift","Touching half-circles flip active particle flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central claim rests on the negative current at low and intermediate disorder being a real property of the disordered lattice rather than a finite-size artifact: the simulations use one box side L=100 (a 10-by-10 array of obstacles), and finite-size checks are reported only for the isolated-obstacle probability, not for the current Jx itself.","fun_headline_variants_meta":{"raw":{"variants":["Random removal flips active particle current","Disordered lattice inverts active matter transport","Missing obstacles reverse active swimmer drift","Touching half-circles flip active particle flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000138,"raw_usage":{"total_tokens":1161,"prompt_tokens":963,"completion_tokens":198,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":142}},"tokens_in":579,"tokens_out":198,"duration_ms":3068,"temperature":1.0,"reasoning_tokens":142,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:59:55.172407+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the simulations with the same obstacle diameter D=10 and periodic boundary conditions in larger boxes, say L=200 and L=400, at a mid-range area fraction such as φ=0.5 and fractions f=0.15, 0.25, 0.35; if the negative minimum of Jx shrinks toward zero, changes sign, or scales with 1/L, the inversion is a finite-size effect rather than a spontaneous reversal.","supporting_citations":[{"cited_title":"In fact, particles sliding on the curved side of the obstacle have their motion rectified along the +x-direction","cited_arxiv_id":null,"evidence_quote":"Establishes that particles slide along the curved side of half-circle obstacles and defines the easy-flow direction; the baseline result this paper's inversion is compared against."},{"cited_title":"Reichhardt, and C","cited_arxiv_id":null,"evidence_quote":"The funnel-lattice rectification result whose positive easy-flow current the disorder-induced inversion is explicitly contrasted with."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the prior observation of current inversion in periodic substrates, which this paper distinguishes from its own disorder-driven inversion."},{"cited_title":"McDermott, C","cited_arxiv_id":null,"evidence_quote":"Shows accumulation of active particles between two bodies, supporting the claim that the spaces between touching obstacles act as traps."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports related trapping of active matter against obstacles, reinforcing the flat-side and curved-side capture picture."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates trapping of active particles by funnel-shaped obstacles as a function of aperture angle, used when interpreting the weaker inversion for 90-degree wedges."},{"cited_title":"Bechinger, R","cited_arxiv_id":null,"evidence_quote":"The review documenting that active particles slide along surfaces, the microscopic ingredient behind the trap-capture argument."}],"review_version":1}