{"id":"ba73cc3e-c3c5-4e8b-80d4-f2109f216804","arxiv_id":"1908.02466","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Direct nanoscale imaging shows quantum Hall edge states in graphene carry paired counterpropagating currents, while the pure magnetic monopole response survives only at singular points of a mixed magnetoelectric phase diagram.","lead":"A SQUID-on-tip microscope acted as both a tiny movable electric charge and an ultra-sensitive magnetic sensor. It imaged electric currents inside graphene's quantum Hall edges and found that every edge state carries two opposite currents that nearly cancel, not the single chiral current usually assumed.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Forward-model inversion is not validated against a no-upstream null model; opposite-sign peaks could be derivative sidelobes.","rationale":"The reader's weakest assumption identifies the Biot-Savart forward model as the main risk, and I agree that the current extraction is the least secure step. My concern sharpens that risk: the published forward-model tests (SM5, Fig. S3) demonstrate that prescribed strip configurations produce the expected Bz' peaks, but they do not test the inverse inference against a null model containing only downstream topological current. Since the derivative kernel itself generates sign-changing sidelobes, the measured opposite-sign features in Fig. 3f could in principle be explained without any upstream current. This is not an accusation of error; the simulations including both currents are compared with the data and appear qualitatively consistent. Rather, the decisive evidence is missing: a residual comparison showing that the no-upstream model cannot fit the data. Because the central qualitative claim actually requires that upstream currents are present, this missing null test is load-bearing. The concern is real but does not overturn the reader's CONDITIONAL verdict; it makes the required condition more specific. The paper has independent support from two simulation approaches and clear sign changes with filling factor, so the likely conclusion is correct, but the current-level evidence should be tightened before the claim is fully accepted.","tokens_in":35550,"tokens_out":9268,"duration_ms":112068,"concrete_test":"Take the line-scan data of Fig. 3f (Device B parameters in SM14) and compute the predicted Bz' using the same COMSOL electrostatic potential and Biot-Savart kernel, but set I_non=0 identically while keeping I_top=sigma_xy E from the same potential. Include the finite SOT vibration amplitude x0 rather than the linearized derivative, and the SOT sensing area. Then compare the residuals at the positions of the red peaks. If the no-upstream model already describes the measured tree pattern within noise, the counterpropagating-current claim is an artifact of the derivative kernel; if the red peaks remain unexplained, the upstream current is required by the data.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the measured Bz'_probe(x)=x0 dBz/dx line profiles across the p-n junction can be read as a map of j_y(x), so that a positive and an adjacent negative peak imply two counterpropagating current strips. This is the load-bearing step. The Biot-Savart kernel is nonlocal: for a single line current at x0, dBz/dx is proportional to (h^2-(x-x0)^2)/((x-x0)^2+h^2)^2, which has a central peak and negative sidelobes at |x-x0|~sqrt(3)h with roughly one-eighth of the main amplitude. SM5 and Fig. S3 validate the forward calculation for prescribed strip arrangements, but no inversion or null test is shown. In particular, the data in Fig. 3f are compared with simulations that already include both I_top and I_non; the paper never asks whether a model with I_non=0 (only downstream topological current) can reproduce the measured red/blue tree pattern through sidelobes and finite-vibration-amplitude distortion. Because the SOT is also a metallic local gate, tip-induced 'wedding-cake' currents could add to the same opposite-sign features, although the simulations include the SOT as a gate. Thus the existence and 'comparable magnitude' of the upstream current is plausible but not yet demonstrated; a quantitative residual test against the no-upstream null model would settle it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports scanning SQUID-on-tip experiments on hBN-encapsulated graphene in the quantum Hall regime, using the same tip as a tunable local electrostatic gate and as a nanoscale magnetometer, with a tuning-fork-induced vibration to record the spatial derivative of the out-of-plane magnetic field, Bz'_ac, interpreted as a local image of the current density j_y. The central claims are: (i) equilibrium quantum Hall edge states carry a pair of counterpropagating currents, a topological downstream current in the incompressible strip and a nontopological upstream current in the adjacent compressible strip, of comparable magnitude; and (ii) the topological mirror-magnetic-monopole response exists only at singular points of the phase diagram, with a nonlinear mixed magnetoelectric effect dominating elsewhere. The evidence combines two independent simulation approaches, semiclassical COMSOL electrostatics with Biot-Savart forward modeling and microscopic tight-binding calculations, together with spatial maps and gate-sweep line cuts across p-n junctions and electrostatically defined edges. The paper also argues that the monopole response is extremely fragile, requiring the entire sample to remain incompressible, and that a single electron charge is sufficient to destroy it in graphene as well as in magnetic topological insulators.","tokens_in":35802,"tokens_out":9657,"duration_ms":102320,"significance":"If established, the results would constitute a substantial advance: the first nanoscale imaging of equilibrium quantum Hall edge currents, the direct observation of the long-predicted nontopological upstream current, and the first phase diagram of the mixed magnetoelectric effect, including the demonstration that the pure mirror-monopole response is confined to isolated singular points. The paper is commendable for combining two independent simulation methods, for explicitly reporting all measurement and simulation parameters in SM14, and for making falsifiable predictions, most notably the absence of the nontopological current for the n=0 Landau level and its presence for higher Landau levels. The monopole-fragility argument for magnetic topological insulators is also a significant and testable conclusion. However, the central edge-current claim currently rests on a forward-model interpretation of Bz'_ac line profiles that has not been tested against the null hypothesis of a single downstream current strip; this is a load-bearing gap that must be closed before the counterpropagating-current claim is fully established.","major_comments":[{"comment":"The central inference that opposite-signed Bz'_ac peaks correspond to two counterpropagating current strips is not tested against the null hypothesis of a single downstream strip. For a line current, dBz/dx is proportional to (x^2 - h^2)/(x^2 + h^2)^2, which has a central peak and opposite-sign sidelobes at |x| approximately sqrt(3)h with about one-eighth of the main amplitude. With h approximately 25-35 nm and vibration amplitude x0 = 35 nm rms in Fig. 3f (SM14), those sidelobes occur at 43-60 nm from the main peak, which is within the range of the reported red-blue pairs. The simulations in Figs. 3g-j already include both I_top and I_non, and Fig. S3 demonstrates resolution only for a prescribed three-strip input; no I_non = 0 forward model is shown. I request a quantitative residual comparison of the Fig. 3f line cuts and the Fig. S7f edge-state scans against a single-strip model, using the exact finite-vibration lock-in response rather than the small-x0 derivative approximation.","section":"SM5, Fig. 3f, Fig. S3"},{"comment":"The quantitative claim that the topological and nontopological currents have 'comparable magnitude' is not supported by an absolute calibration of Bz'_ac to current density; the agreement between data and simulation is described qualitatively as 'in agreement' without residual statistics. Moreover, the p-n junction simulation parameters (top hBN thickness 2 nm, SOT height 32 nm above graphene) differ from the experimental device (Device B, top hBN approximately 11.5 nm, scan height 25 nm above the surface, i.e. approximately 36 nm above graphene), and the electrostatic screening and Biot-Savart kernel both depend on these distances. The paper should provide a calibration chain, a quantitative residual, and a sensitivity analysis over the device parameters and the Landau-level broadening used in the smoothed occupation function before the 'comparable magnitude' assertion is treated as demonstrated.","section":"SM14, Figs. 3f-j and 4i"}],"minor_comments":[{"comment":"The expression Bz'_ac = x0 dBz/dx is an approximation valid for small vibration amplitude; since x0 is comparable to the scan height in several datasets, the paper should state this approximation explicitly and quantify its validity for each measurement.","section":"SM5"},{"comment":"The color scales of the line-cut panels are not shown, which makes it difficult to assess quantitatively the statement that the n=0 Landau level shows 'essentially no' nontopological current; a noise floor or color-bar scale should be provided.","section":"Fig. 3f and Fig. S7f"},{"comment":"The phrase 'directly image' the equilibrium currents should be qualified as 'image under a forward Biot-Savart model,' since the measured quantity is the field derivative, not the current distribution itself.","section":"Abstract and main text"},{"comment":"The derivation of the orbital magnetic moment for Dirac fermions would benefit from an explicit statement of the factor-of-two difference from the parabolic two-dimensional electron gas case and its sign convention, as this ratio is central to the predicted near-cancellation of I_top and I_non shown in Fig. S5.","section":"SM6"}],"recommendation":"major_revision","confidential_remarks":"The paper is strong and likely publishable after the requested null-model test is added. The missing I_non = 0 comparison in Fig. 3f is the only substantive obstacle; I do not see grounds for rejection, but the counterpropagating-current claim should not be accepted without that test."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the headline: this is the first direct imaging of equilibrium currents inside individual quantum Hall edge states, and the evidence for a counterpropagating upstream current is strong enough that I'd bet on it, even though one important control is missing.\n\nWhat's genuinely new: the imaging itself. The SQUID-on-tip with tuning-fork modulation gives Bz' maps that are directly proportional to j_y for narrow strips, and the forward-model simulations in SM5 show the method works for model distributions. The observation of the upstream current in compressible strips—the Geller-Vignale magnetization current—is a first, and the way it flips sign between compressible and incompressible bands as a function of filling (Fig. 2i) is exactly what the theory predicts. The zero-LL exception, where that current is absent, is a satisfying internal consistency check. The mixed magnetoelectric phase diagram in Fig. 4 is also new and the quantum and semiclassical calculations agree well.\n\nThe soft spots are real but fixable. The stress test is right that the paper never runs the null model with I_non = 0. For a single line current, Bz' has sidelobes at roughly one-eighth the main amplitude, so a single downstream strip could in principle fake a counterpropagating partner. But the observed opposite-sign peaks are comparable in magnitude to the main peaks, and the sign-alternating bands in Fig. 2i extend across the whole filling sweep, including regions where the derivative of a single strip would not produce alternating sign at integer and half-integer fillings. Still, the authors should show that a simulation with only I_top fails to reproduce the tree pattern in Fig. 3f. That would close the gap. Second, the absolute current scale is not calibrated; the ratio of the two currents is given but not the absolute amplitude, and no error bars appear on the extracted currents. A careful comparison of measured and simulated line profiles, with residuals, would go a long way. Third, the forward model assumes strip-like current distributions; the data are compared to simulations, not inverted, which is fine, but the paper would be stronger with a statement about the spatial resolution and a test with non-strip-like distributions.\n\nNone of this changes the main conclusion. The physics is consistent with prior theory and the two independent simulation methods agree with the data. The paper deserves a serious referee. If I were the editor, I'd send it out. If I were a referee, I'd ask for the null test and a calibration statement before accepting.","headline":"First imaging of equilibrium QH edge currents is real science; add a null test and calibration before publishing.","tokens_in":36370,"tokens_out":4335,"would_cite":true,"duration_ms":45354,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["73.43.-f","75.85.+t"],"model":"deepseek-v4-flash","headline":"Quantum Hall edge states in graphene carry a pair of counterpropagating equilibrium currents, not a single chiral current, and the pure magnetic-monopole response exists only at isolated tuning points.","keywords":["quantum Hall effect","graphene","magnetoelectric effect","magnetic monopole response","SQUID-on-tip","equilibrium currents","quantum Hall edge states","scanning magnetometry"],"falsifier":"Scan the same electrostatically defined edge at several scan heights (for example 25, 40, and 60 nm) and deconvolve the measured $B_z'(x)$ with the known SQUID-on-tip sensing area: if the inferred $j_y(x)$ reproduces the same strip positions, widths, and currents up to the height-dependent point-spread function, the counterpropagating-current interpretation survives, whereas if the reconstructed pattern changes sign, shifts, or splits with height, the mapping from field derivative to current density is not unique; a second check is the predicted zero of the upstream current at the graphene $n=0$ Landau level in the same junction.","tokens_in":35363,"feed_emoji":"🧲","tokens_out":6428,"duration_ms":66826,"temperature":0.7,"pith_summary":"This paper reports direct nanoscale imaging of equilibrium electric currents in a graphene quantum Hall system, using a SQUID-on-tip that acts simultaneously as a scanning local charge and as a sensitive magnetometer. The central claim is that each quantum Hall edge state carries two counterpropagating currents: a topological downstream current in the incompressible strip and a nontopological upstream current in the adjacent compressible strip, of comparable magnitude, so the commonly assumed single chiral current is only half the story. The paper also claims that the pure mirror-magnetic-monopole response to a point charge occurs only at singular points of the phase diagram when the tip potential is near zero; for almost all parameters the response is a mixed magnetoelectric effect with a diamond-tiled phase diagram. A sympathetic reader would care because equilibrium orbital currents are usually invisible to transport and to probe techniques that measure only potentials, and imaging them locally exposes the microscopic current and energy flow that underpin quantum Hall edge physics.","feed_headline":"Quantum Hall edges carry two counterflowing currents","feed_subtitle":"A SQUID-on-tip images equilibrium currents directly, exposing the hidden upstream partner to every chiral edge current.","key_machinery":"The load-bearing instrument is a SQUID-on-tip: a nanoscale superconducting quantum interference device on a pipette apex, used both as a tunable electric gate and as a magnetometer with roughly 30 nT/Hz$^{1/2}$ sensitivity. A quartz tuning fork oscillates the SQUID-on-tip parallel to the surface with amplitude $x_0$, and the lock-in signal $B_z'(x)=x_0\\,dB_z/dx$ becomes, by the Biot-Savart law for a thin current strip, a peaked function whose height is proportional to the strip's total current, directly mapping $j_y(x)$. The theoretical decomposition is $\\mathbf{j}_{\\mathrm{t}}=\\sigma_{xy}\\mathbf{E}$ in incompressible regions and $\\mathbf{j}_{\\mathrm{nt}}=\\nabla\\times\\mathbf{M}$ with $\\mathbf{M}=n\\boldsymbol{\\mu}$ in compressible regions; both currents are computed self-consistently through electrostatic simulations of potential and density, and quantum-mechanical Dirac-equation simulations reproduce the semiclassical current patterns smeared over a magnetic length.","core_discovery":"The paper's central discovery is that the equilibrium quantum Hall edge state is not a single chiral channel but a current pair. In the incompressible region, where the filling factor is pinned at an integer, the in-plane electric field drives a topological current $\\mathbf{j}_{\\mathrm{t}} = \\sigma_{xy}\\mathbf{E}$ flowing downstream along the edge chirality. In the adjacent compressible region, screening produces a density gradient, and the resulting gradient of the local magnetization $\\mathbf{M} = n\\boldsymbol{\\mu}$ generates a nontopological current $\\mathbf{j}_{\\mathrm{nt}} = \\nabla\\times\\mathbf{M}$ flowing upstream; for graphene Landau levels with $|n|\\ge 1$ this upstream current has magnitude comparable to the downstream one, while for the $n=0$ Landau level it vanishes. The measured images show these pairs directly, and show that the total equilibrium edge current does not grow with added Landau levels but oscillates as pairs of $j_{\\mathrm{t}}$ and $j_{\\mathrm{nt}}$ strips advance toward the edge. The same measurement of the magnetoelectric response to a tip charge yields a diamond-tiled phase diagram of the mixed magnetoelectric effect, with the pure monopole response confined to the vertices.","pith_inferences":["The local balance between $j_{\\mathrm{t}}$ and $j_{\\mathrm{nt}}$ is set by the local electrochemical potential and Landau-level occupation, so the same imaging technique could serve as a noninvasive nanoscale potentiometer or thermometer on quantum Hall edges under bias.","The predicted vanishing of $j_{\\mathrm{nt}}$ at the graphene $n=0$ Landau level is a sharp internal test: if an upstream current is observed there, the magnetization-current mechanism would need revision.","Because the upstream current is invisible to transport and to potential-sensing probes, similar counterpropagating equilibrium currents may be present but unnoticed in other topological systems, including quantum anomalous Hall and fractional quantum Hall edges.","The diamond phase diagram suggests the magnetic monopole is a fine-tuned limit; a natural extension is to map the same diagram in fractional quantum Hall states, where the quantized Hall conductance would change the monopole charge unit."],"forward_implications":["Each equilibrium quantum Hall edge state contains a downstream $j_{\\mathrm{t}}$ strip and an upstream $j_{\\mathrm{nt}}$ strip of comparable magnitude, so the net equilibrium edge current is close to zero and oscillates with filling rather than growing with Landau-level index.","Under an external bias the net downstream current becomes finite while the net upstream current stays zero, but locally both distributions change, offering a route to image nonequilibrium edge reconstruction and energy equilibration.","The pure mirror-magnetic-monopole response occurs only at the vertices of the diamond phase diagram near zero tip potential; elsewhere the response is a mixed magnetoelectric effect classified by two Landau-level quantum numbers.","A single electron charge on the tip, or very weak disorder in magnetically doped topological insulators, is enough to destroy the global incompressible state, making the monopole response extremely fragile.","The $n=0$ Landau level of graphene carries topological current but no nontopological current, in contrast to higher Landau levels, providing a clear internal control for the counterpropagating-current picture."],"supporting_citations":[{"why":"Predicted the counterpropagating currents in the compressible and incompressible regions of a two-dimensional electron gas; the paper's central edge-state claim is the direct imaging of this prediction.","marker":"[4]"},{"why":"Predicted the mirror magnetic monopole response to a point charge via topological surface currents; the paper tests this prediction and confines it to singular points of the phase diagram.","marker":"[2]"},{"why":"Formulated the topological magnetoelectric effect for time-reversal-invariant topological insulators, providing the theoretical basis for the topological current $j_{\\mathrm{t}}$.","marker":"[1]"},{"why":"Introduced the SQUID-on-tip with single-spin sensitivity, the instrument that simultaneously supplies the electric charge and measures the magnetic response.","marker":"[3]"},{"why":"Gave the electrostatics of compressible and incompressible edge strips that underlies the edge-state geometry and the pair-of-currents picture.","marker":"[26]"},{"why":"Analyzed the decay of the topological magnetoelectric effect, supporting the paper's fragility argument that limits the observable monopole response.","marker":"[12]"},{"why":"Provided the quantum Hall wedding-cake potential calculations used for the microscopic quantum simulations of the mixed magnetoelectric effect.","marker":"[31]"}],"fun_headline_variants":["Quantum Hall edges flow both ways","Counterflowing edge currents imaged","Hidden upstream current on every edge","Quantum Hall edge: two currents, not one","Monopole response shows paired edge flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the measured $B_z'(x)=x_0\\,dB_z/dx$ is a faithful local image of the current density $j_y(x)$: that the Biot-Savart forward model, the known sensing area and scan height, and the thin-strip current geometry uniquely convert each peak into one current strip with current proportional to peak height, so that if tip-induced gating, topographic crosstalk, or a non-striplike current distribution corrupts that conversion, the inferred counterpropagating pair could be a deconvolution artifact.","fun_headline_variants_meta":{"raw":{"variants":["Quantum Hall edges flow both ways","Counterflowing edge currents imaged","Hidden upstream current on every edge","Quantum Hall edge: two currents, not one","Monopole response shows paired edge flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00018,"raw_usage":{"total_tokens":1341,"prompt_tokens":1018,"completion_tokens":323,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":261}},"tokens_in":634,"tokens_out":323,"duration_ms":4308,"temperature":1.0,"reasoning_tokens":261,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:43:23.908742+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Scan the same electrostatically defined edge at several scan heights (for example 25, 40, and 60 nm) and deconvolve the measured $B_z'(x)$ with the known SQUID-on-tip sensing area: if the inferred $j_y(x)$ reproduces the same strip positions, widths, and currents up to the height-dependent point-spread function, the counterpropagating-current interpretation survives, whereas if the reconstructed pattern changes sign, shifts, or splits with height, the mapping from field derivative to current density is not unique; a second check is the predicted zero of the upstream current at the graphene $n=0$ Landau level in the same junction.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Predicted the counterpropagating currents in the compressible and incompressible regions of a two-dimensional electron gas; the paper's central edge-state claim is the direct imaging of this prediction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Predicted the mirror magnetic monopole response to a point charge via topological surface currents; the paper tests this prediction and confines it to singular points of the phase diagram."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Formulated the topological magnetoelectric effect for time-reversal-invariant topological insulators, providing the theoretical basis for the topological current $j_{\\mathrm{t}}$."},{"cited_title":"Vasyukov, Y","cited_arxiv_id":null,"evidence_quote":"Introduced the SQUID-on-tip with single-spin sensitivity, the instrument that simultaneously supplies the electric charge and measures the magnetic response."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gave the electrostatics of compressible and incompressible edge strips that underlies the edge-state geometry and the pair-of-currents picture."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Analyzed the decay of the topological magnetoelectric effect, supporting the paper's fragility argument that limits the observable monopole response."},{"cited_title":"Gutiérrez, D","cited_arxiv_id":null,"evidence_quote":"Provided the quantum Hall wedding-cake potential calculations used for the microscopic quantum simulations of the mixed magnetoelectric effect."}],"review_version":1}