An edge signal that standard MFM models miss
Magnetic force microscopy (MFM) maps a sample by scanning a magnetic tip and recording the phase shift of a cantilever oscillating above the surface. On roughly 1 µm diameter Bi2Te3 plates, the paper reports a phase shift of about +3° at the plate edge relative to the HOPG substrate, while the plate interior is attractive at about −0.5° from free oscillation. Reproducibility of the phase determination is about 0.1°.
The puzzle is not that an edge produces contrast, but that the contrast is far larger than published electrostatic and magnetic models predict. Applying those models to the images yields edge currents the authors call unrealistic, and most such models predict a phase shift scaling as 1/z², which the data do not follow. The same strong edge contrast appears for Bi2Se3 and Sb2Te3 plates, but not for MoS2 or Au plates of similar dimensions, which the authors treat as evidence that the enhancement is tied to the topological material rather than to geometry alone.
Two magnetic sources inside the edge channel
The authors propose that the edge contrast comes from currents in the conductive edge channels, fed by two mechanisms. The first is a pseudo-Faraday effect: as the tip oscillates vertically, its moving magnetic field induces a current around the edge loop, much as a changing flux drives a current in a wire. The second is magnetic scattering, in which the tip's localized dipole field disrupts transport in the edge channel and creates an asymmetry between the two counter-propagating directions.
This distinction matters because the two mechanisms respond differently to imaging settings. A Faraday-induced current should scale with oscillation amplitude and tip velocity, while a scattering-related persistent current should depend on how close the tip is to the edge. That difference is what makes multi-parameter imaging, in which amplitude and lift height are varied independently, informative rather than redundant.
Why a single height curve cannot fit the data
A simple model treats the edge as a straight current-carrying wire seen by a magnetic dipole tip, which predicts that the edge phase shift should vary linearly with 1/z³, with a slope set by the current. The measured data instead show two distinct slopes in different lift-height ranges. Fitting the simple model gives about 79 nA for lift heights between 20 and 40 nm, but about 1.04 µA for lift heights between 40 and 80 nm, a shift of two orders of magnitude over a modest height change.
The authors note that the larger apparent current occurs when the tip is farther from the substrate, which they regard as unphysical for simple classical induction. Their reading is that the single-current picture is incomplete and that an additional induced component, the Faraday term, must be added to the persistent-current term. In the limit of vanishing amplitude or frequency, only the persistent-current contribution should remain.
Separating persistent current from Faraday current
To isolate the persistent-current contribution, the authors plot edge phase shift against oscillation amplitude at three lift heights: 100 nm, 150 nm, and 200 nm. Extrapolating each curve to zero amplitude gives a y-intercept that should represent the magnetic phase shift of the persistent current alone, because the Faraday term vanishes as the tip motion goes to zero.
At a lift height of 200 nm, where the authors assume scattering is weakest, this extrapolation yields a persistent current of about 802 nA. They compare that value with theoretical predictions for a flux ratio of about 0.2 of a flux quantum in HgTe, and describe the agreement as within an order of magnitude. This is a consistency check, not a direct measurement of a topological current: the number depends on the analytical model, the assumed tip dipole, and the extrapolation procedure.
What the capacitive background contributes
The tip, substrate, and plate also form a capacitive system, and this electrostatic attraction dominates over the plate surface. The authors model the plate as a dielectric disk with a conductive edge loop and approximate the capacitive phase contrast between HOPG and the plate for lift heights below the plate thickness. That term falls off as 1/z² and also contains a contribution proportional to the square of the oscillation amplitude.
This background is why the edge signal has to be separated carefully from the surface signal. The paper reports that the surface phase shift relative to HOPG approaches about 0.65° as the lift height goes to zero, implying a surface phase of roughly −1.35°, and that the magnetic term expected for a topological surface state is not obvious in the data. The authors suggest this may mean the thicker layered crystals behave as a 2D system, but they present it as an inference rather than a demonstrated result.
Scan direction reveals a velocity-dependent asymmetry
The edge contrast is not uniform around the plate perimeter. The authors observe that the asymmetry appears only along the scan direction and scales with the tip's lateral velocity, and that reversing the scan direction flips which side is stronger. They interpret this as the lateral motion of the tip generating its own time-varying flux through the edge loop, producing opposite currents as the tip approaches and retreats.
This observation supports the Faraday interpretation, because a purely static magnetic or electrostatic interaction would not reverse with scan direction. The paper also notes that at slow scan speeds, averaging the phase shift around the circumference allows the asymmetry to cancel, which is how the averaged edge curves used in the amplitude and height analysis are obtained.
Where the interpretation holds and where it stops
The model is deliberately simple. The authors use a dipole approximation for the tip field and a straight-wire approximation for the edge, and they estimate that assumptions such as the tip dipole angle and the flux-area prefactor could shift the predicted values by a factor of roughly two to three. A multipole expansion suggests second-order corrections are more substantial at lift heights above 35 nm than near 100 nm, and these estimates are folded into the uncertainty bars. The feedback system that adjusts the cantilever position as amplitude changes introduces a further error, estimated at under 10% in phase for small amplitude changes.
The measurements were made in ambient laboratory conditions at about 25°C and 50% relative humidity, on plates 3 to 30 nm thick and 1 to 2 µm in diameter, with HOPG, mica, and thermally grown SiO2/Si as substrates. The authors also describe a calibration issue with the JEOL 5200 instrument, whose software unit conversion they correct by dividing the output by 57.3 to obtain degrees. The central claim is therefore a mechanism proposal supported by order-of-magnitude consistency, not a quantitative proof of topological edge currents.
Why the comparison samples carry the argument
The strongest discriminator in the paper is not the absolute size of the edge phase shift but the contrast between material classes. Bi2Te3, Bi2Se3, and Sb2Te3 plates all show strong edge contrast with the same overall height dependence, while MoS2 and Au plates of similar dimensions do not show the enhanced contrast. MoS2 does show some edge contrast, as established models would predict, but the magnitude seen in the chalcogenide topological insulators is significantly larger.
That comparison is what makes the proposed Faraday-plus-scattering mechanism specific to conductive edge channels rather than a generic tip-sample artifact. For readers evaluating MFM as a probe of topological circuit elements, the practical priority follows from the same evidence: vary lift height and oscillation amplitude together, check whether the edge asymmetry reverses with scan direction, and compare against a topologically trivial plate before attributing an edge signal to topological states.
Sources used in this answer
Magnetic Force Imaging of 2D Topological Insulators
Ambient MFM phase images of Bi2Te3 and Bi2Se3 plates show anomalously large edge contrast that standard contrast models cannot explain; the authors attribute it to a pseudo-Faraday current induced by tip dither plus magnetic scattering from the tip's localized field in the conductive edge channels.