Pith. sign in

REVIEW 3 major objections 8 minor 53 references

Vision transformer based Deep Learning of Topological indicators in Majorana Nanowires

T0 review · 3 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A vision transformer trained on simulated conductance traces can read out the topological state of a Majorana nanowire, including indicators that are not directly measurable.

desk verdict Solid, reproducible ML study of Majorana nanowire diagnostics; the headline P>0.9998 claim outruns the statistics. read the letter →

arxiv 2412.06768 v3 pith:VAVJQLQJ submitted 2024-12-09 cond-mat.mes-hall cond-mat.dis-nn

Jacob R. Taylor , Sankar Das Sarma This is my paper · ORCID
classification cond-mat.mes-hallcond-mat.dis-nn
keywords Majoranazeromodesdisorderednanowiresvisiontransformertopologicalinvariantconductancespectroscopymachinelearningvisibilitylocaldensityofstates
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that a vision transformer trained only on simulated four-terminal conductance measurements can determine the topological state of a Majorana nanowire, including the continuous scattering invariant, the full phase diagram in magnetic field and chemical potential, and two local-density-of-states indicators that experiments cannot measure directly. The motivation is that disorder routinely produces zero-bias conductance peaks that mimic Majorana zero modes, so standard experimental signatures are unreliable. If correct, the method would let experimentalists declare a device topologically non-trivial with an arbitrarily low false-positive probability by tuning a cutoff, and locate the parameter region worth using for braiding. The whole procedure is validated in simulation across low, moderate, and high disorder, with the moderate regime matched to current experimental devices.

What carries the argument

The load-bearing object is a generalized Vision Transformer operating on a three-dimensional input image whose channels are the four conductance traces $G_{LL}$, $G_{RR}$, $G_{RL}$, and $G_{LR}$, and whose axes are bias voltage, magnetic field, and chemical potential. The network applies three-dimensional patching, additive positional encoding, four transformer blocks, and then splits into 100 small multilayer perceptron heads, one per $(B,\mu)$ point, each producing a continuous indicator; a separate network is trained for each indicator. The continuous output lets the user impose a cutoff $C_{\mathrm{cutoff}}$ such that only devices predicted below it are declared topological, converting raw accuracy into tunable confidence. Training data come from numerical solutions of the standard nanowire Bogoliubov–de Gennes Hamiltonian with Gaussian disorder, randomized spin-orbit coupling, and finite-temperature convolution.

What would settle it

Feed the trained network four-terminal conductance traces from a device whose true topology is known independently, for instance by a direct numerical solution of its full Hamiltonian or by braiding measurements, and compare the predicted $TV$ and phase diagram with the exact values; any systematic disagreement would falsify the claim that conductance alone determines topology. A simpler out-of-distribution check is to vary the barrier height or wire length outside the training ranges and see whether the prediction error collapses.

Watch

Extended reading notes

Core claim

The central claim is that the four measured differential conductances of a nanowire, local at each end and nonlocal end-to-end, contain enough information to determine the scattering-matrix topological invariant $TV$, its sign, and its full $(B,\mu)$ phase diagram, even when disorder amplitude, correlation length, and spin-orbit coupling are unknown. The paper further claims that the same conductance inputs determine the LDOS-based operational indicators $I_1$ and $I_2$, which assess whether end-localized Majoranas are usable for fusion and braiding, not merely present. Because $TV$ is predicted continuously rather than as a binary label, a user can set a passing cutoff and make the probability of falsely declaring a trivial device topological arbitrarily small, with reported fidelities above $0.9998$ at the most stringent cutoffs. The authors conclude that conductance data alone should suffice to analyze future experiments.

Load-bearing premise

The whole transfer to experiment rests on the assumption that real nanowire conductance traces are statistically close to the simulated training distribution, meaning the same Hamiltonian form and the same ranges of disorder, spin-orbit coupling, barrier strength, temperature, and wire length.

Editorial extensions

If this is right

  • An experimentalist can use routine conductance measurements to declare a device topological with false-positive probability below $0.0002$, by choosing a sufficiently negative cutoff on the predicted $TV$.
  • The full topological phase diagram in $(B,\mu)$ can be reconstructed from conductance alone, identifying the parameter window where a device should be operated for fusion and braiding experiments.
  • The LDOS-based indicators $I_1$ and $I_2$, which cannot be measured directly, can be inferred from the same conductance data, giving a check on whether end Majoranas are localized and usable.
  • The method replaces arbitrary visual thresholds in protocols like the topological gap protocol with a quantitative, tunable decision rule.
  • Retraining the same architecture with different material parameters would extend the technique to other nanowire platforms without changing the method.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the network's success implies the four conductances carry enough mutual information to fix the scattering invariant; if so, the same architecture should transfer to other topological platforms where transport spectra encode a topological index, such as Josephson junctions or higher-order topological insulators.
  • Editorial inference: because the paper finds prediction errors behave like a Gaussian filtering of the true phase diagram, a low-pass filtered version of the network output could serve as a conservative operating map, with the caveat that smoothing may erase narrow topological patches.
  • Editorial inference: the transfer claim would be directly testable by intentionally training on one material parameter set and testing on conductance traces generated with a moderately different barrier height or wire length; the paper reports no such out-of-distribution test.
  • Editorial inference: since the network never sees disorder parameters, its predictions are only as reliable as the training distribution's coverage of experimental reality; active learning on real devices would be the natural next step.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 8 minor

Summary. The paper trains a Vision Transformer on simulated four-terminal differential conductance traces (GLL, GRR, GRL, GLR) of disordered Majorana nanowires. The network is tasked with predicting the continuous scattering-matrix topological visibility TV, classifying devices as topologically non-trivial, reconstructing the TV phase diagram in (B, μ), and inferring two LDOS-based alternative indicators I1 and I2. The authors report RMSE values between 0.15 and 0.62 across three disorder regimes and claim that by tuning a continuous cutoff the false-positive probability of declaring a device topological can be pushed below 0.0002 (P>0.9998). They conclude that the conductance-only method should be usable for analyzing future experiments.

Significance. If the central claims hold, the method offers a practical way to extract topological information from routine four-terminal conductance measurements, including indicators that are not directly measurable experimentally. The use of a 3D-patched Vision Transformer with a multi-path tree decoder is a nontrivial architectural adaptation, and the simulation pipeline is based on standard KWANT transport calculations with publicly available data-generation code (Ref. [50]). The supervised-learning core is credible: held-out conductance traces are predicted with RMSE values of roughly 0.15-0.5, and the phase diagrams show good qualitative agreement. However, the headline statistical claim about arbitrary confidence is not supported by the finite test set, and the experimental-applicability conclusion goes beyond the evidence provided.

major comments (3)
  1. [§IV.A, Tables II–IV] The claim that the false-positive probability can be pushed below 0.0002 ('P>0.9998') is not statistically supported. The reported entries of 1−FP = 1.0000 correspond to zero observed false positives in a single 10% test split with the cutoff selected on that same test data (Sec. IV.A: 'using our test data, we do not find a single instance'). With roughly 20,000–30,000 total realizations, the test set contains only a few thousand devices, and the number of passing devices at the relevant operating points is a few hundred to at most ~1,500. Zero failures in N trials gives a one-sided 95% upper bound of about 3/N on the failure probability, which is an order of magnitude larger than 0.0002. The abstract's 'up to arbitrary confidence (P>0.9998)' should be replaced by a statement with a confidence interval, or the test set must be enlarged so that zero observed failures actually supports the claimed bound.
  2. [§IV.A, definitions of FN in Tables II–IV] The false-negative metric is misdefined. The table captions state FN = P(TV < Ccutoff | TPred > 0), which conditions on a predicted positive TV, not on the complement of passing (TPred ≥ Ccutoff). The standard false-negative rate for a pass/fail decision is P(TPred ≥ Ccutoff | TV < Ccutoff), i.e., the fraction of truly topological devices that are rejected. The reported '1−FN' values therefore do not quantify the risk that topological devices are discarded, which is exactly the question the text claims to address when discussing whether the method rejects too many devices. Please report a standard confusion matrix with sensitivity and specificity so that the trade-off between false positives and false negatives is unambiguous.
  3. [§V and abstract/conclusion] The conclusion that the technique 'should be usable to analyze future experiments' is an extrapolation beyond the evidence. All tests are performed on simulated conductance traces generated from the same Hamiltonian family, with fixed barrier height (15 mV), fixed length (3 μm), fixed material parameters, and with disorder and spin-orbit coupling drawn from the training ranges. There is no out-of-distribution test, no variation of barrier height or device length, and no experimental data. The network may be learning the specific simulation manifold, and no evidence is given that real device traces lie on that manifold. The claims should be restricted to the simulated parameter ranges, and the manuscript should explicitly state that transfer to different geometries or barrier settings requires retraining or an explicit domain-adaptation test.
minor comments (8)
  1. [Table III] In the full-regime table, the row for Ccutoff = −0.9 lists P(Passing) = 0.0010, which is inconsistent with the gradual progression from 0.3077 at Ccutoff = −0.8 and is never discussed in the text. Please check whether this is a typo and ensure the table matches the narrative.
  2. [§II] The simulated conductance traces are repeatedly called 'measurements,' which can blur the distinction between simulated training data and experimental data. Consider reserving 'measurements' for actual experimental data and using 'simulated conductance traces' for the KWANT outputs.
  3. [§III.A] The rescaling definitions I2 = min(I2^(4) − 1, 1) and I1 = min(I1^(4)/0.1 − 1, 1) are not bounded below as stated in the text ('between 1 and −1'). Unless an additional lower truncation is applied, these quantities can be arbitrarily negative. Please clarify the exact clipping operation.
  4. [§III.C] The network was trained from random initialization, but no information is given about the variance of the reported metrics across random seeds or hyperparameter choices. Given the small test set, reporting seed-averaged results with standard deviations would substantially strengthen the quantitative claims.
  5. [Ref. [49]] The trained model is not publicly released; Ref. [49] states that the authors 'are happy to help' rather than providing a repository. For reproducibility, consider releasing the model weights with a DOI-stable archive.
  6. [Throughout] The phrase 'arbitrary confidence' is used repeatedly (abstract, Sec. IV.A, conclusion). Since the attainable confidence is bounded by finite test statistics and the selected cutoffs, 'tunable confidence' or 'high confidence' would be more precise.
  7. [§III.A] The edge length is denoted ϵ in the text but appears as ξ in the definition of I2; please make the notation consistent.
  8. [Figure captions] Every figure caption begins with a stray '(a)', likely a LaTeX artifact. Please remove it unless subfigures are actually intended.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported 'arbitrary confidence' false-positive probabilities are obtained by tuning the cutoff on the same test data used to report them, so the headline P>0.9998 claim is a selected-subsample frequency rather than an independent prediction.

  1. fitted input called prediction [Section IV.A (Aggregated Topological Invariant), full regime paragraph; abstract]
    "when the Ccutof f is further lowered to −0.8, no devices in our test data are incorrectly declared topological, implying a probability of a topological pass being accurate at a fidelity level of > 0.9998."

    The passing cutoff Ccutoff is chosen by inspecting the very same test data that is then used to report the false-positive probability. The paper lowers the threshold until zero failures are observed in the test set and then reports the resulting frequency (shown as 1.0000 in Tables II–IV) as a predictive confidence bound. This is circular in a statistical sense: the reported 'P>0.9998' is not an independent prediction but a selected-subsample frequency, with the threshold acting as a fitted parameter on the test set. Zero failures in a few hundred passing devices gives a one-sided 95% upper bound on the false-positive probability of roughly 0.3–1%, not 0.02%, so the claimed 0.0002 false-positive probability is forced by test-set selection rather than supported by held-out evidence.

full rationale

The core supervised-learning pipeline is not circular: conductance inputs and indicator targets (TV, I1, I2) are computed independently from the same KWANT BdG Hamiltonian; the network is trained on one set of disorder realizations and evaluated on a held-out 10% split. The labels are not constructed from the conductance inputs, and no target quantity is fed back into the training features. The self-citations to Refs [29–31] supply the definition and motivation for the alternative indicators, but the learned mapping from conductance to those indicators is an independent empirical claim. The one genuine circularity is in the confidence claim. The paper tunes the passing cutoff Ccutoff by examining the same test data used to report false-positive probabilities, then reports zero observed failures as 'P>0.9998'. This is a fitted threshold reported as a predictive guarantee; the abstract's 'arbitrary confidence' inherits the test-set selection. The paper even acknowledges the limited test data in the extremes-regime sentence ('for our (limited) test data'), which undercuts the statistical force of the claim. No other reduction of the derivation to its inputs is present; the transfer-to-experiment weakness is an external-validity concern, not circularity.

Assumptions & free parameters 10 free parameters · 7 assumptions · 0 invented entities

The central result depends on a chain of modeling choices: the BdG Hamiltonian, the disorder distribution, the indicator definitions, and post-hoc thresholds. These are inputs to the ML scheme, not outputs, so they must be independently justified for the method to transfer to experiments.

free parameters (10)
  • Ccutoff (topological passing cutoff) = 0 to -0.95 depending on regime
    Chosen post hoc on the test set to reach P>0.9998; not set a priori and not calibrated on independent data.
  • epsilon (edge/bulk boundary) = 70 nm
    Set as approximately the coherence length from fitting disorder-free LDOS; directly defines the edge and bulk regions for I1 and I2.
  • I2 pass threshold = 1
    Set by observing the ideal no-disorder case; acknowledged by the authors as arbitrary.
  • I1 pass threshold = 0.1
    Set by observing the ideal no-disorder case; acknowledged by the authors as arbitrary.
  • disorder magnitude ranges (sigma_dis) = [0.15,1.05]+[3,4.5] meV extremes; [0.15,4.5] meV full; 1.5 meV moderate
    Chosen to match recent experiment [18] and prior work; determines the training distribution.
  • disorder correlation length range Lc = [20,70] nm
    Chosen by assumption; the authors note that narrowing this range improves results, implying sensitivity to this choice.
  • spin-orbit coupling alpha range = [6.8,9.2] nm*meV
    Chosen based on physical expectations for InAs-Al nanowires; treated as an unknown random parameter in training.
  • barrier height V_L = V_R = 15 mV
    Fixed for all simulations; authors say it can be varied only at the cost of generating new training data.
  • B/mu/Vbias measurement grid = B: 20 points in [0,0.8] T; mu: 5 points in [0.2,0.4] meV; Vbias: 81-151 points in +/-0.05 meV
    Chosen based on current experiments; determines the input image dimensions and coverage of the phase diagram.
  • ViT hyperparameters (patch size, number of blocks, dropout) = 3D patch (10,5,5), 128 filters, 4 ViT blocks, dropout 0.1/0.3
    Chosen by hand and not systematically optimized or ablated; affects the network's capacity and performance.
assumptions (7)
  • domain assumption The BdG Hamiltonian Eq. 1 with parameters m*=0.03me, gamma=0.15meV, g=25, Delta0=0.12meV accurately models experimental InAs-Al Majorana nanowires.
    Invoked in Section II; all labels and inputs derive from this model.
  • domain assumption KWANT scattering-matrix transport with BTK and finite-temperature convolution produces conductance data faithful to experiments.
    Section II; conductance traces used as inputs are simulated, not measured.
  • domain assumption The continuous topological visibility TV (sign of det reflection matrix) is a meaningful label for topological phase in finite disordered wires.
    Section III A; the paper itself notes sign alone is insufficient for short wires, so magnitude and alternative indicators are needed.
  • domain assumption The LDOS-based indicators I1 and I2 from Ref. [31] operationally identify usable end-localized Majoranas.
    Section III A; thresholds are set by clean-limit inspection and acknowledged as arbitrary.
  • domain assumption Gaussian disorder with convolved correlation length and uniform sampling of sigma_dis, Lc, and alpha covers the experimental disorder distribution.
    Section III B-D; training and test data are drawn from this distribution.
  • domain assumption Held-out i.i.d. samples from the same generator estimate real-world generalization.
    Section III C; no experimental or out-of-distribution test is performed.
  • ad hoc to paper Fixed barrier height and fixed material parameters can be held constant without loss of generality.
    Section II; authors state variation is possible only with new training data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Vision transformer based Deep Learning of Topological indicators in Majorana Nanowires." pith.science (2026). https://pith.science/paper/VAVJQLQJ

@misc{pith2026241206768,
  author       = {Pith},
  title        = {Pith review of: Vision transformer based Deep Learning of Topological indicators in Majorana Nanowires},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VAVJQLQJ}},
  note         = {Machine review of arXiv:2412.06768}
}
abstract

1D superconductor-semiconductor nanowires are the leading candidates for topological quantum computation due to their ability to host non-Abelian Majorana zero modes (MZMs). However, the standard methods for identifying MZMs are often inadequate, particularly in the presence of disorder, where many properties considered to be heralds of MZMs are often generated by trivial disorder induced Andreev bound states. Recent works clearly indicate the need for developing new techniques for identifying and diagnosing MZMs. In this study, we utilize a generalized Vision Transformer-based neural network to predict, using tunnel conductance measurements, both whether a device manifests a topological MZMs phase in the presence of disorder, and also to map out the entire topological phase diagram. We show the ability of our method up to arbitrary confidence ($P>0.9998$) in classifying a device as possessing a non-trivial MZM-carrying topological phase for a wide variety of disorder parameters. We demonstrate an ability to predict from conductance measurements alternative (to the extensively used scattering-matrix-invariant topological indicator) Majorana indicators based on local density of states (LDOS). This is relevant since topology may not be uniquely defined by the scattering invariant in short disordered wires. This work serves as a significant advance offering a step towards the practical realization of Majorana-based quantum devices, enabling a deep-learning understanding of the topological properties in disordered nanowire systems. We validate our method using extensive simulated Majorana results in the presence of disorder, and suggest using this technique for the analysis of experimental data in superconductor-semiconductor hybrid Majorana platforms.

Figures

Figures reproduced from arXiv: 2412.06768 by the authors.

Figure 1
Figure 1. FIG. 1: (a) Diagram of the neural network used to [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Comparisons between the expected minimum [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: FIG. 4: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9: Comparison between expected [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 41 canonical work pages

  1. [50]

    Contributors, Vision transformer implementation in torchvision, https://github.com/pytorch/vision/ blob/main/torchvision/models/vision_transformer

    T. Contributors, Vision transformer implementation in torchvision, https://github.com/pytorch/vision/ blob/main/torchvision/models/vision_transformer. py (2025), accessed: 2025-02-14

    work page 2025

  2. [1]

    P(Passing) describes the fraction of devices being declared passing the scheme, namely devices with T P red V < Ccutof f

    is the false negative probability. P(Passing) describes the fraction of devices being declared passing the scheme, namely devices with T P red V < Ccutof f. The root mean squared error between the expected and predicted results was RM S(TV ) = 0.3134. Next we discuss the moderate disorder regime, where it is difficult to determine whether a device is in a...

    work page 2015

  3. [2]

    J. D. Sau, R. M. Lutchyn, S. Tewari, and S. Das Sarma, Robustness of majorana fermions in proximity-induced superconductors, Physical Review B—Condensed Matter and Materials Physics82, 094522 (2010)

    work page 2010

  4. [3]

    R. M. Lutchyn, J. D. Sau, and S. Das Sarma, Ma- jorana fermions and a topological phase transition<? format?> in semiconductor-superconductor heterostruc- tures, Physical review letters105, 077001 (2010)

    work page 2010

  5. [4]

    J. D. Sau, R. M. Lutchyn, S. Tewari, and S. Das Sarma, Generic new platform for topological quantum computa- tion<? format?> using semiconductor heterostructures, Physical review letters104, 040502 (2010)

    work page 2010

  6. [5]

    Y. Oreg, G. Refael, and F. Von Oppen, Helical liquids and majorana bound states in quantum wires, Physical review letters 105, 177002 (2010)

    2010

  7. [6]

    S. D. Sarma, M. Freedman, and C. Nayak, Majorana zero modes and topological quantum computation, npj Quan- tum Information 1, 1 (2015)

    work page 2015

  8. [7]

    Das Sarma, In search of majorana, Nature Physics19, 165 (2023)

    S. Das Sarma, In search of majorana, Nature Physics19, 165 (2023)

    work page 2023

Show all 53 references
  1. [8]

    Results are shown for different cutoff val- ues for which a device will be declared passing

    A summary of these results can be seen below in Table X below: Cutoff 1 − FP 1 − FN P(Passing) 0.0000 0.9074 0.7419 0.8270 -0.3000 0.9394 0.7881 0.7476 -0.5000 0.9563 0.8231 0.6925 -0.7000 0.9698 0.8671 0.6217 -0.8000 0.9762 0.8925 0.5714 -0.9000 0.9865 0.9221 0.4760 -0.9500 0...

  2. [9]

    R. M. Lutchyn, E. P. Bakkers, L. P. Kouwenhoven, P. Krogstrup, C. M. Marcus, and Y. Oreg, Majo- rana zero modes in superconductor–semiconductor het- erostructures, Nature Reviews Materials3, 52 (2018)

    2018

  3. [10]

    A. Y. Kitaev, Unpaired majorana fermions in quantum wires, Physics-uspekhi 44, 131 (2001)

    2001

  4. [11]

    A. Das, Y. Ronen, Y. Most, Y. Oreg, M. Heiblum, and H. Shtrikman, Zero-bias peaks and splitting in an al– inas nanowire topological superconductor as a signature of majorana fermions, Nature Physics8, 887 (2012)

    2012

  5. [12]

    M. Deng, C. Yu, G. Huang, M. Larsson, P. Caroff, and H. Xu, Anomalous zero-bias conductance peak in a nb– insb nanowire–nb hybrid device, Nano letters12, 6414 (2012)

    2012

  6. [13]

    Mourik, K

    V. Mourik, K. Zuo, S. M. Frolov, S. Plissard, E. P. Bakkers, and L. P. Kouwenhoven, Signatures of majo- rana fermions in hybrid superconductor-semiconductor nanowire devices, Science336, 1003 (2012)

    2012

  7. [14]

    Churchill, V

    H. Churchill, V. Fatemi, K. Grove-Rasmussen, M. Deng, P. Caroff, H. Xu, and C. M. Marcus, Superconductor- nanowire devices from tunneling to the multichannel regime: Zero-bias oscillations and magnetoconductance crossover, Physical Review B—Condensed Matter and Materials Physi...

    2013

  8. [15]

    Finck, D

    A. Finck, D. Van Harlingen, P. Mohseni, K. Jung, and X. Li, Anomalous modulation of a zero-bias peak in a hybrid nanowire-superconductor device, Physical review letters 110, 126406 (2013)

    2013

  9. [16]

    Nichele, A

    F. Nichele, A. C. Drachmann, A. M. Whiticar, E. C. O’Farrell, H. J. Suominen, A. Fornieri, T. Wang, G. C. Gardner, C.Thomas, A.T.Hatke, et al.,Scalingofmajo- rana zero-bias conductance peaks, Physical review letters 119, 136803 (2017)

    2017

  10. [17]

    Zhang, M

    H. Zhang, M. W. de Moor, J. D. Bommer, D. Xu, G. Wang, N. van Loo, C.-X. Liu, S. Gazibegovic, J. A. Logan, D. Car, et al. , Large zero-bias peaks in insb-al hybrid semiconductor-superconductor nanowire devices, arXiv preprint arXiv:2101.11456 (2021)

    2021 arXiv

  11. [18]

    P. Yu, J. Chen, M. Gomanko, G. Badawy, E. Bakkers, K. Zuo, V. Mourik, and S. Frolov, Non-majorana states yield nearly quantized conductance in proximatized nanowires, Nature Physics17, 482 (2021)

    2021

  12. [19]

    H. Song, Z. Zhang, D. Pan, D. Liu, Z. Wang, Z. Cao, L. Liu, L. Wen, D. Liao, R. Zhuo,et al., Large zero bias peaks and dips in a four-terminal thin inas-al nanowire device, Physical Review Research4, 033235 (2022)

    2022

  13. [20]

    Aghaee, A

    M. Aghaee, A. Akkala, Z. Alam, R. Ali, A. A. Ramirez, M. Andrzejczuk, A. E. Antipov, P. Aseev, M. Astafev, B. Bauer, et al., Inas-al hybrid devices passing the topo- logical gap protocol, Physical Review B 107, 245423 (2023)

    2023

  14. [21]

    S. Ahn, H. Pan, B. Woods, T. D. Stanescu, and S. D. Sarma, Estimating disorder and its adverse effects in semiconductor majorana nanowires, Physical Review Materials 5, 124602 (2021)

    2021

  15. [22]

    Sengupta, I

    K. Sengupta, I. Žutić, H.-J. Kwon, V. M. Yakovenko, and S.D.Sarma,Midgapedgestatesandpairingsymmetryof quasi-one-dimensional organic superconductors, Physical Review B 63, 144531 (2001)

    2001

  16. [23]

    Pan and S

    H. Pan and S. D. Sarma, Physical mechanisms for zero- bias conductance peaks in majorana nanowires, Physical Review Research 2, 013377 (2020)

    2020

  17. [24]

    J. Liu, A. C. Potter, K. T. Law, and P. A. Lee, Zero-bias peaks in the tunneling conductance of spin-orbit-coupled superconducting wires with and without majorana end- states, Physical review letters109, 267002 (2012)

    2012

  18. [25]

    Das Sarma and H

    S. Das Sarma and H. Pan, Disorder-induced zero-bias peaks in majorana nanowires, Physical Review B 103, 195158 (2021)

    2021

  19. [26]

    D. I. Pikulin, B. van Heck, T. Karzig, E. A. Martinez, B. Nijholt, T. Laeven, G. W. Winkler, J. D. Watson, S. Heedt, M. Temurhan, et al. , Protocol to identify a topologicalsuperconductingphaseinathree-terminalde- vice, arXiv preprint arXiv:2103.12217 (2021)

    2021 arXiv

  20. [27]

    J. R. Taylor, J. D. Sau, and S. Das Sarma, Machine learn- ing the disorder landscape of majorana nanowires, Phys- ical Review Letters132, 206602 (2024)

    2024

  21. [28]

    Thamm and B

    M. Thamm and B. Rosenow, Machine learning optimiza- tion of majorana hybrid nanowires, Physical Review Let- ters 130, 116202 (2023)

    2023

  22. [29]

    Akhmerov, J

    A. Akhmerov, J. Dahlhaus, F. Hassler, M. Wimmer, and C. Beenakker, Quantized conductance at the majorana 24 phase transition in a disordered superconducting wire, Physical review letters106, 057001 (2011)

    2011

  23. [30]

    Das Sarma, A

    S. Das Sarma, A. Nag, and J. D. Sau, How to infer non- abelianstatisticsandtopologicalvisibilityfromtunneling conductance properties of realistic majorana nanowires, Physical Review B94, 035143 (2016)

    2016

  24. [31]

    Das Sarma, J

    S. Das Sarma, J. D. Sau, and T. D. Stanescu, Spectral properties, topological patches, and effective phase dia- grams of finite disordered majorana nanowires, Physical Review B 108, 085416 (2023)

    2023

  25. [32]

    Das Sarma and H

    S. Das Sarma and H. Pan, Density of states, transport, and topology in disordered majorana nanowires, Physical Review B 108, 085415 (2023)

    2023

  26. [33]

    Pan and S

    H. Pan and S. Das Sarma, Disordered majorana nanowires: Studying disorder without any disorder, Physical Review B110, 075401 (2024)

    2024

  27. [34]

    Peeters, T

    C. Peeters, T. Hodge, E. Mascot, and S. Rachel, Effect of impurities and disorder on the braiding dynamics of majorana zero modes, Physical Review B 110, 214506 (2024)

    2024

  28. [35]

    Cheng, R

    M. Cheng, R. Okabe, A. Chotrattanapituk, and M. Li, Machine learning detection of majorana zero modes from zero-bias peak measurements, Matter7, 2507 (2024)

    2024

  29. [36]

    C. W. Groth, M. Wimmer, A. R. Akhmerov, and X. Waintal, Kwant: a software package for quantum transport, New Journal of Physics16, 063065 (2014)

    2014

  30. [37]

    Das Sarma, J

    S. Das Sarma, J. D. Sau, and T. D. Stanescu, Split- ting of the zero-bias conductance peak as smoking gun evidence for the existence of the majorana mode in a superconductor-semiconductor nanowire, Physical Re- view B—Condensed Matter and Materials Physics 86, 220506 (2012)

    2012

  31. [38]

    Y.-H. Lai, J. D. Sau, and S. Das Sarma, Presence versus absence of end-to-end nonlocal conductance correlations in majorana nanowires: Majorana bound states versus andreev bound states, Physical Review B 100, 045302 (2019)

    2019

  32. [39]

    H. Pan, J. D. Sau, and S. Das Sarma, Three-terminal nonlocal conductance in majorana nanowires: Distin- guishing topological and trivial in realistic systems with disorder and inhomogeneous potential, Physical Review B 103, 014513 (2021)

    2021

  33. [40]

    A. Vuik, B. Nijholt, A. R. Akhmerov, and M. Wimmer, Reproducing topological properties with quasi-majorana states, SciPost Physics7, 061 (2019)

    2019

  34. [41]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, et al. , An image is worth 16x16 words: Transformersforimagerecognitionatscale,arXiv preprint arXiv:2010.11929 (2020)

    2020 arXiv

  35. [42]

    J. R. Taylor and S. D. Sarma, Neural network based deep learning analysis of semiconductor quantum dot qubits for automated control, arXiv preprint arXiv:2405.04524 (2024)

    2024 arXiv

  36. [43]

    trivial andreev band mimicking topological bulk gap reopening in the nonlocal conductance of long rashba nanowires

    S. Das Sarma and H. Pan, Comment on “trivial andreev band mimicking topological bulk gap reopening in the nonlocal conductance of long rashba nanowires”, Physical Review Letters 132, 099601 (2024)

    2024

  37. [44]

    B. D. Woods, S. D. Sarma, and T. D. Stanescu, Charge- impurity effects in hybrid majorana nanowires, Physical Review Applied 16, 054053 (2021)

    2021

  38. [45]

    G. W. Winkler, A. E. Antipov, B. Van Heck, A. A. Soluyanov, L. I. Glazman, M. Wimmer, and R. M. Lutchyn, Unified numerical approach to topological semiconductor-superconductor heterostructures, Physi- cal Review B99, 245408 (2019)

    2019

  39. [46]

    Setiawan, C.-X

    F. Setiawan, C.-X. Liu, J. D. Sau, and S. Das Sarma, Electron temperature and tunnel coupling dependence of zero-bias and almost-zero-bias conductance peaks in ma- jorana nanowires, Physical Review B96, 184520 (2017)

    2017

  40. [47]

    G. J. Percebois, A. Lacerda-Santos, B. Brun, B. Hack- ens, X. Waintal, and D. Weinmann, Reconstructing the potential configuration in a high-mobility semiconduc- tor heterostructure with scanning gate microscopy, arXiv preprint arXiv:2308.13372 (2023)

    2023 arXiv

  41. [48]

    Chollet et al., Keras, https://keras.io (2015)

    F. Chollet et al., Keras, https://keras.io (2015)

    2015

  42. [49]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, et al. , Pytorch: An imperative style, high-performance deep learning library, Advances in neural information processing systems 32 (2019)

    2019

  43. [51]

    The authors are happy to help those who contact us to utilize the latest iteration of our model, which is contin- uously evolving as AI models improve

  44. [52]

    Pan, Conductance kwant, https://github.com/ hainingpan/Conductance_Kwant (2025), accessed: 2025- 02-14

    H. Pan, Conductance kwant, https://github.com/ hainingpan/Conductance_Kwant (2025), accessed: 2025- 02-14

    2025

  45. [53]

    T.ChenandC.Guestrin,Xgboost: Ascalabletreeboost- ing system, in Proceedings of the 22nd acm sigkdd in- ternational conference on knowledge discovery and data mining (2016) pp. 785–794

    2016

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.