Abstract
Threedimensional topological insulators are bulk insulators with Z_{2} topological electronic order that gives rise to conducting lightlike surface states. These surface electrons are exceptionally resistant to localization by nonmagnetic disorder, and have been adopted as the basis for a wide range of proposals to achieve new quasiparticle species and device functionality. Recent studies have yielded a surprise by showing that in spite of resisting localization, topological insulator surface electrons can be reshaped by defects into distinctive resonance states. Here we use numerical simulations and scanning tunnelling microscopy data to show that these resonance states have significance well beyond the localized regime usually associated with impurity bands. At native densities in the model Bi_{2}X_{3} (X=Bi, Te) compounds, defect resonance states are predicted to generate a new quantum basis for an emergent electron gas that supports diffusive electrical transport.
Introduction
Threedimensional topological insulators (TI) with the chemical formula M_{2}X_{3} (M=Bi, Sb; X=Se, Te) are characterized by a singleDiraccone surface state^{1,2,3}, and have been widely investigated as model materials for TI surface physics. Shortly after the discovery of the first TI’s, it was established that standard effects of defect disorder such as backscattering and localization are greatly suppressed in TI surface states^{4,5,6,7,8,9,10,11}. Altogether with the intuition that topological properties of band structure should be inherently impervious to weak perturbations, this led to a pervasive characterization that topological interfaces are essentially blind to the presence of nonmagnetic defects. It is only with the recent advent of more detailed nonperturbative numerical investigations and targeted scanning tunnelling microscopy (STM) experiments^{12,13,14,15,16,17,18,19,20,21,22,23,24,25} that elements of this picture have been overturned. It is now known that, rather than being blind to defects, topological Dirac cone electrons actually bind loosely to atomic point defects in energy levels that fall very close to the Dirac point (see states labelled ‘resonance’ in Fig. 1b), and can radically redefine electronic structure relevant to the Dirac transport regime.
In isotropic Dirac models of TI surface states, strongly perturbing nonmagnetic defects are theoretically expected to create circular swave resonance states, such as those plotted in Fig. 1a. Experimentally, these states have been observed in STM measurements of density of states (DOS) immediately above the Dirac point of topological insulator Bi_{2}Se_{3} (see Fig. 1a (left) inset patch). Three bright spots in the STM image represent coherent interference with a slight ≲1% admixture of fwave symmetry (Supplementary Note 2) because of anisotropy not considered in the accompanying simulation. The combined impact of multiple randomly distributed defects on the electronic structure has not yet been evaluated, and is the subject of the present work.
Theoretical treatments of lattice defects are usually carried out in the single impurity approximation, since multiple impurities do not lead to emergent coherent bandstructures. Large modifications to the band structure of nonTI materials are only achieved through heterostructure growth or fractional alloying. The ‘impurity bands’ associated with electrons on minority impurity sites are generally disregarded as being nonideal for electronic transport. These bands exhibit miniscule dispersions, and strong defect potentials or reduced dimensionality cause the electronic states to be Anderson localized^{26,27} over relevant device lengthscales. A typical example is the wellstudied manganese impurity band in Ga_{1−x}Mn_{x}As, whose dispersion across the Brillouin zone is far smaller than its intrinsic width along the energy axis, meaning that the electrons are effectively localized to isolated lattice sites^{28,29}.
In the following, in contrast, we find that uncompensated lattice defects at low densities (ρ<0.1%) in nonamorphous TIs provide a mechanism for band structure engineering that is not shared in a conventional semiconducting system. Theoretically, it is expected that Anderson localization will be forbidden on the TI surface with nonmagnetic impurities^{30,31,32,33,34} and backscattering is suppressed relative to forward scattering^{4,5,9}, leading to greater itinerancy for impurityderived band structure. Moreover, the electronic density of states falls to zero at a surface Dirac point, making it possible for defect resonance states to dominate the absolute density of states. In the following discussions we will show that these conditions enable new defectderived electronic dispersions, providing viable channels for diffusive electrical transport. We also find that these emergent electronic band structures can superficially appear to be gapped, and resemble Dirac point anomalies that have been seen by angle resolved photoemission spectroscopy (ARPES) in TI compounds that have no known time reversal symmetry broken ground state^{35,36,37,38}.
Results
Defect resonance states of Dirac electrons
We will focus on defect resonance states in the popular Bi_{2}Se_{3−x}Te_{x} family of singleDiraccone TIs^{1,2,3}, so as to make quantitative comparisons to STM measurements. However, our results are applicable to other TI surface states with isolated Dirac cones. In this material family, structural point defects tend to occur with a density of ρ≲0.1% per formula unit, and have been found to overlap with the surface state in the outermost quintuple layer of the lattice^{15,39} (see Supplementary Note 1). Intensive characterization of Bi_{2}Se_{3} has found that the density and typerange of defects can be tuned over a broad range by adjusting sample growth conditions^{39}. For Sepoor and stoichiometric synthesis, defects near a cleaved surface are predominantly selenium vacancies positioned in the centre of the outermost quintuple layer of the lattice. Interstitial atoms and replacement defects can be common at the surface of Serich or quenchcooled samples. The chemical potential can be shifted as a separate parameter via methods that include bulk doping, surface dosing, and electrostatic gating^{40,41,42,43}.
Defect resonance states have been observed by STM less than 200 meV above the surface Dirac point, surrounding point defects^{14,15} and at step edges between plateaus differing in height by one quintuple layer^{18,19}. Characteristic local density of states (LDOS) distributions measured by STM at the defect lattice sites are plotted as black curves in Fig. 2b,c. Theoretical analyses of isolated local impurities have shown that these resonance states arise naturally when strong scalar potentials are introduced to a surface occupied by spinhelical TI Dirac cone electrons^{12,13,15,16,17}. For an isolated defect described as a point potential with amplitude U, the resonance state is split from the Dirac point by an energy roughly proportional to −U^{−1}.
In the following, we numerically approach the problem of multiple impurity scattering on the surface states of TIs with a single Dirac cone, which is otherwise theoretically intractable. We use a basis of spinor plane waves^{44} on a discrete hexagonal realspace lattice, with lattice parameter a=4.2 Å corresponding to Bi_{2}Se_{3−x}Te_{x} surface atomic spacings. We have used the impurityfree spinhelical Hamiltonian, H_{k}=v_{0}(k × σ), with Dirac velocity v_{0}=3.0 eV·Å chosen to be intermediate between Bi_{2}Se_{3} and Bi_{2}Te_{3}.
When a large surface with a typical ρ∼0.06% defect density is modelled in this way, we find that ∼2 electronic states per defect are added near the Dirac point (see Supplementary Note 1; Supplementary Fig. 4), resulting in a new DOS peak that becomes increasingly visible as the defect potential grows (Fig. 2a). Isolating the LDOS at defect cores on Bi_{2}Se_{3} reveals that a defect potential of U=−45 eV accurately reproduces the resonance state energy and LDOS line shape obtained in STM studies, which is plotted in Fig. 2b. The very different experimental line shape of a step edge defect can likewise be reproduced by considering the tunnelling barrier generated with a repulsive potential of U=3.8 eV applied to each atom along the step edge path (Fig. 2c). These comparisons are evidence that our numerical approach is able to quantitatively capture the physics observed in experiments.
New delocalized electronic structure
The resonance states associated with this defect distribution reveal a novel structure in the momentumresolved electron annihilation (ARPES) spectral function, modelled in Fig. 2d–f. The resonance itself shows up as a discontinuity in the ARPES spectral function. As the defect potential strength increases, the surface Dirac point (termed D_{0}), as well as the discontinuity associated with the resonance state LDOS maximum (E_{R}) move down in energy (Fig. 3h). When the defect potential crosses a critical value of U≲∼45 eV, a new local maximum appears in the spectrum at K=0 ( point), roughly 50 meV above the D_{0} Dirac point, and will henceforth be referred to as ‘D_{1}’. As the defect resonance state converges on E=0 eV with increasing potential strength, the region between D_{0} and D_{1} begins to qualitatively resemble a band gap, as has been seen at the Dirac point of TI crystals with nonmagnetic defects in refs 35, 36. Unlike a true band gap, which is theoretically disallowed^{6}, the momentumintegrated total DOS distribution contains a local maximum within the seemingly gapped region (Fig. 2a).
A remarkable evolution of the electronic structure also occurs when the defect density is increased (see Fig. 3a–c series). When the defect density is tripled to ρ=0.18%, the faint local maximum D_{1} becomes the dispersion minimum of an upper band that is clearly disconnected from the D_{0} Dirac cone. The D_{0} upper Dirac cone maintains its dispersion for small momenta K≲0.02 Å, out to more than 50 meV from the Dirac point, and becomes dispersionless at larger momenta, as would be expected for a more typical impurity band. For all of these impurity densities, the upper D_{0} Dirac cone dispersion is significantly larger than the energy axis intrinsic width, which is δE∼0.02 and 0.04 eV at half maximum intensity in Fig. 3b,c, respectively.
Higher defect densities are also associated with further lowering of the D_{0} Dirac point energy. An STM map obtained on a Se vacancyrich Bi_{2}Se_{3−δ} surface shows that spatial fluctuations in the energy of the D_{0} LDOS minimum are highly correlated with defect density (Fig. 3e). Simulating the observed Se vacancies with a U=−45 defect potential reproduces the qualitative spatial distribution of the D_{0} LDOS minimum (Fig. 3f). These modelling results can be plotted in the same ±25 meV range as the STM data, showing that the energy scale of fluctuations is accurately captured, even though fine details may be inconsistent because of missing information about defects outside of the STM map, or deeper in the crystal. We note that the spatial distribution of defects observed by STM is very uniform and uncorrelated, and is not readily distinguished from the random defect configurations used in our simulations (see Poisson distribution overlay in Supplementary Fig. 1).
The emergence of new spectral features at high defect density is closely related to the fraction of states near the Dirac point that can be attributed as defect resonance states. In the absence of defects, the total number of states within an energy E of the Dirac point is , where N is the total number of 2D surface lattice sites. Setting N_{E} equal to the approximate number of resonance states (N_{E}=N_{R}≡2ρN) gives a new energy scale, , which we find to have a novel physical significance. As can be see in Fig. 3g,h, we find that a local maximum associated with the emergent D_{1} feature becomes visible when the resonance state energy E_{R} becomes smaller than E*. These simulations also reveal that the resonance energy may be tuned not only by changing the impurity strength, as has been theoretically predicted before, but also by increasing the impurity concentration, a fact that is technologically significant because the impurity strength is usually not tunable. With the experimentally fitted defect potential U=−45 eV, the resonance energy crosses E* when the defect density becomes larger than ρ=0.02%, resulting in the easily visible emergent features in Fig. 3b,c.
To evaluate the ability of the emergent coherent impurity bandlike features to carry mobile charge, we have calculated the twist velocity, v_{θ}, for configurations with 300 × 300 sites (300 sites=126 nm per side). These are shown in Fig. 3a–c, as vectors overlaid at local maxima of the modelled ARPES spectrum. The twist velocity of a quantum state is proportional to the magnitude of a persistent DC current carried by that state in the presence of an electric field (see Methods), and in the scatteringfree limit is equivalent to the group velocity of the electronic band. It is calculated as the gradient of the state energies with respect to a phase twist around the system boundary.
We find that the twist velocities of band features at energies E>0 eV have similar amplitudes, and are approximately proportional to the apparent slope in momentum space, suggesting that all of these features are viable bands for charge transport. However, twist velocities decay towards zero as the upper D_{0} Dirac cone disperses into the seemingly gapped region beneath the D_{1} feature, particularly at momenta of k0.05 Å, and this may greatly limit carrier mobility for certain values of the chemical potential (see evaluation of twist and band velocities in Supplementary Note 1; Supplementary Figs 2 and 3). Velocities are relatively large in the lower D_{0} Dirac cone, where states are energetically distant from the resonance states, and maximal velocities are found immediately beneath D_{0}, where elastic scattering is minimal because of the vanishing density of states near a 2D Dirac point.
For comparison, a similar simulation with magnetic defects instead of a U=−45 eV scalar field is shown in Fig. 3d, representing a case in which the topological protection against Anderson localization is expected to fail^{45}. We have approximated each magnetic defect by a localized J=45 eV exchange field perfectly aligned with the +zaxis, perpendicular to the surface (see Methods). In this magnetic scenario, even though disorder of the magnetic moment vectors has been neglected, twist velocities are still reduced by an additional order of magnitude near the Dirac point because of Anderson localization.
We have also investigated the transport characteristics of the potential disorderinduced band structure by an alternate metric, the participation ratios of the quantum eigenstates. The participation ratio^{46,47} of an electronic wavefunction is the sum of squares of the siteresolved probability densities (see Methods). For itinerant electronic states that can participate in electronic transport, it decays as the inverse of the total number of sites in the system (N^{−1}), for large system sizes. On the other hand, for localized states, the participation ratio converges on a nonzero value as system size increases beyond the localization length of the wavefunction. The participation ratios of the energy eigenstates in the presence of scalar and magnetic disorder are compared in Fig. 4. We find that in the latter case, Anderson localized states around the resonances display large participation ratios that saturate with system size (Fig. 4e), while topological protection against localization ensures that the participation ratios for the comparable scalar potential disorder case vary inversely as the system size, for large system sizes (Fig. 4b). Calculating the twist velocities for larger system sizes shows a similar trend (Fig. 4c,f), in which scalar defect potentials allow diffusive transport (v_{θ}L^{−1}), being topologically protected against Anderson localization, unlike conventional impurity bands, represented here by the same surface states in the presence of magnetic disorder of comparable strength.
Discussion
Impurityinduced carrier control is the cornerstone of the semiconductor industry, and band structure engineering via disorder is the natural successor of such disorderinduced functionalization of electronic materials. In conventional 2D materials, Anderson localization prevents impurity bands from acting as viable carriers for diffusive charge transport. In this work, we have shown that this restriction is ameliorated in the surface states of topological insulators, in the presence of moderate potential disorder that does not break time reversal symmetry. Using numerical techniques verified by comparison to experimental observations, we have demonstrated that a new impurity band structure emerges in the TI surface states in the presence of scalar potential disorder, and that these states remain diffusive and are able to participate in transport processes. We have contrasted this novel behaviour, when topological protection against localization is present, to the more conventional situation when magnetic disorder of comparable strength is present, when Anderson localization prevents diffusive transport in the impurity band.
Our results also shed light on some less understood DOS anomalies on TI surface states that have been reported in the literature. ARPES experiments have previously resolved gaplike features^{35,36,37} and Dirac point elongation^{38} in the spectrum of TI surface states, in the absence of any timereversal breaking effects. The experiment in ref. 37 presents a particularly telling case. In this work, incident photon polarization was tuned to eliminate the dominant evensymmetry photoemission channel from Bi_{2}Se_{3} along the high symmetry − axis. Defect resonance states do not have spatial reflection symmetry because of their random distribution, and would not be intrinsically suppressed by this symmetry factor. This defectsensitive measurement geometry revealed a gaplike dispersion, with the upper Dirac cone bands appearing to converge more than E>50 meV above the Dirac point, as expected from our calculations for the D_{1} feature.
We have shown that these observations may be attributed to the effects discussed above, arising from scalar potential disorder alone. The emergent band structure associated with defects is very narrowly separated from nearby bands in momentum, and will be difficult to cleanly resolve in ARPES experiments. The energyaxis widths of bands in our calculations are 2–3 times narrower than those actually observed in Bi_{2}Se_{3}, indicating that factors not considered here such as phonon scattering and surface inhomogeneity may play a significant role in the low energy physics, and may obscure the role of defectderived states.
Nonetheless, the existence of defect resonance states immediately above the Dirac point of these compounds is well known from STM measurements, and the new physics discussed in this work are direct consequences of the interplay between these resonance states and other established elements of TI surface physics. The numerical evaluations of electron mobility and localization discussed above show that the emergent band structure identified in these simulations is viable as a basis for conductivity, and that band structure engineering is possible via nonmagnetic defects in a TI system, driven by unique factors such as the lack of Anderson localization and the vanishing of DOS near a 2D Dirac point.
Methods
Energetics and state basis at the TI surface
STM data and LDOS curves were obtained by the procedures described in refs 15, 18. The surface state is modelled as a spinhelical 2D Dirac cone occupying a discrete hexagonal realspace lattice with repeating boundary conditions, and is essentially a discretized version of the model in refs 13, 15. The kinetic Hamiltonian is given by H_{k}=v_{0}(k × σ), with velocity v_{0}=3.0 eV chosen to be intermediate between Bi_{2}Se_{3} and Bi_{2}Te_{3}. The real space basis includes two spindegenerate states per lattice site. The kinetic Hamiltonian is diagonalized on this basis, and the resulting momentumspace basis is reduced by applying a high energy cutoff W=0.4 eV to exclude high energy states that would not conform to a Dirac dispersion.
Defects are created by perturbing sites at selected coordinates i with a scalar energy term. as , where n_{i} is the number operator, and the full Hamiltonian H=H_{k}+H_{d} is diagonalized. The convention of creating point defects as triangular threesite clusters with a perturbation strength of U/3 on each site is adopted to give threefold rotational symmetry, as seen experimentally^{14,15}, though calculated spectra do not strongly differentiate between differently sized small site clusters (Supplementary Fig. 6). The Hamiltonian is numerically diagonalized, and the ARPES spectral function is defined as the energy and momentumresolved spectral function of single particle annihilation, convoluted by a 10 meV halfwidth Lorentzian:
This represents the spectral function for photoemission with the zaxis photon polarization, which acts with even reflection symmetry and removes electrons from the dominant p_{z} orbital component of the Bi_{2}Se_{3−x}Te_{x} surface state. Energymomentum dispersions were found to be rotationally isotropic, and have been rotationally averaged to obtain a small spacing between states on the momentum axis. Figures in the paper use the rotationally averaged ‘radial’ momentum (K_{r}), and singleaxis spectral functions can be found in Supplementary Fig. 5. For smaller system sizes considered in Fig. 4, it was necessary to average over several randomly generated defect configurations to achieve convergence of the twist velocity and participation ratio. Additional modelling details are found in Supplementary Note 3.
Metrics of localization
Adding a phase twist θ at the system boundary provides an easy way to identify the degree to which electronic states are localized. This is achieved by the standard method of shifting the momentum (Kspace) basis of the Hilbert space by an offset of , where L_{x} (L_{y}) is the width of the system along the x− (y−) axis. A velocity termed the ‘twist velocity’ can be calculated for individual eigenstates α〉 as v_{θ}(α)=∇_{Kθ}E_{α} (using numerical derivatives), and represents the velocity of a persistent current. The energy and momentumresolved twist velocity (v_{θ}(E, k)) displayed on ARPES spectral intensity maxima in Fig. 3 is calculated via a weighted average of states within ±10 meV of the energy E, as:
Here, represents the free particle eigenstates that have momentum k. When divided by the width of the system L along the indicated axis, the phase twist has units of wave number. The numerical derivative is performed at a small but nonzero twist value of along the direction to break degeneracies and avoid possible finite size nesting effects.
The participation ratio is defined as:
where the sum is over all sites in the system, and n_{i,α} is the LDOS on site i of α〉, which is an eigenstate of the full Hamiltonian.
Data availability
The data and source code that support the findings of this study are available from the corresponding authors on reasonable request.
Additional information
How to cite this article: Xu, Y. et al. Disorder enabled band structure engineering of a topological insulator surface. Nat. Commun. 8, 14081 doi: 10.1038/ncomms14081 (2017).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
 1.
Zhang, H. et al. Topological insulators in Bi_{2}Se_{3}, Bi_{2}Te_{3} and Sb_{2}Te_{3} with a single Dirac cone on the surface. Nat. Phys. 5, 438–442 (2009).
 2.
Xia, Y. et al. Observation of a largegap topologicalinsulator class with a single Dirac cone on the surface. Nat. Phys. 5, 398–402 (2009).
 3.
Chen, Y. L. et al. Experimental realization of a threedimensional topological insulator, Bi_{2}Te_{3}. Science 325, 178–181 (2009).
 4.
Hasan, M. Z. & Kane, C. L. Colloquium: topological insulators. Rev. Mod. Phys. 82, 3045 (2010).
 5.
Qi, X.L. & Zhang, S.C. Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057 (2011).
 6.
Fu, L., Kane, C. L. & Mele, E. J. Topological insulators in three dimensions. Phys. Rev. Lett. 98, 106803 (2007).
 7.
Hsieh, D. et al. Observation of unconventional quantum spin textures in topological insulators. Science 323, 919–922 (2009).
 8.
Moore, J. E. & Balents, L. Topological invariants of timereversalinvariant band structures. Phys. Rev. B 75, 121306 (2007).
 9.
Roushan, P. et al. Topological surface states protected from backscattering by chiral spin texture. Nature 460, 1106–1109 (2009).
 10.
Alpichshev, Z. et al. STM imaging of electronic waves on the surface of Bi_{2}Te_{3}: topologically protected surface states and hexagonal warping effects. Phys. Rev. Lett. 104, 016401 (2010).
 11.
Qi, X.L., Hughes, T. L. & Zhang, S.C. Topological field theory of timereversal invariant insulators. Phys. Rev. B 78, 195424 (2008).
 12.
BlackSchaffer, A. M. & Yudin, D. Spontaneous gap generation on the surface of weakly interacting topological insulators using nonmagnetic impurities. Phys. Rev. B 90, 161413 (2014).
 13.
Biswas, R. R. & Balatsky, A. V. Impurityinduced states on the surface of threedimensional topological insulators. Phys. Rev. B 81, 233405 (2010).
 14.
Teague, M. L. et al. Observation of Fermienergy dependent unitary impurity resonances in a strong topological insulator Bi_{2}Se_{3} with scanning tunneling spectroscopy. Solid State Commun. 152, 747–751 (2012).
 15.
Alpichshev, Z. et al. STM Imaging of Impurity Resonances on Bi_{2}Se_{3}. Phys. Rev. Lett. 108, 206402 (2012).
 16.
BlackSchaffer, A. M. & Balatsky, A. V. Subsurface impurities and vacancies in a threedimensional topological insulator. Phys. Rev. B 86, 115433 (2012).
 17.
BlackSchaffer, A. M. & Balatsky, A. V. Strong potential impurities on the surface of a topological insulator. Phys. Rev. B 85, 121103 (2012).
 18.
Alpichshev, Z., Analytis, J. G., Chu, J.H., Fisher, I. R. & Kapitulnik, A. STM imaging of a bound state along a step on the surface of the topological insulator Bi_{2}Te_{3}. Phys. Rev. B 84, 041104 (2011).
 19.
Dmitriev, A. Yu., Fedotov, N. I., Nasretdinova, V. F. & ZaitsevZotov, S. V. Effect of surface defects and fewatomic steps on the local density of states of the atomicallyclean surface of Bi2Se3 topological insulator. JETP Lett. 100, 398–402 (2014).
 20.
Wang, J. et al. Powerlaw decay of standing waves on the surface of topological insulators. Phys. Rev. B 84, 235447 (2011).
 21.
Biswas, R. R. & Balatsky, A. V. Scattering from surface step edges in strong topological insulators. Phys. Rev. B 83, 075439 (2011).
 22.
Rizzo, B., Arrachea, L. & Moskalets, M. Transport phenomena in helical edge state interferometers: a Green’s function approach. Phys. Rev. B 88, 155433 (2013).
 23.
Fu, Z. G., Zhang, P., Wang, Z. & Li, S. S. Quantum corrals and quantum mirages on the surface of a topological insulator. Phys. Rev. B 84, 235438 (2011).
 24.
Fu, Z. G. et al. Anisotropic FabryProt resonant states confined within nanosteps on the topological insulator surface. Sci. Rep. 4, 5544 (2014).
 25.
Liu, Q., Qi, X.L. & Zhang, S.C. Stationary phase approximation approach to the quasiparticle interference on the surface of a strong topological insulator. Phys. Rev. B 85, 125314 (2012).
 26.
Anderson, P. W. Absence of diffusion in certain random lattices. Phys. Rev 109, 1492–1505 (1958).
 27.
Abrahams, E., Anderson, P. W., Licciardello, D. C. & Ramakrishnan, T. V. Scaling theory of localization: absence of quantum diffusion in two dimensions. Phys. Rev. Lett. 42, 673–676 (1979).
 28.
Ernst, A., Sandratskii, L. M., Bouhassoune, M., Henk, J. & Lüders, M. Weakly dispersive band near the fermi level of GaMnAs due to Mn interstitials. Phys. Rev. Lett. 95, 237207 (2005).
 29.
Okabayashi, J. et al. Angleresolved photoemission study of Ga_{1−x}Mn_{x}As. Phys. Rev. B 64, 125304 (2001).
 30.
Konig, E. J., Ostrovsky, P. M., Protopopov, I. V. & Mirlin, A. D. Metalinsulator transition in twodimensional random fermion systems of chiral symmetry classes. Phys. Rev. B 85, 195130 (2012).
 31.
Fu, L. & Kane, C. L. Topology, delocalization via average symmetry and the symplectic anderson transition. Phys. Rev. Lett. 109, 246605 (2012).
 32.
Ostrovsky, P. M., Gornyi, I. V. & Mirlin, A. D. Interactioninduced criticality in Z_{2} topological insulators. Phys. Rev. Lett. 105, 036803 (2010).
 33.
Ostrovsky, P. M., Gornyi, I. V. & Mirlin, A. D. Quantum criticality and minimal conductivity in graphene with longrange disorder. Phys. Rev. Lett. 98, 256801 (2007).
 34.
Ryu, S., Mudry, C., Obuse, H. & Furusaki, A. Z_{2} topological term, the global anomaly, and the twodimensional symplectic symmetry class of anderson localization. Phys. Rev. Lett. 99, 116601 (2007).
 35.
Xu, S.Y. et al. Hedgehog spin texture and Berry’s phase tuning in a magnetic topological insulator. Nat. Phys. 8, 616–622 (2012).
 36.
Xu, S.Y. et al. Dirac point spectral weight suppression and surface ‘gaps’ in nonmagnetic and magnetic topological insulators. Preprint at http://arxiv.org/abs/1206.0278v3 (2012).
 37.
Jozwiak, C. et al. Photoelectron spinflipping and texture manipulation in a topological insulator. Nat. Phys. 9, 293–298 (2013).
 38.
Miao, L. et al. Quasiparticle dynamics in reshaped helical Dirac cone of topological insulators. Proc. Natl Acad. Sci. 110, 2758–2762 (2013).
 39.
Dai, J.X. et al. Toward the intrinsic limit of the topological insulator Bi_{2}Se_{3}. Phys. Rev. Lett. 117, 106401 (2016).
 40.
Hsieh, D. et al. A tunable topological insulator in the spin helical Dirac transport regime. Nature 460, 1101–1105 (2009).
 41.
Checkelsky, J. G., Ye, J., Onose, Y., Iwasa, Y. & Tokura, Y. Diracfermionmediated ferromagnetism in a topological insulator. Nat. Phys. 8, 729–733 (2012).
 42.
Wray, L. A. et al. Observation of topological order in a superconducting doped topological insulator. Nat. Phys. 6, 855859 (2010).
 43.
Wray, L. A. et al. A topological insulator surface under strong Coulomb, magnetic and disorder perturbations. Nat. Phys. 7, 32–37 (2011).
 44.
Nomura, K., Koshino, M. & Ryu, S. Topological delocalization of twodimensional massless dirac fermions. Phys. Rev. Lett. 99, 146806 (2007).
 45.
Sessi, P. et al. Dual nature of magnetic dopants and competing trends in topological insulators. Nat. Commun. 7, 12027 (2016).
 46.
Bell, R. J. & Dean, P. Atomic vibrations in vitreous silica. Disc. Faraday Soc. 50, 55–61 (1970).
 47.
Thouless, D. J. Electrons in disordered systems and the theory of localization. Phys. Rep. 13C, 93–142 (1974).
Acknowledgements
We are grateful for discussions with W. Wu, P. Hohenberg and Y.D. Chuang. Work at Stanford University was supported by the Department of Energy Grant DEAC0276SF00515. R.R.B. was supported by Purdue University startup funds.
Author information
Affiliations
Department of Physics, New York University, New York, New York 10003, USA
 Yishuai Xu
 , Janet Chiu
 , Lin Miao
 , Haowei He
 & L. Andrew Wray
Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA
 Lin Miao
Massachusetts Institute of Technology, Department of Physics, Cambridge, Massachusetts 02139, USA
 Zhanybek Alpichshev
Department of Physics, Stanford University, Stanford, California 94305, USA
 Zhanybek Alpichshev
 & A. Kapitulnik
Department of Physics and Astronomy, Purdue University, West Lafayette, Indiana 47907, USA
 Rudro R. Biswas
NYUECNU Institute of Physics at NYU Shanghai, 3663 Zhongshan Road North, Shanghai 200062, China
 L. Andrew Wray
Authors
Search for Yishuai Xu in:
Search for Janet Chiu in:
Search for Lin Miao in:
Search for Haowei He in:
Search for Zhanybek Alpichshev in:
Search for A. Kapitulnik in:
Search for Rudro R. Biswas in:
Search for L. Andrew Wray in:
Contributions
Y.X. and J.C. carried out the numerical modelling with assistance from H.H. and guidance from R.R.B. and L.A.W.; Y.X., J.C., L.M., R.R.B. and L.A.W. participated in the analysis, figure planning and draft preparation; Z.A. and A.K. performed and analysed the STM experiments; L.A.W. was responsible for the conception and the overall direction, planning and integration among different research units.
Competing interests
The authors declare no competing financial interests.
Corresponding authors
Correspondence to Rudro R. Biswas or L. Andrew Wray.
Supplementary information
PDF files
 1.
Supplementary Information
Supplementary Figures, Supplementary Notes and Supplementary References.
 2.
Peer Review File
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Further reading

Impurity potential induced gap at the Dirac point of topological insulators with inplane magnetization
Physical Review B (2019)

Efficient ab initio calculations of electrondefect scattering and defectlimited carrier mobility
Physical Review Materials (2019)

Interface effects on the magneticproximityinduced quantized Hall response in heterostructures based on threedimensional topological insulators
Physical Review B (2019)

Electrostatic formation of the Majorana quasiparticles in the quantum dotnanoring structure
Journal of Physics: Condensed Matter (2019)

Connection topology of step edge state bands at the surface of a three dimensional topological insulator
New Journal of Physics (2018)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.