PHYS

ICS

Local optical control of ferromagnetism and chemicalpotential in a topological insulatorAndrew L. Yeatsa,b, Peter J. Mintuna, Yu Panc,d, Anthony Richardellac,d, Bob B. Buckleya, Nitin Samarthc,d,and David D. Awschaloma,b,1

aInstitute for Molecular Engineering, University of Chicago, Chicago, IL 60637; bMaterials Science Division, Argonne National Laboratory, Argonne, IL60439; cMaterials Research Institute, The Pennsylvania State University, University Park PA 16802; and dDepartment of Physics, The Pennsylvania StateUniversity, University Park PA 16802

Contributed by David D. Awschalom, August 14, 2017 (sent for review July 30, 2017; reviewed by Harold Y. Hwang and Christopher Palmstrom)

Many proposed experiments involving topological insulators (TIs)require spatial control over time-reversal symmetry and chemi-cal potential. We demonstrate reconfigurable micron-scale opti-cal control of both magnetization (which breaks time-reversalsymmetry) and chemical potential in ferromagnetic thin filmsof Cr-(Bi,Sb)2Te3 grown on SrTiO3. By optically modulating thecoercivity of the films, we write and erase arbitrary patterns intheir remanent magnetization, which we then image with Kerrmicroscopy. Additionally, by optically manipulating a space chargelayer in the underlying SrTiO3 substrates, we control the localchemical potential of the films. This optical gating effect allowsus to write and erase p-n junctions in the films, which we studywith photocurrent microscopy. Both effects are persistent andmay be patterned and imaged independently on a few-micronscale. Dynamic optical control over both magnetization and chem-ical potential of a TI may be useful in efforts to understand andcontrol the edge states predicted at magnetic domain walls inquantum anomalous Hall insulators.

topological insulators | ferromagnetism | Kerr microscopy |magneto-optical recording | photocurrent microscopy

The unusual topology of electronic bands in some mate-rials can produce quantum states that are stabilized by

symmetry. Topological insulators (TIs) have attracted partic-ular attention for their symmetry-protected surface and edgestates, which may hold promise for applications in spintron-ics and quantum computing (1, 2). Ferromagnetic TIs com-bine a topologically nontrivial band structure with ferromag-netism, which intrinsically breaks time-reversal symmetry (TRS)and therefore may produce unusual electromagnetic phenom-ena such as the topological magnetoelectric effect (3, 4), quan-tized Kerr and Faraday rotation (5, 6), and quantization ofthe anomalous Hall effect (QAHE) (7, 8). However, whilethe quantum states in an ideal TI are protected by symme-try, most topologically nontrivial materials have proven diffi-cult to grow and engineer. In particular, residual bulk con-ductivity, and the tendency of TI materials to degrade duringsemiconductor processing, have impeded many experimentalefforts (9–11). Harnessing the unique physics of these materi-als will require better understanding and control of their mag-netism and disorder, as well as methods to engineer TI deviceswithout degrading the properties that make these materialsunique.

A major challenge in TI research is to identify systems in whichthe energies and symmetries of TI physics can be reliably con-trolled in the laboratory. For instance, a criterion for the QAHinsulator state is for the chemical potential of electrons µ to liewithin an energy gap ∆ in the surface state dispersion, where∆∝ | ~M · n̂|, ~M is the local magnetization and n̂ is the surfacenormal (1, 2). At the edges of a uniformly magnetized film,the top and bottom surfaces meet and therefore ~M · n̂ changessign, causing ∆ to vanish. In sufficiently thin films, the vanish-ing gap produces 1D conductive states that propagate along the

edges of the film, and their quantized conductance is observedas the QAHE (7, 8). Similar edge states are predicted to occuralong magnetic domain walls, where the direction of magnetiza-tion inverts. These domain wall-bound states are of considerablepresent interest (12–16), but they have not yet been observeddirectly. Methods to intentionally pattern the magnetization andchemical potential of TI materials in situ during experiments maybe particularly useful in efforts to detect, and eventually to har-ness, these unique phenomena.

Much research has focused on understanding and minimizingthe sources of both magnetic (17–20) and chemical potential (21–23) disorder in TI materials. However, relatively few techniqueshave been explored to systematically control these parameters.Magnetic control of TI materials has focused largely on intro-ducing TRS-breaking in different layers along the growth axisof thin films by modulation doping of magnetic ions (24) or inmagnetic-nonmagnetic bilayer devices (25). Individual magneticdopants have also been arranged on TI surfaces with STM tech-niques (26).

Spatial control of chemical potential in TI materials is diffi-cult due to material degradation during top gate fabrication andother cleanroom processes (9–11). Previously, we reported anall-optical technique to tune the chemical potential of a varietyof ultra-thin materials grown on SrTiO3 (27). By optically manip-ulating space-charge in the underlying substrate, we created aneffective local back-gate that was strong enough to create p-njunctions in nonmagnetic thin films of (Bi,Sb)2Te3. These fea-tures were persistent and optically erasable, allowing arbitrary

Significance

Ferromagnetic topological insulators (TIs) have promise forapplications in spintronics, metrology, and quantum comput-ing. However, TI materials are fragile and often incompati-ble with nanofabrication techniques. Here, we demonstratepersistent, micron-scale optical control of both magnetizationand chemical potential in Cr-(Bi,Sb)2Te3 grown on SrTiO3. Weshow that this system is uniquely positioned to enable arbi-trary routing of the quantized edge states recently discoveredin magnetic TIs. We also use Kerr and photocurrent micro-scopies to image magnetic inversion dynamics, p-n junctions,and magnetic recordings that we make in these materials. Thiswork may enable dynamic, reconfigurable control of 1D quan-tum channels.

Author contributions: A.L.Y., P.J.M., and D.D.A. designed research; A.L.Y. and P.J.M. per-formed research; Y.P., A.R., and N.S. contributed new reagents/analytic tools; D.D.A.supervised the experiments and data analysis; A.L.Y., P.J.M., and B.B.B. analyzed data;and A.L.Y., P.J.M., Y.P., A.R., B.B.B., N.S., and D.D.A. wrote the paper.

Reviewers: H.Y.H., Stanford University; and C.P., University of California, Santa Barbara.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

1To whom correspondence should be addressed. Email: awsch@uchicago.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1713458114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1713458114 PNAS | September 26, 2017 | vol. 114 | no. 39 | 10379–10383

Dow

nloa

ded

by g

uest

on

May

25,

202

0

Hext

θK ∝MZ

x y

z

A

C

-1000 -500 0 500 1000

16.2

16.4

16.6

16.8

μ0Hext (Gauss)

R xx (k

Ω)

-2

-1

0

1

2

3T = 2.9 KVg = 0

R xy (k

Ω)

-6

-4

-2

0

2

4

6

θK (a.u.)

D

2HC

1 10 100 1000

10

100

MOKE Laser Power (μW)

B

Vg = 0

Vxx , Vxy

μ 0HC

(Gau

ss)

T = 3.2 KVg = 0

Fig. 1. Simultaneous MOKE and AHE measurements. (A) Magnetic hystere-sis loops showing anomalous Hall resistance Rxy (red squares) and Kerr angleθK (blue triangles) as a function of magnetic field in a film of Cr-(Bi,Sb)2Te3

at T = 2.9 K. MOKE was conducted with a spot size of ≈1 µm and a laserpower of≈1 µW. (B) Schematic of the measurement geometry. MOKE mea-surements were performed on the unused leg of the Hall bar to avoid opticaleffects on transport measurements. (C) Magnetic hysteresis in longitudinalresistance Rxx . (D) Coercive field HC of the film from MOKE hysteresis mea-surements as a function of MOKE laser power. The red arrow indicates HC

extracted from AHE hysteresis without illumination.

spatial control of the chemical potential without additional mate-rials or fabrication.

In this article, we demonstrate persistent, reconfigurable opti-cal control over both the electron chemical potential and localmagnetization of ferromagnetic thin films of Cr-(Bi,Sb)2Te3

grown on SrTiO3. First, we use simultaneous magneto-opticalKerr effect (MOKE) and anomalous Hall effect (AHE) mea-surements to establish a correspondence between optical andelectrical measures of TI magnetism. We image the dynamics ofmagnetic coercion in the films using scanning Kerr microscopy,which suggest the presence of a nanoscale domain structure. Wealso use both optical and electrostatic gating techniques to char-acterize both carrier-mediated and van Vleck-mediated ferro-magnetism in the films. Finally, we demonstrate the ability tooptically pattern and reconfigure the local chemical potential andlocal magnetization of the films using a two-color optical andmagneto-optical recording protocol. We characterize these coex-isting patterns with simultaneous scanning Kerr and photocur-rent microscopies, showing that independent spatial control ofboth chemical potential and magnetization can be achieved atthe same time in a ferromagnetic TI on a few-micron scale.

Results and DiscussionWe used simultaneous magneto-optical and magneto-transportmeasurements to characterize the ferromagnetism of thin filmsof Cr-(Bi,Sb)2Te3 at T = 2.9 K. Fig. 1A shows the evolutionof polar Kerr angle θK and anomalous Hall resistance Rxy ina representative 9 quintuple-layer (QL) film of Cr-(Bi,Sb)2Te3

as a function of increasing and decreasing magnetic field. Clear,roughly square-shaped hysteresis is visible in both MOKE andAHE traces, indicative of ferromagnetic ordering with an easyaxis out of the sample plane. We estimate the film has a Curietemperature TC ≈ 18 K (see SI Appendix). Fig. 1B shows themeasurement geometry. Fig. 1C shows the longitudinal resis-tance Rxx as a function of applied field at 2.9 K. The peaks at±HC have been observed in similar materials and are attributed

to increased scattering from magnetic disorder during coercion(28). Although similarly composed materials grown in the sameMBE chamber have shown quantized conductance at T� 1 K(16, 19, 29), the absolute resistance values we observe indicatethat conductivity is dominated by surface- and bulk-conductivechannels (as opposed to quantized edge states) at the tempera-tures achievable in our optical cryostat (T > 2.9 K).

Optical illumination during MOKE experiments could affectthe magnetism of the films. Fig. 1D shows the coercive field HC

extracted from MOKE hysteresis measurements as a function oflaser power at T = 3.2 K and with a spot size of ≈1 µm. Thecoercivity of the film decreases with illumination, fitting well to apower law relation (dashed line) from 20 µW to 2 mW, wherelocal heating is likely to dominate. The trend plateaus at lowpowers, showing good agreement with the coercivity extractedfrom AHE hysteresis in the absence of illumination (red arrow).This suggests that illumination from MOKE measurements con-ducted below ≈2 µW is unlikely to significantly affect the mag-netic properties of the film.

By rastering the MOKE laser spot across the sample at lowpower, it is straightforward to produce spatial images of themagnetization of the films. Fig. 2 shows scanning Kerr micro-graphs of the reversal of remanent magnetization in a thin filmof Cr-(Bi,Sb)2Te3 after 5-s exposures to successively larger fields.The images show a continuous and spatially uniform change inremanent magnetization with no obvious pinning effects nearthe edges of the film. This is most likely an indication that mag-netic inversion occurs on length scales smaller than the resolv-ing power of our apparatus (≈1 µm). This behavior is consis-tent with the submicron-size domains observed by Lachman et al.(19) by scanning nano-SQUID microscopy in similar materials.We see similar behavior in all of the films grown on SrTiO3. Bycontrast, we observe inversion by domain-wall motion on longerlength scales in one very heavily Cr-doped film grown on InP (seeSI Appendix). However, the heavily-doped sample is unlikely toexhibit QAHE.

Field-effect gating is generally needed to position the chemicalpotential within the surface state gap in QAHE studies (8). Weused electrostatic back-gating, as well as a substrate-based persis-tent optical gating effect, to characterize the response of the filmto the field effect. UV illumination produces a persistent field

ΔθK(arb.)

50 μm

12 Gauss 134 Gauss 280 Gauss

426 Gauss 571 Gauss 717 Gauss

Baresub-strate

TIfilm

Fig. 2. Kerr microscopy of remanent magnetization. Zero-field Kerrmicroscopy showing the reversal of remanent magnetization of a uniformlymagnetized Cr-(Bi,Sb)2Te3 thin film after exposure to successively largeropposing fields. The lower right corner shows a region of bare substratewhere the film was scratched away for reference. A zero-field baselineimage at the initial magnetization (not shown) was subtracted from eachimage for clarity, such that the color scale represents the change in rema-nent magnetization of the film ∆θK(H) = θK(H) − θK,initial. The laser powerwas ≈2 µW.

10380 | www.pnas.org/cgi/doi/10.1073/pnas.1713458114 Yeats et al.

Dow

nloa

ded

by g

uest

on

May

25,

202

0

PHYS

ICS

A

-30 0 30 60 90225

250

275

300

325 AHE vs. Vg

MOKE vs. Vg

AHE vs. UV

Back Gate Voltage (V)

Cumulative UV Dose (s)

R xx (k

Ω) R

xy (kΩ)

12

15

18

21

24

0 100 200

2

4

6

8

10

μ0 H

c (Gauss)

Rxx vs. Vg

Rxy vs. Vg

Rxx vs. UV Rxy vs. UV

BCumulative UV Dose (s)

0 100 200

-30 0 30 60 90Back Gate Voltage (V)

Fig. 3. Optical and electrostatic gate response. (A) Zero-field longitudi-nal resistance Rxx and symmetrized Hall resistance Rxy as a function of backgate voltage (red squares and blue triangles) or cumulative exposure to UVlight (green circles and purple diamonds). Zeros are offset to account forhysteresis in the back gate response. The saturation fields for AHE measure-ments were µ0Hext =±1000 G. (B) Coercive field HC extracted from AHE(red squares) and MOKE hysteresis (blue triangles) measurements, plottedas a function of back-gate voltage (bottom axis). The green circles show HC

extracted from AHE as a function of cumulative exposure to UV light (topaxis). The UV intensity was ≈10 mW/m2.

effect in thin films grown on SrTiO3 via electrical polarizationof the underlying substrate, and this effect is functionally equiv-alent to a an electrostatic back-gate (27). Fig. 3A shows the evo-lution of Rxx and Rxy as a function of either Vg (bottom axis) orprevious exposure to UV illumination (top axis). The resistancesevolve similarly with either type of gating. The response is con-sistent with expectations for an initially p-type film that is gatedjust past its Dirac point to a weakly n-type regime at Vg & 35 Vor &60 s of cumulative UV exposure. The similar response of theparameters to either type of gating suggests that any differencein gating strength between the top and bottom surfaces is compa-rable for electrostatic versus optical gating techniques. Becausethe optical gating effect is persistent, all measurements were per-

θK(arb.)

-Hc < Hrecord

< 0

Magneto-opticalrecording

Hinit

> Hc

Initialization

Hext = 0

Kerrmicroscopy

θK ∝ MZ

30 μm

4 K 8 K

12 K 17 K

Baresub-

strate

θK(arb.)

A

D E

B C

TIfilm

Fig. 4. Magneto-optical recording. (A–C) Protocol for magneto-opticalrecording in a magnetic TI. (A) A strong magnetic field (1,000 Gauss) isused to uniformly magnetize the film. (B) Laser illumination (32 µW) locallyreduces the coercivity of the film, allowing its magnetization to be invertedby a weak opposing field (−63 Gauss). (C) The pattern remains persistent atzero field and can be measured with scanning Kerr microscopy at low power(4 µW). (D) Scanning Kerr micrograph of magnetization pattern recorded ina thin film of Cr-(Bi,Sb)2Te3. (E) Magneto-optical recording as a function oftemperature. No effect is seen above TC or from the substrate in areas wherethe magnetic film has been scratched away.

formed in the dark ≈ 60 s after each exposure to avoid any tran-sient effects from illumination of the sample.

Carrier-independent magnetism is important for magnetic TIsto achieve the QAHE because magnetization must remain aftertuning the chemical potential to the surface-state energy gap (7,30). While carrier-mediated magnetism dominates in the morelightly Cr-doped films we have studied (see SI Appendix), in thepresent film we observe magnetism that is largely independentof gate-tuning by either optical or electrostatic effects. Fig. 3Bshows the evolution of HC (extracted from simultaneous AHEand MOKE measurements) as a function of electrostatic back-gating (bottom axis) or previous exposure to UV light (top axis).HC decreases by≈ 30% with increasing Vg (or previous exposureto UV light), consistent with only partially hole-mediated fer-romagnetism. This is consistent with other reports of coexistinghole-mediated and van Vleck-mediated ferromagnetism in mod-erately Cr-doped (Bi,Sb)2Te3 (31, 32). Notably, the values of HC

extracted from AHE and MOKE measurements correspond wellin the absence of a gate effect, but at higher voltages (or UVexposures), MOKE hysteresis measurements indicate a largerHC than AHE hysteresis measurements. This may imply thatoptically excited carriers from MOKE measurements can playa role in the carrier-mediated ferromagnetism of these materi-als, and this contribution is relatively more important when freeholes are otherwise depleted by a back-gate.

Many proposed applications of magnetic TI materials requirespatial control of TRS-breaking. Fig. 4 A–C shows a protocolfor magneto-optical recording (33, 34) in Cr-(Bi,Sb)2Te3. Afterinitialization with a field Hinit >HC , a high-intensity laser spotis used to locally reduce the coercivity of the film, such that aweak external field −HC <Hrecord < 0 will invert the magneti-zation of the film only in the illuminated area. When the laser

Magneto-opticalrecording

Kerr and photocurrentimaging

Iph dμ/dx

25 μm 25 μm Iph(pA)

Local gatepatterning

A

θK MZ

Mag. upMag. down

p-typen-type

x y

z

θK(arb.)

A

D E

B C

-50

50

0

Fig. 5. Simultaneous local control of magnetization and chemical poten-tial. (A–C) Schematic of the patterning and imaging procedure. (A) Threeconcentric circles are recorded in the magnetization of a Cr-(Bi,Sb)2Te3 filmusing the protocol shown in Fig. 4. (B) Focused UV illumination is used tolocally increase the chemical potential of the film in the exposed areas, cre-ating two n-type regions in the otherwise p-type film. (C) Scanning Kerr andphotocurrent microscopy are performed simultaneously to measure boththe local magnetization and local chemical potential gradient of the film.(D) Scanning Kerr micrograph of three concentric circles recorded in themagnetization of a thin film of Cr-(Bi,Sb)2Te3. (E) Scanning photocurrentmicrograph collected simultaneously with D, showing the photoresponse ofp-n junctions patterned in the same area. The image axes and pattern aretilted slightly with respect to the path between the two electrodes. Theimages were taken at zero field, Vg = 0, T = 3 K. A relatively high laserpower of 30 µW was used to produce a clear photocurrent image.

Yeats et al. PNAS | September 26, 2017 | vol. 114 | no. 39 | 10381

Dow

nloa

ded

by g

uest

on

May

25,

202

0

and field are shut off, the magnetic pattern remains and can beimaged with low-power scanning Kerr microscopy. Fig. 4D showsa Kerr micrograph of intentionally patterned magnetization inCr-(Bi,Sb)2Te3. The minimum feature size is ≈ 1 µm. We findthat magnetic patterns remain persistent >24 h unless they areerased with a coercive magnetic field or by raising the tempera-ture above TC . After erasure, no signs of the old pattern appearin subsequent Kerr images, and new patterns may be recorded.Fig. 4E shows magneto-optical recording as a function of temper-ature, indicating that the effect vanishes above TC . Moreover,patterns were written across the edge of the film, and no pat-terning effect was observed on the bare substrate area. The lackof response in these areas rules out other explanations due toeffects in the substrate, such as optical polarization of magneticimpurities (35).

In addition to controlling the local magnetization throughmagneto-optical recording, simultaneous patterning of the localchemical potential can be achieved by confining the substrate-based optical gating effect to a focused spot. Fig. 5 A–C showsa schematic of the patterning technique. First, magneto-opticalrecording is performed as described above to write three con-centric circles in the local magnetization of the film. Then, tworectangular regions are exposed to UV light, which creates aneffective local back-gate, raising the chemical potential of thefilm in those areas and causing them to become n-type. ScanningKerr (Fig. 5D) and photocurrent (Fig. 5E) microscopies are thenperformed simultaneously using the same focused laser spot, asdescribed in the Materials and Methods. These images provide amap of the local magnetization and chemical potential gradientof the film, respectively. The photocurrent micrograph is consis-tent with the presence of local n-type regions in the otherwisep-type film. The image axes and pattern are tilted slightly withrespect to the path between electrodes. The Kerr micrographindicates that the magnetic pattern has survived the optical gat-ing process, highlighting the presence of carrier-independent fer-romagnetism in these films, and the ability of the films to supportpersistent, micron-scale magnetic features regardless of carrierdensity.

Once written, both magnetic and chemical potential patternsare persistent in the dark and show no sign of degradationwhen imaged a day later. However, a relatively high laser power(30 µW) is needed to produce clear photocurrent images, andso the imaging process does relax the UV gating effect (27).Repeated imaging can therefore be used to erase chemicalpotential patterns in the field of view. The elevated laser powerduring magneto-optical recording may also relax the optical gat-ing effect, and so magnetic recording must be performed asthe first step in the process. Despite the elevated power dur-ing photocurrent imaging, the magnetic patterns do not appearto degrade so long as the external field is held at zero. Withinthese constraints, we find that magneto-optical recording andUV gate patterning combine well without noticeable cross-talk in

the Kerr and photocurrent images. The combination of these twoeffects therefore allows independent, reconfigurable optical con-trol of both the local magnetization (TRS-breaking) and chemi-cal potential of a magnetic TI material on a few-micron scale.

Materials and MethodsSamples consisted of 5 to 10 QL films of Crx(BiySb1−y )2−xTe3 grown bymolecular beam epitaxy (MBE) on (111)-oriented SrTiO3. Growth parame-ters and detailed material characterization data are presented in a sepa-rate manuscript (36). The composition of the films is comparable to pre-vious materials grown in the same MBE chamber that showed QAHE atvery low temperatures (T� 1 K) (16, 19, 29). For consistency, we presentdata in the main text from one representative sample, a 9 QL film ofCr0.15(Bi0.5,Sb0.5)1.85Te3 capped with a 4.5 nm layer of Al. Other films arediscussed in SI Appendix. Hall bars were patterned in all films by mechani-cally scratching away the growth layer with a needle, thereby avoiding sam-ple contamination from lithographic processes. The samples were mountedin an optical cryostat, and an electromagnet was used to apply an out-of-plane magnetic field at the sample.

The local magnetization of the film was imaged with scanning Kerrmicroscopy in the polar MOKE geometry. A beam of linearly polarized lightwas directed at normal incidence onto the sample, and the rotation in polar-ization θK of the reflected beam was measured using lock-in techniquesas described in SI Appendix. A reflective microscope objective was used toavoid unwanted Faraday rotation from glass optics (37). Illumination wasprovided by a HeNe laser, with a photon energy (~ ω= 1.96 eV) in the vicin-ity of a known resonance in the MOKE spectrum of similar materials (38).For simultaneous resistance, AHE, and MOKE hysteresis measurements, thelaser spot was positioned on an unconnected leg of the Hall bar to avoidoptical effects on the transport measurements.

Resistance measurements were performed using standard AC lock-intechniques with an excitation current of 100 nA and a reference frequencyof 17 Hz. Electrical contacts to all samples were made with indium and wereohmic. The chemical potential of the films was tuned by applying a voltageto the back side of the substrate or through a persistent, substrate-basedoptical gating effect (27). A fiber-coupled UV (λ= 370 nm) light-emittingdiode (LED) was used to provide illumination for optical gating. For simul-taneous Kerr and photocurrent imaging, a virtual null current preamplifierwas used to measure the photocurrent induced by the MOKE laser spot as itwas rastered across the sample.

Note Added to Proof. After submission of this manuscript, two complemen-tary preprints discussing domain-wall edge states have appeared online(39, 40).

ACKNOWLEDGMENTS. We thank J. van Bree for useful discussions. Thiswork is supported by Office of Naval Research Grants N00014-15-1-2369and N00014-15-1-2370, Air Force Office of Scientific Research Multidisci-plinary University Research Initiative Grant FA9550-14-1-0231, NSF Mate-rials Research Science and Engineering Centers Grant NSF-DMR-1420709,and National Science Foundation Grants NSF-DMR-1306300 and NSF-DMR-1306510. This study is based in part on research conducted at The Penn-sylvania State University 2D Crystal Consortium–Materials Innovation Plat-form (2DCC-MIP), which is supported by NSF Cooperative AgreementDMR-1539916. This material is based on work supported by LaboratoryDirected Research and Development (LDRD) funding from Argonne NationalLaboratory, provided by the Director, Office of Science, of the US Depart-ment of Energy under Contract DE-AC02-06CH11357.

1. Hasan MZ, Kane CL (2010) Colloquium: Topological insulators. Rev Mod Phys82(4):3045–3067.

2. Qi XL, Zhang SC (2011) Topological insulators and superconductors. Rev Mod Phys83(4):1057–1110.

3. Qi XL, Hughes TL, Zhang SC (2008) Topological field theory of time-reversal invariantinsulators. Phys Rev B 78(19):195424.

4. Essin AM, Moore JE, Vanderbilt D (2009) Magnetoelectric polarizability and axionelectrodynamics in crystalline insulators. Phys Rev Lett 102(14):146805.

5. Tse WK, MacDonald AH (2010) Giant magneto-optical Kerr effect and uni-versal faraday effect in thin-film topological insulators. Phys Rev Lett 105(5):057401.

6. Wu L, et al. (2016) Quantized Faraday and Kerr rotation and axion electrodynamicsof a 3D topological insulator. Science 354(6316):1124–1127.

7. Yu R, et al. (2010) Quantized anomalous Hall effect in magnetic topological insulators.Science 329(5987):61–64.

8. Chang CZ, et al. (2013) Experimental observation of the quantum anomalous Halleffect in a magnetic topological insulator. Science 340(6129):167–170.

9. Kong D, et al. (2011) Rapid surface oxidation as a source of surface degradation factorfor Bi2Se3. ACS Nano 5(6):4698–4703.

10. Benia HM, Lin C, Kern K, Ast CR (2011) Reactive chemical doping of the Bi2Se3 topo-logical insulator. Phys Rev Lett 107(17):177602.

11. Wang X, Bian G, Miller T, Chiang TC (2012) Fragility of surface states and robustnessof topological order in Bi 2Se 3 against oxidation. Phys Rev Lett 108(9):096404.

12. Wickles C, Belzig W (2012) Reflectionless transport of surface Dirac fermions on topo-logical insulators with induced ferromagnetic domain walls. Phys Rev B 86(3):035151.

13. Feng Y, et al. (2015) Observation of the zero hall plateau in a quantum anomaloushall insulator. Phys Rev Lett 115(12):126801.

14. Wang YH, et al. (2015) Observation of chiral currents at the magnetic domain bound-ary of a topological insulator. Science 349(6251):948–952.

15. Wang YH, et al. (2015) Retraction. Science 350(6267):1482–1482.16. Liu M, et al. (2016) Large discrete jumps observed in the transition between Chern

states in a ferromagnetic Topological Insulator. Sci Adv 2:e1600167.17. Okada Y, et al. (2011) Direct observation of broken time-reversal symmetry on the

surface of a magnetically doped topological insulator. Phys Rev Lett 106(20):206805.

10382 | www.pnas.org/cgi/doi/10.1073/pnas.1713458114 Yeats et al.

Dow

nloa

ded

by g

uest

on

May

25,

202

0

PHYS

ICS

18. Wang W, et al. (2015) Visualizing ferromagnetic domains in magnetic topologicalinsulators. APL Mater 3(8):083301.

19. Lachman EO, et al. (2015) Visualization of superparamagnetic dynamics in magnetictopological insulators. Sci Adv 1(10):e1500740.

20. Wang W, Chang CZ, Moodera JS, Wu W (2016) Visualizing ferromagnetic domainbehavior of magnetic topological insulator thin films. Npj Quan Mater 1(July):16023.

21. Roushan P, et al. (2009) Topological surface states protected from backscattering bychiral spin texture. Nature 460(7259):1106–1109.

22. Beidenkopf H, et al. (2011) Spatial fluctuations of helical Dirac fermions on the sur-face of topological insulators. Nat Phys 7(12):939–943.

23. Kastl C, et al. (2015) Chemical potential fluctuations in topological insulator (Bi 0.5 Sb0.5 ) 2 Te 3 -films visualized by photocurrent spectroscopy. 2D Mater 2(2):024012.

24. Mogi M, et al. (2015) Magnetic modulation doping in topological insulators towardhigher-temperature quantum anomalous Hall effect. Appl Phys Lett 107(18):182401.

25. Kou X, et al. (2013) Manipulating surface-related ferromagnetism in modulation-doped topological insulators. Nano Lett 13(10):4587–4593.

26. Schlenk T, et al. (2013) Controllable magnetic doping of the surface state of a topo-logical insulator. Phys Rev Lett 110:126804.

27. Yeats AL, et al. (2015) Persistent optical gating of a topological insulator. Sci Adv1(9):e1500640.

28. Checkelsky JG, Ye J, Onose Y, Iwasa Y, Tokura Y (2012) Dirac-fermion-mediated ferro-magnetism in a topological insulator. Nat Phys 8(10):729–733.

29. Kandala A, Richardella A, Kempinger S, Liu CX, Samarth N (2015) Giant anisotropicmagnetoresistance in a quantum anomalous Hall insulator. Nat Commun 6(May):7434.

30. Chang CZ, et al. (2013) Thin films of magnetically doped topological insulator withcarrier-independent long-range ferromagnetic order. Adv Mater 25(7):1065–1670.

31. Kou X, et al. (2013) Interplay between different magnetisms in Cr-doped topologicalinsulators. ACS Nano 7(10):9205–9212.

32. Richardella A, Kandala A, Samarth N (2015) Topological insulator thin films and het-erostructures: Epitaxial growth, transport, and magnetism. Topological Insulators:Fundamentals and Perspectives, eds Ortmann F, Roche S, Valenzuela SO (Wiley-VCHVerlag GmbH & Co. KGaA, Weinheim, Germany), pp 295–329.

33. Chen D, Ready JF, Bernal GE (1968) MnBi thin films: Physical properties and memoryapplications. J Appl Phys 39(8):3916–3927.

34. Lambert CH, et al. (2014) All-optical control of ferromagnetic thin films and nanos-tructures. Science 345(6202):1337–1340.

35. Rice WD, et al. (2014) Persistent optically induced magnetism in oxygen-deficientstrontium titanate. Nat Mater 13(March):481–487.

36. Richardella A, Kandala A, Lee JS, Samarth N (2015) Characterizing the structure oftopological insulator thin films. APL Mater 3(8):083303.

37. Kondo Y, Nakamura Y, Yamakawa K, Ishio S, Ariake J (2012) Development of micro-scopic magnetometer with reflective objective using magneto-optical Kerr effect. JAppl Phys 111(7):07E324.

38. Patankar S, et al. (2015) Resonant magneto-optic Kerr effect in the magnetic topo-logical insulator Cr(SbxBi1-x)2Te3. Phys Rev B 92(21):214440.

39. Rosen IT, et al. (2017) Chiral transport along magnetic domain walls in the quantumanomalous Hall effect. arXiv:1707.08677.

40. Yasuda K, et al. (2017) Quantized chiral edge conduction on reconfigurable domainwalls of a magnetic topological insulator. arXiv:1707.09105.

Yeats et al. PNAS | September 26, 2017 | vol. 114 | no. 39 | 10383

Dow

nloa

ded

by g

uest

on

May

25,

202

0