# Fileset

[ncomms7437.pdf](https://mdr.nims.go.jp/filesets/f0330b15-b239-4770-9b6c-a9186d53e002/download)

## Creator

Patrick Gallagher, Menyoung Lee, Trevor A. Petach, Sam W. Stanwyck, James R. Williams, [Kenji Watanabe](https://orcid.org/0000-0003-3701-8119), [Takashi Taniguchi](https://orcid.org/0000-0002-1467-3105), David Goldhaber-Gordon

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[A high-mobility electronic system at an electrolyte-gated oxide surface](https://mdr.nims.go.jp/datasets/e91ac570-0200-40eb-a24d-843c2f0e2753)

## Fulltext

A high-mobility electronic system at an electrolyte-gated oxide surfaceARTICLEReceived 3 Nov 2014 | Accepted 28 Jan 2015 | Published 12 Mar 2015A high-mobility electronic system at anelectrolyte-gated oxide surfacePatrick Gallagher1, Menyoung Lee1, Trevor A. Petach1, Sam W. Stanwyck2, James R. Williams1, Kenji Watanabe3,Takashi Taniguchi3 & David Goldhaber-Gordon1Electrolyte gating is a powerful technique for accumulating large carrier densities at a surface.Yet this approach suffers from significant sources of disorder: electrochemical reactionscan damage or alter the sample, and the ions of the electrolyte and various dissolvedcontaminants sit Angstroms from the electron system. Accordingly, electrolyte gating is wellsuited to studies of superconductivity and other phenomena robust to disorder, but of limiteduse when reactions or disorder must be avoided. Here we demonstrate that these limitationscan be overcome by protecting the sample with a chemically inert, atomically smooth sheetof hexagonal boron nitride. We illustrate our technique with electrolyte-gated strontiumtitanate, whose mobility when protected with boron nitride improves more than 10-fold whileachieving carrier densities nearing 1014 cm� 2. Our technique is portable to other materials,and should enable future studies where high carrier density modulation is required butelectrochemical reactions and surface disorder must be minimized.DOI: 10.1038/ncomms7437 OPEN1 Department of Physics, Stanford University, Stanford, California 94305, USA. 2 Department of Applied Physics, Stanford University, Stanford, California94305, USA. 3 Advanced Materials Laboratory, National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan. Correspondence and requestsfor materials should be addressed to D.G.-G. (email: goldhaber-gordon@stanford.edu).NATURE COMMUNICATIONS | 6:6437 | DOI: 10.1038/ncomms7437 | www.nature.com/naturecommunications 1& 2015 Macmillan Publishers Limited. All rights reserved.mailto:goldhaber-gordon@stanford.eduhttp://www.nature.com/naturecommunicationsAconventional field effect transistor is controlled by thevoltage on a metal electrode separated from the channelby a thin insulating dielectric. The maximum appliedvoltage is determined by the dielectric breakdown field, beyondwhich the resistance of the dielectric sharply drops, shorting themetal electrode to the channel. For a typical high-qualitydielectric, the breakdown field limits the accumulated carrierdensity to B1013 cm� 2 (ref. 1), although for special cases such asferroelectrics stronger modulation is possible2–4. Electrolytegating circumvents dielectric breakdown by eliminating themetal/dielectric interface: an electrolyte is applied directly to thesurface of interest and polarized, drawing one charged species tothe surface and building a large electric field5. Carrier densitiesB1015 cm� 2 can be induced by electrolyte gating6, facilitatingthe discovery of superconductivity in new parameter regimes7,8and the creation of novel photonic devices9, among otheradvances.While very effective at modulating surface properties, electro-lyte gating also introduces disorder. The deposition of con-taminants on the sample is difficult to control, a problem that iscompounded by the possibility of surface-degrading electroche-mical reactions. Recent studies have further suggested thatchemical modification of the surface of interest, rather thanelectrostatics, is primarily responsible for the marked changes inelectronic properties in some electrolyte-gated systems10–12.Motivated by these challenges, we consider the well-studiedtwo-dimensional electron system (2DES) created by electrolytegating at the surface of strontium titanate (STO)13–19.The transport properties of this surface 2DES closely resemblethose of the 2DES at the lanthanum aluminate/strontiumtitanate (LAO/STO) interface. However, the highest reportedlow-temperature electron mobility in the STO 2DES is about1,000 cm2 V� 1 s� 1, at an electron density of 3� 1013 cm� 2(refs 13,14,16,19); for the same density, the LAO/STO 2DES hasmobility up to 10,000 cm2 V� 1 s� 1 (ref. 20). We demonstratethat by protecting the STO channel with a thin boron nitride(BN) dielectric impermeable to the ions of the electrolyte21,the mobility of the resulting electrolyte-gated 2DESsubstantially increases over a wide density range, surpassing12,000 cm2 V� 1 s� 1 at a density of 4� 1013 cm� 2 in our bestsample.ResultsBN-protected STO samples. Each of our samples consists of asingle crystal of STO partially covered by an atomically flat BNflake (Fig. 1a). The BN flake conforms to the substrate withouttrapping contaminants, as evidenced by the 0.4 nm terrace stepsof the underlying STO seen in the topography of the BN (Fig. 1b).The substrate is masked by a thick insulator except in a Hall bar-shaped channel area (Fig. 1c); the electrolyte induces negligiblecarrier density in the masked regions. In this work, we considerfour BN-covered STO samples—denoted A, B, C and D—withBN thicknesses measured to be 0.6, 1.0, 1.2 and 1.5 nm, respec-tively, by atomic force microscopy (see Supplementary Note 1 forlateral dimensions and thickness measurement details). For eachsample, we collect low-temperature magnetotransport data overmultiple cooldowns at different coplanar gate voltages Vgate.Mobility and carrier density. The striking improvement in 2DESquality with a BN spacer is evident in the magnetotransportproperties of Sample A, which is covered by a 0.6-nm thick flake(Fig. 2a,b). The five cooldowns of Sample A, numbered 1 through5, correspond to different Vgate settings. Although higher Vgatetypically induces higher density, this is not always the casebecause of hysteresis (see Methods) and because of drifting offsetvoltages from electrochemical reactions at the gate electrode. Toextract density and mobility, we perform a simultaneous fit to thesheet resistance rxx and the Hall coefficient RH�rxy/m0H, whererxy is the Hall resistance, m0 is the magnetic constant, and H is theapplied magnetic field. As is typical in the STO 2DES literature,we assume that the magnetotransport behaviour can be describedby two bands22. Although quantum oscillation data suggestseveral bands (discussed below), a two-band description often fitsthe data, providing reliable numbers for average mobility andtotal density (Supplementary Note 2). For LAO/STO, a two-bandfit with four parameters (densities n1, n2, mobilities m1, m2)captures the approximate shapes of rxx and RH, but deviates fromthe data at higher fields22. We encounter the same difficulty: thetwo-band model cannot simultaneously fit the nonsaturatinglinear magnetoresistance and nearly saturated Hall coefficientobserved up to 31 T in our samples (Supplementary Note 2) andin LAO/STO samples23. Inclusion of a third band cannotgenerally reproduce our high-field data, and where a three-bandfit does work, the required densities are unrealistically large,frequently exceeding 1016 cm� 2 with mobility B1 cm2 V� 1 s� 1.We instead fit to a two-band model in which the sheetresistance of each band contains a term linear in applied field:Coplanar gateSTOOhmic contactsInsulatingmaskBN flakeab cFigure 1 | Electrolyte gating with a boron nitride barrier. (a) Schematicrepresentation of a device fabricated on a single crystal of strontiumtitanate (STO). In operation, the entire device is submerged in ionic liquid(not shown), which is polarized by the coplanar gate. (b) Atomic forcemicrograph (topography) of a few-layer boron nitride (BN) flake (left half ofimage) on an STO crystal. STO terraces (0.4-nm steps) run bottom left totop right, and are visible beneath the BN, indicating that the flake conformsto the substrate with few trapped impurities. Scan window is 1 mm by 1mm.(c) Optical micrograph of Sample A, which has a cross-linked PMMA mask(darker brown regions; the relative lightness here is opposite to that in a,where to aid visualization the flake is darker than the mask). The thin BNflake is not visible on STO, but covers the entire opening in the PMMAmask, except near the contacts. Scale bar: 10mm.ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms74372 NATURE COMMUNICATIONS | 6:6437 | DOI: 10.1038/ncomms7437 | www.nature.com/naturecommunications& 2015 Macmillan Publishers Limited. All rights reserved.http://www.nature.com/naturecommunicationsrxx,i¼ 1/niemiþ kiH for i¼ 1, 2 and kiZ0. The linear term couldarise from spatial fluctuations in mobility24,25.Our two-band fits with linear magnetoresistance provide anexcellent match to the data (Fig. 2a,b). These fits exclude thelow-field region, where the magnetotransport properties areaffected by magnetic moments in the STO (ref. 26 andSupplementary Note 2). We find a high-mobility band withdensity n1 between 6� 1012 and 5� 1013 cm� 2 (Fig. 2c) andmobility m1 between 8,000 and 17,000 cm2 V� 1 s� 1 (Fig. 2d), aswell as a low-mobility band with a similar density n2 and mobilitym2 that grows with decreasing n2. The total induced densityntot can reach 9� 1013 cm� 2 (Fig. 2c) with an averagemobility mavg¼ (n1m1þ n2m2)/ntot approaching 8,000 cm2 V� 1 s� 1(Fig. 2d). The average mobility for Cooldown 4 exceeds12,000 cm2 V� 1 s� 1. These mobilities match (for lower densities)and exceed (for higher densities) the highest reported mobilitiesin LAO/STO 2DES20,27, and are 10 times larger than themobilities reported in the literature for electrolyte-gated STO2DES at any carrier density13,14,19. Our conclusions areunchanged if we instead calculate m and n by naively dividingRH by rxx, or if we fit with the four-parameter, two-band model(Supplementary Note 2).Quantum oscillations. Quantum oscillations appear above B3 Tin both rxx and rxy for all cooldowns. The rxy oscillations fromCooldown 1 (Fig. 2e) show a primary oscillation frequency of50 T (Fig. 2f), corresponding to a carrier density near 2�1012 cm� 2. This contrasts with the results of the two-band Halltransport fits, in which both bands are at least 10 times morepopulated. For a typical cooldown, we can identify multiplequantum oscillation frequencies corresponding to densitiesB1012 cm� 2, regardless of the total density measured by the Halleffect. The strongest oscillations thus appear for the lowest Halldensities (see Cooldown 5 in Fig. 2a), as the bands that producequantum oscillations now constitute a substantial fraction of thecarriers. Our findings resemble quantum oscillation data collectedon the highest mobility LAO/STO 2DES, in which multiple bandsof density B1012 cm� 2 show quantum oscillations, and totalHall densities B1013 cm� 2 or lower are required for strongoscillations in rxx (refs 28,29). The presence of low-densityoscillating bands does not strongly impact the shapes of rxx andRH, so the two-band model still captures most of the devicebehaviour (Supplementary Note 2).DiscussionThe maximum mobility that we have achieved in each of our fourBN-covered samples is significantly higher than the maximummobility that we have achieved in any uncovered STO sample(Fig. 3a). The mobility improvement with BN results in part fromthe added separation between the 2DES and the disorderedcharges in the electrolyte. As discussed below, we also expect thatthe BN acts as a barrier to surface-degrading chemical reactionsthat occur during electrolyte gating or during processing. Ourlimited sample size produces enough scatter in the maximummobility as a function of thickness that we cannot identify themain sources of residual disorder.A single layer of graphene is known to be permeable toprotons30 but impermeable to other small chemical species,including He atoms31 and Liþ ions32. Because BN has a latticestructure nearly identical to that of graphene, we anticipate asimilar diffusion resistance for even our thinnest BN barriers. Theenergy barrier to diffusion is so high (ref. 32 calculates 10 eV forLiþ across graphene) that we still expect impermeability withVgate dropped across our BN. An electrolyte-gated gold samplecovered by 6 nm of BN behaved in accordance with theseexpectations: a gold oxide film is readily grown on uncovered3002001000FFT mag (a.u.)10080604020Frequency (T )−0.50.00.5Δd� xy/d� 0H (Ω/T)0.50.40.30.20.1(�0H )−1 (T−1)1482� (103 cm2 V−1 s−1)54321Cooldown index   8642n (1013 cm−2)54321Cooldown index   30252015101050�0H (T)�0H (T)123457060504030201010501 (2.7)2 (2.25)3 (2)4 (1.5)5 (1.8)a b cdefT = 40 mKT = 40 mKCooldown 1� xx (Ω)RH (Ω/T)�avg�1�2ntotn1n2Figure 2 | High-mobility magnetotransport in Sample A over a range of densities. (a) Sheet resistance rxx (symmetrized in field) for 5 cooldowns,labelled 1, 2, 3, 4 and 5, with Vgate¼ 2.7, 2.25, 2, 1.5 and 1.8 V, respectively. Dashed black curves result from two-band fits with linear magnetoresistance,performed simultaneously on the data in a and b. The BN is 0.6 nm thick. (b) Hall coefficient RH�rxy/m0H (symmetrized in field) for the same 5 cooldownsas in a and fits (dashed black curves). (c) Extracted carrier densities n1 and n2 for the two bands for each cooldown; ntot¼ n1þ n2. (d) Extracted carriermobilities m1 and m2 for the two bands for each cooldown; mavg¼ (n1m1þ n2m2)/ntot. (e) Quantum oscillations in drxy/dm0H as a function of inverse appliedmagnetic field for Cooldown 1. For clarity, the signal has been smoothed and a quadratic background has been subtracted. The oscillations commence atB3 T for all cooldowns. (f) Magnitude of the Fourier transform of e, showing a peak at B50 T, corresponding to a density B2� 1012 cm� 2.NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7437 ARTICLENATURE COMMUNICATIONS | 6:6437 | DOI: 10.1038/ncomms7437 | www.nature.com/naturecommunications 3& 2015 Macmillan Publishers Limited. All rights reserved.http://www.nature.com/naturecommunicationsgold samples12, but the BN-covered gold sample was unmodified(Supplementary Note 3). The chemical species responsible for theredox reaction is unknown, but these results nonetheless illustratethat BN can limit chemical reactions during electrolyte gating.An intriguing possibility for electrolyte-gated oxides is that BNbarriers could prevent oxygen removal. Experiments on rutileTiO2 single crystals10 and VO2 thin films11 have found evidencethat oxygen near the crystal surface diffuses out through theelectrolyte, calling into question the relative roles of oxygenvacancy creation and electrostatic carrier accumulation in tuningthe properties of oxide materials. An electrolyte-gating study ofSTO found that injecting oxygen gas into the electrolytesuppresses the source-drain current, which was interpreted asevidence that the otherwise observed carrier accumulation resultsfrom oxygen vacancies33. Another study of STO concluded thatvery high gate voltages are required to create oxygen vacancies,and that the reduced STO system (density B1015 cm� 2) is three-dimensional and remains conductive at zero gate voltage14. Whilewe cannot directly prove the absence of oxygen migration whengating BN-protected STO, we verify that electrostatic carrieraccumulation can account for our data by considering theapparent capacitance between the electrolyte and the 2DES,defined as Capparent¼ entot/Vgate. If electrostatics alone isresponsible for the carrier accumulation, a naive model suggeststhat Capparent should fall below a serial arrangement of twocapacitances: that of the double layer formed by the ions(12 mF cm� 2, ref. 34), and that of the BN dielectric. This yieldsC� 1max ¼ 12 mFcm� 2ð Þ� 1þ 4E0=tð Þ� 1, where 4 is the dielectricconstant of BN and t is the sheet thickness. On the other hand,if carriers accumulate by chemical modification, Capparent isunrestricted.For all BN-covered samples, Capparent falls near or below Cmax,and two orders of magnitude below Capparent for uncovered gold,whose surface is chemically modified by electrolyte gating12. Thecapacitance to the channel from the large coplanar gate, locatedo200 mm away, accounts for the violation of the electrostaticlimit in Samples A (0.6 nm) and C (1.2 nm). Due to thelow-temperature dielectric constant of 25,000 in STO and thefocusing of field lines from the large gate onto the much smallerHall bar35, this capacitance can be as large as several mFcm� 2.We have measured such a capacitance on some samples byzeroing the coplanar gate voltage at low temperature. However,modulating the gate voltage at low temperature appears to causemechanical problems as our ionic liquid droplet unfreezes onwarmup. We therefore did not collect coplanar gate capacitancedata for most samples and cannot quantitatively correct Cmax.Our device geometry exposes some area of our contact metaldirectly to the electrolyte (Fig. 1a), limiting Vgate to about 3 V:above this, chemical reactions readily occur with the contactmetal. This limitation in turn limits the maximum thickness ofBN that can be used to create a metallic STO 2DEG. The lowestdensity for which we have measured a conducting state in ourSTO 2DES is 1013 cm� 2, although the mobility edge may besomewhat lower. To accumulate 1013 cm� 2 electrostaticallyrequires a minimum capacitance of 0.5 mF cm� 2, or a maximumBN thickness of 7 nm. Although we have not studied such thickBN flakes, we have measured several samples which had wrinklesin the BN several nm tall due to the transfer process(Supplementary Note 4). When these wrinkles cut fully acrossthe current path between source and drain, the sample neverconducted between source and drain. Presumably the areabeneath the wrinkles remained insulating, in approximatenumerical agreement with the electrostatic accumulation picture.The BN barrier need not be kept thin if all conductive materialcan be masked. This is often difficult in insulators, since theelectrolyte must create a conductive path between the devicechannel and metal contacts, unless the insulator can bechemically doped near the contacts. For intrinsically metallicsystems, it is straightforward to mask all conductive area (see ourBN on gold sample, Supplementary Note 3). In this case, highervoltages can in principle be applied without chemical reactions,increasing the maximum thickness of BN that can be used for atarget electron density, which may have advantages for certainmaterials. Our technique is easily applied to other systems, andshould enable electrolyte gating experiments that require highcarrier mobility, high carrier density and chemical stability of thesurface.MethodsSample fabrication. Our samples were fabricated on (100) strontium titanatesubstrates from either Shinkosha Co. (Japan) or Crystec GmbH (Germany); thevendor for each sample is specified in Supplementary Table 1. The surfaces ofShinkosha crystals were TiO2-terminated as received. We prepared a nominallyTiO2-terminated surface on the Crystec samples by the method described in ref. 36.A BN flake was transferred onto the STO surface using the water-based processdescribed in ref. 37, followed by an anneal for 4 h at 500 �C in an Ar/O2atmosphere. An ohmic contact pattern was defined in PMMA via e-beamlithography at 10 kV, after which the sample was milled with Ar ions at 300 V toetch away the exposed BN and about 40 nm of the underlying STO. Ohmiccontacts (10 nm titanium, 40 nm gold) were then deposited into the milled trenchesby e-beam evaporation. Finally, an insulating mask with holes to expose the Hallbar and the coplanar gate was patterned using 10 kV e-beam lithography. The maskmaterial was either cross-linked PMMA or sputtered alumina; in both cases anegative process was used so that the channel was not exposed to the e-beam.Low-temperature measurement. Before measurement of each sample, we cleanedthe sample surface of resist residues by a brief exposure to a remote oxygen plasma.We then covered the Hall bar and coplanar gate with a drop of the ionic liquid1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)amide (EMI-TFSI) andplaced the sample inside the vacuum chamber of our cryostat (either a dilutionrefrigerator with base temperature 40 mK or a variable-temperature insert reaching350 mK or 1.5 K). We polarized the electrolyte at around 290 K, in either highvacuum or helium vapour, by applying a voltage to the coplanar gate. On cooling,the polarized electrolyte froze and we collected magnetotransport data up to thehighest available fields (9, 14, or 31 T) via standard lock-in techniques in a current-biased configuration. We typically used an AC source current of 2 mA, which1410621.510.50BN thickness (nm)110100Capparent (μF cm−2)1.510.50BN thickness (nm)AuSTO (chemical)STO (electrostatic)BN/STOa b� avg (103  cm2  V−1 s−1)Figure 3 | Properties of all measured samples. (a) Best average mobilitymavg, extracted from two-band fits with linear magnetoresistance, recordedover all cooldowns at the various boron nitride (BN) thicknesses studied.We have included our highest mobility uncovered strontium titanate (STO)sample (BN thickness zero). Error bars indicate thickness uncertainty in ouratomic force microscope measurements (Supplementary Note 1). (b)Apparent capacitance Capparent¼ entot/Vgate versus BN thickness for allcooldowns on all BN-covered STO samples (filled red circles). The totaldensity ntot is extracted from two-band fits with linear magnetoresistance.Dashed black line is the maximum capacitance Cmax for electrostatic carrieraccumulation. For comparison, we include Capparent for the bare STOsamples from ref. 14; those which were determined to be chemicallymodified (open purple circles) fall above Cmax, while those modulatedprimarily by electrostatics (open red circles) fall below Cmax. We also showCapparent for the uncovered gold sample from ref. 12 (open black square),which falls far above Cmax.ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms74374 NATURE COMMUNICATIONS | 6:6437 | DOI: 10.1038/ncomms7437 | www.nature.com/naturecommunications& 2015 Macmillan Publishers Limited. All rights reserved.http://www.nature.com/naturecommunicationsexceeded the superconducting critical current in all samples, suppressing thesuperconducting features that would otherwise appear for some cooldowns inFig. 2 (see also Supplementary Note 5). The sample was then warmed to near roomtemperature, melting the electrolyte. We always set Vgate to zero in betweencooldowns, which introduces some hysteresis in Vgate.References1. Ahn, C. H., Triscone, J.-M. & Mannhart, J. Electric field effect in correlatedoxide systems. Nature 424, 1015–1018 (2003).2. Ahn, C. H. et al. Electrostatic modulation of superconductivity in ultrathinGdBa2Cu3O7–x Films. Science 284, 1152–1155 (1999).3. Takahashi, K. S. et al. Local switching of two-dimensional superconductivityusing the ferroelectric field effect. Nature 441, 195–198 (2006).4. Boucherit, M. et al. Modulation of over 1014 cm� 2 electrons in SrTiO3/GdTiO3heterostructures. App. Phys. Lett. 104, 182904 (2014).5. Fujimoto, T. & Awaga, K. Electric-double-layer field-effect transistors withionic liquids. Phys. Chem. Chem. Phys. 15, 8983–9006 (2013).6. Yuan, H. et al. High-Density Carrier Accumulation in ZnO Field-EffectTransistors Gated by Electric Double Layers of Ionic Liquids. Adv. Func. Mater.19, 1046–1053 (2009).7. Ueno, K. et al. Discovery of superconductivity in KTaO3 by electrostatic carrierdoping. Nat. Nanotechnol. 6, 408–412 (2011).8. Ye, J. T. et al. Superconducting dome in a gate-tuned band insulator. Science338, 1193–1196 (2012).9. Zhang, Y. J., Oka, T., Suzuki, R., Ye, J. T. & Iwasa, Y. Electrically switchablechiral light-emitting transistor. Science 344, 725–728 (2014).10. Schladt, T. D. et al. Crystal-facet-dependent metallization in electrolyte-gatedrutile TiO2 single crystals. ACS Nano 7, 8074–8081 (2013).11. Jeong, J. et al. Suppression of metal-insulator transition in VO2 by electricfield-induced oxygen vacancy formation. Science 339, 1402–1405 (2013).12. Petach, T. A., Lee, M., Davis, R. C., Mehta, A. & Goldhaber-Gordon, D.Mechanism for the large conductance modulation in electrolyte-gated thin goldfilms. Phys. Rev. B 90, 081108 (2014).13. Ueno, K. et al. Electric-field-induced superconductivity in an insulator. Nat.Mater. 7, 855–858 (2008).14. Ueno, K., Shimotani, H., Iwasa, Y. & Kawasaki, M. Electrostatic chargeaccumulation versus electrochemical doping in SrTiO3 electric double layertransistors. App. Phys. Lett. 96, 252107 (2010).15. Lee, Y. et al. Phase diagram of electrostatically doped SrTiO3. Phys. Rev. Lett.106, 136809 (2011).16. Lee, M., Williams, J. R., Zhang, S., Frisbie, C. D. & Goldhaber-Gordon, D.Electrolyte gate-controlled kondo effect in SrTiO3. Phys. Rev. Lett. 107, 256601(2011).17. Li, M., Graf, T., Schladt, T. D., Jiang, X. & Parkin, S. S. P. Role of percolation inthe conductance of electrolyte-gated SrTiO3. Phys. Rev. Lett. 109, 196803(2012).18. Stanwyck, S. W., Gallagher, P., Williams, J. R. & Goldhaber-Gordon, D.Universal conductance fluctuations in electrolyte-gated SrTiO3 nanostructures.App. Phys. Lett. 103, 213504 (2013).19. Ueno, K. et al. Effective thickness of two-dimensional superconductivity in atunable triangular quantum well of SrTiO3. Phys. Rev. B 89, 020508 (2014).20. Huijben, M. et al. Defect engineering in oxide heterostructures by enhancedoxygen surface exchange. Adv. Func. Mater. 23, 5240–5248 (2013).21. Chuang, H.-J. et al. High mobility wse2 p- and n-type field-effect transistorscontacted by highly doped graphene for low-resistance contacts. Nano Lett. 14,3594–3601 (2014).22. Joshua, A., Pecker, S., Ruhman, J., Altman, E. & Ilani, S. A universal criticaldensity underlying the physics of electrons at the LaAlO3/SrTiO3 interface. Nat.Commun. 3, 1129 (2012).23. Ben Shalom, M., Ron, A., Palevski, A. & Dagan, Y. Shubnikov-De HaasOscillations in SrTiO3/LaAlO3 Interface. Phys. Rev. Lett. 105, 206401 (2010).24. Parish, M. & Littlewood, P. Non-saturating magnetoresistance in heavilydisordered semiconductors. Nature 426, 162–165 (2003).25. Kozlova, N. V. et al. Linear magnetoresistance due to multiple-electronscattering by low-mobility islands in an inhomogeneous conductor. Nat.Commun. 3, 1097 (2012).26. Joshua, A., Ruhman, J., Pecker, S., Altman, E. & Ilani, S. Gate-tunable polarizedphase of two-dimensional electrons at the LaAlO3/SrTiO3 interface. Proc. NatlAcad. Sci. USA 110, 9633–9638 (2013).27. Xie, Y., Bell, C., Hikita, Y., Harashima, S. & Hwang, H. Y. Enhancing electronmobility at the LaAlO3/SrTiO3 interface by surface control. Adv. Mater. 25,4735–4738 (2013).28. McCollam, A. et al. Quantum oscillations and subband properties of thetwo-dimensional electron gas at the LaAlO3/SrTiO3 interface. APL Mater. 2,022102 (2014).29. Xie, Y. et al. Quantum longitudinal and Hall transport at the LaAlO3/SrTiO3interface at low electron densities. Solid State Commun. 197, 25–29 (2014).30. Hu, S. et al. Proton transport through one-atom-thick crystals. Nature 516,227–230 (2014).31. Bunch, J. S. et al. Impermeable atomic membranes from graphene sheets. NanoLett. 8, 2458–2462 (2008).32. Das, D., Kim, S., Lee, K.-R. & Singh, A. K. Li diffusion through doped anddefected graphene. Phys. Chem. Chem. Phys. 15, 15128–15134 (2013).33. Li, M. et al. Suppression of ionic liquid gate-induced metallization ofSrTiO3(001) by oxygen. Nano Lett. 13, 4675–4678 (2013).34. Ohno, H. Electrochemical Aspects of Ionic Liquids (Wiley, 2011).35. Rakhmilevitch, D. et al. Anomalous response to gate voltage application inmesoscopic LaAlO3/SrTiO3 devices. Phys. Rev. B 87, 125409 (2013).36. Connell, J. G., Isaac, B. J., Ekanayake, G. B., Strachan, D. R. & Seo, S. S. A.Preparation of atomically flat SrTiO3 surfaces using a deionized-water leachingand thermal annealing procedure. App. Phys. Lett. 101, 251607 (2012).37. Amet, F., Williams, J. R., Watanabe, K., Taniguchi, T. & Goldhaber-Gordon, D.Insulating Behavior at the Neutrality Point in Single-Layer Graphene. Phys.Rev. Lett. 110, 216601 (2013).AcknowledgementsWe thank Thomas Schladt and Tanja Graf for helpful discussions at an early stage of thiswork and Harold Hwang for a careful reading of our manuscript. Sample fabrication wassupported by the Air Force Office of Science Research, Award No. FA9550-12-1-02520.Sample measurement was supported by the MURI program of the Army Research Office,Grant No. W911-NF-09-1-0398. Development of the ionic liquid gating technique wassupported by the Center on Nanostructuring for Efficient Energy Conversion (CNEEC)at Stanford University, an Energy Frontier Research Center funded by the U.S.Department of Energy, Office of Basic Energy Sciences under Award No. DE-SC0001060.P.G. acknowledges support from the DOE Office of Science Graduate FellowshipProgram. M.L. acknowledges support from Stanford University. J.R.W. and D.G.-G.acknowledge support from the W.M. Keck Foundation. A portion of our samplefabrication and characterization was performed at the Stanford Nano Center (SNC)/Stanford Nanocharacterization Laboratory (SNL), part of the Stanford Nano SharedFacilities. A portion of our measurements was performed at the National High MagneticField Laboratory, which is supported by National Science Foundation CooperativeAgreement No. DMR-1157490, the State of Florida, and the U.S. Department of Energy.Authors contributionsP.G., J.R.W. and D.G.-G. designed the experiment. P.G. fabricated the BN on STOsamples and performed the measurements, with help from M.L., S.W.S. and J.R.W. P.G.,M.L. and D.G.-G. analysed the data. T.A.P. performed the BN on gold experiment. K.W.and T.T. grew the BN crystals. P.G. prepared the manuscript with input from all authors.Additional informationSupplementary Information accompanies this paper at http://www.nature.com/naturecommunications.Competing financial interests: The authors declare no competing financial interests.Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions.How to cite this article: Gallagher, P. et al. A high-mobility electronic system at anelectrolyte-gated oxide surface. Nat. Commun. 6:6437 doi: 10.1038/ncomms7437 (2015).This work is licensed under a Creative Commons Attribution 4.0International License. The images or other third party material in thisarticle are included in the article’s Creative Commons license, unless indicated otherwisein 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/NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7437 ARTICLENATURE COMMUNICATIONS | 6:6437 | DOI: 10.1038/ncomms7437 | www.nature.com/naturecommunications 5& 2015 Macmillan Publishers Limited. All rights reserved.http://www.nature.com/naturecommunicationshttp://www.nature.com/naturecommunicationshttp://npg.nature.com/reprintsandpermissionshttp://npg.nature.com/reprintsandpermissionshttp://creativecommons.org/licenses/by/4.0/http://www.nature.com/naturecommunications title_link Results BN-protected STO samples Mobility and carrier density Figure™1Electrolyte gating with a boron nitride barrier.(a) Schematic representation of a device fabricated on a single crystal of strontium titanate (STO). In operation, the entire device is submerged in ionic liquid (not shown), &!QJ;which is polarized  Quantum oscillations Discussion Figure™2High-mobility magnetotransport in Sample A over a range of densities.(a) Sheet resistance rgrxx (symmetrized in field) for 5 cooldowns, labelled 1, 2, 3, 4 and 5, with Vgate=2.7, 2.25, 2, 1.5 and 1.8thinspV, respectively. Dashed black curves resul Methods Sample fabrication Low-temperature measurement Figure™3Properties of all measured samples.(a) Best average mobility mgravg, extracted from two-band fits with linear magnetoresistance, recorded over all cooldowns at the various boron nitride (BN) thicknesses studied. We have included our highest mobili AhnC. H.TrisconeJ.-M.MannhartJ.Electric field effect in correlated oxide systemsNature424101510182003AhnC. H.Electrostatic modulation of superconductivity in ultrathin GdBa2Cu3O7-x FilmsScience284115211551999TakahashiK. S.Local switching of two-dimensiona We thank Thomas Schladt and Tanja Graf for helpful discussions at an early stage of this work and Harold Hwang for a careful reading of our manuscript. Sample fabrication was supported by the Air Force Office of Science Research, Award No. FA9550-12-1-025 ACKNOWLEDGEMENTS Authors contributions Additional information