In-plane anisotropy in two-dimensional electron gas at LaAlO3/SrTiO3(110) interface
Shen Sheng-Chun, Hong Yan-Peng, Li Cheng-Jian, Xue Hong-Xia, Wang Xin-Xin, Nie Jia-Cai†,
Department of Physics, Beijing Normal University, Beijing 100875, China

 

† Corresponding author. E-mail: jcnie@bnu.edu.cn

Project supported by the Ministry of Science and Technology of China (Grant Nos. 2013CB921701, 2013CBA01603, and 2014CB920903), the National Natural Science Foundation of China (Grant Nos. 10974019, 51172029, 91121012, 11422430, 11374035, 11474022, and 11474024), the Program for New Century Excellent Talents in the University of the Ministry of Education of China (Grant No. NCET-13-0054), and the Beijing Higher Education Young Elite Teacher Project, China (Grant No. YETP0238).

Abstract
Abstract

A systematic study of the two-dimensional electron gas at LaAlO3/SrTiO3(110) interface reveals an anisotropy along two specific directions, [001] and . The anisotropy becomes distinct for the interface prepared under high oxygen pressure with low carrier density. Angular dependence of magnetoresistance shows that the electron confinement is stronger along the direction. Gate-tunable magnetoresistance reveals a clear in-plane anisotropy of the spin–orbit coupling, and the spin relaxation mechanism along both directions belongs to D’yakonov–Perel’ (DP) scenario. Moreover, in-plane anisotropic superconductivity is observed for the sample with high carrier density, the superconducting transition temperature is lower but the upper critical field is higher along the direction. This in-plane anisotropy could be ascribed to the anisotropic band structure along the two crystallographic directions.

1. Introduction

Two-dimensional electron gas (2DEG) formed at the interface between two band insulators LaAlO3 (LAO) and SrTiO3 (STO) exhibits many fascinating properties, such as magnetism,[1] superconductivity,[2,3] and their coexistence.[46] Besides, the observation of the spin–orbit coupling (SOC) in this system provides possibilities to control the orbital motion of electrons by acting on their spins.[79] Initially, the vast majority of the investigations were specific only to the LAO/STO(001) interfaces.[110] However, recent findings have shown that 2DEGs can be formed at interfaces with different crystal orientations, such as LAO/STO(110).[11,12] and LAO/STO(111),[12] opening fresh perspectives to understand the fundamental physics and to offer new functionalities.

Via x-ray linear dichroism spectroscopy, Pesquera et al. recently demonstrated an intrinsic distinctive electronic structure of the 2DEG at the LAO/STO(110) interface due to the different orbital symmetry and hierarchy compared to those of the (001)-oriented interface.[13] Therefore, the physical properties, such as superconductivity and SOC,[1417] could be influenced by the distinctive orbital symmetry and hierarchy. Latest studies on the LAO/STO(110) interfaces revealed distinct superconductivity and Rashba SOC,[16] and the SOC is insensitive to the gate voltage owing to the stronger polarizations of dxy, dyz/dzx orbitals along the confinement direction. Moreover, for the LAO/STO(110) interface, there are two different crystal directions, i.e., [001] and , in the plane of the interface. Recent angle-resolved photoemission (ARPES) studies on SrTiO3(110) surface showed a high electronic in-plane anisotropy along the two specific directions, and the anisotropy is pronounced at lower carrier density when only dyz/dzx-derived ellipsoids are occupied.[18] Hence, the in-plane crystallographic directions of LAO/STO(110) may offer an extra degree of freedom to tune the properties of 2DEG. A few of the latest works mentioned the existence of the in-plane anisotropy,[11,17] however a systematic study of the in-plane anisotropy in the (110) system is still required. Here, we present a careful and systematic research of the influence of the crystallographic direction on both normal state and superconducting properties at the LAO/STO(110) interfaces.

2. Experimental method

The samples with ten unit cells of LAO were grown at two different oxygen partial pressures (PO2), 10−4 Torr and 10−3 Torr, on treated (110)-oriented SrTiO3 substrates by pulsed laser deposition. The details of the growth procedure have been given in Ref. [15]. After preparation, two Hall bar structures were patterned in the same sample by using a standard photolithography technique, the LAO layers and a few STO layers were removed by low energy argon ion milling. For the directional dependence magnetotransport measurements, the Hall bar structures were along [001] and crystallographic directions, respectively. An etched STO surface was insulating, indicating that the conducting channels were at the interfaces. Ohmic contacts between electrode pad and sample interface were made by wire bonding. Magnetotransport properties were measured at temperature T = 2 K and with a magnetic field up to 9 T by a commercial Physical Properties Measurement System (Quantum Design, PPMS) equipped with a rotating sample holder. The gate was applied to the backside of the SrTiO3 substrate by a voltage source and no leakage current (< 1 nA) was detected for ±100 V gate voltages (VG).

3. Results and discussion

The anisotropic out-of-plane magnetoresistance (MR) along two directions, and [001], is shown in Fig. 1. The MR is defined as ΔR/R0 = [R(B) − R(B = 0)]/R(B = 0), where R(B) is the sheet resistance in the magnetic field B. As we can see from Fig. 1(a), for the sample prepared at PO2 ∼ 10−4 Torr, the MR linearly increases with increasing magnetic field for both directions. The anisotropy is not obvious; but for the interface prepared at PO2 ∼ 10−3 Torr, the anisotropy of the MR becomes obvious, as shown in Fig. 1(b). Along the direction, the MR firstly increases and then decreases with increasing magnetic field. These results indicate that the anisotropy along the two directions is pronounced at lower carrier density, which is consistent with the ARPES results on the STO(110) surface, attributed to the occupation of orbitals.[18] In order to effectively exploit the anisotropy of the 2DEG at the LAO/STO(110) interface, we now focus on the sample prepared at high PO2 (10−3 Torr) with a relatively low carrier density.

Fig. 1. The magnetoresistance measured along (black) and [001] (red) directions of LAO/STO(110) for samples prepared under oxygen partial pressures of (a) 10−4 Torr and (b) 10−3 Torr.

Anisotropic MR was investigated for different angle θ between the direction of the interface normal and the magnetic field. For the investigation of angular MR, the magnetic field was rotated from out-of-plane (θ = 0°, i.e., perpendicular to the surface) to in-plane (θ = 90°, parallel to the current) at 2 K. As shown in Fig. 2(a), along the direction, as the magnetic field increases, the MR gradually increases in the low field regime and then starts to decrease in the higher field. There is a local MR maximum, a positive slope in the low magnetic field regime but a negative slope in the high magnetic field regime. Besides, a negative MR appears for θ > 60° and the negative MR regime widens out as θ increases. Moreover, with increasing θ, the maximum of the MR slightly shifts to low magnetic field and the value becomes smaller. Along the [001] direction, the MR decreases as θ increases from 0° to 90°; and the local MR maximum appears at large angles, as shown in Fig. 2(b). The behavior of the in-plane MR is similar, while an obvious difference of MR along two directions can be observed at θ = 0° (perpendicular magnetic field), indicating that the anisotropy is more pronounced in the plane of the interface.

Fig. 2. The magnetoresistance for different angles θ along (a) and (b) [001] directions.

The positive out-of-plane MR is attributed to the orbital effect, while the in-plane MR rules out this effect. The difference of out-of-plane MR between [001] and directions suggests that crystallographic anisotropy plays a critical role in magnetotransport properties. The negative MR appears and becomes visible as the parallel magnetic field increases, as shown in Fig. 2. The negative MR can be counteracted by the positive contribution for the out-of-plane field case. When applying the parallel magnetic field, the suppression of positive MR indicates that the confinement zone of 2DEG is within one carrier mean free path. We assume that only one subband is occupied, then the mean-free path can be naively calculated by using formula ,[19] where l is the mean-free path, h is the Planck constant, e is the electron charge, Rs is the sheet resistance, is the Fermi wave number, and ns is the carrier density. Then at T = 2 K, the mean-free path is found to be about 22 nm along the direction, while it is about 34 nm along the [001] direction, the same order of magnitude compared with that reported by Ben Shalom et al. for (001)-orientation LAO/STO interfaces. It is noted that the mean-free path is shorter along the direction than that of [001], suggesting that scattering becomes stronger along the direction.

The remarkable angular dependence of MR and the appearance of negative MR indicate that the 2D weak localization effect should be considered. The magnetic field dependence of MR provides an important insight into the coupling between spin dynamics and transport. The existence of a local maximum MR and the appearance of a negative MR were also observed in other two-dimensional disordered systems,[2022] which is associated with the weak localization related to the presence of spin–orbit interactions (weak antilocalization).[2326] In LAO/STO heterostructures, the local electric field is enhanced by the accumulation of electrons in the interfacial quantum well, giving rise to the SOC. A strong spin–orbit field has been demonstrated at both (001) and (110)-oriented LAO/STO interfaces.[7,8,16,17]

For the investigation of directional dependence of SOC, we measured the MR as a function of VG, as shown in Fig. 3. Due to the high dielectric constant of SrTiO3,[27] the properties of 2DEG can be effectively tuned by a backgate voltage.[7,8] As shown in Fig. 3(a), the sheet resistances along two directions vary as VG changes. Also the sheet resistance along the direction is more sensitive to the applied electric field and is larger than that along the [001] direction. Figures 3(b) and 3(c) show the gate-tunable MR at different VG along two directions. Along the direction, the MR decreases with increasing VG, the local MR maximum shifts to low magnetic field, while along the [001] direction, the local MR maximum appears only for negative VG.

Fig. 3. (a) The sheet resistance as a function of VG along two directions. The magnetoresistance at different VG along (b) and (c) [001] directions.

The conductance correction due to the weak localization is modified by the SOC. The influence of the SOC can therefore be assessed by properly analyzing the magnetoconductance in the diffusion regime. The magnetic field dependence of the conductance correction normalized by the quantum of conductance as a function of VG is investigated, as shown in Fig. 4. We fit the experimental data by using the Maekawa–Fukuyama (MF) form[8,28]

where Δσ (B) = σ (B) −σ(0) is the correction to the conductance, σ(B) is the magnetoconductance under magnetic field B, G0 = e2/πh is the quantum of conductance, Ψ(x) is defined as Ψ(x) = ln (x) + ψ(1/2+1/x), and ψ(x) is the digamma function. The parameters in Eq. (1) are the inelastic field Bϕ = ħ/(4eDτi) and the spin–orbit field BSO = ħ/(4eDτSO). The parameter γ = BB/4eDBSO is the Zeeman correction factor in which g-factor enters, where μB is the Bohr magneton, D = πħ2nsμ/(m*e) is the diffusion constant, τi and τSO are the relaxation times related to inelastic and spin–orbit terms, respectively. ns is the sheet carrier density and the effective mass of the electron takes m* = 3me at both directions.[29] It is worth mentioning that we also used the Hikami–Larkin–Nagaoka (HLN) formula.[23] to fit the experimental data. By fitting the temperature dependence of the magnetoconductance correction, the MF equation gives more reasonable parameters, especially the BSO, compared with those of the HLN equation.

Fig. 4. The out-of-plane magnetoconductance (normalized by the quantum of conductance) at different VG along (a) and (b) [001] directions. The dark grey dashed lines are the fitting lines by using Eq. (1). (c) The obtained fitting parameters BSO (filled square) and Bϕ (open circle) as a function of VG along (black) and [001] (red) directions. (d) The τSO as a function of VG along two directions, the inset shows the τi versus VG. Comparing τSO with τi at different VG along (e) and (f) [001] directions. Spin relaxation time vs. elastic scattering rate along two directions is shown in the insets of panels (e) and (f). The lines are a guide to the eye.

The evolution of the fitting parameters as a function of VG is shown in Fig. 4(c). Along both directions, the inelastic field Bϕ decreases as VG increases. When increasing VG, the sheet resistance decreases and the carrier density increases, so the inelastic scattering weakens. Along the direction, with increasing VG, the spin–orbit field BSO decreases slightly, while along the [001] direction, the spin–orbit field increases as VG increases. For both directions, the BSO changes slightly with varying VG. This insensitive gate tunable SOC may be attributed to the strong atomic orbital polarizations of dxy and dxz/dyz orbitals along the confinement direction.[16] The corresponding scattering time as a function of VG is shown in Fig. 4(d). Along both directions, τi decreases with increasing VG, and τi is smaller than τSO (Figs. 4(e) and 4(f)), suggesting that the effect of the SOC is stronger compared with the orbital effect of the magnetic field. Moreover, a significant difference of spin–orbit term BSO is observed. The τi is shorter along the direction, suggesting that the weak localization is stronger than that of the [001] direction, which is consistent with the difference of sheet resistivity (Fig. 3(a)). For spin relaxation processes, there are mainly two typical scenarios: the Elliott–Yafet (EY) mechanism and the D’yakonov–Perel’ (DP) mechanism.[3032] In the latter mechanism, the spin relaxation time τSO is expected to be inversely proportional to the elastic scattering time τ. As shown in the insets of Figs. 4(e) and 4(f), τSO is proportional to the inverse of τ over a wide range of VG, characterizing the Rashba spin–orbit interactions. Besides, we note a deviation for VG > 50 V along the [001] direction, which may be due to the same order of τ and τSO. For intuitive comparison, all transport parameters for the sample prepared under the high oxygen pressure are listed in Tables 1 and 2.

Table 1.

Transport parameters along [001] direction.

.
Table 2.

Transport parameters along direction.

.

Moreover, for the PO2 = 10−4 Torr sample, down to mK magnitude of temperature, the superconductivity emerges, while for the PO2 = 10−3 Torr sample, the superconductivity is absent. To our surprise, the superconductivity is anisotropic along the two directions, as shown in Fig. 5. The superconducting transition temperature (Tc) is lower but the out-of-plane critical field (Hc2,⊥) is larger along the direction than that along [001] (inset of Fig. 5). The temperature dependence of Hc2,⊥ is extracted from the MR curves measured at different temperatures, where Hc2,⊥ is defined as the perpendicular magnetic field in which the sheet resistance of interfaces reaches to half of the normal state resistance taken at a given temperature. Generally, Hc2,⊥ should be proportional to Tc. We speculate that the unconventional larger Hc2,⊥ along the direction is due to the effect of disorder.[33,34] The above MR results suggest the strong localization along the direction. The deviation at low temperatures indicates that the effect of disorder overtakes the thermal fluctuation.[35] The mechanism of in-plane anisotropy of superconductivity needs further investigation and will be discussed elsewhere.

Fig. 5. Temperature dependence of the sheet resistance along two directions. The inset shows temperature dependence of the out-of-plane critical field, taken at 50% of normal resistance.

The anisotropic behavior along two directions and [001] at the LAO/STO(110) interfaces may be due to the difference of Ti t2g (i.e., dxy, dyz, and dzx) orbitals. For LAO/STO(110), contrary to (001), the lower energy states have dyz/dzx character.[13] The ARPES results performed on SrTiO3(110) surfaces reveal that the dyz/dzx-derived Fermi surface is anisotropic.[18] For the interface prepared at low PO2 (e.g., 10−4 Torr), the carrier density is high (in our case, ∼ 6 × 1013 cm−2), and then both dxy and dyz/dzx subbands are filled, the anisotropy weakens. However, for the interface prepared at high PO2 (e.g., 10−3 Torr), the carrier density is low (∼ 2.2 × 1013 cm−2), only dyz/dzx derived ellipsoids appear, the anisotropy is obvious. Along the direction, the dyz/dzx like subbands dispersing is weak and the effective electron mass is semiheavy, while the dyz/dzx like subbands dispersing is strong and the effective electron mass is light along the [001] direction.[17,18] The difference of dyz/dzx like subbands along the two directions may influence the properties of 2DEGs at the LAO/STO(110) interfaces.

4. Conclusion

We demonstrate the in-plane anisotropic behavior of 2DEG at LAO/STO(110) interfaces by performing magnetotransport and superconducting measurements. The anisotropy becomes more obvious for samples of low carrier density. We find that the SOC is anisotropic and insensitive to the gate voltage, while the spin relaxation mechanism is the DP mechanism along both directions. The effect of localization is stronger along the direction. Moreover, the anisotropic superconductivity is also observed, lower Tc but larger Hc2,⊥ is shown along the direction. Such crystallographic tunable properties expand the possibilities for electronic engineering of two-dimensional electron gas at the LaAlO3/SrTiO3 interface.

Reference
1Brinkman AHuijben MVan Zalk MHuijben JZeitler UMaan J Cvan der Wiel W GRijnders GBlank D HHilgenkamp H 2007 Nat. Mater. 6 493
2Reyren NThiel SCaviglia A DKourkoutis L FHammerl GRichter CSchneider C WKopp TRuetschi A SJaccard DGabay MMuller D ATriscone J MMannhart J 2007 Science 317 1196
3Caviglia A DGariglio SReyren NJaccard DSchneider TGabay MThiel SHammerl GMannhart JTriscone J M 2008 Nature 456 624
4Bert J AKalisky BBell CKim MHikita YHwang H YMoler K A 2011 Nat. Phys. 7 767
5Dikin DMehta MBark CFolkman CEom CChandrasekhar V 2011 Phys. Rev. Lett. 107 056802
6Li LRichter CMannhart JAshoori R C 2011 Nat. Phys. 7 762
7Shalom M BSachs MRakhmilevitch DPalevski ADagan Y 2010 Phys. Rev. Lett. 104 126802
8Caviglia AGabay MGariglio SReyren NCancellieri CTriscone J M 2010 Phys. Rev. Lett. 104 126803
9Fête AGariglio SCaviglia ATriscone J MGabay M 2012 Phys. Rev. 86 201105
10Ohtomo AHwang H Y 2004 Nature 427 423
11Annadi AZhang QRenshaw Wang XTuzla NGopinadhan KLu W MRoy Barman ALiu Z QSrivastava ASaha SZhao Y LZeng S WDhar SOlsson EGu BYunoki SMaekawa SHilgenkamp HVenkatesan TAriando 2013 Nat. Commun. 4 1838
12Herranz GSanchez FDix NScigaj MFontcuberta J 2012 Sci. Rep. 2 758
13Pesquera DScigaj MGargiani PBarla AHerrero-Martín JPellegrin EValvidares SGázquez JVarela MDix N 2014 Phys. Rev. Lett. 113 156802
14Herranz GBergeal NLesueur JGazquez JScigaj MDix NSanchez FFontcuberta J2013arXiv:1305.2411
15Han Y LShen S CYou JLi H OLuo Z ZLi C JQu G LXiong C MDou R FHe LNaugle DGuo G PNie J C 2014 Appl. Phys. Lett. 105 192603
16Herranz GSingh GBergeal NJouan ALesueur JGazquez JVarela MScigaj MDix NSanchez FFontcuberta J 2015 Nat. Commun. 6 6028
17Gopinadhan KAnnadi AKim YSrivastava AKumar BChen J SCoey J M DAriando Venkatesan T 2015 Adv. Electr. Matters 1500114 1
18Wang Z MZhong Z CHao X FGerhold SStoger BSchmid MSanchez-Barriga JVarykhalov AFranchini CHeld KDiebold U 2014 Proc. Natl. Acad. Sci. USA 111 3933
19Shalom M BTai CLereah YSachs MLevy ERakhmilevitch DPalevski ADagan Y 2009 Phys. Rev. 80 140403
20Komnik Y FAndrievskii V VBerkutov I B 2007 Low Temp. Phys. 33 79
21Kawaguti TFujimori Y 1983 J. Phys. Soc. Jpn. 52 722
22Spirito DDi Gaspare LEvangelisti FDi Gaspare AGiovine ENotargiacomo A 2012 Phys. Rev. 85 235314
23Hikami SLarkin A INagaoka Y 1980 Prog. Theor. Phys. 63 707
24Bergman G 1982 Phys. Rev. Lett. 48 1046
25Lee P ARamakrishnan T V 1985 Rev. Mod. Phys. 57 287
26Stornaiuolo DGariglio SFête AGabay MLi DMassarotti DTriscone J M 2014 Phys. Rev. 90 235426
27Neville R CMead C AHoeneise B 1972 J. Appl. Phys. 43 2124
28Maekawa SFukuyama H 1981 J. Phys. Soc. Jpn. 50 2516
29Mattheiss L 1972 Phys. Rev. 6 4740
30Dyakonov MPerel V1972Soviet Physics Solid State USSR133023
31Elliott R J 1954 Phys. Rev. 96 266
32Yafet Y1963Solid State Phys.141
33Gurevich APatnaik SBraccini VKim K HMielke CSong XCooley L DBu S DKim D MChoi J HBelenky L JGiencke JLee M KTian WPan X QSiri AHellstrom E EEom C BLarbalestier D C 2004 Supercond. Sci. Tech. 17 278
34Bugoslavsky YCohen LPerkins GPolichetti MTate TGwilliam RCaplin A 2001 Nature 411 561
35Fisher D SFisher M PHuse D A 1991 Phys. Rev. 43 130