Introduction

Interfaces between perovskite oxides have drawn much interest in pursuit of novel functionalities. The enticement of this system comes not only from the wide-ranging electronic properties of the parental bulk, but from the fact that they have common (quasi-)cubic structure with a similar lattice constant. This means that high-quality epitaxial interfaces are obtained with a wide variety of combinations of functional oxides, opening the way to pursue new devices by design. The conductivity at the LaAlO3/SrTiO3 (LAO/STO) heterointerface is a striking example of an emergent state at a perovskite heterointerface. This interface shows high mobility electron transport although both constituent layers are band insulators.[1] Various exotic phenomena, including two-dimensional superconductivity, two-dimensional Shubnikov-de Haas oscillations and magnetism are observed at low temperatures. The emergence of these phenomena is regarded to be intimately linked to the carrier distribution and the carrier density n, after several methods have been found to be important in determining n and the electronic properties, such as a variation of the growth conditions, thickness of the LAO layer, surface adsorbates and so on. In order to further push this system to its limits, a scalable and clean way of tuning the carrier density is crucial, in respect of gaining both fundamental understanding of the electron correlations and for device applications. This thesis presents two strategies to modulate the carrier density at the LAO/STO interface: delta-doping and electrostatic modulation.

Delta-doping the LAO/STO interface

In order to achieve scalable carrier density modulation while maintaining high-mobility transport, epitaxial LaTiO3 (LTO) layers were employed as dopant layers. LTO is a perovskite Mott insulator well lattice-matched to STO, known to donate electrons when embedded in STO heterostructures.[2, 3] This structure maintains the consistency of the conduction band originated from Ti 3d-electrons and minimizes unwanted scattering centers which have been a problem faced by using relatively lightly-doped conducting layers attempted so far.[4, 5] All samples with the epitaxial structure LAO (x uc) /LTO (y uc)/STO were grown by pulsed laser deposition (PLD), by depositing an LTO layer with thickness of y unit cells (uc) on a single crystalline STO with TiO2 {001}-termination, then depositing an x-uc-thick LAO [Fig. 1(a) inset]. Focusing on the x = 3 series, the temperature dependence of the sheet resistances is shown in Fig. 1(a) for typical samples with (x, y) = (3, 0.25), (3, 0.5), (3, 1) and (3, 3). All show a monotonically decreasing sheet resistance down to ~ 100 Ω/sq. at the temperature T = 2 K. A nonlinear Hall resistance vs. magnetic field was observed for all samples, which could be well fitted using a two-carrier model where two parallel carrier layers with different carrier densities, n1 and n2, and mobilities, μ1 and μ2, respectively, are assumed, which we ascribe to the spatial variation of the carrier density in the depth direction [3]. n(tot) = n1+n2 and μave = (n1μ1+n2μ2) / (n1+n2) as well as n2 and μ2 are shown for a various y in Fig. 1(b). n(tot) rapidly increased with small y, reaching values of n(tot) ~ 1×10(14) cm(-2), before saturating above y = 0.5 uc. Combining the systematic variation of n(tot) and conductivity with variation of the (x, y) shown in Fig. 2, we could determine the role of the LTO as a dopant layer in this system. In regard to the polar discontinuity between LAO and STO {001}, one of the intriguing possible origins of the Q2DEG in this system, we can consider LTO simply as a polar layer akin to LAO, with alternating {001} layers of (La(3+)O(2-))+ and (Ti(3+)O2(4-))− in the ionic limit. This assignment can be ruled out however, since we would expect that conductivity is observed for x + y ≳ 3 uc, leading to a phase boundary that follows the dashed line in Fig. 2, which is in disagreement with the data. Instead we can consider LTO acting as a dopant layer due to the constituent Ti(3+) ions, which form 3d 1 electrons that can migrate in(tot)he empty 3d 0 orbitals in the STO and make n significantly higher than for typical LAO/STO samples without LTO.

Next we investigated the effect of the LTO insertion on the field-effect response using a gate contact on the back side of the STO substrate. The gate voltage VG dependences of the carrier density and mobility for a (x, y) = (3, 0.25) sample at T = 2 K are shown in Fig. 3. The Hall non-linearity was observed for positive gate voltages. n(tot) shown here is the total density determined after the two-carrier analyses. For this (3, 0.25) sample, a change from n(tot) = 4.9×10(13) cm(−2) at VG = 0 V could be suppressed down to 2.4×10(13) cm(−2) at VG = −100 V, in good agreement with the value expected from a simple capacitor model employing a STO dielectric constant εSTO = 20000, and substrate thickness 0.5 mm, as depicted as a dashed gray line in Fig. 3. These data were quite distinct to those of LAO/STO where a relatively small modulation of n and large change in μ is observed. For VG > 0, both n1 and n2 increased, suggesting spreading of the carrier distribution in(tot)he depth of the substrate, where the mobility is relatively enhanced.

Electric field carrier modulation via top gate

The field-effect transistors (FETs) with Hall-bar structure were made on LAO/STO samples with various LAO thickness using standard optical photolithography and a gold top gate [Fig. 4(a), (b)]. Room temperature transport characteristics of the sample with 7-uc-thick LAO are shown in Fig. 4(c). The drain-source current ID was clearly modulated by the top-gate voltage VGS in a region where ID is independent of VDS, indicating the drain-source channel is pinched off at the saturated current IDSS, similar to the characteristics derived from the commonly known gradual channel model where n is ideally controlled by VGS and μ is assumed to stay constant. IDSS showed the typical quadratic dependence on VGS above the threshold voltage VTH [Fig .4(d)]. At a lower temperature, modulation of n by VG was demonstrated by Hall measurements in a 16-uc-thick LAO sample [Fig. 5]. Interestingly, μ showed a clear increase with decreasing n, which is the opposite tendency to back gate modulation [6]. This could be understood in two ways; one is that population of electrons at the immediate interface, which have lower mobility due to interfacial scattering potential, was reduced by negative VGS, resulting in increase of the average mobility. The other is that the resulting three-dimensional electron density is smaller with smaller VGS, and the reduced electron-electron interactions resulted in a higher mobility. In reality both of these effects are likely to occur simultaneously, and the potential significance here is that systematic n suppression achieved by top gate voltage at all temperature range, providing access to a regime with both smaller carrier density and less disorder, where other methods have experienced difficulties in approaching.

Conclusion

We have demonstrated carrier tuning at the LAO/STO interfaces by delta-doping and top gating the interfaces. By inserting a LTO layer, the carrier density of the LAO/STO interfaces could be increased accordingly to the thickness of the LTO layer, and further modulated by electric field applied to the back gate. Top gating was found to be a robust way in modulating carrier density, showing ideal device characteristics at room temperature, and also clear n modulation at all temperatures. The variation of the electron density and accompanying mobility change of several samples studied in this Thesis at T = 2 K are shown in Fig. 6. The open symbols connected with a line show the modulation by the top/back gate in a single sample whose ungated data are indicated with solid symbols. These data suggest that a wide-range modulation of the carrier density from 8×10(13) to 2×10(14) cm(-2) was achieved by combining the methods established in this Thesis. One intriguing perspective to mention is that top and back gating show opposite trends in the mobility modulation, suggesting that by applying both simultaneously, completely independent control of the carrier density and mobility, or carrier distribution can be achieved, providing a powerful tool to investigate the nature of the LAO/STO interface.

Fig. 1. (a) Temperature dependence of the sheet resistance and (b) the Hall carrier density and Hall mobility for representative samples with a fixed LAO thickness of 3 uc. The inset of (a) shows a schematic of the structure of the samples. Data for LAO (10 uc)/STO are also plotted at y = 0 for reference. Lines are guides.

Fig. 2. Conductivity variation with different LAO and LTO thicknesses. Circles represent metallic conductivity, and crosses insulating behavior. Observed phase boundary is indicated with a solid line. Dashed line shows the predicted phase boundary x + y ≳ 3 uc if the LTO is assumed as a simple polar stack.

Fig. 3. VG dependence of the n(tot) and μ(ave) of the LAO (3 uc)/LTO (0.25 uc)/STO sample extracted from a two-carrier fit at T = 2 K. Dashed gray line shows the slope expected from the capacitance between the interface and gate contact; other lines are guides.

Fig. 4. The FET structure. (a) Top view picture and (b) cross-section view along the dotted line in (a). α denotes amorphous. (c) Room temperature ID-V(DS) curves of the FET device with 7-uc-thick LAO, taken with the 0.2 V steps of V(GS) showed a wide-ranged saturation region. (d) The transfer curve of the same device showed an ideal quadratic behavior of ID when VGS was increased over V(TH) ~ -0.5 V.

Fig. 5. n and μ vs. VGS at lower temperature. V(TH) stayed around -1.8 V at the range of temperature as indicated with fitted dashed lines, suggesting robust n modulation is achieved. Mobility enhancement is clear. Solid lines are guides.

Fig. 6. n and μ achieved in the current study. Solid marks show data ungated, while open ones are their variation by applying top/back gate voltage to a single sample. All data were taken at T = 2 K.

本論文は、ペロブスカイト酸化物のLaAlO3/SrTiO3ヘテロ接合において、LaTiO3の数単位格子層の挿入およびゲート電極の採用により、伝導性界面におけるキャリア濃度およびキャリア移動度を広範に制御できることを実証した研究成果を述べたものである。本文は英文で記され、全7章からなる。

第1章は序論であり、ペロブスカイト酸化物およびそのヘテロ界面に関する従来の物性研究を概観し、界面に特有の性質として発現する特性を利用した新規デバイスの可能性など本研究の背景を説明したうえで、本研究の目的を述べている。

第2章では、本研究で扱うヘテロ界面における物理現象および物質に関する基礎事項を説明している。本来絶縁体である酸化物LaAlO3とSrTiO3のヘテロ界面において導電層が生じる機構、および界面におけるキャリア密度の制御方法に関する基本的アイディアが従来の研究に言及しつつ示されており、さらに有効な制御方法を確立することにより、多彩な応用へ発展しうることを述べている。

第3章では、本研究で用いた実験方法を述べている。酸化物単結晶薄膜を積層する有効な手段としてパルスレーザー堆積法を用いた。試料の電子物性の評価には、電界効果トランジスタ構造における電気伝導特性の測定および軟X線光電子分光測定が用いられ、これらの評価に用いられる装置の機構および原理、また実験結果の解析手法について述べている。

第4章では、LaAlO3/SrTiO3ヘテロ界面のキャリア密度を増加させる方法として、界面に数単位格子のLaTiO3層を電子ドープ中間層として挿入することが有効であることを述べている。LaTiO3層の厚さおよび試料背面のゲート電圧がキャリア密度および移動度に与える変化を精密に測定し、軟X線光電子分光測定による界面近傍の伝導帯下端エネルギー見積もりと併せて、移動度の大きな低下を伴うことなくキャリア密度を増加させうることを示した。

第5章では、LaAlO3層上にAuによるトップゲート膜を形成することで、さらに界面キャリア密度の制御性が向上することを述べている。Auトップゲート膜を有する電界効果トランジスタ構造において、ドレイン電流のゲート電圧依存性およびホール効果測定、さらにLaAlO3絶縁層の容量測定の結果から、ゲート電圧により界面キャリア密度が2桁にわたり良好に制御できることを実証した。

第6章では、前記トップゲート型電界効果トランジスタの低温動作における特性について述べている。低温におけるキャリア密度の変調によって量子輸送現象が発現しうることを論じ、実際にゲート電圧によりキャリア密度変調を実現することにより、キャリア輸送特性に現れる局在・反局在効果を制御できることを示した。

第7章は結論であり、本研究により学術上意義のある新規な知見が得られたことを総括的に述べている。

遷移金属酸化物は多彩な物性を示すことから新奇な機能をもつデバイスへの応用が期待されているが、ヘテロ界面の物性やその制御法に関しては未だ十分に解明されているとはいえない。とくに界面における電子密度の制御は界面の特異な物性の発現とその利用に極めて重要であり、有効な制御方法の確立が期待されている。本研究ではヘテロ界面の高移動度の電子伝導および多彩な電子状態とで特徴付けられるLaAlO3/SrTiO3ヘテロ界面を対象として、界面キャリア密度の有効な制御方法を確立した。

本研究によって確立した手法においては、ヘテロ界面に挿入する中間ドープ層の厚さやゲート電圧によって、界面キャリア密度および移動度を広い範囲で連続的に制御できる点で優れている。他の酸化物の組み合わせによるヘテロ界面にも適用できる一般的な手法と考えられ、今後の酸化物ヘテロ界面の物性研究においても貢献しうるものと期待できる。

なお、本論文の第4章はクリストファー・ベル、疋田育之、ハロルド・ファン、蓑原誠人、組頭広志、尾嶋正治との、第5章及び第6章はクリストファー・ベル、疋田育之、ハロルド・ファンとの共同研究であるが、論文提出者が主体となって実験及び解析を行ったもので、論文提出者の寄与が十分であると判断される。

以上、本論文は、物質科学へ大きく寄与するものであり、よって、博士(科学)の学位を授与できると認められる。