Two-dimensional superconductivity at the surfaces of KTaO3 gated with ionic liquid.

The recent discovery of superconductivity at the interfaces between KTaO3 and EuO (or LaAlO3) gives birth to the second generation of oxide interface superconductors. This superconductivity exhibits a strong dependence on the surface plane of KTaO3, in contrast to the seminal LaAlO3/SrTiO3 interface, and the superconducting transition temperature Tc is enhanced by one order of magnitude. For understanding its nature, a crucial issue arises: Is the formation of oxide interfaces indispensable for the occurrence of superconductivity? Exploiting ionic liquid (IL) gating, we are successful in achieving superconductivity at KTaO3(111) and KTaO3(110) surfaces with Tc up to 2.0 and 1.0 K, respectively. This oxide-IL interface superconductivity provides a clear evidence that the essential physics of KTaO3 interface superconductivity lies in the KTaO3 surfaces doped with electrons. Moreover, the controllability with IL technique paves the way for studying the intrinsic superconductivity in KTaO3.

INTRODUCTION

Since the discovery of superconductivity at the LaAlO3/SrTiO3 (LAO/STO) interface (), oxide interface superconductivity has been attracting increasing interest (–). It is two-dimensional (2D) in nature (, , , –) and can be tuned by applying an electrical gate bias (, , , ), which provides a promising platform to explore the rich phase diagram for 2D superconductors and relevant quantum phase transitions. In LAO/STO, the superconductivity, or, more precisely, the conduction, locates in a thin STO layer near the interface (), and the superconducting transition temperature Tc is comparable with other STO-related superconductors, including the electron-doped bulk STO, () the δ-doped STO thin layer (), and the ionic liquid (IL)–gated STO surface (). This first-generation interfacial superconductivity does not display strong dependence on the STO crystalline orientation (, , , ). However, the newly found superconductivity in KTaO3 (KTO) interfaces (–, , ), with a one order of magnitude higher Tc than LAO/STO, behaves in a very different manner, although KTO shares many common features with STO (, , ).

While STO is the first found superconducting semiconductor known since 1964 (), the first report of superconductivity in KTO was in 2011 (), when a Tc of ~50 mK was observed at KTO(001) surface gated with IL. After that, there was a decade of silence until the recent discovery on a family of KTO interfaces (–, , ). Unusually, the superconductivity of these interfaces exhibits a strong dependence on the KTO surface planes: The (111) () and (110) () interfaces have an optimal Tc of 2.2 and 0.9 K, respectively, a big leap from aforementioned 50 mK (); in contrast, no superconductivity was detected at the (001) interface down to 25 mK (). It is yet to be unveiled why the EuO (or LAO)/KTO interfaces have such a markedly enhanced superconductivity and why the superconductivity exhibits a strong dependence on the KTO crystalline orientation.

A crucial question follows: Does the EuO (or LAO) overlayer play a fundamental role in the occurrence of superconductivity or mainly serve as a way to induce charge carriers (), say, electrons? To address this issue, it is in high demand to explore the electron-doped KTO surface that is not capped with any oxide layer. For this purpose, IL gating is an ideal tool, in which high-concentration electrons can be induced by applying a gating bias, whose underlying mechanism could be electrostatic charge modulation (), oxygen vacancy formation (), or hydrogen insertion (). In this work, we report our IL-gating experiments on KTO surfaces normal to the three principle crystalline orientations. A strong crystalline orientation–dependent 2D superconductivity, which is essentially the same as that observed at the EuO (or LAO)/KTO interfaces, is achieved on KTO surfaces.

RESULTS AND DISCUSSION

Device fabrication

The device fabrication is illustrated sequentially in Fig. 1 (A to E). Using standard optical lithography and lift off techniques, we confined a clean and flat KTO surface (Fig. 1A and fig. S1) to a Hall bar region by coating the other area with a 200-nm-thick amorphous AlO (a-AlO) layer (Fig. 1B). The gating electrodes of a combined metal layer of Ti(10 nm) and Au(50 nm) were deposited on the a-AlO layer (Fig. 1C). Then, we deposited a thin amorphous LAO (a-LAO) layer on six contacting pads of the Hall bar (Fig. 1D). The deposition of the a-LAO layer made the underneath KTO metallically conducting (fig. S3), which ensures better electrical contacts to the central channel when gated by IL. A tiny droplet of the IL N,N-diethyl-N-methyl-N-(2-methoxyethyl)ammonium bis(trifluoromethanesulfonyl)imide (DEME-TFSI) covered the channel and lateral gate electrodes (Fig. 1E). The gate voltage VG was applied as schemed in Fig. 1E. A photo image of a typical device is shown in Fig. 1F. More details in fabricating devices can be found in Materials and Methods.

Fig. 1.

Schematic of gating KTO surfaces with IL.

(A to E) Step-by-step processes for fabricating devices. (A) Clean KTO single-crystalline substrate with flat surface. (B) Confining KTO surface to a Hall bar region by coating the other area with 200-nm-thick a-AlO layer as a hard mask. The a-AlO/KTO interface is highly insulating. (C) Depositing Ti(10 nm) and Au(50 nm) metallic layers to form gating electrodes. (D) Deposing a thin a-LAO layer on six contacting pads of the Hall bar. The deposition was performed in vacuum at room temperature. This process metallizes the underneath KTO surface and thus improves the electric contact to the central Hall bar channel when gated with IL. (E) Putting a tiny drop of IL (DEME-TFSI) on the central area as schemed. Gating configuration and electrical contacts to the device are as labeled. Before putting the IL, the device surface was cleaned by oxygen plasma to remove any residual photoresist. (F) Photo image of a typical device with bonded Al contacting wires.

Emergence of superconductivity

We measured the transport properties of IL-gated KTO surfaces in fabricated devices. Figure 2A shows the temperature dependence of the channel sheet resistance Rsheet of a set of typical samples with three different surface planes (111), (110), and (001), gated at VG = 4.5 V. An overall metallic conduction was observed on these three types of KTO surfaces. At a given temperature, Rsheet decreases in an order following (111), (110), and (001). As demonstrated in Fig. 2B, a superconducting state occurs at low temperatures in these (111) and (110) samples, with a midpoint Tc of 1.66 and 0.94 K, respectively. In contrast, no superconducting transition was detected in the (001) sample down to 0.4 K (Fig. 2B). Hall effect measurements show that charge carriers are electrons. The sheet carrier density nsheet values measured at 2 K for the (111), (110), and (001) samples are 4.37 × 1013, 4.44 × 1013, and 5.21 × 1013 cm−2, respectively (Fig. 2C). The presence of superconductivity is also supported by the following two facts: The zero-resistance state in the IL-gated (111) or (110) surfaces can be completely suppressed by applying a magnetic field perpendicular to the surface (Fig. 2D); a fairly well-defined critical current was observed in the voltage versus current characteristics [Fig. 3A for (111) and fig. S4A for (110)].

Fig. 2.

Transport properties of KTO surfaces gated with IL.

(A) Sheet resistance Rsheet as a function of temperature for a set (denoted by A) of typical samples with three different crystalline orientations, at VG = 4.5 V. (B) A close view of Rsheet(T) values at low temperatures. (C) Hall resistance RHall as a function of magnetic field at T = 2 K. (D) Rsheet as a function of a magnetic field applied perpendicular to the surface, measured at T = 0.5 K. (E and F) The same Rsheet(T) dependences of (111)_A and (110)_A shown in (B), plotted on a [dln(Rsheet)/dT]−2/3 scale. Solid lines indicate the scaling law for a Berezinskii-Kosterlitz-Thouless (BKT) transition. (G) Rsheet(T) for multiple gated KTO(111) samples (denoted by A, B, and C) showing resistive superconducting transitions. The VG and midpoint Tc for each sample are as labeled. The optimal Tc was observed in sample (111)_B when gated and measured in a second run (#2). Note that the process 1(D) (depositing a-LAO) was skipped in fabricating (111)_C.

Fig. 3.

Features of 2D superconductor for the IL-gated KTO(111) surface.

(A) Temperature-dependent voltage-current (V-I) characteristics of the (111)_A sample at VG = 4.5 V. The channel width is 100 μm. (B) The same V-I curves on a logarithmic scale. The long dashed line corresponds to V ∝ I3 dependence and shows that 1.56 K < TBKT < 1.58 K. (C) The upper critical field μ0Hc2 versus temperature. The (squares; field data multiplied by 5 for clarity) and (circles) label the upper critical field perpendicular and parallel to the sample surface, respectively. Dashed lines are fits to the linearized Ginzburg-Landau theory.

The emergence of superconductivity is quite robust. As shown in Fig. 2G, we have observed superconductivity in multiple IL-gated KTO(111) samples. The observed maximum midpoint Tc reaches 2.04 K [see (111)_B, #2], which is comparable with the optimal Tc achieved at the EuO (or LAO)/KTO(111) interfaces (, ). We would like to emphasize that the a-LAO/KTO interfaces only exist on the contacting pads (not the measured channel) (Fig. 1D and fig. S3) and thereby is irrelevant to the emergence of superconductivity on the IL-gated KTO surfaces. The superconductivity occurs in samples without depositing a-LAO as well [see (111)_C].

2D superconductivity

The superconductivity at the IL-gated KTO(111) and KTO(110) surfaces exhibits 2D features. For a 2D superconductor, the superconducting transition is of the Berezinskii-Kosterlitz-Thouless (BKT) universality class. At the vortex-antivortex pair unbinding transition, the current causes the proliferation of vortices, resulting in a V ∝ Iα power law dependence, with α(TBKT) = 3. As shown in Fig. 3B and fig. S4B, for samples gated at VG = 4.5 V, the power law dependence was clearly demonstrated; the BKT transition temperature TBKT is identified to be 1.57 K for (111) (Fig. 3B) and 0.81 K for (110) (fig. S4B), respectively. We also derived the TBKT from the Rsheet(T) characteristics (see Materials and Methods), which gives rise to TBKT = 1.69 K for (111) (Fig. 2E) and TBKT = 1.00 K for (110) (Fig. 2F). These TBKT values, as well as the midpoint Tc derived from the Rsheet(T) curves, agree fairly well with each other.

The 2D nature of the IL-gated KTO surface superconductivity is further revealed by the large anisotropy in magnetoresistance in magnetic fields applied perpendicular and parallel to the KTO surface. A typical set of temperature-dependent magnetoresistance data for the KTO(111) surface gated at VG = 4.5 V is shown in fig. S5, and the upper critical field μ0Hc2 derived from it is shown in Fig. 3C. From the temperature dependences of and , we extract the zero-temperature Ginzburg-Landau coherence length ξGL(0) and the corresponding superconducting layer thickness dsc to be ~24.7 and ~ 9.4 nm, respectively (see Materials and Methods). The analysis of the KTO(110) data for VG = 4.5 V (see fig. S6) gives rise to ξGL(0)~45.8 nm and dsc ~ 17.4 nm. It is noted that, in both (111) and (110) samples, dsc < ξGL(0) is consistent with 2D superconductivity. In addition, the ξGL(0) values are comparable with, while the dsc values are larger than, those of the corresponding EuO (or LAO)/KTO interfaces (–).

VG dependence of transport properties

We examine the VG dependence of transport properties on both IL-gated (111) and (110) surfaces. For the (111) surface, by sweeping VG from 4.5 to 1.75 V, we observed superconductivity at VG ≥ 2.5 V (Fig. 4A), and the midpoint Tc decreases as VG is lowering (Fig. 4B). The sample turned into a metallic or weakly insulating state at VG ≤ 2.0 V and became too insulating to be measured at VG < 1.75 V. Similar observations were made on the (110) surface except for a larger onset VG for superconductivity [3.5 V on (110) versus 2.5 V on (111)] (Fig. 4E). For both surfaces, in the superconducting range, the VG modulation of nsheet is only a minor effect and does not follow that expected from an electrostatic tuning. For example, for the (111) surface, the nsheet measured at 2 K increases, rather than decreases, from 4.44 × 1013 to 5.48 × 1013 cm−2 when VG changes from 4.5 to 2.5 V (Fig. 4C). This suggests that the gating effect is not purely electrostatic. This conclusion is further supported by the device characteristics with a Ag/AgO reference electrode (figs. S8 to S10) (, ). It is noted that we never detected substantial difference in surface morphology (fig. S2) and crystalline structure (fig. S7) in samples before and after IL gating experiments. Meanwhile, the gating-induced conduction vanished when IL was removed. Thus, as well as the electrostatic effect, we cannot attribute the gating effect to the IL gating–induced oxygen vacancies or other electrochemical processes solely. It is mostly likely a hybridized process.

Fig. 4.

Gate voltage VG dependence of transport properties.

(A to D) (111) and (E to H) (110) surfaces. (A and E) Temperature dependence of Rsheet at different VG values. (B and F) Midpoint Tc, (C and G) sheet carrier density nsheet, and (D and H) Hall mobility μHall at different VG values.

Further insights can be gained from the transport data for (111) at VG = 2.0 V and for (110) at VG = 2.5 to 3.0 V. In these metallic states, nsheet decreases notably with decreasing VG (Fig. 4, C and G), as expected from an electrostatic tuning, accompanying with an even more notable increase in the Hall mobility μHall (Fig. 4, D and H). Coming back to the superconducting VG range, we found that the enhancement of Tc with VG is generally accompanied by a reduction of μHall (see μHall at 2 K in Fig. 4, D and H), which is similar to the situation in the electric field–controlled LAO/KTO(111) interface ().

Last, we note that, in both present IL-gated KTO surfaces and previous EuO(or LAO)/KTO interfaces (–, ), the conductance and superconductivity occur in a thin KTO surface layer. Despite the absence of oxide interfaces, the IL-gated KTO surfaces can still be regarded as interfaces between IL and KTO. Therefore, it is not strange that present IL-gated surfaces show similar transport properties as previous EuO (or LAO)/KTO interfaces. However, it is still an open question why the electron-doped KTO surfaces behave so different from the electron-doped bulk KTO () in which superconductivity has not been observed yet. One speculation is that both surface electronic subbands and surface phonon modes could be markedly different from those in bulk and depend on surface planes strongly. Further studies are highly desirable to unveil all these puzzles.

In summary, we have demonstrated that a strong surface crystalline plane-dependent 2D superconductivity can be induced on KTO surfaces by IL gating technique. The very similarities between this surface superconductivity and its cousin, i.e., the one observed in the interfaces between KTO and other oxides (EuO or LAO), provide a clear evidence that the essential physics of KTO interfacial superconductivity lies in the KTO surfaces doped with electrons. As a last remark, we comment that the present and previous experiments (, ) indicate that in a given KTO sample, a relatively low mobility favors the superconductivity, which could be a clue to further tackle the unusual KTO superconductivity.

MATERIALS AND METHODS

The commercial 0.5-mm-thick KTO single-crystalline substrates were purchased from Hefei Kejing Materials Technology Co. Ltd. Hall bar devices were prepatterned onto the surface of the KTO substrates using standard optical lithography and lift off techniques. The ~200-nm-thick a-AlO hard mask (blue regions; Fig. 1B) was grown by pulsed laser deposition at room temperature under 0.01-mbar oxygen atmosphere. The interface between a-AlO and KTO is highly insulating and thus is inactive. The active uncovered KTO surface is confined within the Hall bar area (red regions; Fig. 1B). The width of the central channel is 100 μm for samples (111)_A, (110)_A, and (001)_A and 20 μm for samples (111)_B and (111)_C. The gate electrodes (yellow regions; Fig. 1C) were made by a combined bilayer of 10-nm titanium and 50-nm gold metals, both of which were evaporated by electron beam evaporation. A 20-nm [or 10 nm for (111)_B] a-LAO layer was deposed on the Hall bar contacting pads (gray regions; Fig. 1D). The deposition was made by pulsed laser deposition at room temperature, in 2 × 10−7–mbar background vacuum. Such an interface between a-LAO and KTO is metallic and thus facilitates the electric contacts to the central channel when gated with IL. Before putting IL, the sample surface was treated by oxygen plasma for 20 s to remove any remaining photoresist.

Atomic force microscopy

The atomic force microscopy data were taken using noncontact mode on a Park NX10 system.

X-ray diffraction

The x-ray diffraction data were taken using a monochromated Cu-Kα source on a 3-kW high-resolution Rigaku SmartLab system.

Electrical contacts and transport measurements

The electrical contacts to the devices were made by ultrasonic bonding with Al wires. The transport measurements were carried out in a commercial DynaCool Physical Property Measurement System (Quantum Design) with a 3He insert. A DC current measurement method was used.

IL gating

The electrolyte we used was DEME-TFSI (Kanto Chemical Co. Inc.). For accumulating carriers, the gate voltage VG was applied at 250 K using a Keithley 2611B source meter. For changing VG, we set VG to 0 in low temperature, warmed up the devices to 300 K, and then cooled them down to 250 K where the new VG was applied. In most devices, the applied maximum VG was limited to 4.5 V to avoid severe electrochemical reactions of the IL.

Derivation of TBKT with Rsheet(T) characteristics

The Rsheet(T) data were fitted using Rsheet(T) = R0exp[−b(T/TBKT − 1)−1/2], where R0 and b are material parameters ().

Analysis using Ginzburg-Landau form

The Ginzburg-Landau coherence length, ξGL, is extracted using the linearized Ginzburg-Landau form (, ) , where ∅0 is the flux quantum and ξGL(0) is the extrapolation of ξGL to T = 0 K. For a 2D superconductor, , where dsc is the superconducting layer thickness (, ).