Introduction

The coexistence of superconductivity and ferroelectricity represents a long-standing challenge due to their fundamentally conflicting requirements. Superconductivity demands on a high density of free charge carriers, while ferroelectricity is generally found in insulating materials. The introduction of free charges into such materials screens long-range Coulomb interactions, suppressing ferroelectricity. Although the coexistence of free carriers and polarization has been proposed1 and observed2,3 in polar metals, including polar superconductors4,5, the polarization in these systems is non-switchable. Recently, however, superconductivity and ferroelectricity were observed to coexist in two-dimensional (2D) van der Waals heterostructures6, where ferroelectricity arises from mechanisms distinct from conventional Coulomb interactions.

Oxide interfaces, particularly those based on SrTiO3 (STO)7,8,9 and KTaO3 (KTO)10,11,12,13,14,15, offer a compelling alternative for exploring the interplay between superconductivity and polarization. Both STO and KTO are wide-gap semiconductors (3.2 eV for STO and 3.6 eV for KTO) and quantum paraelectric16,17. Electron-doped STO was the first oxide superconductor discovered18, and its proximity to ferroelectricity inspired the discovery of high-temperature cuprate superconductors19. Quantum ferroelectric fluctuations have been proposed as a potential pairing mechanism in these systems20,21,22,23,24,25,26,27, and a ferroelectric quantum phase transition was observed inside the superconducting dome of STO28. Additionally, when STO is combined with insulating oxides such as LaAlO3 (LAO), 2D superconductivity emerges at the interface8. In STO-based interfaces, superconducting-ferroelectric coexistence has been achieved via 18O substitution19, and ferroelectric-switchable 2D electron gases have been realized through Ca alloying29. Furthermore, the effects of ferroelectric/ferroelastic domains on both conductivity30,31 and superconducting pairing32 have been suggested. However, the full integration of superconductivity with electrically switchable ferroelectric polarization, despite being conceptually proposed29, has remained experimentally unrealized.

In contrast, while superconductivity has not been observed in electron-doped bulk KTO33, recent studies reveal that 2D superconductivity can emerge at KTO surfaces34,35 and interfaces11,12,13. Notably, KTO-based systems exhibit significantly higher superconducting transition temperature (Tc) than STO-based interfaces, making them more promising for exploring the interplay between ferroelectric polarization and superconductivity. Here, we demonstrate the coexistence of superconductivity and ferroelectricity at LAO/KTO interfaces, providing a novel platform for studying this unconventional phenomenon.

Results

Universally enhanced superconductivity and bistability

A noteworthy feature of KTO interface superconductivity is its tunability via gate bias (VG) across KTO13,36,37,38, as illustrated in Fig. 1a. Over the past several years, we have gated more than 100 LAO/KTO samples and identified a universal phenomenon: after gating experiments (where a VG of up to ± 200 V was applied and subsequently removed, referred to as “poling”), the Tc of LAO/KTO exhibited a significant enhancement of ~0.2–0.6 K (see Fig. 1b for selected results). Additionally, the normal-state sheet resistance (Rs) exhibits a bistable characteristic after poling with different bias polarities (Fig. 1c), indicating the presence of ferroelectricity.

Fig. 1: Universally enhanced superconductivity and bistability.
Fig. 1: Universally enhanced superconductivity and bistability.
Full size image

a Schematic illustration of the gating setup for the LAO/KTO interface, with VG applied across the KTO substrate. The polarity of VG is defined relative to the interface. “Poling” refers to the process of applying VG and then removing it. b Enhancement of Tc (ΔTc) for multiple LAO/KTO samples after poling with VG = −180 V. Open black stars: original Tc. Closed blue diamonds: unpatterned LAO/KTO(110) samples. Closed green triangles: unpatterned LAO/KTO(111) samples. Closed red circles: LAO/KTO(111) Hall bar devices. Open red circles: LAO/KTO(111) Hall bar devices where LAO films were deposited using a non-typical procedure (details provided in Methods). c Time-dependent sheet resistance (Rs) at T = 4.5 K measured while VG was switched repeatedly between 0, −180, 0, +180 V. “Origin” denotes the state before any VG was applied. At VG = 0, two distinct Rs states are observed: “0+” and “0−”, corresponding to the states after removing positive or negative VG, respectively.

Ferroelectric hysteresis under gating cycles

To further investigate these phenomena, we examined the transport behaviors of a typical LAO/KTO(111) Hall bar device under continuous gating cycles. For each VG value, both the temperature dependence of Rs(T) and the Hall effect (measured at T = 4.5 K) were recorded. Figure 2a shows the Rs(T) curve before any VG was applied (denoted as “origin”). A metallic behavior, followed by a superconducting transition at Tc = 1.89 K (defined as the temperature where Rs drops to 50% of its normal-state value at 4.5 K), was observed. Figures 2b, c display the Rs(T) curves during a cycle where VG was swept from +180 V to −180V and then back. In both sweeping directions, a clear overall tuning effect consistent with previous studies13,38 was observed: a positive (negative) VG decreases (increases) the normal-state Rs and lowers (raises) Tc. However, in addition to this overall tuning effect, the Rs(T) curves exhibit a strong dependence on the gating history. For example, the three different VG = 0 states (“origin”, “0+”, and “0−”, where “0+” and “0−” represent the VG = 0 state after removing positive or negative VG, respectively) yielded significantly different Rs(T) curves.

Fig. 2: Ferroelectric hysteresis at the LAO/KTO interface.
Fig. 2: Ferroelectric hysteresis at the LAO/KTO interface.
Full size image

Consecutive gating cycles between VG = +180 V and −180 V were performed on a typical LAO/KTO(111) Hall bar device (S1_A#): a Temperature-dependent Rs(T) curve for the “origin” state (before any gating). b, c, Temperature-dependent Rs(T) curves during a single gating cycle: b sweeping VG from +180 V to −180 V; c sweeping VG from −180 V to +180 V. d, e, f Hysteresis loops during consecutive gating cycles: d Rs-VG; e, Tc-VG; f, μ-VG. The pink shading highlights the VG = 0 region. g, h Evolution of Rs-VG loops with temperature: g measured in decreasing temperature order; h measured in increasing temperature order. To improve clarity, the normalized form [Rs(VG)-Rs(0)]/Rs(0) is used, with curves shifted vertically for better visualization.

As summarized in Fig. 2d–f, two notable and correlated features emerge beyond the overall tuning effect. First, during the initial run starting from the “origin” state (indicated by the blue lines), the device undergoes an irreversible change likely associated with polarization formation, corresponding to the universal Tc enhancement. Second, after the initial run, the device exhibits repeatable and pronounced hysteresis in both VG-Rs (Fig. 2d) and VG-Tc (Fig. 2e) loops. Analysis of the VG-Rs loops at different temperatures, measured during both decreasing (Fig. 2g) and increasing (Fig. 2h) temperature orders, reveals that hysteresis begins above 50 K and becomes pronounced below 30 K. This temperature-dependent behavior closely matches previous observations39,40,41 of weak spontaneous polarization (0.04 μC/cm−2) in KTO, where defect-induced polar nanoregions develop macroscopic coherence within this specific temperature range.

Effects of poling magnitude, temperature, and time

Further investigations of the 0± states after poling with VG (denoted as \({V}_{G}^{0\pm }\)) were conducted to gain deeper insights into the observed ferroelectric behaviors. Since LAO/KTO is sensitive to gating history, we ensured an “origin” state in different experiments by using fresh samples when necessary. Figure 3a–d show the effects of poling VG polarity and magnitude. Two identical Hall bar devices (S2_A# and S2_B#) cut from the same LAO/KTO sample were used. Similar experiments were also conducted on a single device (S7#, Supplementary Fig. 1), whose “origin” state was regenerated through a refreshing process (described below).

Fig. 3: Effects of poling VG, temperature, time, and stability.
Fig. 3: Effects of poling VG, temperature, time, and stability.
Full size image

a, b Temperature-dependent Rs(T) maps as a function of negative (VG < 0) and positive (VG > 0) poling VG, respectively. c, d Corresponding carrier density (ns) and mobility (μ) measured at T = 4.5 K. e Temperature-dependent Rs(T) map as a function of poling temperature, using a fixed poling VG = -150 V. f Evolution of the enhanced ΔTc with cumulative poling time, performed at T = 4.5 K with poling VG = ± 120 V on two different samples. The inset shows the same data on an extended time scale. g The poling-induced state remains non-volatile at low temperatures. h The poling-induced state recovers to the “origin” state after being left in ambient conditions for several days. For each experiment, fresh LAO/KTO samples were used as needed to ensure measurements began from the “origin” state.

For each polarity, the VG was gradually increased from 0 to ± 180 V. At each VG, we applied the poling VG at 4.5 K for 3 min, followed by Rs(T) and Hall effect measurements after removing VG to 0. As shown in Fig. 3a, b, and Supplementary Fig. 1a, the Tc increases with the magnitude of poling |VG| for both polarities, and Tc for “0−” is consistently higher than for “0+”. Poling up to ± 180 V, corresponding to ± 3.6 kV/cm, did not saturate Tc enhancement, suggesting that the ferroelectric polarization remains unsaturated. As shown in Fig. 3c, d, and Supplementary Fig. 1b, the increase in Tc was accompanied by a clear decrease in mobility μ (with a slight increase in carrier density ns; however, the change in ns was much smaller than that of μ). This indicates that, as discussed below, the ferroelectric polarization primarily modulates the interfacial potential well rather than directly altering ns. It should be noted that the transport coefficients, ns and μ, derived from Hall measurements on Hall bar devices, represent spatially weighted averages of all contributing layers, as these parameters are inherently depth-dependent along the z-direction.

Poling temperature also plays a critical role. As shown in Fig. 3e, Tc begins to increase sharply when poling temperatures dropped below 30 K, matching the temperature at which VG-Rs hysteresis becomes pronounced (Fig. 2g, h). Notably, significant Tc enhancement was observed after just one sec of poling, with saturation occurring after 10–15 sec of cumulative poling time (Fig. 3f and Supplementary Fig. 2). The tuned states remain nonvolatile at low temperatures (Fig. 3g). Full recovery to the “origin” state can be achieved by leaving the samples at ambient conditions for several days (Fig. 3h and Supplementary Fig. 3).

Induced ferroelectricity in KTO

All these experimental observations support that ferroelectricity emerges in the KTO side of LAO/KTO interfaces. This is unsurprising, as ferroelectricity has previously been suggested in KTO through minimal chemical doping (e.g., Nb, Li)42, oxygen deficiency43, strain44, or randomly distributed defects40,45,46. Notably, even nominally pure KTO samples typically contain substantial defect concentrations (~1017 cm3). These defects promote the formation of polar nanoregions, which collectively drive KTO into a weak ferroelectric state at low temperatures40,46. Further evidence for low-temperature ferroelectricity in LAO/KTO comes from our Raman scattering measurements (Supplementary Note 1 and Supplementary Fig. 4), where the TO2 and TO4 optical modes emerge below ~40 K (~60 K in poled samples) – closely coinciding with the onset of hysteresis in the VG-Rs loops (Fig. 2g, h). The similarities between the observed hysteresis and Raman features in LAO/KTO and those in the intentionally ferroelectric Al/Sr0.99Ca0.01TiO3 interface29 provide additional support for ferroelectricity in LAO/KTO. We attribute the induced ferroelectricity to a field-induced ordering of the polar nanoregions that are present in KTO, which are inherently disordered and lacking long-range correlation in the pristine state.

While VG-induced irreversibility in Rs has been previously observed in SrTiO3-based heterostructures and attributed to charge trapping/detrapping effects47,48,49, the behavior we observe in LAO/KTO is fundamentally distinct. Our system exhibits reproducible and switchable hysteresis that is independent of the initial VG sweep direction–a hallmark of ferroelectricity (see Supplementary Fig. 5). Moreover, the limited magnitude of carrier density modulation in LAO/KTO (Supplementary Fig. 6) is incompatible with charge trapping, which would require significantly larger ns variations47,48. Although ionic migration could theoretically produce switchable polarization, this process is typically slow and thermally activated50,51,52, conflicting with our low-temperature results. Taken together, the observed hysteresis and switching behaviors in LAO/KTO cannot be explained by charge trapping or ionic migration, providing compelling evidence for ferroelectricity as the governing mechanism.

Coexistence of ferroelectricity and superconductivity in the conducting layer

As illustrated in Fig. 1a, superconductivity at KTO interfaces is confined within a thin KTO layer (~5–10 nm thick11,12,13,38), corresponding to the width of the interfacial potential well. The remaining KTO bulk serves as a thick insulator across which VG is applied. While ferroelectric polarization is expected in the insulating bulk, we propose that it also occurs in the conducting layer. If polarization were confined solely to the insulating bulk, as in conventional ferroelectric transistors, its effects would primarily manifest in ns rather than μ, inconsistent with our observations. Moreover, the bistability in transport properties suggests a switchable polarization within the conducting layer, whose built-in field modulates the interfacial potential well. These results lead us to conclude that ferroelectric polarization coexists with 2D superconductivity in the interfacial conducting layer. Although ferroelectricity and conductivity are traditionally considered mutually exclusive, polarization in the conducting KTO layer could emerge via coupling with the ferroelectric KTO bulk, mediated by their shared lattice bonding, while potentially developing distinct interfacial behavior.

Ferroelectric polarization modulating the interfacial potential well

We now address how the presence of ferroelectricity in the conducting KTO layer can explain our experimental observations. Before delving into the specific phenomena associated with ferroelectricity, we reinforce that the transport properties of LAO/KTO are largely governed by the interfacial potential well profile, with μ serving as a key indicator. Previous studies13,36,38 have shown that, particularly when ns is relatively large, gating at KTO interfaces primarily modulates μ rather than ns. As shown in Fig. 2f and Supplementary Fig. 6a, sweeping VG from −180 V to 180 V caused μ to vary from ~20 cm² V⁻¹ s⁻¹ to ~90 cm² V⁻¹ s⁻¹, while ns changed only slightly (and in this device, in a manner opposite to that expected from a simple capacitance effect. See Supplementary Figs. 6b-d for more information). This behavior can be attributed to VG-induced modulation of the interfacial potential well, which alters the spatial distribution of carriers13,53,54. Such modulation influences the “effective disorder”, thereby affecting μ (a narrower potential well reduces μ)13,53,54.

Notably, the hysteresis observed in VG-Rs (Fig. 2d) and VG-Tc (Fig. 2e) loops is mirrored in VG-μ loops (Fig. 2f) but not in VG-ns loops (Supplementary Fig. 6a). Furthermore, the universal Tc enhancement is accompanied by a significant decrease in μ (Figs. 2f, 3c, d, and Supplementary Fig. 1b), which can be attributed to narrowing of the interfacial potential well13,36,38. Therefore, in LAO/KTO interfaces, unlike typical ferroelectric effects, ferroelectricity primarily affects transport by modulating the interfacial potential well profile rather than directly altering ns. We note that the enhancement of Tc with potential well narrowing is a well-established experimental phenomenon in KTO-based interfaces13,38, yet the underlying mechanism remains elusive. Several potential mechanisms may be at play: (i) fractal superconductivity induced by strong scattering55, (ii) enhanced spin-orbit coupling energy38, (iii) increased local three-dimensional carrier density, and (iv) modifications to electron-phonon coupling56. Further studies are needed to discern which of these mechanisms, if any, contribute to the observed enhancement.

Ferroelectric polarization in the interfacial KTO layer has two key effects: (1) it reduces the dielectric constant (ε) (Fig. 4a) and (2) introduces a built-in electric field due to polarization and the associated screening charges (insets of Fig. 4b). KTO, as an quantum paraelectric, exhibits a large ε of up to 4500 at low temperatures17. However, ε decreases dramatically under applied fields57,58,59 and induced polarization39,41. As illustrated in Fig. 4a, after poling, ferroelectric polarization, regardless of its direction, lowers ε39,41, thereby narrowing potential well (solid blue and red lines vs. dashed line, Fig. 4b), which explains the universal Tc enhancement. Bistability naturally arises from switchable polarization, which can either narrow or widen the potential well depending on its direction. As shown in Fig. 4b, the built-in electric field (Ebuilt-in) forming the potential well results from the superposition of two competing contributions: EP from ferroelectric polarization (opposing the gate field direction) and ES from surface screening charges. After VG removal, the persistent residual field–dominated by ES which opposes and exceeds EP–creates two distinct states: (1) In the “0−” state, the net field Ebuilt-in = EP + ES narrows the potential well, enhancing carrier confinement and increasing Tc; (2) conversely, in the “0+” state, the net field broadens the well, reducing Tc.

Fig. 4: Effect of ferroelectric polarization on the interfacial potential well.
Fig. 4: Effect of ferroelectric polarization on the interfacial potential well.
Full size image

a Schematic evolution of the dielectric constant (ε) of KTO under an applied external electric field, reproduced based on experimental results from refs. 39,41. The pink shading highlights the “0±” states, which exhibit a reduced ε compared to the “origin” state. b Potential well modulation (thick lines) and electron envelope wavefunction (thin lines) for original (dashed), “0−” (blue), and “0+” (red) states. Insets show competing fields from EP (ferroelectric polarization) and ES (screening charges). The net field Ebuilt-in = EP + ES modulates the well profiles. Post VG removal, persistent tuning (aligned with VG direction) suggests that | ES | > | EP |.

As shown in the VG = 0 region of Fig. 2e, the Tc enhancement after initial run is over 3 times larger than the Tc difference between the “0+” and “0−” states in subsequent runs. This indicates that the reduction in ε due to polarization has a stronger influence on the potential well profile than built-in field of the polarization alone. The inability to fully recover the origin state by warming the samples above the ferroelectric onset temperature (even to room temperature; see Supplementary Fig. 3a) is likely due to unbalanced screening charges60,61 at the insulating LAO surface persisting in the cryostat environment. Upon cooling, these charges act as a poling field, reinducing polarization.

Our work demonstrates the coexistence of superconductivity and ferroelectricity at LAO/KTO interfaces, offering a unique platform for studying their interplay. The switchable polarization modulates the interfacial potential well by reducing ε and altering built-in field, driving Tc enhancement and bistable transport properties. These findings deepen our understanding of the mechanisms coupling superconductivity and ferroelectricity, and open new possibilities for designing multifunctional quantum devices. Prospectively, the doping tunable ferroelectricity in KTO42 may enable LAO/KTO heterostructures to serve as a programmable quantum platform62,63 for controlling superconducting-ferroelectric coupled states.

Methods

Sample fabrication

LAO/KTO interface samples were fabricated by depositing amorphous LAO films onto 0.5 mm thick KTO single-crystalline substrates using pulsed laser deposition (PLD). A 248-nm KrF excimer laser was employed with a laser fluence of 0.7 J/cm2 and a repetition rate of 10 Hz. A single-crystalline LAO target was used. For typical samples, 7-20 nm thick LAO films were grown at 300 °C in an atmosphere of 1 × 10−5 mbar O2 and 1 × 10−7 mbar water vapor13. After deposition, the samples were cooled to room temperature under the same atmospheric conditions. In addition, “non-typical” samples were prepared by depositing a 2 nm “typical” LAO layer, followed by a 20 nm LAO layer grown at room temperature. The LAO films are highly insulating, with conduction confined to a thin (~5–10 nm thick) KTO layer adjacent to the interface.

Hall bar devices

Hall bar structures were patterned onto KTO substrates using standard optical lithography and lift-off techniques, with ~200 nm thick AlOx films serving as a hard mask35. The AlOx films were deposited by PLD at room temperature under base pressure, with a laser fluence of 2.5 J/cm2. To ensure high insulation in the AlOx-covered areas, the patterned substrates were annealed at 300 °C for 2 h in a flow of 1 bar O2.

Subsequently, LAO films were deposited onto these pre-patterned KTO substrates, forming conducting LAO/KTO interfaces exclusively in the uncovered regions. Each sample contained four identical Hall bar devices. Across different samples, the central Hall bar bridges were fabricated in two sizes: 20 μm in width and 100 μm in length, or 100 μm in width and 500 μm in length.

Electrical contact

Electrical contacts to the conducting LAO/KTO interfaces were established using ultrasonic bonding with Al wires.

Gating and poling

As illustrated in Fig. 1a, a back-gating voltage (VG) was applied between the conducting interface and the bottom silver electrode, with the polarity defined relative to the interface. The value of VG represents the bias applied to the bottom silver electrode. Throughout the gating process, the leakage current was consistently below 10 nA.

The process of applying VG and subsequently setting it to 0 is referred to as “poling”, similar to operations in conventional ferroelectrics. The states “0+” and “0−” represent the VG = 0 state after removing positive and negative VG, respectively. For clarity, we also denote the VG = 0 state after poling with VG as \({V}_{G}^{0\pm }\). For example, “1200+” indicates the VG = 0 state following an applied VG = +120 V. This nomenclature is extended to all cases accordingly.

Transport measurements

Low-temperature transport measurements were performed using a commercial 4He cryostat equipped with a 3He insert (Cryogenic Ltd.). A four-probe DC technique was employed, utilizing a Keithley 6221 current source and a Keithley 2182 A nanovoltmeter. Carrier mobility (μ) and density (ns) were obtained from Hall effect measurements conducted on Hall bar devices.

Raman spectroscopy

The low-temperature Raman measurements were performed in a cryostat (Attocube, attoDRY2100) equipped with a commercial confocal microscope (Horiba, LabRam Odyssey). The sample temperature was controlled in the range of 1.6 K to 300 K. A linearly polarized 532 nm laser was used as the excitation source, with a ∼1 μm beam radius and ∼3.6 mW power at the sample position.