Introduction

Optical control of superconductivity has been the focus of many research works1,2,3,4,5. For instance, optical phonon pumping can be used to facilitate electron pairing3,4,5. Such method, however, usually requires stringent resonance conditions and intense laser pulses. The produced pairing states often have a short lifetime and are only detectable by ultrafast pump–probe measurements. To date, persistent impact on superconductivity generated by continuous-wave (CW) light has only been demonstrated in ionic liquid-gated organic materials1, where the photochemically controlled isomerization of the ionic liquid modulated the doping level in the superconducting layer.

Here, we explore a different strategy to optically manipulate superconductivity in epitaxial heterostructures where the photoactive and superconducting layers are strongly coupled through the interface. By selecting a photoactive substrate that also exhibits strong electron correlations, photoexcitation can yield additional functionalities far beyond doping. The material system we study here is the heterostructure of FeSe/SrTiO3. SrTiO3, well known for its rich structural and electronic phases6,7,8,9, has a 3.2 eV direct optical bandgap. In contrast, FeSe is a layer superconductor that can be isolated to a single layer10. Monolayer FeSe is particularly susceptible to charge transfer doping11,12,13, strain11,14, phonon scattering15,16, and spin related proximity effects17 at the interfaces. The combination of these two materials has led to TC well above the bulk values of FeSe11,18,19. In this system, we show that a higher-TC metastable state in FeSe can be reached by a weak CW UV photoexcitation in SrTiO3 substrate, owing to a polaron related interface polar distortion. Using tailored sequence of UV irradiation and field-controlled dipole re-orientations, the heterostructure can be persistently driven between the metastable state and its ground state, allowing the superconducting zero-resistance to be rapidly turned on and off. This realization of persistent optical switching of high-temperature superconductivity in FeSe/SrTiO3 highlights an interesting route toward the active manipulations of quantum materials.

Results

Superconducting transitions in FeSe persistently controlled by photoexcitations in SrTiO3 substrates

One unit cell (uc) FeSe film (Fig. 1a) was grown on TiO2-terminated insulating SrTiO3 (001) substrate by molecular beam epitaxy (MBE). On top of the FeSe monolayer, ~10 uc FeTe capping layer was grown subsequently. Transport properties of the capped film were characterized shortly after the sample was removed from the ultrahigh vacuum (UHV) growth chamber. A typical set of measurement data is shown in Fig. 1. In dark, the onset temperature (TC) of the superconducting transition was around 24 K (Fig. 1b, black). Illuminated by 3.5 eV ultraviolet (UV) light, TC of the sample was raised to 30 K (Fig. 1b, red). The observed superconductivity enhancement can be produced by very weak UV light fields and is independent on the illumination intensity (Fig. 1c). Photoenergy-dependent measurements were also carried out at 1.5, 2.3, and 3.1 eV (Supplementary Figure 1). These longer wavelength lights produced no observable effect in the transport measurements. In metallic systems with large carrier densities, such as FeSe20, the effects of direct carrier excitations by weak CW light on the electrical properties are usually negligible. Considering the stark contrast of the electrical responses to photons with energies below and above 3.2 eV (i.e., bandgap of SrTiO3), the UV-induced superconductivity enhancement observed here more likely originated from the optical absorption in the SrTiO3 substrate rather than in the gapless FeSe film.

Fig. 1
figure 1

Raising TC by light in FeSe/SrTiO3 heterostructure. a Effects of UV light on FeSe/SrTiO3 were evaluated by transport measurements in van der Pauw geometry. Scanning tunneling microscope (STM) topography images of sample with and without FeTe capping layer are shown on the right. Scale bars indicate 100 nm. b Solid lines: sample resistances measured in dark (black) and under UV illumination (red), showing an increase in TC from 24 K to 30 K. Dashed lines plot the resistances measured with 9 T out-of-plane magnetic field. c Resistances measured at different UV intensities. Despite of the greater normal state resistance modulations at larger light intensities, the TC enhancement is intensity independent

The light-induced transition into superconducting state occurs very fast. While the study on the dynamic switching process is currently limited by the time needed to update the UV diode power supply output (a few ms), the almost instantaneous resistance drops observed upon turning on the UV light indicate that the light switching speed is at least in the kHz range or better. More interestingly, the light-induced superconductivity enhancement in FeSe persisted even after the UV light was turned off (Fig. 2a). As shown in Fig. 2b, following a sharp optical switching, the produced zero-resistance state was very stable in dark at 16 K. This state can remain at least for days when the sample temperature is kept constant. To better characterize the persistency of the light-induced superconductivity enhancement, we performed a multi-loop temperature variation experiment as illustrated in Fig. 2c. In this experiment, the zero-resistance state was first created at 16 K by UV light. After tuning off the UV light, the sample went through a series of thermal cycles to different maximum heating temperatures (TH = 20K, 30K, 40K, …, 300K). After reaching TH, the sample was cooled back to 16 K where its resistance was measured. The resistance at 16 K as a function of TH exhibits a plateau-like behavior with two significant upturns at 40 K and 150 K, both coinciding with structural phase transitions in SrTiO3. Around 40 K, SrTiO3 undergoes a transition into quantum paraelectric state6,21,22,23,24. And at 150 K, a cubic–tetragonal structural phase transition occurs in the surface layers of SrTiO325,26. It’s worth noting that the dwell time at TH had no observable effect on the measurement result. For example, after reaching 200 K, cooling immediately versus staying at 200 K for several hours before cooling yielded the same resistance measurement at 16 K.

Fig. 2
figure 2

Persistent superconducting state generated by UV illumination. a Resistance measured before UV illumination (black, solid), during UV illumination (red, solid), and after turning off the UV light at base temperature (black, dashed). b Sample resistance measured at 16 K as the UV light was switched on and off. Resistance dropped to zero instantly when UV light was turned on and remained zero even after the light was turned off. c To study the retention of this effect, a series of thermal cycles of 16 K → TH → 16 K were performed after turning off the UV light. At the end of each thermal cycle, sample resistance at 16 K was measured and plotted as a function of the maximum temperature (TH) reached during sample heating

Instead of thermal cycling to room temperature, the UV-induced transition to superconducting zero-resistance state can also be rapidly reverted at low temperature by applying bias pulses to the back of the 0.5-mm-thick SrTiO3 substrate. To restore the finite resistance, a negative bias needs to be applied first. As shown in Fig. 3, after five −100 V, 5s voltage pulses, the sample resistance can be raised to a value even larger what was measured prior to UV exposure. This large resistance, however, was decreasing slowly over the time. Then, by applying a positive (5 s, 100 V) voltage pulse, the sample resistance can be quickly stabilized near the pre-exposure value. We note that, the seconds-level back bias switching speed, considerably slower than what was produced by UV light, might be related to the capacitive effects and defect states associated with the thick high-k dielectric substrate. We also note that, the persistent switching effects of back biases are only possible after UV exposure. In as-grown samples, the effects of back biases are completely volatile as discussed in ref. 13. Using the combination of UV light and back bias pulses, FeSe can be switched between superconducting and normal states repeatedly and reliably.

Fig. 3
figure 3

Switching between normal and superconducting states. Black stars plot the 17 K resistance measured before any UV exposure. Purple squares plot the zero resistance of a superconducting state achieved by a brief flash of UV light (with a duration less than 1 s). After applying five pulses of VB = −100 V with 5 s duration to the back of the SrTiO3 substrate, the normal state in FeSe can be recovered with a resistance (orange circles) that is larger than the pre-UV exposure value but slowly decreasing over time. The evolving resistance can be quickly stabilized near the pre-UV exposure value (green triangles) by applying one pulse of VB = 100 V with 5 s duration. The non-volatile switching between normal and superconducting zero-resistance states by UV light and back bias pulses can be repeated many times without any sign of fatigue

Photoexcited carriers transfer between SrTiO3 and FeSe

The mechanism for the observed optical control of superconductivity can be complex, as UV light affects SrTiO3 in many ways. For example, if defects are introduced to SrTiO3 by the substrate pre-annealing step in UHV, the relaxation of defects after carrier excitation can serve as hole traps and produce persistent UV photoconductance27,28. Exposing freshly cleaved SrTiO3 (001) surface to intense extreme ultraviolet (EUV) light can lead to the formation of surface two-dimensional electron gas (2DEG)29,30,31, possibly originating from the creation of either oxygen vacancies or surface reconstructions. Additionally, due to the strong coupling between photocarriers and the lattice in SrTiO3, UV irradiations were also found to excite soft phonons/polarons32,33 and cause persistent phonon softening at low temperatures34. In FeSe/SrTiO3 heterostructures, many of these effects can potentially impact the superconducting behaviors in FeSe through the interface, such as generating charge transfer doping11,12,13, modulating interface electron–phonon coupling15,16, and producing interface lattice distortions14,35,36,37,38,39. Among these possible contributors to the observed UV-induced persistent superconductivity enhancement, we first evaluate the role of photoexcited charge transfer.

As shown in Fig. 4c, without UV illumination (black curve), the magnitude of the Hall resistance RH decreases sharply toward zero below TC due to the Meissner effect. In the presence of light (yellow and red curves), the enhancement of superconductivity is evident by the higher temperature where this sharp drop occurs. Above TC, the Hall resistance under illumination only deviates from the dark value between 50 K and 90 K. Such changes, however, did not persist after turning off the UV light. In this temperature range, the field dependence of the transverse resistance became nonlinear under the UV exposure (Supplementary Figure 3), indicating the formation of mobile electrons with mobilities much higher than the intrinsic carriers in FeSe. This effect was accompanied by a large light-induced magnetoresistance (Supplementary Figure 3) and a surface UV photovoltage that becomes more positive with decreasing temperature (Fig. 4d).

Fig. 4
figure 4

Photoinduced charge transfer and superconductivity related interface polar distortion. a Schematics of the band bending and UV excited charge transfer in the FeSe/SrTiO3 heterostructure at different temperature ranges. b Instead of photodoping, UV-induced SrTiO3 surface polar distortion is more likely the cause of the enhanced superconducting behaviors. A reported double-TiO2 layer terminated SrTiO3 surface structure13, 39 is depicted. c Hall resistances measured as functions of temperature with and without UV illumination. d Temperature dependence of open circuit photovoltage measured between two surface electrodes when the UV light was focused near one of them. e Frequency spectra of the out-of-phase impedance responses at various temperatures during and after the UV exposure. R-C resonances marked by C1 and C2 indicate the emergences of photocapacitance at low temperatures

These features between 50 K and 90 K can be understood by a photoexcited charge transfer at the FeSe/SrTiO3 interface. When electron–hole pairs are generated in SrTiO3 by UV light, due to its upward band bending toward the interface12,13, holes are driven into FeSe, leaving behind electrons in SrTiO3 (Fig. 4a, middle). Such spatial charge separation effectively prevents the electron–hole recombination, giving rise to a positive surface photovoltage and a unique n-type photoconductance in SrTiO3 that was not observed in identically processed bare SrTiO3 substrates within this temperature range (Supplementary Figure 4).

While these results clearly indicate a significant photocarrier transfer at the FeSe/SrTiO3 interface, the resultant hole doping into FeSe is not likely to enhance superconductivity. Several studies have identified the inter-electron-pocket pairing as the most likely mechanism for the high-TC state in FeSe40,41,42, which is expected to be enhanced by electron rather than hole doping11,12,43. Also, light-induced Hall resistance modulation (Fig. 4c) and photovoltage (Fig. 4d) both decreased significantly below 50 K, indicating that the UV-induced interface charge transfer is suppressed in the temperature range that is most relevant to the superconducting transition.

Photocapacitances and photocarriers mediated interface distortions

To identify UV-induced effects in SrTiO3 at temperatures below 50 K that can positively impact the superconducting state in FeSe, AC impedance spectroscopy was performed (Fig. 4e) to detect signals that may not manifest in DC ohmic measurements. Without UV exposure, the sample exhibits an ~3 nF temperature-independent capacitance (Supplementary Figure 5) that is related to the geometric dielectric properties of the sample and contact junctions. Above 100 K, sample capacitance was not affected by the UV light (Fig. 4e, top). Between 50 K and 90 K, a 101 nF level photocapacitance (marked as C1 in Fig. 4a) was measured and associated with the interface charge separation (Fig. 4e, middle). Below 50 K, while the charge transfer is diminished significantly, another much greater (~102 nF) photocapacitance (marked as C2 in Fig. 4a) emerges (Fig. 4e, bottom) and remains strong as temperature falls even below TC (Supplementary Figure 5b).

The observed large photocapacitance provides a valuable hint for understanding the optical superconductivity enhancement. The generation of capacitance relies on the changes of either free or bound charge distributions. Since the free charge separation between different material layers are significantly suppressed at low temperatures, the onset of photocapacitance C2 mostly likely originates from an enhanced material polarizability and the related bound charge formation in response to UV light. Consistent with such assessment, a giant enhancement in dielectric constant was reported in SrTiO3 samples irradiated by UV light at low temperatures44. As discussed above, SrTiO3 undergoes a quantum paraelectric phase transition at low temperatures6,21,22,23,24, where quantum fluctuations associated with zero-point energy prevent the onset of long-range ferroelectric order. In this phase, photoexcited electrons can quadratically couple to the T1u soft mode (relative displacement between the Ti ion and the oxygen octahedra) and directly impact the quantum fluctuations32,33,34,45. In particular, the polarons formed from photocarriers and phonons can serve as effective charge trap to suppress electron–hole recombination and generate local dipole moments44, leading to the large photocapacitance observed in our experiments.

In FeSe/SrTiO3 heterostructure, the interface band bending12,13 gives rise to a large electric field at the interface that points from SrTiO3 to FeSe. In the presence of this field, the alignment of induced dipole moments at the interface may effectively modify the ferroelectric distortions with relative out-of-plane shifts between the Ti and O ions that are well known for both single-46 and double-TiO247 terminated SrTiO3 surfaces. As a result, it is viable that the Se–Fe–Se angle in the FeSe monolayer, a parameter sensitively modulating the electron correlation strength in FeSe35, will be perturbed as well (Fig. 4b). Because of such photocarrier-mediated structural distortion, an enhancement in superconductivity can be produced. Besides of producing a superconductivity enhancement in FeSe and a large photocapacitance in SrTiO3, the T1u polaron related interface polarization will also reduce the band bending and thus suppress the photocarrier transfer between SrTiO3 and FeSe (Fig. 4a, right). This effect well explains the reductions of surface photovoltage and Hall resistance modulations observed below 50 K (Fig. 4c, d).

We note that, the possibility of generating oxygen vacancies (OV) in SrTO3 by UV irradiation was also suggested in several studies30,48. While OVs can have a lower formation energy38 in the double-TiO2 surface layer found in some FeSe/SrTiO3 heterostructures13,39 and are often associated with interface-enhanced superconductivity37,49, they are not likely the origin of the light-induced TC enhancement observed here. First, the creation of OVs typically requires extended exposure time and should be highly dependent on light intensity30,48, both of which are inconsistent with our observations. In addition, since FeSe is prone to oxidation, the removal of oxygen with FeSe in proximity may generate chemical modifications in FeSe that are harmful to its superconductivity.

As shown in Fig. 2, the light-induced zero-resistance state well persisted after the removal of UV light, indicating that the charge trapping by polarons and the interface polar distortion triggered by light is metastable in the quantum paraelectric phase without external perturbation. While the combination of photocarrier, strong electron–phonon coupling, and built-in electric field allows the system to reach such metastable state with higher TC, there can be multiple pathways for the system to go back to the ground state. One of them is through thermal cycling to different structural phases (Fig. 2c). Alternatively, by applying an external field antiparallel with the built-in field, as executed via negative back biases (Fig. 3), re-alignment of the interface dipoles can also quickly excite the system from the UV-induced metastable state, allowing it to either slowly relax toward the ground state or rapidly settle with the aid of positive back bias (Fig. 3).

Effects of the capping layer in restricting the superconducting transition temperature

TC observed by ex situ transport measurements typically varied between samples when they were measured in as-grown state in dark (Fig. 5a, black). Nonetheless, the TC enhancement induced by UV light always saturated at around 30 K (Fig. 5a, red). To understand the origin of such TC saturation, we characterized the superconducting transition in the same sample both before it was capped by FeTe layers using in situ ARPES (Fig. 5b–d) and after it was capped using ex situ transport measurements (Fig. 5a, Supplementary Figure 6). Comparing with the ~50 K gap opening temperature detected by ARPES before capping, the ~30 K onset TC observed in ex situ transport measurements was much lower. Additionally, ARPES data showed that the uncapped film had no hole pocket at the Γ-point, and the n-type doping was little changed as temperature varied (Supplementary Figure 6a, b). In contrast, the ex situ transport measurements performed on capped film revealed a co-existence of electrons and holes and a strongly temperature-dependent doping type (Supplementary Figure 6c).

Fig. 5
figure 5

Capping layer related TC bottleneck. a Ex situ resistance of three capped samples (S1, S2, S3) measured with and without UV exposure. While the ex situ TC measured in darkness varied between different samples, the UV enhancement of TC always saturated at around 30 K, which was still much lower than what was measured in uncapped films, indicating the leading bottleneck of TC in capped samples is likely related to the capping layer. b, c In situ ARPES spectra of uncapped film (sample S3 before capping) taken across Γ-point (b) and M-point (c) along the directions as marked in the 2D intensity plot (d). e, f Spectra across M-point measured at 10K (e) and 70K (f). g, h Temperature-dependent energy distribution curves (EDCs) (g) and symmetrized EDCs (h) at the Fermi crossing, showing the superconducting gap opening between 50 K and 60 K

Since the FeTe capping layer itself typically had a much lower conductance comparing with the FeSe samples (Supplementary Figure 6d), electrical shunting caused by the capping layer is expected to be weak. Instead, the capping layer as well as the environmental factors more likely introduced a profound modification to the electrical properties of the FeSe monolayer underneath, which restricted the highest TC that can be reached by manipulating the FeSe/SrTiO3 interface alone. To evaluate the full capability of the UV-induced superconductivity enhancement, alternative capping methods (Supplementary Figure 8) and/or in situ measurements with light excitations needs to be explored in the future.

In conclusion, we have shown that a brief exposure to a weak (~101 µW cm−2) 3.5 eV CW UV light can raise the superconducting TC in FeSe/SrTiO3 heterostructure and generate zero-resistance state that persists in dark for at least days. We attribute this effect to the strong photocarrier–phonon coupling in SrTiO3 and the resultant metastable polar lattice distortion occurred at the FeSe/SrTiO3 interface. Quick and non-volatile switching between this metastable state and the as-grown ground state can be achieved by UV photoexcitation and field-controlled interface dipole re-orientations. An ex situ TC upper limit of ~30 K, even with the UV enhancement, was observed. This TC bottleneck is a result of the second interface formed between FeSe and the capping layer. To further enhance the ambient superconducting performance of FeSe, this interface deserves as much as attention as the interface formed with the substrate. The FeSe/SrTiO3 heterostructures also exhibited very large photovoltage and photocapacitance that are consequences of interfacial effects, showing a promising potential for the implementations of designed heterostructures from 3D oxides and 2D layered materials in optoelectronic applications. Most importantly, the capability of manipulating the superconducting state by photoexcitations in the substrate demonstrated a promising venue for realizing active controls in correlated materials.

Methods

SrTiO3 substrate preparation

The insulating single crystal SrTiO3 (001) was cleaned with deionized water for 30 min then chemical etched with buffered-oxide etchant for 1 min. It was subsequently thermal annealed in a tube furnace under O2 flow for 4 h to obtain a TiO2-terminated surface. The substrate was then transferred into the ultrahigh vacuum (UHV) MBE chamber and annealed at 600 °C for 30 min. After those treatment, the SrTiO3 (001) surface becomes atomically flat with well-defined step-terrace structure, as shown in Supplementary Figure 2a.

FeSe films and FeTe capping layer preparation

The growth of monolayer FeSe films were carried out on SrTiO3 (001) substrates in an UHV system (base pressure ~1 × 10−10 Torr) that integrates two MBE chambers, a room temperature scanning tunneling microscope (STM), a low temperature (5 K–300 K) STM and an angle revolved photoemission spectroscopy (ARPES). For the growth of monolayer FeSe films, the substrate was held at around 400 °C, and Fe and Se were supplied via separate Knudsen cells with a flux ratio of around 1:10. The film, as shown in Supplementary Figure 2b, was then annealed at ~520 °C for 2–3 h and in situ transferred into RT-STM and ARPES. For ex situ electrical transport measurements, around 10 ML FeTe films were deposited on 1 ML FeSe/SrTiO3(001) by co-evaporating Fe and Te with a flux ratio of around 1:4, and the substrate was kept at ~300 °C. The growth rate is 0.5 ML/min for both FeSe and FeTe growth.

STM and ARPES measurements

In situ STM imaging with a chemically etched tungsten tip was used to monitor surface morphology of SrTiO3 (001), FeSe and FeTe films at room temperature. ARPES was carried out with a Scienta DA30 analyzer and He discharge lamb (hv = 21.218 eV). The energy resolution was set at ~ meV, as shown in Supplementary Figure 6. The angular resolution is 0.3°. The Fermi level was determined by measuring the Ag film on Si (111) substrate (shown in Supplementary Figure 7).

Magnetotransport and photocapacitance measurements

Ex situ magnetotransport measurements with and without UV illumination was performed using a Quantum Design Physical Property Measurement System (PPMS). Photocapacitance measurements were performed in the same PPMS system but using an external lock-in amplifier for the AC current sourcing and voltage detection. Indium electrical contacts were mechanically attached to the sample surface in Van der Pauw geometry at room temperature in the atmosphere. 3.5 eV UV light used in the experiments was generated by a light emitting diode (LED) and collimated by lenses.

Photovoltage measurements

Photovoltage measurements were performed in a Montana instrument Cryostation system. 3.5 eV UV light used was generated by a LED source and focused to the sample surface near one electrode by lenses. Voltage between illuminated and unilluminated electrodes was measured by a Keithley DC nanovoltmeter.