Dongxing
Zheng
ab,
Junwei
Zhang
ac,
Xin
He
a,
Yan
Wen
a,
Peng
Li
ad,
Yuchen
Wang
b,
Yinchang
Ma
a,
Haili
Bai
b,
Husam N.
Alshareef
a and
Xi-Xiang
Zhang
*a
aKing Abdullah University of Science and Technology (KAUST), Physical Science and Engineering Division (PSE), Thuwal 23955-6900, Saudi Arabia. E-mail: xixiang.zhang@kaust.edu.sa
bTianjin Key Laboratory of Low Dimensional Materials Physics and Processing Technology, Institute of Advanced Materials Physics, Faculty of Science, Tianjin University, Tianjin 300072, P. R. China
cKey Laboratory of Magnetism and Magnetic Materials of Ministry of Education, School of Physical Science and Technology, Lanzhou University, Lanzhou, 730000, PR China
dState Key Laboratory of Electronic Thin Film and Integrated Devices, University of Electronic Science and Technology of China, Chengdu 610054, China
First published on 11th August 2022
The high-mobility two-dimensional electron gas (2DEG) generated at the interface between two wide-band insulators, LaAlO3 (LAO) and SrTiO3 (STO), is an extensively researched topic. In this study, we have successfully realized reversible switching between metallic and insulating states of the 2DEG system via the application of optical illumination and positive pulse voltage induced by the introduction of oxygen vacancies as reservoirs for electrons. The positive pulse voltage irreversibly drives the electron to the defect energy level formed by the oxygen vacancies, which leads to the formation of the insulating state. Subsequently, the metallic state can be achieved via optical illumination, which excites the trapped electron back to the 2DEG potential well. The ON/OFF state is observed to be robust with a ratio exceeding 106; therefore, the interface can be used as an electrically and optically erasable non-volatile 2DEG memory.
Previous studies on the interface between the epitaxial, wide-bandgap insulators, LAO and non-polar STO, have revealed that it is either a high-mobility electron conductor or an insulator, depending on the atomic stacking sequences and thickness of the LAO layer.16,20,25,26 For practical applications, it is desirable to have the ability to modulate the physical properties of the 2DEG by applying external stimuli. Although modulation of the carrier density and mobility of the 2DEG system has been realized via electric field gating,2,15,16 optical illumination,3–5 and ferroelectric gate effect,27 comprehensive, non-volatile, and reversible control of all physical properties is coveted as it would immensely benefit the practical applications of 2DEG systems.
In this study, a defect energy level was induced by the introduction of oxygen vacancies that serve as reservoirs for electrons. By applying an appropriate voltage, high-mobility electrons in the 2D QW were driven to the defect energy level, leading to the formation of a non-volatile and completely insulating state. Conversely, the trapped electrons were excited to 2DEG QW via optical illumination once again, leading to the formation of the non-volatile conducting state. Thus, a metal–insulator transition (MIT) co-modulated by voltage gating and optical illumination was achieved, which also exhibited a high resistance ratio between the insulating and conducting states.
It was reported that the conducting interfacial area in LAO/STO is located ∼7 nm below the LAO layer within the STO substrate.28 Thus, HR-STEM equipped with EELS was used to study the conducting mechanism of the LAO/STO heterostructure as its spectrum-collection area is smaller than 1 nm2 and can be precisely controlled. In situ EELS were collected at the LAO layer (highlighted in blue), the interfacial area (highlighted in black), and the region 15 nm away from the interface (highlighted in red). The corresponding areas are schematically depicted in the inset of Fig. 1e. EELS provides information on both the O and Ti states from the O-K (O 1s to 2p) and Ti-L (Ti 2p to 3d) core edges, which is useful in the characterization of the oxygen vacancies in the LAO/STO 2DEG system.23,29–33 The peaks in the Ti spectrum correspond to the excitation of electrons from the inner 2p3/2 (L3 edge) and 2p1/2 levels (L2 edge) to the narrow unoccupied 3d band, resulting in two narrow peaks. They are further doubled by the crystal-field split induced by the surrounding O atoms,29 namely, the lower energy t2g and higher energy eg levels. In Fig. 1e, the Ti spectrum (highlighted in red) collected at the region located away from the interface was observed to exhibit a typical Ti spectrum in the STO layer with doubled L2 and L3 peaks.29 For the Ti spectrum (black) collected at the interfacial area, the intensity of the L2 t2g peak was observed to have weakened significantly. The weakening of the L2 t2g peak corresponds to the transfer of Ti4+ to Ti3+ for Ti ions near the interface, which was induced by the introduction of oxygen vacancies.30Fig. 1f depicts the O 1s spectrum collected in different areas, as indicated in the inset of Fig. 1e. The O spectrum (highlighted in blue) collected at the LAO layer is a typical O 1s spectrum for LAO.32 The A1 and A2 peaks correspond to the interaction of the O atom with the two nearest Ti atoms—A1 is correlated to the Ti t2g orbital and A2 is correlated to the Ti eg orbital. Compared to the Ti 2p spectrum, the difference between the spectra collected at the interface area and the region located 15 nm from the interface within STO is insignificant, except for a weak difference in the intensity. This can be attributed to two reasons. (1) There are three times as many O sites as Ti, one oxygen vacancy nominally transfers two electrons to the Ti 3d bands; consequently, a slight change in oxygen can be amplified in the Ti 2p spectrum. (2) The sample was annealed in an oxygen atmosphere (1 Torr) at 500 °C, leading to the minimization of the oxygen vacancies. Thus, it is difficult to acquire much information about the oxygen vacancies based on O 1s EELS.
Consistent with previously published works,3,21,24,34–36 oxygen vacancies were found in the interfacial region via in situ EELS analysis. Based on the EELS analysis results, we proposed a model to realize reversible metal–insulator transition (MIT) in the LAO/STO 2DEG system via electric field gating and optical illumination. As depicted in Fig. 2a, the oxygen vacancies induced defect energy levels were supposed to be slightly below the Fermi energy level and the electrons were confined to 2DEG QW. In this case, the sample exhibited a normal metallic 2DEG transport behavior. When a sufficiently large positive pulse voltage (with a 300 ms pulse width) was applied at the gate, the conduction band was bent down and the electrons escaped from 2DEG QW to the defect energy levels induced by oxygen vacancies (Fig. S2b†). This was verified to be an irreversible process by removing the applied pulse voltage.36–38 After the removal of the applied positive pulse voltage, the electrons were observed to remain in the oxygen vacancies induced defect energy levels, and the device was adjusted to an insulating state (Fig. 2b). Subsequently, optical illumination was applied to excite the electrons back to 2DEG QW; following which, the system will revert to the metallic state, as depicted in Fig. 2c. A detailed analysis of the influence of voltage gating and optical illumination was illustrated in Fig. S2 and S3.† Thus, as is shown in Fig. 2d, the reversible modulation between conducting and insulating states can be theory achieved by a combination of positive pulse voltage and optical illumination.
To verify the model, electric transport properties measurements were performed under different conditions. Fig. 3a shows the R–T curves, first, the sample was cooled from 200 to 3 K using a constant gating voltage, Vg of 0, 20, 50, 80, and 130 V (gating cooling), and then, the R–T curves were measured during the warming process from 3 to 200 K without gating voltage. In the second scenario, the sample was cooled without gating, and then, a pulse gating voltage (200 V, 300 ms) was applied at 3 K. The R–T curve was also measured by warming the sample from 3 to 200 K without gating voltage. In the Vg = 0 V cooling process, the R–T curve exhibits an overall metallic behavior. The upturn at low temperatures may be related to the Kondo effect induced by the interplay between the localized Ti 3d orbital moment and unlocalized electrons.12 As Vg was increased, the resistance was observed to significantly increase over the entire temperature range. When Vg was increased above 50 V, the behavior of the R–T curves was observed to change dramatically, and the low-temperature resistance was observed to increase beyond the measurement limit of our instruments at Vg = 130 V. This observation indicates that the electrons are transferred from 2DEG QW to the defect energy levels by the gradual positive gating. It may also suggest the existence of an energy gap between the 2DEG QW and defect energy level, which blocks the transfer of free electrons between these two states. The existence of an energy gap is supported by the avalanche-like sharp drop in the resistance that was observed at 91 K in the curves corresponding to Vg = 130 V (gating cooling) and Vg = 200 V (zero-gating cooling). This avalanche-like effect is expected to be associated with thermally the assisted jumping of electrons from the defect energy levels to 2DEG QW and the decreasing dielectric constant of STO substrates with the increasing temperature (Fig. S8†). Based on the avalanche-like sharp drop at 91 K, we estimated the energy gap at 91 K to be as the thermal energy of KBT = 7.83 meV. Above 91 K, the sample was observed to recover its metallic behavior but with a much higher resistance, indicating that not all the electrons could return to the QW state due to the existence of the energy gap.
To gain a deeper understanding of the gating effect on the electric transport properties of the heterostructure, the R–V curve without a pre-gating process was measured at 2 K under a constant current of 1 μA, as is shown in Fig. 3b. The dependence of R on Vg can be roughly divided into three sections: the conducting (highlighted in blue), transition (highlighted in red), and insulating sections (highlighted in dark yellow). Within the conducting section spanning 0–50 V, where the resistance was observed to decrease rapidly with increasing Vg. This was followed by the transition section at Vg ≥ 50 V, which exhibited a much slower decreasing relationship. The effect of a back-gate voltage on the LAO/STO interface was observed to be twofold. (1) It added electrons to QW when the gate voltage was increased from zero and removed electrons from QW when it was decreased. (2) It controlled the shape of the energy band of QW by tilting the conduction band profile within the STO substrate.16,37–39 As schematically depicted in the inset with a blue frame, the application of a small positive gating voltage (which was not sufficient to bend the conduction band to the defect energy level) induced electrons to be simply added to QW and allowed for a more conducting state. During the transition process, a further increase in the positive voltage caused the conduction band of the STO substrate to bend down, leading to an overlap between the conduction band and defect energy levels. This effect was observed to induce the escape of electrons from 2DEG QW to the conduction band of the STO substrate, nearly causing a saturation in resistance (as depicted in the inset with the red frame).37,39 The final process was the insulating section. When the voltage was decreased from 200 V, the resistance gradually increased and finally reached the measurement limit (10 MΩ). When the gating voltage was reverted to zero, the device attained a completely insulating state as the electrons remained at the defect energy level (the resistance of the sample was too high to be measured using a multimeter with a measurement range of 200 MΩ).
After the above-mentioned process, the sample remained insulating at a low gating voltage situation (Vg < 25 V) and became conducting at a high voltage gating situation (Vg > 25 V) (see Fig. S4a†). This suggests the possibility of resistance modulation between the high and low resistance states by varying Vg. In Fig. 3(c), the Vg of 0 V and 70 V were chosen to realize the ON (low R) and OFF (high R) states. The corresponding band structures for the ON and OFF states are shown in the bottom-left and top-left insets, respectively. To verify the reversibility of the ON/OFF states, we repeated the measurement of R at Vg = 0 and 70 V for 1000 times. The results demonstrated very good reversibility even after 1000 cycles. Further, the influence of a negative pulse voltage pre-gating process on conduction states was also investigated. Our results demonstrated that the negative pulse voltage pre-gating process does not exert significant influence on the conduction states as electrons return to 2DEG QW following the negative pulse voltage pre-gating process. Further details can be found in Fig. S5 and S6.†
Although the ON/OFF states can be achieved via a back-gate gating method, the ON state was observed to be volatile, which requires a sustainable voltage to bend the conduction band. To achieve a nonvolatile ON state, the trapped electrons need to be transmitted back to 2DEG QW. As depicted in Fig. 2c, this is achieved via the illumination method. Fig. 3d presents the influence of white light (wavelength of 400–900 nm) illumination at 3 K. The green zone illustrates the resistance of the initial metallic state, which is cooled from 300 K without a pre-gating process. Following the application of a 95 V voltage, the sample becomes more conducting (cyan zone). However, it becomes insulating when the voltage is removed (gray zone). It maintains the insulating state unless the sample is illuminated again. As is evident from the yellow region (under illumination), the resistance decreases sharply, and then, the sample transfers to a metallic state following illumination over a short period of time. During the illumination process, a 95 V voltage is applied over a duration of 1 s, and the sample is immediately tuned to an insulating state after the gating process. However, this state could be switched to the ON state using the illumination condition. Our results demonstrate that reversible modulation of the metallic and insulating states could be achieved by a combination of optical illumination and positive voltage gating processes.
To understand the physics underlying the above observations, we measure the Hall resistance at different gating voltages and temperatures (3–30 K) to gain more information about the free electrons. The samples are first cooled from room temperature to 3 K, and then, a positive pulse voltage (200 V, 300 ms) and negative pulse voltage (−40 V, 300 ms) are used to pre-gate the sample. In Fig. 4a, following the positive voltage pre-gating process, the sample becomes insulating. Thus, each of the voltage-dependent Hall resistances is measured by applying a consistent voltage capable of tuning the sample into a conducting state. A linear dependence of the Hall resistance on the applied magnetic field is observed with a negative slope, which indicates the electron-dominated conducting behavior. The voltage-dependent carrier density and electron mobility deduced from the Hall effect measurement is depicted in Fig. 4a and b, respectively. In Fig. 4a, the carrier density increases monotonically with increasing voltage, and although the mobility of electrons increases slowly from Vg = 0 to 50 V, it is observed to increase sharply and attain a maximum at Vg = 73 V. With further increase in Vg, the mobility decreases sharply and then increases to a nearly constant value at approximately Vg = 80 V. The non-monotonic behavior of electron mobility is attributed to two factors—the carrier density and trapping of electrons induced by band-bending. At low gating voltages (Vg < 50 V), which are insufficient to bend the conduction band to the defect energy level, the electrons were localized at the defect energy level, leading to a nearly zero electron mobility and an extremely slow increase in the carrier density gating voltage increasing from 50 to 73 V, the localized electrons were delocalized due to the overlap between the conduction band and the defect energy levels. The higher the gating voltage, the stronger overlap will be. Thus, both the electron mobility and carrier density were observed to increase with increasing gating voltage. However, the significantly increased electron density also induces scattering between the electrons. Thus, the electron mobility decreases with increasing gating voltage for gating voltages exceeding 73 V. In Fig. 4b, in the negative voltage case, the electrons in the 2DEG QW are gradually emptied by the negative voltage pre-gating process (Fig. S2, S5 and S6†), thus both the carrier density and mobility are observed to decrease with increasing negative voltage.
The mapping of the temperature- and voltage-dependent carrier density and electron mobility for a sample pre-gated using positive voltage (200 V) have been illustrated in Fig. 4c and d. In Fig. 4c, the high carrier density state was observed to exist within the high voltage and low temperature region. The carrier density decreased monotonically with increasing temperature and decreasing voltage. This temperature-dependent behavior was primarily dominated by the decrease in the permittivity of STO with increasing temperature (Fig. S8†).40–42 In contrast, the voltage-dependent behavior was attributed to the voltage-induced band bending.39 In Fig. 4d, the electron mobility exhibits strong temperature- and voltage-dependent behavior. In this case, the electron mobility was primarily determined by scattering. Two types of scattering were observed in the system—electron–phonon and electron–electron scattering. The former is temperature dependent, while the latter is voltage dependent. Usually, electron mobility is known to decrease with increasing temperature due to the enhancement of electron–phonon scattering. However, in this system, the mobility exhibits a mountain-like peak in the range between 73–80 V at approximately 5 K, indicating the dominance of electron–electron scattering at low temperatures. When the temperature was raised above 8 K, the mobility was primarily influenced by electron–phonon scattering. Thus, electron mobility was observed to decrease with increasing temperature.
Footnote |
† Electronic supplementary information (ESI) available: Method, temperature dependence of resistance, metal–insulator transition mechanism. Temperature dependent R–V curves, R–V curves, reversible switching, temperature dependent dielectric constant, effect of illumination, time dependence of resistance. See DOI: https://doi.org/10.1039/d2nr01582j |
This journal is © The Royal Society of Chemistry 2022 |