Plasma‐Induced 2D Electron Transport at Hetero‐Phase Titanium Oxide Interface

Abstract Interfaces of metal oxide heterojunctions display a variety of intriguing physical properties that enable novel applications in spintronics, quantum information, neuromorphic computing, and high‐temperature superconductivity. One such LaAlO3/SrTiO3 (LAO/STO) heterojunction hosts a 2D electron liquid (2DEL) presenting remarkable 2D superconductivity and magnetism. However, these remarkable properties emerge only at very low temperatures, while the heterostructure fabrication is challenging even at the laboratory scale, thus impeding practical applications. Here, a novel plasma‐enabled fabrication concept is presented to develop the TiO2/Ti3O4 hetero‐phase bilayer with a 2DEL that exhibits features of a weakly localized Fermi liquid even at room temperature. The hetero‐phase bilayer is fabricated by applying a rapid plasma‐induced phase transition that transforms a specific portion of anatase TiO2 thin film into vacancy‐prone Ti3O4 in seconds. The underlying mechanism relies on the screening effect of the achieved high‐density electron liquid that suppresses the electron‐phonon interactions. The achieved “adiabatic” electron transport in the hetero‐phase bilayer offers strong potential for low‐loss electric or plasmonic circuits and hot electron harvesting and utilization. These findings open new horizons for fabricating diverse multifunctional metal oxide heterostructures as an innovative platform for emerging clean energy, integrated photonics, spintronics, and quantum information technologies.


Atomic force microscopy measurements
The surface morphologies of the LAO substrate and the TiO2 layers were characterized by atomic force microscopy (AFM, FSM-Nanoview 1000 AFM), as depicted in Fig. S1.The LAO substrate exhibited an atomically smooth single-crystalline surface with a root-mean-squared roughness (Sq) of 1.3 Å (Fig. S1A).The Sq increased to 5.1 Å after the sputtering of 50 nm-thick TiO2 (Fig. S1B), and further increased to 1.4 nm after post-annealing (Fig. S1C).These results indicate that the post-annealing facilitated the crystallization of TiO2.

Mechanism of plasma induced phase transition
There are two key elements to successfully preparing TiO2/Ti3O4 hetero-phase bilayer using PIPT: a dazzling glow ball and the addition of a small amount of hydrogen.This differs from previous reports and implies different mechanisms.The dazzling glow ball in PIPT should belong to a type of striation (Fig. 1D), representing the resonant absorption of RF electromagnetic waves by ionized gas, so the energy density within the glow ball will greatly exceed that of a typical diffuse glow discharge.In contrast, diffuse plasmas cannot produce Ti3O4 even after 30 min of treatment in our experiment.Once the plasma creates Ti 3+ and oxygen vacancies on the surface, the resulting free electrons will interact with the RF electromagnetic waves, producing a localized heating.This heating effect may accelerate the phase transition of TiO2 to Ti3O4 in the lower layer, creating more free electrons which generate more heat.Anion vacancy-type metal oxides (e.g.TiO2) always have some equilibrium oxygen vacancies, the concentration of which can be controlled by adjusting the temperature and oxygen partial pressure during the synthesis process. 1 Oxygen vacancies and Ti 3+ in TiO2 can be achieved through hydrogenation during heat treatment. 2Hydrogen can stabilize dangling bonds; otherwise, treated TiO2 would rapidly fade under ambient conditions.4][5] Prolonged heating leads to the uniform distribution of oxygen vacancies across entire samples, making it difficult to form heterojunctions on the nanoscale.[5] The lattice mismatch between the Ti3O4 and TiO2 phases induces strain in both layers.As proposed by the Materials Project, the Ti3O4 is in the tetragonal I4/mmm space group, and has a = b = 0.413 nm and c = 0.818 nm, 6 very close to that of the anatase TiO2.There is still a 10% of lattice mismatching between the 2 layers.The relaxed state of TiO2 has a d(100) spacing of 0.378 nm, whereas the relaxed state of Ti3O4 has a d(100) spacing of 0.413 nm.Consequently, the TiO2 layer experiences tensile stress along the [100] direction, resulting in a slight upward curvature and an expansion of its upper half d(100) spacing to 0.385 nm.Conversely, the Ti3O4 layer undergoes compressive stress along the [100] direction, leading to a reduction of its d(200) spacing to 0.193 nm, as indicated in Fig. 1H.

Thickness and uniformity of the TiO2/Ti3O4 film
The thickness of the Ti3O4 layer is evaluated by 3 ways, a) ~27 nm as fitting the XRR data (Fig. S3A), b) 25 nm as exhibited by TEM images taken over a 5-μm zone (Fig. S3B-E), c) 25 nm as shown by STEM images (Fig. 1).The uniformity of the thickness of the Ti3O4 layer is verified by the van der Pauw measurement, in which a uniform film is mandatory.Otherwise, the measured Hall voltages would be in significant error when flipping the polarity of the magnetic field.In another word, the uniformity of the thickness of a film can be guaranteed as long as a reliable mobility is measured with the van der Pauw method.First, the intensities of Ti-L2,3 edges in the EELS spectra were normalized within energy window 453 -470 eV.Then, a linear combination of the EELS spectra of pure TiO2 and Ti2O3 for Ti-L2,3 edges was conducted with linear least-square fitting.Considering the quite similar EELS spectra for Ti 2+ and Ti 3+ , especially at the Ti-L2,3 edge, it is reasonable to take the Ti2O3 as a reference. 7The Ti 2+ and Ti 3+ can be treated as a whole.The weight of the combination is the fraction of Ti 3+ (Ti 2+ ) and Ti 4+ ions.The EELS spectrum of Ti 3+ (Ti 2+ ) was collected from commercial trigonal Ti2O3 with R3̅ c corundum structure (Sigma-Aldrich, 99.99%), because commercial tetragonal Ti3O4 is currently unavailable.The EELS spectrum of Ti 4+ ions originated from laboratory sputter-deposited anatase TiO2.
Fitting confidence of the sheet resistance vs. temperature Since the e-e interaction mechanism takes effect over the whole temperature range, the resistance dependence on lnT and T 2 is effective all the way above 90 K.The fitting result using the formula   =  0 +  1  2 +  2  5 +  3   We thus found a perfect linear dependence of ∆() on lnT, as shown in Fig. 4D in the manuscript, which is also shown below (Figure S5).We further note that the relationship between the corrections to conductivity ∆ and resistance ∆ is ).
Overall, although there are only 7 data points below 160 K, the fitting result to 20 points over the full temperature range is still very reliable.

2D versus potential 3D transport in TiO2/Ti3O4
In theory, the resistance (or conductance) dependences on temperature for different localization and dimensions, which are summarized in literature, 8 are listed in Table S2.
Table S3.Resistance (or conductance) dependences on temperature for different localization and dimensions.

Weak localization
Strong localization In experiment, the dependence of sheet resistance on temperature was first measured to obtain the logarithmic dependence (lnT).Then, applying the Ioffe-Regel criterion Fl = 4.07 >>1, one can ascribe the electron transport of the sample to the weak localization regime.According to the theory described in the Ref. 8, also summarized in Table S2, the electron transport is 2D.In addition, it is also verified the electron transfer and interface electron accumulation of the TiO2/Ti3O4 heterostructure through the EELS scanning (Section of Electron transfer across the heterointerface in the main text), which assisted in explaining the 2D transport of electrons at the TiO2/Ti3O4 interface.
Finally, the fundamentally different electric properties of TiO2/Ti3O4 and Ti3O4 can differentiate the transport properties.The high mobility (around 10 cm 2 V -1 s -1 ) is only possible for crystalline TiO2 according to the data in literature, and the mobilities of amorphous and defective titanium oxide are at least one order of magnitude lower.In Figure S6, we compare the electrical properties (μ vs n) of TiO2/Ti3O4 and reported TiO2 materials.Obviously, that the mobility reported in this work is comparable to the mobility of crystalline TiO2 and one order of magnitude higher than that of amorphous TiO2.
Moreover, The Ti3O4 has an essentially different electrical transport mechanism from the TiO2/Ti3O4 hetero-phase bilayer.As shown in Figure S7, the resistances of them have the opposite dependence with temperature.Above room temperature, the TiO2/Ti3O4 still behaves as a 2DEL, but the Ti3O4 behaves like a semiconductor.
Overall, it is concluded that the 2D transport is indeed the dominant transport mechanism in TiO2/Ti3O4.Why there is no short circuit by the Ti3O4 layer in the overall conduction: The resistance of Ti3O4 decreases with increasing temperature, and the same is true for γ-Ti2O3, which is just the opposite for TiO2/Ti3O4.If the Ti3O4 (a semiconductor) participates in conduction, the resistance temperature dependence above 160 K will be more complicated than the formula: A component of exponential decrease with temperature should be involved at least.Therefore, it is reasonable to conclude that the Ti3O4 layer does not participate in conduction, and the short circuit is unlikely to occur.The reason might be depletion of Ti3O4 in the heterostructure, despite itself is conductivity.

1 𝑇(
Eq. 1 in the manuscript) shown in Figure S4 presents a very low Chi square and a close-to-one coefficient of determination (R square), indicating a properly used fitting model and successful fitting.

Figure S4 .
Figure S4.Fitting results of the sheet resistance vs. temperature.

Figure S5 .
Figure S5.The correction to the conductivity of a 2D Fermi liquid system linearly fitted with lnT