Observation of superconductivity in structure-selected Ti2O3 thin films

The search for new superconductors capable of carrying loss-free current has been a research theme in condensed matter physics for the past decade. Among superconducting compounds, titanates have not been pursued as much as Cu2+ (3d9) (cuprate) and Fe2+ (3d6) (pnictide) compounds. Particularly, Ti3+-based compounds or electron systems with a special 3d1 filling are thought to be promising candidates as high-TC superconductors, but there has been no report on such pure Ti3+-based superconducting titanates. With the advent of thin-film growth technology, stabilizing new structural phases in single-crystalline thin films is a promising strategy to realize physical properties that are absent in the bulk counterparts. Herein, we report the discovery of unexpected superconductivity in orthorhombic-structured thin films of Ti2O3, a 3d1 electron system, which is in strong contrast to the conventional semiconducting corundum-structured Ti2O3. This is the first report of superconductivity in a titanate with a pure 3d1 electron configuration. Superconductivity at 8 K was observed in the orthorhombic Ti2O3 films. Leveraging the strong structure-property correlation in transition-metal oxides, our discovery introduces a previously unrecognized route for inducing emergent superconductivity in a newly stabilized polymorph phase in epitaxial thin films. Researchers have discovered a titanate -based single-crystal film with a rare structure that allows superconductivity to emerge at low temperatures. Titanium (III) oxide, Ti2O3, has garnered recent interest for solar and energy harvesting devices because its electrons are easily excited by electromagnetic radiation. A team led by Tom Wu from University of New South Wales in Sydney, Australia and Shuai Dong from Southeast University in Nanjing, China, report that these electrons have a very high density comparable to metals and are strongly coupled with lattice in constrained thin films. The researchers used a sapphire substrate to direct the growth of nanometer-thick Ti2O3 layers into a solid-state structure not normally found in free-grown crystals. One particular combination of growth temperature and film thickness yielded an orthorhombic crystal phase that exhibited an unusual semiconductor-to-superconductor electronic transition when cooled to 8 K. Polymorph-dependent superconductivity was discovered in orthorhombic-structured Ti2O3 thin films, which is in strong contrast to the conventional semiconducting trigonal corundum-structured Ti2O3. Leveraging on the strong structure-property correlation in transition-metal oxides, our discovery introduces a hitherto unrecognized route towards inducing superconductivity in 3d1 electron systems via polymorph engineering.


Introduction
Exploring structure-property correlations is a frontier of material science, which enables researchers to understand and harness the emergent properties of advanced materials. Functional oxides have attracted attention due to their strong structure-property correlation. [1][2][3][4][5][6] For example, enhanced multiferroic properties have been discovered in epitaxial-stabilized tetragonal BiFeO 3 7 and hexagonal TbMnO 3 films, 8 and these structures are absent in the bulk counterparts. Superconductors have zero electrical resistance and magnetic flux expulsion below certain critical temperatures. Motivated by both scientific curiosity and application demands, extensive efforts have been devoted to seeking new, high-T C superconductors and exploring the origin of superconductivity. However, emergent superconductivity in thin films with newly stabilized polymorph phases has not been reported. Recently, nanosized, trigonal corundum Ti 2 O 3 phase has attracted much attention owing to its narrow bandgap and efficient photothermal effect. 9 Here, we report the discovery of a strong structure-property correlation in Ti 2 O 3 polymorphs; i.e., superconductivity emerges in orthorhombic Ti 2 O 3 films epitaxially stabilized on sapphire substrates, while the trigonal bulk phase is semiconducting.
Most milestones in superconductivity research have been set by the discoveries of new materials such as cuprates 10,11 and Fe-based superconductors. [12][13][14][15] In addition, there are some notable transition-metal-oxide superconductors, such as tungsten bronzes with T C values up to 6 K. 16,17 Within the 3d-transition-metal series, Ti 3 + -based compounds are predicted to be promising candidates for high-T C superconductors as a result of their special filling (3d 1 and S = 1/2), 18 but superconducting titanates have been limited to materials like electrondoped SrTiO 3 (T C ≈0.41 K), 19 non-stoichiometric titanium monoxide TiO x (T C ≤ 2.3 K) 20,21 and ternary oxide Li-Ti-O (T C ≈14 K). [22][23][24] Recently, interest in superconducting titanate thin films has resurged, and superconductivity in TiO, 25 γ-Ti 3 O 5, 26,27 and Ti 4 O 7 26,27 was reported. It is important to note that these titanates feature Ti 2+ or mixed (Ti 3+ /Ti 4+ ) valence states, and there has been no report of superconductivity in a "pure" Ti 3+ (3d 1 ) system.
Little progress has been made over the past several decades in the search for binary-oxide superconductors, which might possess attractive characteristics such as structural simplicity, high tenability and environmental stability. Among the few superconducting binary oxides reported to date, monoxide NbO exhibits a very low T C of 1.38 K. 28 Among sesquioxides, Zn-doped and Sn-doped In 2 O 3 were reported to be superconducting at 3.3 K 29 and 4 K, 30 respectively. The recent advent of thin film synthesis techniques offers a new route towards achieving superconductivity at high transition temperatures in epitaxial films and heterostructures. For example, in one pioneering work, superconductivity was reported to be above 100 K in single-layer FeSe films, 31 while bulk FeSe shows superconductivity below 8 K. 32 Inspired by these discoveries, we postulate that investigating titanites, particularly Ti 3+ -containing Ti 2 O 3 , in a thin film form might open a new door in the search for new oxide superconductors.
In this work, we report the discovery of superconductivity in newly stabilized, epitaxial, orthorhombicstructured Ti 2 O 3 thin films grown on α-Al 2 O 3 substrates, which demonstrates the strong structure-property correlation in such strongly correlated electron systems. Although six different lattice variants of Ti 2 O 3 were predicted, i.e., corundum, Rh 2 O 3 (II)-type, perovskite, post-perovskite, α-Gd 2 S 3 -type and Th 2 S 3 -type, only corundum and Th 2 S 3 -like types have been reported experimentally. 33 Surprisingly, we found that the structure of Ti 2 O 3 thin films can be selected by tuning the growth conditions; particularly, a hitherto unreported orthorhombic Rh 2 O 3 -like structure was discovered. Furthermore, although the conventional corundum-type Ti 2 O 3 is a regular, narrow-bandgap semiconductor, the orthorhombic thin films are a new oxide superconductor with a T C of 8 K. In fact, this is, to our knowledge, the first report demonstrating emergent superconductivity in thin films of a certain material while absent in the bulk counterpart. Our discovery offers a new route to inducing superconductivity in artificial epitaxial thin films and heterostructures by modifying their polymorph crystalline structures.

Materials and methods
Fabrication of Ti 2 O 3 thin films Ti 2 O 3 films with a thickness range of 40-280 nm were deposited on (0001) sapphire substrates with a size of 5 × 5 × 0.5 mm 3 using pulsed laser deposition (PLD). A corundum-phase Ti 2 O 3 target (Sigma-Aldrich, 99.99%) was used. Before deposition, the pressure of the PLD chamber was lower than 3.0 × 10 −9 Torr. The deposition was performed at temperatures from 500 to 900°C with a 248 nm laser (KrF, Coherent Inc.). The energy density of the laser on the target was varied from 2 J/cm 2 to 3 J/cm 2 . The distance between the target and substrates was fixed at 5.6 cm. More synthesis details are given in Table S1.

X-ray diffraction (XRD)
X-ray diffraction patterns were recorded using a Bruker D8 ADVANCE diffractometer (used for powder samples) and Bruker D8 DISCOVER high-resolution diffractometer (used for thin film samples) operated at 40 kV/40 mA and 35 kV/50 mA, respectively. Both diffractometers were equipped with Cu Kα radiation source and LynxEye detector.

Raman spectroscopy
Raman measurements were performed with a confocal micro-Raman system (Horiba Aramis/Jobin Yvon HR800) at ambient conditions. The wavelength of the excitation solid-state laser is 473 nm. The laser beam was focused on thin film samples using a ×100 objective lens (×50 objective for powder samples) with a numeric aperture (NA) of 0.90, and the spot size was approximately 1 µm. A typical laser power of 0.5 mW was used to avoid sample heating. The spectra were calibrated using a standard silicon sample.

X-ray photoelectron spectroscopy (XPS)
XPS measurements were carried out on a custom-built ultra-high-vacuum system with a Mg Kα (1253.6 eV) excitation source. The vacuum of the analysis chamber was approximately 2 × 10 −9 Torr, and the spectrum was calibrated according to the O 1s XPS spectra (530.3 eV) before the first experiment on the Ti 2 O 3 -LT sample. Before the XPS experiments, the samples were sputtered for 27 min to remove contaminated surface layers.

Transmission electron microscopy
A focused ion beam (FEI Helios 400S) was used to prepare the cross-section samples for the S/TEM analyses. A protecting Pt layer was deposited on the top of the Ti 2 O 3 thin film. S/TEM lamella with a thickness of~100 nm was transferred to a half-moon FIB grid by an in situ lift-out method, followed by low-energy Ar-ion cleaning in a Fischone Nano Mill 1040. The damaged surface layer was removed, and the final thickness of the sample was ~20 nm. The cross-sections of Pt/Ti 2 O 3 /sapphire structures were characterized using a scanning transmission electron microscope (Titan Mono-Probe Cs, FEI). Electron diffraction patterns were obtained with a Titan ST electron microscope operated at 300 kV. HAADF-STEM images were collected with an aberration-corrected Mono-Probe Cs scanning transmission electron microscope (Titan, FEI) operated at 300 kV using a convergence semi-angle of 21 mrad. The inner collection semi-angle of the HAADF was 70 mrad.

Transport and magnetic measurements
The electrical transport and magnetic measurements were performed using a Quantum Design physical property measurement system (PPMS) and a Quantum Design superconducting quantum interference device, respectively. For the transport measurements, electrodes with 30 nm of titanium and 100 nm of platinum with a size of 0.5 × 0.5 mm 2 were deposited by sputtering. Gold wires were used to connect the samples to the PPMS puck. In the magnetic measurements, a quartz holder and a straw were used to support the sample when the magnetic field was parallel and vertical to the film surface, respectively. Data were recorded at 10 Oe during warming after being zero-field cooled from 300 to 2 K. A ceramic tweezer was used to handle the samples during all measurements.

First-principles calculations
The electronic structure calculations were performed based on a generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof-revised (PBEsol) function, 34 as implemented in the Vienna ab initio Simulation Package. 35,36 Two types of structural relaxations were performed: (1) full relaxation of lattice constants and inner atomic positions and (2) relaxation of inner atomic positions with only fixed experimental lattice constants. The Hubbard U was applied to Ti d orbitals. In addition, hybrid functional calculations based on the Heyd-Scuseria-Ernzerhof exchange 37 were also performed to correctly describe the bandgaps. Nonmagnetic states were considered. The phonon structure and e-ph coupling were calculated using the PWSCF program of the Quantum-ESPRESSO distribution based on density function perturbation theory. 38

Thin-film deposition and structural characterizations
The epitaxial Ti 2 O 3 thin films were deposited on α-Al 2 O 3 (0001) substrates in a high vacuum (<3.0 × 10 −9 Torr) using PLD. Experimental details are provided in the Supplementary Materials. XRD was used to examine the crystal structure and phase of the as-deposited films. The diffraction patterns of Ti 2 O 3 thin films grown at 600°C (sample denoted as Ti 2 O 3 -LT) and 900°C (sample denoted as Ti 2 O 3 -HT) on α-Al 2 O 3 substrates are shown in Fig. 1a and Supplementary Figure S1. Cross-sectional transmission electron microscopy (TEM) was used to determine and calibrate the film thickness in this work (Supplementary Figure S2). Apart from the substrate peak, the peaks at 39.06˚and 83.78˚in the 600°C grown sample (Ti 2 O 3 -LT) agree well with the (0006) and (00012) planes, respectively, of the known corundum (trigonal) structure of Ti 2 O 3 (R3c space group, #167, a = b = 5.15`Å, c = 13.61 Å). Surprisingly, the XRD peaks of the 900°C grown sample (Ti 2 O 3 -HT) are located at different locations of 37.90˚and 80.86˚, indicating a phase different from that of Ti 2 O 3 -LT. Note that based on the XRD peak positions, the Ti 2 O 3 -HT phase can be well distinguished from the TiO phase. 25 Furthermore, we carefully explored the PLD synthesis parameters and determined the conditions for growing Ti 2 O 3 -LT and Ti 2 O 3 -HT phases (Supplementary Figure  S3). We should note here that both Ti 2 O 3 -LT and Ti 2 O 3 -HT thin films are stable. No change was observed in their appearance and structure after being stored under ambient conditions for over half a year (Supplementary Figure  S4).
As shown in Fig. 1b, the Raman modes observed in the Ti 2 O 3 -LT film are consistent with the D 34 point-group symmetry of the corundum structure (Supplementary Figure S4). 39 On the other hand, for the as-deposited Ti 2 O 3 -HT film, only four peaks were observed, indicating a different lattice structure and higher lattice symmetry. Raman data also allow us to exclude the existence of titanium oxides other than Ti 2 O 3 . Specifically, the comparison of the Raman spectra of Ti 4 O 7 and Ti 2 O 3 -HT (Supplementary Figure S5) reveals the absence of any Ti 4 O 7 phase in our samples.
Atomic force microscopy (AFM) was used to investigate the surface morphology of Ti 2 O 3 films, and in an area of 1 µm × 1 µm, Ti 2 O 3 -LT and Ti 2 O 3 -HT have similar surface roughnesses of~0.45 and~0.40 nm, respectively. However, as shown in Fig. 1c 42 Finally, electron energy loss spectroscopy 43,44 and synchrotron-based X-ray absorption spectroscopy (XAS) 43 were used to further confirm the valence state of Ti 3+ in both phases. As shown in Fig. 1f, the XAS spectra collected from Ti 2 O 3 -LT and Ti 2 O 3 -HT films are identical and consistent with that of the standard corundum Ti 2 O 3 bulk. 45 Meanwhile, the Ti 3+ valence state was confirmed by XPS and XAS in the Ti 2 O 3 -HT sample, which also reveals a difference from that of previously reported TiO, 25  We should note here that there are thin (2-3 nm) transition layers at the interfaces. These transition layers were also observed in some similar cases in previous reports, such as epitaxial TiO 25 and MgB 2 46 films on Al 2 O 3 . They can be regarded as buffer layers in the epitaxial growth. [47][48][49] The transition layer observed at the Ti 2 O 3 /Al 2 O 3 interface may not have the stoichiometric composition of Ti 2 O 3 due to oxygen diffusion during the film-growth process. 46 However, determining the composition and structure of this intermediate layer is outside the scope of this work.
The HAADF-STEM images taken from two different zone axes (Fig. 2a, e) Fig. 2a, e, which are consistent with the simulated electron diffraction data (Supplementary Figure S6). The inverse FFT (IFFT) images in Fig. 2b, f also confirm the corundum structure of Ti 2 O 3 -LT.
The HAADF-STEM images of the Ti 2 O 3 -HT/Al 2 O 3 sample, as shown in Fig. 2c, g, suggest a structure that is different from that of corundum Ti 2 O 3 -LT. Close-up views of the film and substrate lattices are shown in the IFFT images in Fig. 2d, h. In fact, the structure of Ti 2 O 3 -HT can be reconstructed from that of Ti 2 O 3 -LT (Supplementary Figure S7). The unit cells of Ti 2 O 3 -LT (R3c space group, #167) and Ti 2 O 3 -HT (Immm space group, #71) are shown as insets in Fig. 2e, g, respectively.
The unit cell parameters of Ti 2 O 3 -HT are a = 9.385 Å, b = 4.422 Å, and c = 2.809 Å, and a structural comparison with α-Al 2 O 3 and Ti 2 O 3 -LT is given in Fig. 3. Excellent agreement was obtained between the experimental HAADF-STEM images and the projections of the reconstructed structure, as shown in Fig. 2d, h (Supplementary Figure S8). Furthermore, as shown in Supplementary Figure  When the PLD growth temperature is higher (900°C), the Ti and O atoms have higher kinetic energies to rearrange themselves into the orthorhombic Ti 2 O 3 -HT phase to achieve a smaller mismatch with the substrate. Figure  S11) was found for the Ti 2 O 3 -HT samples at approximately 370 K, which is much lower than that reported for the corundum-structured Ti 2 O 3 (500 K). 50 Hall effect measurements on the Ti 2 O 3 -LT and Ti 2 O 3 -HT thin films are shown in Supplementary Figure S12. Ti 2 O 3 -LT is a ptype semiconductor, which is consistent with previous work, 51 whereas Ti 2 O 3 -HT is an n-type semiconductor. More importantly, the carrier concentration of Ti 2 O 3 -HT is on the scale of 10 22 /cm 3 , approximately one order of magnitude higher than that of Ti 2 O 3 -LT, which is the same order as that of normal metal. 52 The most important discovery is that the Ti 2 O 3 -HT samples were superconducting at low temperatures (Fig. 4a), while the Ti 2 O 3 -LT samples remained  Fig. 2e, g, respectively. The small red spheres are oxygen, and the large blue and cyan spheres are titanium and aluminum, respectively. Scale bars in a, c, e and g are 2 nm semiconducting (Fig. 4b). One unique property of the superconducting samples of Ti2O3 is the semiconducting transport observed in their high-temperature normal state, which is different from the behavior of common superconductor. In addition to the metal-insulator transition at~370 K in Ti 2 O 3 -HT, a semiconductorsemiconductor transition (SST) was observed at~110 K. A similar SST was observed in Mg[Ti 2 ]O 4, 53 which is also a Ti 3+ system. The SST in Mg[Ti 2 ]O 4 was believed to be a result of Ti-Ti dimerization instability, and the doublewell Ti-Ti bond potential at the crossover changes from localized to itinerant electronic behavior. 53 Most likely, the existence of superconductivity in the 3d 1 system of Ti 2 O 3 is related to the semiconducting normal state and SST, which warrants further investigation.

A metal-insulator transition (Supplementary
To exclude the possibility of superconductivity exclusively existing in the intermediate layer and to confirm its bulk nature, we studied the thickness dependence of the superconductivity in the orthorhombic Ti 2 O 3 -HT samples. Ti 2 O 3 -HT phase films were grown on α-Al 2 O 3 (0001) substrates with thicknesses from~10 tõ 280 nm. As shown in Fig. 5a, we found that the critical transition temperature (Tc: onset) increases with the thickness and saturates at approximately 8 K when the   Table S1 thickness is larger than 150 nm (Fig. 5b). When the sample thickness decreases to 10 nm, no superconductivity is observed above 1.8 K (the lowest measurement temperature). In addition to the thickness, the growth temperature is another important factor influencing the Tc of the orthorhombic-structured Ti 2 O 3 samples. As shown in Table S1 and Supplementary Figure  S13, samples (with the same thickness) deposited at lower temperatures exhibit lower Tc values, which could be a result of their lower degree of crystallinity. In fact, as summarized in Fig. 5c, the PLD growth temperature is the most critical factor determining the phase and superconductivity in Ti 2 O 3 samples.
As shown in Fig. 6a, T C ≈8 K was obtained for Ti 2 O 3 -HT (sample HT-3 with a thickness of 168 nm), which to our knowledge is the highest T C ever observed in binary-oxide superconductors. Figure 6b shows that a magnetic field can suppress the superconducting transition temperature of Ti 2 O 3 -HT. These behaviors were also observed in Ti 2 O 3 -HT films grown on other types of substrates (Supplementary Figure S14), which indicates that the superconductivity is an intrinsic property of the orthorhombic polymorph of Ti 2 O 3 . The contour plot shown in Fig. 6c summarizes the variation in the sample resistance as a function of the magnetic field and temperature, showing a phase diagram with a sharp phase boundary between the superconducting and semiconducting regimes.
As shown in Fig. 6d, the critical field, H C , increased with the decreasing temperature. Below 4 K, the critical field was larger than 9 T. In Fig. 6e, the critical transition fields at different temperatures are fitted to the Ginzburg-Landau (GL) equation: 54 H c2 (T) = H c2 (0) (1 − t 2 )/(1 + t 2 ), where t denotes a reduced temperature, T/T c . From the fitting, the upper critical field at zero temperature was estimated to be 21.9 T and the critical temperature to be 8.4 K.
The superconductivity was further confirmed by measuring the diamagnetic property of the Ti 2 O 3 -HT samples. As shown in Fig. 6f, the magnetization appears saturated at low temperatures, and the superconducting volume fraction is estimated to be more than 99% at 2 K (Supplementary Figure S15), indicating the clear bulk nature of the observed superconductivity. The smaller T C observed in the magnetization measurement suggests that the superconductivity between 6 and 8 K could be filamentary. In addition, the broadened transition shown in Fig. 6f may imply the existence of some defects that formed during the film deposition.

Theoretical calculations
The physical properties of both Ti 2 O 3 phases were further elucidated using density functional theory (DFT) calculations. More details of the DFT calculations can be found in the Supplementary Materials. The electronic structures of Ti 2 O 3 -LT and Ti 2 O 3 -HT are shown in Fig. 7a, b, respectively. To simulate the experimental conditions, we adopted the experimental lattice constants and relaxed the atomic positions based on the generalized gradient approximation with Hubbard U (GGA+U). However, as shown in Fig. 7c, the GGA+U calculations led to metallic states for both Ti 2 O 3 -LT and Ti 2 O 3 -HT (except in the large U limit for Ti 2 O 3 -LT), which was different from the experimental semiconducting results. Therefore, newly developed hybrid functional calculations were employed to recalculate the electronic structures based on the GGA+U relaxed structures. As shown in Fig. 7a, b, both Ti 2 O 3 -LT and Ti 2 O 3 -HT were predicted to be semiconductors with very narrow bandgaps close to the experimental values (0.108 eV and 0.124 eV, respectively, estimated from the optical absorption data in Supplementary Figure S16). As expected, both the highest valence band and lowest conducting band around the  Most importantly, the superconductivity of Ti 2 O 3 was observed only in the orthorhombic-structured Ti 2 O 3 film and not in the corundum-structured bulk phase. Thus, the superconductivity observed in Ti 2 O 3 is strongly polymorph-dependent. To our knowledge, only a few materials show polymorph-dependent superconductivity, including tin, 55 zirconium, 56 and CuSe 2.

57
For tin, the β-phase (known as white tin) is superconducting, while the α-phase (known as gray tin) tin is semi-metallic. However, a recent breakthrough was made by Liao et al. 58 upon the observation of superconductivity in few-layer stanene (ultrathin α-phase tin). We believe that the exploration of polymorph dependence will attract more interest in the field of superconductor research.
In the current stage, it is difficult for us to identify the mechanism for superconductivity in Ti 2 O 3 -HT films, which may be different from that in the Cu 2+ (3d 9 ) and Fe 2+ (3d 6 ) systems. One unique feature of the orthorhombic-structured Ti 2 O 3 -HT phase is its high carrier concentration (~10 22 /cm 3 ), which is comparable to that of normal superconducting metals. 52 The Bardeen-Cooper-Schrieffer (BCS) theory can describe the quantum-mechanical many-body state of interacting electrons in superconducting metals by considering electron-phonon (e-ph) coupling. We calculated the phonon-related properties of unstrained, nonmagnetic Ti 2 O 3 -HT, and the data for the phonon dispersion, total phonon DOS, and frequency-dependent e-ph coupling, λ (ω), are shown in Fig. 7d, e, f, respectively. Remarkably, the total e-ph coupling constant, λ, is 1.114, which is large enough to enable electrons to form Cooper pairs (for MgB 2 with T C = 40 K, λ = 0.73). 59 In contrast, much weaker e-ph coupling was obtained for Ti 2 O 3 -LT (Supplementary Figure S18, λ = 0.46). The BCS superconducting transition temperature was estimated using the standard Allen-Dynes McMillan equation. As shown in Fig. 7g, using a typical Coulomb pseudopotential, μ* = 0.1, and electron DOS at the Fermi level, a high T C above 24 K can be estimated for Ti 2 O 3 -HT. The deviation between the experimental T C and the calculation result might be a result of overestimating the electron DOS in the calculation or extrinsic factors such as structural defects and electronic inhomogeneity. Fig. 6 Superconductivity observed in the orthorhombic-structured Ti 2 O 3 polymorph. a Resistance versus temperature data measured on the HT-3 and LT-2 samples. b Temperature dependence of the resistance (HT-3) measured under magnetic fields ranging from 0 to 9 T. c Superconducting phase diagram in the contour mapping plot using the data in (b). d Magnetic-field dependence of the resistance measured at fixed temperatures ranging from 2 to 10 K (HT-3). e Relationship between the critical magnetic field and the transition temperature. The line is the GL equation fitting. f Diamagnetic shielding measurement performed on the HT-3 sample. The temperature dependence of the zero-field-cooling (ZFC) and field-cooling (FC) moment was measured in a field of 10 Oe applied perpendicular to the film surface Here, we should note that because of the spin degree of freedom in Ti 2 O 3 ; i.e., Ti ions have a spin of 1/2, superconducting mechanisms beyond the conventional BCS mechanism involving strong quantum fluctuations of magnetism, such as that in the case of cuprates and ironpnictides/selenides, may play important roles. In fact, if magnetism is considered in the calculation, the ground state is antiferromagnetic for both Ti 2 O 3 -LT and Ti 2 O 3 -HT polymorphs (Supplementary Figure S19). Recently, efforts have been devoted to study the interplay between magnetism and superconductivity in cuprates, 60 Fe-based superconductors 61 and 2D oxide interfaces. 62,63 Interestingly, similar to our case, Ti 3+ has been reported to play an important role in the coexistence of magnetism and superconductivity at the LaAlO 3 /SrTiO 3 interface. 62,63 Therefore, further studies to elucidate the superconductivity mechanism in orthorhombic-structured Ti 2 O 3 -HT with 3d 1 electrons will be interesting and may help boost the T C in this class of materials.

Conclusion
In summary, we demonstrated the selective stabilization of a new orthorhombic phase of Ti 2 O 3 in thin films grown on sapphire substrates. Moreover, as an indication of a strong structure-property correlation, the orthorhombic Ti 2 O 3 films showed unexpected superconductivity at 8 K, which is absent in trigonal, corundum-structured Ti 2 O 3 . Our calculations revealed stronger e-ph coupling in orthorhombic-structured Ti 2 O 3 compared to that in the bulk corundum phase. We envision that a further increase in the superconducting temperature could be accomplished by doping and pressure. From the perspective of fundamental science, the discovery of superconductivity in this 3d 1 electron system may provide a new platform to investigate titanite-based superconductors, which are complementary to the well-explored cuprates (3d 9 ) and iron-based (3d 6 ) systems. Importantly, Ti 2 O 3 is earthabundant, nontoxic and chemically stable, which are essential merits for potential power and communication . c Bandgaps calculated using both GGA+U and HSE methods. The HSE method clearly gives a better description of the experimental data. d Phonon dispersion spectrum, e Total phonon DOS, and f Eliashberg function, α 2 F(ω) (solid line), and frequency-dependent electron-phonon coupling, λ(ω) (dashed line), calculated for relaxed Ti 2 O 3 -HT. g T C of the two Ti 2 O 3 phases calculated using different μ* values applications. Furthermore, our study presents a novel path toward inducing emergent physical phenomena in thin films with newly stabilized polymorph phases, which is promising for advancing material exploration and device technologies.