Room temperature Mott metal-insulator transition in V2O3 compounds induced via strain-engineering

Vanadium sesquioxide (V2O3) is an archetypal Mott insulator in which the atomic positions and electron correlations change as temperature, pressure or doping are varied giving rise to different structural, magnetic or electronic phase transitions. Remarkably, the isostructural Mott transition in Cr-doped V2O3 between paramagnetic metallic and insulating phase observed in bulk has been elusive in thin film compounds so far. Here, via continuous lattice deformations induced by heteroepitaxy we demonstrate a room temperature Mott metal-insulator transition in 1.5% Cr-doped and pure V2O3 thin films. By means of a controlled epitaxial strain, not only the structure but also the intrinsic electronic and optical properties of the thin films are stabilized at different intermediate states between the metallic and insulating phases, inaccessible in bulk materials. This leads to films with unique features such as a colossal change in room temperature resistivity (DR/R up to 100,000 %) and a broad range of optical constant values, as consequence of a strain-modulated bandgap. We propose a new phase diagram for pure and Cr-doped V2O3 thin films with the engineered in-plane lattice constant as a tuneable parameter. Our results demonstrate that controlling phase transitions in correlated systems by epitaxial strain offers a radical new approach to create the next generation of Mott devices.


I. INTRODUCTION
Vanadium sesquioxide (V 2 O 3 ) is an example of a Mott insulator presenting both non-isostructural and isostructural metal-insulator transitions (MIT) as a function of temperature, doping and pressure 1-3 and has attracted great interest for applications.V 2 O 3 undergoes a non-isostructural MIT at 160 K upon cooling from a corundum paramagnetic metallic (PM) phase to a monoclinic antiferromagnetic insulating phase (AFI) 4,5 , yielding a resistivity change of 7 orders of magnitude 1 .In contrast to this low-temperature MIT, the isostructural Mott MIT between paramagnetic metallic (PM) and insulating (PI) phases occurs close to room temperature (RT) and has been observed for a limited range of Cr doping (0.5% -1.7%) concentration 6 .This transition has been observed in bulk crystals with a resistivity change of 3 orders of magnitude and a variation in a-axis lattice parameter of 1% within the same corundum structure 1,2 .
The bulk phase diagram of V 2 O 3 as a function of temperature, doping and applied pressure has been well established based on extensive experimental data from single crystals [7][8][9][10] , and supported by a number of theoretical calculations that successfully predicted the density of states (DOS) of the different phases [11][12][13] .In thin films, various deviations from the bulk phase diagram have been observed.For example, recently it has been shown that the AFI phase in thin films is characterized by stripe domains related to different lattice distortions 14 .a) Electronic mail: mariela.menghini@imdea.org.b) Electronic mail: jeanpierre.locquet@kuleuven.be.
Then, the low-temperature PM-AFI transition has been reported by many groups 6 , while the isostructural RT Mott transition involving PM-PI phases -despite greater interest for applications -has been missing in most studies 6,15,16 .This suppression of the RT MIT has been attributed to the clamping effect of the substrate in epitaxial films 16 or to structural disorder and small grain size present in polycrystalline films 6,15 .Especially, epitaxial strain has so far shown its strong effect on the phase transitions in the vanadate family, which can range from a change in transition temperature 17 or direction dependent electrical properties 18 in VO 2 films to a complete suppression of the low-temperature phase transition 19 in V 2 O 3 films.
Salev et al. 20 demonstrated that the low-temperature MIT can actively be modulated through a voltage-actuated PMN-PT [Pb(Mg,Nb)O 3 -PbTiO 3 ] ferroelectric actuator, which transfers switchable ferroelastic strain into the epitaxial V 2 O 3 film and leads to a giant nonvolatile resistive switching of ∆R/R ~1,000%, while the effect at room temperature is only ∆R/R ~18%.Querr é et al. 21observed non-volatile resistive switching in Cr-doped V 2 O 3 films, where conducting filaments form as a result of a local electric field driven lowtemperature Mott transition.These recent achievements have pointed out the great potential of Mott insulator thin films with large resistivity changes across phase transitions for application in sensors and memory devices.However, triggering the metal-insulator phase transition at room temperature would be even more desirable for their implementation in devices.
The RT MIT has been observed in polycrystalline thin films of Cr-doped V 2 O 3 by optical transmission measurements while electrical resistivity didn't show a transition 6 .Epitaxial V 2 O 3 thin films showed a local, pressure-induced, transition at RT by means of a contact AFM tip 22 whereas a systematic study by varying the oxygen doping concentration in epitaxial films didn't show the RT MIT 23 .These results indicate that the RT MIT can be stabilized within a narrow window of conditions, which need to be systematically explored.So far, a generalized phase diagram for V 2 O 3 thin films does not exist but would be highly desirable to gain fundamental understanding about the interplay between epitaxial strain and Mott transition and to use these films in practical applications.
Here, we present a systematic strain-engineering study of 1.5% Cr-doped and pure V 2 O 3 thin films.Our approach is based on the use of precisely engineered oxide heterostructures as artificial substrates to systematically tune the in-plane lattice parameter in the pure and 1.5% Cr-doped V 2 O 3 thin films within and beyond the reported structural states in the bulk.The in-plane lattice parameters of the engineered substrates were gradually adjusted between 4.943 Å and 5.037 Å by varying the precise composition of the thin film buffer layer.Detailed chemical analysis by means of Energy Dispersive X-ray spectroscopy (EDX) confirmed the chemical composition of the buffer layers and the V 2 O 3 films.Extensive structural studies including High Resolution X-Ray Diffraction (HRXRD), Reciprocal Space Maps (RSMs) and High Resolution Transmission Electron Microscopy (HRTEM) revealed the in-plane lattice parameters of the heterostructures and the amount of induced stress.Temperature-dependent electrical resistivity curves were measured using the Van Der Pauw (VDP) configuration.In addition, optical properties were addressed by Spectroscopic Ellipsometry (SE) to illustrate how physical properties in the V 2 O 3 thin films can be fully controlled by epitaxial strain.
The results show an effective transfer of strain achieved by heteroepitaxy leading to a Mott MIT in a wide range of temperatures around RT in 1.5% Cr-doped and pure V 2 O 3 compounds.We observed a colossal change in room temperature resistivity (RTR) with ∆R/R up to ~100,000% as well as a significant change in optical constants associated with the strainmodulated bandgap of these Mott insulators.Based on these results, a novel phase diagram of V 2 O 3 thin films was proposed with the RT in-plane parameter as variable.These results demonstrate that meticulous strain engineering is a powerful tool to tune physical properties of strongly-correlated electron systems and also allows accessing novel states unaccesible in bulk.The stabilization and control of the MIT at RT in 1.5% Cr-doped and pure V 2 O 3 thin films is an important achievement towards implementation of Mott materials in devices.

A. Thin film growth
All the thin film layers were deposited on (0001)-Al 2 O 3 substrates by oxygen-assisted molecular beam epitaxy (MBE).The vacuum chamber had a base pressure of 10 −9 Torr and it was equipped with RHEED to confirm the epitaxial growth in situ.The oxide heterostructures consist of a 40 nm Cr 2 O 3 epitaxial layer followed by a 30 nm thick (Cr 1 − y Fe y ) 2 O 3 solid solution (y = 0 corresponds to a 40 nm Cr 2 O 3 only).To guarantee that all V 2 O 3 film compounds have the same chemical interface with the heterostructure, an ultrathin 3 nm Cr 2 O 3 layer was deposited on the (Cr 1 − y Fe y ) 2 O 3 alloys with y > 0. Finally, the 60 nm 1.5% Cr-doped and pure V 2 O 3 films were deposited on the artificial substrates composed by the oxide heterostructures.

B. Structural characterization
The structural properties of the samples were characterized by means of high-resolution x-ray diffraction (XRD) with a Panalytical X'pert Pro diffractometer using monochromatic K α1 radiation of a Cu anode.Out-of-plane measurements in the θ /2θ configuration were done around the symmetric peak (0006) and 2-axes measurements -reciprocal space maps (RSM) -around the asymmetric peak (1 0 1 10) were performed to extract in-plane and out-of-plane lattice spacings.Besides, the structural quality and composition were also addressed by transmission electron microscopy (TEM) and Energy dispersive X-ray spectroscopy (EDX), respectively.TEM and EDX analysis were performed using a probelens corrected JEOL ARM200F scanning transmission electron microscope operating at 200 kV and equipped with a cold-field emission gun and a Centurio energy dispersive xray spectrometer.The TEM samples were prepared by means of a focused ion beam (FIB).

C. Electrical and optical characterization
Temperature-dependent resistivity measurements in the Van der Pauw (VDP) configuration were performed in an Oxford Optistat CF2-V cryostat using a Keithley 4200-SCS parameter analyzer.The temperature sweep rate was of 1 K per minute.Finally, spectroscopic ellipsometry (SE) measurements from 70 meV to 0.5 eV in the MIR and from 0.5 eV (NIR) to 6 eV (UV) were performed at SENTECH Instruments GmbH in Berlin with a SENDIRA and a SENresearch 4.0 SER 850 DUV ellipsometers, respectively.The ellipsometer angles ψ and ∆ were measured at room temperature using three different angles of incidence (65 • , 70 • and 75 • ).Independent measurements on the layers in the oxide heterostructures were performed before the 1.5% Cr-doped and pure V 2 O 3 films were deposited and the data were fitted applying the Tauc-Lorentz oscillator model 51 .Then, the measurements were repeated on the stack with the top 1.5% Crdoped or pure V 2 O 3 films and the data were fitted applying the same optical model used for analyzing the layers in the heterostructure.

III. RESULTS AND DISCUSSION
Systematic straining of 1.5% Cr-doped and pure V 2 O 3 films was achieved by using oxide heterostructures consisting of several epitaxial thin films with corundum structure.The thickness of the 1.5 % Cr-doped and pure V 2 O 3 layers was kept to 60 nm.As there is a large lattice mismatch of > 4% between Al 2 O 3 and Cr-doped and pure V 2 O 3 16 , a 40 nm buffer layer of Cr 2 O 3 was first deposited.As demonstrated previously 24 , Cr 2 O 3 buffer layer can grow almost relaxed on Al 2 O 3 and enables the subsequent growth of V 2 O 3 structurally decoupled from the substrate with a lattice mismatch of about 0.1%.
In addition, Cr 2 O 3 has the additional advantage that it can be alloyed with Fe. (Cr 1 − y Fe y ) 2 O 3 also forms a corundum structure with the in-plane lattice parameter that can be changed by more than 1% (between a = 4.959 Å and a = 5.035 Å) 25 by varying the Fe content y.On top of this (Cr 1 − y Fe y ) 2 O 3 layer, an ultrathin Cr 2 O 3 layer (3 nm) was deposited that remained coherent/pseudomorphic with the underlying (Cr 1 − y Fe y ) 2 O 3 layer.This ultrathin Cr 2 O 3 layer ensured that the 1.5% Cr-doped and pure V 2 O 3 films deposited on the buffer stack experienced exactly the same chemical interface so that possible chemical effects of the underlying layer can be excluded.Figure 1e shows a cross-section EDX map obtained from a 1.5% Cr-doped V 2 O 3 film grown on the buffer stack described above.The color code linked to different Cr concentration in the layers reflects the layout of the final heterostructure used for this study.The Cr doping of 1.5% was specifically chosen to ensure the presence of a single PI phase at RT while in pure V 2 O 3 a single PM phase is expected 1,2 .This allows us to address the effect of strain engineering on both phases involved in the Mott transition.
The in-plane and out-of-plane lattice parameters of the strained (doped and undoped) V 2 O 3 films were determined by means of extensive HRXRD analysis.Figures 1a and b show the RSMs of the 1.5% Cr-doped and pure V 2 O 3 films, measured around the (1 0 1 10) peak of each layer in the stacks.They clearly demonstrate that all layers in the stacks are single-phase, epitaxial and preserve the corundum structure.No secondary phases were detected.The excellent structural quality is evidenced by the presence of finite-size oscillations in out-of-plane θ /2θ XRD measurements (see Fig. S1 in the supplementary information).
The (1 0 1 10) peak positions of the Cr 2 O 3 /(Cr 1 − y Fe y ) 2 O 3 /Cr 2 O 3 stack -and the lattice parameters (a and c) -confirm that the engineered buffer has an in-plane lattice parameter effectively decoupled from the substrate.The blue and red dots in the RSMs in Fig. 1a and b mark the bulk lattice parameters of the PM and PI phases, respectively.By increasing the Fe content in the (Cr 1 − y Fe y ) 2 O 3 , the lattice parameters a and c of the 1.5% Cr-doped and pure V 2 O 3 films gradually move from the blue dot position (corresponding PM with a PM , c PM ) to the red dot position (corresponding PI with a PI , c PI ).For the Fe concentration of y = 1 (Fe 2 O 3 ), the lattice parameters go even beyond the PI bulk value.
In-plane lattice parameters of the films are calculated based on the RSM data and summarised in Fig. 1c as a function of the Fe content y together with the corresponding values for the buffer layers.The data show that the 1.5% Cr-doped and pure V 2 O 3 films roughly follow the in-plane lattice parameter evolution of the buffer layers.It is important to note that no systematic widening of the RSM peaks is observed as the Fe content changes (except for the cases with y = 1) in the buffer layers.Moreover, by using the Scherrer equation for the outof-plane diffraction peaks we found that the crystalline coherence is almost the same as the thickness extracted from x-ray reflectivity measurements (only an average difference of 3%) evidencing a good uniformity of the films lattice structure.
These results indicate that the designed buffer layers are sufficient to grow films with a well-defined lattice constant across the whole film.In the case of y = 1 the observed RSM peak broadening could be due to local inhomogeneities or oxygen vacancies that can be more favorable due to the larger tensile strain in these films.
The in-plane lattice parameter of the (Cr 1 − y Fe y ) 2 O 3 layer (light-blue squares) increases linearly with y (only the data point for y = 1 deviates from this linear trend), evidencing that the produced buffer effectively tailors the amount of induced strain as a function of Fe content.From the in-plane lattice parameters, the resulting in-plane strain was calculated as ε in-plane = (a film -a bulk )/a bulk , with a bulk equal to a PI (a PM ) for 1.5% Cr-doped (pure) V 2 O 3 films.Figure 1d shows the evolution of the c/a ratio as a function of the strain in the Cr-doped (left) and pure (right) V 2 O 3 films.The continuous decrease of c/a between the PM and PI bulk values (marked by blue and red line, respectively) with ε in-plane indicates that the c-axis adjusts to the imposed in-plane deformations.The measured changes in the c/a ratio of our strained films slightly deviate from what is expected considering only an elastic deformation: The Poisson ratio for the PI phase in 1.5% Cr-V 2 O 3 for ε in-plane = -0.8yiels c/a = 2.831 26 while we find c/a = 2.822 experimentally.The most remarkable result is the continuous change in c/a between the PM and PI phases, in contrast to the discontinuous one observed in single crystals 2 .This deviation indicates that the bulk and thin films of doped and pure V 2 O 3 behave differently allowing us to explore states with intermediate lattice parameters that are not stable in bulk samples.As highlighted in Fig. 1d, in pure V 2 O 3 films only tensile strain is possible because the Cr 2 O 3 layer has nearly the same in-plane lattice parameter as the PM phase, and alloying with Fe leads to larger in-plane parameters.On the contrary, both compressive and tensile strains are accessible for the 1.5% Cr-doped V 2 O 3 films, which should be in the PI phase in the unstrained bulk phase.
Figure 1e shows a representative color-coded EDX map of the entire stack illustrating the sharp interfaces present across the different layers in the oxide heterostructure.Comparable results were obtained for 1.5% Cr-doped (pure) V 2 O 3 films on the different buffers.In particular, systematic EDX line-scans across the interface between the ultrathin 3 nm Cr 2 O 3 layer and the 1.5% Cr-doped and pure V 2 O 3 layers confirmed that there is no Cr diffusion into the top vanadium oxide layer.The Cr content in the Cr-doped films was estimated to be 1.5% ± 0.3%.Also the HRTEM image and corresponding Fast Fourier Transform (FFT) (shown in the inset) demonstrate the single-crystalline character of the top layer across its whole thickness.The lattice spacings determined by HRTEM were in agreement with those extracted from HRXRD (Fig. 1c).
The temperature-dependent electrical resistivity curves R(T) of the 1.5% Cr-doped and pure V 2 O 3 films are shown in Fig. 2a and b, respectively, along with the reference curves (grey dashed lines) originating from 1% Cr-doped and pure V 2 O 3 single crystals 27 .Note that the (Cr 1 − y Fe y ) 2 O 3 layers are insulating in agreement with the band gaps of 2.1 eV and 3.2 eV for (Fe 2 O 3 ) and (Cr 2 O 3 ) 28 , respectively, therefore the transport measurements solely probe the V 2 O 3 thin films.
The low temperature (LT) transition can be seen in all samples, except in those with the largest in-plane lattice parameter (dotted red curves in each figure), which correspond to the highest Fe content y = 1 in the (Cr 1 − y Fe y ) 2 O 3 buffer.The T MIT (marked with star symbols in Fig. 2a and b) was determined from the inflection point in the dLnR(T)/dT curve and increased from 150 K to 180 K with the in-plane lattice parameter in both series.This is similar to the change in T MIT upon Cr doping in bulk (see for example the change in T MIT in the bulk 1.5% Cr-doped and pure V 2 O 3 samples shown in Figs.2a and b).Nevertheless, there are differences between the strained films -at constant doping -and bulk samples as a function of doping, such as the lack of sharp transitions and significantly reduced resistance jumps at the transition temperature.These differences can be due to several reasons including the in-plane clamping of the film lattice parameters that hampers the structural phase transitions as well as small variations in composition such as Cr-doping and oxygen content.It has been previously reported that increase of tensile strain 29 or oxygen deficiency 30 can cause an increase of RTR and a decrease of resistance ratio across the MIT.Therefore, possible formation of oxygen vacancies could explain the crossing of the R(T) curve for the most strained films (red dotted curves) with the one of films with smaller lattice parameter (yellow dotted curves).
The PM-PI transition close to RT is, as expected, completely absent in the 1.5% Cr-doped V 2 O 3 films for all inplane parameters, most probably due to the clamping effect 16 .Nonetheless, a significant change of the RTR occurs when changing the in-plane lattice parameter of the films.This trend is also observed in the pure V 2 O 3 thin film series (Fig. 2b). Figure 2c highlights the RTR values as a function of the in-plane lattice parameter for both film series, which clearly shows a PM to PI transition triggered by epitaxial strain.The most striking result is that an in-plane lattice deformation of ~1% can trigger a ∆R/R as large as 100,000 % in both the strained 1.5% Cr-doped (yellow circles) and the pure (green circles) V 2 O 3 films.To our knowledge this is the largest change in RTR achieved by epitaxial strain in V 2 O 3 compounds 20,22,32 .
The RTR of the 1.5% Cr-doped V 2 O 3 film with the smallest in-plane lattice parameter (a = 4.957 Å) is closer to the bulk RTR value of the PM phase of pure V 2 O 3 (marked by the grey cross PM undoped ) than to the one of the PM phase of 1% Crdoped V 2 O 3 compound (marked by the blue cross PM 1% Cr ).This low RTR value suggests that the metallic state achieved by strain in 1.5% Cr-doped V 2 O 3 thin films is mostly determined by a collective modification of the crystal lattice and electronic band structure in contrast to bulk samples, where the non-perfectly homogeneous distribution of Cr dopants can lead to electron scattering and to higher RTR.
Optical studies have been invaluable for understanding the underlying physics in V 2 O 3 systems 7,[34][35][36][37][38] because electron correlation effects appear as distinct signatures in the optical conductivity spectra of metallic and insulating phases and the Mott MIT is accompanied by large spectral weight changes 33,39 .The optical conductivity spectra are shown in Fig. 3a for the strained 1.5% Cr-doped (top) and pure (bottom) V 2 O 3 films.The features (labelled from i to iv) are assigned to the optical transitions depicted in Fig. 3b.The scheme in Fig. 3b shows the density of states (DOS) of metallic (PM) and insulating (PI) phases based on the one-band Hubbard model 33 , in which the vanadium 3d band is composed of the quasiparticle peak (QP) around the Fermi level (in the PM phase, defining feature i) and the incoherent bands (so-called Low Hubbard Band (LHB) and Upper Hubbard Band (UHB)) emerging at E -E F = ± U/2, each of width W and separated by the Coulomb energy U. Note that the energy scale of the optical transitions (ii) and (iii) are in the order of U/2 and U, respectively, and the Mott gap in the insulating state is in the order of ∆* = U -W.
Although in the case of V 2 O 3 the realistic multi-orbital scenario of the V 3d bands (i.e., Hubbard bands and QP are composed of two different orbitals) makes the DOS and optical transitions much more complex 36,40 , the optical responses across the Mott MIT can be explained semiquantitatively based on the one-band Hubbard model for the optical conductivity 33 .The real part of the optical conductivity σ 1 , defined as σ 1 (ω) = 2ε 0 nk, with ω the angular frequency and ε 0 the vacuum permittivity 41 , has been calculated from the optical constants (index of refraction, n and extinction coefficient, k) derived by Spectroscopic Ellipsometry (SE) experiments performed at RT in an energy range from 0.5 eV to 6 eV (see supplementary information for details).
Consistent with the one-band Hubbard model 33 , a signature of the Drude conductivity (feature i, associated with optical transitions within the quasiparticle (QP) peak in the PM phase) is spotted at low energies in both film series with the smallest in-plane lattice parameters.The effects of electron correlation appear in the visible range as features ii (~1.4 eV in film spectra) and iii (~2.8 eV in film spectra), associated to optical transitions between the QP peak and the Hubbard bands, and between LHB and UHB, respectively.Finally, the peak around 4 eV (labelled as iv) observed in all films is related to transitions from the oxygen 2p bands.
As can be seen in Fig. 3a, the features below 2 eV (i and ii) in σ 1 significantly decrease, when moving from the PM to the PI state triggered by epitaxial strain, owing to the spectral weight transferred between the Hubbard bands and the QP.The drastic decrease in σ 1 in this low energy range for the insulating films is directly related to the suppression of the QP and suggests the gradual opening of a gap across the Mott MIT in both strained V 2 O 3 thin film series.
The data in Fig. 3a also reveal isosbestic points near 2.1 eV in both series (marked by circles).Such a point is a characteristic feature of Mott transitions 42 , which has been observed in V 2 O 3 thin films, single crystals 36,37 and VO 2 films 38,43 but for temperature-dependent optical conductivity spectra, which intersect at the same σ 1 and energy values due to redistributions of spectral weight associated with a change from itinerant to localized electrons.Consequently, the presence of an isosbestic point in the RT optical conductivity spectra of both strained film series further suggests that a Mott phase transition can be triggered by continuous lattice deformation induced by heteroepitaxy.In Fig. 3c the optical conductivity spectra extended down to the mid-infrared (MIR) regime of the 1.5% Cr-doped and pure V 2 O 3 films with the smallest and largest in-plane lattice parameters are shown.For comparison, the optical conductivity curves obtained from single crystals (1.1% Cr-doped V 2 O 3 single crystal in the PM (220 K) and PI (300 K) phases and pure V 2 O 3 single crystal in the AFI (100 K) phase) 7,34 are added.These bulk data illustrate that there is only a small drop in spectral weight below 1 eV linked to the opening of a band gap of ∼ 0.12 -0.2 eV 44 for the Mott PM-PI transition in bulk.In contrast, the changes in the strained films are spectacular with a significant spectral weight drop and opening of much larger band gap of up to ∼ 0.5 eV.There has been no comparable precedent report in the literature.The gradual decrease of σ 1 for energies below 1 eV as the lattice parameter increases, and the films become more insulating, indicates that the large spectral weight at low energies might not be an intrinsic property of the PI phase.In addition, it is remarkable that the evolution of σ 1 of the more insulating films resembles that of the reference V 2 O 3 single crystals in the AFI phase, for which a large band gap of 0.6 eV opens 45 .Our optical data (features ii and iii) indicate that U in the films is of the same order as in bulk.Therefore, a bandwidth reduction as a function of a-axis expansion is possibly the origin of the larger gap in the PI phase in films with a > 4.985 Å, as the Mott gap is The modulation of the bandgap, associated with intermediate states stabilized across the Mott transition, allows a considerable range of tuning of the optical constants at RT.For example, Fig. 3d displays the optical constants n (top) and k (bottom) at the photon energy of 0.8 eV -which corresponds to a wavelength of 1550 nm typically used in telecommunications-across the Mott transition at RT in both series.At this photon energy, n varies between ∼ 2.3 and ∼ 3 while k changes between ∼ 2.2 and ∼ 0.5 as a function of the in-plane lattice parameter.These trends and values are similar to those reported for the temperature driven MIT in VO 2 43,46 , however, the transition in the latter occurs at about 70 • C whereas our films show a transition at RT, which is beneficial for device applications.It also has to be noted that the optical constants for V 2 O 3 compounds have not been reported previously.For example, the n and k values of the 1.5% Cr-doped V 2 O 3 films in the whole NIR-UV energy range can be found in Fig. S2.4 of the supplementary information.
The structural, electrical and optical results reported here for pure and Cr-doped V 2 O 3 thin films indicate that the wellknown temperature -doping phase diagram (see Fig. 4(top)) for bulk V 2 O 3 can not be used to represent the different phases achieved via strain engineering.Figure 4(bottom) shows the proposed phase diagram where temperature and in-plane lattice parameter are the variable axes.The absence of a PM-PI phase transition with temperature in pure and 1.5% Cr-doped V 2 O 3 thin films suggests that the 1st order Mott PM-PI transition line (dashed line in Fig. 4(top)) in the bulk phase diagram should be replaced by two straight lines comprising a region of intermediate states between the stable PM and PI ones.These intermediate states are characterized by lattice parameter, RTR and optical constant values between the bulk PM and PI ones.Although these results point to the stabilization of homogeneous intermediate states, local techniques will be necessary to investigate whether there is phase coexistence at nanometric scales.These novel intermediate phases, accessible thanks to the systematic control of strain by heteroepitaxy, can be very interesting for applications based on V 2 O 3 compounds.For example, the changes in resistivity vs lattice parameter could be exploited by inducing lattice deformations when applying an electric field to a nearby piezoelectric material 20 or by depositing the material on a membrane, the effect can be used as a piezoresistive transduction mechanism.
The observation of a PM-PI phase transition by strain engineering in a wide range of temperatures (around room temperature) shows that the Mott transition can be solely driven by lattice deformation.In contrast, the temperature-driven transition to the AFI phase (yellow and green circles in Fig. 4(bottom)) is preserved for the whole range of in-plane lattice parameters studied in this work 47 .Notably, the temperature dependence of the boundary between the PM (or PI) and AFI phases resembles the one for the bulk samples.The robustness of both PM-AFI and PI-AFI transitions against substrate clamping can be related to the considerably larger thermodynamic enthalpy of formation (∆H = 2.03 kJ/mole and 2.36 kJ/mole, respectively) as compared to the enthalpy of formation for the PM-PI (∆H = 0.15 kJ/mole) transition 48,49 .

IV. CONCLUSION
In this work, the direct connection between the structure, electronic and optical properties has been exploited to control the Mott MIT phase transition at room temperature in V 2 O 3 thin film compounds.Intermediate states between PM and PI phases, which are not accessible in bulk, are realized by "smart" heteroepitaxy based on fabrication of tailored artificial substrates.It is demonstrated that the Mott MIT can be induced at RT both, in pure and Cr-doped compounds, by strain engineering, leading to films with a colossal ∆R/R ∼100,000 % and a broad range of optical constants.These results demonstrate that Cr doping is actually not the key control parameter for the RT Mott MIT, in contrast to the observations in the bulk.Moreover, strain engineering leads to a more homogeneous modification of Mott materials compared to chemical doping, where the modification occurs locally on specific sites.
The observed unique electrical and optical properties are attributed to the strain-modulated bandgap beyond the bulk value of 0.1 eV.Our data suggest that the significant spectral weight at low energies in the bulk PI phase, traditionally considered as a hallmark of Mott physics [11][12][13]50 , might not be an intrinsic feature but an artefact owing to local inhomogeneities. X-ay absorption and Photo-emission spectroscopy studies together with theoretical calculations considering the lattice deformations and local structures, could give more detailed insight into the changes in the electronic bandstructure and the origin of the modulated bandgap in the strained films.Finally, the stabilization of a Mott PM-PI transition, in both pure and Cr-doped V 2 O 3 thin films, around room temperature and the possibility to modify the structural, electrical and optical properties in a continuous manner can be used in more practical concepts such as pressure sensors, resonators, acoustic light modulators, and photonic devices in the near-infrared range.

FIG. 2 .FIG. 3 .
FIG. 1. Structural properties of strained 1.5% Cr-doped and pure V 2 O 3 films.a, b RSMs around the (1 0 1 10) reflections transformed to in-plane and out-of-plane lattice spacings of the 1.5% Cr-doped and pure V 2 O 3 films grown on the different (Cr 1 − y Fe y ) 2 O 3 /Cr 2 O 3 heterostructures.The PM and PI bulk lattice parameters are indicated by blue and red dots, respectively.c In-plane lattice parameters of the (Cr 1 − y Fe y ) 2 O 3 alloy layers in the oxide heterostructures (ligh blue squares) and the corresponding ones for the 1.5% Cr-doped (yellow circles) and pure (green circles) V 2 O 3 films.d c/a ratio as a function of the in-plane strain (ε in-plane ) induced by heteroepitaxy onto the 1.5% Cr-doped and pure V 2 O 3 films.In c and d, PM and PI bulk values are indicated by the blue and red lines, respectively.e (Left) Colorcoded EDX mapping of a stack representative of the 1.5 % Cr-doped V 2 O 3 film series and (Right) HRTEM image of the top layer and the corresponding Fast-Fourier Transform (FFT).