Tuning electronic and magnetic properties through disorder in V2O5 nanoparticles

We report on the synthesis and characterization of V2O5 nanoparticles grown using a sol–gel method at different calcination temperatures. We observed a surprising reduction in the optical band gap from 2.20 to 1.18 eV with increasing calcination temperature from 400 to 500 °C. Raman and X-Ray diffraction measurements indicated slight changes in the lattice parameters induced by the growth process. However, density functional theory calculations of the Rietveld-refined and pristine structures revealed that the observed optical gap reduction could not be explained by structural changes alone. By introducing oxygen vacancies to the refined structures, we could reproduce the reduction of the band gap. Our calculations also showed that the inclusion of oxygen vacancies at the vanadyl position creates a spin-polarized interband state that reduces the electronic band gap and promotes a magnetic response due to unpaired electrons. This prediction was confirmed by our magnetometry measurements, which exhibited a ferromagnetic-like behavior. Our findings suggest that oxygen vacancies play a crucial role in band gap reduction and the promotion of a ferromagnetic-like response in an otherwise paramagnetic material. This provides a promising route to engineer novel devices.

www.nature.com/scientificreports/ In this paper, we prepared V 2 O 5 nanoparticles (NPs) using the sol-gel method 35 at various calcination temperatures (T CAL ). The T CAL regulates the particle sizes and general properties of the NPs. We characterized the samples with X-ray diffraction, Raman and UV/vis spectroscopies, Scanning Transmission Electron Microscopy (STEM), and Vibrating Sample Magnetometry (VSM). We report a remarkable reduction of around 1 eV of the optical band gap of V 2 O 5 NPs when increasing T CAL . By performing DFT calculations with Hubbard (U) 36 and van der Waals (D3) 37,38 corrections, we found that structural changes do not reduce the gap as much as observed experimentally. Instead, when including oxygen vacancies, the simulations show that a spin-polarized interband state appears, which decreases the band gap to values close to the experimental ones and induces a magnetic state. VSM measurements showed that the samples have a ferromagnetic-like behavior, which is probably attributed to oxygen vacancies and surface effects. Therefore, we show that the band gap and magnetism of V 2 O 5 NPs can be tuned with T CAL through oxygen vacancies, which might be relevant for future technological applications.

Methods
Experimental details. Vanadium pentoxide (V 2 O 5 ) nanoparticles were synthesized using a non-aqueous sol-gel route 35 . A volume of 0.5 ml of vanadium oxytrichloride (VOCl 3 ) of analytic degree was mixed with 20 ml of benzyl alcohol and stirred for 5 h at room temperature. The obtained blue solution was left to stand for 48 h and then heated at the target calcination temperature T CAL for 12 h 35 . The T CAL used were 400, 425, 438, 450, 475, and 500 °C for the first set of samples. Subsequently, with the second set of samples, the T CAL used were 460, 500, and 540 °C. The first set of samples was used for the optical measurements while the second one was for magnetic measurements. The particle size change when increasing T CAL 39,40 . X-ray diagrams of the first set were measured using a diffractometer PANalytical X'PERT PRO MPD with a Bragg Brentano configuration and an X-ray source Cu-Kα1 with 1.54060 Å wavelength. Raman spectra for the first set were obtained with a HORIBA XploRA One Raman spectrometer with shift values between 100 and 1200 cm −1 . A laser light source of 638 nm was used in these measurements. The Raman spectrum peaks were analyzed utilizing Voigt distributions with Fityk software 41 . Microscopy analysis was performed using the TES-CAN LYRA3 in a Scanning Transmission Electron Microscope (STEM) setup. Absorbance spectra were recorded with a spectrophotometer Analytik Jena Specord 50 Plus between 300 and 1100 nm. The sample powders were dissolved in isopropanol, obtaining an unsaturated solution 20,42 . Finally, magnetometry measurements were performed in powder samples in a vibrating sample magnetometer from LakeShore.
Computational details. The calculations were based on density functional theory using the software Vienna Ab initio Simulation Package (VASP) [43][44][45][46] adopting the exchange and correlation functional as parametrized by Perdew-Burke-Ernzerhof (PBE) 47 under the Generalized Gradient Approximation (GGA). The Projector Augmented Wave (PAW) method 48 was used to describe electron wave functions. An energy cutoff of 500 eV was sufficient for expanding the plane wave basis set. The orbital potentials method 36 (DFT + U) was used to correct for the electron correlation in localized vanadium 3d orbitals, using U = 4 eV and J = 0 eV, as suggested by Scanlon et al. 31 for V 2 O 5 . The van der Waals DFT + D3 method 37,38 was also used to correct the weak interactions between the V 2 O 5 layers. PyProcar 49 and VASPKIT 50 were used for the post-processing of VASP output data. The calculations were performed in bulk, so the characteristics of NPs, such as surface effects or interactions between them, were neglected (this approach has been successful with other systems 21,51 ).
We first applied the computational method to the pristine α-V 2 O 5 Pmmn structure with lattice parameters 17 a = 3.564 Å, b = 11.512 Å, and c = 4.368 Å with a k-point sampling of 8 × 8 × 8. Using the experimental X-ray structures as inputs, we calculated the band structures and the absorption spectra under the independent particle approximation increasing the number of bands by four times the default value. Then, we simulated systems with three concentrations of oxygen vacancies: 1.1%, 2.5%, and 5.0%. These concentrations were achieved by removing one O1 oxygen from pristine supercells of sizes: 3 × 1 × 3 for 1.1%, 2 × 1 × 2 for 2.5%, and 2 × 1 × 1 for 5.0%. These systems were studied with spin-polarized calculations. The first Brillouin zone was sampled by 3 × 8 × 3, 4 × 8 × 4, and 4 × 8 × 8 k-point meshes for 1.1%, 2.5%, and 5.0% systems, respectively. Atomic positions were relaxed maintaining constant the experimental lattice parameters until the force on every ion was less than 0.01 eV Å −1 . Oxygen vacancies in O2 and O3 positions were also investigated with the same methods as the O1 site.

Results and discussions
Figure 1a-c shows representative X-ray diffractograms of the samples. Using the Rietveld refinement, we identified a good overall match to a pure α-V 2 O 5 phase 52-54 with the most intense peaks of crystallographic planes (001), (110), and (040) indicated in Fig. 1a. Following the Scherrer method 55 , we estimated the primary particle sizes of the samples, which varied with T CAL and are of the order of 60 nm (see Supplementary Table S1). STEM images confirmed the order of magnitude of the particle sizes (see Supplementary Figure S1). Figure 1d-f shows Raman spectra of the same samples. The Raman peaks from 100 to 1200 cm −1 fully coincide with those of the α-V 2 O 5 phase 52,56 . Figure 1i shows a non-monotonic behavior of the unit cell volume (V cell ) as a function of T CAL . The V cell increased from T CAL 400 to 438 °C and decreased at 475 °C. The V cell varied by a maximum of 0.17% of the pristine value of 179.2 Å 3 reported in the literature 17 . Raman results also have a non-monotonic behavior after a Voigt peak fitting. Figure 1h shows the normalized Raman intensity for three representative peaks as a function of T CAL . Raman intensities have local minimums at 438 °C and the higher and smallest T CAL .
The non-monotonic behavior of the Raman intensities, and V cell can be associated with structural changes induced by the synthesis process. Since the Raman intensities are proportional to the square of the polarizability derivative with respect to vibration coordinates 57 , it is expected that for larger bond distances, the Raman peaks become less intense. This can be noticed in the sample at T CAL = 438 °C because it has the larger vanadyl www.nature.com/scientificreports/ distance so the weaker vanadyl Ag 991 cm −1 Raman mode (see Fig. 1d). In addition, the volume at 438 °C has a maximum because this sample has the largest c-lattice parameter. The Raman frequency of the vanadyl mode as a function of T CAL also followed a similar trend as V cell , which indicates that the fabrication process produces slight structural changes, especially at 438 °C (see Supplementary Figure S2). We investigated the correlations between the crystal structure and the optical properties of the V 2 O 5 NPs by measurements of the optical absorption at room temperature. Figure 2a-c shows the absorption curves as blue lines. The optical band gap E g opt was determined using the Tauc plot method 24 to these curves assuming an indirect transition. The results are summarized in Fig. 2d. Notably, the optical band gap decreases considerably when increasing T CAL . The total change between the T CAL = 400 and 500 °C is close to 1 eV. These values appear to be extremely small compared to values reported by V 2 O 5 in bulk using the same Tauc plot technique 58 . Additionally, at T CAL = 438 °C, there is a minimum of E g opt . This minimum corresponds to the sample with the greater vanadyl distance and unit cell volume (see Fig. 1). This may suggest that local structural changes may control the optical properties. However, it appears that the overall decrease of the optical band gap cannot be solely explained by the structural parameters obtained from the X-ray and Raman measurements. To identify the contributions to the optical band gap due to the lattice structure, we carried out first-principles calculations.
Computational results. We first calculated the band structure of the V 2 O 5 pristine crystal (see Supplementary Figure S3), which shows that vanadium 3d orbitals (especially d xy ) predominate the conduction band and that there is one d-band slightly separated from the others. We found an indirect electronic band gap of E g ele = 2.42 eV, which is in good agreement with experimental 23 (2.33 eV) and theoretical data 59 . Then, we used the refined X-ray crystal structures at different T CAL as inputs and calculated their band structures (see Fig. 3ac). Overall, the band structures show a similar behavior compared to the pristine one. From the band structures, we calculated the corresponding E g ele values. Figure 2d shows how E g ele compares with the corresponding experi- www.nature.com/scientificreports/ mental values (see the orange squared symbols). The E g ele values versus T CAL match moderately with the experimental results at T CAL = 400 °C. Furthermore, the calculations appear to capture the minimum E g ele at 438 °C. However, the overall values calculated for the band gap are much higher than the ones observed experimentally. One reason may be due to the presence of excitons which may change the optical band gap considerably 33 . We then proceeded to calculate the optical band gap from the calculations. We calculated the optical absorption curve under the independent particle approximation (see green curves in Fig. 2a-c) and employed the Tauc plot method to estimate the optical band gap, as in the experimental case. Figure 2d show the calculated band gap versus T CAL (green diamond symbol). All values are close to 2.8 eV with little change. Furthermore, they are far away from the ones obtained experimentally. Therefore, the structural changes observed in X-ray diffraction are minor and do not fully explain the behavior of the band gap with T CAL . Similarly, Kang et al. 60 found that augmenting the temperature of α-V 2 O 5 film samples while measuring, largely modifies their optical properties, but structural modifications do not completely explain these changes.
A reason that might explain the differences between theory and experiment is the existence of oxygen vacancies 13,14,61 . Exposing vanadium pentoxide samples to 1 min of air creates oxygen vacancies, which changes the material properties 62 . Thus, it is unlikely that the various synthetic methods to prepare this material, including the sol-gel synthesis 35 , produce V 2 O 5 with total purity but instead with a certain degree of oxygen vacancies. Accordingly, the experimental properties available in the literature, such as the band gap, might be affected by this situation. Therefore, we simulated three V 2 O 5 systems with 1.1%, 2.5%, and 5.0% concentrations of oxygen vacancies. The energy required to remove an O1 oxygen required about 2 eV less than the O2 or O3 types (see Supplementary Table S2). This result is consistent with previous reports 31 . Thus, we focused our calculations on the O1 defect.
After relaxation, the new geometries of the 1.1% and 2.5% systems formed a new bond between the vanadium vacancy site (V vac ) and the O1 of an adjacent layer (see inset of Fig. 3e and Supplementary Figure S4). These results are similar to the ones of Scanlon et al. 31 , who also found a new bond between layers. The 5.0% system had slight local changes since in this case there is no adjacent O1 oxygen (see Supplementary Table S3 for specific distances and angle variations). These geometrical changes might be explained by the chemical bonds formed. The V-O1 bond is analogous to the vanadyl group VO 2+ , which appears in many inorganic species of this metal.  63 . The vanadium coordination number in these molecules is five. When an O1 atom is removed in V 2 O 5 , the atom V vac (which acts as a Lewis acid) becomes four-coordinated, forming the less stable square geometry. Therefore, V vac binds to O1 layer2 and recovers the coordination number of five and the square pyramid structure (recall atom labels from Supplementary Figure S4). The relaxed oxygen-deficient structures also showed a change in their symmetry operations. The symmetry of these systems shifted from pristine 22 Pmmn D 13 2h to Pm C 1 s , as calculated using FINDSYM 64 . We also simulated the X-Ray diffractograms of these systems using the VESTA 65 software and found no difference against the pristine structure (see Supplementary Figure S5). Thus, we confirmed that X-Ray diffraction cannot detect the studied vacancy concentrations.
When including oxygen vacancies, another relevant aspect to consider are electric charges. The process of releasing one oxygen from V 2 O 5 creates a vacancy site and lefts 2 electrons in excess in the lattice. According to the spin densities in the inset of Fig. 3e (and Supplementary Figure S4), the two released electrons localize mainly on V vac . Thus, one might assign to V vac an oxidation state of + 3 for 1.1% and 2.5% systems or + 4 for the 5.0% system. Such inferences correspond to the fact that signals of V 4+ and V 3+ are observed in X-ray photoelectron spectra when heating V 2 O 5 thin films under an ultrahigh vacuum chamber 66 . The results also align with other calculations 67 . In addition, V cell tends to expand with oxygen vacancies because reduced cations have larger ionic radii 68 , and V cell was indeed greater than the pristine value of 179.2 Å 3 for various T CAL (see Fig. 1i). Figure 2e shows the simulated Tauc plot relation for the 2.5% oxygen-deficient system, and Fig. 3d presents its electronic band structure. The band structure shows an interband state which reduces ~ 1 eV the band gap compared to the pristine value. Similarly, the simulated Tauc plot curve of this system has a straight segment at low energies associated with a band gap energy of 1.45 eV. Thus, both the electronic and the Tauc-simulated band gap give values that are much closer to the experimental ones at high T CAL . These findings are also in agreement with X-ray and ultraviolet photoelectron spectroscopies signals of an interband state at 1.3 eV above the valence band of thin film V 2 O 5 samples 66,69 .  www.nature.com/scientificreports/ The interband state is also below Fermi energy and is populated by excess electrons. This suggests that V 2 O 5 acts as an n-type semiconductor, which agrees with the literature 5,70 . Figure 3d and Supplementary Figure S6 present partial densities of states of the oxygen-deficient systems and reveal that the interband state is mainly composed of V vac 3d orbitals, specifically d yz and d xz ones for the 1.1% and 2.5% systems. For the 5.0% system, two interband states resulted. The band farther from the valence is mainly made by the d xz orbital, and the one closest to the valence is made by the d z 2 orbital. Similar results were obtained by Blum et al. 14 and Laubach et al. 69 . These results support the hypothesis that mainly oxygen vacancies and, to a less extent, structural changes are the origin of the reduction of the band gap with T CAL .
Moreover, the calculations on the oxygen-deficient systems show that their interband state is spin-polarized, so they have a magnetic behavior. To verify this, we performed VSM measurements of samples at T CAL = 460 and 500 °C since the optical band gap showed greater changes at high temperatures. Figure 3e shows the magnetization as a function of the external magnetic field. Indeed, the results indicate that V 2 O 5 NPs have a ferromagneticlike behavior. The saturation magnetization was considerably greater at 460 °C than at 500 °C. This implies that T CAL tunes the magnetic properties of V 2 O 5 NPs. However, further studies should be performed to fully understand how the magnetization scales with T CAL , including a complete range of temperatures.
It is known that many nonmagnetic metal oxides exhibit ferromagnetic ordering when they form NPs because of the interactions between spin moments resulting from oxygen vacancies at their surfaces 71 . Therefore, the V 2 O 5 ferromagnetic-like state observed experimentally might be ascribed to oxygen vacancies on the surfaces of the NPs. Our magnetic results also align with other experimental evidence such as the ferromagnetic response found in oxygen-deficient thin film V 2 O 5 systems 27 . The results are also supported by other theoretical calculations that indicate that the ferromagnetic solution is more stable than the nonmagnetic or antiferromagnetic ones in oxygen-deficient V 2 O 5 systems 25 .
The magnetic results were understood based on the O1 type of oxygen vacancies, but O2 and O3 defects may also exist. The O1 vacancy requires the least energy and is the most probable to be present 25,31 (see Supplementary  Table S2). Nevertheless, we also studied the electronic structure of the O2 and O3 systems for 5.0% and 2.5% concentrations. Intriguingly, the calculations show that oxygen vacancies at these positions do not produce a spin imbalance and thus do not contribute to the material's magnetic properties (see Supplementary Figure S8). This is attributed to the distinct chemical environment of the O1 vacancy compared to the O2 and O3 ones. In the O1 vacancy, the excess charge is localized mainly on one vicinal vanadium atom. However, with the O2 and O3 types, the electrons in excess are delocalized mainly on the two or three vicinal vanadium atoms. Therefore, the magnetic properties observed experimentally are only ascribed to O1 vacancies.
On the other hand, the saturation magnetization might be used to approximate the concentration of oxygen vacancies. Because each oxygen vacancy adds 2 µ B to the system, there is an exact linear relationship between the magnetization per molecular formula (M) and the oxygen vacancy percentage (C), which is M = αC, with α = 0.1 µ B , between the concentrations studied. This fact was also previously reported in the literature 25 . Assuming that the NPs are spherical and have a size equal to the average primary particle sizes of Supplementary Table S1, we can roughly estimate the C on the surfaces of the NPs using their experimental magnetization (see Supplementary Note for more detail). Following this reasoning, we estimate concentration values of 0.25% and 0.02% for T CAL = 460 and 500 °C, respectively. These values appear to agree with the percentual changes in V cell . As mentioned previously, oxygen vacancies are expected to increase V cell , and the cell volume of the samples at T CAL = 450 and 500 °C expanded by 0.12% (see Fig. 1i). This is the same order of magnitude of the concentration estimated with the magnetization at T CAL = 460 °C.
With these results, a question arises about how the concentration of oxygen vacancies relates to T CAL . Intuition suggests that a greater T CAL can increase C since a higher T CAL provides more thermal energy that can promote reduction-oxidation reactions, but these reactions may also result in the recombination of vacancies due to thermal energy. If we assume a direct relationship between C and V cell , Fig. 1i indicates that C varies nonmonotonically with T CAL . This suggests that at higher temperatures, the fabrication process may not promote as many vacancies as at intermediate temperatures. In addition, the C value estimated with the magnetization reduced from 0.25% at 460 to 0.02% at 500 °C. Thus, the dependence of oxygen vacancy concentration with T CAL is probably not straightforward, and more studies are required to address this correlation.

Conclusions
We have synthesized V 2 O 5 nanoparticles using a non-aqueous sol-gel method at different calcination temperatures and characterized their properties using structural, optical, and magnetic techniques. Our results demonstrate that the calcination temperature plays a crucial role in determining the properties of the nanoparticles. The nanoparticles' optical band gap decreased from 2.20 to 1.18 eV with an increase in calcination temperature from 400 to 500 °C. While the structural changes induced by calcination were found to be minor, our DFT + U + D3 calculations suggest that the reduction in the band gap is mainly due to the presence of oxygen vacancies at the vanadyl position that induce a spin-polarized interband state, resulting in the promotion of unpaired electrons. Furthermore, our magnetometry measurements reveal that the nanoparticles exhibit a ferromagnetic-like behavior, which may be attributed to oxygen vacancies. However, further studies are needed to understand the details of the magnetism observed and to quantify the number of vacancies induced by the growth process. Our study highlights the importance of oxygen vacancies in modifying the band gap and magnetic properties of V 2 O 5 nanoparticles and suggests that these properties could be harnessed for developing novel devices for photonics and neuromorphic computing.