Interconversion of intrinsic defects in $\mathrm{SrTi}{\mathrm{O}}_3\left(001\right)$

Photoemission features associated with states deep in the band gap of $n-\mathrm{SrTi}{\mathrm{O}}_3\left(001\right)$ are found to be ubiquitous in bulk crystals and epitaxial films. These features are present even when there is little signal near the Fermi level. Analysis reveals that these states are deep-level traps associated with defects. The commonly investigated defects—O vacancies, Sr vacancies, and aliovalent impurity cations on the Ti sites—cannot account for these features. Rather, ab initio modeling points to these states resulting from interstitial oxygen and its interaction with donor electrons.

Figure 1

Reflection (a, b) high-energy electron diffraction (15 keV) and (c) low-energy electron diffraction (69 eV) for $n-\mathrm{SrTi}{\mathrm{O}}_3\left(001\right)$ after oxygen plasma cleaning.

Figure 2

VB spectra for $\mathrm{SrN}{\mathrm{b}}_x\mathrm{T}{\mathrm{i}}_{1-x}{\mathrm{O}}_3$(001) measured at photon energies of 6000 eV with $x=0.029$, and at 1487 and 21.2 eV with $x=0.010$ (same sample). The probe depths are $\sim 1, \sim 11$, and $\sim 32\mathrm{nm}$ for $h\nu =21.2$, 1487, and 6000 eV.

Figure 3

Core-level photoemission spectra for bulk $\mathrm{SrN}{\mathrm{b}}_x\mathrm{T}{\mathrm{i}}_{1-x}{\mathrm{O}}_3$(001) measured at x-ray energies of 6000 and 1487 eV, and at photoelectron take-off angles $\left({\theta }_{\mathrm{t}}\right)$ of ${90}^{\circ }$ (normal emission) and ${10}^{\circ }$ for $h\nu =1487\mathrm{eV}$.

Figure 4

(a) Simulations of the Ti $2p$ spectrum for $\mathrm{SrN}{\mathrm{b}}_x\mathrm{T}{\mathrm{i}}_{1-x}{\mathrm{O}}_3$ for $0.01\le x\le 0.10$ based on the assumption that every Nb donor results in one $\mathrm{T}{\mathrm{i}}^{3+}$ cation, as required for charge neutrality in a defect-free crystal. The simulations are linear combinations of spectra from pure, bulk $\mathrm{SrN}{\mathrm{b}}_{0.001}\mathrm{T}{\mathrm{i}}_{0.999}{\mathrm{O}}_3\left(001\right)$ and epitaxial $\mathrm{LaTi}{\mathrm{O}}_3\left(001\right)$. The difference spectra (simulation minus pure STO) show a transfer of intensity from the nominally $\mathrm{T}{\mathrm{i}}^{4+}$ binding energies to that for $\mathrm{T}{\mathrm{i}}^{3+}$ in both spin-orbital channels. (b) Comparison of the simulated and measured spectra for $x=0.01$.

Figure 5

VEELS measurements conducted with (a) 19 and (b) 38 mrad collection angles $\left(\beta \right)$, respectively, for a range of sample thicknesses. The spectra have been normalized to the signal integral. Simulated low-loss spectra for (c) 19 and (d) 38 mrad collection angles, respectively, for a range of sample thicknesses.

Figure 6

(a) VB and (b) gap spectra excited with 21.2 eV photons for bulk STO(001) crystals with a range of doping levels, as well as for epitaxial films. A 14-nm film of nominally undoped STO was deposited on $p$-Ge(001) by conventional MBE. The deposition conditions that were used lead to the incorporation of some $V_{\mathrm{O}}$. The h-MBE film is 70 nm $\mathrm{L}{\mathrm{a}}_{0.01}\mathrm{S}{\mathrm{r}}_{0.99}\mathrm{Ti}{\mathrm{O}}_3\left(001\right)$ grown on Nb:STO(001).

Figure 7

(a) VB and gap spectra excited with 21.2 eV photons for undoped STO(001) as cleaned and after anneals at $600{}^{\circ }\mathrm{C}, 700{}^{\circ }\mathrm{C}, 800{}^{\circ }\mathrm{C}$, and $900{}^{\circ }\mathrm{C}$ in UHV, and one at $600{}^{\circ }\mathrm{C}$ in activated ${\mathrm{O}}_2$. Also shown are the accompanying Ti $2p_{3/2}$ spectra excited with 1487 eV x rays. Insets: Expanded gap spectra (left) and a fit to Voigt functions (red) for the spectrum measured after the third anneal at $900{}^{\circ }\mathrm{C}$ (right). Slight changes in binding energy caused by variations in band bending (within $\sim 0.15\mathrm{eV}$ of flat band) have been removed to facilitate comparison. (b) Expanded view of experimental gap spectra. (c) Calculated DOS for select defect configurations broadened with Gaussian functions using a width of 0.5 eV. This panel shows how the DOS transform as a result of the defect reactions shown to the right. The solid magenta and solid cyan features in the purple DOS result from localized and delocalized electrons from $V_{\mathrm{O}}$, respectively. The numbers by the arrows are the energies gained in the indicated defect conversion reactions.

Figure 8

(a) Local structures for ${\mathrm{O}}_2$ at an empty anion site (${{\mathrm{O}}_2}^{2-}$) and (b) a pair of ${{\mathrm{O}}_{\mathrm{int}}}^{2-}$ that result from ${{\mathrm{O}}_2}^{2-}$ trapping $2e^{-}$ from nearby $V_{\mathrm{O}}$ and dissociating; Sr, Ti, and O are shown as green (large), blue (medium), and red (small) spheres. (c) Surfaces of constant spin density for the $2{{\mathrm{O}}_{\mathrm{int}}}^{2-}+{[3\times V_{\mathrm{O}}]}^{2+}$ configuration indicate that three electrons are localized in ${V_{\mathrm{O}}}^{+}$ and one electron is at the conduction-band minimum.

Figure 9

Densities of states calculated (a, b) for the neutral Sr vacancy and (c, d) for the Sr vacancy with two trapped electrons using (a, c) PBEsol and (b, d) HSE density functionals. The unoccupied states near the top of the valence band are marked with $h^{+}$ (a, b). Upon trapping two electrons per Sr vacancy, these states become occupied and produce a small perturbation of the DOS near the VB maximum (c, d), but do not create an in-gap state.

Figure 10

Comparison of VB and gap spectra for stoichiometric and $\sim 10\%$ Sr poor STO homoepitaxial films grown by hybrid MBE.

Figure 11

Deep in-gap state count rate plotted against the $\mathrm{Sr}{\mathrm{O}}_x$ peak area normalized to the lattice peak area taken from Sr $3d$ spectra for the annealing sequence for undoped STO(001) summarized in Fig. 3. The numbers indicate the steps in the annealing sequence shown in Fig. 3.

Figure 12

Densities of states calculated for a hydrogen atom trapped in $\mathrm{SrTi}{\mathrm{O}}_3$. The H atom binds to a lattice oxygen, forming an $\mathrm{O}{\mathrm{H}}^{-}$ species and an electron that occupies a state close to the bottom of the CB. DOS calculated using (a) PBEsol and (b) $\mathrm{PBEsol}+U$ for $U_{\mathrm{Ti}}=4\mathrm{eV}$ and (c) $U_{\mathrm{Ti}}=8\mathrm{eV}$ indicate that this electron state is degenerate with the CB minimum. This electronic structure is further supported by the DOS obtained using single-point HSE calculation for the structure preoptimized using (d) $\mathrm{PBEsol}+U$ ($U_{\mathrm{Ti}}=8\mathrm{eV}$).

Figure 13

Hall effect measurement measured at room temperature for a $\mathrm{SrN}{\mathrm{b}}_{0.015}\mathrm{T}{\mathrm{i}}_{0.985}{\mathrm{O}}_3\left(001\right)$ crystal from MTI.