The blue emission at 2.8 eV in strontium titanate: evidence for a radiative transition of self-trapped excitons from unbound states

ABSTRACT The origin of the blue emission in SrTiO3 has been investigated as a function of irradiation fluence, electronic excitation density, and temperature using a range of ion energies and masses. The emission clearly does not show correlation with the concentration of vacancies generated by irradiation but is greatly enhanced under heavy-ion irradiation. The intensity ratio of the 2.8 and 2.5 eV bands is independent of fluence at all temperatures, but it increases with excitation rate. The 2.8 eV emission is proposed to correspond to a transition from conduction band states to the ground state level of the self-trapped exciton center. GRAPHICAL ABSTRACT IMPACT STATEMENT A novel mechanism attributes the origin of the intriguing blue band in SrTiO3 to a non-localized transition of the self-trapped exciton center from conduction band states to the ground level.


Introduction
Strontium titanate (SrTiO 3 ) is a model transition metal oxide, receiving intensive attention due to its many applications. This multifunctional perovskite ceramic has remarkable physical and chemical properties, including high temperature superconductivity, photocatalytic behavior and 'colossal' magnetoresistance [1][2][3][4]. It is often considered to constitute the basis for oxide-based microelectronics [5]. The electronic and optical behavior is presently an active and controversial field of research [6][7][8][9]. The luminescence spectra under a variety of excitation sources, including UV light [8][9][10][11][12][13][14][15][16][17][18], X-rays [19], electrons [20] and ion-beams [21,22], show three main bands centered at 2.0 eV (red), 2.5 eV (green) and 2.8 eV (blue). A detailed study using ion-beam irradiation experiments has concluded that the red band is due to d-d transitions between an excited level in the conduction band (CB) with mostly 3d(t 2g ) character and an in-gap 3d(e g ) level associated with an electron selftrapped as Ti 3+ adjacent to an oxygen vacancy (Ti 3+ -V O center) [23][24][25][26][27][28], as illustrated in Figure S1 (see supplementary data online). It is generally accepted that the green 2.5 eV emission band is associated with a triplet-singlet optical transition of a self-trapped exciton (STE) [8,9,19,29]. The situation for the 2.8 eV band is more controversial. A number of proposals on the origin of this emission have been advanced, including its relationship with oxygen vacancies [9][10][11]. Several authors [11,12,17] have proposed that the blue band may be associated with a transition from the conduction band minimum (CBM) to the in-gap level of a self-trapped hole (STH) [30]; whereas others [6] suggest a transition from a self-trapped polaron level at oxygen vacancies to the valence band (VB). Due to disagreement and inconclusive identification on the origin of the blue emission band peaked at 2.8 eV, it appears relevant to tackle this puzzling scenario with an alternative strategy using ion beam induced luminescence or ionoluminescence (IL). Ion excitation has several key advantages over other excitation sources, which have not been, so far, sufficiently recognized: 1) it is a real-time in-situ technique with a very broad energy excitation spectrum in contrast to laser pulses; 2) it allows for an adjustable balance between the generation of point defects (associated mostly with elastic collisions) and electronic excitation, which can be modified through the choice of mass and energy of the incident ion; and 3) it has an easily adjustable excitation rate to investigate the role of electronic excitation density on the emissions.

Materials and methods
High-purity, epi-polished, stoichiometric SrTiO 3 (001) single crystals, provided by MTI Corporation Ltd., were irradiated in the Ion Beam Materials Laboratory (IBML UT-ORNL) at the University of Tennessee, Knoxville [31]. The irradiation setup along with the temperature control and the spectroscopic characterization have been described previously [21][22][23]32,33]. The relevant irradiation parameters, including electronic excitation densities and total number of oxygen vacancies generated per incident ion [34], are summarized in Table 1 (also Supplemental Material). Note that the electron-hole densities induced in this work are comparable to those induced in pulsed laser experiments.

Results and discussion
The 2.8 eV band is clearly enhanced relative to the 2.5 eV band with increasing excitation rate, a key finding of this work, when the emission spectra corresponding to irradiation with 8 MeV O and 18 MeV Cl ions are compared. Figure S2 (see supplementary data online) illustrates the representative emission spectra with several incident ions at low fluence ( ∼ 10 11 cm −2 ) and different temperatures, together with the decomposition into three Gaussian bands, as previously described [21,22]. The above results appear similar to those obtained in SrTiO 3 under pulsed laser irradiation using different pulse energies [12], where it was found that the integrated light emission intensity grows linearly with excited carrier density up to around 5 × 10 19 cm −3 (excitation density below 1 mJ/cm 2 ), where it starts reaching saturation.
The kinetics of the 2.8 eV band as a function of irradiation fluence offers new insights into the origin of this emission. Figure 1 illustrates the evolution of the 2.8 and 2.5 eV emission band yields under irradiation with several incident ions at different temperatures. The emission yield of the 2.8 eV blue band (Y B ), as well as that for the 2.5 eV green band (Y G ), rapidly increases with fluence (even though difficult to appreciate and evaluate in the figures) and reaches an essentially constant (steady-state) level after a fluence of ≈ 10 11 cm −2 , typically associated with the establishment of a steady-state concentration of electron-holes (e-h) pairs. The initial evolution of such bands is considerably faster than that for the 2.0 eV band (not shown) confirming previously reported results [21][22][23] that associated this band to the generation of isolated vacancies during irradiation. The corresponding evolution of the emission yield for the green band is also shown in Figure 1 for comparison purposes. The similar evolution for both bands suggests a close correlation between them. In order to explore more quantitatively this correlation, Figure 2 plots the ratio, ρ B/G = Y B /Y G , between the integrated peak areas of the blue and green luminescence bands as a function of ion fluence at various temperatures. The ratio remains relatively constant as a function of fluence, with a small dispersion in the data of about 1% for lighter ions and 5-10% for the case of heavier ions and lower temperatures. These deviations can be expected due to heavy band overlap and the rapid evolution of the yields with fluence, especially at low temperatures. This figure confirms that ρ B/G is independent of fluence at these temperatures. Moreover, this  ratio ρ B/G clearly increases with the excitation rate, i.e. the 2.8 eV band is enhanced relative to the 2.5 eV band as the electronic excitation density increases, corroborating the case of laser pulse excitation [12][13][14]. Based on these novel results, we attribute the 2.8 eV band to optical transitions between CB levels and the localized STE ground state level. The CB levels become densely populated during the strong electronic excitation provided by heavy-mass and high energy ion-beam irradiation, which may account for a significant emission at 2.8 eV. This model states that both the blue and green bands are ascribed to transitions of the STE center, from either unbound (2.8 eV emission) or bound (2.5 eV emission) excited states, in accordance with the electronic levels scheme for all the optical transitions considered in Figure 3. The coexistence between these two types of transitions can only be achieved under high electronic excitation rates, such as those achieved by pulsed-laser or ion-beam irradiation. In accordance with this new model, the position of the localized excited STE level with respect to the CBM should match the difference between the energies of the blue (E B ) and green (E G ) emissions (leaving aside lattice relaxations), i.e. around 0.3-0.35 eV. On the other hand, the position of the ground state above the maximum of the VB for the two (green and blue) transitions would also be around E B -E G ≈ 0.35 eV (E g ≈ 3.25 eV being the energy gap of SrTiO 3 ), suggesting a rather symmetrical position of the two STE levels inside the forbidden gap. Possible non-radiative Auger transitions that have been suggested by lifetime and integrated light intensity measurements for the 2.8 eV emission [12,17,18] are also indicated in Figure 3.
For a further quantitative interpretation of the kinetics, a definite model is needed. We assume that the main decay channel for the free carrier density is the formation of STEs through bimolecular e-h recombination followed by their associated green (2.5 eV) emission. Simultaneously, a monomolecular decay channel for free carriers to STEs levels accounts for the blue emission. We ignore any irradiation-induced defects, such as oxygen vacancies. This assumption is justified as long as the concentration of carriers, either electrons (e) or holes (h), is much higher than that for defects. This is expected for sufficiently high overall excitation rates G, around or above 10 20 cm −2 s −1 , comparable to those obtained under high power light pulses (see Table 1). This coexistence of bimolecular and monomolecular processes to account for the data on the emission lifetimes and yields as functions of excitation rate has been previously suggested by previous workers [13,14]. However, the strong overlap between the blue and green emission bands prevents quantitative separation of the two decay channels and the assignment of each to definite emissions (either green or blue). Previously reported data [13] on the photoluminescence spectra of pure and Nb-doped SrTiO 3 , which showed a clear enhancement of the blue emission over the green one, are consistent with our current ion-beam results and the present model. Due to the electron donor character of Nb, the doped samples contain a much higher free electron concentration and, therefore, the corresponding blue emission for these samples is strongly enhanced. For un-doped samples, a simple model, described below, predicts that the ratio, ρ B/G , between the two emission yields at 2.8 and 2.5 eV should evolve with the square-root of the excitation rate. For the ion fluxes considered in this work ( ∼ 10 11 cm −2 s −1 ), the ion trajectories and subsequent excitation processes are essentially uncoupled, temporally and spatially, so that one can treat the effects of various incident ions as independent events. Therefore, in order to meaningfully compare the role of the different ions and energies on the ratio ρ B/G , one should use the excitation rate corresponding to a single ion, i.e. g = N e−h (see Table 1). This assumption is also reasonable given the fast dynamics (the lifetime, τ , for the 2.8 eV emission is on the nanosecond timescale) of the optical processes. In order to simulate the kinetics, one may proceed as follows: the evolution of the total carrier (e, h) and STE populations per incident ion in the irradiated volume, ignoring spatial variations along the ion projected range (R p ), can be described by the following rate equations: with α being the recombination rate of carriers into STEs, β the light emission probability per unit of time Within this simple scheme, the overall yields for the green (Y G ) and blue (Y B ) emissions will be (ignoring possible self-absorption of the light): Y B ∝ N 0,e N STE ∝ g 1 2 g (monomolecular channel) (5) and the ratio between the yields is finally obtained as, Thus, at constant temperature the ratio should evolve with the excitation rate following a square-root dependence, as g 1 2 . Figure 4 shows that the yields ratio ρ B/G reasonably fits a square-root dependence on the singleion excitation rate, N e−h , which is consistent with the prediction of the above kinetic analysis. Therefore, the data confirms the physical model proposed for the STE and supports the assignment of the 2.8 eV emission to free carrier-STE recombination. One should remark that the situation under ion beam irradiation is not strictly equivalent to that under laser pulse excitation, since the individual photons are essentially coupled and cannot be separated in space or time.

Conclusion
In summary, a key result of the present study is that, the blue (2.8 eV) and green (2.5 eV) luminescence emissions are both related to transitions of the STE center, involving either localized states (2.5 eV), or pairs of localized and band states (2.8 eV). This represents a novel feature associated to the high electronic excitation densities achieved by both pulsed laser and ion beam irradiations. A new model for the blue emission is proposed consisting of a radiative transition from un-bound (CB) states to the ground STE level. Consequently, the two emissions (green and blue) are associated with the same localized center and imply the STE annihilation. The analysis presented here provides a unified description of the emission mechanisms and demonstrates the unique potential of ionoluminescence to unravel the rich and complex variety of effects associated with electronic carriers and oxygen vacancies in SrTiO 3 .