Temperature-dependent excitation spectra monitored at 2.71eV and 2.17eV are shown in Fig. 4a and b, respectively. As described above, the excitation consists of 4f transitions of Gd3þ and CT transition of TaO3 4 group. For the high-energy band, the shape of the excitation spectra changes significantly with temperature. The excitation efficiency around 4eV decreases faster than that of 4.53eV. A broad band at about 3.5eV also appears at higher temperature, which may be caused by some defects. While for the low-energy band, the excitation spectra are rather similar to each other. Some impurity-related peaks located at 3.53eV and 3.39eV
can be observed at higher temperature, which are corresponding to the 4f transitions of Eu3þ impurities (emission spectra excited by photons around these peaks show the characteristics of Eu3þ emission). The appearance of these peaks is consistent with the situation observed in the temperature-dependent emission spectra. Fig. 5a and b shows the luminescence decay curves monitored at 2.88 eV and 2.16 eV at some selected temperature points, respectively. The decays of both bands show a two-exponential manner in the full temperature range. It can be written as follows: IðtÞ¼a1expðt=τ1Þþa2expðt=τ2ÞþC ð2Þ The fit results are summarized in Tables 1 and 2. The two- exponential behavior implies that the excited states experience additional nonradiative channel accompanying the radiative recom- bination [23]. For the high-energy band, both the fast component and slowcomponent decreasewith increaseof the temperature.The decay of the high-energy band becomes very faster above 150K, as shown in Fig. 5a, which agrees well with the temperature depen- dence of PL intensity. Abrupt increase of the lifetime at low temperature evidences the presence of a pair of closely spaced energy levels involved in the photoexcited dynamics [24,25]. For the low-energy band, the decays change gradually with increasing temperature, with the fast component changing slightly around 10nsandtheslowcomponent decreasingfrom661nsto124 ns.We calculate the average lifetime using the following equation [26]: τave ¼R tIðtÞdt R IðtÞdt ð2Þ where I(t) denotes the PL intensity at time t corrected for the background. The average lifetime of 2.7eV band at 20 K is calculated to be 14μs, much longer than that of the 2.2eV band with the average lifetime of 673 ns. Apparently, the decay rates of both bands are significantly different, which gives the further evidence that these two bands originate from different luminescent centers. The general trends of the decays can be outlined but the measured temperature-dependent lifetime may not be completely alike the real situation due to the slow response of our instrument (97ns). Oxygen vacancy is the most obvious defect we can think of for the reasons that the crystal is grown in anoxic atmosphere and oxygen vacancies in many oxides emit broad luminescence [23,27]. To confirm whether the luminescence is related to oxygen vacancy, we investigated the annealing effect on the luminescence of GTO crystal. The air annealing effect had been reported in Ref. [13] and the effectof H2 annealing wasfurtherstudiedinthis work,as shown in Fig. 6. The intensity of these samples can be classified in the following order: IairoIunannealedoIH2 However, the difference of these spectra looks very small. As mentioned above, the lumines- cence of GTO crystal is severely thermally quenched. Therefore, this cannot be treated as reliable evidence to confirm whether and how much oxygen vacancies contribute to the luminescence of GTO crystal. Further experiments need to be taken to clarify it.