Effect of the Pr³⁺ → Gd³⁺ energy transfer in multicomponent garnet single crystal scintillators

After systematic study of Lu₃Al₅O₁₂ (LuAG)-based single crystal scintillators, for the review see [1], a new material concept based on single crystal multicomponent (Gd,RE)₃(Ga,Al)₅O₁₂ host (RE = Lu, Y) was proposed. Doped by Ce³⁺, such a host lattice with balanced combination of the admixed Gd³⁺ and Ga cations, demonstrated ultrahigh light yield exceeding 50 000 phot MeV⁻¹ [2–5], this value is by 30–40% more than the value of the best LYSO : Ce scintillators reported so far. Development of electronic band structure, lanthanide 4f–5d and charge transfer energies with the material composition in the family of multicomponent garnets has also been reviewed recently [6]. Given the high density and effective atomic number, chemical stability and favourable combination of scintillator characteristics which can be composition-tailored to a particular application, this family of scintillators is currently considered in a number of applications in medical imaging, high-energy physics, hi-tech industrial devices, oil-well logging and radiation monitoring in the ambient. LuAG host appeared particularly flexible as for the doping to achieve efficient scintillation. In the bulk crystal form it has been doped not only by Ce³⁺, Pr³⁺ but also with Sc³⁺ [7, 8], while in the thin film crystalline form prepared by liquid phase epitaxy also the Bi³⁺ luminescent centre could be introduced [9]. The latter thin film form appears prospective for microimaging under x-rays [10] and synchrotron beam diagnostics [11].

The latest generation of Ce³⁺-doped scintillators based on aluminum garnets (LuAG : Ce) shows the high intrinsic scintillation efficiency and light yield up to 26 000 phot MeV⁻¹ [12, 13]. However, their performance is always negatively influenced by shallow electron traps delaying an energy transfer to the Ce³⁺ emission centres. As a result, the scintillation decay contains a considerable amount of slow components in submicrosecond range [14, 15]. Admixing Ga into the aluminum garnet matrix reduces the trapping effect [16]. This is because the antisite defect-related shallow electron traps fall into the bottom of the conduction band [17]. The Gd³⁺ admixture ensures a sufficient separation between the conduction band edge and 5d₁ excited state of Ce³⁺ and prevents the unwanted ionization of excited state of Ce³⁺ ion around room temperature (RT) [18–20]. Combined effect of the Gd³⁺ and Ga admixture into LuAG host thus results in substantial acceleration of scintillation response and the mentioned enormous light yield increase. In addition, total manufacturing cost of these so called multicomponent garnet crystals with general chemical formula (Gd,Lu,Y)₃(Ga,Al)₅O₁₂ is even lower compared to pure LuAG or (Lu,Y)₂SiO₅ (LYSO)-based scintillators. Despite more expensive raw materials (Ga₂O₃) there is considerable energy saving in the crystal growth process due to more than 200 degrees lower melting point of Gd₃Ga₃Al₂O₁₂ compared to LuAG and LYSO which also extends 2–3 times the lifetime of expensive iridium crucibles and hot-zone insulation ceramics.

Pr³⁺-doped LuAG single crystals were introduced in 2005 [21] and 2006 [22], prepared by micro-pulling-down and Czochralski techniques, respectively. Scintillation characteristics of LuAG : Pr were studied by several groups [23–30] and found very promising due to favourable combination of fast scintillation decay dominated by 20 ns component, light yield approaching 20 000 phot MeV⁻¹ and excellent energy resolution below 5% at 662 keV. Moreover, its scintillation parameters do not deteriorate significantly up to 300 °C [31] which makes it particularly suitable e.g. for geophysical explorations. However, in the case of the Pr³⁺-doped Gd³⁺-containing multicomponent garnets, a rather discouraging situation has been found. Although the optimized host compositions mentioned above provided the values of light yield above 50 000 phot MeV⁻¹ in the case of Ce³⁺ doping, the analogous Pr³⁺-doped ones show light yield values only up to 4000–4500 phot MeV⁻¹ [20, 32], i.e. more than ten times lower.

In this paper we investigate the processes influencing the 5d₁ excited state of Pr³⁺. In particular, we monitor the radiative de-excitation, energy transfer, thermal quenching and thermally induced ionization in the (Gd,RE)₃(Ga,Al)₅O₁₂ host by optical methods. Due to the flexibility, cost-effective character and speed of micro-pulling-down method [33] such a comparative study could be done for a number of single crystal samples which is preferable with respect to experiments usually done with powder samples.

All undoped and Pr³⁺-doped (Gd,RE)₃(Ga,Al)₅O₁₂, RE = Y, Lu, single crystal samples, see the list in table 1 and other compositions throughout the text, were prepared by micro-pulling-down method [3, 4, 33] in Japanese laboratories. Plates of 0.5–1 mm thickness were cut from parent rods and polished up to an optical grade.

Absorption spectra were measured by the Shimadzu 3101PC spectrometer in the 190–800 nm range. Radioluminescence (RL) and photoluminescence (PL) spectra were excited by x-ray (40 kV, 15 mA) tube (Seifert Gmbh) and D₂-lamp (Heraus, D200F), respectively. Detection within 200–750 nm was performed by the custom made 5000M model fluorometer Horiba Jobin Yvon equipped with photon counting detector TBX-04 (IBH Scotland). Fast PL decay of Pr³⁺ in the nanosecond time scale was excited by the hydrogen-filled flashlamp as well as the 281 nm nanoLED source and detected by the time-correlated single photon counting method. Slow decays in the ms time scale were measured under a microsecond xenon flashlamp excitation by the multichannel scaling method. The Oxford Instruments Optistat cryostat (www.oxford-instruments.com) as well as Janis closed cycle refrigerator (www.janis.com) were used to obtain temperature dependences of selected luminescence characteristics within 8–500 K. Time-resolved excitation spectra in submicrosecond time scale were measured at SUPERLUMI experimental station in VUV–UV spectral range with the Hamburg synchrotron radiation facility (HASYLAB at DESY, Notkestr. 85, 22607 Hamburg, Germany). All spectra were corrected for spectral distortions caused by experimental setups.

As was mentioned in the introductory part, the admixture of Gd³⁺ dramatically diminishes the intensity of 5d₁–4f emission of Pr³⁺ in RL spectra [20, 32]. In figure 1 the example of RL spectra of the undoped and Pr³⁺-doped multicomponent garnets are shown. The undoped sample spectrum is dominated by the Gd³⁺ line emission transitions from ⁶I_x and ⁶P_x at 279 nm and 314 nm, respectively, while that of Pr³⁺-doped Gd_0.5Lu_2.5Ga₂Al₃O₁₂ shows a typical broad band near UV emission from 5d₁ excited state to ³H_x and ³F_x 4f multiplets of Pr³⁺ with a tint of Gd³⁺ emission at 314 nm.

Zoom In Zoom Out Reset image size

By varying the Gd³⁺ concentration, we found the most intense 5d₁–4f emission of Pr³⁺ under x-ray excitation for the host composition of Gd_0.5Lu_2.5Ga₂Al₃O₁₂ and relatively low Pr³⁺ concentration within 0.1–0.2%. In figure 2, which includes the absorption spectrum of the undoped Gd₃Ga₂Al₃O₁₂ sample, we note the Gd³⁺ absorption transitions ⁸S_7/2 → ⁶P_x, x = 3/2, 5/2, 7/2 peaking at 302.7 nm, 307.9 nm and 313.7 nm, respectively. These absorption lines are considerably overlapped with the maximum of Pr³⁺ 5d₁–³H₄ emission in (Gd,Lu)₃Ga₂Al₃O₁₂ host: it is peaking at about 302 nm in Gd_0.5Lu_2.5Ga₂Al₃O₁₂, see figure 1, and is supposed to shift down to 318 nm in Gd₃Ga₂Al₃O₁₂ host taking into account the shift of Ce³⁺ emission maximum in analogous hosts (see figure 7 in [3]). Similar peak positions hold also for the yttrium-admixed hosts. Such an overlap can in principle enable a nonradiative energy transfer from 5d₁ level of Pr³⁺ to ⁶P_x 4f excited state levels of Gd³⁺, and would effectively deplete the former centre excited state, if the ions are sufficiently closely spaced. As is clearly evident in three PL spectra in figure 2, there is very little residual 5d₁–4f emission of Pr³⁺ under excitation into 4f–5d₁ absorption band of Pr³⁺ at 280 nm and emission intensity of Gd³⁺ line at about 313 nm strongly diminishes with Gd³⁺ concentration.

Zoom In Zoom Out Reset image size

In order to confirm the energy transfer from 5d₁ emitting level of Pr³⁺ to ⁶P_x 4f excited state level of Gd³⁺ we carried out an extended decay kinetics study for a variety of the samples. In undoped samples, the intensity and decay time of Gd³⁺ emission decreases by two orders of magnitude with increasing Gd³⁺ concentration, see figures 3(a) and (b), respectively. Such concentration dependence clearly points to concentration quenching in Gd³⁺-sublattice. It is further confirmed by the increasing efficiency of energy transfer from Gd³⁺ to accidental Tb³⁺ impurity (see figure 3(a)) due to (Gd-Gd ... -Gd)_n energy migration [34]. In Pr³⁺-doped samples, the nanosecond decay time of Pr³⁺ is clearly shortened with the increase of Gd³⁺ concentration (figure 4) and in the case of the highest Gd³⁺ content becomes hardly measurable at the existing experimental setup. It provides clear evidence for Pr³⁺ → Gd³⁺ energy transfer mentioned above.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

This effect is evident in all sets of the samples, as seen in table 1, where the decay data are grouped by values of Ga content that is kept constant, and Gd³⁺ content increases for every set. In several cases, the decay curves possessed non-exponential character and could be only fit by several exponential components. In such cases, the mean decay time τ_m was introduced for consideration, defined by the equation (table 1, values in bold):

$$\begin{equation} \tau_{\rm m} ={\sum {A_{i} \tau_{i}^{2}}}\big/{\sum {A_{i} \tau_{i}}} \end{equation} \tag{ 1 }$$

Noticeably shorter decay time of the dominant Pr³⁺ decay component (see table 1) with respect to the previously known value of about 20 ns at RT in LuAG [19] confirms an efficient depopulation of 5d₁ level of Pr³⁺ by an additional nonradiative pathway.

As mentioned above the energy transfer Pr³⁺ → Gd³⁺ is enabled due to the overlap of 5d₁–³H₄ emission transition of Pr³⁺ with the ⁸S_7/2 → ⁶P_x absorption lines of Gd³⁺. It is confirmed by the above shown decay kinetics features. With the increase of Gd³⁺ concentration the decay curve of Pr³⁺ accelerates and becomes more non-exponential, see figure 4. Such behaviour is typical of the multipole interaction-driven energy transfer between a donor and acceptor centres described by the Dexter model [35].

The energy transfer process could also be proved by the time-resolved excitation spectra. In the steady-state excitation spectrum the contributions from Gd³⁺ and Pr³⁺ ions could not be resolved because of the spectral overlap of Gd³⁺ 4f–4f and Pr³⁺ 5d–4f emissions around 310–315 nm. Therefore, the time-resolved excitation spectrum for Gd³⁺ emission was measured by means of the slow (micro-to-millisecond) decay curves integrals measured at the emission wavelengths in the region of interest with the step of several nanometers, see figure 5(a). The slow Gd³⁺ 314 nm decay is evidently excited also outside of the Gd³⁺ sharp absorption peaks related to ⁸S_7/2 → ⁶D_x, ⁶I_x and ⁶P_x transitions around 250 nm, 275 nm and 310 nm, respectively. Moreover, it is clearly excited within the 4f–5d_1,2 absorption bands of Pr³⁺ at about 280 and 240 nm, which is a direct confirmation of Pr³⁺ → Gd³⁺ energy transfer. The Gd³⁺ concentration was chosen rather low to avoid a Gd–Gd migration. Because of accelerated 5d–4f decay of Pr³⁺ in Gd³⁺-containing multicomponent garnet hosts, see table 1, it was possible to make such an experiment and measure analogous excitation spectra also in deeply submicrosecond time scale using different time windows at SUPERLUMI station at DESY synchrotron, Hamburg [36]. The result is displayed in figure 5(b). Monitoring the emission wavelength around 314 nm the fast emission window of 2–11 ns will dominantly collect the Pr³⁺ fast (prompt) emission, while the slow emission window of 34–64 ns will dominantly collect the slow emission of Gd³⁺. Indeed, also in the latter case, the 4f–5d_1,2 Pr³⁺ excitation bands are clearly visible around 280 nm and 240 nm, respectively, analogously to the situation in figure 5(a). Furthermore, we note an efficient and fast energy transfer from the host to Pr³⁺ ions, which is evidenced in figure 5(b), by sharply increasing signal below 210 nm where the band edge of the host is, see the absorption spectrum in the inset.

Zoom In Zoom Out Reset image size

In order to study a thermal stability of the Pr³⁺ 5d₁ excited state, we measured a temperature dependence of the nanosecond decay time in Lu₃Ga₂Al₃O₁₂ host, see figure 6(a). The onset of the decay time shortening takes place around RT, as in the Ce³⁺-doped samples [3, 16]. Consequently, we also measured the delayed recombination decay to resolve the influence of two possible processes, thermal quenching and/or thermal ionization of the Pr³⁺ 5d₁ excited state. Integrals of the delayed recombination decay curves, see [20, 23] for technical details of experiment, recorded at different temperatures directly show the temperature dependence of the intensity of delayed recombination decay process (figure 6(b)). Comparing the data in figures 6(a) and (b) one can see that the strongly increasing delayed recombination signal within 300–400 K points to a dominant influence of thermally induced ionization, as in the case of Ce³⁺-doped multicomponent garnets [20]. It has been shown in two other recent experimental studies that the admixture of gallium in the garnet composition significantly decreases the ionization barrier of 5d₁ excited state of Ce³⁺ [37, 38]. Thermal quenching possibly occurs at yet higher temperatures. Given the very onset of nanosecond decay time shortening at about 250 K, figure 6(a), the increase of the delayed recombination intensity from RT down to 200 K (figure 5(b)) can be due to the fact that thermally induced ionization process is modulated by the role of shallow traps present in the material [39]. Residual signal below 100 K is most probably a temperature-independent contribution [40] due to quantum effects between the excited luminescence centre and adjacent host lattice defect [41].

Zoom In Zoom Out Reset image size

The processes depleting the 5d₁ excited state of Pr³⁺, namely the radiative de-excitation, energy transfer, thermal quenching and thermally induced ionization were studied in a large set of multicomponent undoped and Pr³⁺-doped (Gd,RE)₃(Ga,Al)₅O₁₂, RE = Lu,Y, garnets with wide concentration range of Gd³⁺ and Ga cations. Deteriorated light yield values observed in the case of Pr³⁺-doped Gd³⁺-containing multicomponent garnets in earlier studies are due to resonance of the 5d₁–³H₄ emission transition of Pr³⁺ and ⁸S–⁶P_x absorption transitions of Gd³⁺ within 302–314 nm, which creates an additional nonradiative pathway from the 5d₁ state of Pr³⁺. Despite the efficient energy transfer from the host to the Pr³⁺ centre, this unwanted phenomenon, further intensified by the concentration quenching in the Gd³⁺-sublattice, strongly degrades scintillation performance. Composition engineering can, in principle, be a way to set the above mentioned transitions of Pr³⁺ and Gd³⁺ off resonance. As a result, an unwanted depletion of 5d₁ Pr³⁺ excited state would reduce and the light yield of the material would improve.

This research was supported by Czech project of MEYS, KONTAKT II, no LH12150 and EC under FP7 Programme-research with synchrotron radiation, project II-20100033. The support of the funding program for the next generation world-leading researchers, Japan society for promotion of science (JSPS) is also gratefully acknowledged. This work is also supported by JSPS, Grant-in-Aid for Exploratory Research (AY) and by the funding program for the Health Labour Sciences Research Grant, The Ministry of Health Labour and Welfare.