Up to now, the X- and γ-ray detection performance of undoped and Tl-doped Cs3Cu2I5 single crystals were reported based on small size samples, from micrometer to millimeter scale.14, 26, 27, 31 However, for practical radiation detection applications, it is critically important to grow large volume single crystals with high optical quality by the melting method, such as the Bridgman method. Based on the differential scanning calorimetry, both of Cs3Cu2I5 and Cs3Cu2I5:Tl show severe thermal supercooling and are slightly incongruent (Supplementary Fig. S2), which causes the melt spontaneous nucleation easily into multiple grains and inclusions. Then, temperature dependent crystal cell parameters were calculated by in-situ synchrotron powder X-ray diffraction Rietveld (Supplementary Fig. S3 and Table. S1-2). A weak thermal expansion anisotropy was characterized. During the cooling process after crystal growth completion, difference in lattice thermal expansion along crystallographic axes will cause the accumulation of thermal stress in the crystal ingots which results in crystal cracking. To prevent the thermal and constitutional supercooling effect, a long ampoule capillary and large thermal gradient were introduced for the optimization of crystal growth in Bridgman method. Large the inclusion- and crack-free Tl-doped Cs3Cu2I5 single crystal ingot with diameter of 7 mm and length over 30 mm were produced (Fig. 2a). The measured transmittance of Cs3Cu2I5 after cutting and polishing at the maximum emission wavelength (445 nm) is 77.8%, close to the theoretical value of 82.4% (Supplementary Fig. S4-5). For Cs3Cu2I5:Tl, there is even smaller difference between the theoretical (81.9%) and measured values (78.6%) at 505 nm.
To elucidate the scintillation emission mechanism of Cs3Cu2I5:Tl, the neutron powder diffraction (NPD) technique was used to illustrate the crystal structure configuration, in particular the site occupation of Tl dopants. Figure 2b shows the NPD pattern collected on energy-resolved neutron imaging instrument (ERNI) at room temperature together with the fitting results. By Rietveld refinement on the diffraction data, detailed structural information could be obtained, such as lattice parameters, atomic positions, especially the position of Tl. The neutron diffraction Rietveld refinement fitting plots can be appreciated in Fig. 2a, and the final refined structural parameters are summarized in Table S3. The Rietveld analysis shows slightly decreased lattice parameters, a = 10.1693(3) Å, b = 11.6495(3) Å, c = 14.3572(5) Å, of Cs3Cu2I5:Tl compared with the parent one Cs3Cu2I5.32 Given that the ionic radii of Tl+ is smaller than that of Cs+, (1.59 Å of Tl+ and 1.74 Å of Cs+)33, the lattice contraction confirms the successful entrance of Tl into the lattice. To further determine the position of Tl+, we tried different refinement models with Tl replacement for Cs, Cu and I. The refinements show that Tl enters the 8d site of the lattice, i.e. the Cs2 position in Table S4, which is indicated by a slightly improved fitting and decreased χ2. Fitting with other models yields either negative occ values or positive values with considerable errors.
We continue to study the photo-physics in Tl-doped Cs3Cu2I5. Based on previous results, there are two kinds of emission observed at room temperature: (i) STE emission peaking at 445 nm (2.79 eV) associated with spatially isolated [Cu2I5]3− clusters, and (ii) Tl-related emission peaking at 510 nm (2.43 eV).14, 27 The contribution of Tl-related luminescence increased with the Tl-doping concentration, but the emission or excitation wavelengths have no change. A slight acceleration of STE emission and a slight deceleration of Tl-related emission can be observed, which is due to the signal superposition of the two broad emissions (Supplementary Fig. S6). 34 Here, as shown in Fig. 2d, two other emissions peaking at 415 nm (2.99 eV) and 550 nm (2.25 eV) were observed at low temperature. To understand the origin of the observed emission centers, their temperature dependence of photoluminescence spectra and decay kinetics were systematically studied. All four observed emission bands can be ascribed to three luminescence centers (see Fig. 2d):
Emission I: Tl + ions: Photoluminescence (PL) and excitation (PLE) spectra of emission Ⅰ are shown in Fig. 2e. Upon monitoring the emission I at 415 nm, two broad excitation bands are observed at 8 K. The weak excitation band range from 200 to 280 nm is associated with the emission II. The main excitation band range from 280 to 350 nm consists of four peaks at 298, 310, 324, and 334 nm, related to the emission I. Under 300 nm and 329 nm excitation at 8 K, the dominant emission I peaks at 415 nm with a full-width at half-maximum (FWHM) of 0.15 eV. The Stokes shift is 0.72 eV. When the measurement temperature rises to 20 K, emission I is significantly reduced, and fully quenched at 40 K (Supplementary Fig. S7). The features of emission I, namely, its PLE and PL spectral position, Stokes shift, and FWHM of the emission peak are very similar to the Tl+ emission at 400 nm in CsI host.35, so that we ascribe the emission I to the Tl+ center. However, it is quenched at much lower temperatures (in CsI host the onset of 400 nm band quenching is above 160 K) and its decay, described by two-exponential fit with 0.8 and 1.65 µs decay times at 8 K, see Fig. S8 in Supplementary, is much different from that of 400 nm emission in CsI:Tl which shows about 10 ns and 1 ms decay times in this temperature range. It is most probably due to different local symmetry of Tl+ ion in Cs3Cu2I5 host and possible interaction with energy levels of Cu+ situated at the bottom edge of conduction band.
Emission II: Self-trapped excitons from host lattice: The normalized PL and PLE spectra of emission II are shown in Fig. 2f. Upon monitoring the emission wavelength of 445 nm, the excitation spectrum at 77 K consists of a peak at 292 nm and a shoulder at 270 nm. There is also an almost invisible weak excitation peak at 330 nm. A redshift of these excitation peaks can be observed when increasing temperature to 300 K. In contrast, with the increase of temperature from 77 to 490 K, the peak position of emission II is nearly the same. The emission II has a broad FWHM of 0.29 eV, and a large Stokes shift of 1.56 eV at 77 K. These are typical features of the STE emission.34 The formation of STEs needs intense electron-phonon coupling, which is characterized by the Huang-Rhys factor (S). In Supplementary Fig. S9, the value of S is estimated as 104 by using the Eq. 1. It indicates a strong electron-phonon coupling and supports the STE emission origin.36 The spectral characteristics of emission Ⅱ in the Tl-doped Cs3Cu2I5 are identical to STE emission in undoped Cs3Cu2I5 (Supplementary Fig. S10), which is strongly confined in individual [Cu2I5]3− clusters.32, 37 The measurement of temperature-dependent PL decay of STE emission was performed. The PL decay time as a function of temperature is shown in Fig. 2g. At 8 K, its decay is single exponential with decay time of 2.07 µs, see in Supplementary Fig. S11, and the onset of the thermal quenching occurs around 280 K. The inset in Fig. 2g shows the decay profile at room temperature as an example, in which the single exponential fitting was used. According to the Arrhenius equation (Eq. 2), the exciton binding energy (Ea) is calculated as 340 ± 12 meV. It is consistent with the value of 335 meV reported for the STE in undoped Cs3Cu2I5.38
Emission III: TI-bound excitons: In addition to Tl+ emission and STE emission, there are two other emissions peaking at 510 nm (2.44 eV) and 550 nm (2.26 eV) observed well-resolved at the low temperatures (20–80 K) in Cs3Cu2I5:Tl. These two emissions have very similar characteristics, including the emitted energy, FWHM, Stokes shift, and decay time at low temperatures, except for the thermal stability. They are assigned as emission III. To study the two emissions, the PL-PLE mapping at 77 K was measured and plotted in Supplementary Fig. S12. When monitoring the emission wavelength of 510 nm, the excitation spectrum at 77 K consists of two peaks at 325 and 310 nm. The Stokes shift is 1.36 eV. For the emission wavelength of 550 nm, it is more likely to be excited at 340 nm. The PLE and PL spectra of emission Ⅲ as a function of temperature are shown in Fig. 2h Similarly with the STEs emission, the excitation peaks of emission III show a redshift when the measurement temperature increases. Since emission III consists of two emission peaks superimposed, we identify the two emissions by bimodal Gaussian fitting of photon energy. Around 220 K, the emission peak at 550 nm is already difficult to identify in PL spectra, see in Supplementary Fig. S13. The S and phonon frequency (ℏω) are evaluated as 62 and 14.6 meV, respectively (Supplementary Fig. S14). At 8 K, the PL decay for the emission at 510 nm is 5277 ns, and that is 5095 ns for the emission at 550 nm (Supplementary Fig. S15). When the temperature increases to 77 K, the decay times of the two emissions are accelerated differently. The decay profiles at different emission wavelengths (from 490 to 570 nm) were measured when excited at 330 nm (Supplementary Fig. S16). All profiles are single-exponential (It is a sum of two similarly decaying components, but it cannot be identified by the double exponential function because of the similarity of the decay times and overlap of emissions.). The decay times are 2160 ns monitoring at 490 nm and 3 190 ns monitoring at 570 nm. This indicates that the emission at 510 nm accelerates about 50% faster than that at 550 nm in the 8-77K temperature range.
The temperature dependence of PL decay profiles of emission Ⅲ were measured when monitoring λem = 510 nm and λem = 570 nm from 77 to 490 K. The decay constants of emission III as a function of temperature are plotted in Fig. 3i. Similarly, to PL spectra, the signal associated with 550 nm emission is almost unrecognizable at 220 K. Based on Arrhenius equation (Eq. 2), the exciton binding energy is calculated of 56 meV, indicating a lower exciton thermal stability. For the emission at 510 nm, the value is 432 meV, and it remains stable at room temperature. Its decay time at room temperature is 706 ns. The significant difference between the two temperature ranges (The decay time acceleration of emission at 510 nm is faster in 8–77 K and slower in 77–220 K compared to that of emission at 550 nm) suggests that the two emission centers are not completely independent. The features of large Stokes shift, broad emission peak, high Huang-Rhys factor, and long decay time suggest the nature of trapped exciton emission for emission III. However, it is clear that the decay time evolution trend as a function of temperature of emission III is different from the host STE emission. It slowly decreases from 77 to 350 K, and then rapidly decreases at higher temperature. This situation closely resembles the temperature dependence decay kinetics of the ns2 ions (such as Tl+, Pd2+, and Bi3+) bound exciton emission.39, 40, 41, 42 At the lowest (8–20 K) temperatures, it clearly show thermally stimulated bidirectional transitions between both the adiabatic potential excited state (APES) minima (Fig. S15) which closely resembles the situation in CsI:Tl.35 Thus, the emission III can be ascribed to the two configurations of exciton trapped around an adjacent Tl ion (Tl-bound exciton), similar to the Tl-bound exciton emission observed in CsI:Tl.40, 43 Furthermore, based on the NPD refinement results mentioned above, the Cs site with eight-fold coordination is more favorable for Tl doping. The thermally stimulated transitions between the 510 nm and 550 nm occur at much lower temperatures compared to CsI:Tl, similarly to the 410 nm emission of Tl+, which points to significantly lower energy barriers between their APES minima.