1. Introduction

The initial reports on quantum cutting (QC) in Pr³⁺-doped fluorides not only experimentally demonstrated the Dexter theory of ‘photon splitting’ in single activator [1,2], but opened up the study on efficient luminescent materials with quantum yield (QY) greater than unit. This interesting process, involving the absorption of a vacuum ultraviolet (VUV) photon and the sequential emission of two visible photons, can be potentially applied in lighting and display [3,4]. Based on the insight on visible QC mechanisms of Pr³⁺, another activator like Mn²⁺ [5], Eu³⁺ [6], Cr³⁺ [7], Er³⁺ [8] is further codoped to resonantly convert the near-UV photon emission of Pr³⁺ (first-step radiation ~400 nm, invisible to the human eye) to the green/red photon emission, where an appreciable QY was evaluated to be 143% for Pr³⁺/Mn²⁺ couple [5]. As an exciting scenario, efficient visible QC of Gd³⁺/Eu³⁺ couple was realized via two Eu³⁺ activators resonantly absorbing the two part energies of VUV-excited Gd³⁺ separated by ³P_j intermediate level [9,10]. This Gd³⁺/Eu³⁺ visible QC with QY ~190% has been strongly promoting the development of QC in experiment (various reports in other systems) and in theory (resonant energy transfer (ET) mechanism, sensitization, etc.) [4,9,11].

Recently the QC study was extended to near-infrared (NIR) emission region, which, with QY more than 100%, potentially enhance the photo-response of solar cells by cutting an incident UV-blue photon into two NIR photons [11–14]. Within the series of RE³⁺/Yb³⁺ (RE = Tb, Tm, Pr, Er, Ho, Nd) codoping systems [15–25], only the Pr³⁺/Yb³⁺ species, at room temperature (RT), yields efficient NIR-QC at low Yb³⁺ content (QY ~200%) via resonant ET from Pr³⁺ to Yb³⁺ [19,23]. Additionally inspired by the ET mechanisms in Pr³⁺/Mn²⁺ [5], Gd³⁺/Eu³⁺ [9,10], etc., it is concluded that the occurrence of resonant QC essentially requires the donor ions feature effective intermediate levels like Pr³⁺: ¹D₂ and Gd³⁺: ⁶P_j for visible QC, and Pr³⁺: ¹G₄ (lying at ~10000 cm⁻¹) for NIR-QC in Pr³⁺/Yb³⁺. However, in stark contrast, Er³⁺ with ⁴I_11/2 ~10000 cm⁻¹ intermediate level still cannot make NIR-QC efficient in Er³⁺/Yb³⁺ due to the nonradiative quenching of Er³⁺: ⁴I_11/2 at RT, especially in hosts with maximum phonon energy (ħω_max) more than 180 cm⁻¹ (ħω_max of bromides) [20,26]. In practice, the stable intermediate levels of donors, closely dependent on the vibration frequency of host lattice, determine if the resonant NIR-QC occurs efficiently as codoped with acceptors like Yb³⁺. Thus the Pr³⁺, as a resonant donor, should feature unique spectroscopic properties in the NIR region around 1000 nm (feasibly resonant to Yb³⁺ absorption), and the resonant ET mechanisms of NIR-QC should not be simply declared for Pr³⁺/Yb³⁺ systems by ignoring the influence of phonon energy. Despite the importance of these issues, the multiplexed NIR emission bands around 1000 nm and ET mechanisms of Pr³⁺ are not well understood indeed.

In the current paper, we report on the observation of several NIR emission bands around 1000 nm in Pr³⁺-doped oxyfluoride glass-ceramics (GCs) containing CaF₂ nanocrystals. The involving emission and ET mechanisms, as well as the two-step NIR-QC (splitting an absorbed blue photon into two NIR photons with QY greater than unit), are investigated in details according to the photoluminescence and time-resolved spectra. These Pr³⁺-doped GCs are the resonant NIR-QC candidates once Yb³⁺ codoped into, thereby potential to be the downconverting layer to increase Si solar cell efficiency.

2. Experimental

Pr³⁺-doped oxyfluoride glasses were prepared by melting the well-mixed batches of raw materials with nominal molar compositions of 45SiO₂-25Al₂O₃-10Na₂O-20CaF₂-xPrF₃ (denoted as PGPr_x; x = 0.05, 0.1, 0.5, 1), and the GCs, denoted as GCPr_x relative to PGPr_x, are obtained via a further heat treatment (see all the details in [27]). Microstructures of the samples were identified by means of X-ray powder diffractometer (Philips PW1830, Cu Kα) and transmission electron microscope (JEM-2010) fitted with the selected-area electron diffraction. Absorption spectra were recorded in 300-2400 nm on a Perkin-Elmer Lambda 900 UV/VIS/NIR spectrometer.

Emission and excitation spectra were measured using Edinburgh FLS920 fluorescence spectrophotometer with a 450 W xenon lamp as excitation resource and by single photon counting technology. Decay curves and time-resolved emission spectra are detected with FL920 system equipped with a microsecond μF900 xenon lamp as the pulsed excitation resource and by time-correlated single photon counting technology. Herein, a Hamamatsu R928 red-sensitive photomultiplier tube (PMT) was applied for visible emission measurements in the wavelength range of 400-780 nm, and a liquid nitrogen-cooled R5509-72 PMT was employed for NIR fluorescence measurements in 770-1150 nm.

3. Results and discussion

The characterization and analysis of the fabricated PGPr_x and GCPr_x were systematically performed by Chen et al. in our group [27], which indicated that the CaF₂:Pr³⁺nanocrystals were well crystallized from the glass matrix. Moreover, the GCPr_0.1 is demonstrated to have the optimal fluorescence properties under excitation of blue photons in 430-480 nm [27], which is just utilized for our current research.

Figure 1 shows the visible-NIR emission spectra of GCPr_0.1 under excitation of 440 and 585 nm, respectively. As shown in Fig. 1(a), a series of visible emissions of Pr³⁺ are typically obtained with peaks at 482 nm (³P₀→³H₄), 524 nm (³P_1,0→³H₅), 605 nm (¹D₂→³H₄/³P₀→³H₆),642 nm (³P₀→³F₂), and 725 nm (³P₀→³F_3,4) [27–34], respectively. Interestingly, within the 800-1100 nm wavelength range, there also exist several NIR emission bands that overlap with each other [Fig. 1(a)]. Reviewing some previous reports, the broad emission peak at ~873 nm with a distinct shoulder was always ascribed to the ³P_1,0→¹G₄ transition of Pr³⁺ only [27–30], and the other broad emission band at ~1040 nm is simply taken as one single emission from ¹D₂→³F_3,4 or from ¹G₄→³H₄ of Pr³⁺ [29–31,34]. Indeed the energy differences of ³P_1,0→¹G₄ (10928~11522 cm⁻¹), ¹D₂→³F_3,4 (9817~10291 cm⁻¹), ¹D₂→³F₂ (~11636 cm⁻¹), and ¹G₄→³H₄ (~9997 cm⁻¹) of Pr³⁺ are almost equivalent such that their corresponding NIR radiations are well overlapped and hardly distinguished from each other [27–32]. However, as an important result in the current paper, the multiple NIR fluorescent compositions of Pr³⁺ in 800-1100 nm should comprise the electronic transitions of ¹D₂→³F₂ peaked at 873 nm, ³P₀→¹G₄ at 915 nm, ¹G₄→³H₄ at 990 nm and ¹D₂→³F_3,4 at 1040 nm, respectively. Notably, the overlapped emission bands in 934-1150 nm (curve a) can be feasibly fitted by a sum of two Gauss functions, as plotted in Fig. 1(a). The fitted curve (curve 0) is well consistent with the experimental curve a (Goodness of fit value, R², ~0.9979), and two broad fluorescence fitting compositions emerge with maxima at about 990 nm (peak 1, curve 1) with full width at half maximum (FWHM) about 62 nm and that at 1040 nm (peak 2, curve 2) with FWHM about 55 nm.

 

Fig. 1 Visible-to-NIR emission spectra of GCPr_0.1 under excitation of (a) 440 nm and (b) 585 nm, respectively. In Fig. 1(a), curve a represents NIR emission band centered at ~1040 nm in 934-1150 nm, curve 0 is Gauss fit of curve a, and curves 1 and 2 are Gauss fit peaks 1 and 2, respectively. In Fig. 1(b), curve a′ refers to NIR emission at ~1040 nm in 927-1150 nm, curve 0′ is Gauss fit of curve a′, and curves 1′ and 2′ are Gauss fit peaks 1′ and 2′, respectively. In comparison with visible emissions, the NIR emissions of Pr³⁺are extremely feeble as measured under the same conditions, which shown in Fig. 1(a) was recorded by enlarging bandwidth of monochromator slits for clarity.

Download Full Size | PDF

To further explore the NIR fluorescence of Pr³⁺, spectrum of GCPr_0.1 is relatively detected upon 585 nm excitation [Fig. 1(b)]. As expected, all the NIR emission bands are still observed except the distinct emission shoulder at 915 nm (³P₀→¹G₄). The FWHM of 873 nm broad emission band is estimated to be about 55 nm. Besides, the 990 nm emission band becomes a non-negligible shoulder superposed on the dominant 1040 nm emission band, as shown in Fig. 1(b). A sum of two Gauss functions is also used to fit the NIR emission band in 927-1150 nm (curve a′), revealing that the fitted curve (curve 0′) is well in line with the experimental curve a′ (R² ~0.9986), and the two broad fluorescence compositions are gained with peaks at ~990 nm (peak 1′, curve 1′) and at ~1040 nm (peak 2′, curve 2′), respectively. These results well agree with the deductions in Fig. 1(a) and would be spectroscopically validated by the following time-resolved fluorescence.

Energy level diagram of Pr³⁺ in Fig. 2 schematically illustrates the proposed NIR emission and ET mechanisms in GCPr. Following fast nanoradiative relaxation (NR) from the excited ³P₂ state, the energy of populated ³P₀ states not only yields various visible photons to excite ³F_j and ³H_j states, but also radiates NIR photon at about 915 nm via the ³P₀→¹G₄ transition. Due to lower phonon energy environment around Pr³⁺ in CaF₂ nanocrystals (325 cm⁻¹) [27], the below ¹D₂ state would be populated mainly by cross-relaxation (CR) between ³P_1,0→¹D₂ (~4600 cm⁻¹) and ³H₄→³H₆ (~4200 cm⁻¹), CR1 [34–36]. Then the energy in excited ¹D₂ state would effectively de-excite to the lower ³F₂ and ³F_3,4 via photon emitting at about 873 and 1040 nm, respectively. Meanwhile, once Pr³⁺: ¹G₄ state effectively populated by the ³P₀→¹G₄ or by a feasible CR of ¹D₂→¹G₄ (~6840 cm⁻¹) + ³H₄→³F₃ (~6600 cm⁻¹), CR2 [27,34–36], it can be de-excited to the ³H₄ ground state by generating NIR photon around 990 nm. Here a two-step NIR-QC occurs from Pr³⁺: ³P₀ excited state with ¹G₄ intermediate level: the first-step ³P₀→¹G₄ at ~915 nm (step 1) and the second-step ¹G₄→³H₄ at ~990 nm (step 2) [1,37]. In sharp contrast, as GCPr excited at 585 nm (Fig. 2), only the ¹D₂→³F₂, ¹D₂→³F_3,4 and ¹G₄→³H₄ transitions are induced following the population of ¹D₂ and ¹G₄ states, respectively.

 

Fig. 2 Simplified energy-level diagram of Pr³⁺ illustrating the visible-to-NIR emission mechanisms, significantly undergoing the process of a sequential two-step NIR-QC: 1) the first-step ³P₀→¹G₄ at 915 nm, and 2) the second-step ¹G₄→³H₄ at 990 nm. Solid arrows represent optical transitions and short-dotted arrows are NR processes.

Download Full Size | PDF

To study the ET dynamics of Pr³⁺, time-resolved emission spectra of GCPr_0.1 are measured in 770-1150 nm under pulsed light excitation of 440 and 580 nm, respectively. As shown in Fig. 3(a), only one NIR emission band around 915 nm emerges initially (~7.5 μs), directly visualizing that the ³P₀→¹G₄ transition of Pr³⁺ does occur preferentially once the ³P₀ state nonradiatively populated by the energy in excited ³P₂ state. With delay time increase, the 915 nm emission intensity increases rapidly but with the band broadening asymmetrically, which is caused by the weak contribution of another NIR fluorescence band around 900 nm. Meanwhile, another emission peak at ~1040 nm begins to appear weakly at delay time of 8.0 μs, and its intensity rises markedly as delay time increases [Fig. 3(a)]. At delay time ~10.5 μs, the 915 nm emission reaches its maximum but with a distinguishable shoulder at about 873 nm, and also, the NIR emission band around 1040 nm intensely presents with a continuous increment trend [Fig. 3(a)]. Furthermore, as delay time increases from 19.5 μs, the fluorescence shoulder at ~873 nm becomes more obvious, and its intensity rise strongly relative to the 915 nm emission intensity. These observations suggest that, with the gradual growth of delay time, the lower ¹D₂ state would be populated nonradiatively from its above excited states, and does yield 873 nm and 1040 nm photons mainly via the ¹D₂→³F₂ and ¹D₂→³F_3,4 of Pr³⁺, respectively. Till delay time ~25.5 μs, the 873 nm emission band becomesdominant only with a quite weak fluorescence shoulder at ~915 nm, and finally at longer delay time ~57.5 μs, only the 873 nm emission can be recorded. On the other hand, an extra emission shoulder at ~990 nm presents at delay time around 19.5 μs, and, by increasing delay time, it exhibits more obvious increment tendency even if it is completely superposed with the intense 1040 nm emission band [Fig. 3(a)]. These results reveal that the later appearance of 990 nm emission does originate from the ¹G₄ state excited effectively by the first-step ³P₀→¹G₄ or NR from the above ³P_j and ¹D₂ states of Pr³⁺, as schematically depicted in Fig. 2.

 

Fig. 3 Time-resolved emission spectra of GCPr_0.1 upon (a) 440 and (b) 585 nm pulsed light excitation, respectively. In the inset of Fig. 3(b), curve a′′ indicates the corresponding emission band in 920-1150 nm recorded at 19.5 μs delay time, curve 0′′ is Gauss fit of curve a′′, and curves 1′′ and 2′′ are Gauss fit peaks 1′′ and 2′′ for curve a′′, respectively.

Download Full Size | PDF

Besides, as comparatively shown in Fig. 3(b), there does not present the 915 nm emission and any other obvious NIR emissions at the initial delay time ~7.5 μs upon 585 nm pulsed light excitation, which, absolutely distinct from that in Fig. 3(a), further validates that the emission peak at ~915 nm is resulted from the ³P₀→¹G₄ transition only. At longer delay time of 8.0 μs, the two NIR luminescence bands centered at 873 and 1040 nm emerge clearly, and then their intensities both raise with an almost identical trend as delay time increases, as plotted in Fig. 3(b). These phenomena confirm that the NIR emissions at 873 nm and 1040 nm do originate from the same excited ¹D₂ state, easily dedicated to the ¹D₂→³F₂ and ¹D₂→³F_3,4 transitions of Pr³⁺, respectively. Well consistent with that shown in Fig. 3(a), the emission shoulder around 990 nm becomes more obvious in Fig. 3(b) as decay time increases. Furthermore, the broad NIR emission band in 920-1150 nm (curve a′′) recorded at delay time ~19.5 μs is fitted by a sum of two Gauss functions: as exhibited in the inset of Fig. 3(b), the fitted curve (curve 0′′) well matches with the experimental curve a′′ (R² ~0.9926), and two broad peaks are derived with maxima at about 990 nm (peak 1′′, curve 1′′) and 1040 nm (peak 2′′, curve 2′′), respectively. These results demonstrate that the NIR emission at ~990 nm just comes from the lower ¹G₄→³H₄ transition of Pr³⁺, completely overlapping with the broad 1040 nm emission band as an indistinguishable shoulder [Figs. 1(a-b)].

Figure 4 shows fluorescence decay curves of 482, 873, 915, 990 and 1040 nm of GCPr_0.1 under pulsed light excitation of 440 nm. Interestingly, the decay curves of 482 and 915 nm basically have an identical single-exponential behavior with shorter lifetime, those of 990 and 1040 nm generally display single-exponential features with longer lifetime, whereas that of 873 nm obviously shows nonexponential characteristic. The fitting function, calculated decay time, R-squared (R²), FWHM, and the corresponding electronic transitions of Pr³⁺ are summarized in Table 1. It can be found that: i) the calculated lifetime of 482 nm (~14.8 μs) is almost equivalent to that of 915 nm (~13.5 μs) [33,38], which further confirms that, same to the 482 nm characteristic emission from the ³P₀ state, the 915 nm emission originates from the excited ³P₀ state via the ³P₀→¹G₄ [35]; ii) the lifetime of 1040 nm emission is typically estimated to be ~134.1 μs [33] and that of 873 nm have a similar long decay component of 129.1 μs (notably the short 29.6 μs decay component should be attributed to the overlapped emission like about 915 nm from Pr³⁺:³P₀ state as Pr³⁺activators occupy the E′, O, P, Q sites in the CaF₂ cluster structures [38]), suggesting that the broad emission bands centered at 873nm and that at 1040 nm dominantly result from the ¹D₂→³F₂ and ¹D₂→³F_3,4, respectively [29–31,35]; iii) the fluorescence shoulder located at about 990 nm has a distinct lifetime of about 107.8 μs, which is contributed from the ¹G₄→³H₄ of Pr³⁺ [36] (noted that the 990 nm decay curve performs the slight deviation of a mono-exponential decline likely due to the partial contribution of broad 1040 nm emission). Additionally, there present comparably long temporal evolutions within both 990 and 1040 nm decay dynamics, which confirms the ET dynamics from the excited ³P₀ state to the lower-lying ¹D₂ and ¹G₄ states of Pr³⁺, respectively [33]. By comparing the radiative lifetimes of ³P₀ state of Pr³⁺ (50 μs) calculated by Judd-Ofelt parameters [27], the energy efficiency of ³P₀ state is obtained to be about 30% in GCPr_0.1, which is much low readily because of the efficient CR1 between Pr³⁺ ions, multiphonon relaxation caused in the partial Pr³⁺ undoped into CaF₂ host lattice, poor crystallinity of CaF₂:Pr³⁺ nanocrystals, defects and/or impurities around Pr³⁺ ions. Correspondingly, the numerous optimization works should be further done during this GC fabrication for its potential downconverting layer to enhance the photo-response of solar cell.

 

Fig. 4 Luminescence decay curves of GCPr_0.1 monitoring at 482, 873, 915, 990 and 1040 nm under excitation of 440 nm pulsed light. The dot lines represent the experimental data, and the solid lines are the fitting results.

Download Full Size | PDF

Table 1. Decay curves of emission peaks, fitting function, R-squared (R²), calculated decay time, electronic transitions and FWHM of Pr³⁺ in GCPr_0.1 upon 440 nm pulsed light excitation

View Table

Herein the distinct ET dynamics of Pr³⁺ in the GC sample are analyzed to comprise the sequential two-step NIR-QC of Pr³⁺ with ¹G₄ intermediate level, i.e., the ³P₀→¹G₄ step 1 and the ¹G₄→³H₄ step 2. Generally, QY can be determined by the ratio of number of photons re-emitted over the number of photons that are being absorbed, just involving visible regime (${\eta }_{Vis}$) and NIR regime (${\eta }_{NIR}$) for NIR-QC system [4,15]. Using Judd-Ofelt parameters (Ω₂ = 2.10 × 10⁻²⁰ cm², Ω₄ = 2.88 × 10⁻²⁰ cm², Ω₆ = 2.01 × 10⁻²⁰ cm²) calculated from the absorption spectrum of GCPr_0.1 (see details in Ref [27].), the internal QY of GCPr_0.1,${\eta }_{{\mathrm{Pr}}^{3+}}$, can be theoretically evaluated in terms of luminescence branching ratios (β) [2,23,37],

4. Conclusion

In this paper, Pr³⁺-doped transparent GCs containing CaF₂ nanocrystals are prepared by the melting-and-quenching technique. The well overlapped NIR fluorescence bands of Pr³⁺ in 800-1100 nm are distinguished by photo emission and time-resolved spectra. Two-step sequential two-photon NIR-QC takes place feasibly with ¹G₄ acting as an intermediate level in the GCPrx samples via the ³P₀→¹G₄ transition at about 915 nm and the sequential ¹G₄→³H₄ transition at about 990 nm. The total internal QY of the two-photon NIR-QC reaches approximately 104% in this Pr³⁺ singly doped GC sample. Undoubtedly, the two-step NIR-QC transitions of Pr³⁺ would become the resonant ET routes once Yb³⁺ codoped into the system. It is believed that the new insight into ET dynamics and NIR luminescence mechanisms of Pr³⁺ would provide the rational principle and feasible approach to design and obtain the high efficient NIR-QC materials co-activated by Pr³⁺/Yb³⁺ couple.