Identification of Dy 3+ = Dy 2+ as Electron Trap in Persistent Phosphors

Laser excitation and x-ray spectroscopy are combined to settle a long-standing question in persistent luminescence. A reversible electron transfer is demonstrated, controlled by light and showing the same kinetics as the persistent luminescence. Exposure to violet light induces charging by oxidation of the excited Eu 2 þ while Dy 3 þ is simultaneously reduced. Oppositely, detrapping of Dy 2 þ occurs at ambient temperature or by infrared illumination, yielding afterglow or optically stimulated luminescence, respectively.

Luminescent materials or phosphors are crucial for various technological applications, ranging from environmentally friendly lighting [1,2] to specialized applications in medicine and optoelectronics [3][4][5][6][7]. A subclass are those that feature energy storage capabilities, for which not all of the energy absorbed upon excitation is immediately emitted as photoluminescence. A fraction is used to transfer a charge carrier from the luminescent center to a so-called trap, generating a metastable state.
The lifetime of these metastable states can range over many orders of magnitude (quantified by the so-called trap depth). The long lifetimes of deep traps are exploited to date minerals or archeological artifacts, or for dosimeters, e.g., in medical imaging plates where red or infrared (IR) light is used to detrap charge carriers and induce the radiative decay of the luminescence activator, a process called optically stimulated luminescence (OSL) [8,9]. For luminescent materials with more shallow traps, ambient temperature can be sufficient to slowly deplete the traps (in minutes to hours), giving rise to the process of long afterglow or persistent luminescence [10][11][12][13]. This process can be accelerated by heating the phosphor, giving rise to thermoluminescence (TL), the analysis of which allows us to experimentally extract information on trap depths or trap depth distributions [14,15]. Afterglow materials hold the promise of new applications such as glow-in-the-dark traffic signalization [16], damage monitoring [17], radiation-free medical imaging [3,18], theranostics [19,20], cell characterization [21], or nanothermometry [22]. The storage capabilities need nevertheless to be improved in order to enable these advanced applications by the optimization of existing materials or the development of new ones [12,13].
In the 1990s it was found that some Eu 2þ based phosphors acquire intense persistent luminescence upon codoping with a trivalent lanthanide ion (Ln 3þ ), often Dy 3þ [23]. This triggered the discovery of a group of phosphors that can still be regarded as state of the art. Among them, the strontium aluminates (SrAl 2 O 4 ∶ Eu 2þ ; Dy 3þ and Sr 4 Al 14 O 25 ∶Eu 2þ ; Dy 3þ ) and the silicate Sr 2 MgSi 2 O 7 ∶Eu 2þ ; Dy 3þ take the most prominent places [10,12]. As the trapping capacity is significantly increased upon Dy 3þ addition, it is tempting to identify the trivalent codopant as a trap. To date, no hard evidence has however been found that confirms or rejects a valence state change for Dy 3þ . The reason for this standstill is the difficulty to detect such transient phenomenon in common structural and optical experiments, and the intrinsic complexity of these materials, making theoretical studies scarce and hard to correlate with experiments such as TL [24,25].
TL experiments have suggested that distributions of trap depths are present, rather than one discrete trap [15,26]. However, TL does not infer the chemical nature of the traps, and the distribution can be due to multiple intrinsic defects that induce trapping [27], but equally well to multiple nonequivalent Dy 3þ centers in the crystal [28] distributed at different distances from the Eu 2þ activator [29]. Detailed experimental studies of SrAl 2 O 4 ∶Eu 2þ ; Dy 3þ suggest that trapping is local, i.e., close to the Eu 2þ activator rather than a delocalization of the electron toward the conduction band [26]. A limited number of x-ray absorption spectroscopy studies have been reported on Eu 2þ , Ln 3þ codoped aluminates and silicates [30][31][32][33][34][35]. Most studies failed to find reversible valence state changes, except for Korthout et al. who uncovered a partial oxidation of Eu 2þ to Eu 3þ in SrAl 2 O 4 ∶Eu 2þ ; Dy 3þ [34]. Exposure to x rays does not only lead to radioluminescence (RL) from Eu 2þ but also to charge trapping, as witnessed by the slow increase of RL intensity (the so-called charging) after switching on the x-ray beam and by the afterglow when switching off the x-ray beam. Traps were thermally cleaned to ensure that the electron transfer was reversible, i.e., that the observed oxidation was not due to a degradation process. This confirmed that indeed an electron is released from the Eu 2þ ion during the charging (trapping) process, but no convincing information could be extracted on where it is trapped [34]. Carlson et al. recorded oxidation of Eu 2þ in Sr 2 MgSi 2 O 7 ∶Eu, but on a much longer timescale (hours) than the typical charging (seconds to minutes [26,34]), indicating that this was presumably caused by degradation, rather than the filling of traps [32].
Here, the valence state changes of Eu and Dy in Sr 4 Al 14 O 25 ∶Eu 2þ ; Dy 3þ persistent phosphors are recorded via resonant inelastic x-ray scattering (RIXS) [ Fig. 1(a)]. It is found that a fraction of the Dy 3þ is reduced to Dy 2þ along with the oxidation of Eu 2þ to Eu 3þ , and that this reversible electron transfer is indeed responsible for the persistent luminescence.
The situation for a pristine Eu,Dy codoped persistent phosphor is represented in Fig. 1(b)(i), where Eu 2þ and Eu 3þ coexist with Dy 3þ . Charging by blue to ultraviolet light [ Fig. 1 oxidation of a part of the Eu 2þ while the redox partner, Dy 3þ or another defect, gets reduced. This metastable state is long-lived and is slowly depleted at room temperature in the case of persistent phosphors [ Fig. 1(b)(iii)], giving rise to afterglow or TL when the electron is transferred back to the Eu center. Infrared light has a similar effect by electronically exciting the trapped electron, giving rise to OSL [ Fig. 1 X rays are ideal to probe the oxidation states of the dopants, but they simultaneously induce fast trapping, requiring measures to overcome this beam effect.
To slow down the x-ray induced charging, the sample was mounted on a rotating stage (spun at 2700 rpm), with the x-ray beam hitting 5 mm off axis. This allows us to probe a larger sample volume and reduces the number of photons per unit volume per unit time. Indeed, the time to reach a steady state (the equilibrium between the trapping effect of x rays and the detrapping processes) is hugely enlarged, passing from a few milliseconds to tens of seconds, enabling measurement of the time evolution of the x-ray induced charging.
In addition, the density of trapped charges can be controlled by illuminating the sample with violet or IR lasers (Fig. 1). This gives rise to the situations represented schematically in Figs. 1(b)(vi) and 1(b)(vii) where the effect of the lasers is superimposed on the effect of the x rays. The use of the external light sources hence allowed us to affect a fraction of the charges and create a differential effect on top of the unavoidable charging induced by x rays. : monochromatic x rays strike the rotating sample, the x-ray fluorescence is analyzed by a spectrometer in Rowland geometry, using an array of spherically bent crystal analyzers and an avalanche photodiode. The sample is irradiated by violet or IR lasers and the radioluminescence (RL) is recorded. The schematic representation in (b) illustrates how light of different wavelengths, temperature, and x rays induce electron transfer and hence affect the Eu 2þ -Eu 3þ (bluered) and Dy 2þ -Dy 3þ (green-yellow) equilibria in the phosphor. (c) Camera image of the rotating sample under steady-state conditions. The x-ray beam hits the sample at point 1. Normalized radioluminescence spectra are displayed (d), measured on three different spots on the rotating sample [indicated in (c)], featuring Eu 2þ 4f 6 5d 1 → 4f 7 broadband (400-570 nm) and Dy 3þ 4f 9 → 4f 9 line emissions at around 480, 575, 675, and 760 nm, following the 4 F 9=2 → 6 H J (2J ¼ 15, 13,11,9) transitions, respectively.
Finally, we could connect the x-ray measurements with the luminescence processes by collecting the RL emitted by the sample under x-ray irradiation. The parallel analysis of the two datasets was fundamental to ensure that the observations for Eu and Dy are indeed connected to the persistent luminescence process.
Our experiments were conducted on commercial Sr 4 Al 14 O 25 ∶Eu; Dy powders at beam line ID26 of the European Synchrotron Radiation Facility (ESRF) in Grenoble, France [36,37]. The sample was diluted in boron nitride, pressed to a pellet and mounted on a rotating sample stage. The Ln L III absorption edges were used to probe the oxidation states of Eu and Dy. The rotating sample was positioned in the x-ray beam (Fig. 1). The incident photon energy was scanned across the Eu and Dy L III edge. The maximum of the characteristic L α1 emission was selected by an x-ray emission spectrometer based on Rowland geometry, resulting in high energy resolution fluorescence detected (HERFD) x-ray absorption near edge structure (XANES) spectra. Here, the main advantage of using the HERFD mode to collect XANES is its superior energy resolution which increments the sensitivity of the technique to small spectral variations [38][39][40]. Technical details are given in the Supplemental Material (SM) [41].
The spinning sample (16°=ms) allows us to record spectral changes upon continuous x-ray irradiation on a millisecond timescale [ Figs. 1(c) and 1(d)]. The color of the luminescence changes, depending on the position, and hence on the time after irradiation [ Fig. 1(d)]. This is analyzed in detail by recording the luminescence spectra at different positions (1, 2, 3) on the spinning pellet, corresponding to delays of 0, 2, and 20 ms after irradiation, respectively. Upon x-ray irradiation (position 1), different spectral components are clearly visible: the blue-green broadband 4f 6 5d 1 → 4f 7 emissions from Eu 2þ in Sr 4 Al 14 O 25 peaking around 495 nm [42,43], as well as characteristic Dy 3þ inter-4f 9 transitions, predominantly in the yellow. The first few milliseconds after irradiation (position 2), the parity and spin-allowed Eu 2þ emission already decayed (order of 1 μs) and the slow-decaying Dy 3þ emission dominates the spectrum. The Dy 3þ decay is clearly visible on the picture in Fig. 1(c). After 20 ms (position 3), also the excited Dy 3þ ions have largely decayed and the afterglow remains, showing a dominant Eu 2þ 4f 6 5d 1 → 4f 7 emission. It is hence clear that the x rays are indeed able to induce the persistent luminescence. Figure 2(a) shows HERFD XANES spectra, measured at the Eu L III edge. These spectra feature two relatively sharp peaks, so-called white lines, that correspond to the resonant 2p 6 4f N → 2p 5 4f N 5d 1 electronic transition of Eu 2þ (N ¼ 7, around 6.975 keV) and Eu 3þ (N ¼ 6, around 6.983 keV) [44,45]. The characteristic 7-8 eV separation between Eu 2þ and Eu 3þ white lines, and more generally between Ln 2þ and Ln 3þ white lines [46,47], is often used to acquire quantitative and qualitative information about the concurrence of lanthanide oxidation states [44,45].
It was previously found that Eu 2þ oxidizes to Eu 3þ in SrAl 2 O 4 ∶Eu; Dy upon x-ray irradiation on the same timescale as the charging of the persistent luminescence [34]. Therefore, it is clear that the x rays, which are used as a probe for the lanthanide oxidation states, induce a net trapping effect, where a steady-state regime with a higher ½Eu 3þ ∶½Eu 2þ ratio, compared to a pristine sample, is established under irradiation [ Fig. 1(b)(v)]. In order to push this dynamic equilibrium in both directions, laser stimulation was applied on top of the x-ray irradiation: violet light to further increase ½Eu 3þ ∶½Eu 2þ [ Fig. 1

(b)(vi)]
and IR light to provoke detrapping, counteracting the effect of the x rays, thus lowering the ½Eu 3þ ∶½Eu 2þ ratio [ Fig. 1(b)(vii)]. Figure 2 shows the effect of the laser stimulation. The blue curve, for violet stimulation, presents a higher ½Eu 3þ ∶½Eu 2þ ratio than the black curve, for IR stimulation. Interestingly, a similar but opposite effect was found for the same experiment, performed at the Dy L III edge. The amount of Dy 3þ goes slightly down upon violet stimulation (yellow curve), compared to IR stimulation (black curve). Concurrently, a small increase of the spectral intensity in the region where the Dy 2þ white line is expected can be observed. Figures 2(e) and 2(f) display the differences between the HERFD XANES spectra, measured upon violet and IR stimulation, more clearly illustrating the valence changes.
The Dy 2þ white line lies at lower energy than the Dy 3þ white line, in the region where the Dy 3þ preedge features are also observed [ Fig. 2(d)]. The latter, which typically show a richer fine structure, are due to weakly allowed 2p 6 4f 9 → 2p 5 4f 10 transitions [48][49][50][51]. In our experiment, both features have a comparably low intensity. In order to confirm that the changes below the Dy 3þ white line are indeed due to the creation of Dy 2þ , RIXS planes (see SM [41]) were measured under both violet and IR illumination and compared to RIXS planes of the preedge region (Fig. 3). The Dy 3þ preedge structure is clearly seen in the RIXS plane of Dy 2 O 3 in the range of Ω ¼ 7785 eV, but it is broadened for the phosphor sample. The difference of the violet and IR illuminated RIXS planes confirms that the spectral changes correspond to a decrease of the amount of Dy 3þ and an increase of Dy 2þ : two broad bands are found, consistent with the rise of a Dy 2þ white line, rather than with changes in the Dy 3þ preedge region. This explains why the Dy 3þ preedge structure is obscured for the phosphor sample as the x rays inevitably create some Dy 2þ in the phosphor [see also Fig. 1(b)]. These observations clearly prove the existence of the following photoinduced electron transfer, in accordance with the scheme in Fig 1(b). To obtain reliable estimates for the quantity of Eu and Dy that changes oxidation states, the HERFD XANES spectra were fitted [52][53][54][55] with two Gaussians, one for each white line, and a smooth step function, representing the electronic transitions toward the dense set of unbound states, 2p 6 4f N → 2p 5 4f N ϕ 1 (see also SM [41]). The result for the difference spectra is shown in Figs. 2(e) and 2(f).
The area under the white line is associated with the concentration of Ln ions [44][45][46]. By exploiting that the total number of Ln ions remains constant throughout the experiment, it was verified that difference spectra of Figs. 2(e) and 2(f) correspond to 3% of all Eu and 2% of all Dy. Given the higher concentration of Dy in the phosphor (as verified by energy-dispersive x-ray spectroscopy [56]; see SM [41]), quantitative agreement is found between the valence changes of Eu and Dy within the experimental uncertainties. This number is of the same order of magnitude as a prior estimate for the number of Eu ions changing valence state, based on optical spectroscopy experiments on SrAl 2 O 4 ∶Eu; Dy [57].
The connection between the electron transfer and the persistent luminescence is substantiated by measuring their time evolution (Fig. 4). An optically cleaned phosphor, i.e., for which all traps were first emptied by OSL, is irradiated by the x-ray beam from t ¼ 0. The oxidation states of the lanthanide ions are sampled every second by recording the maximum of the white line for every oxidation state separately (see SM [41]). Figure 4(a) shows that the x-ray induced oxidation of Eu 2þ and reduction of Dy 3þ occur on the same timescale of about 20 s. After this, a dynamic equilibrium is reached where the forward and backward electron transfers, Eq. (1), occur at the same rate. This relatively long timescale is the consequence of the sample being mounted on a rotating stage, uniformly probing a  sample area which is ≈280 times larger than for a static sample (with a rotation period of ≈22 ms, each measured point is averaged over several rotations). For a static sample, the dynamic equilibrium would have been reached after 0.1 s, presumably explaining why no valence state changes were found in prior investigations [30][31][32][33]35]. The charging of the persistent luminescence, monitored by the increasing Eu 2þ emission, follows indeed the same temporal behavior as the electron transfer [ Fig. 4(b)]. In the same time frame of the charging, the Dy 3þ RL intensity shows a small decrease, presumably due to the partial reduction of the Dy 3þ ions [Eq. (1)].
The variations in luminescence are also visible in Fig. 4(d), showing the rotating sample at several time delays after switching on the x-ray beam. The commashaped signature of the Dy 3þ RL appears instantaneously, while the blue persistent luminescence builds up more slowly (see SM [41] for the associated intensity profiles [58,59]). Complementarily, the Dy 3þ RL disappears almost immediately after switching off the x-ray beam [ Fig. 4(e)], leaving the blue afterglow visible for several hours. The first two frames confirm that the intensity of the blue Eu 2þ luminescence that is probed during this experiment is due solely to the afterglow; i.e., the blue intensity along the race track does not change abruptly when the excitation is switched off.
In summary, a dedicated RIXS study is performed on Sr 4 Al 14 O 25 ∶Eu; Dy persistent luminescent materials in order to elucidate the electron transfer that underlies the energy storage process. While the x rays allow us to accurately probe the oxidation states of both dopants in real time, they unavoidably disturb the equilibrium electron occupation, prohibiting the unambiguous assessment of the persistent luminescence mechanism unless measures are taken to overcome the beam effect. A novel measurement approach is demonstrated, where illumination with visible and near-IR light are used to manipulate the electron occupations, in combination with an effective method to spread the applied x-ray dose by rapidly spinning the sample. Excitation by violet light induces trapping, creating metastable Eu 3þ -Dy 2þ centers, while IR light provokes detrapping. Quantitatively, 3% of the Eu ions and a corresponding number of Dy ions are affected, in agreement with the optically determined storage capacity. The characteristic time of the metastable Eu 3þ -Dy 2þ state coincides with the charging of the persistent luminescence. This is hence the first confirmation of the popular, yet debated, model where Dy 3þ acts as the main electron trap in persistent phosphors. J. J. J. and P. F. S. acknowledge the Ghent University Special Research Fund via the projects BOF/PDO/2017/ 002101 and GOA-Enclose, respectively. We acknowledge the European Synchrotron Radiation Facility (ESRF) for the allocation of beamtime (beamline ID26).