Observation of efficient population of the red-emitting state from the green state by non-multiphonon relaxation in the Er3+–Yb3+ system

The rare earth Er3+ and Yb3+ codoped system is the most attractive for showcasing energy transfer upconversion. This system can generate green and red emissions from Er3+ under infrared excitation of the sensitizer Yb3+. It is well known that the red-emitting state can be populated from the upper green-emitting state. The contribution of multiphonon relaxation to this population is generally considered important at low excitation densities. Here, we demonstrate for the first time the importance of a previously proposed but neglected mechanism described as a cross relaxation energy transfer from Er3+ to Yb3+, followed by an energy back transfer within the same Er3+–Yb3+ pair. A luminescence spectroscopy study of cubic Y2O3:Er3+, Yb3+ indicates that this mechanism can be more efficient than multiphonon relaxation, and it can even make a major contribution to the red upconversion. The study also revealed that the energy transfers involved in this mechanism take place only in the nearest Er3+–Yb3+ pairs, and thus, it is fast and efficient at low excitation densities. Our results enable a better understanding of upconversion processes and properties in the Er3+–Yb3+ system. Cross-relaxation energy back transfer between Yb3+ and Er3+ ions has been shown to be significant in red upconversion luminescence. The mixed rare-earth dopant system of erbium (Er3+) and ytterbium (Yb3+) ions is commonly used to achieve infrared-to-visible upconversion luminescence. Under infrared excitation, the Yb3+ ions absorb the incident light and consequent energy transfer to the Er3+ ions results in red and green emission. Now, Jiahua Zhang and co-workers from Changchun Institute of Optics, Fine Mechanics and Physics in China have conducted luminescence spectroscopy of Y2O3 doped with Er3+ and Yb3+ and discovered that the new energy pathway only takes place between nearest Er3+–Yb3+ pairs. In addition, they also found that this process is faster and more efficient than multiphoton excitation and makes a major contribution to red upconversion.


INTRODUCTION
Infrared to visible upconversion luminescence has been extensively studied for its fundamental value 1-3 and its various potential applications in upconversion lasers, 4 bioimaging, 5 etc. The codoping of Er 31 and a high concentration of sensitizer Yb 31 forms the most attractive energy transfer upconversion (ETU) system. Under infrared (980 nm) excitation of the sensitizer Yb 31 , this system can generate green and red upconversions originating from the 4 S 3/2 R 4 I 15/2 and 4 F 9/2 R 4 I 15/2 transitions of Er 31 , respectively. Unlike the green upconversion, the red upconversion benefits from several possible excitation mechanisms. 6,7 Multiphonon relaxation (MPR) from the upper 4 S 3/2 state and ETU from the lower intermediate 4 I 13/2 state are generally considered dominant at low infrared excitation densities because other mechanisms involving three photon processes 6,7 become important only at high infrared excitation densities, 8 which is not the topic of this work.
The MPR is not the only mechanism for populating the 4 F 9/2 from the 4 S 3/2 ; a non-MPR mechanism was proposed earlier, 8 but it has not been considered important since then. This mechanism involves two sequential energy transfers between Er 31 and Yb 31 . The first step is a well-known cross-relaxation (CR) energy transfer from Er 31 in the 4 S 3/2 excited state ( 4 S 3/2 R 4 I 13/2 ) to Yb 31 in the ground state ( 2 F 5/2 r 2 F 7/2 ), 9 resulting in the excitation of Er 31 4 I 13/2 and Yb 31 2 F 5/2 . The subsequent step is an energy back transfer from the Yb 31 ( 2 F 5/2 R 2 F 7/2 ) excited by the CR to Er 31 in the 4 I 13/2 state ( 4 F 9/2 r 4 I 13/2 ) to promote the excitation of Er 31 4 F 9/2 . Hence, the CR can be divided into two parts: one is followed by the energy back transfer, CRB, and the other is not followed by the energy back transfer, CRNB. The Yb 31induced green emission quenching of Er 31 by the CR has been widely recognized. 9,10 However, the population of the 4 F 9/2 by the CRB from the 4 S 3/2 has rarely been studied or valued in both photoluminescence (PL) and upconversion luminescence (UCL).
In this article, we report an observation of the CRB in cubic Y 2 O 3 : Er 31 , Yb 31 . We find the CRB can be more efficient than MPR and can even make a major contribution to the red UCL. To the best of our knowledge, this is the first time that the CRB has been found to be important to the red emission both in the PL and UCL of the Er 31 -Yb 31 system.

Sample preparation
The cubic Y 2 O 3 :0.002Er 31 , xYb 31 (x50, 0.04, 0.1, 0.2, 0.3) samples were prepared by the normal firing precursor method. 2 The firing precursor is more favorable to achieve uniform and highly crystallized samples than the traditional solid-state reaction. The starting aqueous solutions, Y(NO 3 ) 3 ,Yb(NO 3 ) 3 , and Er(NO 3 ) 3 , with corresponding mole ratios were mixed and stirred vigorously to form a homogeneous solution. The samples were obtained after being dried at 100 6 C for 6 h and then calcined at 1600 6 C for 6 h. The expression of Y 2 O 3 :0.002Er 31 , xYb 31 in the present paper means the formula Y 220.0022x Er 0.002 Yb x O 3 . The low Er 31 and high Yb 31 concentrations applied in this work were used to achieve a real ETU system and suppress the interaction among Er 31 ions for simplifying the ETU processes.

Spectroscopy measurements
Steady state PL and UCL spectra were measured using an EI-FS920 fluorimeter with a xenon lamp as an excitation source for PL and a CW 980 nm laser diode as an excitation source for UCL. The decay curves of PL and UCL were detected using a Triax 550 spectrometer (Jobin-Yvon) and recorded by a Tek-tronix digital oscilloscope (TDS 3052), while a 10 ns pulsed laser with tunable wavelengths from an optical parametric oscillator pumped by a Nd:YAG laser (spectra-physics, GCR 130) was used as an excitation source. In energy level lifetime measurements, the excitation wavelengths were tuned to 520 nm for Er 31 4 S 3/2 , 650 nm for Er 31 4 F 9/2 and 4 I 11/2 , 1480 nm for Er 31 4 I 13/2 and 980 nm for Yb 31 2 F 5/2 . The lifetime is defined as the area under the decay curve with normalized initial intensity. The UCL spectra under pulse excitation at 980 nm were detected using a USB4000 spectrometer (Ocean Optics), which gives the time-integrated intensities.

RESULTS AND DISCUSSION
Observation of the CRB in PL Figure 1 shows PL spectra of Y 2 O 3 :0.002Er 31 , xYb 31 under 520 nm excitation of Er 31 2 H 11/2 , which can rapidly relax to the 4 S 3/2 due to their proximity in energy. One can observe that the red (660 nm) to green (560 nm) emission intensity ratio (R/G) increases considerably with increasing x. A similar result upon Er 31 4 F 7/2 excitation with 488 nm was also observed in Y 2 O 3 :Er 31 , Yb 31 nanocrystals, 11 and an interaction between two excited Er 31 ions ( 4 F 7/2 R 4 F 9/2 and 4 F 9/2 r 4 I 11/2 ) 7 was considered the main excitation mechanism of the red-emitting 4 F 9/2 state. In the present work, the 4 F 7/2 is not excited and the Er 31 concentration is low, so the above mechanism is not feasible. We found through experimentation that the R/G ratio remains unchanged if La 31 or Lu 31 is doped instead of Yb 31 , indicating that the 4 S 3/2 -4 F 9/2 MPR rate hardly changes with rare earth doping. Therefore, we attribute the Yb 31 -induced R/G increase to the effect of CRB, which benefits from well-matched energy level structures between Er 31 and Yb 31 , as sketched in the insert to Figure 1. The CRB may take place within the same Er 31 -Yb 31 pair, meaning that the Yb 31 ion excited by the CR transfer from Er 31 transfers its energy back again to the same Er 31 ion. The CRB rate in this case is independent of excitation densities, similar to the MPR. If the CRB takes place not within the same Er 31 -Yb 31 pair, i.e., the excitation energy of Yb 31 is not retransferred to the same Er 31 but transferred to any other Er 31 ion in the 4 I 13/2 state, then the CRB rate in this case is dependent on the concentration of the other Er 31 ions and thus dependent on excitation densities. Therefore, the CRB may become important only at high excitation densities. In the present work, the Yb 31 -induced R/G increase can be observed under weak 520 nm excitation using a grating monochromator with a xenon lamp. We found that the enlarged R/G ratio in the presence of Yb 31 is independent of excitation densities using the xenon lamp as an excitation source. Based on the CRB models mentioned above, we consider that the CRB takes place within the same Er 31 -Yb 31 pair in the samples in this work.
Time evolutions of PL are measured for distinguishing the CRB from the MPR, as shown in Figure 2. With increasing x, the appreciable shortening of the 4 S 3/2 lifetime (t 4 ) ( Figure 2a) is due to the CR, 9,10 and the small shortening of the 4 F 9/2 lifetime (t 3 ) ( Figure 2b) is due to weak coupling 12 of Er 31 4 F 9/2 R 4 I 15/2 with Yb 31 2 F 5/2 r 2 F 7/2 . In the time evolution of the red emission after the 4 S 3/2 is populated (Figure 2c), we find that the samples containing Yb 31 appear to undergo a fast build-up process, unlike the Yb 31 -free sample that appears as a normal slow rising edge. 13 Accordingly, the time evolution patterns in the presence of Yb 31 shown in Figure 2c are the combined patterns of the fast build-up component and the slow build-up component. If one normalizes the initial emission intensity of the fast build-up component ( Figure 2c) and denotes the total area under the time evolution curve by T 3 , the proportion of the emitted red photons by the fast build-up component of the total emitted red photons can be calculated by t 3 /T 3 . This proportion is found to be consistent with the proportion of the increment in the R/G ratio (Table 1), written as a512(t 3 /t 3,0 )(R/G) 0 /(R/G), where the subscript 0 identifies the case of x50 and (t 3 /t 3,0 )(R/G) 0 is the R/G ratio contributed only by MPR in the presence of Yb 31 . The result indicates that  A new excitation mechanism in Er 31 -Yb 31 system JH Zhang et al 2 the CRB is a fast process. Accordingly, there must be a fast CR. However, a corresponding fast decay of the green emission is not detected (Figure 2a). We thus deduce that the fast green emission may be completely quenched by the fast CR. This deduction is further evidenced in the following section.
Efficiency of the CRB in PL After the 4 S 3/2 is populated upon 520 nm excitation, the 4 S 3/2 undergoes depopulation by radiative transition, MPR, CRNB and CRB, with respective efficiencies of g r4 , g MPR43 , g CRNB and g CRB . Figure 3 shows the comparison of PL spectra under 520 nm excitation with that under 650 nm excitation of the 4 F 9/2 . Here, the red intensity under 650 nm excitation is normalized to that under 520 nm excitation. Under 520 nm excitation, one can find that the green intensity decreases rapidly with increasing x, followed by simultaneous intensity enhancement of Er 31 4 I 13/2 R 4 I 15/2 emission at 1530 nm and Yb 31 2 F 5/2 R 2 F 7/2 emission at 1040 nm compared to that under 650 nm excitation. The feature is an explicit indication of the CRNB process. In the absence of Yb 31 , the Er 31 4 I 11/2 R 4 I 15/2 emissions for both 520 nm and 650 nm excitations can be observed at 1000 nm; they coincide because the population of 4 I 11/2 by 4 S 3/2 R 4 I 9/2 , 4 I 11/2 radiative transitions for 520 nm excitation is very small with respect to that from the 4 F 9/2 by cascade MPR. In the presence of Yb 31 , the Er 31 4 I 11/2 R 4 I 15/2 emission is too weak to be identified, and instead, the Yb 31 emission is dominant. This is due to a highly efficient energy transfer from Er 31 4 I 11/2 to Yb 31 2 F 5/2 , 14 which is also evidenced by the very weak 1530 nm emission of Er 31 under 650 nm excitation. In addition, one may find that the Yb 31 emission intensities decrease with increasing x in the x range of 0.1-0.4 for 520 nm excitation. The decline is due to concentration quenching because the critical concentration of Yb 31 was observed to be 7.36% in the Y 2 O 3 host. 15 We evaluate g r4 , g MPR43 , g CRNB and g CRB based on the observed increase in the Er 31 1530 nm emission, Yb 31 1040 nm emission and Er 31 R/G ratio for 520 nm excitation. Considering the distinct increment (I 1 2i 1 ) in 1530 nm emission intensity in the presence of Yb 31 , one may also observe a small increase for x50. The small increase is attributed to the feeding by the 4 S 3/2 R 4 I 13/2 radiative transition, which has a big branch ratio (b50.27). 16 Hence both the CRNB and the 4 S 3/2 R 4 I 13/2 radiative transition contribute to the increase. Regarding the experimentally observed x-independent 4 I 13/2 lifetime, we have, as described in Supplementary Equations (S1)-(S4) where I i and i i are emission intensities from the ith state of Er 31 upon 520 nm and 650 nm excitations, respectively. These emission intensities on the right side of Equation (1) are dependent on x.
For the increment (I d 2i d ) in Yb 31 emission intensity, the CRNB is the only contributor, while the rest intensity (i d ) is populated by quasiresonant energy transfer from Er 31 4 I 11/2 with a transfer efficiency close to 1 due to the high Yb 31 concentration. The 4 I 11/2 can be populated from the 4 F 9/2 via the 4 I 9/2 by cascade MPR with an efficiency of almost 100%, regarding a very short 4 F 9/2 lifetime (approximately 20 ms) compared with its radiative lifetime (600 ms). 16 As a result, the population of the 4 F 9/2 by the MPR and the CRB from the 4 S 3/2 finally and completely reaches the 4 I 11/2 . Hence, the excitation source of the rest Yb 31 emission is the MPR and the CRB from the 4 S 3/2 . Then, we have where the subscript d denotes Yb 31 .
The red emission is populated by MPR and CRB from the 4 S 3/2 , so the R/G ratios simply satisfy Using Equations (1)-(3) with g r4 1g MPR43 1g CRNB 1g CRB 51, we calculated the efficiencies, as shown in Figure 4a. We obtained g MPR43 /g r4 52.08, which yields g r4 532% for cubic Y 2 O 3 :Er 31 . The measured green PL intensities (square) follow the tracks of g r4 well, but it decreases faster than t 4 with increasing x (Table 1). This feature further implies the existence of an undetected fast decay in the green emission, as indicated in the above section, like the case of the Perrin model. 17 We may consider an 'active sphere' centered at an Er 31 ion so that its green emission can be completely quenched by a Yb 31 ion located within the sphere through the CR. If the number of the effective cation sites within the sphere is n, the fraction (f) of Er 31 ions that  have no Yb 31 in the sphere satisfies the binomial distribution: f5(12x/2) n . Considering that the green luminescence yield (I 4 or g r4 ) can be regarded as the area under its decay curve including the undetected fast component, then f can be obtained either by (I 4 / I 4,0 )(t 4,0 /t 4 ) or (g r4 /g r4,0 )(t 4,0 /t 4 ) from the decay picture of the green emission containing a fast component with a lifetime close to 0 and a slow one with a lifetime of t 4 , weighted by 12f and f, respectively, as described in Supplementary Fig. S1 and Equation (S5). In cubic Y 2 O 3 , there are two Y sites with respective C 3i and C 2 symmetries. The C 3i has an inversion center on this site; the 4f-4f transition of a rare earth ion is forbidden. Therefore, only the C 2 site is effective. A C 2 site has four of the first nearest C 2 sites, four of the second nearest C 2 sites and six of the next nearest C 2 sites. The fitting results point to n58, as shown in Figure 4b, indicating that the green emission of an Er 31 can be completely quenched only by an Yb 31 located in the nearest site of the Er 31 ion. As a result, the nearest Er 31 -Yb 31 pairs exhibit completely quenched green emission and produce the undetected fast decay of the green emission. As we have observed, the CRB is a fast process; thus, the CRB takes place only within the nearest Er 31 -Yb 31 pairs. Of course, the occurrence of a fast CRNB is also expected in the nearest Er 31 -Yb 31 pairs. Accordingly, the distant Er 31 -Yb 31 pairs can only perform the slow CRNB to reduce t 4 . In Figure 4a, over 30% of the CR is the CRB upon 520 nm excitation, and the CRB becomes more important than the MPR for populating the red-emitting state from the green state as xo0.1. The efficient CRB benefits from the nearest Er 31 -Yb 31 pairs that can compete with the energy transfer from the Yb 31 ions excited by the CR to other Yb 31 ions outside the pairs. The nearest Er 31 -Yb 31 pair ensures that the CRB takes place within the pair because the energy back transfer from Yb 31 to other Er 31 undergoes a longer interaction distance due to the low Er 31 concentration used in this work. Hence, no dependency of the CRB on excitation densities is observed in the present work. The Perrin model, like CRB, may be governed by the exchange mechanism described by Inokuti and Hirayama. 17 In CRB, the CR transfer from Er 31 ( 4 S 3/2 R 4 I 13/2 ) to Yb 31 ( 2 F 5/2 r 2 F 7/2 ) and the back transfer from Yb 31 ( 2 F 5/2 R 2 F 7/2 ) to Er 31 ( 4 F 9/2 r 4 I 13/2 ) suffer from an energy mismatch of ,1700 cm 21 and ,1500 cm 21 , respectively, which is more than twice as large as the cutoff phonon energy of 600 cm 21 in Y 2 O 3 . The performance of the energy transfers thus requires emission of two or three phonons to make up for the large energy mismatch. In spite of requirement of several phonons, the high Yb 31 concentration and the nearest Er 31 -Yb 31 pairs make the transfers efficient.
Effect of the CRB on UCL Figure 5 shows the UCL spectra for Y 2 O 3 :0.002Er 31 , 0.1Yb 31 upon pulse and CW infrared (980 nm) excitations. For comparison, its PL spectrum is also presented. The phenomenon 11,18-20 that the R/G ratio in PL is much less than that in UCL is evident. Ignoring the CRB, it was naturally deduced from the phenomenon that the ETU from 4 I 13/2 makes the major contribution to the red UCL. 18,19 If the CRB is operative, the above deduction could be possibly invalid because of the following reasons: the nearest Er 31 -Yb 31 pairs can be preferentially excited to the Er 31 4 S 3/2 state by two sequential energy transfers from Yb 31 in ETU, as shown in the insert to Figure 5. Meanwhile, the CRB enables the nearest Er 31 -Yb 31 pairs to only have a red emission in the visible spectral region. In the PL measurement, however, each Er 31 ion has the same probability to be excited to the 4 S 3/2 state by ground state absorption. Based on the analysis above, a much smaller R/G ratio of PL than that of UCL is still achievable if the CRB makes the major contribution to the red UCL. In other words, the observation of a much smaller R/G ratio for PL than for UCL cannot exclude the major role of the 4 S 3/2 state in populating the red-emitting state in the upconversion of the Er 31 -Yb 31 system. In our experiment, the PL and UCL temporal behaviors of Y 2 O 3 :0.002Er 31 , 0.1Yb 31 (Figure 6 A new excitation mechanism in Er 31 -Yb 31 system JH Zhang et al 4 demonstrate that the 4 F 9/2 is populated mainly from 4 S 3/2 in ETU, as analyzed below. From Figure 6, the Yb 31 emission decays exponentially (with a decay time of t d 5840 ms), indicating the rapid diffusion limited energy transfer. The decay time of the green UCL is 391 ms, and that of the red is 412 ms. These UCL decay times are much longer than t 4 (66 ms) and t 3 (20.9 ms) (Table 1), exhibiting the typical effect of ETU. Thus, the UCL decay function is mainly determined by the product of the decay functions of the Yb 31 2 F 5/2 and Er 31 intermediate states. The green UCL decay time approaches t d /2, reflecting that Er 31 4 I 11/2 , as the intermediate state for the green upconversion, has the same decay time as Yb 31 2 F 5/2 because the two levels are thermally coupled in case of rapid energy transfer between them, as predicated earlier. 14 The approximation of the red UCL decay time to the green one indicates that the 4 F 9/2 is populated mainly from the 4 S 3/2 under pulse infrared excitation; otherwise, the red could have a decay time close to t d (840 ms) because of the quite long lifetime (t 1 58.4 ms) of the intermediate 4 I 13/2 state. Now, we want to determine if the 4 F 9/2 is still populated mainly from the 4 S 3/2 under CW infrared excitation.
In Figure 5, one may find that the R/G ratio in UCL for CW excitation, (R/G) UCL @CW, is larger than that for pulse excitation, (R/ G) UCL @pulse. The smaller (R/G) UCL @pulse arises from the low build-up process of the 4 I 13/2 state (Figure 6), as described in Supplementary Equation (S10), which slows down the ETU from the 4 I 13/2 to the 4 F 9/2 after pulse infrared excitation. Taking the different R/G ratios into account, the percentages of the red component R 4 , populated from the 4 S 3/2 , and R 1 , from the 4 I 13/2 , can be evaluated.
If the 4 S 3/2 and the 4 F 9/2 states are populated under infrared excitation in ETU, we denote the mean emitting efficiencies of the two states by g G and g R , respectively. While the mean depopulating efficiency from the 4 S 3/2 to the 4 F 9/2 via MPR and the CRB is denoted by g 43 , the R/G ratio in UCL is where g G , g R and g 43 are unchanged for CW or pulse infrared excitations. In view of the high Yb 31 and low Er 31 concentrations, we assume here the simplest possible model; (i) ETU undergoes the rapid diffusion limited energy transfer from Yb 31 that can be described by an average energy transfer rate; (ii) the ground and excited states absorption of Er 31 is neglected. We also assume the 4 F 7/2 can rapidly and completely relax down to the 4 S 3/2 state due to their proximity in energy.
From the diagram of ETU processes inserted in Figure 5, the population flow from the 4 I 11/2 to the 4 F 7/2 is C d5 n d n 2 and that from the 4 I 13/2 to the 4 F 9/2 is C d3 n d n 1 . Under CW excitation, the population flow is time independent, and thus, R 4 @CW5g R g 43 C d5 n d n 2 and R 1 @CW5g R C d3 n d n 1 , where C di is the coefficient for energy transfer from donor Yb 31 to the ith state of Er 31 , and n d and n i are populations of Yb 31 2 F 5/2 and Er 31 in the ith state, respectively. Then, the R 1 /R 4 ratio for CW infrared excitation is expressed as where W 21 is the sum of the 4 I 11/2 -4 I 13/2 radiative and MPR rates, which satisfies W 21 n 2 5n 1 /t 1 for a steady state excitation. Under pulse infrared excitation of Yb 31 , we detect the upconversion luminescence yield over time. Based on the rate equations for describing ETU in the Er 31 -Yb 31 system, as in Supplementary Equations (S7)-(S13), we obtain From Figure 5, the R/G ratio in UCL for CW excitation is 1.2 times as large as that for pulse excitation, Combining Equations (4)- (7) with t d 5840 ms and t 1 58.4 ms gives (R 1 /R 4 )@pulse50.02 and (R 1 /R 4 )@CW50.22 for Y 2 O 3 :0.002Er 31 , 0.1Yb 31 . These values mean that only 2% of the red UCL is populated from the 4 I 13/2 and 98% from the 4 S 3/2 under pulse infrared excitation, and for CW infrared excitation, 18% of the red UCL is populated from the 4 I 13/2 and 82% from the 4 S 3/2 . The main role of the 4 S 3/2 in populating the red UCL reflects a strong CRB. Meanwhile, the small contribution of the ETU from the 4 I 13/2 is partially due to a small value of C d3 /C d5 that is calculated to be 0.124g 43 using Equation (5). In this calculation, a W 21 of 218 s 21 is used, which is determined from our measured 4 I 11/2 lifetime (t 2,0 ) of 2.73 ms in Y 2 O 3 :0.002Er 31 and the reported 16 intrinsic 4 I 11/2 lifetime of 6.03 ms and 4 I 11/2 R 4 I 13/2 radiative branch ratio of 0.108. The small C d3 /C d5 ratio is well understood in view of a large energy mismatch (,1500 cm 21 ) in the transfer from Yb 31 ( 2 F 5/2 R 2 F 7/2 ) to Er 31 ( 4 F 9/2 r 4 I 13/2 ) relating to C d3 , while the energy transfer from Yb 31 ( 2 F 5/2 R 2 F 7/2 ) to Er 31 ( 4 F 7/2 r 4 I 11/2 ) relating to C d5 is quasi-resonant.
Obviously, a high Yb 31 concentration and preferential excitation of the nearest Er 31 -Yb 31 pairs in ETU can promote the CRB process. Meanwhile, increasing the population ratio of Er 31 4 I 11/2 to 4 I 13/2 is beneficial to ETU from the 4 I 11/2 to the 4 F 7/2 , and it thus can promote the contribution of the CRB to the red UCL against the ETU from the 4 I 13/2 to the 4 F 9/2 . Therefore, the host material with low cutoff phonon energy likely has a pronounced CRB process like Y 2 O 3 . However, this population ratio is usually reduced in nanomaterials because the 4 I 11/2 -4 I 13/2 MPR is strongly enhanced by coupling with OH 2 groups on the large surface of the nanomaterials. 21

CONCLUSIONS
We observed a pronounced CRB process for populating the Er 31 redemitting state from its green state in cubic Y 2 O 3 :Er 31 , Yb 31 . We found the CRB can be more efficient than MPR and can even make the major contribution to the red emission in both PL and UCL. The CRB takes place only in the nearest Er 31 -Yb 31 pairs, and thus, it is a fast and efficient process at low excitation densities. The present research methods may be applied to a general Er 31 -Yb 31 system, and the identification of the importance of the CRB in other material hosts is expected.