Evidence for energy transfer induced by superexchange in LaF 3 : Pr3+

2014 Direct measurement of fluorescence quenching rates for various Pr3+ pairs in weakly doped LaF3 are reported. Since few pairs only are quenched, the transfer mechanism is short range. From the rate values, the shape of the fluorescence decay for heavily doped samples is predicted. The good agreement with the observed decay proves the interest of the connection between microscopic and macroscopic measurements. Finally arguments are developed to prove that the transfer mechanism is superexchange. J. Physique LETTRES 43 (1982) L-745 L-753 Classification Physics Abstracts 78.5020137855 ler NOVEMBRE 1982, Among the possible interaction mechanisms for optical excitation, exchange is often discounted in rare earth inorganic systems although Birgeneau pointed out that it could be as efficient as the other mechanisms [1]. Frequently, a multipolar electric interaction is postulated, the nature of which is deduced from the shape of the donor fluorescence decay using the Inokuti-Hirayama formula [2]. When used for randomly doped crystals, that method implies a statistical average over the possible positions of the acceptors around the donors and is subject to the hypothesis that a single mechanism is active. It is well known that the conclusions deduced from this « macroscopic » approach are questionable, the fit between theoretical and experimental decay curves never being perfect Improvements of this approach either by Dornauf and Heber [3] and Siebold (*) Laboratoire associ6 au C.N.R.S. Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyslet:019820043021074500 L-746 JOURNAL DE PHYSIQUE LETTRES and Heber [4] with a « discrete model » instead of a continuous model, or by Hegarty et al. [5] who deduce the microscopic transfer rate between two nearby ions from macroscopic results, need nevertheless the hypothesis that a single mechanism is active. It has been shown recently that microscopic transfer rates can be measured directly when a class of pairs of ions can be selectively excited [6]. Using weakly doped crystals, pairs are well « isolated » and their intrinsic behaviour is observed. From these measurements, it was inferred that the transfer mechanism could not be the same for weakly and strongly coupled pairs. The possible co-existence of various mechanisms has been considered by Watts [7] who showed, by numerical calculation of the quadrupole-quadrupole and dipole-dipole coupling strength between Yb3 + and Er3 +, that the former dominates the latter for ions separated by less than 9 A. The results reported here, again based on the direct study of the properties of pairs, show unambiguously that a short range interaction dominates the dipole-dipole interaction and is responsible for the quenching previously attributed to dipole-dipole coupling from macroscopic measurements [8]. They are then used to predict, with success, the behaviour of more heavily doped samples : starting from microscopic values macroscopic experiments are explained. The system studied, LaF3 : Pr3 +, is a very well known system whose energy levels have been tabulated in detail by Carnall et al. [9]. Most of the energy transfer experiments made with it have been presented and analysed by Yen in a review paper on energy transfer in rare earth ions in crystals [10]. The present paper is centred on the quenching of the 3po fluorescence. The first study of this quenching by Brown et al. [11] has shown that it results from cross relaxation between one Pr3 + excited and one nearby unexcited Pr3 +. This quenching has recently been studied by the Madison group with modem experimental techniques. Lai et al. [12], by a cooperative absorption, excite two neighbouring Pr3 + ions in PrF3 with a single photon, putting one ion into the 3p 0 state, the other into the 3 F2 state. Ions excited into 3 Po have then one less quenching channel, the nearby ions excited into 3 F2 being not able to cross relax with them. By measuring the change of the lifetime of these ions with respect to that of normally excited ions, the cross relaxation rate between two neighbouring ions is deduced. Hegarty et al. [8], using Inokuti and Hirayama formula, deduce also the value for this rate within the hypothesis of dipole-dipole coupling. To reach this conclusion, they must subtract from the fluorescence signal an exponential component they attribute to ions in « non quenching sites ». In this work the values of the cross relaxation transfer rates responsible for the quenching of 3p 0 fluorescence are directly measured for various pairs after their selective excitation. They are then used to show that the interaction has a short range and to predict the quenching for more heavily doped crystals. The fact that these measurements are made with a 0.1 % doped crystal in which there is no macroscopic quenching (the main line, due to isolated ions has a long lifetime) is not a paradox but shows the peculiarity of the method. 1. Experimental results. Figure 1 shows the absorption spectrum around the main 3H4-+ 3po transition. It is now well established that satellites are due to pairs of ions [6]. By monitoring the fluorescence induced by a narrow dye-laser pulse, it is possible to record time resolved excitation spectra as shown in figure 1 or to observe the fluorescence decay after a selective excitation of a given satellite as given in figure 2. For the f line, the slow component associated with the central line has been subtracted from the actual signal. The decay for lines a and f is exponential for more than two decades, indicating that all the pairs associated with one line have the same dynamical properties. Values of the decay times r are given in table I. It has not been possible, for lines c, d, e, to subtract the contribution of the wing of the central line because the decay times are too close. However, since the decay time in wings of the main line is shorter than that of lines c and d (this is deduced from the comparison of the two excitation spectra shown in figure 2), its measurement gives an upper limit for lines c and d. This limit is given in table I. L-747 ENERGY TRANSFER INDUCED BY SUPEREXCHANGE Fig. 1. The lower curve is the absorption spectrum of a LaF3 : 0.1 % Pr3 + sample around the 3H4 -+ 3p 0 transition. The two other curves are time resolved excitation spectra of the 3 Po fluorescence with two different delays after the laser pulse. Note that the gain for the 40 us spectrum is twice that for the 5 us spectrum. The reduction of the decay time in the wing of the main line, already pointed out for heavily doped samples [13] as well for 1 D2 fluorescence in weakly doped samples [14] will be discussed elsewhere. As first proposed by Brown et al. [ 11 ], the reduction of the 3 Po lifetime results from a cross relaxation between an excited and an unexcited ion. From the measured lifetimes r and the radiative lifetime To = 48 x 106 J.1s of the isolated ions [15] it is easy to calculate the cross relaxation rate w c for each satellite by The choice of to as the radiative lifetime for the ions of the pairs is justified by the fact that the relative intensities of the fluorescence lines corresponding to the transitions from 3 Po towards the levels of the sublying multiplets are the same whether the central line or a satellite is excited. If the perturbation created on a Pr3 + by another nearby Pr3 + changed significantly the oscillator strengths of the transitions, it would be surprising that this change would be the same for all the transitions. These results are now used to determine the nature of the coupling responsible for the observed cross relaxation. As it will appear below, this determination will result from strongly convergent evidence. The only remaining problem lies in the difficulty of associating a definite class of pairs with each satellite. This problem has already been pointed out [6] and its resolution would need uniaxial stress experiments as those made in ruby [16]. L-748 JOURNAL DE PHYSIQUE LETTRES Fig. 2. 3p 0 decay of a LaF3 : 0.1 % Pr3 + sample following a selective excitation: 8 of line f, 0 of line a, . of lines c, d, and the central line. The two other curves correspond to lines b and e. Note that for f line, the contribution of the wing of the main line has been subtracted. Table I. Values of the decay time 1: and of the cross relaxation rate W ~ for the various pairs associated with the satellite of figure 1. The determination of the nature of the coupling between the two Pr3 + ions of a pair is based on three arguments now explained in detail. 1.1 VARIATION OF THE TRANSFER RATE WITH THE DISTANCE. The sequence of the transfer rates of table I, after normalization is 1-0.26-0.01-0.006. From the LaF~ structure [17], the possible pairs can be characterized by the couple (n-R) when n is the number of sites at a given distance R from a La site. The list of these couples, starting from the lowest R, is Pairs in the same bracket are nearly equivalent but those belonging to classes having n = 2 have a symmetry operator which inter-changes the two ions. This operator is a two-fold axis for the first and the third bracket, an inversion centre for the second bracket. If a dipole-dipole L-749 ENERGY TRANSFER INDUCED BY SUPEREXCHANGE mechanism is supposed for the cross relaxation, the rates would vary along the R 6 following normalized sequence : 1-0.81-0.75-0.64-0.10-0.089 and if a dipole-quadrupole transfer mechanism is supposed, the sequence would be R ‘ 8 i.e. : 1-0.75-0.68-0.55-0.048-0.040. Even if a satellite would correspond to more than one class of pairs, it is clear that these two sequences are incompatible with the sequence of the observed rates. This is the first argument for rejection of dipoledipole and dipole-quadrupole mechanism. ’ 1.2 ORDER OF MAGNITUDE. From the energy levels of LaF3 :Pr3+ [9], the more probable quenching processes for the 3p 0 fluorescence are : the former being assisted by the emission of an energy of 265 cm-1 as phonons, the latter needing the absorption of a phonon of 71 cm 1. At low temperature, only the former is active. From the general theory developed by Holstein et al [ 18] it is easy to show that the transfer rate takes the form