1. Introduction

Er-doped silica-based materials are of strategic interest in the field of optical communication technology for their application as active elements in photonic devices. In particular, this is related to the sharp room temperature luminescent emission at 1540 nm of the Er³⁺ ions that matches the window of minimum loss in silica optical fibers. Nonetheless, the major limitation to the realization of efficient Er-based optical devices is given by the small cross-section for Er³⁺ excitation (typically of the order of 10⁻²⁰ cm²) [1–3]. To improve the rare-earth excitation efficiency, the interaction with fluorescence sensitizers, such as other rare-earth ions [4, 5], Si nanostructures [6–9] or metallic species [10–15] has been exploited.

As regards metal nanoparticles, the sensitization effect for the Er³⁺ luminescence has been recently demonstrated for ultra-small Au nanoclusters (made by less than 20 atoms), incorporated in Er-doped silica matrices [12, 15]. In such experiments, ion implantation has been used to introduce both erbium and subsequently gold in the silica matrix. Post-implantation thermal treatments have been performed to recover the implantation-induced damage, to activate the Er luminescence and to promote the Au nanocluster formation.

An important point that is still open concerns the possible synergistic role of the defects induced by the subsequent Au implantation on the Er luminescence. Indeed, when the ion enters the substrate a wealth of complex defective and/or charged configurations are generated along the collisional cascade [16]. The physics and thermodynamics of such defects and their interactions under thermal annealing can largely influence the functional properties of the system. In particular, they could in principle affect both the rare-earth sensitization process and its luminescence quenching. The aim of the present work is therefore to investigate this aspect, by decoupling the sensitization mechanism triggered by the molecule-like Au clusters from the defects creation during Au implantation. For this reason, Er and Xe co-implantations have been performed following the same approach used to produce the Er-Au co-implanted samples investigated in Refs. [12, 15]. Xe has been chosen as co-dopant in order to release to the silica matrix an implantation damage as similar as possible to that induced by Au implantation. The Er³⁺ emission properties of the Er-Xe co-implanted samples have been characterized by integrated and time-resolved photoluminescence (PL) measurements and correlated to structural characterizations of the local environment around the Er centers obtained by extended x-ray absorption fine structure (EXAFS) spectroscopy.

2. Experimental

To prepare the samples, silica slabs (Herasil 1 by Heraeus) have been sequentially implanted with Er ions at three different energies (50, 100 and 190 keV), with a total fluence of (6.6±0.9)×10¹⁴ Er⁺/cm², as measured by Rutherford Backscattering Spectrometry (RBS) using a 2 MeV ⁴He⁺ beam. In this way, an almost flat, 70 nm thick, Er profile has been obtained with an Er concentration of about 10²⁰ Er/cm³ (that is below the threshold for concentration quenching of the Er emission [2]). After the Er implantation, thermal annealings at different temperatures in the range 300–1100 °C have been performed for 1h in N₂ atmosphere. The Er-implanted slab annealed at 800 °C has been subsequently implanted with Xe ions following the same triple energy scheme (50, 90 and 150 keV) to match the Er profile, at two different fluences, (7.9±0.9)×10¹⁴ Xe⁺/cm² and (7.8±0.9)×10¹⁵ Xe⁺/cm², respectively, as measured by RBS. The highest Xe⁺ fluence was chosen to have an expected implantation damage on the Er-implanted silica similar to the one induced by the Au⁺ implantation on the same system as in Ref. [12]. In that work the measured Au/Er concentration ratio was about 9. According to simulations obtained with the SRIM-2010 code [17], we evaluated that at the energies used in the present work the nuclear and electronic energy losses of Xe⁺ in silica are about 30% lower than the corresponding quantities for Au⁺ in silica. Therefore to better reproduce the implantation-induced damage we used a Xe⁺ fluence about 30% higher with respect to the Au⁺ one, resulting in a Xe/Er ratio of about 12. On the other hand, the lowest investigated Xe⁺ concentration allowed to study the effect of an order of magnitude lower implantation damage on the Er³⁺ photoluminescence. Thus, the samples were labeled according to the level of damage induced by the Xe or Au implantations, i.e., HD (high damage) corresponds to a concentration ratio Er/Xe=1/12 (comparable to Er/Au=1/9) and LD (low damage) corresponds to Er/Xe=1/1.2.

Post Xe-implantation thermal treatments have been carried out for 1h in N₂ atmosphere in the temperature range 300–800 °C. The Er³⁺ emission properties at 1540 nm have been investigated by integrated and time-resolved photoluminescence measurements performed both with resonant and not-resonant excitation, using, respectively, the 488 nm and 476.5 nm lines of a cw Ar laser mechanically chopped at 5 Hz.

The morphological and compositional characterization of the samples was performed in cross-section by transmission electron microscopy (TEM) with a FEI TECNAI F20 (S)TEM field-emission gun (FEG) microscope operating at 200 keV, equipped with an EDAX energy dispersive x-ray spectrometer (EDS) for compositional analysis and a scanning TEM (STEM) attachment with a high-angle annular detector for dark field STEM imaging (HAADF-STEM).

Er L₃-edge x-ray absorption spectra were recorded at the Italian beamline GILDA of the European Synchrotron Radiation Facility-ESRF (F) in fluorescence mode. The double crystal monochromator was equipped with two (111) Si crystals; the harmonics rejection was achieved by a couple of Pd-coated mirrors working at an incidence angle of 3.6 mrad. A grazing incidence geometry was used (incidence angle ∼2 deg) to increase the (low) fluorescence signal from Er, keeping at the same time the signal of the elastic scattering from the matrix as low as possible [18]. The x-ray absorption spectrum of Er₂O₃ crystalline powder was also measured in transmission mode as standard reference. The EXAFS analysis was performed with the FEFF8-FEFFIT package [19]: the Er-O and Er-Si backscattering amplitudes and phase shifts were calculated for Er₂Si₂O₇ and ErSi₂ clusters. The ${\text{S}}_0^2$ parameter of the standard EXAFS formula [20] was estimated by fitting the spectrum of the crystalline Er₂O₃. k²-weighted EXAFS spectra were Fourier-transformed in the range k=2–8 Å⁻¹ and the analysis was performed in the R-space. Since the first and the second coordination shells give rise to signals that partially overlap into the R-space (especially for low annealing temperature), the analysis was carried out in the R=1–4 Å range, i.e. considering altogether both the first and the second shell contributions, that are practically the whole EXAFS signal; the fit was based on the single scattering approximation; the use of multiple scattering paths did not lead to a significant improvement of the fit quality. As far as the first shell is concerned, as discussed in the following, in most of the cases a simple fit based on a single Er–O distance failed and an additional Er–Si coordination was needed to reasonably reproduce the experimental data. To fit the second shell we considered that O and Si atoms in the first shell are likely bonded to the silica matrix, i.e. to Si and O atoms, respectively; for this reason in the second coordination shell we have considered again a double population of O and Si atoms. We remark that for all the samples the main signal is due to O atoms in the first shell and Si atoms in the second one, that form a sort of Er silicate-like complexes. To reduce the number of fitting parameters, for the minor components of the signal, i.e. Si atoms in the first shell (Si^I) and O atoms in the second one (O^II), the ratio of the coordination numbers CN of the two shells CN(Si^I)/CN(O^II) was set the same for all the spectra. For the same reason, the same Debye-Waller factor was attributed to the two coordinations in the first shell (similarly for the second shell), estimated by the first shell fitting of the spectra from the 800 °C-annealed samples; this value (0.013±0.002 Ų) is pretty similar to the one obtained with full EXAFS spectrum analysis of the Er site in crystalline silicon [21]. In all cases the degrees of freedom in the fit were 4–5. A similar analysis performed on k¹- or k³-weighted EXAFS spectra resulted less accurate but essentially in agreement with the first shell data presented in the following.

3. Results and discussion

In order to evaluate the effect of Xe implantation on the Er³⁺ emission properties, both the PL intensity at 1540 nm and the lifetime of the Er³⁺ emission of the Er+Xe co-implanted samples have been measured as a function of the annealing temperature and compared to the case without Xe implantation (i.e., Er implantation only). The results are shown in Fig. 1. For all the samples, the PL emission exhibited a single exponential decay which has been fitted to determine the lifetimes reported in the inset.

 

Fig. 1 PL intensity as a function of annealing temperature for the Er-implanted (white triangles) and the Er+Xe co-implanted samples (black squares for the Er/Xe LD and half-black circles for the HD samples, respectively). The measurements have been performed with resonant excitation at 488 nm. The inset shows the corresponding emission lifetimes. The legend is the same as for the PL intensity.

Download Full Size | PDF

The evolution with the annealing temperature of the PL intensity of the samples implanted with Er only is reported in Fig. 1 as open triangles. In this case, no PL signal can be detected (within the sensitivity of our set-up) in the as-implanted and low temperature annealed samples. The onset of detectable PL emission occurs at temperatures above 600 °C and it starts to increase reaching a maximum value at 900 °C. For annealing at 1000 °C a further decrease of the PL intensity is observed. Correspondingly, the PL emission lifetimes (see inset) are characterized by a progressive increase in the whole temperature range, reaching values of about 12–14 ms which are typical of properly coordinated Er³⁺ emitting centers [2]. The onset of PL emission above 600 °C and the concomitant increase of PL intensity and lifetime in the 600–900 °C temperature range can be related to (i) a progressive recovery of the implant-induced defects in the silica matrix and (ii) to the formation of the full octahedral coordination of six oxygen atoms around the emitting Er³⁺ ion. Indeed, from one side, irradiation defects as dangling bonds, vacancy complexes or bond distortions in the glass [22] may act as nonradiative recombination centers giving rise to nonradiative decay processes involving energy transfer from the Er³⁺ ion towards acceptor states in the host and quenching the luminescent emission. On the other side, an incomplete oxygen shell (i.e., under-coordinated Er³⁺ ion) due to Er-O bond breaking and/or oxygen displacement under the collisional cascade can prevent a proper PL emission. After that the maximum PL intensity and lifetime is attained, the subsequent decrease of the PL signal observed at 1000 °C can be related to the onset of Er clustering phenomena that decrease the concentration of Er³⁺ active centers. Such clustering effect does not influence the lifetime that sightly increases due to a further recovery of the matrix defects.

As mentioned in the Experimental section, Xe implantations have been performed on silica slabs previously implanted with Er and annealed for 1h in N₂ atmosphere at 800 °C. For this annealing condition the Er-implanted samples exhibit a PL intensity close to its maximum value. Nonetheless, after Xe implantation at both investigated fluences such luminescent emission is completely suppressed and annealing treatments at temperatures above 600 °C are necessary to restore it. Moreover, the annealing evolution of the PL emission after Xe implantation at both fluences is comparable with the one measured from the Er-implanted samples, as shown in Fig. 1. An analogous behavior is observed also for the emission lifetimes (see inset in Fig. 1). In this case, only the samples annealed at the highest temperatures, that is 700 °C and 800 °C, exhibited a sufficiently intense luminescent emission to provide reliable estimations of the lifetimes from time-resolved PL measurements. For both fluences the measured lifetimes are close to those obtained in the Er-implanted samples. Only the Er/Xe LD co-implanted sample annealed at 700 °C exhibited a slightly longer lifetime with respect to the corresponding Er-implanted one, which can be ascribed to a reduced level of damage.

As a comparison, in Fig. 2 we have reported the PL intensity as a function of the annealing temperature for the Er/Xe HD and the Er/Au HD co-implanted samples under resonant excitation (488 nm). The results show the dramatically different temperature evolution exhibited by the Er+Au co-implanted samples with respect to that of Er+Xe co-implanted ones. In fact, in case of Er/Au HD co-implanted silica, the PL intensity is enhanced by the formation of very small Au aggregates (occurring up to T = 600 °C) that transfer part of the absorbed energy to the Er ions with a size-dependent efficiency (3–20 atoms clusters being the most efficient), as we showed in Refs. [12, 15]. Moreover, when the population of these efficient Au sensitizers is over (due to cluster growth, T>600 °C) the PL intensity decreases to values comparable to those of Er/Xe HD samples, in which the energy transfer process does not occur, as we will show below. This last result in particular allows to rule out the contribution of the implantation defects as the main source of energy transfer to Er³⁺ ions.

 

Fig. 2 Comparison of the temperature evolution of the PL intensity for the Er/Xe HD and Er/Au HD co-implanted samples under resonant excitation.

Download Full Size | PDF

Indeed to get deeper information on the luminescent behavior of Er+Xe co-implanted samples, PL measurements have been also performed in out-of-resonance excitation condition (pumping with the 476.5 nm line of the Ar laser). The results are reported in Fig. 3(a), which shows the PL emission spectra around 1540 nm, both with resonant and not-resonant excitation, of the Er/Xe LD and Er/Xe HD co-implanted samples annealed at 800 °C. For both pumping conditions, the spectra of the Er-implanted sample annealed at 800 °C (that is the reference sample before Xe implantations) resulted very similar to those of the Er/Xe LD and HD co-implanted samples (annealed at the same temperature). Control measurements have been performed for all the samples implanted with Xe at both fluences and annealed at the different temperatures. With out-of-resonance excitation, no evidence of 1540 nm luminescent emission has been obtained independently of the annealing conditions and the Xe implantation fluences, i.e., independently of the implantation-damage level in the samples. This is a conclusive evidence of the absence of any energy transfer process (at the explored excitation wavelengths) in the Xe-implanted samples, contrary to what was found for the Au-implanted samples. Therefore, the present result further supports the conclusion that molecule-like Au clusters are required to achieve a broadband energy transfer to Er³⁺ ions [12, 15].

 

Fig. 3 (a) in-resonance (488 nm) and out-of-resonance (476.5 nm) Er³⁺ PL emission spectra around 1540 nm of the Er/Xe HD and LD samples annealed at 800 °C; (b) cross-sectional HAADF-STEM image of the Er/Xe HD sample annealed at 800 °C; (c) BF-TEM cross section of the Er/Au HD sample annealed at 800°C; (d) comparison of the optical absorption spectra of the samples annealed at 800°C.

Download Full Size | PDF

We would like to stress the structural and morphological similarity between Xe-implanted and Au-implanted Er-codoped silica. Figure 3(b) shows the cross-section image of the Er/Xe HD sample annealed at 800 °C taken in STEM high-angle annular dark field (HAADF-STEM) mode. In this technique the image contrast is proportional to the atomic number therefore the bright spots below the surface can be interpreted as Xe clusters (bubbles) with size of about 2 nm. No lattice fringes can be obtained in high-resolution mode (HR-TEM) nor diffraction rings in electron diffraction. This is an indication that the Xe aggregates are amorphous clusters or gas bubbles. Indeed, large Xe bubbles can be formed in silica under high fluence Xe implantation [23], e.g., above 10¹⁶ ions/cm². Their formation is generally attributed to the low solubility of noble gas atoms which therefore tend to segregate in the damaged silica. This relatively low solubility of Xe is also typical of Au in silica. For comparison, Fig. 3(c) shows the bright-field (BF-TEM) cross section image of the sample Er/Au HD annealed at the same temperature. Crystalline spherical Au nanoclusters were detected with size 1.6 ± 0.5 nm, i.e., very similar to that of the Xe bubbles of the Er/Xe HD sample.

Xenon in form of dispersed atoms or as a neutral cluster is not expected to have specific absorption or emission bands in the VIS-NIR region [24]. On the contrary, charged ${\text{Xe}}_n^{+}$ clusters can have (size-dependent) absorption bands close to 1.5 eV, 2 eV and 3 eV, as shown for n ≤ 30 free cationic clusters [24]. We did not find evidence of Xe-related absorption bands, in particular close to the excitation wavelengths used in the present works (476.5 nm and 488 nm): in Fig. 3(d) a comparison between the absorption spectra of the Er/Xe HD and Er/Au HD annealed at 800°C is reported. While the Er/Xe HD sample exhibits an absorption spectrum very similar to the pure Er reference sample, in the Er/Au HD the onset of the interband transitions absorption and of the surface plasmon resonance of small Au nanocluster is clearly visible. At lower annealing temperatures, Er/Xe HD samples do not exhibit major differences with respect to the optical absorption of the 800 °C-annealed ones.

Despite the close similarity between Er/Xe HD and Er/Au HD samples in terms of both damage produced and clustering of the implanted species, the effect on the energy transfer process is dramatically different as shown in Fig. 2. Moreover, the PL results indicate that the defects induced in the silica matrix by the Xe implantations behave as nonradiative de-excitation channels for the Er³⁺ emission as well as those produced by the Er implantation itself. Similar post-implantation treatments are needed to restore the luminescent properties of the samples, independently of the implantation fluence, at least in the range explored in this work.

To investigate the nature of the implantation induced damage on the Er site, we performed EXAFS analysis at the Er absorption edge to correlate the above PL results with the evolution of the local atomic structure around Er ions. The EXAFS spectra of the Er/Xe HD co-implanted samples are reported in Fig. 4(a) and compared with the spectrum of an Er-implanted silica annealed at 800°C; in Fig. 4(b) the Fourier transform (FT) moduli are shown together with the best fit curves. For all the samples, the EXAFS signal (Fig. 4(a)) -damped at high k values-is typical of a light backscatterer, i.e., mainly oxygen in this case. The EXAFS spectra show some differences in the k=5.5–8 Å⁻¹ region: in particular, in this range the signal of the as-implanted sample is very low, while it progressively increases at higher annealing temperatures. Correspondingly, the main peak of the Fourier transform modulus related to the Er-O coordination (Fig. 4(b)) is highly asymmetric for the as-implanted sample, while the shoulder on its right-hand side (at R≈2–2.5 Å) progressively fades at higher temperature leading to an almost symmetric peak upon annealing at 800 °C. These observations can be quantified only if a full analysis extended up to at least k=8.0 Å⁻¹ is performed and for relatively low-noise data.

 

Fig. 4 (a) EXAFS spectra of the Er/Xe HD co-implanted samples, compared to the spectrum of an Er-implanted silica annealed at 800 °C; (b) k²-weighted Fourier transform (in the range k=2–8 Å⁻¹) moduli (markers) and fits (solid line) of the spectra in (a); in (a,b) the annealing temperature is reported. (c,d) Fourier-filtered (in the range R=1–4 Å) signal (markers) superimposed to the best fit curve (solid line) for the as-implanted (c) and 800 °C-annealed samples (d); the signals that contribute to the fit are also reported (O in the first shell + Si in the second one and Si in the first shell + O in the second one).

Download Full Size | PDF

To properly fit the first shell EXAFS signal, including the shoulder discussed above, two different coordinations have to be considered: in fact, the use of an asymmetric distance distribution alone failed to reproduce the experimental data. The shorter coordination is the dominant signal and comes from the nearest O shell; the second one is responsible for the right-hand side shoulder of the first peak in the Fourier transform moduli, and is related to a light backscatterer, such as Si and/or O. This signal is a minor part of the whole first shell contribution to the EXAFS spectrum, so the data analysis alone could not lead to strong conclusive arguments about the nature of the backscatterer (O or Si, or even a coexistence of both). Nevertheless, we remark that: (i) this second contribution to the first shell is thermodynamically unstable and vanishes at high annealing temperature and (ii) the ion implantation process determines a depletion of O atoms in the implanted region, so that the formation of dopant-Si bonds is favored. For these reasons, this contribution to the first shell is more likely due to Si atoms. Similarly to this case, it has been shown that the local site of Er ions co-implanted with O in crystalline Si exhibit a main Er–Si coordination that is progressively substituted by the more stable Er–O coordination upon annealing [25] (as a comparison, the heat of formation ΔH_f of Er₂O₃ is 1898 kJ/mol [26], to be compared with ∼85 and ∼63 kJ/mol of ErSi and ErSi₂, respectively [27]). Another fact that supports this picture is that a first attempt to fit the first shell experimental data with O atoms at two different distances led to a Debye-Waller factor higher for the spectra of the samples annealed at higher temperature, that is unlikely to happen.

About the second shell, it is known that Er atoms bonded to O atoms embedded into silicon, silicon oxide and silica-based matrix exhibit a second shell coordination of Si atoms of the matrix [21, 28, 29]; in the present case, we consider that O atoms (Si atoms) in the first shell are likely bonded to Si atoms (O atoms) of the matrix, with the result that the second coordination shell (signal in the range R = 2.7–4 Å) is formed of both Si and O backscatters, connected to Er ions via O and Si species, respectively. In Fig. 4(c) and 4(d) data and fit are shown in the Fourier-filtered k-space for the as-implanted and 800 °C-annealed samples, respectively; the main signal from O in the first shell (+ Si in the second one, Er–O^I+Er–Si^II) and the minor contribution from Si in the first shell (+ O in the second one, Er–Si^I+Er–O^II) are reported in the two cases: it is evident that the contribution from Er-Si^I+Er–O^II coordination becomes negligible upon high temperature annealing.

For the first shell, the results of the EXAFS analysis are reported in Table 1. For the as-implanted Er/Xe HD sample, the average Er site is composed of about 5 O atoms and about 2 Si atoms. The Er–O distance is significantly shorter than in the Er₂O₃ and Er₂Si₂O₇ crystals (where R = 2.23–2.31 Å), but is in the range of distances usually observed in Er-doped silica-based systems [28, 30, 31] especially when the preparation conditions are far from equilibrium. Moreover, the Er–Si distance is significantly shorter than in Er-silicide crystals (that is about 3.00 Å in ErSi₂, 2.80–3.17 Å in ErSi); nevertheless, theoretical analysis on Er+O-doped crystalline Si indicates Er–O and Er–Si first shell distances of 2.18 Å and 2.60 Å, respectively, similar to the value found in the present case [32]. Upon annealing, the number of O atoms in the first shell slightly increases, and Er ions reach practically the full octahedral coordination, i.e. the standard six-fold coordination of Er³⁺ in crystals, upon annealing at T = 800 °C; correspondingly, the number Si atoms in the first shell significantly decreases and becomes negligible upon annealing at 800 °C (see Figs. 4(d) and 5). So, the presence of both Si and O atoms in the first shell, observed upon annealing at T < 800 °C likely suggests that the electronic configuration around the Er site may be different from that of Er ions octahedrally coordinated with just O atoms.

Table 1. Results of the EXAFS analysis for the first shell for Er and Er/Xe HD samples; N is the coordination number, R the interatomic distance. The Debye Waller factor σ² has been set to (13 ± 2) ×10⁻³Å² for all the samples and for both O- and Si-coordination, as estimated by the first shell fitting of the spectra from the 800 °C-annealed samples

View Table

As far as the signal from the second coordination is concerned, the analysis indicates that, on the average, the second shell of atoms surrounding Er ions is composed of 3–5 Si atoms and a number of O atoms that, being linked to the first Si coordination (see details in the Experimental section), from about 3 progressively vanishes by increasing the annealing temperature. In all cases the Er–Si^II distance is 3.7 ±0.1 Å, similar to what found in Er+O-implanted crystalline silicon [21], while the Er–O^II distance is 3.1±0.1 Å. Moreover, the EXAFS signals of the two samples annealed at 800 °C, i.e. with and without Xe implantation, are identical (see Fig. 4(a) and Table 1). These results indicate that the Er-Si coordination in the first shell is related to the defects formed upon Xe implantation: at high temperature annealing the matrix and the Er site recover from radiation damage and, correspondingly, the Er–Si coordination vanishes and only the octahedral coordination with O atoms remains.

It is worth noting that the EXAFS results for the Er/Xe LD as-implanted sample are very similar to those of the Er/Xe HD as-implanted sample indicating that, within the EXAFS sensitivity, the Er site does not depend on the Xe implantation fluence, at least for the fluence range explored in this work.

In Fig. 5 the thermal evolution of the Er–O^I and Er–Si^I coordination numbers in the first shell are plotted together with the PL intensity for the Er/Xe HD sample. A good direct correlation between PL efficiency and the Er–O^I is shown, indicating that the main contribution to the PL intensity comes from a properly structured Er³⁺ environment, i.e., when the fully octahedral configuration of the oxygen atoms around each erbium ion is reconstructed by the thermal annealing, with the consequent fading of the Er–Si^I contribution, the Xe-implanted Er-doped silica recoveries a PL efficiency (both intensity and lifetime) very similar to the nonimplanted Er reference sample.

 

Fig. 5 First shell coordination number for the Er–O^I and Er–Si^I components as a function of the annealing temperature in the sample Er/Xe HD (left axis). On the right axis, the evolution of the Er/Xe HD photoluminescence intensity I_PL.

Download Full Size | PDF

4. Conclusion

We investigated the thermal evolution of the Er³⁺ PL emission in silica under the influence of implantation with Xe ions at different fluences. In particular we found that Xe implantation is remarkably different with respect to an equivalent Au implantation that is carried out so as to release approximately the same level of implantation damage. In particular, Xe is not able to produce energy transfer to Er³⁺ under non-resonant excitation, at variance with the Au case, despite the close similarity in terms of the size of the precipitates evidenced by TEM under Xe-or Au-implanted samples. This result clearly demonstrates that the energy transfer responsible for out-of-resonance excitation of Er³⁺ cannot be ascribed to implantation induced defects but is actively triggered by ultra-small molecule-like Au clusters, which absorb the non-resonant excitation and efficiently transfer it to Er³⁺ emitting centers.

To confirm this picture, we followed the evolution of the Xe implantation-induced damage around the Er³⁺ site in silica. EXAFS analysis demonstrated the presence of two different short-range coordinations in the Er site: one Er–O coordination at R = 2.14–2.18 Å, that is in the typical range of distance for Er³⁺ in silica, and one longer coordination with Si atoms at R = 2.7–2.8 Å. In particular the Er–Si coordination progressively fades as the annealing temperature is raised. A second coordination shell indicates that O and Si atoms in the first shell are further bonded with the matrix network. The monotonic increase of the PL intensity as a function of the annealing temperature in the Xe-implanted samples is found to be correlated to the increase of the Er–O coordination number and to the decrease of the Er–Si coordination, suggesting that these two configurations greatly affect the Er emission efficiency recovery.