1. Introduction

Erbium-doped silicon-based materials are very important for silicon (Si) photonics due to their unique properties [1–6]. Erbium (Er) ions can emit at 1.54 µm in nearly all kinds of matrices, which corresponds to a minimum in the loss spectrum of silica optical fibers. However, the Er concentration in most host materials is limited to about 10²⁰ cm^-3 due to the low solubility of Er, leading to a small optical gain [7]. Er compounds, such as oxides or silicates, in which Er is a major component rather than a dopant, can have an Er concentration as high as 10²² cm^-3 [8–11]. Furthermore, it has been reported that all Er ions are optically active in Er silicates [8]. Therefore, Er compounds, especially Er silicates, have been attracting much attention. Wang et al. optimized the ytterbium (Y) addition concentration in Er_xY_2-xSiO₅ and theoretically calculated the optical gain at 1 mm length Er_xY_2-xSiO₅ (x = 0.1) waveguide to be above 10 dB [12]. Ning et al. improved the crystal quality of single-crystal erbium chloride silicate nanowires and obtained a net optical gain exceeding 100 dB cm^-1 [9]. Although no optical gain has been reported yet, Er₂Si₂O₇ is one of the best candidates for optical gain based on the lifetime-density products [13]. As the same with other rare earth disilicates, Er₂Si₂O₇ shows various polymorphs, named y, α, β, and γ, etc [14,15]. Previous research has shown that the luminescent properties of the polymorphs can be quite different [16,17]. Therefore, it is important to investigate the luminescent properties of different polymorphs and, more importantly, how to control the formation of these polymorphs, which requires a full understanding of its influencing factors. The phase transformation of Er₂Si₂O₇ synthesized by solid state reaction has been investigated, but the formation conditions are quite different from those for deposited Er-Si-O films [18].

In this paper, we report a detailed study of the influences of annealing temperatures, annealing atmospheres, and chemical compositions on the phase transformation of Er silicates formed in Er-Si-O films and the luminescent properties of different polymorphs. The correlation between the structures and the luminescent properties of different polymorphs will also be demonstrated.

2. Experiment

Er-doped silicon oxide films with thicknesses of about 110 nm were deposited on rotating p-Si (100) substrates heated at 300 °C by reactive magnetron co-sputtering using Si and Er targets in a reactive atmosphere of Ar and O₂ mixed gas. The film compositions were obtained by changing the radio frequency (RF) power applied to the Er target from 70 W to 135 W and by keeping constant the RF power applied to the Si target at 130 W. After deposition, the films were thermally treated for 1 h in flowing nitrogen or oxygen, at temperatures ranging between 900 °C and 1150 °C to form erbium silicates.

Rutherford backscattering spectrometry (RBS) measurements were carried out with a 2.02 MeV ⁴He ion beam at a scattering angle of 165°. X-ray diffraction (XRD) analyses were performed on a Japan Rigaku D/max-2550pc X-ray diffractometer by Cu Kα radiation with the grazing angle of 0.5° or 1°. Raman spectra were obtained with a Bruker SENTERRA compact Raman microscope. The excitation wavelength was 532 nm and the excitation power was 20 mW. Transmission electron microscopy (TEM) images including high resolution TEM (HRTEM) images were carried out with a Tecnai F20G2 microscope. An Edinburgh Instruments FLS-920 fluorescence spectrophotometer, with a 980 nm laser as the excitation source, was employed in the steady-state photoluminescence (PL) measurements.

3. Results and discussion

Table 1 displays the chemical compositions measured by RBS of the as-deposited films deposited with different RF powers. The films are labeled as sample 1 to 5. Uniform concentrations of Er, Si and O have been detected along the film thickness. The Er content increases with the RF powers applied to the Er target. It should be noted that the O content also changes with the RF power of Er target because the change of the RF power is very likely to affect the kinetics of the deposition process leading to some fluctuation of oxygen content. Besides, the fluctuation of the instrument can also result in slight differences in oxygen content. In sample 3, the Er:Si:O atomic ratio is close to that in Er₂Si₂O₇ except for an excess of O. In sample 4, the Er:Si ratio is 1.8 : 1, which is similar to Er₂SiO₅. The phase compositions of the samples treated with different annealing procedures are also listed in Table 1, which will be discussed in detail.

Table 1. Chemical compositions of the as-deposited films and the XRD results for the samples annealed under different conditions

View Table | View all tables in this article

The XRD spectra of the samples after different thermal treatments are presented in Fig. 1. The spectra of all the samples annealed at 900 °C as well as sample 1 annealed at 1000 °C and 1150 °C show no diffraction peaks (the spectrum of sample 1 annealed at 1150 °C is shown in Fig. 1 as an example), demonstrating the amorphous nature of the films. Crystalline structures start to appear above the annealing temperature of 1000 °C except for sample 1. The variation of the diffraction peaks in these samples indicates that different Er silicate polymorphs have formed. The spectra of sample 4 and sample 5 annealed at 1000 °C in N₂ correspond to monoclinic X1-Er₂SiO₅, space group P2₁/c (Joint Committee for Powder Diffraction Standards (JCPDS) file number 70-3279). And the spectra of sample 3 and sample 4 annealed at 1150 °C in N₂ correspond to monoclinic β-Er₂Si₂O₇, space group C2/m (JCPDS file number 89-1291). Apart from these spectra, the others do not fully correspond to any of the known Er silicates from a comparison between their XRD patterns and literature data about Er₂Si₂O₇ and Er₂SiO_5. However, the similar ionic radius of Er³⁺, Y³⁺, and Tm³⁺ (0.89 Å and 0.90 Å for Er³⁺ and Y³⁺ in 6-fold coordination, and 0.99 Å and 1.00 Å for Tm³⁺ and Er³⁺ in 8-fold coordination, respectively [19]) makes it reasonable to use y-Y₂Si₂O₇ (space group P2₁/m, JCPDS file number 74-1994) and α-Tm₂Si₂O₇ (space group P-1, JCPDS file number 31-1994) as references. So the peaks at 19.0°, 22.0°, 29.3°, 37.6°, 38.4°, 44.2°, and 44.9° are attributed to monoclinic y-Er_­2Si₂O₇ and the peaks at 13.5°, 14.8°, 20.4°, 21.3°, 27.2°, 30.0°, 30.9°, 31.4°, 32.2°, 33.3°, 35.5°, 41.3°, 42.6°, 43.2°, and 44.1° are attributed to triclinic α-Er₂Si₂O₇.

 

Fig. 1. XRD spectra of sample 1 annealed at 1150 °C in N₂, sample 2 to sample 5 annealed at 1000 °C and 1150 °C in N₂ and sample 3 and sample 4 annealed at 1150 °C in O₂.

Download Full Size | PDF

As presented in Table 1, no Er silicates were obtained in sample 1, indicating that the formation of Er silicates requires higher Er concentrations. All the other films remain amorphous at 900 °C and have been crystallized at 1000 °C, demonstrating that the crystallization temperature of Er silicates in our films lies between 900 °C and 1000 °C. As reported, the temperature stabilities of Er₂Si₂O₇, in ascending order, are y-Er₂Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇. Therefore, the annealing at 1150 °C in N₂ turns most of the α-Er_­2Si₂O₇ and y-Er_­2Si₂O₇ into β-Er₂Si₂O₇ in sample 2 and sample 3. When the Er:Si atomic ratio is close to or above 2:1 (sample 4 and sample 5), X1-Er₂SiO₅ will form at 1000 °C. X1-Er₂SiO₅ is not stable at the annealing temperature of 1150 °C and consequently, turns into Er₂Si₂O₇. It should be noted that the transformation from Er₂SiO₅ to Er₂Si₂O₇ requires excess Si. However, it seems hard to be accomplished according to the chemical compositions of sample 4 and sample 5. This problem will be discussed later. All the Er silicates in samples 2, 3, 4, 5 have turned into Er₂Si₂O₇ after the annealing at 1150 °C in N₂ except that the containing polymorphs are different. Sample 2 and sample 5 are composed of a mixture of α-Er_­2Si₂O₇ and β-Er₂Si₂O₇ while only β-Er₂Si₂O₇ was formed in sample 3 and sample 4. This is mainly attributed to the different chemical compositions of the samples since the annealing temperature remains the same. As shown in the RBS results, the deviation from stoichiometric Er silicates is very large for sample 2 and sample 5. We argue that the excess of Si or Er will restrain the transformation from low-temperature phase to high-temperature phase, which leads to the remaining of α-Er_­2Si₂O₇ in sample 2 and sample 5. In addition to annealing temperature and chemical composition, annealing atmosphere also has a great influence on the formation of Er silicates. The phase compositions of sample 3 (y-Er_­2Si₂O₇ and α-Er_­2Si₂O₇) and sample 4 (X1-Er₂SiO₅ and α-Er_­2Si₂O₇) annealed in O₂ are quite different from those annealed in N₂ at the same annealing temperature of 1150 °C. It indicates that low-temperature phases tend to form during annealing in O₂. In other words, the phase transformation temperature is increased. It is worth noting that the phase composition of sample 4 annealed at 1150 °C in O₂ reveals that X1-Er₂SiO₅ will directly transform to α-Er_­2Si₂O₇ with increasing temperature without y-Er_­2Si₂O₇ as the intermediate phase. Since some of the samples consist of two polymorphs, it is important to obtain the proportion of each polymorph for the following analysis. However, this task is hard to accomplish by semi-quantitative XRD analysis due to the lack of reference intensity ratio (RIR) for the α-polymorph and the large error caused by the orientation of the Er silicate crystallites in our films. Therefore, we utilized another method to roughly calculate these data, which will be introduced later.

Figure 2(a) shows the cross-sectional TEM image of sample 1 annealed at 1150 °C in N₂ in which many large Er-rich clusters are observed. On the one hand, the Er concentration of this sample is relatively low, so crystallization of Er silicate is unable to occur. On the other hand, the concentration of Er has exceeded the solid solubility limit of Er in SiO₂. Therefore, Er ions accumulate to form Er-rich clusters and these clusters keep growing through Oswald ripening with annealing time prolonged. The SAED (selected area electron diffraction) pattern of this sample is shown in Fig. 2(b). No crystalline structures can be observed in the film, demonstrating that the Er concentrations in the Er-rich clusters are still not high enough for the crystallization of Er silicates. Figure 2(c) shows the cross-sectional TEM image of sample 3 annealed at 1150 °C in N₂. No polycrystalline structure can be observed in the image, indicating that the grain size is very large in the sample. The grain size was estimated to be at least tens of microns because no sign of another grain can be observed in the TEM specimen prepared by ion milling. Figure 2(d) presents the HR-TEM image and the corresponding SAED pattern of this sample. The lattice spacing of the selected area is 4.63 Å, which corresponds to the (001) plane of β-Er₂Si₂O₇. Figure 2(e) shows the cross-sectional TEM image of sample 5 annealed at 1150 °C in N₂. It is obvious that the crystal quality of this sample is much worse than that of sample 3, and a polycrystalline structure with the average grain size in the order of 100 nm is observed. The HR-TEM image and the corresponding SAED pattern are shown in Fig. 2(f). The lattice spacing is 5.73 Å, which corresponds to the (101) plane of α-Er₂Si₂O₇. The large difference in grain size between sample 3 and sample 5 is due to the different chemical compositions of the films. During annealing, crystalline clusters nucleate homogeneously in the film or heterogeneously on discontinuities. In order to reduce the free energy, clusters with smaller size tend to shrink, while those with sizes larger than the critical size tend to grow. In sample 5, the Er content is much larger than the amount of Er required to form Er₂SiO₅ or Er₂Si₂O₇, so the Er atoms in excess will be ejected during the grain growth. This will lead to an accumulation of Er around the grain and a long-distance diffusion process will be needed to continue the grain growth. Therefore, further growth of the grains will be restrained. Figure 2(g) shows the cross-sectional TEM image of sample 3 annealed at 1150 °C in O₂. The grain size is much smaller than that in sample 3 annealed at the same temperature in N₂. A SiO₂ layer as thick as 140 nm was formed between the silicate layer and the Si substrate. The formation of this SiO₂ layer requires continuous supply of O from the atmosphere and diffusion of O from the silicate layer to the Si substrate. It indicates that the Er-Si-O film is always O-rich during the crystallization of Er silicates, and the O in excess will accumulate at the grain boundary preventing the grain from growing very large. Moreover, the grain boundaries provide diffusion channels for O atoms due to the larger diffusion rate thereof. The HR-TEM image and the corresponding SAED pattern are presented in Fig. 2(h). The lattice spacing is 3.05 Å, which corresponds to the (021) plane of y-Er₂Si₂O₇.

 

Fig. 2. Cross-sectional TEM image (a) and SAED pattern (b) of sample 1 annealed at 1150 °C in N₂. Cross-sectional TEM images and HR-TEM images with SAED patterns as insets of sample 3 annealed at 1150 °C in N₂ (c)(d), sample 5 annealed at 1150 °C in N₂ (e)(f), and sample 3 annealed at 1150 °C in O₂ (g)(h).

Download Full Size | PDF

The accumulation of excess atoms around the Er silicate grains accounts for not only the change of grain size in different samples but also the increase in polymorph transformation temperature when the chemical composition of the film deviates from the stoichiometry of Er silicates or the film is annealed in O₂. These excess atoms generally exist in the form of amorphous SiO₂ or amorphous Er₂O₃. Due to the various crystalline structures, different Er silicate polymorphs show different densities. In other words, the polymorph transformation of Er silicate will result in volume changes. When the Er silicate grains are surrounded by SiO₂ or Er₂O₃ phase, the change of grain volume will introduce stress into the material, which increases the energy barrier for the polymorph transformation of Er silicate. In other words, higher annealing temperatures are required.

It should be noted that a thin SiO₂ layer exists beneath the silicate layer even though the annealing was performed in a nitrogen atmosphere. This can be explained by the chemical compositions of the films. As exhibited in Table 1, the O contents of the films are much larger than the amount of O required to form Er silicates. The O atoms in excess will be ejected and diffuse to the interface to react with the Si substrate during annealing. This also explains the question raised above why Er₂Si₂O₇ is formed at 1150 °C in sample 4 and sample 5 despite the lack of Si. It is because the oxidized Si substrate can react with Er₂SiO₅ to form Er₂Si₂O₇.

In the above discussion, we only considered the Er silicate phase. However, other phases, such as SiO₂ or Er₂O₃, should also exist in sample 2 and sample 5 because the chemical compositions of these two samples seriously deviate from the stoichiometry of Er silicates. As demonstrated above, the O excess in the films will react with the Si substrate forming a SiO₂ layer. Without the consideration of this SiO₂ layer, the upper film can be regarded as a SiO₂+Er₂O₃ system and the phases formed can be determined through the phase diagram of the SiO₂+Er₂O₃ system [20]. For sample 2, The Er₂O₃:SiO₂ ratio in the upper film is 1:3.5, so the phases formed are Er₂Si₂O₇ and amorphous SiO₂. For sample 5, the Er₂O₃:SiO₂ ratio is 3:2, which correspond to a mixture of Er₂SiO₅ and Er₂O₃ in the phase diagram. When the sample is annealed at 1000 °C, the XRD spectrum shows no other phase except X1-Er₂SiO₅ demonstrating that the remaining Er₂O₃ phase is amorphous. When the sample is annealed at 1150 °C, the SiO₂ layer cannot be neglected because it will react with the Er₂SiO₅ phase to form Er₂Si₂O₇ and the Er₂O₃:SiO₂ ratio can no longer be considered as 3:2. Assuming that all the oxygen excess has reacted with the Si substrate, the Er₂O₃:SiO₂ ratio becomes 3:4. The Si content is still not high enough to make all the Er atoms exist in the form of Er₂Si₂O₇. Besides, a thin SiO₂ layer can still be observed in Fig. 2(e) demonstrating that the SiO₂ layer has not been consumed completely. However, the XRD spectrum and the Raman spectrum both show that no Er₂SiO₅ phase is remained after the annealing at 1150 °C. Therefore, we argue that all the existing Er₂SiO₅ has transformed into Er₂Si₂O₇ while the rest of the Er atoms still exist in the form of amorphous Er₂O₃. It should be noted that the mixture of Er₂Si₂O₇ and Er₂O₃ does not exist in the phase diagram because the phase diagram shows conditions in thermodynamic equilibrium state, which can be different from the real situation.

Raman spectroscopy was also used to characterize the structure of the films. Figure 3 presents the Raman spectra of sample 3 annealed at 900 °C, 1000 °C and 1150 °C in N₂ and sample 4 annealed at 1000 °C in N₂. The Raman peak at 520 cm^-1 is attributed to the Si substrate. It should be noted that there are two polymorphs existing in sample 3 annealed at 1000 °C. Therefore, the Raman spectra are measured at two different sites (site A and site B). As shown in Fig. 3(a), there is only one line located at 520 cm^-1 in the Raman spectrum of sample 3 annealed at 900 °C, which corresponds to the Raman shift of the Si substrate. No any peaks related to the film appear, demonstrating the amorphous nature of the film. All the other films show multiple Raman peaks besides the line at 520 cm^-1. The different Raman shifts in these films correspond to different Er silicate polymorphs. Therefore, a corresponding relationship can be established between the lines of Raman shifts and the XRD results. The corresponding Er silicate polymorphs are also provided in Fig. 3. Detailed investigations of the Raman shifts of different polymorphs have been carried out in other rare earth silicates such as yttrium and lutetium while no literature data can be taken as a reference in Er silicates [21,22]. However, we can roughly analyze the Raman spectra based on the crystal structures of different polymorphs. α-Er₂Si₂O₇ contains chain-like (Si₃O₁₀) groups plus additional (SiO₄) tetrahedra. X1-Er₂SiO₅ contains isolated (SiO₄) tetrahedra and non-silicon-bonded oxygen ions. The existence of (SiO₄) tetrahedra in both polymorphs leads to some similarity in the Raman spectra of α-Er₂Si₂O₇ and X1-Er₂SiO₅. Both y-Er₂Si₂O₇ and β-Er₂Si₂O₇ are composed of (Si₂O₇) dimers. However, the diortho groups in y-Er₂Si₂O₇ show a Si-O-Si bond angle of about 133 ° while the Si–O–Si angle in β-Er₂Si₂O₇ is 180 °, leading to entirely different lines of Raman shifts.

 

Fig. 3. Raman spectra of sample 3 annealed at 900 °C (a), 1000 °C (b), 1150 °C (c), and sample 4 annealed at 1000 °C (d). (e) OM image of sample 3 annealed at 1000 °C and (f) spatial Raman map of the peak located at 845 cm^-1.

Download Full Size | PDF

As mentioned above, the distributions and proportions of different polymorphs are important for analyzing the samples but difficult to obtain by XRD analysis. However, optical microscopy (OM) combined with Raman spectroscopy can be used to estimate these data in our films due to the following reasons. First, the thicknesses of our films are as thin as 100 nm, so only one crystallite exists along the longitudinal direction, which can be confirmed by the TEM images. Second, on the basis of the same thickness, different polymorphs show different colors because of the difference in their refractive indexes. Sample 3 annealed at 1000 °C was taken as an example. Figure 3(e) shows the OM image of the sample. Blue clusters can be observed in the matrix. Site A and site B where the Raman spectra in Fig. 3(b) are taken are also shown in the OM image. Figure 3(f) displays the corresponding spatial Raman map of the peak located at 845 cm^-1 for the sample. The variation of this Raman peak demonstrates that the blue clusters in the OM image correspond to α-Er₂Si₂O₇ and the matrix corresponds to y-Er₂Si₂O₇. Therefore, the proportion of α-Er₂Si₂O₇ can be estimated by calculating the area percentage of the blue clusters. This estimation method is also feasible for the other films. Five OM images were acquired at different sites in each sample to obtain the average proportions. We estimate that the proportions of α-Er₂Si₂O₇ in sample 3 annealed at 1000 °C and sample 5 annealed at 1150 °C in N₂ are at least twice as large as the proportion of the other polymorph while the proportions of α-Er₂Si₂O₇ in sample 2 annealed at 1000 °C and 1150 °C in N₂ are much less than the proportion of the other polymorph.

PL measurements were performed on the samples in order to characterize their luminescent properties. Figure 4 shows the PL spectra of the samples annealed at 1000 °C and 1150 °C. Multiple peaks can be observed in the PL spectra due to the Stark splitting of energy levels of Er³⁺ in a crystalline environment. The peak positions and the shapes of the spectra are different for different samples. According to the XRD results, both sample 3 annealed at 1150 °C and sample 4 annealed at 1000 °C consist of single polymorphs, β-Er₂Si₂O₇ and X1-Er₂SiO₅ respectively. Therefore, the PL spectrum with the main peak at 1539 nm and sub-peaks at 1520 nm, 1530 nm, 1546 nm, 1551 nm and 1566 nm is attributed to β-Er₂Si₂O₇ and the PL spectrum with the main peak at 1530 nm and a sub-peak at 1560 nm is attributed to X1-Er₂SiO₅. As shown in the XRD and the Raman results, both sample 2 and sample 5 annealed at 1150 °C are composed of a mixture of α-Er₂Si₂O₇ and β-Er₂Si₂O₇. However, the shape of the PL spectrum of sample 2 annealed at 1150 °C is identical to that of β-Er₂Si₂O₇, and no emission peaks attributed to β-Er₂Si₂O₇ can be observed in the spectrum of sample 5 annealed at 1150 °C, so we attribute the PL spectrum of sample 5 annealed at 1150 °C with the main peak at 1529 nm and sub-peaks at 1546 nm and 1556 nm to α-Er₂Si₂O₇. It can be seen that the PL spectrum of α-Er₂Si₂O₇ is very similar to that of X1-Er₂SiO₅, but the width of its emission peak is much larger. After the annealing at 1000 °C, sample 2 and sample 3 show almost the same PL spectrum except that the relative intensity of the peak at 1527 nm is larger for sample 3. Both of these samples consist of a mixture of α-Er₂Si₂O₇ and y-Er₂Si₂O₇. However, neither of them shows the main emission peak of α-Er₂Si₂O₇ at 1529 nm. Considering the little difference between 1527 nm and 1529 nm and the larger width of the 1529 nm peak of α-Er₂Si₂O₇, we attribute the larger relative intensity of the peak at 1527 nm in sample 3 to the larger proportion of α-Er₂Si₂O₇ and the PL spectrum of sample 2 with the main peak at 1536 nm and sub-peaks at 1517 nm, 1527 nm, 1548 nm, 1556 nm and 1573 nm to y-Er₂Si₂O₇. It is well known that the chemical environments of Er ions have a great influence on the emission wavelength of Er due to nephelauxetic effect. The red shift of the emission wavelength increases with the decrease of the coordination number and the decrease of the Er-O distance [23]. The crystal structures of β-Er₂Si₂O₇ and X1-Er₂SiO₅ have been investigated in detail [24,25]. And the cell parameters of α-Er₂Si₂O₇ has been demonstrated [26]. However, no complete crystallographic data of α-Er₂Si₂O₇ and y-Er₂Si₂O₇ has been reported. Therefore, we assume that the atomic positions of Er, Si, O in α-Er₂Si₂O₇ and y-Er₂Si₂O₇ are the same as those of Tm, Si, O in α-Tm₂Si₂O₇ and Y, Si, O in y-Y₂Si₂O₇ [27,28]. The assumption is validated by the facts that the XRD peaks of α-Er₂Si₂O₇ and y-Er₂Si₂O₇ are in good agreement with those of α-Tm₂Si₂O₇ and y-Y₂Si₂O₇ and the ionic radius of Er³⁺, Y³⁺, and Tm³⁺ are almost the same. The cell parameters of y-Er₂Si₂O₇ were refined using the Rietveld analysis. The coordination numbers of y-Er₂Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇ are 6, 8 and 6, respectively. The sequence of Er₂Si₂O₇ sorted by average Er-O distance in descending order is α-Er₂Si₂O₇ (2.40 Å, 2.39 Å, 2.47 Å and 2.41 Å for the four nonequivalent Er sites), y-Er₂Si₂O₇ (2.34 Å and 2.28 Å for the two nonequivalent Er sites) and β-Er₂Si₂O₇ (2.26 Å). Therefore, the red shift of the peak wavelength is the largest for β-Er₂Si₂O₇, followed by y-Er₂Si₂O₇, and the smallest for α-Er₂Si₂O₇. X1-Er₂SiO₅ is different from the above Er₂Si₂O₇ polymorphs because its two nonequivalent Er sites has different coordination numbers, which are 7 and 9. The Er-O distances of the Er sites are slightly different from those of α-Er₂Si₂O₇. Considering both the coordination numbers and the Er-O distances, the difference between the main emission peaks of X1-Er₂SiO₅ and α-Er₂Si₂O₇ should be very small. The larger width of the main emission peak of α-Er₂Si₂O₇ is due to its larger number of nonequivalent Er sites.

 

Fig. 4. PL spectra of sample 2 to sample 5 annealed at 1000 °C and 1150 °C.

Download Full Size | PDF

Different Er silicate polymorphs also lead to different PL efficiencies, but it is hard to obtain the differences between these polymorphs by directly comparing the PL spectra, because the amount of Er silicates, the defect concentrations and the crystal qualities are entirely different for different samples. Nevertheless, we can use the proportions of different polymorphs in these samples, which have been acquired by the combination of Raman spectroscopy and OM images, to make the comparison indirectly. In sample 3 annealed at 1000 °C, the proportion of y-Er₂Si₂O₇ is much smaller than that of α-Er₂Si₂O₇, but the PL spectrum of this sample mainly consists of the PL peaks from y-Er₂Si₂O₇, so it is concluded that y-Er₂Si₂O₇ yields much stronger PL intensity than α-Er₂Si₂O₇ under the same volume. In addition, the volume ratio of β-Er₂Si₂O₇ to α-Er₂Si₂O₇ in sample 5 annealed at 1150 °C is similar to the volume ratio of y-Er₂Si₂O₇ to α-Er₂Si₂O₇ in sample 3 annealed at 1000 °C. However, no PL from β-Er₂Si₂O₇ can be observed in the PL spectrum of sample 5 annealed at 1150 °C. It indicates that the PL efficiency of β-Er₂Si₂O₇ is smaller than that of y-Er₂Si₂O₇. Therefore, we demonstrate that y-Er₂Si₂O₇ shows the largest PL efficiency among the three polymorphs.

Figure 5 presents the PL decay curves of sample 2 and sample 4 annealed at 1000 °C and sample 3 annealed at 1150 °C. These samples show different PL lifetimes due to different phase compositions and different crystal qualities. Table 2 summarizes the PL lifetimes acquired by exponential fit of the samples shown in Fig. 4. As demonstrated, some of the samples are composed of two polymorphs. However, only sample 3 annealed at 1000 °C can be fitted by the sum of two exponentials, which indicates that the PL from the minority polymorph is too weak to detect in the other films. From a comparison between the PL lifetimes and the phase compositions of different samples, it is concluded that y-Er₂Si₂O₇ shows the longest PL decay, followed by α-Er₂Si₂O₇, β-Er₂Si₂O₇, and X1-Er₂SiO₅ in descending order. It should be noted that the lifetime of α-Er₂Si₂O₇ in sample 5 annealed at 1150 °C (10 µs) is larger than that in sample 3 annealed at 1000 °C (6 µs) because higher annealing temperatures lead to better crystal qualities. X1-Er₂SiO₅ shows the smallest PL lifetime due to the higher Er concentration. As for polymorphic Er₂Si₂O₇, the difference in PL lifetimes is determined by the crystal structures of different polymorphs. The average Er-Er distances for y-Er₂Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇ are 3.92 Å, 3.79 Å and 3.53 Å, respectively. Smaller Er-Er distances result in stronger concentration quenching, and thus shorter PL lifetime. In addition, the Er sites of y-Er₂Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇ are of C_i, C₁ and C₂ symmetry, respectively. The higher symmetry of Er sites in y-Er₂Si₂O₇ also contributes to the longer lifetime thereof. The symmetry of Er sites in β-Er₂Si₂O₇ is better than that in α-Er₂Si₂O₇, so the PL lifetimes of these two polymorphs are similar despite their large difference in Er-Er distance.

 

Fig. 5. PL decay curves of sample 2 and sample 4 annealed at 1000 °C and sample 3 annealed at 1150 °C. The red lines represent single-exponential fits.

Download Full Size | PDF

Table 2. PL lifetimes of the samples

View Table | View all tables in this article

Considering the change of defect concentrations under different annealing temperatures, the exact ratio between the PL lifetimes of different polymorphs should be calculated using lifetimes from samples annealed at the same temperature. Then we can obtain that τ_y : τ_α : τ_β : τ_X1 ≈ 2.2 : 1 : 0.9 : 0.6. Accordingly, the ratio among the Er lifetime-density products (LDP) of these polymorphs is LDP_y : LDP_α : LDP_β : LDP_X1 ≈ 3.1 : 1.5 : 1.3 : 1.2. The magnitude of LDP represents the capability of light amplification at 1.54 µm for Er-doped materials [13]. Therefore, the y-polymorph is the best candidate for optical gain. However, the PL lifetime of the y-Er₂Si₂O₇ in our films is still not long enough compared with that of the erbium chloride silicate nanowires [9]. As we know, the PL lifetime can be increased by higher annealing temperatures due to the reduction in defect concentrations. Under such circumstances, we should focus on how to obtain single y-polymorph while increasing annealing temperatures. As demonstrated above, the phase transformation from y-Er₂Si₂O₇ to higher temperature polymorphs can be restrained by the deviation of the chemical composition from the stoichiometry of Er silicates and the annealing in an O₂ atmosphere. Therefore, we can increase the Er or O content in the Er-Si-O film as well as anneal the films in O₂ in order to obtain single y-polymorph at higher temperatures, but this process must be carefully designed because the phase transformation of Er silicate can be very sensitive to these factors. Besides, modification of the RF sputtering parameters such as decreasing the sputtering rates and increasing the substrate temperatures may also lead to a reduction in defect concentrations.

4. Conclusion

In conclusion, we have systematically investigated the effects of chemical compositions, annealing temperatures and annealing atmospheres on the formation of X1-Er₂SiO₅, y-Er_­2Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇ in Er-doped SiO₂ films and the luminescent properties of these polymorphs. We demonstrate that higher annealing temperatures transform the Er silicates into polymorphs with higher temperature stability. However, the annealing in O₂ atmosphere and the deviation of the chemical compositions of the Er-Si-O films from the stoichiometry of Er silicates will restrain this process. Therefore, control of the formation of different polymorphs can be realized by modifying the three factors. The crystal structures of the Er silicate polymorphs have a strong correlation with the PL peak wavelength, the PL efficiency and the PL lifetime. The wavelengths of the PL peaks are 1530 nm, 1536 nm, 1529 nm, and 1539 nm for X1-Er₂SiO₅, y-Er_­2Si₂O₇, α-Er₂Si₂O₇ and β-Er₂Si₂O₇, respectively. We demonstrate that y-Er₂Si₂O₇ yields the largest PL efficiency, the longest PL lifetime, and thus the largest potential optical gain among these polymorphs.