1. Introduction

New concepts and technologies for digital data acquisition, processing and transportation have strongly accelerated the information age. As a consequence, we have seen an explosive increase in information transfer capacity, i.e. an average growth of 58% per year [1]. In this regard, the development of materials for optical data transmission and amplification has been of continuous urgency. Traditional rare earth doped devices such as erbium doped fiber amplifiers (EDFAs) will not, on the long run, suffice to satisfy the demand because of inherent limitations in gain bandwidth, typically to not more than 100 nm [2,3]. This has provoked intense research on novel types of amplifier materials with broader emission bandwidth, at best covering the complete telecom window of about 1200-1600 nm (ultra-dry optical fiber). Various types of transition and post-transition metal doped materials have been studied for this purpose [2–6].

Among these, NIR emitting bismuth-doped glasses are presently considered as one of the most promising candidates. For their intriguing spectral properties, within only two years, research on Bi-doped devices has rapidly progressed from the first demonstration of lasing to efficient all-fiber optical amplifiers and lasers [6–9].

While numerous glasses have now been studied as matrix material for Bi-doping, the most relevant work is still focusing on silica fiber [6–8,10,11]. Preforms are typically fabricated by MOCVD and subsequently drawn into fiber at above 2000°C. This fabrication temperature is much higher than the boiling point of either pure bismuth (1564°C) or bismuth oxide (1890°C) and unavoidably leads to depletion of Bi-species from the fiber, resulting in lower residual dopant concentration and a concentration gradient across the fiber diameter [6,10]. For example, Razdobreev and Bigot reported a residual concentration of only ~50 ppm of Bi in doped silica fiber [11]. Hence, several tens or even hundreds of meters of bismuth fiber are required to realize lasing. The inhomogeneous distribution of bismuth along the fiber core will deteriorate the quality of the final output beam. It appears that these two intrinsic problems can only be overcome by abandoning the principle fabrication process. This requires matrix glasses with acceptable optical performance which can be processed at significantly lower temperature, e.g. lead germanate glasses [12] or lithium zinc aluminosilicate glasses, which may accommodate several mol-% of Bi-species [13].

Optical amplification in Bi-doped germanate glasses was observed by Zhou et al. [9], following the first principle demonstration of NIR-photoluminescence from that materials class by Peng et al. [3]. While the presence of Al³⁺ was first thought to be required for NIR emission, it was later found that also introducing Pb²⁺, Ga³⁺, B³⁺, or Ta⁵⁺ leads to similar luminescence behaviour [2,14,15]. It is now understood that these “codopants” are hence not direct contributors to NIR emission in Bi-doped glasses [14]. Nevertheless, introducing these species typically enhances NIR emission intensity, partly due to their role in providing the redox environment for stabilizing active Bi centers [2,12,16–19]. Noteworthy, while NIR emission typically requires red/IR excitation, UV-VIS to NIR conversion was reported for Bi-doped germanate glasses [20]. Jiang et al. investigated the influence of melting temperature and atmosphere in lithium gallium germanate glass [21]. Beyond matrix composition and processing conditions, co-doping with secondary optically active species provides a third means to improve emission efficiency and/or to adjust excitation and emission schemes. For this reason, co-doping of Bi-doped NIR emitting glasses has received some attention. E.g., the effect of Yb³⁺ and Tm³⁺ co-dopants has recently been studied in Bi-containing phosphate and silicate glasses, and energy transfer was reported from Yb³⁺ to Bi and Bi to Tm³⁺ [22–25]. Kuwada et al. reported on Er³⁺-co-doping which lead to NIR fluorescence with FWHM of 420 nm upon 800 nm excitation [26]. The latter appears indeed a very promising result, but the underlying interaction mechanism remains unclear.

For its optical and processing properties as well as the absence of any toxic species, the glass forming system of Al₂O₃-GeO₂ appears as an particularly interesting candidate for the production of active fiber waveguides and amplifiers. Regarding practical application as laser gain medium, resistance against thermal quenching, excited state absorption and appropriate pumping schemes would then be very important information. These are, however, at present not available for this particular system. In the present work, we therefore focus on a thorough investigation on the influences of temperature, pump power and excitation wavelength on NIR photoluminescence from Bi-species in Al₂O₃-GeO₂ glass. Based on previous observations in silicate glasses, we then include Er³⁺ into the system to further broaden the emission band. The mechanism of energy transfer between Er³⁺ and Bi will be discussed in detail.

2. Experimental

Glasses with molar compositions of (95-x-y)GeO₂·5Al₂O₃·xBi₂O₃·yEr₂O₃ (x=0, 0.3, 0.5, 1.0; y=0, 0.5) were prepared by conventional melting and quenching. The content of alumina was fixed at 5 mol.% since this value was previously identified as optimum for the formation of stable glasses [16]. In the following, sample nomenclature is GAxByE. E.g., GA0.5B0.5E for x=0.5 and y=0.5. High purity GeO₂ (99.999%), Er₂O₃ (99.99%), analytic reagent Al₂O₃ and Bi₂O₃ were selected as raw materials. Batches of 20 g for each sample were weighed and thoroughly mixed in an agate mortar. Each batch was melted at 1540°C in a high-purity alumina crucible for 20 min in air, subsequently cast onto a stainless steel plate and finally annealed at 600 °C for 2h. All obtained samples were visually transparent and bubble free. From the glass slabs, individual specimens were cut and polished for optical analyses.

Absorption spectra were recorded with a JASCO V-570 spectrophotometer. NIR emission and lifetime at room temperature were obtained with the same setup as reported in Ref. 20. A Xe short-arc lamp (150W) was employed as excitation source (760 to 980 nm). The fluorescence signal was detected with an InGaAs detector. A modulated 808 nm laser and a Tektronix TDS 3052 oscilloscope were used to obtain temporal decay curves. Low temperature NIR emission spectra and lifetime were measured on an Edingburgh FLS 920 fluorospectrometer between 10 and 300 K. At each temperature, three measurements were done to obtain the mean lifetime of the excited states.

Upconverted emission from Er³⁺ was measured with a Horiba Jobin Yvon Triax 320 fluorometer, using a 1 W 808 nm laser diode as pump source at room temperature. Raman spectra were measured on a Spex 1877 spectrometer equipped with an Ar ion laser (514.5 nm). Refractive indices were measured with an Abbé refractometer and density was obtained by the traditional Archimedes method.

3. Results and discussion

Usually, with increasing temperature, also network vibrations become stronger, enhancing the interaction between dopants and the local crystal field. Such stronger interaction leads to increasing probability of nonradiative energy transfer, what reflects in a broader luminescence spectrum, a lower level of the lowest excited state (increase of splitting gaps between excited states) and, hence, thermal quenching, reduced excited-state lifetime and red-shifting emission peak. The picture, however, appears more complicated for Bi-doped germanate glasses (Fig. 1 ).

 

Fig. 1 (A) Emission spectra of GA1.0B0E glass upon 500 nm excitation at different temperatures; (B) Fluorescence intensity and mean lifetime of GA1.0B0E as a function of temperature. Dynamic data were obtained for excitation at 500 nm, recording emission at 1200 nm.

Download Full Size | PDF

Figure 1 represents temperature-dependent emission spectra (A) and fluorescence lifetime (B) of sample GA1.0B0E. Noteworthy, luminescence decay does not follow a single exponential function. As a result of thermal quenching, gradual heating from 10K to 300K results decreasing NIR emission intensity by ~19%. Unexpectedly, however, the emission peak blue-shifts with increasing temperature (i.e., from 1200 nm to 1185nm). Furthermore, rather than broadening, FWHM of the emission band decreases from 300 nm to 291 nm. Finally, the mean lifetime of fluorescence remains almost constant. The latter observation appears similar to previous findings on Bi-doped lithium aluminosilicates [27]. All three seemingly contradicting phenomena reflect the complex nature of NIR emission from Bi-doped materials [6,19,28,29]. In a first summary, Fig. 1 provides clear evidence for good resistance to thermal quenching.

Previously, we have shown that within the spectral range of 200-760 nm, varying the excitation wavelength has no significant effect on the position of the NIR emission peak (i.e., 1070-1180nm [20], ). In contrast, further decreasing the excitation energy (i.e., 760 nm to 980 nm) reveals a very different picture where the position of the emission peak now clearly depends on excitation wavelength (Fig. 2A ). That is, starting with an emission peak at 1170 nm (760 nm excitation), a red-shift to 1215 nm is observed when the excitation wavelength is increased to 780 nm. This shift continues to 1240 nm when further increasing excitation wavelength to 800 nm. For excitation at 800-840 nm, the emission peak remains unchanged. When, however, increasing excitation wavelength yet further to 860 nm and beyond, an emission shoulder first appears at around 1090 nm which subsequently grows into a full emission peak, seemingly at the expense of the 1240 nm peak. When exciting at 930 nm, the 1240 nm emission fully disappears, leaving only the peak at 1090 nm for excitation of up to 980 nm. In accordance with the observed multi-exponential decay, this complex behaviour implies that there are several active emission centers present in the glass. Raman spectroscopic analyses of GA0.5B0E reveal the presence of multiple structural units, particularly tetrahedral [GeO₄] (resonance frequency at ~415 cm⁻¹), and octahedral [GeO₆] (343, 540 and 600 cm⁻¹) with bridging and non-bridging oxygen species (see also, e.g., [30]). This structural setting provides a variety of ligand configurations to the optically active Bi-species. In a first consideration, the following conclusions can be drawn: (1) photoemission at 1090 nm is associated with absorption at 250-350nm, 550-640nm, ~720nm, and 930-980nm, respectively; (2) photoemission at 1140 nm follows absorption at ~220 nm, 500 nm and 740 nm, respectively; (3) photoemission at 1170 nm results from absorption at 400-450 nm and 750-760 nm, respectively, and (4) photoemission at 1240 nm results from absorption at 800-840 nm. Exciting at a wavelength where different emission centers absorb light consequently leads to ultrabroad luminescence. For example, excitation at 870 nm produces emission with FWHM of up to 377 nm. This is similar to results obtained by Hughes et al. for Bi-doped lead germanate glasses [12].

 

Fig. 2 (A) Emission spectra of GA0.5B0E for different excitation wavelengths (760-980 nm, labels). Artifact peaks marked with “*” result from scattered excitation light. (B) Dependence of integrated emission intensity of GA0.5B0E on excitation power density (W/cm²), using a 808 nm diode laser.

Download Full Size | PDF

In Fig. 2B, we consider the dependence of emission intensity on pump power. For Bi-doped silica glass, Kalita et al. reported increasing excited state absorption for ~800 nm pumping, dominating the spectral range of 900-1200 nm (covering the Bi-related emission peak) [31]. In the present case, we do not find any sign of luminescence saturation with increasing excitation intensity. Rather, we find an approximately linear increase with pump power density up to 600 W/cm² (808 nm). This might indicate that germanate matrices provide less severe excited state absorption and, therefore, a more suitable matrix for NIR-active Bi-species.

Figure 3A represents the absorption spectra of GA0B0.5E, GA0.5B0E and GA0.5B0.5E. Er³⁺-related absorption peaks locate at 366,380,407,452,488,522,544,653,795,978 and 1530nm (GA0B0.5E, curve 1 in Fig. 3A). They can readily be assigned to the transitions from ground state ⁴I_15/2 to (²G_7/2, ²K_15/2, ²G_9/2), ⁴G_11/2, ²H_9/2, (⁴F_3/2, ⁴F_5/2), ⁴F_7/2, ²H_11/2, ⁴S_3/2, ⁴F_9/2, ⁴I_9/2, ⁴I_11/2, and ⁴I_13/2, respectively, and result in a pale pink color (inset). Singly Bi-doped and Er³⁺/Bi co-doped samples exhibit the typical reddish color (inset, Ref [20]). The latter samples exhibit broad absorption peaks at ~500, 700, 800, and 1000 nm, consistent with Refs [3,14–16]. Interestingly, Er³⁺/Bi co-doping leads to a slight decrease of Bi-related absorption, indicating that the presence of Er₂O₃ affects partitioning and redox state of Bi-species. A similar observation has been made for Bi-doped Y₂O₃ - Al₂O₃ - SiO₂, where Bi-related absorption was found to decrease with increasing concentration of Y₂O₃ [32].

 

Fig. 3 (A) Absorption spectra of Bi singly-doped and Er/Bi co-doped samples. Inset: photographs of GA0B0.5E and GA0.5B0E [20]. (B) Emission spectra of GA0.5B0E, GA0.5B0.5E, GA0B0.5E, and GA0.5B1.5E upon 808 nm pumping. Artifact peaks marked with “*” result from second order diffraction of the pumping laser.

Download Full Size | PDF

After excitation at 808 nm, singly-doped GA0.5B0E exhibits a broad emission band centered at 1275 nm (Fig. 3B). Er³⁺ was originally introduced as co-dopant to enhance the photoluminescence intensity on the red tail of the Bi-associated emission band, i.e. in the range of 1400-1600 nm. Co-doped specimens exhibit the according additional emission peak at 1530 nm (Fig. 3B). At the same time, co-doping appears to provoke a blue-shift and a notable decrease in intensity of the of the Bi-related emission band. The blue-shift may be understood as a result of a decreasing amount of active Bi-centers in accordance with a previous study [16] as well as with the absorption data shown in Fig. 3A. In addition, the decrease in Bi-related emission intensity may further partially result from the competition between Er³⁺ and Bi-related absorption of the pump light. The intensity of the Er³⁺-related emission band (1530 nm) appears about 10-times more intense for co-doped samples as compared to the singly-doped sample (Fig. 3B). FWHM of co-doped GA0.5B0.5E is about 442 nm, a little higher than the value found in Er³⁺/Bi co-doped silicate glasses [26]. The ratio between emission intensities of Er³⁺ and Bi, R_Er/Bi, can be tuned by changing their content ratio (Table 1 ).

Table 1. Intensity ratio R_Er/Bi, ⁴I_13/2 lifetime (τ_Er) of Er³⁺, mean lifetime (τ_Bi) of Bi-luminescence and energy transfer efficiency (η_Er→Bi and η_Bi→Er) in (95-x-y) GeO₂ · 5Al₂O₃ · xBi₂O₃ · yEr₂O₃ (x = 0, 0.3, 0.5, 1.0, 1.5; y = 0, 0.5, 1.5) glasses.

View Table

To elucidate Er³⁺↔Bi interaction, luminescence decay was considered. Decay of Er³⁺ was clearly single-exponential. In this case, the lifetime τ_Er was obtained from a simple fit of the decay data to a single-exponential equation. In contrast, emission from Bi centers did not follow single-exponential decay (as noted before, see also [2,16]). The mean lifetime τ_Bi was hence calculated from

Increasing the content of either Er³⁺ or Bi always leads to reduction of τ_Er and increase of τ_Bi. This reflects that energy transfer occurs predominantly from Er³⁺ to Bi rather than in the other direction. With increasing Bi, η_Er→Bi increases from 28.91% to 49.31%. This seems contradictory to the observed increase in the intensity of the Er³⁺-related emission band after co-doping. To dissolve this discrepancy, the electronic structure of both species must be considered (Fig. 4 ). Consulting Fig. 3A, absorption at 808 nm greatly increases after co-doping. That is, the absorption coefficient rises from about 0.5 cm⁻¹ for GA0B0.5E to about 2.46 cm⁻¹ for GA0.5B0.5E (compared to a value of 2.75 cm⁻¹ in singly-doped GA0.5B0E). This is taken as one of the main reasons for the enhancement of Er³⁺-related emission intensity: pumping at 808 nm lifts electrons to the excited state at 12376 cm⁻¹, corresponding to ⁴I_9/2 in Er³⁺ and the second lower excited state ES2 in Bi. In a first assumption, we may estimate that the exitons partition equally on both states, i.e. half occupy ⁴I_9/2, half ES2. If interaction between both states was neglected, Bi-related emission from GA0.5B0.5E would then approximately decrease to ~50% as compared to singly-doped GA0.5B0E. At the same time, Er³⁺-related emission would be ~2.5 times more intense as compared to GA0B0.5E. In reality, for GA0.5B0.5E, Bi-related emission was found to decrease by about 20% as compared to GA0.5B0E (Fig. 3B), accompanied by the already noted 10-fold increase in Er³⁺-emission. This is taken a clear evidence for direct interaction between ES2(Bi) and ⁴I_9/2 (Er³⁺).

 

Fig. 4 Simplified energy level diagram of Er³⁺ and Bi, and dominant mechanism of interactions in co-doped Al₂O₃-GeO₂ glasses. Since the nature of Bi-related NIR emission remains unclear, Bi energy levels are defined by the absorption spectrum (curve 2 of Fig. 3A). The ground state is labeled by GS. The four excited states are ES1, ES2, ES3 and ES4, corresponding to the absorption peaks at 1000, 800, 700 and ~500 nm, respectively. Non-radiative processes are indicated by dotted lines, absorption transitions by solid red lines with upward arrow. Radiative transitions marked by solid lines with downward arrows. Predominant cross-relaxation energy transfer processes are denoted CR1, CR2, CR3 and CR4.

Download Full Size | PDF

Bi-codoping leads to a decrease in τ_Er from 5.05 ms to 3.17 ms, and an increase in τ_Bi from 410 μs to 421 μs (Table 1). Noteworthy, ⁴I_11/2 and ⁴I_13/2 of Er³⁺ are partially overlapping ES1 of Bi (Fig. 4). This provides two principle paths for depopulating ⁴I_13/2 and simultaneously populating ES1 of Bi: (1) energy exchange between ⁴I_11/2 and ES1 by cross relaxation CR1, and (2) energy exchange between ⁴I_13/2 and ES1 by cross relaxation CR2 (Fig. 4). Decreasing τ_Er results in decreasing probability of excited state absorption on ⁴I_13/2, and, consequently, suppresses upconverted emission from Er³⁺. The latter is supported by the experiment as shown in Fig. 5 .

 

Fig. 5 Upconverted green emission from Er³⁺ in (95-x-y)GeO₂·5Al₂O₃·xBi₂O₃·yEr₂O₃ (x=0, 0.3, 0.5, 1.0; y=0, 0.5) glasses under 808 nm 1W laser pumping.

Download Full Size | PDF

Judd-Ofelt analyses were conducted on the basis of the absorption spectrum of GA0B0.5E. They indicate that the probability of spontaneous electric dipole relaxation is 54.5 and 553.8 s⁻¹ for the transitions of ⁴I_13/2→⁴I_15/2 and ⁴S_3/2→⁴I_15/2, respectively. The latter is about 10 times larger than the former, implying that the upconversion process is dominant in the singly-doped sample. Emission bands are consequently detected at 543 nm and 521 nm (curve 1 in Fig. 5, assigned to transitions from ⁴S_3/2 and ²H_11/2 to ⁴I_15/2 [34]). Co-doping with Bi leads to a sixfold decrease in the intensity of these bands, tunable via the amount of Bi. Addition of Bi further leads to a predominance of the emission peak at 521 nm at the expense of the 543 nm transition. This adds further complexity to the assumed mechanism of interaction between Er³⁺ and Bi: both transitions fall within the 500 nm absorption band of Bi (ES4), giving rise to two more non-radiative decay paths, CR3 and CR4 (Fig. 4).

An important parameter for a potential laser gain medium is the product of stimulated emission cross section and emission lifetime, σ_em × τ_Bi, being inversely proportional to the laser threshold. σ_em was therefore estimated from the Füchtbauer - Landenburg formula, and σ_em×τ_Bi was calculated. σ_em and σ_em×τ are about 0.65 × 10⁻²⁰cm² and 2.74×10⁻²⁴cm²s, respectively, for sample GA0.5B0.5E. This value is comparable to that of lead germanate glass [2], and larger than 1.4×10⁻²⁴cm²s as found in Ti³⁺-doped sapphire. The examined glass might hence provide an interesting material for broadband optical amplification or as a widely tunable laser source.

4. Conclusions

In summary, the effects of temperature, pump power and excitation wavelength on near-infrared photoluminescence from Bi singly-doped and Er³⁺/Bi co-doped aluminogermanate glasses were presented. It was shown that in comparison to conventional silica or silicate matrices, the examined material exhibits superior resistance to thermal quenching and significantly less pronounced excited state absorption for pumping at 808 nm, suggesting improved suitability as matrix material for Er³⁺/Bi optical amplifiers and fiber lasers. By selecting the optimal excitation wavelength, photoemission can be initiated from multiple active centers in parallel. This results in an emission bandwidth FWHM of more than 370 nm. Er³⁺/Bi co-doping is further presented as an effective means to significantly enhance emission intensity from Er³⁺ centers by suppressing red-to-green upconversion. This indicates a new route towards improving the performance of Er-based optical devices (EDFAs and fiber lasers). The mechanism of Er³⁺↔ Bi energy transfer was examined in detail. Adjusting the molar ratio between both species provides an effective tool for tuning the emission scheme and further increasing emission bandwidth.