1. Introduction

For applications in the near-infrared spectral range (NIR), other glass systems besides silica have demonstrated growing interest [1,2]. Germanate glasses exhibit several advantageous properties such as high chemical, mechanical and thermal stability [1]. They offer a variety of material properties for numerous applications, ranging from low-loss optical fiber for telecommunication with a wider transmission window to photonic crystal fibres, sensors and amplifiers for the NIR [1,3]. With addition of heavy metal oxides such as bismuth- and lead-oxides, the transparency of germanate glasses can be extended into the mid-infrared up to 6 µm [4,5]. Here, Bismuth(III)oxide has the advantage of intoxicity in comparison to lead oxide. Incorporated in glasses, Bi₂O₃ can be either network modifier or network former, depending on other glass components and concentration, and can therefore further contribute to chemical, mechanical and thermal stability. In the presence of conventional glass formers such as Ge⁴⁺, Si⁴⁺, B³⁺ and P⁵⁺, Bi³⁺ coordinates mostly in BiO₃-pyramidal and BiO₆-octahedral units which, due to the fact that Bi³⁺ ions are highly polarizable, form a secondary network structure [6]. When bismuth is present in other forms than the 3⁺-state, it is typically expected to act as a network modifier which will be incorporated into the glass lattice by breaking-up the network structure and, therefore, reduce the glass transition temperature.

The redox state in which bismuth is present in an oxide matrix is highly sensitive to the (structural) environment, the basicity of the matrix components and various processing parameters. In principle, a large number of species is possible, ranging from Bi³⁺ as the predominant species in common applications to lower redox states such as Bi⁺ and Bi²⁺ [2,7–9], or higher oxidized species such as Bi⁵⁺ [10]. In addition, various subvalent species and ion clusters [11–13], Bi⁰ or metallic bismuth particles have been reported. The question as to which species are dominant in a given material has significant consequences for the material's optical properties, which makes it crucial to delicately control the redox conditions [14]. For example, decreasing the redox state of bismuth towards the formation of metallic nanoparticles has been known as an easy method to tune glass color from yellow to reddish tints to deep black [15]. To prevent the formation of lower-valence bismuth states during glass fabrication, several methods have been explored, such as calcination of the precursors, different annealing temperatures, shorter melting times and the use of different crucible materials [2,16,17]. Another possibility to prevent reduction is through oxidants, MeO_x, which provide oxygen to the glass system while the cation is embedded in the network. The use of an oxidant provides a major advantage in comparison to the other methods as it needs no specific and time consuming process variation or secondary batch treatment procedures. The oxidant used in this work was CeO₂ [6,18].

By adding the ceria(IV) oxide to bismuth-germanate glasses, the bismuth ion keeps the redox state of the raw material Bi₂O₃, which is Bi³⁺. Only small molar percentages are needed to avoid the darkening of the glass. Particularly for the fabrication of highly transparent bismuth-germanate optical fiber, it is important to understand the structural influence of cerium species and the quantitative effect of the decolorization of bismuth-germanate glasses.

However, also the lower redox states of bismuth are of some interest as they may show unique luminescence behavior in the visible and NIR range. If this could be further exploited, it could have significant implications for fiber lasers and optical amplifiers [19]. For example, after the first observation of ultrabroad NIR luminescence in Bi-doped silica glass by Murata et al. [20], NIR emission has been demonstrated in many glass systems, including several germanate glasses [17,21,22]. Despite the obvious interest for optical fiber devices, there is still no consensus regarding the origin of the NIR-luminescence [23], although most authors attribute it to lower redox states of bismuth. From a technical point of view, the darkening of the glass, which is associated with bismuth being in low oxidation states, is a disadvantage for transmittance in the visible and needs to be chosen in accordance to the application. Ideally, the redox state would be manipulated so that the required optical transmission is maintained, but also advantageous luminescence behavior is achieved.

The wavelength and the intensity of the emitted light depend on the local structure surrounding the bismuth ions as a result of ligand field strength. That is, coordination of the cations, distortions in the structural units around the cation and (phonon-assisted) energy transfer may influence the excitation and emission wavelengths of the active ion. Luminescence from Bi³⁺ typically occurs in the UV-blue spectral range, Bi²⁺ emits in the red-NIR. Both species can be excited in the UV to blue, depending on the glass system. Excitation wavelengths of 250 – 450 nm are commonly used [8,9]. Ultrabroad NIR emission follows a rather complex excitation scheme where various excitation bands occur throughout the UV-NIR spectral range [2,21,24]. As already noted, it is not exactly clear which redox state causes this emission, but it is assumed that the NIR luminescence originates from lower redox states such as Bi⁺ or ion clusters [2]. For example, Meng et al. showed that with increasing basicity of the glass, the NIR emission disappeared [25]. As a higher basicity favors a higher valence state of bismuth, it is likely that the emission in the NIR occurs from lower redox states of bismuth. It has also been shown that the use of reducing agents could improve the emission intensity as long as no metallic particles are generated [2,14]. As an empirical rule, the optimal excitation wavelength for the species Bi^x (with x = 0; 1 + ; 2 + ) is typically found in the range of 500 – 800 nm, whereas the emission band lies usually in the range of 1000 – 1700 nm [10–26] (however, also mid-infrared emission has been reported for, e.g., chalcogenide matrices) [27,28]. These spectral ranges demonstrate great interest for a variety of communication, amplification and sensing applications.

In this paper, we investigate the influence of cerium(IV)oxide on the optical properties of bismuth-germanate glasses and the change of the redox state of Bi. Glasses with varying cerium concentration were analyzed to gain insight into the trade-off between optical transparency (bismuth oxidation) and NIR luminescence (bismuth reduction).

2. Experimental

2.1 Sample preparation

Glass samples of the composition (mol.%) (70.5-x)GeO₂ – 24.5Bi₂O₃ – 5WO₃: xCeO₂, with x = 0; 0.1; 0.2; 0.5; 1; 2 were produced from batches of oxide powders. Anhydrous GeO₂, Bi₂O₃, WO₃ and CeO₂ with a purity of 99.99 + % were weighed ( ± 5 μg) and transferred into alumina crucibles. In a first step, calcination was done in a conventional muffle furnace (Nabertherm; HT16) at 500 – 700 °C for 1h. Afterwards, the same furnace was heated at a rate of 3 K/min to a temperature of 1200 °C and held at this temperature for 40 min. The obtained glass melt was then poured into preheated (450 °C) brass moulds. Finally, glass slabs were annealed at 490 °C for 1 h, using another muffle furnace (Linn, LM 31227) and cooled to room temperature at a rate of approximately 2 K/min. Samples were subsequently cut into thin disks (with a thickness of 2.45 - 2.7 mm) and mechanically polished in several steps to optical quality with a final polishing slurry grain size of 5 µm.

2.2 Spectroscopic characterization

To characterize the redox state of the as-melted glasses, the presence of Ce³⁺ was analyzed by electron spin resonance spectroscopy (EPR) and optical spectroscopy. EPR was conducted at a frequency of 9.8 GHz (Bruker, ESP 300E) on powder samples. Optical absorption data was collected over the spectral range of 200 nm to 2500 nm in a UV-NIR spectrophotometer (Perkin Elmer, Lambda 950), equipped with a 150 mm integration sphere. Fourier-transform infrared (FTIR) spectroscopic measurements (Nicolet, Impact 420) were performed in transmission mode in the region of 2500 nm – 10000 nm. All FTIR spectra were normalized and corrected for background signals, Fresnel reflection and surface imperfection prior to calculation of the absorption coefficient. Static and dynamic fluorescence spectra were recorded on a high-resolution spectrofluorometer (Horiba Jobin Yvon, Flourolog 3-22), equipped with double monochromators in excitation and emission. A static Xe lamp (450 W) and a Xe flashlamp (75 W) were used as excitation sources for static and dynamic analyses, respectively. The slit size of both the excitation monochromator and the emission monochromator was 5 nm for all measurements. The NIR-signal was recorded with a thermoelectrically cooled InP/InGaAs-photomultiplier tube (NIR-PMT, Hamamatsu H10330A-75), operating in the spectral range of 900 - 1700 nm.

3. Results and discussion

3.1 Optical absorption

Ceria(IV) doping has a pronounced effect on the color of the examined glass system (Fig. 1(a) ). In a first inspection, all glasses show high optical quality with no visible striae, bubbles, inclusions or cracks. With increasing the content of CeO₂, the visual coloration first decreases from black to deep red to light yellow at x = 0.2. For x > 0.2, the tint re-intensifies to deep orange and brownish red at x = 2.0. This phenomenology it attributed to the overlapping effects of bismuth oxidation (x ≤ 0.2), increasing absorption associated to Ce³⁺ species (x ≥ 0.2) and potential secondary, more complex redox relations which lead to the formation of further coloring bismuth species. Ce³⁺ is formed through reduction of Ce⁴⁺ in the Bi^(3-i)+/Ce⁴⁺ redox pair (i = 1...3). While Ce⁴⁺ is EPR inactive, the occurrence of Ce³⁺ can be detected using EPR spectroscopy. Ce³⁺-related defect centers shows weak bands at around 1400 G (g ~4.1) and 3000 G (g ~2.6) [29]. Qualitative EPR analyses have therefore been employed to explore the change in the cerium redox state in the glasses with different cerium concentration. No Ce³⁺-related signals could be detected in the CeO₂-free glass, whereas both mentioned bands were found to increase significantly with increasing CeO₂ doping. This demonstrates that a significant part of the doped Ce⁴⁺ species is reduced to Ce³⁺ during glass melting. Optical basicity and redox activities are not reported for the glass system of this study. Using estimates for the partial molar optical basicity of GeO₂, BiO₂, WO₃ and CeO₂, Λ_GeO2 ~0.7, Λ_Bi2O3 ~1.19, Λ_WO3 ~1.04 and Λ_CeO2 ~1.01 [30–32], we derive a relatively high value for the optical basicity of Λ ~0.83 for the employed series of glasses. In a further approximation, we assume that the absolute amount of bismuth is large compared to the absolute amount of cerium, and that the redox pair of Bi^(3-i)+/Ce⁴⁺ behaves similar to the pair of Tl⁺/Ce⁴⁺. For the latter, redox ratios have been reported for silicate matrices, log[Tl⁺/Tl³⁺] ~13 - 20Λ and log[Ce³⁺/Ce⁴⁺] ~5.4 - 8.3Λ at 1400 °C [33].

 

Fig. 1 Photograph of (mol%) (70.5-x)GeO₂ – 24.5Bi₂O₃ – 5WO₃: xCeO₂ glass samples for varying x (a) and corresponding optical absorption spectra (b). The photograph was taken under ambient illumination.

Download Full Size | PDF

Formally, we further assume the symbolic reaction

Optical absorption spectra of the glasses with varying CeO₂ dopant concentration are shown in Fig. 1(b). In the absence of CeO₂ (x = 0), a strong absorption band occurs in the spectral range of 500 – 600 nm with peak intensity of 10.7 cm⁻¹. This band is assigned to lower valency Bi-species [2]. With increasing CeO₂ concentration, the intensity of this band decreases, i.e. to 2.3 cm⁻¹ at x = 0.1, and cannot be detected anymore at x = 0.2 or higher. For larger CeO₂ concentrations, the absorption edge shifts to longer wavelengths, i.e. the optical Band gap decreases from ~2.5 eV to ~1.5 eV (determined by the Tauc method). The sample with x = 0.2 exhibits the largest Band gap. This reflects that in sample x = 0.0 and x = 0.1 the absorption of Bi²⁺ and lower valence states of bismuth is dominant. On the other hand, at larger CeO₂ content, absorption is determined by the presence of cerium and associated charge transfer reactions which lead to the red-shift of the lower absorption edge and, in particular, 4f–5d transitions in Ce³⁺ ions [34].

Figure 2(a) represents FTIR spectra of the glass samples with x = 0; 0.1; 0.2; 0.5; 1; 2. All spectra comprise a strong absorption band related to hydroxyl groups (OH^-, ~2.8-3.5 μm) and at least three further bands. The latter are readily assigned to transitions from the ground state of Ce³⁺ to the Stark levels of ²F_7/2 [35]. Of the four possible bands, three can be detected unambiguously at ~3.8, 4.2 and 4.65 µm. The fourth is superimposed by the OH^--background absorption and becomes visible only after background subtraction (see inset of Fig. 2(b)). The evolution of the peaks with increasing CeO₂-doping is in very good agreement with the expectation of an increasing amount of Ce³⁺. For a quantitative evaluation, the x = 0 - spectrum was subtracted from each individual spectrum, assuming that OH^- and further background absorption (including absorption from eventual bismuth species) is not affected by the nominal CeO₂ - content in the considered wavelength regime (inset of Fig. 2(b)). Spectra were then integrated over the spectral range of 2.5-5.2 μm. As shown in Fig. 2(b), the obtained band area follows well a linear dependence. At constant sample thickness and constant extinction coefficient, this means that the amount of Ce³⁺ increases linearly with increasing absolute CeO₂ content. The latter is taken as a sign that within the considered range of dopant concentrations, there is no saturation or concentration-dependence of the redox reaction symbolized in Eq. (1).

 

Fig. 2 FTIR absorption spectra of (mol.%) (70.5-x)GeO₂ – 24.5Bi₂O₃ – 5WO₃: xCeO₂ glass samples for varying x (a), and integrated area of the 2.5 - 5.2 μm absorption band, corrected by the absorption spectrum of the undoped sample (b). Labels in (a) indicate absorption bands of Ce³⁺ and OH^- groups. The inset of (b) shows the spectra after subtraction of the absorption spectrum of the undoped sample, assuming that residual absorption originates solely from Ce³⁺.

Download Full Size | PDF

The absorption by Ce³⁺ is relatively high, up to 2.6 cm⁻¹ at ~4.65 μm. This needs to be considered in potential applications, as the absorption bands of Ce³⁺ lie within the intrinsic transmission window of the glass (i.e. at wavelengths shorter than the edge at 5.5 µm). OH^--related absorption can be controlled via the water content of the glass, i.e. by employing dry raw materials and adjusting process conditions.

3.2 Photoluminescence

Bismuth and its lower valence states, such as Bi²⁺ and Bi⁺ show luminescence in the visible and NIR spectral ranges [7–28,36]. In Fig. 3 , luminescence spectra of the glasses are shown as a function of CeO₂ content. Using an excitation wavelength of 368 nm, a broad emission band occurs around 650 nm (orange-red, Fig. 3(a)). This band is assigned to the transition of ²P_3/2(1) → ²P_1/2 in Bi²⁺, following non-radiative decay of ²P_3/2(2) to ²P_3/2(1) [8,9]. The signal of Ce³⁺ can be neglected as it shows emission at smaller wavelengths, i.e. 300 – 500 nm [37], this also lies in the absorption edge of these glasses and does not overlap with the emission of Bi²⁺ in the visible. The intensity of this band exhibits a maximum at x = 0.1, assumedly as a consequence of incomplete oxidation of Bi^(2-i)+ to Bi²⁺. With further increasing CeO₂-content, the amount of Bi²⁺ species is reduced by oxidation to Bi³⁺, leading to a continuous decrease in the intensity of the 650 nm emission band. Additionally, intrinsic quenching occurs due to overlaying absorption by Ce³⁺ (Fig. 1) in the same spectral range. Above an absolute CeO₂ content of 0.5 mol%, the intensity of VIS luminescence reduces drastically. This picture is confirmed by the evolution of NIR luminescence with CeO₂ content (Fig. 3(b)). For x ≤ 0.1, a broad NIR luminescence band occurs over the spectral range of 1000 - 1600 nm (note that at wavelengths > 1600 nm, the sensitivity of the employed PMT detector drops sharply). As noted before, the origin of this NIR emission band remains disputed. For the present case and considering the shape of the spectrum, we may assume that NIR luminescence originates from more than one redox state of bismuth. That is, for Bi²⁺, NIR emission at ~950 nm may result from electronic excitation of to the state of H, followed by non-radiative relaxation to X₂ and radiative decay to the ground state of ²P_1/2. The energy levels and their position concerning level I and X₂ have been taken from [38]. For Bi⁺, the electrons can be excited from ³P₀ to ³P₂ (or ¹S₀), relax to ³P₁ and decay radiatively to the ground level. These transitions are shown schematically in the energy level Fig. 4 . The intensity of the NIR band decreases already in the presence of a small amount of CeO₂, i.e. for x = 0.1, and vanishes completely for x > 0.1, indicating that practically all species of bismuth with valence < 2 + have been oxidized to higher valence, and that also Bi²⁺ species are gradually erased. The presence of metallic Bi nanoparticles has been ruled-out for the present case by analysis through transmission electron microscopy and X-ray diffraction.

 

Fig. 3 Luminescence spectra of (mol.%) (70.5-x)GeO₂ – 24.5Bi₂O₃ – 5WO₃: xCeO₂ glass samples for varying x at an excitation wavelength of 368 nm (a) and 500 nm (b), respectively.

Download Full Size | PDF

 

Fig. 4 Simplified energy level diagram for Bi²⁺ (left) and Bi⁺ (right). The arrows indicate potential excitation and relaxation paths.

Download Full Size | PDF

Exemplary emission decay curves for VIS and NIR luminescence, respectively, are given in Fig. 5 . Decay clearly does not follow a single exponential function. This indicates that for both cases (VIS as well as NIR emission), the optically active species are present in more than one lattice configuration, where different ligand situations result in different relaxation times. The best fit of the experimental data was obtained with a double exponential equation,

 

Fig. 5 Exemplary decay spectra of the 70.5GeO₂ – 24.5Bi₂O₃ – 5WO₃. The solid line represents the best fit of the data to a double exponential function, Eq. (2).

Download Full Size | PDF

Values of 2.6 and 10.6 µs are obtained for τ_1,2, respectively, of the Bi²⁺-related VIS luminescence (Fig. 5(a)). This corresponds-well to common expectations, where the lifetime of Bi²⁺-related luminescence hardly exceeds 20 µs. For example, a value of ~10.5 μs has been reported for Bi²⁺ ions in crystalline Bi₁₂GeO₂₀ [39], ~11-13 μs in crystalline strontium borate [9], or ~20-25 μs in crystalline alkaline earth phosphates [7]. For NIR emission, the lifetime measurement yields values of 6.8 and 27.4 µs for τ_1,2, respectively. Assumedly as a result of Bi-Bi interaction, these values are significantly lower than those which were previously reported for bismuth-doped alumina germanate glasses with low bismuth concentration [22], but are comparable to reports for bismuth cluster species [11].

4. Conclusions

Luminescence from cerium-doped 70.5GeO₂ – 24.5Bi₂O₃- 5WO₃ glasses occurs from Bi²⁺ (VIS) and lower valence states (NIR). The intensity of both emission bands can be controlled via the amount of CeO₂, where the latter acts as an oxidant for the transformation of bismuth in lower valence states to Bi³⁺. The parallel formation of Ce³⁺ leads to a trade-off in optical transparency in the visible spectral range. Optimal VIS transparency is obtained for a nominal CeO₂ concentration of x ~0.1-0.2 mol%. Ultrabroad NIR photoemission is erased at x > 0.1 mol%. Compared to lead germanate glasses, the intoxicity of bismuth(III)oxide provides a significant advantage for glass processing and application in optical devices. Using CeO₂ as an oxidant prevents formation of strongly coloring bismuth species. At appropriate composition and redox conditions, ultrabroad NIR luminescence from bismuth centers can be obtained without further dopants, but could be modified by the addition of co-dopants. This provides interest for optical amplifiers and lasers.