3.1 UV absorption

Fig. 1a shows vacuum UV transmission spectra of an undoped borosilicate (BS) glass sample with high purity (analysed Fe content < 1 ppm) and two samples doped with 20 ppm iron melted under reducing and oxidising conditions. Fig. 1b demonstrates the fit of two charge-transfer bands for Fe³⁺ with Gaussian shape. The dominating band has a maximum at 222 nm (45 000 cm^–1) and a very large bandwidth (full width at half maximum), 11 000 cm^–1. But a second smaller band (width ∼5500 cm^–1) with a maximum at 255 nm (39 200 cm^–1) with lower intensity was necessary for a well fitting of all measured spectra of the iron-doped glass samples melted under oxidising conditions. Fe²⁺ could not be detected in this sample with an error lower than 10% . On the contrary, large amounts of Fe³⁺ were found (15 to 70%) in all doped BS samples melted under reducing conditions. Fig. 1c illustrates a typical fit of one charge transfer band of Fe²⁺ with a maximum at 215 nm (46 500 cm^–1) and a small band width, ∼5500 cm^–1, besides the two bands for Fe³⁺. In comparison with the analysed Fe content by ICP–OES, chemical analysis and the fitted optical absorption spectra, the specific absorption coefficients for Fe³⁺ and Fe²⁺ in the borosilicate glass matrix (BS) were determined and compared with those in phosphate (P) and fluoroaluminate (FP) samples from previous papers 〚3, 4, 11, 12, 20〛 (Table 2).

Figs. 2–4 show the vacuum UV spectra of undoped and Sn doped glass samples investigated and possible fits for a d → s transition of Sn⁴⁺ with a maximum near the intrinsic VUV edge besides s → p transitions for Sn²⁺, which were found in all tin doped samples with high accuracy. In fluoroaluminate (FP) and phosphate (P) glass samples doped with tin, the electronic transitions of tin ions detected by optical absorption and luminescence spectroscopy are shifted to shorter wavelength, that means higher energies (Figs. 2 and 3). The amounts of Sn⁴⁺ and Sn²⁺ depending on the host glass were estimated by Mössbauer spectroscopy 〚21〛. The values obtained are summarised in Table 3.

3.2 UV-radiation-induced effects in iron-doped glass samples

Fig. 5 demonstrates the typical effect of photo-oxidation of Fe²⁺ by UV lamp irradiation in fluoride glass samples melted under reducing conditions, FP with ∼15 ppm Fe²⁺, on UV transmission (a), and the kinetic curves (b) depending on exposure time with increasing phosphate content. No change of Fe³⁺ species could be detected by UV irradiation of fluoride and phosphate glass samples 〚4〛, in contrast to silicate glass samples.

Fig. 6 shows the maximal defect formation curves by different UV lamp irradiation in iron-doped borosilicate glass samples melted under reducing (a) or oxidising (b) conditions. The influence of 193 nm laser irradiation with increasing pulse energy density on BS glass samples with 20 ppm Fe is given in Fig. 7. The kinetic curves are very different, depending on reducing (Fig. 7a) or oxidising (Fig. 7b) melting conditions. Figs. 8a and b illustrate the fitting procedure used in comparison with EPR 〚20〛 analysis in iron doped BS glass samples. Nearly 10 ppm (Fe²⁺)⁺ hole centre and mainly boron-related electron centre defects were attributed in samples melted under oxidising conditions (Fig. 8a). Approximately 6 ppm (Fe³⁺)^– electron centre combined with boron–oxygen related hole centre (BOHC) and also boron-related electron centre defects (B-EC) were found in samples melted under reducing conditions (Fig. 8b).

Figs. 9–13 give an overview of the UV lamp and laser-radiation-induced defects in different tin-doped glass samples. Tin-doped fluoroaluminate glass samples, FP/50 ppm Sn as Sn²⁺, are very stable against UV lamp irradiation. No significant change of the tin absorption was detected in fluoride glass samples, only photo-oxidation of ∼3 ppm Fe²⁺ from the initial value, 〚Fe²⁺〛₀ ∼ 15 ppm, as impurity (Fig. 9a and b). In contrast to phosphate, Sr(PO₃)₂/50 ppm Sn with ∼35 ppm as Sn²⁺ and Fe’ < 5 ppm, and borosilicate glass samples, BS/200 ppm Sn with ∼30 ppm as Sn²⁺ and ∼170 ppm as Sn⁴⁺, in which strong decrease of Sn²⁺ absorption also occurs with UV lamp irradiation. In FP glass samples with ∼50 ppm Sn²⁺, tin defect generation was found only with 193-nm excimer laser irradiation, which correlates with a very high negative radiation-induced absorption < 200 nm and positive values > 200 nm (Fig. 10). Defect formation at 193 nm is typical for Sn²⁺-related defects and at 450 nm for phosphorus oxygen-related hole-centre defects (POHC).

Fig. 11 compares the tin related defect formation in Sr(PO₃)₂/50 ppm Sn glass samples by different laser irradiation wavelength at 193 and 248 nm. Defect formation at ∼196 nm is typical for Sn²⁺-related defects and at ∼500 nm for POHC defects, which are much higher with 193-nm than with 248-nm laser irradiation. Figs. 12 and 13 give an overview of typical UV-radiation-induced defects in borosilicate glass samples doped with 200-ppm tin. The influence of 248-nm laser irradiation depending on pulse number and energy density per pulse is shown. At high energy density, 350 mJ cm^–2 per pulse, a very fast change of the UV transmission in a opposite way is observed already after 100 pulses: an increase of the transmission below ∼230 nm and a decrease between 230 and 400 nm. At further irradiation, the increase of transmission < 230 nm is continued slowly, but for wavelengths > 230 nm, an increase in transmission is also found (Fig. 12a). The kinetic curves for the typical defect formation, depending on the pulse number and energy density per pulse, are demonstrated at λ = 206 nm, as the maximum of the Sn²⁺ absorption, in Fig. 12b, and at λ = 250 nm, in Fig. 12c. The defect generation at 250 nm increases very fast, in the same way as the Sn²⁺ absorption decreases, reaches a maximum after ∼100 pulses, but then, after further irradiation, this defect generation decreases again. Both effects increase with increasing energy density. Figs. 13a, b and c show a first assumption of a peak simulation for the 248-nm laser-induced defect generation at the highest energy density, 350 mJ cm^–2 per pulse. The measured data (lines) could be well fitted with Gaussian functions (dashed lines) with a correlation coefficient > 0.999, and in good agreement with EPR measurements. After 100 pulses, the Sn²⁺ absorption bands are decreased and a lot of bands, accordingly EPR measurements mainly boron-related EC and HC defects, are formed, Fig. 13(a). Between 100 and 5000 pulses, further decrease of the Sn²⁺ bands connected with an increase of the boron-related defect bands is observed (Fig. 13b). As a result, after 5000 pulses (Fig. 13c), only a few weak intrinsic bands with maximums at 255, 280 and 340 nm besides the strong negative Sn²⁺ bands could be detected.

In the EPR spectra of tin-doped BS glass samples three additional signals, with g-values at 1.9775, 1.9866 and 1.9977, were found in all irradiated samples. The nature of these defects is still unknown, but their intensity was increased with increasing tin content.

4.1 Charge transfer absorption of Fe³⁺ and Fe²⁺ complexes in different glasses

The three glass types investigated (Table 1) have very low optical basicity, Λ 〚10〛, and a large intrinsic UV transmission. But their structure is very different 〚4, 5〛 and so their influence on charge transfer absorption of Fe³⁺ and Fe²⁺ complexes is also different. The dominating UV absorption bands are caused in all three glass types by Fe³⁺ complexes. The measured spectra could be well fitted with Gaussian shape bands. In the fluoride phosphate glass FP with the lowest optical basicity, Λ_Pb = 0.38, and the highest ionic bonding, the band maximum, λ = 260 nm, has the lowest energy (38 500 cm^–1 = 4.8 eV), the smallest bandwidth (8000 cm^–1), and the lowest specific absorption coefficient, ϵ_260 nm = 0.18 cm^–1 ppm^–1. The charge transfer should occur from the surrounding ligands to the central Fe³⁺ ion. In the case of FP, ligands are mainly fluorine in six-fold coordination, but oxygen ligands are possible too. The charge-transfer process in glass can be regarded as the promotion of a nonbonding p electron, on the ligand, to the lowest unoccupied energy orbital sited on the metal ion 〚10〛.

In phosphate glasses, Sr(PO₃)₂ and ultraphosphate 〚4〛, the maximum for the Fe³⁺ charge transfer band (λ = 240 nm) is shifted to a higher energy (41 700 cm^–1 = 5.2 eV), a higher intensity (ϵ_240nm = 0.20 cm^–1 ppm^–1) and a larger bandwidth (10 000 cm^–1). These phosphate glasses possess a higher optical basicity, Λ_Pb = 0.47, and a higher covalent bonding than the FP glass.

The borosilicate glass (BS ∼ Duran) with nearly the same low calculated optical basicity 〚20〛, like the phosphate glass, shows another behaviour for the Fe³⁺ charge-transfer bands, due to other sites for Fe³⁺. A very strong and broad band with a maximum at 222 nm (45 050 cm^–1 = 5.6 eV) is dominating (Fig. 1) and in good agreement with the data in high-purity soda–silica glass 〚13〛. A very high specific absorption coefficient, ϵ_222nm = 0.30 cm^–1 ppm^–1, was determined. But a second smaller band with a maximum at 255 nm with lower intensity (ϵ_255nm = 0.06 cm^–1 ppm^–1) was necessary for a well fitting of all measured spectra of iron doped BS glass samples. These Fe³⁺ charge transfer bands could be determined in all glasses investigated with high accuracy 〚2–4〛. Further transitions for Fe³⁺ and Fe²⁺ are assumed near the intrinsic UV edge of the glasses (Fig. 1a).

Exact determination of the Fe²⁺ charge transfer bands is more difficult. Only in FP glasses was it possible to shift the redox ratio completely to the Fe²⁺ state 〚2–4, 12, 20〛 and to determine the bands with high accuracy (Table 2). However, in phosphate and silicate glasses, the values for the Fe²⁺ charge transfer bands determined have a larger error. The main band maximum is between 215 and 220 nm, with a much lower intensity (ϵ_λ: 0.006 to 0.03 cm^–1 ppm^–1) than those for Fe³⁺.

4.2 Electronic transitions for Sn²⁺ and Sn⁴⁺

Three or four electronic s → p transitions for Sn²⁺ with 5s² configuration were detected by optical absorption and luminescence 〚20, 22〛 spectroscopy shifted to longer wavelength in the range FP → P → BS glass samples with increasing covalent bonding (Table 3). These electronic transitions are parity allowed and depends strongly from the host glass (Figs. 2–4). Three transitions could be well fitted in glass samples with only a few ppm Sn²⁺ (Figs. 2b and 3b); with higher Sn²⁺ content, a fourth band at longer wavelength was found (Table 3). The highest intensity should have the allowed ¹S₀ → ¹P₁ transition with a maximum at ∼175 nm in FP, ∼180 nm in P and at 206 nm in BS glass samples. The ¹S₀ → ³P₂ is forbidden (ΔJ = 2), but can obtain some intensity by coupling with vibrations, assuming for the fitted band with a maximum at 192 nm in FP, at 200 nm in P and at 216 nm in BS glasses. A third band, with a maximum at 217 nm in FP, at 220 nm in P and at 240 nm in BS samples and low intensity was fitted assuming for the ¹S₀ → ³P₁ transition, which is spin-forbidden but can also be observed, due to spin-orbit coupling, which mixes the spin triplet and singlet states 〚20, 22〛. The intensity ratios detected for the four transitions in all tin-doped glass samples investigated were approximately 10:5:1:0.1. The specific absorption coefficients, ϵ_λ, were calculated with chemical analysis, absorption and luminescence spectroscopy and Mössbauer measurements on samples doped with higher tin contents (0.5–1%) and melted under the same conditions 〚21〛. Only 10–15% of the total tin content was found in the Sn²⁺ state in BS samples. The extrapolation to very low Sn concentration is still a problem, and so the absolute values for the absorption coefficients are not so sure.

In the case of FP and P glasses, the Sn⁴⁺/Sn²⁺ ratio could be completely shifted to Sn²⁺ by melting under reducing conditions in a carbon crucible at temperatures > 1000°C 〚21〛. The s → p transitions are shifted in the vacuum UV region (Figs. 2 and 3) and these measurements have a larger error (∼30%) (Table 3).

For the Sr(PO₃)₂ glass sample doped with 50 ppm Sn, exists only an estimated value of ∼30 ppm Sn²⁺. The Sn²⁺ luminescence excitation and emission spectra 〚20〛 are also strongly dependent on the host glass. Narrow bands were measured in BS and very broad bands in FP samples different from those in P samples, due to the influence of the various local structure (surrounding ligands). But all measured spectra could be also well fitted with three or four Gaussian bands.

In all glasses investigated, a d → s electronic transition band with low intensity for Sn⁴⁺ in the VUV was found overlapped with the intrinsic edge (Table 3).

4.3 Radiation-induced effects in iron doped glasses

The influence of iron on radiation-induced effects in phosphate and fluoride phosphate glasses was described in previous papers 〚2, 4, 14, 15〛. Photo-oxidation of Fe²⁺, forming (Fe²⁺)⁺ hole centres (HC) and phosphorous related electron centres (EC) as a single photon process, was found with UV-lamp and laser irradiation in both glass types, phosphates (ultra- and meta-phosphates) and fluoride phosphates melted under reducing conditions (Fig. 5a and b). But the time dependence to reach the saturation value, for which a maximum of 50% of the initial Fe²⁺ content has been photo-oxidised, was very different. The measured kinetic curves could be well fitted with the function:

High-purity borosilicate glass samples, BS, with Fe < 10 ppm, are very stable against UV-lamp and 248-nm laser irradiation 〚6, 20〛. Therefore, the formation of UV-radiation-induced defects increases drastically with increasing iron content, especially in samples melted under reducing conditions (Figs. 6–8). BS samples for technical use normally have an impurity content of 100 to 200 ppm Fe. These glasses melted under reducing conditions have a good transmission in the UV–B region (280–320 nm) but unfortunately also a high solarisation (Fig. 6a and b), which increases with shorter wavelength irradiation (HOK > XeHg lamp). The kinetic curves for the maximal defect generation in Fe-doped BS samples melted under reducing conditions at λ_max ∼ 250 nm (Fig. 6a) could be well fitted with the following functions:

• • for HOK-lamp irradiation

  $$\mathrm{y}={\mathrm{y}}_0+\mathrm{a}\ln \mathrm{x}+\mathrm{b}{\left(\ln x\right)}^2+\mathrm{c}{\left(\ln x\right)}^3$$

• • for XeHg-lamp irradiation

  $$\mathrm{y}={\mathrm{y}}_0+\mathrm{a}\left[1-\exp \left(-\mathrm{b}x\right)\right]+\mathrm{c}\left[1-\exp \left(-\mathrm{d}x\right)\right]$$

In the case of samples melted under oxidising conditions (Fig. 6b), the maximal positive defect generation at λ_max ∼ 275 nm, depending on the exposure time x, could be fitted for both lamps with the same function (3), and for the negative defect generation at λ_max ∼ 218 nm with a simpler function (4):

This means that the mechanisms for the radiation-induced processes are very difficult to recognise, because they are very complex. The defects can change. With HOK and laser irradiation, the formation curves can pass through a maximum or a minimum, due to change of photo-oxidation into photoreduction of the iron species, which is strongly dependent on the irradiation wavelength and the energy density (Figs. 6a and 7b). This makes the extrapolation of the defect generation for long-time irradiation very difficult, but increases the defect generation in silicate glasses with increasing iron content generally. Most kinetic curves can be well fitted with the function:

The designation of the generated defect was assumed by correlation of the induced optical bands, EPR analysis combined with thermal annealing experiments 〚20〛. It was found that photo-oxidation of Fe²⁺ is the first dominating step in all samples melted under reducing conditions. That means Fe²⁺ absorbs an UV photon and loses an electron to form stabilised (Fe²⁺)⁺ defect hole centres. The electron is trapped, forming boron- and silicon-related electron centres (Fig. 8a and b). The maximum for the induced (Fe²⁺)⁺–HC absorption band is at the same position as that for the Fe³⁺ charge-transfer transition, at 222 nm (Fig. 8a). The fitted maximums for boron- and silicon-related EC are assumed at 240 and 280 nm. In BS glass samples with low alkali contents, boron-related EC defects dominate. The superposition of the induced bands leads to a maximal radiation-induced absorption at ∼250 nm. The photo-oxidation rate is very high and increases with the energy density and the excitation efficiency for the Fe²⁺ species. The last is higher for the HOK, λ > 190 nm, than for the XeHg lamp, λ > 230 nm. In samples doped with 100 or 200 ppm Fe, the saturation value is not reached after 100 h irradiation with the XeHg lamp (Fig. 6a). The fitting results have shown that only 15 to 40% of the initial Fe²⁺ content was photo-oxidised. With HOK lamp irradiation, then an another process is initiated: the photoreduction of Fe³⁺, forming a (Fe³⁺)^– defect electron centre by electron trapping and a boron-related hole centre (BOHC, λ_max ∼ 280 nm), and also a boron-related electron centre (B-EC, λ_max ∼ 240 nm), was detected as the dominating process in all samples melted under oxidising conditions (Figs. 6b, 7b, and 8b). The total content of the photo-reduced (Fe³⁺)^– is very low, only 5 to 10% of the initial Fe³⁺ content; e.g., in the sample, BS/20 ppm Fe_ox, a maximum of 2–3 ppm Fe³⁺ was reduced to (Fe²⁺)^– by HOK or laser (248 nm) irradiation (Figs. 6b, 8a and b). No significant photoreduction could be detected with XeHg-lamp irradiation and course-selective excitation of photoelectrons, λ > 230 nm; the highest values were found with X-rays, for which an avalanche of electrons is released 〚20〛. Irradiation with the 193-nm laser, near the UV edge 〚5, 16〛, generates also intrinsic defects of BS in the visible region and an efficient fast photoreduction of Fe³⁺ (Fig. 6b). The results of the optical spectroscopy were consistent with those of the EPR analysis 〚20〛, also for the Fe³⁺ signal at g ∼ 4.3.

The initial reaction for the solarisation should be the release of an electron. In the case of oxidised melted samples, it is not clear where does this electron come from? High-purity BS samples, Fe < 1 ppm, are very stable against UV lamp and laser (248 nm) irradiation (Fig. 4), but solarisation increases with increasing the iron content. It could be also possible that traces of Fe²⁺ (< 10% of the total Fe content) in the oxidised samples are responsible for the first step, the release of an electron, as the initiator for the photoreduction of Fe³⁺. But there has to be still an influence of the surrounding glass matrix, because this process was not found in fluoride and phosphate glasses 〚4, 14, 15, 17, 20〛. In these glasses, Fe³⁺ species were not able to trap an electron and stabilise a (Fe³⁺)^– defect electron centre. This is in contrast to the chemical reduction of Fe³⁺ to Fe²⁺ in the glass melt, which occurs very easy in FP and P systems 〚2–4〛. In the melts, a relaxation of the local structure for stabilising the reduced species is possible, whereas in the glass at room temperature this relaxation is not possible. The Fe³⁺ complexes are very stable and not involved in radiation-induced processes. They can be used as filters for increasing the radiation resistance. This is the main difference between fluoride and phosphate glasses, on the one hand, and silicate and borosilicate ones, on the other hand; in these latter, both Fe²⁺ and Fe³⁺ species take part into UV-radiation-induced processes, due to the local glass structure.

The photo-oxidised (Fe²⁺)⁺ defects in all glass samples investigated can also be formed with thermal treatment of irradiated glasses and no recovery was detected, in contrast to the photoreduced (Fe³⁺)^– defects, which were annealed to Fe³⁺ by thermal treatment below the T_g temperature.

4.4 Radiation-induced effects in tin doped glasses

Tin doped FP glasses (Figs. 2a, 9a and b) are very stable against UV lamp irradiation, in comparison to tin-doped P and BS glasses. The HOK lamp spectrum starts at very short wavelengths, > 190 nm, for which an excitation of the Sn²⁺ electronic transitions is possible in all glasses investigated (Fig. 2–4). But, in FP glasses, no significant change in Sn²⁺ absorption bands with UV lamp irradiation was detected, but only photo-oxidation of Fe²⁺ as an impurity (Fig. 9a and b). This means that the fluoride glass structure prevents a tin-related defect generation by UV lamp irradiation. With 193-nm laser irradiation, pulse energy density 200 mJ cm^–2, a fast Sn²⁺-related defect generation in FP glass samples was found (Fig. 10). The dependence of the measured kinetic curves on pulse number could be well fitted with the following equation (6):

Sn⁴⁺ species, which absorb in the shorter UV range (Fig. 2b) should also participate in the defect generation process. A model of defect generation in silicate glasses, doped with As, Sb, Sn, Pb or Ti, very similar to the mechanism of defect generation in FP glasses doped with Pb, is suggested 〚6, 15, 20〛. It can be explained by a two-step absorption process. This includes absorption of UV photons, an energy transfer process between the excited species, introduced by the dopants, and the precursors. The existence of Sn²⁺ or other ions, e.g., As³⁺, Sb³⁺, Pb²⁺, with s² electronic configurations, which also show luminescence transitions, is very important for the defect generation. The disappearance of the s² ions, which can be detected by the radiation-induced decrease in luminescence, explains the loss of induced extinction after passing through a maximum. In the case of tin-doped glasses, where the redox ratio Sn²⁺/Sn⁴⁺ can vary in a wide range between Sn²⁺ and Sn⁴⁺, both species can be involved in radiation-induced processes in all three glass types investigated.