1 Introduction

Phosphate glasses are of great technological interest due to specific thermal properties, i.e. low glass transition and dilatometric softening temperatures, and high thermal expansion coefficients, which make them ideal as low-temperature sealing glasses 〚1–5〛. The problem is their low chemical durability, often associated 〚6〛. However, the nitridation ability of these glasses under ammonia atmosphere, demonstrated as soon as 1982 〚7〛, is a suitable way to improve their chemical resistance.

The structure of metaphosphate glasses consists of –PO₄–PO₄– chains of tetrahedra, interconnected through modifier cations. The middle tetrahedra, named Q²-groups according to Lippma et al.’s notation 〚8〛, have two bridging oxygens (BO), bonded to neighbouring phosphorus atoms, and two non-bridging oxygens (NBO).

During nitridation in ammonia, P–N bonds form at the expense of P–O bonds, nitrogen substituting within PO₄ tetrahedra for both bridging and non-bridging double-bonded oxygens to form PO₃N and PO₂N₂ tetrahedra, as proved by ³¹P Nuclear Magnetic Resonance (NMR) 〚9, 10〛. Moreover, X-ray Photoelectron Spectroscopy (XPS) results have shown that nitrogen atoms exist as two-coordinated, –N=(N_d), and three-coordinated, –N<(N_t) species 〚10–12〛. As a result of the nitrogen/oxygen substitution, the cross-linking density and the covalency of the glass network increase, which improves effectively the glass durability.

Two different models have been proposed to explain the nitridation mechanism. Reidmeyer et al. 〚13〛 have assumed that the N/O substitution is a random process. On the opposite, Le Sauze et al. 〚9〛 have described the nitridation of the Li_0.5Na_0.5PO₃ glass by oxynitride microdomains growing progressively at the expense of oxide ones, where PO₂N₂ tetrahedra appear by substituting preferentially bridging oxygens shared by a PO₄ tetrahedron and a PO₃N tetrahedron. In both cases, it is concluded that non-bridging oxygens coordinating modifier cations are not substituted.

The higher field strength of divalent compared to monovalent modifier cations results generally in higher values of melt viscosity, which leads to a limited nitridation, for example when BaO is introduced in the starting glass formulation. PbO is an exception as its introduction provides a low melt viscosity and thus enables incorporation of high nitrogen contents, as shown by Pascual et al. 〚14, 15〛. Moreover, the presence of Pb²⁺ cations improves the glass chemical durability while maintaining low softening points 〚16〛, thus making lead-containing glasses appropriate materials for soldering applications.

This paper deals with the thermal nitridation in flowing ammonia of alkali and alkali–lead metaphosphate glasses. In addition to a comparison between the nitridation kinetics, ³¹P Magic Angle Spinning (MAS) NMR and Double Quantum (DQ)-MAS NMR, and XPS results have been used to compare the structural evolution resulting from progressive nitrogen incorporation.

2 Experimental

The oxide-base glass compositions 50 Na₂O·50 P₂O₅ (‘NaPO’), 25 Li₂O·25 Na₂O·50 P₂O₅ (LiNaPO) and 12.5 Li₂O·12.5 Na₂O·25 PbO·50 P₂O₅ (‘LiNaPbPO’), in mol%, were prepared by melting, according to a previously described procedure 〚17, 18〛. Phosphorus oxynitride glass samples, ‘NaPON’, ‘LiNaPON’ and ‘LiNaPbPON’, were prepared by thermal treatment of the base glasses in flowing anhydrous ammonia, at temperatures ranging from 600 to 750 °C for times up to 96 h, as previously described 〚17, 18〛.

Nitrogen analyses were carried out in a nitrogen/oxygen LECO TC-436 analyser by the inert gas fusion method, with a maximum deviation in the N/P atomic ratio of ±0.02. The nitrogen content was confirmed by direct chemical analysis as NH₃, according to the Grekov method 〚19〛.

³¹P MAS NMR spectra were recorded on a Bruker ASX 100 spectrometer operating at 40.48 MHz (2.34 T). The pulse length was 3.0 μs (π/2) and 120 s delay time was used. The spinning rate was 15 kHz. The ³¹P spectra were fitted to Gaussian functions, in accordance with the chemical shift distribution of the amorphous state. The precision of the relative component determination was ±0.5%.

³¹P Double Quantum MAS NMR measurements were performed at 2.34 T, using a previously described procedure 〚20, 21〛. A 4 mm MAS probe was used at a MAS frequency of 12.5 kHz. The BaBa (back to back) sequence was applied, using TPPI (time proportional phase increment) 〚20–22〛. A 3-μs pulse was used (π/2), and a 5 s recycling delay. Sixty-four t₁ increments of 80 μs were used, with 32 accumulations per slice.

XPS measurements were performed on a CAMECA RIBER SIA 200 electron spectrometer for multi-technique surface analysis system. This system is equipped with a MAC 2 CAMECA RIBER double stage cylindrical mirror electron-energy analyser. The photon source was a CAMECA SCX 700 dual anode X-ray source. A non-monochromatized Al Kα X-ray source (hν = 1486.6 eV) was used as the excitation source in all cases. A spectrometer energy calibration was made by using the Au 4f^7/2 and Cu 2p^3/2 photoelectron lines. Due to surface electrostatic charges detected for some samples and resulting in variable retarding effects, all the energy scales corresponding to the XPS spectra reported in this paper were normalized from the energy position of the C_1s photoelectron line of atmospheric carbon (CH₂)_n: 285 eV. The photoemission measurements were carried out on a series of samples before and after removing the surface carbon contamination by ion sputtering. The sputtering was performed for 10 min in the UHV (ultra high vacuum) analysis chamber with a 0.6 keV Ar⁺ ion beam (30 mA cm^–2). The elemental composition of the samples was evaluated using the integrated areas of the core-level peaks O_1s, C_1s, P_2p, Pb_4f, Na_1s and N_1s. The Li atomic concentration could not be measured due to the very low value of the ionisation cross-section for this element.

3 Results

3.1 Nitridation kinetics

Fig. 1 shows the progressive nitrogen enrichment of NaPON, LiNaPON and LiNaPbPON glasses as a function of nitridation time at T = 700 °C. For the three different base glasses studied, the nitrogen content reaches an asymptotic limit, which is different for each composition. The time after which this limit is reached, i.e. the nitridation rate, depends also on the starting glass composition.

3.2 ³¹P MAS NMR

The ³¹P MAS NMR spectra of the NaPO₃, Li_0.5Na_0.5PO₃ and Li_0.25Na_0.25Pb_0.25PO₃ base glasses contain a single peak at –20.7, –21.4 and –23 ppm, respectively. According to Sato et al. 〚23〛, this resonance corresponds to Q²-type sites. From the previously reported chemical shifts, Q²(Na): –19 ppm 〚23〛, Q²(Li): –23 ppm 〚9〛 and Q²(Pb): –24.5 ppm 〚24〛, and considering an average distribution of the cations around the Q² sites, we calculate Q² (Li, Na) and Q² (Li, Na, Pb) values of –21 ppm and –22.8 ppm for the Li_0.5Na_0.5PO₃ and Li_0.25Na_0.25Pb_0.25PO₃ glass compositions, respectively, in agreement with the experimental results.

As N/P increases, two new resonances can be observed in the ³¹P MAS NMR spectra of the nitrided glasses in addition to the Q²-type resonance mentioned above, which are assigned to PO₃N (∼–10 ppm ) and PO₂N₂ (∼0 ppm ) tetrahedra 〚9, 10, 24〛. As an example, Fig. 2 shows the ³¹P NMR spectra of LiNaPbPON glasses.

For the three studied series, PO₃N and PO₂N₂ tetrahedra appear at the expense of PO₄ tetrahedra. Fig. 3 shows the variation of the relative proportions of the three kinds of tetrahedra as a function of the N/P ratio.

The PO₄ percentage (Fig. 3a) does not decrease at the same rate as the N/P ratio increases. For N/P < ∼0.2, the proportion of remaining PO₄ tetrahedra is similar for the three glass series. At higher N/P, it follows the sequence NaPON > LiNaPbPON > LiNaPON.

As shown in Fig. 3b, PO₃N tetrahedra appear as soon as nitrogen is incorporated within the glass network, and, in all cases, their proportion versus N/P increases asymptotically up to a maximum value, which depends on the starting glass composition. In the alkali phosphate glasses, the percentages of PO₃N are similar for N/P < ∼0.3, then diverge at higher nitrogen contents to reach a maximum around 36% and 29% in NaPON and LiNaPON glasses, respectively. In LiNaPbPON glasses, much higher percentages of PO₃N are observed from the beginning of nitridation, with a final value of ∼40%.

The three glass series show also a different evolution of their PO₂N₂ tetrahedra when the nitrogen content increases (Fig. 3c). In NaPON and LiNaPON glasses, the percentages of PO₂N₂ only coincide at low nitrogen contents (N/P ∼ 0.15), their increase following a straight line for NaPON with a maximum around 41% for N/P ∼ 0.7, whereas much higher percentages, around 56% for N/P = 0.55, are reached in LiNaPON glasses.

On the other hand, in LiNaPbPON glasses the PO₂N₂ proportion remains close to zero for N/P < ∼0.2, then increases rapidly up to a maximum of ∼34% for N/P = 0.69. Whatever the considered system, the slope of the corresponding variation curve is never decreasing, even at high nitrogen contents.

3.3 ³¹P DQ-MAS NMR

The ³¹P DQ MAS-NMR spectrum of the Li_0.5Na_0.5PO_2.17N_0.55 nitrided glass sample, which is the LiNaPON nitrogen-richest composition, is presented in Fig. 4a. In addition to an auto-correlation peak A–A, at approximately –40 ppm, which is characteristic of a dipolar connectivity between PO₄ tetrahedra, it indicates connectivity between PO₃N and PO₂N₂ tetrahedra (auto- and cross-correlations at –5, –15 and –22 ppm). It reveals also A–B and A–C cross-correlations, i.e. the existence of direct links between pure oxygenated PO₄ tetrahedra and partially nitrided ones.

The ³¹P double quantum spectrum of the Li_0.25Na_0.25Pb_0.25PO_2.25N_0.5 glass (Fig. 4b) shows several differences compared to the LiNaPON glass. It shows a contribution on the diagonal at –21 ppm in the MAS dimension, which can be assigned to PO₄–PO₄ auto-correlation. The two off-diagonal resonances at –22 and –8 ppm can be attributed to cross-correlation peaks due to connectivity between PO₄ and PO₃N tetrahedra. Two other off-diagonal resonances are also detected at –8 and –5 ppm and attributed to PO₃N–PO₂N₂ cross-correlations. At high nitrogen content, N/P = 0.69, the ³¹P DQ-MAS NMR spectrum (Fig. 5) shows that some PO₄–PO₄ connections still remain along with off-diagonal peaks at –7.5 and –3 ppm, corresponding to PO₃N–PO₂N₂ connections. No auto-correlations PO₃N–PO₃N or PO₂N₂–PO₂N₂ are observed in the LiNaPbPON glasses, even at the highest N/P ratios. In addition, a weak resonance is also detected on the isotropic projection, with a chemical shift of +4 ppm in the MAS dimension, which is attributed to the presence of Q¹ sites of pyrophosphate-type 〚25〛.

3.4 N_1s XPS

Fig. 6a,b shows the N_1s XPS spectra of two representative oxynitride glass compositions, Li_0.25Na_0.25Pb_0.25PO_1.97N_0.69 (Fig. 6a) and Li_0.5Na_0.5PO_2.17N_0.55 (Fig. 6b), which can be decomposed into two Gaussians located at ∼ 397.7 and ∼ 399.4 eV. They have been previously assigned to the two different bonding states of nitrogen, i.e., respectively, –N=(N_d) nitrogen atoms, bonded to two phosphorus atoms, and –N<(N_t) nitrogen atoms, bonded to three phosphorus atoms 〚11, 12, 26, 27〛.

The variation of the N_d/P and N_t/P atomic ratios versus increasing N/P (as obtained by XPS) is shown in Fig. 7a and 7b for the three glass series. Both types of nitrogen are always present from the beginning of nitridation, and in all cases the variation of N_d/P or N_t/P follows the same trend: the N_t atoms are first predominant before to become in a minority, because the slope of their variation curve decreases progressively with increasing N/P, whereas that of N_d regularly increases.

As can be seen in Fig. 8, the variation of the N_t/N_d ratio depends on the nature of the glass. Whereas a practically linear decrease is observed for the NaPON glasses, the mixed-alkali glasses, LiNaPON and LiNaPbPON give a similar variation, with in both cases a fast decrease during the first nitridation stage.

4 Discussion

Fig. 1, which represents the nitridation kinetics, clearly shows a dependence on the base glass composition. On the one hand, the time after which the maximum nitrogen content is reached is shorter for LiNaPON and LiNaPbPON glasses, which have lower glass transition temperatures, and therefore lower melt viscosities, than NaPON glasses 〚18, 23〛, which is in accordance with a diffusion-controlled nitridation mechanism. On the other hand, the value of the maximum is clearly a function of the glass composition. In particular, although the nitridation rate is higher in the less viscous LiNaPON glasses than in NaPON glasses, their maximum nitrogen content is lower.

For the three systems, the proportion of PO₃N tetrahedra as a function of the nitrogen content (Fig. 3b) reaches an asymptotic maximum value and, since that of PO₄ tetrahedra decreases regularly along the whole oxynitride domain (Fig. 3a), it looks like that, from a certain N/P value, PO₂N₂ tetrahedra form directly at the expense of PO₄. The interconnections between the different P(O,N)₄ tetrahedra, observed by DQ-MAS NMR, imply existence of both pure oxide and oxynitride microdomains within the glass network. So, the limit in the PO₃N percentage can be explained by a nitrogen/oxygen substitution affecting preferentially bridging oxygen atoms shared by a PO₄ tetrahedron and a PO₃N tetrahedron at the boundaries between microdomains. It results in a simultaneous formation of both a PO₃N and a PO₂N₂ units. In other words, formation of one PO₂N₂ tetrahedron at the expense of a PO₃N is accompanied by formation of another one PO₃N at the expense of a PO₄, and, consequently, the number of PO₃N tetrahedra remains constant. According to this scheme, the N/O substitution would not be a random process during the intermediate and final nitridation stages. However, a purely random substitution cannot be excluded during the first stage.

This nitridation process turns different in LiNaPbPON glasses, which is attributable, of course, to lead atoms. The fast increase in the proportion of PO₃N tetrahedra with increasing N/P, together with an absence of PO₂N₂ before the maximum PO₃N percentage is reached, indicates their preferential formation at the beginning. In addition, there are neither PO₃N–PO₃N nor PO₂N₂–PO₂N₂ connections in the corresponding DQ–MAS NMR spectra. The scheme that we propose in this case 〚24〛 assumes that, at first, PO₄ tetrahedra located in the vicinity of lead atoms are preferentially nitrided and rapidly transformed into PO₃N. Only once all of these PO₄ are converted into PO₃N units, the other PO₄ tetrahedra as well as the PO₃N are substituted to form PO₃N and PO₂N₂ units, respectively. The high PbO concentration of the base glass along with the high coordination number of Pb²⁺ ions would explain the fast increase in PO₃N tetrahedra, as well as their higher percentage at the highest N/P ratios than in the NaPON and LiNaPON glasses.

XPS results show that the N_t species are systematically predominant at the beginning of nitridation. However, the N_t/N_d ratio decreases as a function of nitrogen content, showing a different behaviour depending on glass composition (Fig. 8). The decrease is faster for LiNaPbPON glasses. As it has been already proved 〚17〛, the N_t/N_d ratio does not depend on the melt viscosity, but it should be better correlated with structural aspects induced by the nature of modifier cations. At the beginning of nitridation, the formation of N_t nitrogen atoms, assumed to only affect bridging oxygen atoms (BO), proceeds more rapidly, probably because it is easy to replace BO atoms after the cut of P–O–P bonds within the metaphosphate chains. On the opposite, formation of N_d nitrogen atoms requires both bridging and non-bridging (NBO) oxygen atoms 〚12〛. The progressive decrease observed in the formation rate of N_t atoms is explainable by steric effects resulting from the higher cross-linking they induce compared to N_d atoms. The faster decrease of N_t/N_d in LiNaPON and LiNaPbPON than in NaPON glasses could indicate that the greater structural complexity of a multi-cation glass network, associated to the high coordination number of large modifier cations such as Pb²⁺, is favourable to the formation of N_d species. Further work is now in progress to clarify the structural role of lead by introducing different PbO contents in the Li–Na–Pb–P–O–N system.

On the other hand, the presence in the lead-containing glasses of PO₃N tetrahedra as the only nitrided structural units up to N/P ∼ 0.2 confirms that these PO₃N tetrahedra form from the two kinds of nitrogen, N_d and N_t, since both species coexist from the very beginning of nitridation. An ideal situation for a complete understanding of the nitridation mechanism would be, of course, to closely correlate the NMR and the XPS data, for example to be able to differentiate a PO₃N tetrahedron containing a N_d atom from a PO₃N containing a N_t atom, and to distinguish also between the three different possible PO₂N₂ tetrahedra.

5 Conclusion

In this study, a comparison was made between three different series of oxynitride phosphate glasses, NaPON, LiNaPON and LiNaPbPON, concerning their ability to nitridation and their structure, investigated by NMR and XPS. The results show that, if all the glasses are composed of PO₄, PO₃N and PO₂N₂ coexisting tetrahedral units, with both two- and three-coordinated nitrogen atoms, they can present significant differences which originate either from different melt viscosities or from structural aspects related to the nature of the modifier cations.

The nitridation mechanism proposed for LiNaPON glasses, assumed as an initial random N/O substitution followed by a progressive growing of nitrided microdomains from oxide ones, changes with introduction of lead. In this case, it is thought that the nitridation is not random since its beginning. The Pb²⁺ ions induce nitrogen/oxygen substitution in all the PO₄ tetrahedra surrounding lead atoms, thus giving rise only to PO₃N tetrahedra before any formation of PO₂N₂.

The other relevant difference between these systems is the faster decrease of the N_d/N_t ratio in LiNaPON and LiNaPbPON than in NaPON glasses. This could indicate that the greater is the structural complexity of the glass network and the steric hindrance introduced by large modifier cations such as Pb²⁺, the easier is the formation of N_d nitrogen atoms.

Acknowledgements

This work was financed by MCYT (Spanish Ministry of Science and Technology) through the project MAT 2000-0952-C02-01, and the Integrated Actions CSIC-CNRS 2000FR0009 and HF-2001-124. The authors wish to specially thank Drs. L. Montagne and R. Berjoan for the NMR and XPS experiments.