Chemical behavior of aluminum and phosphorus during dissolution of glass fibers in physiological saline solutions.

The dissolution of textile glass fibers of four different compositions has been investigated at 37 degrees C. In these glasses, which are isolation type, the P2O5 contents scatter between 0 and 2 wt% and Al2O3 from 0.12 to 3.4 wt%. Both static (30-mg fibers; 250-ml solution) and dynamic (50-mg fibers; 40 ml/day flow rate) conditions with or without bubbling of a gas mixture (95:5, N2-CO2) have been used. Two physiological solutions, one proposed by Kanapilly and the other by Scholtze, were used. After each run (1, 3, 7, 14, and sometimes 30, 62 days) the solutions were analyzed for B and Si by inductively coupled plasma (ICP), the weight losses were determined, and the residual solid were observed by scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy (XPS). Static runs give a better agreement between measured and calculated weight losses from solution analyses than dynamic experiments. SEM examinations indicate diameter reduction and formation of a hydrated Si-rich layer. Sometimes hollow tubes, suggesting the detachment of these layers, are observed. XPS and energy dispersive X-ray (EDX) analysis indicate the formation of a veneer of calcium phosphate for the most rapidly dissolving glass. In other cases an Al increase is observed at the solid solution interface. Whatever experimental conditions are used, the relative dissolution rates of the four glasses are identical. The kinetics may be modeled with variable dissolution rates from initial high values to final low ones. The latter reflect the very low solubility of the residual product.


Introduction
The idea that the risk of carcinogenesis related to fiber inhalation is decreased if these materials are rapidly dissolved in the lung is now widespread. For this reason, the in vitro dissolution rate of glass fibers in saline solutions has been measured by many authors (1)(2)(3)(4)(5). However, uncertainties remain about the chemical composition of the leached layer at the interface solid-solution specifically when glasses contain variable amounts of Al and P. These data, which can be supplied by surface analysis, are necessary to form a realistic picture of the dissolution of the fibers.
Among the parameters that can modify the dissolution rate, changes in the glass composition greatly affect the clearance of fiber. This study aimed to characterize modifications of the chemical durability of standard isolation fiber glass when the amount of Al is reduced and when extra components, such as P, have been added in This  very small amounts, and also to compare the dissolution processes of various glasses in physiological saline solution and to establish a general kinetic model.

Materials and Experimental Procedure Materials
Glasses of four compositions were studied (Table 1). One is a commercially available isolation glass. The others are derivatives with lower percentages of A1203 and added percentages of P205. Fibers of uniform diameter (10 pm) were obtained by the textile process from a one-hole bushing in the Research Center of Saint-Gobain (Aubervilliers, France). Two physiological saline solutions derived from the Gamble's solution (6) were used. The first, referred to as solution L, was developed by Kanapilly (7) and used by Leineweber (8) and Feck (3). The second, solution W, which contains more calcium ions, was used by Scholze and Conradt (1). In both cases, 0.1% of formaldehyde was added to prevent microorganism proliferation. The chemical composition of solutions is shown in Table  2. The initial pH of both solutions ranged from 7.6 to 8.0.

Experimental Procedures
The dissolution of glass fibers was studied under three experimental leaching conditions: static mode, unbuffered dynamic mode, and buffered dynamic mode.
In the first mode, fibers approximately 20 mm long were laid on a large teflon grid in a teflon waterproof vessel. The weight of fibers was 30 mg and the volume of solution was 250 ml, to reach only approximately 60% saturation with respect to amorphous silica, which in the case of complete dissolution of the fibers would be about 140 mg/l (9). The vessels thermostatted at 37°C ± 10C and the runs lasted from 1 to 62 days. At the end of the experiment, fibers were removed from the solution, rinsed in deionized water, and air dried before weighing. The solution was filtered on a 0.22-pm pore diameter filter, and analyzed by inductively coupled plasma (ICP) or atomic absorption spectrometry (AAS).
In the second mode, 50 mg of fibers approximately 5 mm in length, were laid on a 0.22 ,um pore diameter filter in a cell volume of 1.2 ml thermostatted at 37°C + 1°C. The solution was delivered by a peristaltic pump at a flow rate of 40 ml/day. After each run, the fibers were removed from the filter for determination of weight loss and analysis. The fluid collected during the leaching was analyzed by ICP or AAS. The experimental conditions in the third mode were identical except that the pH in the supply container was buffered by bubbling through the solution a N2-CO2 gas mixture (95:5 by volume). pH measurements were made after each run, and for dynamic leaching the pH in the supply container was checked during the experiments. SEM analysis was made either directly on fibers or on polished sections of fibers embedded in epoxy resin. Fresh samples were used for each experiment.

Experiments in Static Mode
The weight loss with time of four fibers treated in the Kanapilly solution (L) is shown in Figure 1. As expected, the two more soluble fibers have the lowest alu-[ C3 minum content (CM2 and CM19), while CM19 comparing CM19 with CM2, and CM25 * CM2 5 with C3, it is evident that phosphate -1-60increases the dissolution rate. During wl (%) leaching, the pH increased slightly due to icon (in per-ion exchange, but it remained at around solution L. 8.8. For the most rapidly dissolving glass s the line of (CM 19) the weight losses at 14, 30, and 62 days were virtually the same. SEM observa-tions showed a regular decrease in fiber diameter for CM19 and CM2 with the reaction time, reaching a diameter of about 2 gm, corresponding to 95% weight loss.
At this stage (CM 19, t = 14 days), irregular shapes were observed on polished sections of epoxy resin-embedded fibers (Figure 2a). Some fibers were completely leached, and energy dispersive X-ray (EDX) analyses of such fibers indicate that the residual product is essentially siliceous, strongly enriched in aluminium. The Al/Si atomic ratio, which was 0.003 in the initial fiber, has risen to 0.13. In contrast, the C3 and CM25 fibers after 14 days' treatment, showed on polished sections, a fine hydrated layer surrounding an unaltered circular core (Figure 2b). The thickness of this layer is less for C3 glass (0.2 pm) than for CM25 (0.5 pm). Dissolved silicon and boron in the solutions were plotted against weight loss (Figures 3,4). The correlation between the fraction of dissolved silicon and the relative weight loss is good, indicating that the contribution of the leached layer is negligible compared with the fraction of dissolved glass when weight loss is less than 80%. At higher values, a slight discrepancy is observed indicating that Experiments are in static mode. The curve is the line of unit slope.  face analysis by X-ray photoelectron spectroscopy (XPS) relative to a layer <10 nm thick, are given in Table 3; B, Ca, Mg, and P are rapidly leached from the surface veneer, but Na remains in the interface at about half of the initial Na/Si ratio in the glass. This is probably due to the high concentration of Na4 ions in the solution, rather than to precipitation of Na-salts. In contrast, Al is markedly enriched for all glasses and for each duration. The surface compositions of the residual products determined by XPS are not in complete concordance with those observed by EDX. The XPS approach suggests an "albite-like" gel composition with Al/Si and Na/Si atomic ratios at about A:3 tO A:4. (The composition of crystallized albite is NaAlSi3O8.) The EDX data indicate lower Al/Si and Na/Si atomic ratios. Actually, there is probably a concentration gradient, with the interface being richer in aluminium and sodium.

Experments in Unbuffered Dynamc Mode
The mean pH of the effluent solution is 8.8, similar to the pH observed in experiments in the static mode. The weight losses observed in solution L in both dynamic and static modes for three glasses are similar ( Table 4). The kinetic curves, however, are more irregular in dynamic mode experiments, probably due to a more or less high degree of leaching of the fibers in the cell, related to the variation of the flow within the fiber mat. SEM observations reveal a * C3 variety of morphologies for the residual * CM29 product of CM2 and CM19 fibers. For [!iMJ 5 example, the CM19 glass that has been leached for 28 days shows tubular shapes, whose diameter is close to that of the origi-L/S250 nal fiber. Apparently the dissolution of the fiber continues within this residual outer losses obained layer, possibly from the base of the cylinstatic mode in der. The chemical composition of these tubes is not well established, but EDX analysis indicates a high concentration of calcium, confirmed by XPS  ments, it appears that in the latter, an accumulation of ions (Ca and P) can occur )ers leached in around the fiber, leading, in a few cases, to iments are in the formation of a sparingly soluble phase. are duplicated.
Runs using the solution W richer in Ca2+ ions, confirm the formation of surface calcium phosphate which reduces the dissolution rate of the most soluble glass (CM19). In this regard, Figure 5 clearly illustrates the role of the composition of the solution. XPS analysis indicates an apatite-like composition (P/Ca atomic ratio near 0.6) of the surface of such altered fibers.

Experiments in Buffered Dynamic Mode
Runs were performed using the Ca2+-rich solution W, buffered by bubbling through CO2. The pH of the effluent solution was only 0.2 pH units lower than it would have been without buffering by bubbled CO2. The weight loss against duration for four glasses is shown in Figure 6. Runs at 14 days have been duplicated to test the reproducibility of the experiments. The most rapidly soluble glass (CM19) displays wide variation of the weight loss while for other glasses the results were much more reproducible.
Surface analysis by XPS (Table 6) reveals the development of calcium phosphate (with an apatite-like P/Ca ratio) in the Si-rich outer layer of the CM19 glass. For other fibers, the Si gel outer layer is highly enriched in aluminium, but the sodium concentration is halved in comparison to the initial concentration in the glass. SEM observations on polished sections reveal a variety of morphologies among dissolved fibers. C3 and CM25 glasses show an unchanged core surrounded by a leached layer. For CM25 the mean diameter of the core at 14 days is 6.2 gm. From the bulk diameter (core + layer = 7.4 ,um) weight loss of 45% can be calculated assuming identical Si volume concentrations in glass and leached layer. This is identical with the result obtained by weighing the altered fibers. CM2 fibers ( Figure  7a) have a leached layer of about 0.3 pm (compared with 0.6 pm for CM25) and a residual core that is not completely circular, resulting, perhaps, from a local higher pH in stagnant zones between the fibers. From the mean diameter of these fibers, about 5 pm, a weight loss of 75% can be calculated, which is in good agreement with the figure of 71% obtained by weighing. EDX analysis indicates a higher concentration in Al in the leached layer for CM25 glass. This enrichment is less marked for CM2 glass probably because the leached layer is not so thick.

Dissolution Rate Modeling
Fibers are assumed to have a uniform initial diameter of 10 pm and the area of the extremities is neglected. The dissolution rate parameter k is supposed to vary from an initial rate value k0, corresponding to the glass, to the dissolution rate constant of the residual product k,. The shift from ko to ki can be described by an exponential expression as developed by Barkatt (10). k=ko e-a,'+ k (l-e-a') k in,um/hr [1] where oc is a constant in h-1 Limit conditions are: t -> k ->ko;t -oo k -->k, The dissolution rate parameter k is expressed as the thickness dissolved per time using radial coordinates k dr [2] dt where r is the radius of the fiber in gm.
Combining equations 1 and 2 gives -dr = [ko e-t+ k, (1 -eat)] dt [3] The integration of the equation 3 gives the variation of the radius of the fiber with time r = R [4] where Ro is the initial radius of the fiber. This relation can also be expressed as weight loss (_ M) 1/2 k,t _(kok, )(1-ee-at) [5] where MO is the initial weight of the fiber and AM the weight loss at time t. [6] M j1/2

MO) Ro
Where the dissolution rate parameter of the residual product k, is very low in respect to the initial dissolution rate k, Equation 6 would describe the dissolution kinetics of a sparingly soluble glass at the early stage of dissolution. Equation 7 models the dissolution kinetics of rapidly dissolving glass that leaves an insoluble fraction. The higher the parameter a, the more rapid is the variation of the dissolution rate with time.
M1 =MO r1 ko [8] The mass of the residual fraction, M, is given by: A nonlinear least square method was used to fit equation 9, deduced from equation 7, with weight loss data for the static mode and the L solution. and C=cte (c=1) [9] The computed values for A, B, and C parameters in equation 9 for the four glasses are given in Table 7. The fit is good for all glasses (Figure 8), although with more experimental data the computations would be more accurate.

Conclusion
From the data obtained from the residual fibers by XPS and SEM, we can postulate the dissolution mechanisms of the fibers in physiological saline solutions. Two types of solid-solution interactions may be considered. In the first, the unaltered core of the fiber is surrounded by a leached Si-gel layer and the diameter of the fiber decreases regularly with time. The dissolution process slows down greatly only when the diameter of the fiber becomes very small. In the second type, concentrations of dissolved calcium and phosphate, initially in the solution and resulting from glass dissolution, exceed the solubility product of an apatite-like calcium phosphate. This leads to the precipitation on, or the formation into, the Si-rich surface layer of a sparingly soluble calcium phosphate, which effectively reduces the measured weight loss. Moreover, the rate of dissolution of the inner glass probably would be lowered due to the protective effect of a calcium phosphate layer.
Only the interaction of the first type can be easily modeled because the geometric parameters of the leaching device are then not so important. If no precipitations of new phases occur during leaching, the dissolution rate of the fiber is, after an initial stage where ion exchange predominates, controlled by the solubilization of the hydrated silica-rich layer developed around the fiber. The composition of this amorphous layer does not remain constant because some cations, particularly Al ions, remain concentrated in the layer near the interface. In early stages, Al ions are only concentrated in the outer region of the hydrated layer. During leaching more and more Al ions are taken up by the leached layer and the thickness of the region of high concentration of Al increases with time. Finally, a residual product is formed, seen by XPS to be a sparingly soluble amorphous alumino-silicate gel with additional sodium. The gel at the interface has an albite-like composition (NaAlSi3O8) while the bulk is essentially siliceous with less aluminium than in albite. This result strongly suggests that there is a gradient of concentration of aluminium and sodium from the interface to the bulk of the residual product.
In conclusion, the roles of Al and P are well established in these experiments. During in vitro aqueous corrosion, aluminum remains in the residual Si-gel layer, probably creating new chemical bonds with silicon, when the dissolution-polymerization process occurs, resulting in a more stable Si-Al-gel. Phosphorus, which is important for the formation of the glass network, is easily leached in saline solution; this leads to the fragilization of the glass network during the dissolution process by increasing the hydrolysis of the silica. Under special conditions, when Ca and P04 ion concentrations are high, precipitation of calcium phosphate take place on the surface of the fibers. Although any correlation with in vivo experiments remains questionable (11), it is probable that the initial dissolution processes of Al and P would be similar to those observed in vitro. However, surface accumulation of Al in vivo may well be counteracted by Al complex formation by chelating organic molecules present in physiological media which could facilitate the total dissolution of the leached residue.