Astron. Astrophys. 364, 282-292 (2000)

4. Experimental

4.1. Preparation and sample characterization

For the annealing experiments, amorphous silicate materials have to be produced. The following methods for their production have been applied:

4.1.1. Magnesium silicate glass ()

Silicate melts have been produced from a mixture of magnesium carbonate and silica powder in enstatite stoichiometry. The melts were shock-quenched ( 1000 K/s) by pouring the melt through spinning copper rollers. Glassy sheets were obtained of about 100 to 130 thickness. The samples did not show any indications for phase separation or crystalline nucleation.

For powder experiments, the sheets were ground to irregularly shaped particles of about 5 to 10 diameter. Detailed studies of glassy silicate materials have already been published in preceding papers and will not be repeated here (Jäger et al. 1994; Dorschner et al. 1995; Mutschke et al. 1998; Jäger et al. 1998). Glassy Mg-silicates have been considered as a reliable approximation expected for the grain material in space. In IDP's and cometary grains, primitive silicate glass grains (GEMS) of assumed preaccretional origin also occur (Bradley 1994; Bradley et al. 1999a).

4.1.2. Nanometre-sized particles ( and smoke)

Amorphous magnesium silicate smokes were obtained by laser ablation of and targets in 1 atm oxygen atmosphere (Nd:YAG-laser, 6 J/pulse, wavelength 1064 nm).

Using a Jeol 300 kV transmission electron microscope, two differently-sized smoke particle species have been observed in both the and the smoke. The smaller species, about 10 to 50 nm in diametre, had been remarkably magnesium deficient. In the case of the smoke, their composition had been close to . In the smoke, the particles showed a Mg/Si ratio ranging from at least 0.5 to 1. In contrast, the second particle species ranging in size from 0.1 to 2 has been found to be rich in magnesium. The chemical composition of the biggest ones, about 1 to 2 micrometres in size, had been close to that of the ablated target, i.e. the Mg/Si ratio was 1 for the and 2 for the smoke. Generally, the smoke particles consisted of a mixture of rather nonstoichiometric magnesium silicates varying from particle to particle and even inside the particles. Weak absorption features that can be observed in the MIR spectra of both smokes indicate that some forsterite microcrystallites have formed in the predominantly amorphous smoke particles during the laser ablation process (see Fig. 5 and Fig. 7). Independent of the size, the particle shapes were nearly spherical indicating that the particles have most probably experienced a melting phase.

As a frequent characteristic property, most of the particles contained voids that are assumed to have formed by outgassing and by contraction of the originally molten droplets. Formation of voids can also be triggered during observation in the TEM, indicating that a certain amount of gas is still trapped in the amorphous structure and can be released by irradiation (see Fig. 1).

Fig. 1. TEM images of laser-ablated smoke. Spherical micrometre- and nanometre-sized smoke particles are composed of nonstoichiometric Mg-silicates. In the particles with diametres ranging from 10 to 50 nm, voids are contained frequently.

4.2. Annealing experiments

The annealing experiments with glassy and smoke-like silicates have been performed in a Nabertherm HT 04/17 oven in an oxygen atmosphere of 1 atm to prevent evaporative magnesium loss that was observed during annealing in vacuum (Rietmeijer et al. 1986; Karner & Rietmeijer 1996; Hallenbeck et al. 1998).

From the annealed samples of powders embedded in KBr and polyethylene (PE) pellets, IR transmission spectra have been obtained. A Bruker IFS 113 FTIR spectrometer has been used.

4.2.1. Bulk and micrometre-sized glass

Glassy sheets have been annealed as a bulk material at temperatures of 1165, 1121 and 1080 K. The surface evolution was monitored by scanning electron microscopy (SEM) and IR spectroscopy.

Heterogeneous nucleation and crystallization started at the sheet's surface to create a growing orthoenstatite layer. The layer thickness was measured by SEM of the cross section after thermal treatment. To derive kinetic constants, the layer's thickness was measured as a function of annealing time (see Fig. 2). Additionally, time-lag effects of nucleation were observed before crystal growth started. The velocity of linear crystal growth was calculated from the layer thickness (see Table 2). Its values correspond well to the annealing times that have been obtained for powdered glass as will be outlined below.

Fig. 2. Thickness growth of the orthoenstatite layer on a sheet of glass with increasing annealing time for different temperatures.

Table 2. Kinetic data of crystallization of bulk glass

For small particles, shorter annealing times are expected compared with the time for the bulk samples because of the strongly increased particle surface. To investigate the thermal evolution of such small particles, the bulk glass sheets have been ground using an agate mill.

Annealing experiments have been carried out at temperatures ranging from 1121 to 1000 K. In Fig. 3, the evolution of the mass absorption coefficient (MAC) has been monitored as a function of annealing time at 1121 K. Except the experiment at 1000 K, the glass was converted into orthoenstatite as could be verified by IR spectroscopy and XRD analysis. Crystallization has not been observed within 50 h at 1000 K in the IR spectrum, but weak indications of forsterite and tridymite formation have been found in the XRD spectrum.

Fig. 3. Evolution of the MIR spectrum of glass particles (f). At T=1121 K, the annealing times for the spectra are indicated. For comparison, the enstatite spectrum (a, from Jäger et al. 1998) has been added. For clarity, the spectra are vertically shifted: a, b, c, d, e by +3500, +2500, +2000, +1500, +1000 , respectively.

Using Eq. (1) and , the activation energy has been calculated from the annealing times (see Table 3). It is evident that the values of show a systematic decrease with growing annealing temperature. Therefore, Eq. (1) with the used value of is not fully appropriate to describe the annealing process. We proved by plotting ) that the exponential relation fits the measured data. However, the plot indicated that the value of the constant has to be 4 to 8 orders of magnitude larger. From our limited number of experiments, we cannot determine the constant exactly. In addition, the constant probably depends on the chemical composition and has to be determined for each smoke/glass independently. For the sake of consistency to Lenzuni et al. (1995), in this paper, we will further use their value. However, so long as the constant has not been determined experimentally, one should bear in mind that the numerical value of the activation energy depends on the estimation for .

Table 3. Annealing time and activation energy determined for glassy micrometre-sized particles using Eq. (1) and the constant

The annealing times obtained for glass powder are compatible to the growth velocities that have been obtained from the bulk glass. For example, at 1080 K, the growth velocity was determined to be 72 nm/min. At this velocity, surface crystal growth would fully crystallize a particle of 5 m in diametre in approximately 70 min. This is in the range of the annealing time measured for the glass particles. We conclude that the velocity of crystal growth does not depend on the particle size. In contrast, nucleation strongly depends on the surface area available and therefore determines the time-lag before crystal growth starts.

The opacity of untreated and annealed powder has been obtained in the FIR range up to . After 1 hour annealing at 1121 K, a distinct drop in opacity has occurred below 200 . An additional drop was observed in the wavenumber range after 2 hours of annealing (see Fig. 4, spectra b and c).

Fig. 4. Comparison of the IR spectrum of thermally untreated glass powder (a) with the spectra of glass powder that has been annealed at T=1121 K for 1 and 2 h (b, c).

Using Eq. (4), the spectral indices of powders have been estimated by least-square fits of the data in the range from 200 to 50 . The parameters are and for the untreated amorphous powder. For the sample that was annealed at 1121 K for 2 hours, the parameters are and . As can be seen in Fig. 4, a further opacity drop occurs between 1 and 2 h of annealing. That drop may be caused by the continuous growth of the biggest crystalline regions that "consume" smaller nanocrystals. Another effect could be the "healing" of lattice defects in the nanocrystals.

4.2.2. Annealing of smoke

The smoke has been annealed at 1000 K to investigate the behaviour as a function of annealing time. Using IR spectroscopy and TEM, significant structural evolution of the smoke particles has not been observed prior to crystallization. In detail, TEM imaging showed that the texture with voids and rims was still preserved (see Fig. 1). Smoke evolution could not be characterized by a distinct stall phase as was observed by Hallenbeck et al. (1998). In contrast to the laser-ablated smoke described in this study, the Hallenbeck smoke was prepared at 770 K from Mg metal placed inside a furnace tube and a 1:1:4 flowing gas mixture of , and He at a total pressure of 80 Torr. Although both laser-ablated smoke and smoke produced in a gas-flow-reactor are nonstoichiometric condensates, the latter is dominated by micrometre-sized entities of pure silica with a nanometre-sized mantle of chaotic Mg-silicate whereas our laser-ablated smoke consists of different-sized particles that are Mg-enriched or depleted according to their grain size. There is a basic compositional difference. For these reasons, both products cannot be compared with each other. That means that the stall phase need not be expected in all varieties of smoke particles.

Our results are shown in Fig. 5. The structure in the wide 10 m profile of spectrum d indicates that some predominantly amorphous particles already contained forsterite microcrystallites. The main component at 1050-1100 has been assigned to amorphous silica and the shoulder at 900 to forsterite. The peaks of spectrum a in the 10 m range at 843, 890, 960 and 1000 can be identified with those of forsterite, whereas a shoulder at 1090 has been assigned to amorphous silica.

Fig. 5. Evolution of the MIR spectrum of smoke after annealing at 1000 K up to 30 h. The annealing times are indicated. For clarity, the spectra are vertically shifted: a, b, c by +1500, +1000, +750 , respectively.

As Fig. 5 demonstrates, rapid crystallization sets in not earlier then after 21 h of annealing. Therefore, the true annealing time used to calculate the activation energy has to be in the range between 21 and 30 h and is estimated to be 25 . Using Eq. (1) and the mean frequency , the activation energy of the smoke has been determined to be = 42040 150 K.

In the FIR, a significant opacity drop occurs between 250 and 75 . At 283 and 325 , tridymite features have been observed (see Fig. 6, spectrum b). Hence, the `30 h' smoke consisted of polycrystalline forsterite, tridymite and amorphous silica (see Fig. 6 and Fig. 9).

Fig. 6. Comparison of the IR spectrum of thermally untreated smoke (a) with the spectrum of smoke that has been annealed at T=1000 K for 30 h (b).

Annealing experiments using laser ablated smoke from a natural pyroxene mineral were already carried out by Brucato et al. (1999). In contrast to our study, the peaks of the annealed samples fall nearly at the same positions as for the crystalline pyroxene, i.e. the smoke changed into crystalline pyroxene whereas our smoke transformed into forsterite and silica. Presumably, the presence of iron promotes pyroxene formation; but iron-free Mg-silicate smokes change into forsterite and silica. The activation energies given by Brucato et al. (1999) (47500 K) are not comparable to the ones determined here since they are based on a different definition of the annealing time. In contrast to the Lenzuni et al. (1995) perception, they define the anneling time, when the IR spectra do not change any longer, i.e. when the sample is crystallized completely.

4.2.3. Annealing of smoke

The smoke has been annealed at 1000 K up to 30 h and at 1206 K for 1 h. The results are presented in Fig. 7 and Fig. 8. The IR spectra proved the smoke to be converted into crystalline forsterite. The values of the MAC of annealed smoke coincide to the MAC of synthetic forsterite (Jäger et al. 1998). Because the smaller particle species had been magnesium deficient, some amorphous silica and MgO as well as crystalline has to be contained in the smoke; but the weak amorphous features are superimposed by the strong crystalline peaks.

Fig. 7. Evolution of the MIR spectrum of smoke after annealing at 1000 K up to 30 h. The annealing times are indicated. For clarity, the spectra are vertically shifted: a, b, c, d, e by +5000, +4000, +3000, +2000, +1000 , respectively.

Fig. 8. Comparison of the IR spectrum of thermally untreated smoke (a) with the spectrum of smoke that has been annealed at T=1000 K for 30 h. (b)

Fig. 9. The IR spectrum of annealed and smoke (spectra a, b) compared with that of a- (c) and c- (d) (Jäger et al. 1998). For clarity, spectra a is vertically shifted by . The spectra c and d are in arbitrary units.

As can be seen in Fig. 7, the smoke starts to evolve significantly towards a polycrystalline material already after an annealing time of 1 h. From this spectra, we estimate the annealing time to be in the range between 1 and 2 h. Using Eq. (1) and the mean frequency , the activation energy of the smoke has been determined to be = 39100 400 K.

In the FIR, a significant opacity drop occurs below 250 (see Fig. 8). The overall opacity falls below that of the annealed smoke. Due to the higher initial magnesium content, most of the smoke has been transformed into crystalline forsterite. Otherwise, less amorphous silica and MgO is left that would significantly raise the FIR absorption.

4.2.4. Summary of and smoke evolution

From our results, we conclude that the annealing of Mg-silicate smokes in oxygen-rich environments leads to polycrystalline c-. If the Mg/Si-ratio falls below 2, the excess Si appears as tridymite and/or amorphous . The laser-ablated smoke had been rather inhomogeneous with two particle species, a small nanometre-sized fraction (10 to 50 nm in diametre, Mg-deficient) and a micrometre-sized fraction (0.1 to 2 in diametre, Mg-enriched). For these reasons, the presence of silica or crystalline and MgO had to be expected even in the smoke of forsterite stoichiometry.

In the TEM, a profound difference between the annealing behaviour of the and the smoke could not be observed. During annealing at 1000 K up to 30 h, the smaller-sized smoke particles did not show observable changes. Although these Mg-poor particles seemed almost unchanged, electron diffraction analyses unvailed their partly crystalline structure. Hence, structural order has developed on a nanometre scale (see Fig. 10). These results are essentially the same for both the and the smoke. The percentage of crystalline forsterite depends on the initial Mg/Si ratio of the particles that had been different for the and the smoke. At 1206 K, nanocrystals have been observed showing that long range order has developed (see Fig. 11). The Mg-enriched micrometre-sized particles had been almost totally crystalline as was monitored by electron diffraction. Their outer appearance as sperical droplets did not change. According to their inital composition (Mg/Si 1 for the smoke and 2 for the smoke), they most probably consisted of a mixture of crystalline forsterite and amorphous/crystalline silica or forsterite.

Fig. 10. TEM image of smaller-sized particles of the smoke after annealing up to 30 h at 1000 K. The Mg-poor particles that are characterized by a Mg/Si-ratio of about 1 seem to be structurally unchanged. However, they are partly crystalline as was checked by electron diffraction analyses.

Fig. 11. TEM image of smoke that has been annealed at 1206 K for 1 h. Forsterite single crystals are embedded in a matrix made up of polycrystalline forsterite, crystalline and amorphous silica and MgO.

From the experiments, we found a significant decrease with increasing Mg/Si-ratio in the annealing time and therefore in the activation energy. For the smoke, we found = 42040 150 K. For the smoke, has been determined to be 39100 400 K. Because of the Mg-deficiency in the smoke, the diffusion rate of Mg-atoms limits the growth rate of the forsterite nanocrystals - increasing the macroscopically measured annealing time.

4.2.5. Annealing of amorphous silica nanoparticles

Amorphous silica is a primary condensate formed from MgO- or FeO- vapour at non-equilibrium conditions (Rietmeijer et al. 1999).

To investigate the evolution of pure condensates, nanoparticles of precipitated silica (commercial product, Fisher Scientific Ltd.) with diametres of 10 to 50 nm have been annealed at temperatures between 1150 and 1350 K. Their thermal evolution has been monitored by the appearance and the sharpening of the crystalline cristobalite and tridymite features in the infrared spectra (see Fig. 13 and Fig. 12). IR and XRD analysis proved cristobalite to be the major component. Peak positions and the full widths of half maximum (FWHM) of spectral features have been listed in Table 4.

Fig. 12. Evolution of the MIR spectrum of silica that has been annealed at 1220 K. The annealing times are indicated. For clarity, the spectra have been shifted vertically: a, b, c, d, e by +7500, +6000, +4500, +3000, +2500 , respectively. has been approximated to 4.5 h.

Fig. 13. Comparison of the IR spectrum of amorphous (a) with that of annealed silica (b) (T= 1220 K, 5 h).

Table 4. Crystalline features in annealed silica (T `tridymite feature', C = `cristobalite feature' (Hofmeister & Rose 1992, Nyquist 1997))

At a temperature of 1220 K, the annealing time has been estimated to be 4.5 0.5 h. Using Eq. (1) and , the activation energy has been calculated to be = 49190 150 K. The diffusion constant D has been calculated as well (see Eq. (3))

In Eq. (3), the constant a has been approximated to be Å from the volume of the basic molecule forming the lattice (). is the molecular weight of , u the atomic mass unit and the mass density of a-. Using the diffusion constant D and Eq. (2), the size of locally ordered structures can be estimated. At the annealing time of 4.5 h at 1220 K, it equals Å. However, electron diffraction analyses showed that nanometre-sized crystals are present in the sample. Thus, the application of the formulas given in Sect. 3 is limited to the nucleation and should not be applied to the further growth of nanometre-sized crystals.

Crystallization of amorphous silica nanoparticles is associated with a sharpening of absorption features in the MIR and an opacity drop in the FIR (see Fig. 12 and Fig. 13). The mass absorption coefficient in the FIR has been approximated by a power law (using Eq. 4). The spectral indices of silica have been estimated in the wavenumber range from 100 to 15 . For a-, the parameters are and . The parameters are and for the silica that has been annealed at 1220 K for 5 hours.

© European Southern Observatory (ESO) 2000

Online publication: December 15, 2000
helpdesk.link@springer.de