2 Experimental details

As a starting point in this work we have chosen an amorphous silicate manufactured so as to have the same stoichiometry as the Mg-rich pyroxene end member enstatite. We shall in time be broadening the scope of our experiments to include other silicate compositions (including iron silicates) and will report the results of these investigations in subsequent papers. Estimates of the Mg/Si ratio for Comet Halley, obtained from the PUMA-1 mass spectrometer on board the Vega-1 spacecraft, suggest a value between $\sim$0.5 and $\sim$1.6 (Jessberger et al. 1988; Lawler et al. 1989). The choice of amorphous MgSiO₃ as a starting material is therefore broadly typical of the type of composition that may be expected in many comets.

2.1 Sample manufacture

The silicate sample used in these experiments was produced using the well known gel desiccation method (Sabatier 1950; Day 1974) according to the following prescription:

1.

0.1 M solutions of the soluble metal salts MgCl₂ and Na₂SiO₃ (sodium metasilicate) were mixed in the correct stoichiometric ratio to produce MgSiO₃.

2.

The white suspension thus formed was left to settle for two days. The excess liquid was then decanted off and the remaining suspension centrifuged and washed with distilled water. This centrifuging/washing cycle was repeated three times.

3.

The resulting gelatinous precipitate was then dried in air over a hot plate (100-150$^{\circ }$C) to yield lumps of a white glassy-looking solid.

The silicate forms as a hydrated precipitate according to the reaction,

$${\mbox{Na}_{2}\mbox{SiO}_{3} + \mbox{MgCl}_{2} + \mbox{H}_{2}\mbox{O} \rightarrow}$$

$${\mbox{MgSiO}_{3}.x\mbox{H}_{2}\mbox{O} + 2\mbox{NaCl} + y\mbox{H}_{2}\mbox{O}}$$ (1)

with the sodium salt being removed from the sample by the centrifuging/washing cycle.

To prepare the sample for presentation to both the synchrotron beam and IR spectrometer, the raw silicate was ground by hand in a mortar and pestle to give a fine-grained powder. SEM micrographs showed the particles to be typically $\sim$100 $\mu$m with some as large as several hundreds of microns, while a limited number of particles also occupy the range from a few microns to a few tens of microns. Despite observing a substantial mass reduction during annealing, the SEM data showed no evidence of a change in particle size between processed and unprocessed samples. This method of preparation does produce particles that are rather large in comparison to the size of the cometary grains supposed to be responsible for the bands observed in the 10-20 $\mu$m spectra of such bodies (Brucato et al. 1999a). However, we think this difference is not of primary importance for the scope of the present work. Although Rietmeijer et al. (2002) have observed a size dependency in the annealing behaviour of silicate smokes, the particle sizes involved their experiments were significantly smaller than in the work discussed here. We believe therefore that the structural changes discussed in this paper are unlikely to be strongly dependent on particle size. However, we acknowledge that this point is deserving of further attention and we plan to investigate this possibility in future experiments.

2.2 Synchrotron X-ray powder diffraction

Based on the parallel beam optics of Parrish et al. (1986), the Daresbury Laboratory synchrotron radiation source (srs) station 2.3 diffractometer used in our experiment was originally constructed for ambient high-resolution powder diffraction studies (Cernik et al. 1990; Collins et al. 1992). Located 15 m tangentially from a 1.2 T dipole magnet in the 2 GeV electron storage ring (Munro 1997), it receives X-rays in the range 0.7-2.5 Å. These are filtered by a water cooled Si(111) channel-cut single crystal to give a monochromatic beam incident at the centre of the two circle ($\theta$ and 2$\theta$) diffractometer. The sample furnace used in the diffraction experiment is mounted on the diffractometer's $\theta$-circle with a flat-plate sample holder inside the device coincident with the centre of the $\theta$-circle allowing Hart-Parrish diffraction geometry to be achieved (Hart & Parrish 1986). This makes the diffraction optics insensitive to changes in sample height, which is essential as small movements are inevitable when the sample is heated. The whole furnace assembly is enclosed in a stainless steel body, with incident and diffracted X-rays passing through kapton entrance and exit windows allowing measurements to be made during annealing. The diffracted beam passes through a parallel foil assembly on the 2$\theta$ arm and is detected using an enhanced dynamic range scintillation counter.

The furnace itself is based on a design by Debrenne et al. (1970), details of which are given in Tang et al. (1998). The sample crucible is made of molybdenum, chosen for its high melting point and induction characteristics, while a 1 mm deep and 15 mm diameter pressed platinum former is placed on the crucible to hold the sample and to prevent possible chemical reaction between sample and crucible at high temperatures. Heating is via a water-cooled 2 kW RF copper coil regulated by a Eurotherm 900 controller. Sample temperature is measured by a tungsten-rhenium thermocouple placed at the sample/crucible assembly. The operational temperature range is 290-2000 K with a heating response time of $\sim$30 s. Even at high temperatures, $\pm$1 K stability is achieved in under a few minutes. In order to ramp our sample up to the 1000 K annealing temperature, the furnace temperature was increased in steps of $\sim$200 K over a period of approximately 5 min, with the approach to 1000 K being made in progressively smaller increments to avoid overshoot of the target temperature. Data collection began as soon as the temperature stabilised at 1000 K (approximately 2-3 min).

Normally $\theta$-$2\theta$ XRD patterns are collected by synchronised rotation of the $\theta$ and 2$\theta$ arms. However to prevent the sample falling from the holder at high angle, the $\theta$ circle was fixed so that the sample was inclined at 10$^{\circ }$ to the horizontal incoming beam and the diffraction intensities corrected accordingly. An X-ray wavelength of 1.2995 Å (calibrated against a Si powder standard) was selected as a good compromise between peak incident flux and the requirement for low wavelength for increased X-ray reciprocal wave vector k-space sampling within the silicate. For a given X-ray wavelength, $\lambda$, this is given by

$$k=\frac{4\pi\sin\theta}{\lambda},$$ (2)

where $\theta$ is the X-ray scattering angle. The significance of this relation is that the whole of the material's reciprocal space can only be sampled by the X-ray beam as the wavelength approaches zero. For srs station 2.3, the flux delivered to the diffractometer by the station's monochromator decreases significantly towards low wavelengths, thus in order to obtain good counting statistics at such short wavelengths, the integration time per point would have to be increased accordingly. This would have the additional effect of lowering further the temporal resolution of the experiment, or would have to be compensated for by reducing the angular range (which would also limit the available k-space), or by reducing the angular resolution which would reduce the k-space resolution and hence the long-range resolution in real-space (long-range oscillations in k-space correspond to short-range variations in real-space). Given these considerations and the pragmatic choice of wavelength we were able to employ scan parameters of a 2$\theta$ step size of 10 mdeg and detector integration time of 1 second per point throughout the experiment, giving a scan time of approximately 75 min for an angular range 15-60$^{\circ }$ 2$\theta$.

2.3 IR spectroscopy

In order to perform spectroscopic measurements, the samples were prepared according to the standard pellet technique as described by Borghesi et al. (1985) and Bussoletti et al. (1987). The technique is based on the dispersion of a known quantity of sample in an IR transparent matrix (KBr in our case). The resulting mixture is then compressed into a solid pellet and the sample spectrum recovered by comparison with a pellet of pure KBr.

Figure 1: Typical profile of the measured annealing temperature as a function of time for the samples used in the IR spectroscopic measurements. $T_{\rm set}$ is the nominal target temperature while $T_{\rm eff}$ is defined as 85% of $T_{\rm set}$ (in $^{\circ }$C). Annealing begins at time t=0 when the furnace is switched on. The effective temperature $T_{\rm eff}$ is reached at time t=t1 and the target set point temperature $T_{\rm set}$ is reached at t=t2. The furnace is switched off at t=t3 and allowed to cool naturally. Finally the sample temperature crosses $T_{\rm eff}$ at t=t4. The annealing time is thus defined as the time the sample has spent at or above $T_{\rm eff}$.

With the obvious exclusion of the unprocessed sample, all the others were annealed in a Carbolite furnace, model CTF 12/65, capable of reaching a maximum temperature of 1200 $^{\circ }$C. The sample compartment consists of an alumina tube approximately 700 mm in length and 75 mm in diameter, with the sample being placed at the centre of the tube on a small alumina plate. In addition to the standard furnace thermocouple we placed an additional NiCroSil/NiSil thermocouple directly in contact with the sample plate so that the annealing temperatures quoted are, to within a very close approximation, the actual temperatures experienced by each sample. During annealing, the furnace was evacuated to prevent possible reaction of the hot sample with atmospheric gases. Annealing began only when the pressure inside the furnace was less than 10^-4 mbar. After an initial increase, the pressure remained between $2 \times 10^{-5}$ and $5\times 10^{-5}$ mbar for most of the time. To further reduce the likelihood of interaction with the atmosphere, the sample was allowed to cool to ambient temperature before opening the furnace. Unfortunately the Carbolite furnace has a significant thermal inertia with no cooling system fitted. The typical annealing profile of the measured temperature at the sample plate as a function of time is shown in Fig. 1. The furnace is switched on at time t=0 and switched off at time t₃ once the nominal set-point temperature, $T_{\rm set}$, has been maintained for a required period of time $\Delta t=t_{3}-t_{2}$. As can be seen from the figure, the rise time for the temperature is relatively short (between 30 and 45 minutes) depending on the value of  $T_{\rm set}$, whilst the cool-down time to ambient temperature can take more than 24 hours. The annealing times quoted in this paper therefore have been defined as the interval between the time t₁ at which the furnace reaches 85% of of the target temperature $T_{\rm set}$ (measured in $^{\circ }$C) and the time, t₄, when the furnace subsequently cools back down to the same temperature. We are aware that such a definition may be viewed as somewhat arbitrary and that a better determination of this important parameter may be desirable. Nevertheless in the present discussion it can probably be accepted without any major disadvantage for two reasons. Firstly we have used only two basic annealing times which, in terms of the scope of the present discussion, can be simply defined as short (from 2 to 4 hours) and long (from 20 to 24 hours). Secondly the data we have obtained appears to depend quite strongly on temperature while the influence of the annealing time seems to be much weaker.

Following removal from the furnace the samples were embedded in KBr, with a pure KBr pellet also being manufactured at the same time to allow immediate subtraction of the matrix contribution to be made for spectra collected under similar conditions. The spectra themselves were recorded using a Spectrum 2000 Perkin-Elmer FT-IR spectrometer. The single beam Michelson interferometer was equipped, for the spectral range of interest (4000-400 cm^-1; 2.5-25 $\mu$m), with a KBr beamsplitter, a FR-DTGS (Fast Recovery Deuterated TriGlycine Sulphate) detector and a wire coil source at 1350 K. A resolution of 4 cm^-1 was selected as being high enough for the convenient detection of solid state absorption features, while the measured interferograms were directly transformed in spectra using a FFT (Fast Fourier Transform) algorithm that forms part of the Spectrum for Windows software package supplied by Perkin-Elmer for operating the whole instrument.