Freezing Experiments on Clay Gels - American Chemical Society

J. Swenson. Department of Applied Physics, Chalmers University of Technology,. S-412 96 Go¨teborg, Sweden. A. C. Hannon and S. M. King. ISIS Facility...
0 downloads 0 Views 102KB Size
5562

Langmuir 2000, 16, 5562-5567

Freezing Experiments on Clay Gels H. L. M. Hatharasinghe and M. V. Smalley* Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, U.K.

J. Swenson Department of Applied Physics, Chalmers University of Technology, S-412 96 Go¨ teborg, Sweden

A. C. Hannon and S. M. King ISIS Facility, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire OX11 0QX, U.K. Received November 15, 1999. In Final Form: March 8, 2000 A three-component system consisting of butylammonium vermiculite, butylammonium chloride, and water was studied by X-ray and neutron diffraction experiments in freezing cycles between -5 and +5 °C. In liquid water, all the samples studied were in a colloidally swollen gel phase, with interlayer separations between the clay platelets of the order of hundreds of angstroms. Upon freezing of the water, the colloidal state collapsed into a tactoid phase with an interlayer separation of d ) 19.4 Å. This phenomenon was observed throughout a wide range of clay concentrations r and salt concentrations c. The phase transition was observed to be reversible, the gel phase always being recovered upon warming through the freezing point of water. Such a reversible phase transition between swollen and collapsed clay mineral phases may be important in the weathering of rocks in freezing cycles.

Introduction The weathering of rocks in freezing cycles is an important problem in geology and physical geography. It is largely a problem of the phase behavior of a threecomponent clay-salt-water system. We have used X-ray and neutron diffraction experiments to investigate freezing cycles between -5 and +5 °C for a model system consisting of butylammonium vermiculite, butylammonium chloride, and water. Vermiculites are expandable clay minerals that occur naturally as macroscopic (cm3) crystals. They consist of stacks of negatively charged sheets, with charge-balancing counterions between the sheets. In the presence of water or other polar solvents, the counterions solvate, forcing the clay layers apart. With certain counterions, notably butylammonium (C4H9NH3), macroscopic swelling occurs upon soaking in water or dilute solutions of the counterion.1 These gels, which are ideal colloids (with d-values along the swelling axis in the range 50-500 Å), undergo a temperature-induced phase transition to a crystalline phase with a d-value of 19.4 Å on heating.2,3 The thermodynamic nature of this transition has been established.2 The crystalline (d ) 19.4 Å) phase is not identical to the original wet macroscopic crystals, which also have a d-value of 19.4 Å. Tactoids, clumps of plates containing about 10 silicate layers, are formed as described in ref 4. Neutron scattering has been used to map out the {r, c, Tc} phase boundary of the system, where r is the volume fraction of clay (the sol concentration), c the salt concentration, and Tc the upper critical solution temperature.5 (1) Walker, G. F. Nature 1960, 187, 312. (2) Smalley, M. V.; Thomas, R. K.; Braganza, L. F.; Matsuo, T. Clays Clay Min. 1989, 37, 474. (3) Braganza, L. F.; Crawford, R. J.; Smalley, M. V.; Thomas, R. K. Clays Clay Min. 1990, 38, 90. (4) McCarney, J.; Smalley, M. V. Clay Min. 1995, 30, 187.

Here we report the discovery that the clay gels also have a lower critical solution temperature (LCST) in the butylammonium vermiculite system. Experimental Section The vermiculite crystals were from Eucatex, Brazil. The preparation of the butylammonium vermiculite crystals for the experiments is described in detail elsewhere.2,3,5 The crystals were stored in a molar butylammonium chloride solution prior to the swelling experiments. The butylammonium vermiculite crystals were washed with distilled water at 60 °C before drying on filter paper.5 Crystals in the wet crystalline state (d ) 19.4 Å) were individually weighed, and the volume of a crystal in its fully hydrated state was calculated using the density (F ) 1.86 g cm-3). Three n-butylammonium chloride concentrations, 0.01, 0.03, and 0.1 M, were used. Solutions were prepared by dissolving the required mass of n-butylammonium chloride in D2O for the neutron-scattering experiments and H2O for the X-ray experiments. It was necessary to swell the crystals in D2O rather than H2O solutions for the neutron diffraction experiments because of the large incoherent scattering cross section of hydrogen, which would otherwise have obscured the scattering of interest. Three types of diffraction experiments were performed, in different ranges of the momentum transfer vector Q, small-angle (low-Q) neutron diffraction experiments on the LOQ instrument at ISIS, wide-angle (high-Q) neutron diffraction experiments on the LAD instrument at ISIS, and laboratory X-ray diffraction experiments in an intermediate Q-range. In the LOQ experiments, a single vermiculite crystal was placed into a quartz sample cell of internal dimensions 1.0 × 1.0 × 4.5 cm and an appropriate amount of solution (typically 2.5 cm3) was added to prepare r ) 0.01, 0.1, and 0.3 samples. The cells containing the three components were sealed with Parafilm and allowed to stand at 7 °C for 2 weeks prior to the experiments, to ensure that full equilibrium swelling had been achieved.5 The points on the phase diagram of the three-component system (5) Williams, G. D.; Moody, K. R.; Smalley, M. V.; King, S. M. Clays Clay Min. 1994, 42, 614.

10.1021/la991494k CCC: $19.00 © 2000 American Chemical Society Published on Web 05/20/2000

Freezing Experiments on Clay Gels

Figure 1. Phase diagram of the three-component system of clay (butylammonium vermiculite), salt (butylammonium chloride), and water at constant T and P. The solid lines illustrate qualitatively the phase boundaries at T ) 4 °C, P ) 1 atm. The left-hand wedge (region IV) represents a threephase region of crystalline clay, solid salt, and saturated salt solution, and the central wedge (region III) represents a twophase region of crystalline clay and salt solution. Regions I and II are the one-phase (I) (all components mixed) and two-phase (II) (gel in equilibrium with clear supernatant fluid) regions of colloid stability. The crosses mark the points studied in the neutron diffraction experiments. studied are shown in Figure 1.6 These points overlap with those studied by both Williams et al.5 and Swenson et al.7 in previous neutron-scattering studies of the UCST in this system. The timeof-flight small-angle scattering instrument LOQ has been described in detail elsewhere.8 A white beam of neutrons with wavelengths in the range between λ ) 2.2 and 10 Å was used, and the incident beam was collimated by passage through 8 mm wide, 2 mm high rectangular slits. The samples were mounted on a temperature-controlled 20 position sample changer. Neutrons scattered by the gel samples were recorded on a two-dimensional area detector situated 4.1 m behind the samples, covering the approximate Q-range between 0.01 and 0.2 Å-1, and on a high-angle bank of detectors whose center subtended an angle of 20° to the transmitted beam, covering the approximate Q-range between 0.2 and 1.6 Å-1. The quartz sample cells used were practically transparent to neutrons at the wavelengths utilized on LOQ, and the small-angle neutronscattering from D2O was of low intensity over the Q-range studied. Subtraction of the background scattering, after the appropriate transmission corrections were made, was found to have a negligible effect on the scattering patterns, which were dominated by scattering from the gels. One experiment, at c ) 0.1 M, r ) 0.1, was carried out on the LAD instrument at ISIS, described in detail elsewhere.9 On this occasion, swelling was also carried out for 2 weeks in a quartz jar, but immediately prior to the experiment the gel stacks were removed from the soaking solution and put into a flat plate sample can made from a TiZr alloy. The can produces almost no coherent neutron scattering. The internal dimensions of the can were 55 mm × 19 mm × 2 mm, and the gels were placed in the can with the clay layers parallel to the 55 mm × 19 mm faces of the cell. To ensure that the sample remained in this orientation, individual gels were selected that were about 2 mm thick. Six gels were used, of suitable dimensions to pack together as efficiently as possible to maximize coverage of the neutron beam. The soaking solution that had been in equilibrium with the gels was then added to give a volume fraction of about 0.1. The sample can was sealed with a stainless steel cap containing a Teflon washer and mounted on a He gas closed cycle refrigerator. The samples were (6) Smalley, M. V. Langmuir 1994, 10, 2884. (7) Swenson, J.; Smalley, M. V.; Thomas, R. K.; Crawford, R. J. J. Phys. Chem. B 1998, 102, 5823. (8) Heenan, R. K.; King, S. M. LOQ Instrument Handbook; RAL Report RAL-TR-96-036; Rutherford Appleton Laboratory: Didcot, 1996. (9) Howells, W. S. LAD Instrument Handbook; Report RAL-86-042; Rutherford Appleton Laboratory: Didcot, 1986.

Langmuir, Vol. 16, No. 13, 2000 5563 oriented such that the clay platelets were perpendicular to Q at the scattering angle 10°. Spectra were recorded separately for each group of detectors at the angles 5°, 10°, 20°, and 35°. Thus, using only these low scattering angles we ensured that the momentum transfer vector Q was approximately perpendicular to the clay platelets for all angles. The incident wavelength was in the range 0.1-6.0 Å. The data from each detector group were corrected separately for background and container scattering, absorption, multiple scattering, and inelasticity effects and normalized against the scattering of a vanadium rod following the procedure described in ref 10. The corrected individual data sets obtained at each angle were then combined to obtain a large Q-range and to improve the statistics. For each data set we only used the Q-range that agreed with the other data sets in the overlapping Q-region. Finally, the combined data set was normalized to a proper structure factor S(Q) for the direction perpendicular to the clay platelets. The X-ray measurements were carried out using a Philips Xpert θ-θ X-ray diffractometer. In this geometry, both the X-ray source and detector arm subtend an angle θ° to the free surface of the sample. The normal operating conditions were 50 kV and 30 mA. An X-ray beam produced by a molybdenum target that passes through a Zr filter placed over the divergence slit was used for all experiments. The beam consists of two emission lines at 0.7096 and 0.7136 Å and was treated as one unresolvable wavelength at 0.7107 Å. A programmable slit system was used to control the dimensions of the incident and scattered beams. The widths of divergence and antiscatter slits were set as automatic and controlled by changing the illuminating length (generally 2 mm). The height of the receiving slit was 0.1 mm. The samples were mounted on a temperature-controlled diskshaped sample holder, which was connected to a cryostat, and the level of the sample surface was adjusted before the inner and outer sample chambers were fixed to the cryostat. The data were collected over the 2θ range 0.5°-30°, which corresponds to the Q-range 0.05-4.5 Å-1, with a step size of 0.1°. X-rays scattered by the samples were recorded on a two-dimensional area detector and analyzed using the software PC-APD.

Results (a) LOQ: Small-Angle Neutron-Scattering Data. Three salt concentrations c ) 0.01, 0.03, and 0.1 M and three sol concentrations r ) 0.01, 0.1, and 0.3 were studied, with three samples used at each of the nine {r, c} points, to allow for sample-to-sample variability.5 At the salt concentration for which most structural data is available,5 c ) 0.1 M, the gel expands to 6 times the volume of the original vermiculite crystal (d ) 19.4 Å) at r ) 0.01 so the volume of the gel phase at equilibrium is only a few percent of the total volume, with over 90% occupied by the supernatant fluid. However, for r ) 0.1, a similar expansion leads to equilibrium conditions in which the gel occupies about 60% of the total volume. In ref 7, preliminary data between 0 °C and Tc were given on the system at r ) 0.4, for which the clay can soak up all of the available solution, resulting in a system with no supernatant fluid. In practical problems, such higher clay concentrations will be important. Typical results of this study are shown for the case r ) 0.3, c ) 0.1 M in Figure 2. The low-angle bank data in the approximate Q-range between 0.02 and 0.25 Å-1 displayed in Figure 2a show the disappearance of the gel peak upon freezing from +5 to -5 °C, and the high-angle bank data in the approximate Q-range between 0.2 and 1.6 Å-1 in Figure 2b show that this is accompanied by the appearance at Q ) 0.32 Å-1 of the first-order Bragg peak of the d ) 19.4 Å crystalline phase of the clay. The essential feature of the experiment was that similar behavior was observed at all the {r, c} points studied. (10) Howe, M.; Howells, W. S.; McGreevy, R. L. J. Phys.: Condens. Matter 1989, 1, 3433.

5564

Langmuir, Vol. 16, No. 13, 2000

Hatharasinghe et al.

Figure 3. LAD data at r ) 0.1, c ) 0.1 M. The low-temperature gel-tactoid phase transition was discovered to occur simultaneously with the formation of ice. (a) shows the structure factor of a r ) 0.1, c ) 0.1 M butylammonium vermiculite gel at T ) +5 °C, and (b) shows the pattern obtained at T ) -5 °C. In (b) the gel has collapsed locally to its tactoid phase (which produces a peak at Q ) 0.32 Å-1), and the solution has crystallized at the same temperature.

Figure 2. (a, top) LOQ low-angle bank data at r ) 0.3, c ) 0.1 M. The pattern obtained at T ) +5 °C is shown by the open circles and, that at T ) -5 °C by the closed circles. (b, bottom) LOQ high-angle bank data at r ) 0.3, c ) 0.1 M. The pattern obtained at T ) +5 °C is shown by the open circles and that at T ) -5 °C by the closed circles. Table 1. Average d-Values Observed in the LOQ Experiments As a Function of the Sol Concentration r and the Salt Concentration c, at T ) +5 °C d, Å c, M

r ) 0.01

r ) 0.1

r ) 0.3

0.01 0.03 0.1

330 240 130

250 170 110

110 100 75

In all cases it was possible to measure the d-value along the swelling axis in the gel phase quite accurately from the position of the maximum in intensity of the welldefined first-order Bragg peak (see Figure 2a) observed at +5 °C. The results are shown in Table 1. The trend to smaller d-values at the higher salt and sol concentrations is clear. The excellent agreement with the results at r ) 0.01 and 0.1 previously reported by Williams et al.5 confirms that the batch of Eucatex vermiculite crystals used in the current study of the LCST has the same characteristics as that used in the previous study of the UCST.5 The data at r ) 0.3 were at a similar sol concentration to those studied by Swenson et al.7 but show one qualitative difference. It is clear that with 10 parts water added to 4 parts clay a homogeneous expansion of the whole sample from a d ) 20 Å phase can lead to a maximum d-value of 70 Å in the gel. The d-values shown in the right-hand column of Table 1 are substantially

higher than this, so there must have been regions of unexpanded crystal in both of these samples. In ref 7, these regions show up clearly in the diffraction patterns, whereas here (see Figure 2b) there seems to be no crystalline phase at +5 °C. The reason for this is simply due to the different geometries of the two experiments. In the current experiment, the samples were initially arranged with the clay layers parallel to the ground and upon cooling no crystalline peaks were observed even at -5 °C in this geometry because the angle subtended by the Bragg peak and the ground is approximately 20°. The samples had to be tilted by 10° with respect to the ground in order to observe the appearance of the crystalline Bragg peak, and insufficient beam time was available to check the +5 °C patterns at this angle. The tilting of the samples also meant that good thermal contact with the aluminum block controlling the temperature was partly lost, so we were unable to identify the exact temperature at which the phase transition occurred in the LOQ experiments. This was determined in the X-ray experiments descibed later. It was also impossible to identify the formation of ice in the LOQ experiments because the maximum Q-value of the instrument lies below the Q-value of the first Bragg peak of ice. This was performed in the X-ray experiments and in the LAD experiment described below. (b) LAD: Wide-Angle Neutron-Scattering Data. A single experiment was performed at c ) 0.1 M, r ) 0.1 using the much higher Q-range of the LAD instrument. The normalized structure factor of the gel at +5 °C is shown in the Q-range between 1 and 6 Å-1 in Figure 3a. In this Q-range we do not see the gel Bragg peak, but rather the diffuse scattering from the gel phase that has been interpreted elsewhere.7,11 The pattern changed completely at -5 °C, as shown in Figure 3b. The intense peak at Q ) 0.32 Å-1 (note the difference in intensity (11) Swenson, J.; Smalley, M. V.; Thomas, R. K.; Crawford, R. J.; Braganza, L. F. Langmuir 1997, 13, 6654.

Freezing Experiments on Clay Gels

scales between parts a and b of Figure 3) is the first-order Bragg peak of the tactoids identified in the high-angle bank data in the LOQ experiments, and the even more intense peak at Q ) 1.7 Å-1 is the first Bragg peak of ice. The full diffraction pattern obtained was compared with those from known ice structures12 and was found to correspond to the ordinary hexagonal ice phase one would expect to obtain on cooling a simple salt solution or pure water through its freezing point. As far as it was possible to tell from the 10 °C temperature jump employed, the ice peaks appeared simultaneously with the appearance of the clay tactoid peak. This strongly suggested that the clay gel phase is thermodynamically unstable in the solid solvent. Having determined from the higher Q-range of the LAD instrument that ordinary ice is formed in the “double” phase transition of the clay and solvent systems, the sharpness and reversibility of the transition were then investigated in laboratory X-ray experiments. Before describing these, we note that the high penetrating power of the neutrons means that they sample the entire bulk of the material, so we can be sure that surface effects are not playing any significant role in the phase behavior. (c) X-ray Diffraction Data. The θ-θ X-ray diffractometer used was ideal for our investigation as we could see both the transition from the gel phase to the tactoid phase of the clay (at high salt and sol concentrations) and the formation of ice. The clay volume fraction was held constant at r ) 0.1 for two reasons; first, to obtain a d-spacing in the gel phase that would be observable within the Q-range of the instrument and, second, because the sample container of the X-ray instrument was too small to conveniently study low volume fractions of clay. Before the measurements were taken, the system was kept at each temperature for at least half an hour to ensure that thermal equilibrium had been achieved. The diffraction patterns obtained from a sample with c ) 0.1 M, r ) 0.1 (the same conditions as for the LAD experiment), at five temperatures between 2 and -2 °C with 1 °C steps, are shown in Figure 4a as a plot of I(Q) vs Q. At 2 °C (dotted line), there appears to be a weak peak at Q ) 0.18 Å-1 corresponding to a real space correlation length of 35 Å. In view of the LOQ result that d ) 110 Å at c ) 0.1 M, r ) 0.1, we interpret this as the weak third-order Bragg peak of the gel phase, with the lower order Bragg peaks outside the low Q-range of the instrument. At 1 °C (solid line) the overall intensity of the scattering in the Q-range 0.1-0.4 Å-1 increases and then decreases again at 0 °C (open circles). The peak also becomes less distinct and seems to shift toward lower Q, indicating an expansion of the gel phase as freezing is approached. However, we cannot substantiate such a conclusion in advance of further experiments in a lower Q-range. What is definitely clear is that there is no peak at 0.32 Å-1 down to 0 °C, in liquid H2O. At -1 °C, the diffraction patterns from the gel phase collapsed entirely and the Bragg peak typical of the crystalline phase appeared at Q ) 0.32 Å-1, equivalent to a d-spacing of 19.4 Å. The second Bragg peak at Q ) 0.64 Å-1 is also clearly visible. Further scans below -1 °C were identical to that at -1 °C, showing that the phase change was complete within 1° of temperature and an interval of 1 h. The phase transition was found to be reversible by cycling through the freezing point between 3 and -3 °C, and the transition was always completed within a degree of the freezing point. At two temperatures above and below the phase transition temperature (at 1 and -1 °C), the sample was scanned over a higher Q-range up to 4.5 Å-1. (12) International Tables for Crystallography, 4th ed.; Hahn, T., Ed.; Kluiver: Dordrecht/Boston/London, 1995.

Langmuir, Vol. 16, No. 13, 2000 5565

Figure 4. X-ray data at r ) 0.1, c ) 0.1 M. (a, top) In the Q-range up to 1.0 Å-1, the appearance of the peaks at Q ) 0.32 and 0.64 Å-1 at -1 °C shows the formation of the tactoids between 0 and -1 °C. In (b, bottom) the patterns at -1 and +1 °C have been extended to Q ) 4.5 Å-1 to show the appearance of the ice peaks at -1 °C. The clay peaks are marked by open circles (O).

The results are shown in Figure 4b. It is clear that the scattering pattern at -1 °C is very different from that at 1 °C, with sharp Bragg peaks appearing in the higher Q-region between 1.5 and 4.5 Å-1. The peaks marked with an open circle (O) are the higher order Bragg peaks arising from the crystalline clay layers, and the other peaks are from the crystalline ice. For comparison, diffraction patterns were obtained for pure water. It was evident that the Bragg peak positions for ice were the same in both cases, indicating that the structure of ordinary ice12 is not perturbed by the presence of the clay and the salt. We should note that the limit of sensitivity of our experiments restricts us to saying that the phase transitions are simultaneous only in the sense that they both occur between -1 and 0 °C (in either direction). More subtle variations, of the order of ( 0.1 °C, would not be detected in the current experiments. According to DebyeHu¨ckel theory,13 the depression of the freezing point of pure water is 0.18 °C in a 0.1 M uni-univalent electrolyte solution. We would expect the clay to cause a further small depression in the freezing point, as discussed below. Within these limits, the temperature where both the freezing transition and the gel-tactoid phase transition occur in the system of a clay colloid in a salt solution can be concluded to be the freezing point of the soaking solution. (13) Debye, P. J. W.; Hu¨ckel, E. Phys. Z. 1923, 24, 85.

5566

Langmuir, Vol. 16, No. 13, 2000

Figure 5. X-ray data at r ) 0.1, c ) 0.03 M. There were no essential changes in behavior observed at c ) 0.03 M. (a) shows the gel-tactoid transition between -1 and 0 °C, and the higher Q data in (b) shows the appearance of the ice peaks at -1 °C.

The UCST is sensitive to the salt concentration of the soaking solution.3,5 To investigate the effect of the salt concentration on the LCST, an experiment was performed for a 0.03 M salt concentration. Figure 5 shows the patterns obtained for a gel with c ) 0.03 M, r ) 0.1 at five temperatures between 2 and -2 °C with 1 °C steps. As in the case of 0.1 M, at temperatures above the freezing point, we can see the scattering from the gel phase in the Q-range from 0.1 to 0.4 Å-1. At -1 °C, the diffraction pattern from the gel phase collapsed and the Bragg peak typical of the tactoid phase appeared at Q ) 0.32 Å-1, showing that the phase transition of the clay had again occurred between 0 and -1 °C. Further scans below -1 °C were identical to that at -1 °C, showing that the phase change was complete within 1 h. (d) Summary. Piecing together the evidence obtained in the three different types of diffraction experiments, we can state that (i) the gel phase collapses to the tactoid phase simultaneously with the freezing of the water into crystalline ice, (ii) the phase transition temperature is unaffected by the presence of the clay, being equal to the value expected for the pure salt solution, (iii) the structure of the ice formed is unaffected by the presence of the clay, and (iv) the behavior is essentially the same for a wide range of salt concentrations c and sol concentrations r for the three component clay-salt-water system. Discussion The importance of the butylammonium vermiculite system as a model system for colloid stability lies mainly

Hatharasinghe et al.

in the identification of the crystalline, or tactoid, phase (region III in Figure 1) with the primary minimum state (d ≈ 1 nm) and the identification of the gel phase (regions I and II in Figure 1) with the secondary minimum state (d ≈ 10-100 nm).6,14 It has previously been noted that the reversibility, sharpness, and reproducibility of the phase transition at the UCST show that the transition is a true thermodynamic one.2,3,5 All the same considerations clearly apply to the new phase transition at the LCST that we have described here. In the former case, it was possible to directly measure the enthalphy change at the UCST, which was found to be approximately 5 J/g of crystalline vermiculite. It would be very difficult to measure the enthalpy change at the LCST directly, because of the simultaneous occurrence of the freezing transition of the solvent, whose enthalpy change is 3 orders of magnitude greater per gram of water.15 Nevertheless, from the diffraction experiments described here, the structure of the clay system changes in a very similar way at both phase transition temperatures. A crucial feature of the discussion is that simple butyammonium salt solutions are well represented as ideal ions.16,17 The dissolution of such salts into water (which we could represent as a cut along the bottom edge of the triangular phase diagram shown in Figure 1) gives colligative properties such as the depression of freezing point and elevation of boiling point that can be calculated from Debye-Hu¨ckel theory.13 The thermodynamics of such two-component solutions are well understood. In adding a clay with a counterion identical to that of the simple salt solution, we create a three-component system whose thermodynamic properties have been less thoroughly investigated. Let us imagine holding c fixed at some convenient value, say c ) 0.01 M, and consider the addition of clay to the salt solution as a psuedobinary system (which we could represent as a vertical cut across the center of Figure 1). In the sense that we can view the dispersal of the clay into the aqueous medium (i.e., gel formation) as analogous to the dissolution of salt into water, we would expect to see phenomena typical of the two-component salt-water system, like depression of freezing point, brought about by “dispersal” of the clay layers into the liquid solvent (water). It is clear that the solid solute (i.e., the clay) does not dissolve in the solid solvent (ice) and that the solid solute does dissolve (i.e., gel) into the liquid solvent. The explanation for all of these phenomena is clearly a thermodynamic one; the secondary minimum state is thermodynamically stable over a wide range of {r, c, T} conditions in this model system. We do have to be careful in the way we apply the definition of a phase to the butylammonium vermiculite system. According to Gibbs,18 a phase is defined as any homogeneous and physically distinct part of a system that is separated from other parts of the system by definite boundary surfaces. Because the gel can be lifted out of the supernatant fluid on a spatula, it clearly justifies description as a phase in the latter sense, but it is inhomgeneous on the nanometer-to-micron (colloidal) length scale. It can only be defined as homogeneous on the macroscopic length scale. The same considerations apply to the tactoid phase, where what we have described as tactoids are positive (14) Smalley, M. V. Mol. Phys. 1990, 71, 1251. (15) Book of data, 2nd ed.; Harrison, R. D., Ed.; Nuffield: Harmondsworth, Middlesex, U.K., 1973. (16) Desnoyers, J. E.; Arel, M. Can. J. Chem. 1967, 45, 359. (17) Krishnan, C. V.; Friedman, H. L. J. Phys. Chem. 1979, 74, 3900. (18) Gibbs, J. W. Collected Works; Yale University Press: New Haven, CT, 1948.

Freezing Experiments on Clay Gels

tactoids, according to Bernal and Fankuchen.19 The role of attractive and repulsive forces in the formation of tactoids has been discussed perceptively by Langmuir,20 and our results necessarily have a large impact on the theory of colloid stability. On the other hand, there are massive practical applications in stabilizing colloids against flocculation with the temperature.21 Such wider questions are beyond the scope of the present paper. Here we note that sedimentation problems in lakes, the role of drilling mud in the petroleum industry, and the role of clay lining in landfill (19) Bernal, J. D., Fankuchen, I. Nature 1936, 138, 1051; 1937, 139, 923. (20) Langmuir, I. J. Chem. Phys. 1938, 6, 873. (21) Everett, D. H. Basic Principles of Colloid Science; Royal Society of Chemistry: London, 1988.

Langmuir, Vol. 16, No. 13, 2000 5567

barriers at waste disposal sites are mainly affected by the phase behavior of aqueous clay colloid systems. Our results impact indirectly on these problems and directly on the behavior of rocks in freezing and thawing cycles in geological weathering processes.

Acknowledgment. This work was financially supported by Unilever plc. One of the authors (H.L.H.M.) thanks Unilever plc for a studentship in support of the work, and another (J.S.) thanks the Swedish Natural Science Research Council for their support. We thank Mike Cresswell for technical support of the X-ray experiments. LA991494K