J. MATER. CHEM., 1991,1(6), 965-970 Time-resolved Powder Neutron Diffraction Study of Thermal Reactions in Clay Minerals David R. Collins,*" Andrew N. Fitch" and C. Richard A. Catlowb a Department of Chemistry, University of Keele, Staffordshire ST5 5BG, UK Davy Faraday Research Laboratory, Royal Institution of Great Britain, 21, Albemarle Street, London WlX 4BS, UK Thermally induced reactions in three contrasting clay samples have been studied, in situ, using time-resolved powder neutron diffraction. The dehydroxylation of kaolinite was observed in the range 440-600 "C, and the initial stages of mullite formation at 950-1000 "C. A multi-component clay material composed of disordered kaolinite and illite, with excess quartz, known as Etruria Marl, dehydroxylated in the range 400-600 "C and began to recrystallize to mullite at 1000 "C.The quartz-cristobalite phase transition in this material was observed at 1150 "C. An Mg-vermiculite was shown to dehydrate in stages, giving stable hydrates with d-spacings of 14.4, 11.7, 10.3 and 9.4 A. The vermiculite dehydroxylated at 400-850 "C and began to recrystallize to enstatite at 1000 "C. Keywords: Kaolinite; Vermiculite ; Etruria Marl ; Time-resolved powder neutron diffraction Time-resolved powder neutron diffraction (TRPND) is ide- ally suited to studying the dehydration and dehydroxylation of clay minerals as well as the formation of high-temperature anhydrous crystalline phases. The technique allows us to monitor, in situ, the structural changes of a crystalline material under non-equilibrium conditions, by the rapid acquisition of a series of complete diffraction patterns.In addition, the technique also monitors the hydrogen atom content of a material, irrespective of crystallinity, through the time evolution of the incoherent background intensity. In this work the structural changes are thermally induced, although pressure induced changes have also been studied using TRPND.' The thermal reactions of layer silicates have been studied by other in situ techniques, such as thermal analysis and spectroscopy (see the recent review by Brindley and Lemai- tre2). However, these methods do not yield information pertaining directly to structural changes. Previous diffraction experiments have not been performed in situ, and have involved heat treatment and quenching, and then subsequent analy~is.~Although such experiments have provided valuable information regarding the identification and stability ranges of phases, they cannot give a detailed record of thermal evolution and may indeed fail to detect certain short-lived phases.In contrast, TRPND has the potential to give a complete record of structural changes, allowing in principle, the identification of all crystalline phases. The potential of possible applications of TRPND in solid- state chemistry are numerous (see Pannetier4). The viability of studying thermally induced dehydration reactions in inor- ganic crystalline materials by TRPND has been demonstrated previously on Inverse Weberites.' Indeed we note that the latter study was performed on the same diffractometer DlB, as used in the present study.We describe experiments on three different clay samples, namely a pure kaolinite, a pure Mg-vermiculite and a material known as Etruria Marl. The latter is of particular industrial importance as it is used as a raw material for the manufacture of bricks. We have collected data for all three samples between room temperature and ca. 1150 "C observing various stages of water loss and the formation of high-temperature anhy- drous phases. Experimental Starting Materials and Sample Preparation Kaolinite A well crystalline kaolinite known as the soft variety of Georgia kaolinite was used [Clay Minerals Society Source Clay (KGa-1) Washington County, Georgia, USA].6 No sample preparation was needed as the kaolinite was already of suitable powder form.Etruria Marl The Etruria Marl (from Staffordshire, UK) is a multi-phased material containing disordered kaolinite and illite with quartz and minor iron oxide phases. The material was already of suitable crystal size and no sample preparation was necessary. Mg-Vermiculite Obtained from Llano, Texas, USA, this vermiculite is of particularly high crystallinity and ordering and has been used in previous X-ray diffraction ~tudies.~.~ It occurs in macro- scopic flakes ca. 10 mm x 10 mm x 3 mm, with a pronounced cleavage parallel to the layers. An electric herb mill (SEB mini hachoir) was used to reduce the particle size of the material to a suitable powder form.Such a procedure was deemed preferable to grinding the material with a pestle and mortar as the latter is believed to damage the crystal ~tructure.~ A more uniform crystal size was achieved by sieving the powder. In order to ensure complete saturation with Mg interlayer cations, the sample was refluxed with solutions of MgC1, at 70 "C for 7 days. During this period the solutions were changed daily. After the final reflux the excess chloride ions were removed by washing with distilled water, using silver nitrate as a test to ensure their complete removal. The sample was partially dried in a desiccator over silica gel and sub- sequently placed in a second desiccator over a water bath to ensure complete hydration.Neutron Diffraction Experiments All the data were collected on the two-axis diffractometer D1 B, at the Institut Laue Langevin, Grenoble, France. The instrument is equipped with an 3He/Xe position-sensitive J. MATER. CHEM., 1991, VOL. 1 detector containing 400 cells spanning an 80" range in 28. Individual diffraction patterns were collected at intervals of 150s at a wavelength of 2.52 A. At the beginning of the experiment an A1203 standard was run to correct for the zero-point error of the diffractometer. To collect data over as wide a temperature range as possible we used a furnace equipped with a.niobium heating element which allowed the samples to be heated under vacuum up to 1150 "C.The samples were held in a 10 mm x 50 mm open- ended niobium sample can. However, because of the strong scattering of neutrons by niobium, data collected using the niobium furnace and sample can contained large contami- nation peaks. Owing to a programming error the kaolinite sample was held at 610 "C for ca. 3 '/z h before continuing to heat up to 1150 "C. All the samples were held at this tempera- ture for 2 h before cooling. During this time we continued to collect data until the temperature had cooled to ca. 200 "C, at which point we were able to open the furnace and change the sample. Experimental details of the angular ranges, temperatures and heating rates over which the data were collected are given in Table 1.The temperatures correspond to those of the thermocouple placed in the samples and are believed to be accurate within a few degrees. However, there exists the possibility of temperature gradients within the sample. Interpretation of Data The diffraction patterns for each of the different samples are represented conveniently as a 3D plot, where intensity is expressed as a function of 28 and time or temperature. These data are analysed subsequently by fitting an appropriate peak- shape function to the Bragg reflections and a second-order polynomial to the incoherent background, using the computer program ABFFIT." Thus we are able to monitor accurately the position, FWHM and intensity of the Bragg reflections together with the incoherent background intensity as a func- tion of time and temperature. In practice the intensity of the incoherent background is determined by measuring the average intensity over a small angular range (e.g.So), where no reflection occurs. From this, the time evolution of the hydrogen atom content can be monitored. However, the observed incoherent background intensity contains contributions not only from the hydrogen atoms, but also those due to instrumental background, air scattering, and scattering by amorphous material in the sample. An approximate correction, to obtain the incoherent scattering intensity due solely to hydrogen, can be made by subtracting off the incoherent background intensity measured for the fully dehydrated material. This approach ignores effects caused by changes in the amorphous content of the sample, and the temperature dependence of the incoherent background arising from the sample and furnace.Table 1 Summary of experimental details" heating rate/ temperature samples 28 range/" "C min-' range/ "C kaolinite 5-85 3.75b 20-1 150 Etruria marl 5-85 3.75 30-1 150 Mg-vermiculite 5-85 1.8 40-250 4.0 250-1 150 Wavelength =2.5 18 A;collection time for individual diffraction pat- terns = 150 s. Owing to a programming error the temperature of the kaolinite sample was held constant at 630 "C for ca. 200 min. The thermal evolution of a crystalline phase can be moni- tored by the intensity of one or more of its Bragg reflections. This neglects the effects on the thermal evolution of the Debye-Waller factors. However, the approximation is believed to be reasonable for low-angle isolated reflections over a limited temperature range.5 The amount of each phase in the sample at a given temperature can then be estimated by measuring the intensity of one or more Bragg reflections and then normalizing to the intensities from pure phase reference sample^.^ It is very difficult to determine reliably the relative amount of each phase in a sample from medium-resolution powder diffraction data, especially when non-crystalline phases are present. Instead we have limited our analysis to expressing the time/temperature evolution of each phase in each sample rather than attempting to apportion the total time dependence of the diffraction pattern to precise contributions from individ- ual phases.Using a suitable Bragg reflection we are able to calculate the ratio Illmax,for each phase, where I is the intensity at a given temperature and I,,, is the maximum intensity of that reflection. Finally we note that the observed peak positions, FWHM and intensities are dependent on the time/temperature resol- ution of the experiment. Results and Discussion Kaolinite The 3D plot of the kaolinite data is shown in Fig. 1 and the thermal evolution profile in Fig. 2. At 25 "C (during furnace evacuation and prior to heating) there is a small decrease in the incoherent background which we believe represents the loss of a small amount of grain-boundary water.Indeed since there is no corresponding reduction in the kaolinite intensity, this loss of hydrogen cannot originate from within the kaolin- ite crystal structure. The reduction in the intensities of the kaolinite 001 and 002 reflections and the incoherent back- ground show the disappearance of kaolinite, due to dehydrox- ylation, begins at 440 "C and is complete at 600 "C. The variation of the d-spacings of the 001 and 002 reflections with temperature between 25-600 "C is linear, giving a thermal expansion coefficient along c* of 1.60 x K-I. Following dehydroxylation but before the onset of mullite recrystalliz- ation, we observe an amorphous rneta-kaolin region. Although the majority of the hydrogen has been removed during dehydroxylation there still appears to be small amounts present in the amorphous rneta-kaolin region as suggested previously by IR experiments."-13 The removal of hydrogen does not appear to be complete until ca.900 "C.The formation of mullite begins at 950 "C. We note that this is a somewhat lower temperature than previous reports of mullite forma- tion.14-16 However, it is possible that the early stages of formation of the Bragg peaks attributed to mullite, are in fact the poorly understood spinel phase, which is rep~rted~,~.'~-'' to be stable in the range 900-1000 "C. Since the structural details of the latter phase have not been fully determined, it is not possible to distinguish between the two phases using these data.We note, however, that the intensities of the peaks referred to as the mullite 110 and 210 reflections show local maxima at 990 "C; and we suggest that this effect may be due to the presence of the spinel phase which is replaced by mullite at temperatures above 990 "C. There is, however, no further evidence for this suggestion and, of course the increase with temperature of the Debye-Waller factors of the 1I0 and 210 reflections may significantly affect their observed inten- si ties. J. MATER. CHEM., 1991, VOL. 1 12003 I 001 time/min 002 Fig. 1 3D plot of diffraction patterns for kaolinite. Inset shows the time-temperature profile 0.6OD8l "1i I I0.0 A t A A A A A A A A A A AA AA A AA &A AA A Fig.2 Thermal evolution profile for kaolinite. Incoherent background (-), kaolinite 001 (0)and mullite 110 (A)intensities Etruria Marl The multi-component 3D diffraction pattern of the Etruria Marl (see Fig. 3) is more complex than that of kaolinite. The 001 kaolinite reflections in the Etruria Marl are poorly defined, probably as a consequence of poorer crystallinity. However, the variation in d-spacing of these reflections as a function of temperature is similar to those observed in the pure kaolinite sample. The absence of any illite peaks, particularly the 10 8, 001 reflection, suggests that this sample of Etruria Marl contains little illite, or the illite which it does contain is of exceptionally poor crystallinity.The presence of quartz is, however, demonstrated by its strong 101 reflection. The thermal evolution profile for Etruria Marl is shown in Fig. 4. At 25 "C we observe a sharp decrease in the incoherent background, which we attribute to the loss of grain-boundary water. We note that the intensities of both the 002 kaolinite and quartz 101 reflections increase with reduction of the incoherent background intensity. It is possible that the enhancement of the kaolinite and quartz reflections is a direct consequence of the removal of the grain-boundary water. The reduction in hydrogen-atom content will lead to a reduction in the number of neutrons scattered incoherently, thereby allowing more neutrons to interact coherently with the afore- mentioned crystalline phases. After the grain-boundary loss the incoherent background remains constant up to 400 "C, at which point dehydroxylation begins.The simultaneous reduction in the intensity of the kaolinite 002 reflection and the incoherent background is accompanied by a further increase in the quartz 101 intensity. The latter continues to increase until the dehydroxylation process is complete at 600 "C. Although the increase in intensity of the quartz 101 reflection is probably a consequence of the reduction of the kaolinite and incoherent background intensities, it may poss- ibly represent the a-B phase change. The position of the quartz 101 reflection as a function of temperature indicates normal thermal expansion up to 570"C, whereupon the d- J.MATER. CHEM., 1991, VOL. 1 Fig. 3 3D plot of diffraction patterns for Etruria Marl. C =cristobalite, K =kaolinite, M =mullite and Q =quartz. Inset shows the time- temperature profile spacing abruptly ceases to increase and remains constant. We suggest the point of this abrupt change in behaviour corre- 1.o sponds to the ct-fl quartz phase transition which has pre- viously been observed'' at 573 "C. Between 600 and 1000 "C we observe an amorphous region, which we believe represents 0.8 rneta-kaolin. As with the pure kaolinite sample, small amounts of hydrogen remain trapped until temperatures in excess of 800 "C. The appearance of the mullite at 1000 "C is slightly later than that from the pure kaolinite sample and is consistent 0.6 with the previous observations that mullite formation from a x-E disordered kaolinite is retarded compared with that of a well 2 At 1150 "C we observed the quartz- ordered ka~linite.'~,~'-~~ cristobalite phase transition.The intensity of the quartz 1010.4 reflection decreases as that of the cristobalite 111 reflection increases. 0.2 Mg-VermiculiteA0 0.0 0 A 3D plot of the Mg-vermiculite data is shown in Fig. 5. The 1 I variation in the position of the 001 vermiculite reflection as 0 400 800 1200 a function of temperature, determined by fitting a Gaussian TI"C to the aforementioned reflection, is shown in Fig. 6. The Mg- Fig. 4 Thermal evolution profile for Etruria Marl. Incoherent back- vermiculite initially exhibits a basal spacing of 14.4 A, which mullite 110 (A),quartz 101 (.) andground (-),(.),kaolinite 001 probably represents a two-sheet hydrate.24 This steadily cristobalite 111 (0)intensities decreases to 14.0 8, at 80"C, whereupon the basal spacing J. MATER.CHEM., 1591, VOL. 1 0 200 Fig. 5 3D plot of diffraction patterns for Mg-vermiculite. Inset shows the time-temperature profile 15 13 5 cn .-0 0 P-0 811 LL 9 0 200 400 600 800 1000 TI"C Fig. 6 Position of 001 reflection of Mg-vermiculite, as a function of temperature determined by fitting one Gaussian collapses rapidly to 11.7 A, giving a one-sheet hydrate. We then observe a small decrease in d-spacing to 11.5 8, at 170 "C.Following this we observe a further abrupt collapse of the layers giving a reflection with a d-spacing of 10.3 8, at 200 "C. Since such a d-spacing is of an intermediate value between 400 600 800 timejmin those typical of one- and zero-sheet hydrates, its interpretation needs further consideration as it is difficult to form a physical picture of hydrates containing a non-integral number of interlayer water sheets. The 10.3 8, d-spacing may correspond to the 002 reflection of an ordered arrangement of alternating zero- and one-sheet hydrates. This would yield a similar basal spacing as that observed in the 20.6 8, phase reported in a previous study of Mg-~ermiculite.~~ Alternatively it may corre- spond to the average position of a 001 d-spacing arising from a randomly interstratified distribution of one- and zero-sheet hydrates.In order to distinguish between these two possibil- ities, analysis of higher-order 001 reflections is needed. Since these higher-order reflections are extremely weak and poorly defined, owing to the high incoherent background intensity, such an analysis was not possible. However, as the 10.3 8, reflection is very broad (see Fig. 7) we suggest the interstratifi- cation is probably random. Following this we observe a steady decrease in the d-spacing to 10.0 8, at 700 "C,whereupon it falls more rapidly to 9.3 A, yielding a fully collapsed structure at 800 "C. Shortly after, dehydroxylation destroys the 001 reflection. Analysis of the FWHM of the fitted Gaussian as a function of temperature (Fig.7) reveals very significant broadening during the periods of rapid layer collapse at 80 "C and 170- 200 "C. The broadening probably arises through extensive local variation in hydration states within the sample. After the second layer collapse, the FWHM of the fitted Gaussian is much larger than that fitted to the two- and one-sheet hydrates. We suggest that this broadening, which corresponds to a d-spacing of 10.3 A, probably indicates a disordered 970 3.0 8 2.6 8 2.2 m a m8 0 8 2 1.8 1.4 1.o 8 'IL 8. m '8 0.6 0 200 400 600 800 I000 T/"C Fig. 7 FWHM of 001 reflection of Mg-vermiculite as a function of temperature determined by fitting one Gaussian interstratification of one- and zero-sheet hydrates.Subsequent analysis of the data during the second layer collapse at 160-200 "C,by fitting two Gaussian functions to the 001 reflections, revealed peaks with d-spacings of 11.4 and 10.4 A. These are characteristic of a one-sheet hydrate and a randomly inter- stratified mixture of one- and zero-sheet hydrates, respectively. The first-layer collapse at 80 "C was too rapid to allow the identification of different hydration states. The thermal evolution profile (see Fig. 8) shows dehydration to occur between 25 and 200 "C. There then follows a period up to 400 "C where the incoherent background intensity and the intensity of the vermiculite 001 reflection remain constant.Above 400 "C the incoherent background begins to fall indi- cating the onset of dehydroxylation. Dehydroxylation and 1.o 0.8 0.6 2-2 0.4 0.2 0.0 0 400 800 1200 T/"C Fig. 8 Thermal evolution profile for Mg-vermiculite. Incoherent back- ground (-), and enstatite 420 (V)intensities (.)Mg-vermiculite 001 J. MATER. CHEM., 1991, VOL. 1 hence the removal of the Mg-vermiculite is complete by 850°C. At ca. 1000°C we observe the abrupt growth of the 420 enstatite reflection. From these data it appears that the final stages of dehydroxylation and .the onset of enstatite formation are not simultaneous as has been suggested pre- viously.2 However, we acknowledge that as a result of residual hydrogen, the relatively high incoherent background may obscure the initial stages of the formation of the 420 enstatite reflection.Conclusion This work has demonstrated the use of TRPND to study dehydration, dehydroxylation and high-temperature phase changes in three contrasting clays. We observe structural changes and changes in the hydrogen-atom content during dehydration and dehydroxylation. In addition the formation of mullite is observed from the kaolinite and Etruria Marl samples, whereas enstatite recrystallizes from the Mg-ver- miculite. We thank the I.L.L. for the use of the neutron-beam facilities and SERC for financial support during the experiment. We are indebted to Dr. P. G. Slade for providing the sample of vermiculite. D.R.C. gratefully acknowledges receipt of a stud- entship from Steetley Brick and Tile Ltd.References 1 Z.L.L. Annu. Rep., 1985, 59. 2 G. W. Brindley and J. Lemaitre, in Chemistry of Clays and Clay Minerals, ed. A.C.D. Newman, Mineralogical SOC., London, 1987, p. 319. 3 F. Onike, G. D. Martin and A. C. Dunham, Mater. Sci. Forum, 1986, 7, 73. 4 J. Pannetier, Chem. Scr., 1986, 26A, 113. 5 Y. Laligant, G. Ferey and J. Pannetier, Chem. Scr., 1988, 28, 101. 6 A. A. Hassanipak and E. Eslinger, Clays Clay Miner., 1985, 2, 99. 7 H. Shiroza and S. W. Bailey, Am. Miner., 1966, 51, 1124. 8 P. G. Slade, P. A. Stone and E. W. Radoslovich, Clays Clay Miner., 1985, 33, 51. 8 P. G. Slade, P. A. Stone and E. W. Radoslovich, Clays Clay Miner., 1985, 33, 51. 9 P. G. Slade, personal communication, 1987.10 A. Antoniadis, J. Berruyer and A. Filhol, Acta Crystallogr., Sect. A, 1990,46,692. 11 V. Stubican, Miner. Mag., 1959, 32, 38. 12 V. Stubican and R. Roy, J. Phys. Chem. Zthaca, 1959,65, 1348. 13 R. Pampach, Polka Akad. Nauk, Prace Mineralogiczne, 1966, 6, 53. 14 M. Bulens and B. Delmon, Clays Clay Miner., 1977, 25, 271. 15 M. Bulens and B. Delmon, Bull. SOC. Chim. Belg., 1977, 86, 405. 16 J. Lemaitre, M. Bulens and B. Delmon, in Proc. Znt. Clay Conf., Mexico City, Applied Publishing, Wilmette, Illinois, 1975, 539. 17 G. W. Brindley, Prog. Ceram. Sci., 1963, 3, 3. 18 G. W. Brindley and M. Nakahira, J. Am. Ceram. SOC., 1959, 42, 311. 19 H. Yamadu and S. Kimura, J. Ceram. Assoc. Jpn., (YogkoKyokai Shi), 1962, 70, 65. 20 A. F. Wright and M. S. Lehmann, J. Solid State Chem., 1981, 36, 371. 21 H. D. Glass, Am. Miner., 1594, 39, 193. 22 H. M. Richardson, in X-Ray Zdentijcation and Crystal Structures of Clay Minerals, ed. G.W. Brindley, Mineralogical SOC., London, 1951, p. 76. 23 K. Tsuzuki, J. Earth Sci., 1961, 9, 305. 24 G. F. Walker, Clays Clay Miner., 1959, 4, 101. Paper 1/02056K; Received 1st May, 1991