ELSEVIER Engineering Geology 54 (1999) 187-194 Experience from preparation and investigation of clay microstructure Roland Pusch * 70 Lani, Sweden Geodevlopment AB, 2 Abstract ‘Clays without expanding minerals can be prepared so that their microstructural constitution is preserved. If expansive clays are treated in the same fashion there is an obvious risk of very significant distortion of the particle ‘arrangement because of drying, expansion and disintegration. I is also a risk that impregnated clay becomes too soft to be cut for microscopy, but experience shows that embedding in a suitable plastic monomer with subsequent polymerization gives suficently hard specimens to make ultramierotomy applicable. Impregnation with this substance requires that the pore water is first removed, which can be done in several ways as deseribed and discussed in this paper. The thickness of the thin section should be as small as possible for transmission electron microscopy, Le. a ew hundred angstroms, but thicker sections can be used for certain microstructural analyses, © 1999 Elsevier Science BLY. All rights reserve. Keywords Clay: Microscopy: Microstructure: Microtomy: Preparation 1. Introduction ‘The physical properties of clays depend on their microstructure, and its correlation with transport and rheology has been the scope of a large number Of investigations in soil mechanics and pedology. Microstructural quantification requires prepara- tion of representative two-dimensional pictures for the evaluation of void size and density, which excludes the application of scanning microscopy. Impregnation of the clay must be carried out to prepare hard specimens from which thin sections ccan be obtained using ultramicrotomy, atom mill- ing or conventional grinding. The required thick- PTeL: 4 86-46-286-2400, fax: + 46-46-286-2405, Email addres: puschegeodevelopment.ideon se (Pusch) ness depends on the purpose of the microstructural investigation Itis essential that the preparation does not alter the clay microstructure significantly. Minor changes related to a change in basal spacing of stacks of smectite lamellae may not be avoidable, but the overall pattern of varying void size and clay density must not be appreciably changed. It is hence necessary that the pore water be replaced by an impregnation liquid that can enter and fill the voids in approximately the same fashion as water. A second requirement is that the hardened impregnation material must give the elay a strength that permits sectioning without causing mechanical disturbance, Cutting has to be carried out by use of very sharp glass or diamond knives, Still, where the edge hits quartz, feldspar or heavy minerals, these may not be intersected or truncated but moved by the propagating knife, causing distur- (0013-7952,99)8~ se front matter © 1999 Elsevier Science B.V. Allright reserved Ul: $0013-7952(99)00073-3ee R Pusch | Engineering Geology $4 (1999) 187-196 ances such as folding and disruption of the clay matrix, Such defects will not be caused by alterna- tive preparation techniques like abrasion by neut- ron bombardment or atomic milling, but these may cause other problems, Quantitative interpretation of micrographs requires a definition of relevant microstructural parameters involving the determination of charac- teristic features, like the void size and density. The selection and use of parameters this paper 2. Preparation techniques 21, The traditional acrylate method Geological and medical disciplines were the first to utilize light microscopy and electron microscopy for examining thin sections. Blectron transmission microscopy (TEM) of clays based on experience from biology was introduced in about 1960, and its applicability has been assessed on several occa- sions (Pusch, 1967, 1998). A first attempt was to use acrylate plastics and similar substances, like Vestopal, Araldite, Epon and Durcupan, to replace the pore water. In recent years other materials, cg. SPURR (Smart and Tovey, 1982) and LR White hard resin (Ferrow and Roots, 1989), have been successfully used. For acrylate preparation, fon which we focus in the present paper, the following procedure has been widely used (Pasch, 1967) A prismatic specimen with a base area of a few square millimeters and a length of about 10 mm is cut from the water-saturated sample such that its orientation in situ can be defined, It is placed in a solution consisting of 50% by weight of ethyl alcohol and 50% distilled water for 30min fol- owed by emplacement in 90/10 ethyl alcohol/water and 99.59% ethyl alcohol for $ min each. Itis then transferred to a solution consisting of 85% by weight of butyl methacrylate and 15% methyl methacrylate for 45min, the process being repeated twice. Finally, the specimen is transferred to a solution of 98% monomer and 2% 2.4-dichlorbenzoylperoxide (EMW) catalyst for 90 min, Polymerization is obtained by heating to 60°C for 15h. While the traditional acrylate method works well for clays with illite and chlorite as dominant clay minerals, special measures have to be taken when treating expanding clays. Thus, very signifi- cant expansion and disintegration take place if the specimens are not effectively confined, 22, The ‘direct’ acrylate method The absence of pressure gradients means that stepwise replacement of the respective pore water solutions takes place by dlfusion, which is a rather slow process. The whole preparation sequence is therefore very time-consuming and attempts have been made to speed it up by drying the clay and then saturating it with the aforementioned solution consisting of butyl/methyl methacrylate and 2d-dichlorbenzoylperoxide catalyst in one step only. The disturbance caused by saturating the clay with monomer according to this simpler method may be less severe than that caused by the repeated exposure to diferent solutions of the traditional technique. However, the influence of drying prior to saturation may cause significant disturbance and needs to be considered. Dehydration by heat- ing clay specimens at 105°C or by freeze-drying causes shrinkage of the stacks of flakes by complete loss of interlamellar water, but freeze-drying pro- duces less overall shrinkage and partial preserva- tion of the continuity of the particle network. The difference is small for dense smectites but can be considerable for low densities, Slow drying at RTT 40-50% followed by quick vacuum treatment gives approximately the same disturbance as freeze- dlrying, atleast for high densities, and this method is promising for the preperation of clays with a density at saturation of more than 1700 kg/m. Fig. | illustrates the influence of interlamellar dehydration. Subsequent saturation with the ‘organic impregnation solution is estimated to bring the particle system into almost the same pattern as at water saturation, as concluded from measure- ment of the swelling pressure (Fig. 2)Pusch | Enginering Geology 54 (1999) 187-194 " Fig, 1, Structural effect on smestite clay by drying (Pusch and Karnland, 1986): (1) wet condition; (II) deed 10000 4 i i “Tine, dys Evolution of swelling presure by dry MX-80 sample absorbing the impregnation liquid aecording to the “direct method. The ultimate sweling pressure is practically the same 8 that produced at water saturation, 23. Thickness of clay section The ideal thin section is a two-dimensional plane. The thicker the section the smaller the possibility of identifying voids included in the clay ‘matrix, as illustrated in Fig. 3. On the other hand, a reduced thickness makes the section fragile and ‘experience shows that 200-400 A is a practical minimum when using ultramicrotomy for the cut- ting operation. The question of what thickness one DIRECTION OF OBSERVATION 189 ‘can accept depends on the required resolution and need for discerning microstructural details, High resolution transmission electron. micro- scopy (HRTEM) for revealing interlamellar spac- ing and the nature of interparticle contacts requires extremely thin sections, which are preferably pro- duced by atomic milling. This technique can give section thicknesses down to 10-20 A. To quantify microstructure one has to define what its components are. From a practical point of view, the fractions of permeable and imperme- able parts of representative sections through the original clay, and also of parts with different diffusion capacity, are primary microstructural parameters. They are related to the density varia- tion and the problem is hence to find out how the section thickness affects the possibility of evaluat- ing the density from examination of micrographs (Pusch, 1998), Fig. 4 shows schematic cross-sections of clay films with different thickness and with voids of different size. The issue is to find out how the electron absorption expressed in terms of “greyness” is related to the density. Very dark parts are strongly energy adsorbing microstructural compo- nents, like solid, non-smectite minerals or very dense smectite aggregates, while very bright parts represent open voids filled with impregnation sub- stance. Intermediate degrees of greyness represent clay gels of different density. A fundamental law in photography (Windisch, 1937) expresses the ratio of greyness G of the photographic plate, ie. the micrograph, and the radiation energy amount £ which is the product of the radiation intensity / and time of exposure a G=AE=logit DIRECTION OF OBSERVATION EMBEDDING SUBSTANCE Fig. 3, Influence of section thickness on the interpretation of the microstructure.190 Pusch | Enginering G J Oo ney 54 (1990) 187-194 Oo CS Co!) Ee Fig, 4, Clay fils withciferent thickness prepared from clay with the same void population, Lowest: Element with defined geometry Arrows indicate radiation (light or eleetron beam)Pusch | Engineering Geology $4 (1999) 187-194 191 radiation energy, (radiation intensity, f=time. The greyness Gis inversely proportional to the density p and thickness d of the film which, for constant j and 1, gives the expression: -B(pd)-* @ where B=constant, p=density, d=film thickness Distegarding geometzy-related optical distor tions and disturbance from diffraction and absorp- tion by the void-filing impregnation substance, the greyness of the micrograph intensity on the exit side of the single element in Fig. 3 BIL (od) @) ‘Comparing a film with 3 jim thickness with one ing 30 4m thickness the ratio of the G values is 10. To make a direct quantitative comparison of the G values possible, the radiation energy must hence be 10 times higher for the 30 um film thick- ness. This may have an impact on the crystal lattice stability, especially the location of alkali 24. Selection of section thickness While identification of the most detailed fea- tures of clay crystallites requires a resolution power that only films thinner than 500-1000 A can pro- vide, certain quantitative microstructural analyses can be made by using significantly thicker films Thus, if the objective is to interpret diferent degrees of greyness in terms of density variations, itis possible to do so if calibration has been made by analyzing ultrathin sections. For artificially prepared smectite clays using clay powder that hydrates under confined conditions, sufvient information may be obtained by examining 10 ym films. Hence, clays of this sort with a dry density 0 1700 to 2000 kg/m* are commonly characterized by the presence of one 3-5 jum wide channel filled with soft gels per 1000 um? cross-section area. Applying this to the lowest picture in Fig. 4, a 5 jum wide channel located in the center of the film ‘would mean that the thickness of the dense matrix (d, +4.) would be about 0% of that of the adja- cent homogeneous clay matrix. This means that the greyness of the microgeaph where the channel is located is 50% of that of the rest of the film, which is suflicient to indicate the presence of such a channel or local void. However, no detailed information on the channel structure and gel filling can be obtained. Furthermore, one cannot distin- guish between a case with one channel or void and ceases with two or several thinner channels or voids located on top of each other. A 30 ym clay film ives even less information on density variations, while films with a thickness of about S ym provide considerably more information on density varia- tions than the 10 ym sections. 3. Practical example 3.4. Clay material ‘An ongoing study of Tertiary Friedland Clay from Germany is taken here as an example of how a microstructural analysis can be performed, using the ‘direct acrylate’ method. The soil has a content of particles ner than 2 jm and of 40-50% expand= ing minerals. The clay was prepared by drying, grinding and compressing the clay powder to density that would have been 1910 kg/m? at water saturation, 3.2. Microscopy ‘Typical pictures ofthe clay are shown in Fig 5, while Fig. 6 shows their digitalized equivalents. The lower pictires inthe figures are micrographs of a 10ym section oblained by grinding and Polishing and used for examination by use of ordinary light microscopy. while TEM was used to obtain the micrograph representing the other specimen, which was 400 A thick and prepared by Utramicrotomy. The magnification was 100% for the light micrograph and 2200% for the 1 amicrograph 3.3, Microstructural quantification ‘The TEM picture reveals voids with a width down to about 0.1 jm and one can identify large, gelfilled or open voids with a size of up to about 20m. The typical pattern of curved stacks of1 R Purch | Engineering Geology $4 (1999) 187-196 Fig. 5. Micrographs of anifcally prepared Frisiland Clay. Upper: TEM pictare of 400 A section (length of bar is 5 um). Lower: Light mseroseopepictuse of 10 yz thik aetion (length of bars 100 um).R Pusch Engineering Geology $4 (1998) 187-194 198 Fig. 6. Digtalzed micrographs of artifcaly prepared Friedland Clay. Upper: TEM picture of 400 A section (length of ba eS um). Lower; Light microscope picture of 10pm thick section (length of ar at lower right corner is $0 pn). The colors represent eiferent densities: White (open voids), green (sot clay), ed (intermediately dense ley), black (very dense clay or non-smecite grains)194 R Pusch | Engineering Geology $4 (1999) 187-196 smectite lamellae is clearly discerned, and one sees black stripes which represent shear zones of aligned, densely grouped smectite stacks, As expected, the light microscope picture does not show any detailed microstructural features, except for the shear zones. However, variations in density and homogeneity can be observed and quantified, TThus, this latter micrograph shows the presence of about 10 voids with a width of 20-30 ym and three voids with a width of at least 50 pm. Voids smaller than 104m are embedded in the clay section as demonstrated by Fig. 4 and appear as local regions of low greyness. The digitalized version of the micrographs can be used for a number of statistical analyses. The graphs shown here were obtained by scanning the micrographs by use of the OFOTO code and then converting them into colored pixel patterns by applying the code GRAPHICAL CONVERTER ‘on an Apple Power PC 6100. Editing the pictures by using only four colors makes it possible to distinguish between open voids (white), gelfilled voids (green), smectite aggregates of intermediate density (red), and very dense aggregates or solid grains (black). Both pictures give approximately the same color distribution, i. white + green form- ing 47% of the total area for the TEM picture and 52% for the light microscope picture. 4, Discussion and conclusions Preparation of smectitic soil materials for microstructural analysis appears to be possible by using the same acrylate embedding technique that has been used for non-smectite clays for many years. Certain preparation-induced changes in the detailed arrangement and constitution of smectite particles cannot be avoided. However, since the swelling pressure recorded in the impregnation phase is almost the same as at saturation with water, and since hydraulic data evaluated from micrographs fit actually recorded values, itis con- cluded that clay micrographs obtained in the described fashion are relevant ‘A very important issue isthe selection of section thicknesses. It is concluded that accurate micro- structural analyses with respect to the density distribution can be based on TEM pictures of sulliciently large sections. Information on the scale dependence of the microstructure, respecting the frequency of bigger voids, requires analysis of larger section areas, which makes light micro- scopy us Acknowledgements ‘Some of the research forming the basis of this paper has been performed in the current EU Project ‘Microstructural and Chemical Parameters of Bentonite Determinants of Waste Isolation Efficiency’, Contract No. FI4WCT9S0012. The author expresses his gratitude to the Commission and to SKB, the Swedish Nuclear Fuel and Waste ‘Management Co. References Ferrow, E, Roots, W. 1989. A preparation technique for TEM ‘specimens, application to. synthetic Me-hlonte. Bur 1, 815-819. 1967. A tecigu for investigation of clay miero- ‘tructse. J. Microsc. 6, 963-986, Pusch, R, 1998, Evolution of clay microstructure, Proe. It ‘Workshop on Microstructural Mogelling, Land, 12-14 October, 1-38 Pusch, R, Karan, ©., 1986. Aspects ofthe physical state of ‘smecite-adsorbed water, SKB Technical Report TR 46-25. Swedish Nuclear Fuel and Waste Management Co, Stockh Smart, P, Tovey, NK, 1982, Blectron Microscopy of Sole and ‘Sediments, Clarendon Press, Oxford, pp. 60-10. Windisch, 0, 1957. Die Neue Fotosshnle. Heering-Verag, Haraburg, p, 123