AK Mundegar, Ove Arup & Partners, formerly Imperial College, London
This paper presents some of the findings of a laboratory study into the effects of platy mica particles on the behaviour of sand. The study has been conducted in the light of difficulties in construction of river training works for the Jamuna Bridge project in Bangladesh. The recent geology at this location is dominated by micaceous sand sediments that have brought about a number of challenges for designer and contractor alike.
It will be seen that mica can dramatically reduce the strength of sand, established methods of evaluating mica content have proven unreliable, and significant investment is required to improve our understanding of these materials before we can engineer with them efficiently.
The study formed part of a MSc dissertation by the author carried out at Imperial College during 1997 under the supervision of Professor DW Hight.
The Jamuna Bridge project
The Jamuna river is the fourth largest in the world, flowing through Bangladesh from north to south as shown on Figure 1. With an ever changing course and widths typically between 15km and 40km, this natural obstacle has had a major influence on the economic and demographic development of the country. In 1994 construction began on the Jamuna Bridge at a 14km wide stretch of the river. Half of the pounds450M contract comprised river training works designed to keep the river under the bridge.
One of the interesting engineering issues to arise during construction involved excavations near the toe of river banks. It was found that such activities were leading to successive shallow landslips progressing up slope and resulting in the loss of large volumes of material. This occurred on slopes as shallow as 1 in 5. Such a quick failure suggests that the slope is in a meta-stable condition and therefore movement occurs under a slight increase in shear stress. Such failures are brief and no significant drainage may occur prior to failure. Therefore the slope is judged to be essentially undrained at the time of failure.
During the site investigation it was discovered that the sediments in which the slopes were to be formed comprised micaceous sand. The mica is present in a flaky, plate-like form as compared to the rotund sand particles. This is illustrated in two-dimensions in Figure 2.
A variety of methods were adopted for the assessment of the mica content of sand samples with varying degrees of success. These included flotation and grain counting. Ranges of results were obtained suggesting typical mica contents of up to 40%. Figure 3 shows a magnified specimen of the micaceous sand from the Jamuna site. The darker areas are mica particles.
It is clear that counting the number of mica grains and sand grains would be a labour intensive and error ridden activity. This proved to be the case on the Jamuna project with re-counts showing significant variation. Flotation can be similarly problematic as not all of the mica may separate from all of the sand.
Current project objectives
It has been the aim of this study to add to the existing knowledge on how mica affects the behaviour of sand. A better understanding of the material may help avoid problems such as those experienced on the Jamuna Bridge project.
In general the effects of mica on porosity, compressibility and shear strength have been addressed. The effects on porosity and compressibility have been reported in the classic, yet largely neglected work of Gilboy (1928). This work was repeated to provide a frame of reference for the subsequent direct simple shear and triaxial compression testing.
The vast difference in the structure of a sand mass and mica mass is illustrated in the loosely dumped dry density of the sand and mica used in this study. The unit weight of the dry sand was 16kN/m3 in comparison to a mere 2kN/m3 for the dry mica. The significant conclusions from early parts of the work may be summarised as
the inclusion of mica in a sand mix has a profound effect on the porosity, and that porosity increases with mica content. An increase in mica content increases the susceptibility of the material to loading induced settlement (ie material stiffness is reduced).
Due to limited space the supporting results and triaxial test results are not discussed further here. This paper will simply highlight some of the results from constant volume direct simple shear tests in order to illustrate the significance of the mica inclusions.
As this study is set in the context of the Jamuna Bridge project it was desirable to employ similar materials to those found at the site. The grading of the parent material used in all parts of the current laboratory study lies inside the envelope for the Jamuna sand. A similar mica grading to that of the parent sand was also employed.
Simple shear testing
The simple shear apparatus (SSA) (described by Bjerrum and Landva, 1966) was implemented to investigate the effects of the mica on the shearing behaviour of the sand. Figure 4 is an idealisation of the sample undergoing simple shear during a test. The equipment is believed to impose stress conditions on a specimen closer to that pertinent to certain field cases than say that imposed in triaxial compression or extension. These cases include certain parts of potential failure surfaces under excavations or fills and material local to a pile foundation (with due account for specimen orientation) and, with relevance to this work, natural slopes.
By employing a computer controlled servo-feedback system the normal stress was controlled to maintain a constant volume. Such a test may be considered as representative of undrained tests in spite of the fact that the samples are tested dry (Dyvik et al, 1987)
Key results from simple shear testing
Figure 5 shows the results of a constant height direct simple shear test on a sample of pure sand (ie 0% mica content). Initially the sample was subjected to consolidation under a steadily increasing normal stress which was maintained at 100kPa for 10 minutes to allow any creep to reduce to a negligible level prior to shearing. It is seen that the pure sand initially contracts in shear then goes through a transition from contractant to dilatant behaviour. This phenomenon is termed phase transformation. Additional shearing serves only to increase the material strength and the test was stopped at a shear strain of near 20%. For comparative purposes the shearing resistance at a shear strain of 10% is given as 45kPa.
Figure 6 contrasts the results from tests on a 1% mica content sample and the pure sand. The 1% of mica has the effect of completely suppressing the sand's dilatant behaviour and introduces collapse. for comparative pur- poses the shearing resistance at a shear strain of 10% is given as 15kPA. This represents a reduction in undrained shearing resistance of about 65%. It is worth noting that both samples had similar gradings, similar void ratios and were prepared in the same manner. This leaves the presence of the small mass of platy particles as the source of such a dramatic alteration in behaviour.
Figure 7 shows results for pure sand (0% mica), 1%, 2.5%, 5%, 7.5%, 10%, 20% and 40% mica content samples sheared at constant height. The 40% mica content sample, similar to the intrinsic behaviour of lightly overconsolidated clay in undrained shear, tries to contract and even at shear strains of 17% shows no signs of dilatancy. As the mica content is increased from 1% a transition in material response is observed towards that of the 40% mica sample. Intermediate mica contents show a little dilation near failure. Closer examination in conjunction with the shear stress-shear strain plots suggest that at relatively small strains (up to 3% or 4%) it is the mica that dictates the overall behaviour of the composite material. However, beyond these strains and towards failure the sand begins to dominate and dilatancy becomes evident. The fact that the trend is not evident in the 40% mica test suggests that between 20% and 40% mica a transition takes place whereby the mica changes the failure line of the composite material. No dilatancy is observed and the material may be considered to have reached a critical state. In this study the pure sand was found to have an internal angle of shearing resistance of about 30o compared with 24o for the 40% mica sample.
Summary of key points from simple shear testing
Only 1% mica content (by mass) is sufficient to completely suppress any tendency for the sand to dilate. Further increases in mica content cause samples to exhibit transitional response, towards that of high mica content samples, with dilation near failure.
At low mica contents the sand fraction controls failure, demonstrated by a failure line consistent with that of pure sand. At high contents a shallower failure line demonstrates that the sand alone no longer controls failure. The transition between these characteristics is between 20% and 40% mica for the materials tested.
It is important to keep in mind that percentages of mica by mass can give a false impression of the amount of mica flakes present per sand particle.The mica used in this study has an aspect of ratio of around 50 and a particle grading roughly equivalent to that of the sand. This means that there are approximately 25 mica particles for each sand grain. This is illustrated in Figure 8. Therefore, in a 1% mica-sand mix there are about 25 mica particles in every 100 total. (This assumes comparable specific gravities and that the volume of an idealised sand sphere is half that of an equivalent cube taken up by 50 mica flakes side by side).
It is always important to have a good quality engineering geological appraisal of a site prior to design of works. Amongst other things, one must ascertain the size, shape and nature of any fines present in sands and especially the presence or otherwise of mica flakes. Gilboy's conclusion earlier this century that a system of soil classification is erroneous should it neglect the effects of plate-like particles has been reinforced by this study.
It has been demonstrated that percentage contents by mass can be misleading due to common misconceptions that the mass will translate into an equivalent number of particles due to similar gradings. In fact, for similarly graded materials 1% by mass may be around 25% by number of particles. This is believed to be a more pertinent description of the sand-mica mix. This poses a question on how to formulate consistent sample descriptions. Previous commercial work has shown the difficulties in separating mica flakes from the parent sand and therefore grain counting has been resorted to on occasion. This is of course time consuming and therefore expensive, not to mention open to significant human error. Due to the influence of small changes in mica content such errors may not be practically tolerable.
The laboratory investigation has revealed the dramatic effect of even 1% mica content (by mass) on the sand being tested. The sand was a dilatant, high strength material which had its large strain (approximately 20%) undrained strength reduced by a factor of about 3 by the addition of the platy particles. Such significant behaviour differences due to the inclusion of such small amounts of mica gives cause for concern over undrained events in the field in such sediments. Excavations as took place with serious consequences on the Jamuna Bridge project are examples of such events.
Undertaking geotechnical work in micaceous sands will continue to be highly challenging without investment in further work on these materials.
The dissertation which provided the basis for this paper formed part of the Imperial College Soil Mechanics MSc course undertaken by the author and funded by the Engineering & Physical Sciences Research Council. The author is indebted to the support from all staff and research workers at Imperial college and also at City University where the laboratory testing was undertaken. The supervisor for the research was Professor David Hight whose enthusiasm and encouragement were tireless. The mica samples used were generously donated by Microfine Minerals & Chemicals.
Bjerrum L and Landva A (1966) Direct simple shear testing on a Norwegian quick clay. Geotechnique 16 (1) :1-20.
Dyvik R, Berre T, Lacasse S and Raadim B (1987). Comparison of truly undrained and constant volume simple shear tests. Geotechnique 37 (1): 3-10.
Gilboy G (1928). The compressibility of sand-mica mixtures. Proc ASCE Transactions 54, pp 555-568.