I was interested to read Paul Wheeler's conversation with Professor Andrew Schofield in your August issue in which the latter deplores the use of the word 'cohesion' in soil mechanics. I believe the word 'friction' is also inappropriate.
In a typical cohesionless soil at working stress levels, the volume occupied by an interparticle contact zone (the zone formed by the merging of adjacent particles at their point of contact due to interparticle forces) is typically 106 - 109 times smaller than that occupied by the surrounding particles and voids. Consequently contact zone stresses exceed the yield stress of the minerals of which the particles are composed. A more appropriate description of the mechanical behaviour of the contact is that it is viscous in nature rather than frictional. Since at working stress levels it is likely the behaviour of a soil element would reflect that of the contact zones, one would expect soils to exhibit viscous behaviour in tests.
Analysis of data from creep tests carried out at differing stress levels suggests that soil deformation conforms to a viscous law in that the strain rate at a given time after loading increases as the stress level increases. In the case of triaxal tests, stress level may be defined as the ratio q/p (deviator stress divided by mean effective pressure). This is because an increase in p tends to increase the size of the contact zones by compression and hence their resistance to viscous flow, while increase in q tends to increase the shear component of force applied to the contacts, thereby encouraging flow.
The viscous nature of soil behaviour explains why the results from a series of conventional triaxial tests indicate a linear relationship between q at failure and p. Such tests are usually strain-controlled and consequently each sample is subjected to the same strain rate at failure. It follows from the viscosity law that the q/p ratio at failure is the same for all samples irrespective of the cell pressure used in each case.
The following approach would enable a designer to use a conventional stability analysis while recognising the viscous nature of soil deformation. Creep tests would be carried out on soil samples at differing stress levels (ie q/p values) in order to associate a working stress level value with the creep strain which could be accepted during the lifetime of a proposed soil structure. The use of this sub-failure value in a conventional stability analysis should then lead to a safe solution.
References may be found in my book Rheology and soil mechanics, published in 1984.
honorary research fellow
School of the Built Environment
Coventry CV1 5FB