Re-appraisal of Terzaghi’s soil mechanics Andrew Schofield, Emeritus Professor, Cambridge University

1. Re-appraisal of Terzaghi’s soil mechanicsAndrew Schofield, Emeritus Professor, Cambridge University • “Terzaghi and Peck” versus “Taylor” (Goodman p 213) • Civil engineering plastic design • Continuum of grains at repose (i) Coulomb’s and (ii) Rankine’s errors • Yielding of a saturated soil paste • Conclusion

2. D. W. Taylor (1900-55) Associate Professor, MIT • K. H. Roscoe taught hıs students to respect D.W.Taylor • Thex- y =xin “Fundamentals of soil mechanics” led us to an understanding of the mechanics of soil as an elastıc-plastic continuum

3. “Terzaghi and Peck” versus “Taylor” (1948) • Taylor's “interlocking” theory (1948) • Review of Taylor’s manuscript • John Wiley & Sons reply to Terzaghi • Critical state flow of grains without damageRoscoe, Schofield, and Wroth (1958)

4. Taylor's “interlocking” theory (1948) i x sand in y shear box • Workis x=x+y so strength is (friction) plus (interlocking) /=+y/x. y / y/x  x x x Increase of water content on slıck slıp planes shows that thıs applıes to“true” cohesıon of over-consolıdated clay

5. Peck’s review of Taylor’s manuscript • “I am convınced that the theorıes of soıl mechanıcs and the results of laboratory tests serve only to guıde the engıneer toward a recognıtıon of the factors whıch may affect the desıgn and constructıon of a real project” • from review sent to Wiley by R B Peck July 31 1944 quoted from page 213 “Karl Terzaghi; the Engineer as Artist” R E Goodman (1999)

6. John Wiley & Sons reply to Terzaghi... (Taylor’s book) will be published by one of our competitors if we do not take it. Under the circumstances, we see nothing to do but publish it.However, as I said in the first paragraph of this letter, we believe that each book will be judged on its own merits, and certainly we have no fears for the success of (Terzaghi & Peck).E P Hamilton (President) December 17, 1946

7. Roscoe, Schofield, and Wroth (1958) • q=p • Triaxial test paths approach steady flow in critical states with aggregates of grains at constant v specific volume • As strain increases v and p are constant at a Critical-state (v p) v= { v+lnp} =  wet dry

8. Critical state flow of grains without damage • Competent aggregate of selected sand grains flows in critical states v = v +  ln p = with no dust or damage • Soıl paste is unchanged in mixing or yieldingon the “wet” side, v> 

9. Plastic design in civil engineering • Construction without plastic ductility • Plastic design of a steel frame, Baker (1948) • Plastic design of structures • Ductility and continuity in soil mechanics • Strains by the associated plastic flow rule

10. Construction without plastic ductility Ductility can save life. The 1995 bomb at the Oklahoma Federal Centre, and similar damage in the 1999 Turkish earthquake, show the risk of brittle behaviour

11. Plastic design of a steel frame, Baker (1948) • Cambridge text book example plastic design of shelter to resist floor load 20 lb/sq.ft falling 9 ft in bombed house; Mother and 3 children survived WW II 250kg bomb in Falmouth, UK

12. Plastic design of structures • Small imperfections causes big local stress concentrations in elasticanalysis of steel frames • In practice plasticyield of steel relieves high stress • Ductility of steel gives safety, rather than high yield strength • Claddıng breaks up but framework survıves

13. Ductility and continuity in soil mechanics • A paste of soil saturated with water is plastic, (from the Greek word  plassein to mould, as in moulding pottery from clay). • An aggregate of separate hard grainsın a crıtıcal state behaves as a ductile plastic continuum. • Plastic design guıdes us to select construction materials and methods; soil is not plastic and ductile if over compacted to high peak strength

14. Strains by the associated plastic flow rule ip jp i i j (i j) j In plastic flow, as a body yields under combined stresses i j with strain increments ip jp, the flow vector is normal to the yield locus at (i j). For stability the product of any stress increment vector (i j) and the plastic strain rate flow vector may not be negative; iip + jjp> 0.

15. Calladine’s associated plastic flow (1963) • Yield loci for paste with v = (const) on wet side of Critical-states, satisfy the associated flow rule dpdv+dqd=0 • The Original Cam-clay locus was based on this plus Thurairajah’s dissipation function

16. A continuum of grains • Some historical dates • Belidor and Navier • Coulomb’s error • Rankine Activeslope at angle-of-repose • Drained angle-of-repose slope • Flow of grains with elastic energy dissipation • Elastic-plastic strains of aggregates of grains • Undrained and drained ultimate strength

17. Some historical dates • Coulomb, at school in Mezieres, learned friction theory from a text book written by Belidor in 1737 (reprinted with notes by Navier in 1819) and a Dutch concept of (cohesion) = (adhesion). In his 1773 paper he reported new rock strength data • Terzaghi (1936), in “A fundamental fallacy in earth pressure computation”, rejected Rankine’s theory of limiting statics of granular media, (Sokolovski), for lacking consideration of strains

18. Belidor’s friction hypothesis (1737) • Belidor attributed sliding friction coefficients of 1/3 to the hemispherical geometry of roughness • Navier (1819) called Belidor’s theory très-fautive but he offered no alternative to it.

19. Navier (1819); a footnote in his edition of Belidor

20. Coulomb’s soil (1773) Friction • Coulomb defined soil internal friction as the angle of repose d of drained slopes Grand rock face Canyon • soil slope

21. Coulomb’s soil (1773) Cohesion • In Coulomb’s rock tests, cohesion in shear was slightly greater than adhesion in tension, so he considered it safe to design with tension data • His wall design assumed that newly compacted soil has zero cohesion error

22. Terzaghi interprets Hvorslev’s (1937) shear box tests • Terzaghi fitted “true” cohesion and friction to peak strengths found by Hvorslev in shear box tests, normalısıng them by equıvalent pressure. wet sıde of crıtıcal states

23. A point Terzaghi missed in interpreting test data cs wet side Hvorslev’s data ended at a critical state point. Terzaghi should have asked Hvorslev why he put equations in space where there were no peak strengths. Filling the space meant that he asked no questions about the wet side of critical states v= { v+lnp} > 

24. Alternative strength components in soil paste • For Belidor (and Navier) the 2 soil strength components were (cohesion) + (interlocking = friction) • For Terzaghi (and Mohr) the 2 soil strength components were (truecohesion) + (truefriction) • Critical State Soil Mechanics has only 2 strength components (interlocking = cohesion) + (friction); it is a theory for dust with (truecohesion) = (zero)

25. Rankine Activeslope at angle-of-repose i Stress on a sloping plane  z  d  z cos d

26. Rankine Activeslope at angle-of-repose ii Stresses on sloping planes and on vertical planes are conjugate. Rankine hypothesised  that d is a limiting angle z  for both vectors of d stress, and also that both these planes slip.

27. Rankine Activeslope at angle-of-repose iii Slip lines are lines of constant length. If vertical lines had constant  length, all slope material z  would move forward d horizontally. If we accept Belidor’s error, (friction) = (dilation), no work is done or dissipated . Rankine (1851) should have deduced that slip planes are not planes of limiting stress. Terzaghi called Rankine’s earth pressures “fallacy”. Let us replace Rankine’s “loose earth” by an elastic-plastic continuum.

28. Drained angle-of-repose slope i Stresses on sloping planes and on vertical planes remain conjugate in a plastic  continuum. Instead of z d two sets of slip planes d in a Rankine Active zone r a let us have many ‘triaxial test’ cylinders in constant volume shear, giving plastic flow at all depths z

29. Drained angle-of-repose slope ii  z d d r a For q=(a -r) and p=(a+2r)/3 in triaxial tests, and q/p=3(ar)/(a+2r)=6sind/(3–sind)==(const), a continuum with (a/r)=(1+sind)/(1-sind)=(const), has constant slope angle das q and p increase, without the assumption of slip in two directions. Cırcle dıameters ıncrease wıth depth z.

30. Flow of grains with elastic energy dissipation • Elastic energy is lost on wood surfaces as fibre brushes spring free; Coulomb (1785) • Frameworks of soil grains carry load (after Allersma). Elastic energy is stored and lost as frameworks buckle

31. Elastic compression and swelling states with specific volume v, spherical pressure p, fitv = {v + lnp} Plastic compression fits v = {v + lnp} ,  are constants Elastic-plastic strains of aggregates of grains Plastic slope Elastic slope Taylor (1948) data

32. Elastic compression of aggregate fitsv={v+ lnp} A yield locus defines how elastically compressed grains yield when sheared  line shift v=vp gives plastic volume change (hardening) Elastic-plastic strains of aggregates of grains v =vp plastic volume change loci Roscoe and Schofield (1963)

33. Plastic compression is explained by  lines v line cs  line(-) lnp • Elastic compression  linesin plot of v=v+lnp against lnp go past the cs line v=v+(-)lnp= and yield at a  line. • Plastic compression in tests is observed to fit predicted stable yielding in (-) gap of v> lines

34. Undrained and drained ultimate strength • q=p • Crıtıcal States • Undrained strength c=cu with v=const., cu=/2exp{(-v)/ } • Drained strength in p=const. tests = =d=sin-13/(1+6/) • See Schofield and Wroth (1968) CSSM

35. Fall cone tests of mixtures of clay and silt Plasticity index IP is loss of water content for strength increase by factor of 100 (triaxial test data; Lawrence MPhil 1980) 80gm 240gm v v ln(penetration) • Fall cone tests with 80 and 240gm cones give v=lnp=ln3 • If pPL = 100 pLL then IP= 1.71  (from CSSM)

36. Yielding of a saturated-soil paste • Taylor / Thurairajah (1961) dissipation function • Paste mechanics Original Cam-clay (1963)

37. Taylor’s dissipation x- y =x (note ,x are orthogonal) Undrained and drained triaxial test data, including data of change of elastic energy, fit a function pdvp + qd = pd (p,dare orthogonal) Taylor / Thurairajah (1961) dissipation function

38. Original Cam-clay (1963) q/p=1-ln(p/pc) q cs(dv,d ) v q = p pcp dpdv + dqd = 0 associated flow pdv + qd =  p d dissipation function cs dv/d = -(dq/dp)= -(q/p). Introduce =q/p so d/dp=1/p(dq/dp-q/p)= -/p. Hence ln p d= -dp/p. When integrated this gives /=1-ln(p/pc).

39. Original Cam-clay (figure from my 1980 Rankine Lecture)

40. Original Cam-clay (1963) +(-) (-) q/p=1v   q/p=0  S 1 ln (p/pc)

41. Interım conclusions • Coulomb’s zero cohesion “Law” is confirmed by data on the wet sıde of crıtıcal states • Terzaghı’s Mohr-Coulomb error ıs clear • Map of soil behaviour (Schofield 1980) • Centrifuge work of TC2 up to 1998 • Choice between two liquefaction hypotheses

42. Coulomb’s zero cohesion “Law” is confirmed • Cam-clay model fits test data on the wet side of critical, which confirms Coulomb’s “law” that newly disturbed soil paste has zero cohesion • (CSSM figure; paste data (kaolin-clay)+(rock-flour) (Lawrence1980))

43. Terzaghı’s Mohr-Coulomb error • Terzaghi and Hvorslev wrongly claimed that true cohesion and true friction in the Mohr-Coulomb model fits disturbed soil behaviour. Geotechnical practice using Mohr-Coulomb to fit undisturbed test data has no basis in applied mechanics. • Critical State Soil Mechanics offers geotechnical engineers a basis on which to continue working. • The original Cam-clay model requires modification to fit effects of anisotropy and cyclic loading. Good centrifuge tests of soil-paste models achieves this.

44. Map of soil behaviour (Schofield 1980) Regimes of soil behaviour 1 1 ductile plastic 2 2 dilatant rupture 3 3 cracking 3 2 1 (fracture with high hydraulic gradient causes clastic liquefaction) A centrifuge test of a model made of soil paste will display integrated effects in behaviour mechanisms

45. Choice between two liquefaction hypotheses A A Casagrande Boston There is a unique critical void ratio and a risk of liquefaction in any embankment built with higher void ratio CVR

46. Choice between two liquefaction hypotheses B B Casagrande Buenos Aires Even a dense sand if heavily loaded can liquefy. Reject both A andB. Sand yields, it is stable, on the wet side of critical states Figure from Schofield and Togrol 1966

47. Centrifuge work of TC2 up to 1998 • We should claim a fundamental significance for centrifuge tests of models made of reconstituted soil, and explain how our tests can correct some errors that were made in Harvard. If it led to serious discussions in Istanbul, it would be good for Terzaghi’s Society. • A concluding comment on the Report of TC2 to the Istanbul Conference, Schofield (1998) Lecture in “Centrifuge 98 Vol 2 ”- IS Tokyo

48. Terzaghi’s low expectation for applied mechanics was in error when he said at Harvard (1936)...(the) possibilities for successful mathematical treatment of problems involving soils are very low • When I asked Bjerrum “What should Universities teach in soil mechanics?” he replied “Universities should not teach soil mechanics; they should teach mechanics” ( teaching in the spirit of K. H. Roscoe) • ISSMGE should correct error. We all should teach Plasticity and Critical State Soil Mechanics and promote centrifuge model tests with soil paste

49. Coulomb’s purpose in teaching soil mechanics • j’ai tâché autant qu’il m’a été possible de rendre les principes dont je me suis servi assez clairs pour qu’un Artiste un peu instruit pût les entendre & s’en servir • Teton photo from US Dept of Interior Bureau of Reclamation