Behavior of Five Large Spread Footings in Sand

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

BEHAVIOR OF FIVE LARGE SPREAD

FOOTINGS IN SANDa

Discussion by Nilo C. Consoli3

and Fernando Schnaid4

The authors present results from a series of load tests per-formed on five square spread footings in sand that are valuablefor examining several factors including creep settlement andcreep rate effects. In particular the paper describes an empir-ical method to estimate the time-dependent behavior of a sandfrom a plot of log s/s1 versus log t/t1, as shown in Figs. 11–14. The slope of a straight line regression gives the creep ex-ponent n, regarded by the authors as a material property. Thediscussers present data from a homogeneous residual soil site(Schanid et al. 1995; Consoli et al. 1998) that show the creepexponent n not increasing continuously with increasing rela-tive load (Q/Qu).

We would like to argue on the basis of recent data fromfooting loading tests carried out at Federal University of RioGrande do Sul, Brazil, on a lightly cemented sandy-silt redclay. Typical results are shown in Figs. 25 and 26, plotted inthe same fashion as proposed by the authors in their originalpaper. Results shown in Fig. 25 for monotonic loading dem-onstrates the existence of three distinct regions on a creep ex-ponent (n) versus relative load (Q/Qu) plot—a first region as-sociated to an exponent n close to zero in which creep effectscould be ignored for increasing monotonic loading; a secondregion in which n values, and therefore the creep rates, in-crease with increasing relative load up to a given threshold;after such a limiting value there is a drastic change in responsewith the exponent n approaching a constant final value (despitethe observed scatter). In the case of unloading-reloading cycles(Fig. 26), there is a drop in n values during unloading, whichis followed by a subsequent increase of n when loading isresumed. There are no significant changes in the slope ob-tained in the first load stage and in the reload stage during acycle. As a first insight it is suggested that creep effects arenot unique and will produce different patterns in the elasticand plastic ranges.

The effects of creep from relatively small to intermediatestrains on soil behavior are less understood than the effects onthe yielding and strength characteristics. The study of the time-dependent deformation property of sands, gravels, residualsoils, and soft rocks is indeed limited (e.g., Murayama et al.1984; Pestana and Whittle 1998). However, the discussers feelit would be appropriate to interpret footing response in Briaudand Gibbens’ work, as well as in their own research, withinthe context of viscous-elastic-plastic theory. At early loadingstages, soil elements below the plate respond predominantlyas a linear elastic material. Creep may influence even thesmall-strain response of the soil below the plate, but this effectappears to be small and might be related to n values close tozero, as suggested by Tatsuoka et al. (1999) and observed inour own data for plates tested under monotonic loading (Fig.25). Note that since residual soils always exhibit some struc-ture because of cementation among particles, creep is likely tobe initially inhibited. Values of n close to zero are also ob-served during the unload-reload cycle shown in Fig. 26, when

aSeptember 1999, Vol. 125, No. 9, by Jean-Louis Briaud and RobertGibbens (Paper 13680).

3Assoc. Prof., Dept. of Civ. Engrg., Federal Univ. of Rio Grande doSul, Brazil.

4Assoc. Prof., Dept. of Civ. Engrg., Federal Univ. of Rio Grande doSul, Brazil.

940 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGIN

J. Geotech. Geoenviron. E

FIG. 25. Creep Exponent versus Relative Load for 1.0 mSquare Footing

FIG. 26. Creep Exponent versus Relative Load for 0.7 mSquare Footing

the soil responds essentially as a linear elastic material. Theelastic zone developed by preloading appears to reduce resid-ual displacements for subsequent live loads that are keptwithin this elastic zone. At intermediate strains, there will besome elements that are already experiencing plastic strains,and at these load stages the dependency to viscous effects isprogressively greater. The creep rate increases with increasingaverage stress ratio s1/s3 and becomes significant as the stressstate approaches the failure state on soil elements below theplate. With increasing average shear strains, viscosity can nolonger be ignored and creep acceleration might occur for in-creasing load stages. This effect can be generally representedby the increase in the creep exponent n with increasing relativeload, as suggested in the original paper. The generalization oflarge strains within the main bulb of stresses below the footingimposes a plastic response to the footing under increasing

EERING / OCTOBER 2000

ng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

load. As the elements loaded in the active zone approach stressratios close to that at failure, the foundation starts to exhibit alarge creep compression that increases progressively to a con-stant final creep exponent, the creep exponent being definedby (1) of the original paper.

In conclusion, we feel that the issue of predicting the long-term displacements of structures is far from being properlyunderstood. Interpretation of results as proposed by the authorsprovides a simple framework for a qualitative assessment ofcreep settlement, but (a) it is premature to consider the creepexponent n as a material property, and (b) the use of empiricalcorrelations requires independent judgment in practical appli-cations.

APPENDIX. REFERENCES

Consoli, N. C., Schnaid, F., and Milititsky, J. (1998). ‘‘Interpretation ofplate load tests on residual soil site.’’ J. Geotech. and Geoenvir. Engrg.,ASCE, 124(9), 857–867.

Murayama, S., Michihiro, K., and Sakagami, T. (1984). ‘‘Creep charac-teristics of sands.’’ Soils and Found., 24(2), 1–15.

Pestana, J. M., and Whittle, A. J. (1998). ‘‘Time effects in the compres-sion of sands.’’ Geotechnique, London, 48(5), 695–701.

Schnaid, F., Consoli, N. C., Cudmani, R. O., and Milititsky, J. (1995).‘‘Load-settlement of shallow foundations in structured unsaturatedsoils.’’ Proc., 1st Int. Conf. on Unsaturated Soils, A. A. Balkema, Rot-terdam, The Netherlands, 999–1004.

Tatsuoka, F., Jardine, R. J., Lo Presti, D., Di Benedetto, H., and Kodaka,T. (1999). ‘‘Theme lecture: Characterizing the pre-failure deformationproperties of geomaterials.’’ Post Conf. Proc., 14th Int. Conf. on SoilMech. and Found. Engrg., A. A. Balkema, Rotterdam, The Netherlands,2129–2164.

Discussion by D. R. Phatak5

and H. B. Dhonde6

The authors have carried out experiments on five largespread footings in sand. Incidentally, this paper has given achance to the discussers to present a novel method followedby them to predict load-settlement characteristics of prototypefootings based on observations made on single footing. Thisdiscussion is written to illustrate the method to the engineersand prove it for the observation set provided by Profs. J. L.Briaud and R. Gibbens.

The method is based upon the principles of dimensionalanalysis. To formulate the requisite equation the discussers firstuse the experimental load-settlement data of the 1.0 m squarefooting; we have

V = {F, g, f, S, D, B, L} (26)

where V = set of all the parameters involved in the phenom-enon; F = load, which is a dependent parameter; and all othersare the independent parameters, wherein the symbols have thesame meaning as in the original paper.

The number of the parameter is 7. Writing dimensions ofboth the sides of (26),

22 22 2V = [MLT , ML T , 0, L, L, L, L]

Let P = force = MLT 22; therefore V = [P, PL23, 0, L, L, L,L], thus m, number of independent primary dimensions = 2.Employing Buckingham’s P-theorem, the dimensionlessgroup P-term will be 7-2 = 5. Selecting a set R from V insuch a way that none of the variables in the set have the same

5Prof., Civ. Dept., Government Engineering Coll., Pune-5 (MS), India.6Postgrad. Student, Civ. Dept., Government Engineering Coll., Pune-5

(MS), India. E-mail: [email protected] or [email protected]

JOURNAL OF GEOTECHNIC


TABLE 6. Comparison between Calculated and MeasuredLoads at 25 mm Settlement

Trial no.(1)

Footing size(m 3 m)

(2)

Measured load@ S = 25 mm

(MN)(3)

Calculated load by (30)@ S = 25 mm

(MN)(4)

1 0.991 3 0.991 0.85 0.842 1.492 3 1.505 1.5 1.543 2.496 3 2.489 3.6 2.934 3.004 3 3.004 4.5 3.75 3.016 3 3.023 4.5 4.52

TABLE 5. Trial-and-Error Procedure with Values of Constantsfor (29)

Trialno.(1)

b1

(2)b2

(3)b3

(4)b4

(5)b5

(6)

L.H.S.of (29)

(7)

R.H.S.of (29)

(8)

1 1.2 3 105 0.8 20.4 20.5 20.5 54.84 52.402 3 3 105 0.8 20.4 20.5 20.5 54.84 62.103 2.5 3 105 0.8 20.4 20.5 20.5 54.84 57.504 2.25 3 105 0.8 20.4 20.5 20.5 54.84 55.045 2.22 3 105 0.8 20.4 20.5 20.5 54.84 54.736 2.23 3 105 0.8 20.4 20.5 20.5 54.84 54.84

dimensions nor any dimensionless parameters, we have setR = {F, g, D}; repeating variables from set R are chosen as{F, g}, {g, D}{F, D}. We can write, P1 = {F, g, f}, P2 ={F, g, S}, P3 = {F, g, D}, P4 = {F, g, B}, and P5 ={F, g, L}. Let P1 = Fa 3 gb 3 fc. Therefore, M 0 L0 T 0

= (M1 L1 T 22)a 3 (M 1 L22 T 22)b 3 (M 0 L0 T 0)c. Equating theindices, a 1 b = 0, a 2 2b = 0 and 22a 2 2b = 0. Thisleads to a = 0, b = 0, and c = 1. Substituting these values ofthe indices, we get P1 = f.

Similarly, solving for other P-terms we get P2 = Fg21S23,P3 = Fg21D 23, P4 = Fg21B 23, and P5 = Fg21L23. Thereforewe can write,

21 23 b 21 23 b 21 23 bP = Fg D = b (f) 3 (Fg S ) 3 (Fg B )3 1 2 3 4

21 23 b3 (Fg L )5 (27)

b 23b 3 23b 23b 1/(12b 2b 2b )3 4 5F = g [b f 3 S 3 D 3 B 3 L ] (28)1 2 3 4 5

Now using guiding point as the load at 25 mm settlement ofthe footing of size 0.991 3 0.991 m, D = 0.711 m, g = 0.0155MN/m3, and f = 327, for which the measured load F = 0.85MN, and substituting these values in (28), we have

3 b 23bF = 0.85 = 0.0155 MN/m 3 [b (327) 3 (0.025)1 2 3

3 23b 23b 1/(1 2 b 2 b 2 b )3 4 53 (0.711) 3 (0.991) 3 (0.991) ]4 5 (29)

By trial-and-error method the values of constants b1, b2, b3,b4 , and b5 from (29) are found out. While deciding the valuesand signs of the constants, the behavioral patterns of the in-dependent variables and their effect on the dependent param-eters should be considered. Table 5 below illustrates this pro-cedure. Using the values from trial No. 6 and substituting in(29), we have

0.33 0.5 1.25 0.63F = 169gf S D (BL) (30)

To check the validity of (30), the discussers have used an-other data of 1.492 3 1.505 m footing with D = 0.762 m, g= 0.0155 MN/m3, S = 0.025 m, and f = 327, for which theexperimental measured load (F) is 1.5 MN. Substituting theabove data in R.H.S. of (30) calculated load is 1.54 MN asagainst measured 1.5 MN. Eq. (30) is thus validated. Table 6below shows the comparison of the calculated load by (30)and the measured load at 25 mm settlement.

Table 6 clearly shows that the prediction of load by for-

AL AND GEOENVIRONMENTAL ENGINEERING / OCTOBER 2000 / 941

ng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

mulation of equation by this method is reasonably close. It isto be further noted that the theoretical methods such as Briaud(1992), Burland and Burbidge (1985), Peck and Bazarra(1967), Terzaghi and Peck (1967), etc., have always shownlarge divergence from the field observations, as is evident fromTable 4 of the original paper, whereas the proposed methodshows less divergence. The advocated method, which is es-sentially based on the observations of load-settlement responseof two footings, has predicted the load-settlement character-istics of various other footings correctly. Therefore, the dis-cussers advocate this method. The discussers thank the authorsbecause it has given them the chance to explain their methodand to prove it for the data provided by the authors.

EFFECTS OF PLACEMENT METHOD ON

GEOTECHNICAL BEHAVIOR OF

HYDRAULIC FILL SANDSa

Discussion byJames A. Sladen,5 Member, ASCE

The authors have presented a most interesting case historyconcerning the geotechnical behavior of hydraulic fill sand.The case history illustrates very well the practical difficultiesthat such issues as nonuniformity can present to the geotech-nical engineer struggling to interpret site investigation datawithin the framework of available tools. These tools often de-rive from a mixture of theoretical and laboratory studies, bothof which are generally idealized to some degree.

aOctober 1999, Vol. 125, No. 10, by K. M. Lee, C. K. Shen, D. H. K.Leung, and J. K. Mitchell (Paper 14039).

5Dir., S P Associates, 143–145, Worcester Rd., Hagley, West MidlandsDY9 0NW, U.K.



This discussion concerns only one aspect of the paper, theassessment of in situ sand density and, in particular, the degreeof confidence that can be placed in available methods. Knowl-edge of the in situ density, or void ratio, and stress level arefundamental to any assessment of likely behavior of recentlyplaced sands. Although other factors are potentially of rele-vance, unless these conditions can be established with somedegree of confidence, some methods of behavioral prediction,particularly those concerning potential for liquefaction, maybe difficult to apply.

The authors compare the dry densities determined from Ma-zier tube samples with those from in situ nuclear-density-gauge and sand-cone tests and note that the differences in drydensities obtained from these tests are generally less than 6%.Based on this comparison they conclude that the Mazier tubesamples are generally of high quality and can be consideredreasonably undisturbed samples. The discusser considers thisto be somewhat misleading as, from the authors’ Table 1, thedifference between the maximum and minimum dry density isitself only about 18% of the maximum dry density so that adifference of 6% corresponds to a difference in relative densityof about 33%. Such a difference may be significant with regardto some potential behaviors. This is illustrated in Fig. 19, inwhich the data from the authors’ Fig. 7 have been replotted interms of relative density. The results of the nuclear densitygauge would appear to be particularly erratic. This does notnecessarily mean that the Mazier tube samples are not of highquality but it does perhaps mean that this has not been dem-onstrated. The average relative densities for the top 4 m fromthe three methods are given in Table 5.

At any one depth, the variation can be much greater, asshown in Fig. 19. Although some apparent variations in rela-tive density may reflect nonuniformities in sand grading, etc.,and consequent variations in the maximum and minimum ref-erence states, such variations would be real and would con-tribute to potential difficulties in interpreting the results of insitu density measurements. The above can be contrasted withrelative densities of Mazier tube samples from below the watertable, where densities are believed to be lower, averagingabout 50%, as shown on the authors’ Fig. 17. The authorsdescribe the fill above sea level as medium dense to verydense, based on both SPT and CPT data. They attribute thisrelatively dense state to three factors, as follows:

FIG. 19. Comparison of Relative Density Determined by Various In Situ Measurement Methods


ng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

TABLE 5. Density from Three Methods

Method(1)

Average relative density(%)(2)

Nuclear density gauge 30Sand cone 29Mazier tube 17.5

• Compaction brought about by seepage• Possible overconsolidation due to seepage• Mechanical compaction due to leveling operations

The mechanics of the first two factors and why they wouldonly be operative above the water table are not clear to thediscusser. Seepage per se is not usually associated with com-paction. Excess water in fill above sea level where a wide areais being filled must at least partly drain downward so thatseepage would occur both above and below the water table—as it also would, presumably, as a result of tidal effects. Theoverconsolidation ratio due to tidal fluctuations would be ex-pected to be most pronounced just below the water table. Thethird factor is certainly important and is widely viewed as areason for generally better densities in fill placed above water,but the differences on the West Kowloon site seem particularlymarked. Whatever combination of factors is causing the higherrelative densities above the water table, it is apparently notalways operative, as is evidenced by the results of the otherin situ tests at the trial pit location.

The availability of a site-specific chamber calibration for theCPT is a major advantage and should allow interpretation ofrelative density from CPT data with a relatively high degreeof confidence.

There would appear to be a major discrepancy, one recog-nized by the authors, between the CPT data presented and theother data concerning in situ density, within the top 4 m, forwhich detailed information is provided. For example, a CPTtip resistance of over 20 MPa is recorded at 4 m depth, wherea total stress of the order of 80 kPa would be anticipated.Based on the authors’ Table 4, sand with a relative density ofeven 80%, would be expected to achieve a tip resistance of20 MPa only at an effective stress level of the order of 300kPa. The authors state that the CPT data for partially saturatedsands cannot be referred to the same density-cone tip resis-tance relationships as would be expected for saturated soilsbecause of the negative pore pressure (due to capillarystresses) developed on the partially saturated sand fill, whichwould affect the effective confining stress acting on the sandparticles. If this were true and capillary stresses could resultin such a large apparent discrepancy, then there would be se-rious implications for the usefulness of the CPT for any ap-plication above the phreatic surface.

The general influence of effective stress level on tip resis-tance in sand is, however, reasonably well understood, and theauthors’ chamber calibration data illustrate this influence di-rectly for the West Kowloon sand. For the Ticino sand, Baldiet al. (1986), based on regression analysis, have presented arelationship between relative density Dr, qc and mean effectivestress p9, which can be rewritten as

ln q = (D 1 21.1 ln p9 1 199)/38.3 (2)c r

where Dr is in percent, and qc and p 9 in kPa. This relationshipis entirely empirical but it allows the general influences ofstress and density on qc to be separated in at least an approx-imate manner. A similar regression analysis of the ‘‘free-field’’data presented by the authors for the West Kowloon sand re-sults in the following approximate relationship:

JOURNAL OF GEOTECHNIC


ln q = (D 1 21.9 ln p9 1 118.1)/31.4 (3)c r

Although this does not have the theoretical rigor of the Sal-gado et al. (1997) method used by the authors, inspectionshows that it ‘‘predicts’’ qc within the same levels of confi-dence, and, for simplicity, it is adopted here to illustrate thepotential magnitude of capillary effects.

The exact magnitude of potential increased mean effectivestresses due to capillary effects is not known, although usuallythese are not considered large for sands. Very much an upperbound could be considered to be given by the height abovethe phreatic surface times the density of water. Also unknownis the value of the ratio of vertical to horizontal effective stress,Ko. To accommodate these uncertainties, the density profileinferred from the CPT data is shown in Fig. 19 for two con-ditions: firstly, without any allowance for capillary effects andassuming a typical ‘‘normally consolidated’’ value of Ko of 0.5,and secondly, assuming the above-mentioned upper boundcapillary effect and with Ko = 1.0, which might correspond toan overconsolidation ratio of the order of 4. These conditionscould be assumed to bracket the potential range of in situeffective stress conditions. Nevertheless, as can be seen, theCPT data for both conditions imply much higher relative den-sities than the other in situ tests. It can be concluded that thediscrepancy is not due solely, or even predominantly, to cap-illary effects.

In view of the above observations, the discusser would beinterested in the authors’ opinions concerning possible otherexplanations for the apparent discrepancies between the vari-ous in situ tests and the confidence levels that can be placedin relative density determinations in such circumstances.

APPENDIX. REFERENCE

Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M., and Pasqualini, E.(1986). ‘‘Interpretation of CPTs and CPTUs. Part 2: Drained-penetra-tion of sands.’’ Proc., 4th Int. Geotechnical Sem. on Field Instrumen-tations and In Situ Measurements, Singapore, 143–156.

Closure by K. M. Lee,6 C. K. Shen,7

D. H. K. Leung,8 and J. K. Mitchell9

The writers wish to thank the discusser for his interest inthe paper and for his valuable comments. It is noted that theissues raised in the discussion are largely concerned with theassessment of in situ sand density and the degree of confidenceof the interpreted results. These issues would definitely pro-vide additional insights into the problem. However, the fol-lowing comments are made to clarify some of the main issues.

First, the Mazier samples were taken from a location ap-proximately 5 m away from the test trench. It was observedfrom the site that, even at such a close proximity (relative tothe size of the reclaimed land), nonuniformity and segregationof the hydraulic fill sands at the two sampling locations werenoticed. At any given depth, variations in grade-size distribu-tion and consequently variations in the maximum and mini-mum dry densities are expected to appear at these locations.Thus, rd, max and rd, max shown in Table 1 represent only theaverage properties of the bulk sample. It would be unrealistic

6Asst. Prof., Dept. of Civ. and Struct. Engrg., Hong Kong Univ. ofSci. and Technol., Clear Water Bay, Hong Kong.

7Prof., Dept. of Civ. and Struct. Engrg., Hong Kong Univ. of Scienceand Technology, Clear Water Bay, Hong Kong.

8Geotech. Engr., Ove Arup & Partners (HK) Ltd., 56/F Hopewell Ctr.,183 Queen’s Rd. East, Hong Kong.

9Univ. Distinguished Prof., Dept. of Civ. Engrg., Virginia PolytechnicInst. & State Univ., Blacksburg, VA 24061-0105.

AL AND GEOENVIRONMENTAL ENGINEERING / OCTOBER 2000 / 943

ng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

to apply these averaged values to the interpretation of relativedensities at different sampling locations.

Secondly, the writers agree on the fact that the sample dis-turbance effect of the Mazier samples are difficult to quantify,and comparison of dry density alone is not a highly reliablemeans to indicate the quality of the samples. However, becauseof the nonuniformity and segregative nature of the fill sand,as stated previously, relative densities at different samplingdepths cannot be simply deduced from the bulk values shownin Table 1. At best, the information that we have on the Maziertube samples can only indicate that the samples were reason-ably undisturbed.

Thirdly, the CPT and SPT values (and consequently, thecompacted density) of the fill above sea level were found tobe significantly higher than that below sea level. The mechan-ical compaction due to leveling operations would undoubtedlyincrease the relative density of the sand fill. The ‘‘compaction’’effects by seepage can be considered to develop in two as-pects: (1) The downward drainage of water in the sand-waterdischarge above sea level would induce downward seepagepressure on the sand fill. Along the sand-fill slope formed bythe discharge, individual sand grains settle in a turbulent cur-rent while the sand mass is subjected to a downward seepagedue to the drainage effect of the sand body. Each grain willfinally rest in a stable position, forming a sand body with alower void ratio than under the submerged deposition position;and (2) Segregation in hydraulic fills above the water levelwill inevitably take place, with the coarse fractions mainlysettling in the upper parts of the slope and the finer fractionscarried by the seepage water and deposited at the lower layersbelow water level. In the end, the top layers of the fill willconsist of coarser material than the lower layers close to thewater level. Furthermore, the tidal fluctuations, as suggestedby the discusser, also induced overconsolidation effect on thesand fill above sea level. However, the most significant effectis induced by the leveling operation of bulldozers, which workon the slope to guide the sand-water mixture flow. Usuallythese processes result in cone-tip resistance qc values in therange of 10 to 20 MPa, as suggested by Verkerke and Volbeda(1991). However, it should be noted that, if compaction is notproperly conducted by bulldozers or other leveling equipment,qc values as low as 4 to 5 MPa have been reported (Shen etal. 1997) at the West Kowloon reclamation site.

Lastly, Fig. 12 of the original paper shows the grain-sizedistribution of the bulk sample for calibration chamber testingas well as the grain-size distributions of the Mazier samplesat different depths. The variability in particle-size distributionsis clearly evidence for the hydraulic fill sands, as observed inthe figure. Thus, it would be unrealistic to adopt an empiricalcorrelation qc-Dr-p9 based on the bulk sample and apply it tothe data obtained from the test trench.

Variations in placement density and grain size must be rec-ognized in hydraulic fill because different placement methodsare employed during the filling process and the segregationeffect of coarse and fine particles after hydraulic placement.CPT sounding is a direct and effective way for assessing theplacement density of hydraulic sand-fill in the field. To inter-pret the penetration test results, site specific calibration cham-ber tests can be used to provide the qc-Dr-p9 relationship. How-ever, because of the cost constraints for most hydraulic fillingprojects, only limited batches of bulk samples are used for thecalibration chamber testing. The established qc-Dr-p9 relation-ships may not be truly reflecting the potential variability ofthe grain-size distributions as what occurred in a hydraulic fillsite. Engineering judgment, therefore, must be exercised inorder to interpret the CPT penetration results. One of the ur-gent researches for the interpretation of CPT soundings forhydraulic fill sites is to establish the correlation of qc-Dr-p9



FIG. 10. Sulfate Attack of Concrete Patio (Photograph by EricLind)

with grain-size distribution of the soils, at least for a givenmineralogy of sand. Such an exercise could be used to assistengineers in the interpretation of CPT results for segregatedhydraulic fill sands.

SULFATE ATTACK ON

CEMENT-STABILIZED SANDa

Discussion by Robert W. Day,4 Fellow, ASCE

The authors have prepared an excellent practice-oriented pa-per on sulfate attack of a cement-treated base. Also, Rollingsand Rollings should also be congratulated on their book Geo-technical Materials in Construction (1996), which is the bestreference material on this topic.

In terms of damage, the sulfate attack on the cement-treatedbase seems minor compared with some of the damage ob-served by the discusser. Some of the most significant sulfatedamage observed by the discusser is in Las Vegas, Nevada,which as a high surface concentration of soluble sulfate. Forexample, sulfate damage to a concrete patio and walkway areshown in Figs. 10–12. The salts in the soil can also attacksteel reinforcement or posttensioning foundation cables, suchas shown in Fig. 13. Even more remarkable, the salts canliterally disintegrate surface aggregate (Fig. 14).

Using Type V portland cement and a low water-cement ratio

aMay 1999, Vol. 125, No. 5, by Raymond S. Rollings, J. Pete Burkes,and Marian P. Rollings (Paper 15979).

4Chief Engr., American Geotechnical, 5764 Pacific Center Blvd., Ste.112, San Diego, CA 92121.


ng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

FIG. 11. Close-Up View of Fig. 8 (Photograph by Eric Lind)

FIG. 12. Sulfate Attack of Concrete Sidewalk (Photograph byEric Lind)

does not appear to be totally effective in preventing concretedeterioration. The discusser believes that this could be due tothe arid environment, where water-borne sulfate is drawn upinto the concrete and then deposited on crack faces when thewater evaporates. This causes the sulfate to become concen-trated on the crack faces, leading to accelerated deterioration.

One possible solution (in addition to using Type V cementand a low water-cement ratio) would be to isolate the concretefrom the capillary process. Thus a capillary break consistingof open-graded gravel could be placed below the concrete. Thegravel should extend outside the limits of the concrete so

JOURNAL OF GEOTECHN

J. Geotech. Geoenviron

FIG. 13. Salt Attack of Posttensioning Cables

FIG. 14. Salt Deterioration of Aggregate (Note: Originally, En-tire Ground Surface Contained Open-Graded Gravel)

that the edges are also protected. Gravel must be selected thatwill not disintegrate in the presence of salts, such as shown inFig. 14.

Closure by Raymond S. Rollings,5

J. Pete Burkes,6 and Marian P. Rollings7

The writers would like to thank the discusser for his kindcomments, and we found his descriptions and photographs ofsulfate attack on conventional portland cement concrete in LasVegas most interesting. The discusser observed that Type Vcement did not always protect against sulfate attack damagein the severe exposure conditions found in Las Vegas. DePuy(1994) and ACI Committee 201 (1991) recommend that forvery severe exposure one should use Type V cement combinedwith a pozzolan or slag that has been shown in appropriatelaboratory tests to be effective in mitigating sulfate attack.

5Res. Civ. Engr., U.S. Army Engineer Waterways Experiment Station,3909 Halls Ferry Rd., Vicksburg, MS 39180.

6Res. Petrographer (retired), U.S. Army Engineer Waterways Experi-ment Station, 3909 Halls Ferry Rd., Vicksburg, MS.

7Res. Civ. Engr., U.S. Army Engineer Waterways Experiment Station,3909 Halls Ferry Rd., Vicksburg, MS.

ICAL AND GEOENVIRONMENTAL ENGINEERING / OCTOBER 2000 / 945

. Eng. 2000.126:940-942.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Ufr

gs -

Uni

vers

idad

e Fe

dera

l Do

Rio

Gra

nde

Do

Sul o

n 06

/10/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

Very severe exposure conditions were defined to be sulfateconcentrations of greater than 10,000 ppm in water or over2% by mass in soil. Consequently, the discusser’s observationsupports these recommendations that we may encounter sulfateexposure conditions where simple reliance on sulfate-resistingcements (Type II and V) may not be sufficient to achieve theneeded levels of protection.

While damage from sulfate attack of conventional portlandcement concrete and of the cement-stabilized base course de-scribed in this paper are both caused by the formation of et-tringite, there is one insidious difference. Sulfate attack of con-ventional portland cement concrete can be reduced or evenprevented by reducing the alumina content of the portland ce-ment (the basis for Type II and V sulfate-resisting cements).The alumina is a necessary ingredient in the formation of et-



tringite. However, if clay minerals are present in a cement-stabilized soil, the necessary alumina can be obtained from theclay and the cement-stabilized soil will be vulnerable to sulfateattack regardless of whether sulfate-resistant cements are used.

The discusser’s reference to the breakup of porous aggregatemay refer to a process involving sulfate that is physical andnot chemical and that is similar to that of ASTM C 88, ‘‘TestMethod for Soundness of Aggregate by Use of Sodium Sulfateor Magnesium Sulfate.’’ An interesting description of the var-ious types of sulfate attack can be found in Hime and Mather(1999).

APPENDIX. REFERENCE

Hime, W. G., and Mather, B. (1999). ‘‘Sulfate attack? Or is it?’’ Cementand Concrete Res., 29, 789–791.


ng. 2000.126:940-942.

Behavior of Five Large Spread Footings in Sand

Documents