Influence of Particle Properties and Initial Specimen State on

One-Dimensional Compression and Hydraulic Conductivity

Jason T. DeJong1 and G. Geoffrey Christoph2

Abstract: Particle crushing can adversely affect geotechnical system performance; examples include clogging in wells, pile shaft
capacity degradation, and postconstruction settlements. The generation of fines results in volumetric compression and a reduction hydrau-
lic conductivity, which is important for geotechnical systems whose performance is directly dependent on pore pressure dissipation,
groundwater flow, or hydraulic pumping. Knowledge of hydraulic conductivity change is poorly understood due to limited experimental
data, and an ability to predict this change is lacking. The role of single particle properties, initial specimen state conditions, and loading
conditions on the evolution of hydraulic conductivity with particle crushing was examined experimentally. Specimen response exhibited
an overshoot behavior and the convergence to a unique condition independent of initial relative density, gradation, and particle shape. The
hydraulic conductivity decreased by 2–3 times before specimen yield, and by 2–3 orders of magnitude after specimen yield. Empirical
correlations were developed to estimate the change in hydraulic conductivity given the initial permeability and select crushing parameter
values at the stress level of interest.
DOI: 10.1061/共ASCE兲1090-0241共2009兲135:3共449兲
CE Database subject headings: Crushing; Hydraulic conductivity; Compression; Sand; Particles.

Introduction groundwater flow, and/or hydraulic pumping. The objective

herein is to examine the role of single particle properties, initial
Particle crushing has particularly significant effects on clogging in specimen state conditions, and loading conditions on the evolu-
wells 共Valdes 2002兲, in pile capacity degradation 共Lu 1988兲, and tion of hydraulic conductivity with particle crushing. Insights
in postconstruction settlements 共Barksdale and Blight 1997兲. Con- gained are used to quantify changes in hydraulic conductivity for
ditions favorable to crushing depend on single particle properties a range of conditions and then to develop simple predictive meth-
共shape, size, hardness兲, initial specimen state conditions 共grain ods to estimate change.
size distribution, emin, emax, Cn, etc.兲, and loading conditions
共stress magnitude, compression versus shear兲. Crushing increases
as particle angularity, particle size, grain size distribution unifor-
Material Properties
mity 共poor gradation兲, void ratio, stress magnitude, and shearing
Uniformly graded Ottawa 20-30 and Q-Rok sands were used in
increase, and as particle hardness and particle coordination num-
the testing program 共Table 1兲. Single particle crushing tests were
ber decrease. performed using a setup similar to Nakata et al. 共2001兲. The
Particle crushing significantly alters engineering properties. single particle strength was estimated using 共Nakata et al. 2001,
Datta et al. 共1979兲 and others have observed the internal angle of McDowell and Bolton 1998兲
friction to decrease with particle crushing. The constrained modu-
lus has been shown by Chuhan et al. 共2002兲 to change throughout ␴sp = Fsp/D250 共1兲
loading. Chuhan et al. 共2002兲 and Valdes and Caban 共2006兲 ob- where ␴sp = single particle strength; Fsp = force at particle failure;
served particle crushing to result in a decrease in hydraulic con- and D50 = mean particle size. The average particle shape, charac-
ductivity by up to two orders of magnitude. terized using aspect ratio and roundness parameters 共Table 1兲, was
The reduction in hydraulic conductivity due to particle crush- determined for each soil before and after single particle crushing.
ing is of particular importance for geotechnical systems whose Untested Ottawa sand was more rounded than the angular Q-Rok
performance is directly dependent on pore pressure dissipation, blasting sand. However, the postcrushing particle shapes were
more similar between the two soils due to the rounded Ottawa
1
Associate Professor, Dept. of Civil and Environmental Engineering, particles splitting into subparticles with clean shear planes at the
Univ. of California, One Shields Ave., Davis, CA 95616 共corresponding breakage faces. More angular Q-Rok particles would break into
author兲. E-mail: jdejong@[Link] multiple pieces with similar shapes, resulting in similar character-
2
Former Graduate Researcher, Dept. of Civil and Environmental En- istics pre- and postcrushing.
gineering, Univ. of Massachusetts, Amherst, MA 01003.
Note. Discussion open until August 1, 2009. Separate discussions
must be submitted for individual papers. The manuscript for this technical
note was submitted for review and possible publication on December 5, Testing Methods and Program
2007; approved on May 15, 2008. This technical note is part of the
Journal of Geotechnical and Geoenvironmental Engineering, Vol. 135, A high-stress stainless steel oedometer equipped with rigid filters
No. 3, March 1, 2009. ©ASCE, ISSN 1090-0241/2009/3-449–454/ and plumbing required for hydraulic conductivity testing under
$25.00. stress was developed. The vertical stress was applied through a

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009 / 449

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]
Table 1. Summary of Material Properties
Particle
yield
Index properties stress Particle shape preloading Particle shape postloading

D50 ␴sp
Material G sa 共mm兲b C ub C cb eminc emaxd 共MPa兲e Aspect ratiof Roundnessg Aspect ratiof Roundnessg
Ottawa 20-30 2.64 0.74 1.06 1.16 0.51 0.74 125.5h 1.22 1.25 1.45 1.39
关45.2兴i 关11.3兴i 关21.7兴i 关25.1兴i 关15.8兴i
Q-Rok 2.65 0.75 1.47 1.40 0.80 1.14 47.1 1.34 1.46 1.41 1.45
关62.1兴i 关21.8兴i 关11.4兴i 关20.2兴i 关12.3兴i
a
ASTM D854.
b
ASTM D2487-93.
c
ASTM D4254-91, Method B.
d
ASTM D4253-93, Method 2A.
e
␴sp = Fsp / D50
2
.
f
Aspect ratio= major axis length/minor axis length.
g
Roundness= 共perimeter兲2 / 共4␲ area兲.
h
Average value.
i
关coefficient of variation兴.

hydraulic piston, and load and displacement readings were ob- distribution due to asperity breakage. The process continues as the
tained using a 445 kN load cell and a LVDT. A specimen diameter stress level is increased, resulting in the yield stress being less
of 63.5 mm and height of 20 mm was used. Specimens were pre- distinct than for the rounded sand. All sands approach a similar
pared at relative densities of 35, 50, 65, 80, and 95% with an void ratio of about 0.30 at 100 MPa.
accuracy of ⫾3% using an air pluviator. Under a seating load The influence of initial relative density on the yield stress was
共⬃0.5 MPa兲, the specimen was saturated by slowly flowing three distinct. An “overshoot” phenomenon occurs where specimens
pore volumes of deaired deionized water through the specimen. with a higher initial relative density 共lower e兲 will maintain elastic
Falling head hydraulic conductivity tests 共ASTM 2004兲 were con- behavior to higher stresses before the yield stress is reached. This
ducted to stress levels of 5, 10, 20, 40, 55, 70, 80 and/or is shown clearly for the Ottawa sand in Fig. 1. This overshoot is
100 MPa. A grooved platen located at the specimen ends allowed due to specimens with a higher initial relative density having a
water to disperse uniformly throughout the specimen. A series of higher coordination number, and therefore lower particle-particle
meshes 共#30, #100, #400兲 were used to minimize flushing of fine contact stresses than specimens at a lower relative density for a
particles out of the specimen. Parametric tests confirmed that flow given stress level. Therefore elastic behavior is sustained at the
was equal across the specimen and that the fines generated from particle level to higher global stress levels. The presence of the
crushing were not washed out of the specimen or accumulated “overshoot” phenomena is less defined in the behavior of angular
against the wire mesh 共Christoph 2005兲. Tests were performed at particles because particles undergo asperity breakage at very low
every combination of stress level and relative density. Test results stress levels. At stress levels above the yield stress “convergence”
terminated at different stress levels that directly overlaid each of the void ratio to a unique trend that is independent of initial
other, and therefore only trends from tests to 100 MPa are pre- relative density occurs.
sented herein.
Evolution of Crushing Parameters
Response to One-Dimensional Loading The effect of crushing can be captured through the evolution of
conventional soil index properties and specific crushing param-
The testing program was designed to capture the dependence of eters 共Fig. 2兲. Index properties used include the percent fines 共par-
specimen response and hydraulic conductivity evolution during ticles less than 0.075 mm be mass determined via mechanical
one-dimensional 共1D兲 loading on stress magnitude, initial speci- sieving兲, the average grain diameter 共D50兲, and the coefficient of
men state, particle properties, and crushing parameters. Compos- uniformity 共Cu = D60 / D10兲. Crushing parameters used include pa-
ite plots presenting the evolution of volumetric strain and void rameters based on the change in particle size corresponding to a
ratio with increasing stress is shown in Fig. 1. certain percent passing 关B = 共D15兲i / 共D15兲 f where 共D15兲i is 15%
finer by mass particle size initially and 共D15兲 f is 15% finer by
mass particle size after crushing 共Lee and Farhoomand 1967兲兴,
Specimen Response
the change in percent of particles passing a specified particle size
As the stress level increases, an initial elastic response is followed 关L90 defined as the increase in percent passing the sieve on which
by yielding and the accumulation of significant volumetric strain 90% of the original precrushed soil was retained 共Leslie 1975兲兴,
共or a decrease in void ratio兲. The rounded Ottawa particles exhibit and the change in shape of the grain size distribution 关Br = Bt / B p
an initially elastic response and the yield stress is very distinct where Bt is the total crushing defined as the area between the pre-
共about 40 MPa for 65% relative density兲. The more angular and postcrushing grain size distributions and bounded by the
Q-Rok sand undergoes irrecoverable 共plastic兲 strain at lower number 200 sieve size 共0.075 mm兲 and B p is the crushing poten-
stress levels 共⬍5 MPa兲, which is verified by changes in grain size tial defined as the area above the initial grain size distribution

450 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]
Fig. 1. Response to 1D compression to 100 MPa at select initial relative densities. Additional tests to intermediate stress levels followed identical
trends.

down to the number 200 sieve size 共0.075 mm兲 共Hardin 1985兲兴. Specimen Response
Additional indices were examined in Christoph 共2005兲.
The decrease in hydraulic conductivity with increasing stress
When comparing the same soil type at different relative den-
magnitude 共and decreasing void ratio兲 is presented in Fig. 3. For
sities, all soils begin at the same condition and converge to a
the rounded Ottawa sand a distinct change in hydraulic conduc-
value independent of initial relative density for all effective grain
tivity occurs at the onset of specimen yield 共40 MPa for the Ot-
diameters, including D50. When comparing different soils, the
tawa sand at 65% relative density兲. For the more angular Q-Rok
rounded Ottawa particles show trends of having a more stable
and Unimin sands hydraulic conductivity decreased even at low
structure than the angular Q-Rok sands. The effect of crushing on
stress levels 共10 MPa兲, which correlates to when particle crushing
the coefficient of uniformity, Cu, highlights the progression of all is initiated and fines are generated. For all soils and glass beads,
soils from being poorly graded to well graded. hydraulic conductivity decreased by about 2–3 times before
The evolution of crushing parameters also tracks changes to specimen yield, and by 2–3 orders of magnitude after specimen
soil properties, with varying degrees of success. The B parameter yield. As a result, the initial relative density has a relatively small
共Lee and Farhoomand 1967兲 detects the onset of crushing in the effect on the overall changes to hydraulic conductivity that occur
more rounded Ottawa sand 共20 MPa兲, crushing occurring at very when stress levels are sufficiently high to induce particle crush-
low stress levels 共⬍5 MPa兲 in the Q-Rok sand, and the conver- ing. For the more rounded Ottawa sand there is a convergence in
gence of all sands at high stress levels. The L90 共Leslie 1975兲 hydraulic conductivity at about 100 MPa while the more angular
parameter also adequately captures the onset of particle crushing soils converge after all specimens exceed their yield stress
for both Ottawa and Q-Rok sands. Hardin’s Br parameter 共Hardin 共⬎70 MPa兲.
1985兲 performs very similar to the L90 criteria.

Evolution of Hydraulic Conductivity with 1D Evolution of Index Properties and Crushing

Loading Parameters with Hydraulic Conductivity
The evolution of hydraulic conductivity is directly linked with
The evolution of hydraulic conductivity during 1D compression changes to index properties and particle crushing parameters. The
loading was experimentally captured and results then used to de- increase in percent fines due to crushing results in a decrease
velop simple methods to predict the change in hydraulic conduc- hydraulic conductivity due to a reduction in pore volume, and in
tivity based on the initial specimen hydraulic conductivity and particular constriction of the pore throats. The relationship ap-
specimen changes that occur during loading. pears to be linear on a semilog plot 共Fig. 4兲. Given the primary

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009 / 451

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]
 Fig. 2. Evolution of index parameters and crushing parameters versus vertical stress

Fig. 3. Evolution of hydraulic conductivity versus vertical stress and void ratio

452 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]
 Fig. 4. Evolution of index and crushing parameters versus hydraulic conductivity

dependence of hydraulic conductivity on the smaller particle frac- Prediction of Hydraulic Conductivity Change with
tion, more general parameters such as D50 and Cu do not provide Particle Crushing
unique trends. Other parameters such as D10 provide clearer
trends, but the drawback with the utilization of D10 is that with It is practically desirable to estimate changes in hydraulic conduc-
crushing the value can become less than 75 ␮m, requiring a hy- tivity due to particle crushing as it would enable estimation of
drometer test to be performed. hydraulic conductivity adjacent to driven piles, for example,
Crushing parameters tend to track the evolution of hydraulic where soil particle crushing occurred during pile installation.
conductivity better as the parameters are specifically constructed Laboratory measurement of the grain size distribution and the
to capture changes relative to the initial grain size distribution. As hydraulic conductivity at the initial preloading condition can be
evident in Fig. 4, the Lee and Farhoomand criteria correlates with readily obtained with conventional laboratory tests. With slightly
changes in hydraulic conductivity in a linear manner in semilog more effort the change in grain size distribution due to particle
space. Both the Leslie 90 and Hardin criteria 共Figs. 4兲 also track crushing that occurs at elevated stress levels can also be obtained.
hydraulic conductivity well, albeit in a nonlinear trend in semilog In light of this, empirical correlations that enable estimation of the
space. change in hydraulic conductivity considering the initial condition

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009 / 453

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]
and degree of particle crushing were developed. In all correla- 3. Hydraulic conductivity decreased by about 2–3 times before
tions, results are normalized to the initial specimen hydraulic con- specimen yield, and by 2–3 orders of magnitude after speci-
ductivity to indirectly account for initial grain size distribution men yield. The effect of initial relative density was minor
and relative density. and the index parameters and crushing criteria evaluated cor-
The change in hydraulic conductivity from an initial condition related well to conductivity changes.
to a secondary condition can be estimated using the percent fines 4. Empirical correlations were developed to estimate the change
generated as follows: in permeability given the initial permeability prior to loading
log共k/ki兲 = − 0.173 ⫻ 共F兲 共2兲 and select crushing parameter values at the stress level of
interest. All correlations perform well 共R2 ⬎ 0.92兲, with the
where k = permeability value of interest 共at an elevated stress correlations using percent fines and Hardin’s criteria being
level兲; ki = initial permeability prior to particle crushing; and F preferred.
= percent fines inserted as a percentage. This equation provides an The results of this study provide new insight into fundamental
R2 value of 0.95 for the data obtained in this study. It is noted that mechanisms governing sand crushing under high stresses and pro-
the equation is fitted so that no fines exist at the initial reference vide an approach to estimate the change in permeability in sands
condition. due to crushing.
The utilization of the Lee and Farhoomand 共1967兲 B parameter
yields
log共k/ki兲 = − 0.388 ⫻ 共B − 1兲 共3兲 References
where B = 1 at the initial condition 共no crushing has occurred兲
共R2 = 0.93兲. The L90 criteria 共Leslie 1975兲 provides a similar de- ASTM. 共2004兲. Annual book of standards, Vol. 04.08, West Consho-
gree of accuracy 共R2 = 0.92兲 with hocken, Pa.
Barksdale, R. D., and Blight, G. E. 共1997兲. “Chapter 8—Compressibility
log共k/ki兲 = − 10−5 ⫻ 共L90兲3 共4兲 and settlement of residual soils.” Mechanics of residual soils,
Finally, the Hardin 共1985兲 criteria provides the highest accuracy Balkema, Rotterdam, The Netherlands, 95–154.
Christoph, G. 共2005兲. “Influence of particle properties and initial speci-
共R2 = 0.96兲 using the following:
men state on the specimen response to one-dimensional loading at
log共k/ki兲 = − 12.9 ⫻ 共Br兲2 共5兲 high stresses.” MSCE thesis, Univ. of Massachusetts, Amherst, Mass.
Chuhan, F. A., Kjeldstad, A., Bjorlykke, K., and Hoeg, K. 共2002兲. “Po-
The above equations have been developed based on all experi- rosity loss in sand by grain crushing—Experimental evidence and
mental data obtained as part of this study 共Christoph 2005兲. These relevance to reservoir quality.” Mar. Pet. Geol., 19, 39–53.
sands are poorly graded and do not contain fines in their initial Datta, M., Gulhati, S. K., and Rao, G. V. 共1979兲. “Crushing of calcareous
preloaded condition. It is expected that the equations above, par- sands during shear.” Proc., 11th Annual Offshore Technology Conf.,
ticularly those based on the percent fines and Hardin’s criteria, are Vol. 3, Houston, 1459–1467.
amenable to a broader variety of initial soil conditions than those Hardin, B. O. 共1985兲. “Crushing of soil particles.” J. Geotech. Engrg.,
tested herein. 111共10兲, 1177–1192.
Lee, K. L., and Farhoomand, I. 共1967兲. “Compressibility and crushing of
granular soil.” Can. Geotech. J., 4共1兲, 68–86.
Conclusions Leslie, D. D. 共1975兲. “Shear strength of rockfill.” Physical Properties
Engineering Study No. 526, South Pacific Division, Corps of Engi-
Particle crushing from the single particle level, to specimen state, neers, Sausalito, Calif.
to the effect of crushing on hydraulic conductivity has been ex- Lu, B. T. D. 共1988兲. “Effects of grain crushing on axial behavior of driven
amined. The following observations were made: piles in calcareous sand.” Proc., Int. Conf. Engineering for Calcare-
1. An overshoot behavior occurs where specimens with a ous Sediments, R. J. Jewel, and M. S. Khorshid, eds., Balkema, Rot-
higher initial relative density 共higher coordination number兲 terdam, The Netherlands, 231–238.
were able to sustain stresses to a higher yield stress than McDowell, G. R., and Bolton, M. D. 共1998兲. “On the micro mechanics of
crushable aggregates.” Geotechnique, 48共5兲, 667–679.
could be sustained by less dense specimens. This overshoot
Nakata, Y., Kato, Y., Hyodo, M., Hyde, A., and Murata, H. 共2001兲. “Mi-
behavior was not observed in the angular soils where crush-
croscopic particle crushing of sand subjected to high-pressure one-
ing of asperities is initiated at low stress levels. dimensional compression.” Soils Found., 41共1兲, 69–72.
2. Convergence of the void ratio versus stress relationships for Valdes, J. R. 共2002兲. “Fines migration and formation damage—
specimens at all initial relative densities converge above the Microscale studies.” Doctoral dissertation, Georgia Institute of Tech-
maximum yield stress. In angular sands this occurred rela- nology, Atlanta.
tively rapidly while in rounded sands the overshoot behavior Valdes, J. R., and Caban, B. 共2006兲. “Monitoring the hydraulic conduc-
delays convergence to higher stresses. tivity of crushing sands.” Geotech. Test. J., 29共4兲, 322–329.

454 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2009

Downloaded 11 Jun 2012 to [Link]. Redistribution subject to ASCE license or copyright. Visit [Link]