United States
                    Environmental Protection
                    Agency
Risk Reduction
Engineering Laboratory
Cincinnati, OH 45268
                    Research and Development
EPA/600/SR -93/099  August 1993
&EPA      Project Summary

                    Parameters Affecting  the
                    Measurement  of  Hydraulic
                    Conductivity  for Solidified/
                    Stabilized  Wastes
                    D. J. Conrad, S. A. Shumborski, L. Z. Florence, and A. J. Liem
                      A series of experiments conducted
                    at the Alberta Environmental Centre ex-
                    amined the variation in hydraulic con-
                    ductivity (K) within and among three
                    matrices formed by steel mill baghouse
                    dust treated with 8%, 9% and 10% Nor-
                    mal Portland Cement at a water/cement
                    ratio of 1:1 Within the 8% and 9% ma-
                    trices, test gradient (i) and back pres-
                    sure (P) were combined into 3x3 fac-
                    torial treatments. Commercially avail-
                    able equipment was modified to allow
                    sensitive and continuous monitoring of
                    hydraulic conductivity. A permeant-ma-
                    trix interaction was indicated by K de-
                    creasing with  time at a rate that in-
                    creased with higher cement contents.
                    After hydraulic conductivity testing, the
                    samples were  examined by scanning
                    electron microscopy and energy dis-
                    persive x-ray analysis. A cement hy-
                    dration product, identified as ettringite,
                    had formed in the solidified/stabilized
                    waste pores. This product reduced hy-
                    draulic conductivity by two orders of
                    magnitude by restricting  conducting
                    pores. Four to seven weeks of testing
                    were required before  an  acceptable
                    equilibrium had been reached and sta-
                    tistical comparisons among the i x  P
                    treatments were made. Within each ma-
                    trix, gradient was statistically the most
                    significant parameter accounting for
                    60% of the variation in results. The
                    response to gradient was different than
                    that observed with clay and soil-liners
                    in the literature. The overall mean hy-
                    draulic conductivity of the 8% matrix
                    (10 ± 5 x 10'6 cm.sec1) was significantly
                    greater (p<0.01) than that of the 9%
matrix (0.06 ± 0.03 x 106 cm.sec1). Tem-
poral effects, gradient and cement con-
tent were identified as important fac-
tors  affecting  hydraulic  conductivity
measurements and must be considered
in regulatory tests. Bulk density was a
useful quality control criterion for mini-
mizing sample variance  within  each
matrix.
  This Project Summary was developed
by EPA's Risk Reduction Engineering
Laboratory, Cincinnati, OH, to announce
key findings of the research project
that is fully documented in a separate
report of the  same  title  (see Project
Report ordering information at back).

Introduction
  Solidification/stabilization processes are
widely used  to  treat wastes before their
disposal, especially those containing heavy
metals. These processes reduce the po-
tential for release of such  contaminants
by removing free water, i.e., forming a
monolith (thereby reducing the surface
area available for leaching)  and by reduc-
ing contaminant solubility by alkaline pre-
cipitation or incorporation into cement hy-
d rat ion products.
  One of the routes of contaminant re-
lease is through dissolution and flow
through the  bulk of the treated waste.
Hydraulic  conductivity,  defined below by
Darcy's law,  is thus important:
           K =
               Q
               iA
 where K is hydraulic conductivity (cm. sec'1),
 Q is flow rate (cm3, sec1) through a cross-
 sectional area of A (cm2), and i is gradient
                                                                     Printed on Recycled Paper

-------
(dimensionless), defined as head  loss or
pressure drop (cm H2O) per unit length of
the solid (cm).
   Hydraulic conductivity and permeability,
which are often incorrectly interchanged,
are related by:
                 KU,
             K = -
where K is permeability (cm2), u, and p are
absolute viscosity (g.cnr'.sec'1) and density
(g.crrr3) of the fluid,  respectively,  and g is
gravitational acceleration (980 cm.sec'2).
  When there is no fluid-solid interaction,
permeability is an intrinsic and useful prop-
erty of a solid: the flow rates of  different
fluids through it can be computed from its
permeability and the properties of the liq-
uids.  As shown in the  report as  summa-
rized  here,  however, this is  not necessar-
ily the case with solidified/stabilized waste.
Moreover, since it is not the solid  property
per se, but  the flow  rate of  aqueous
permeant through it which  is of  interest,
hydraulic conductivity will be used herein.
  The available literature deals  predomi-
nantly with clay and soil liners. Test and
instrument  parameters,  such  as satura-
tion, gradient and back pressure, as well
as temporal effects  and sample prepara-
tion, have been identified as being impor-
tant factors in hydraulic conductivity mea-
surements. The corresponding information
on solidified/stabilized  waste is, however,
practically non-existent. Nor is there suffi-
cient  information that  addresses the  dif-
ferences  between clay or soil liners and
solidified/stabilized  waste, such  as those
in compressive strength  and permeant-
matrix interactions.
  This report  deals with the effects  of
parameters affecting the measurement of
hydraulic conductivity of solidified/stabilized
waste. The study was  undertaken to form
bases for the development of a regulatory
test method - to improve intra- and inter-
laboratory precision -  and  for correlating
accelerated laboratory test results to those
occurring under field conditions.
  The scope of the investigation was lim-
ited to:
  •  one waste, steel mill baghouse dust,
    treated with Normal Portland Cement
    at different formulations to produce a
    range  of hydraulic conductivities
    typical  of those achieved by  other
    commercial solidification/stabilization
    processes, and
  •  the  following  test  and instrument
    parameters:  sample  preparation,
    temporal effects, gradient and back
    pressure.
  New equipment was acquired and modi-
fied to allow for sensitive,  accurate, and
continuous flow measurements. Particular
 attention was given to minimize variance
 due to sample preparation, statistical meth-
 ods,  and  experimental designs.  Electron
 microscopy, chemical analyses, and mea-
 surements of physical properties were con-
.ducted to  assist in the interpretation of the
 observed  phenomena.

 Materials, Methods and
 Procedures

 Sample Preparation
   Equal portions of  steel mill baghouse
 dust and silica sand (42, 41, and 40 wt.%)
 were treated with equal portions of Nor-
 mal Portland Cement and water (8, 9 and
 10 wt.%).  The products are referred to as
 8%, 9% and  10% matrices,  respectively.
   The waste and the treatment were cho-
 sen to be  those commonly encountered in
 practice. The formulations, including  the
 use of sand, were selected  to  produce
 samples with a range of hydraulic conduc-
 tivities typical of solidified/stabilized wastes
 (10~6 to 10~8 cm.sec'1). To obtain samples,
 the dry ingredients were mixed, water was
 added and mixed, and the product was
 compacted into  cylindrical  plastic molds
 (7.6 cm diameter by  15.2 cm long) with
 the  use  of  a standard  method.  The
 samples were cured  for at least 28 days
 at 23°C and  a minimum relative humidity
 of 95%.
   After curing, the samples were removed
 from  the molds and then trimmed to  en-
 sure parallel  end  faces. The bulk density
 of each sample was computed from mea-
 surements of its  mass and dimensions.
 For each matrix, the mean bulk density of
 all the samples prepared was computed.
 Only  those  samples  with bulk densities
 within 0.5% of the mean were used.

 Sample Characterization
   In  addition to bulk density and water
 content, the  specific gravity of the dried
 samples (true density) was measured, from
which porosity and  degree of saturation
 could be computed. To determine the ef-
fect of hydraulic conductivity testing, these
destructive measurements were made on
 samples after testing as well as on com-
 panion  samples  that had not undergone
testing. The  unconfined  compressive
 strengths  of  these  companion samples
were  also  measured.

 Equipment
  Three flexible-wall  permeameters were
 used.  The equipment featured a  novel
 method for measuring permeant flow rate,
 using the  movement  of a piston inside a
permeant  interface. This piston was con-
 nected to  a linear variable  displacement
transducer, the output of  which was digi-
tized  and logged in a data logger. The
inlet and outlet flow rates were simulta-
neously measured.
  Two major  modifications  were made.
The first was the replacement of the trans-
ducers to achieve 0.001-mm accuracy over
a 30-mm range.  The second was the use
of a bladder interface to isolate the pneu-
matic  pressurization system from the
permeameter cell water.

Hydraulic Conductivity
Measurements
  After vacuum  saturation  was  applied,
the inlet and outlet flow rates were com-
puted using volume measurements,  from
the calibrated  piston movement, and the
corresponding elapsed  times. Hydraulic
conductivity was computed when the inlet
and the outlet flows  were within 5%.

Other Measurements
  Electron microscopic analyses were per-
formed on a Hitachi  S510* scanning elec-
tron microscope (SEM), a Hitachi X-650
(SEM) with energy dispersive X-ray spec-
trometer, and  a Hitachi H-600  scanning
transmission electron microscope (STEM)
equipped with  a  Kevex Be window X-ray
detector. Chemical analyses  of the waste
and the exuded permeant were performed
with the use of ion coupled spectroscopy,
atomic absorption spectroscopy,  colorim-
etry and potentiometric titration.

Experimental Design
  The effects of  sample formulation,  time
and instrument  parameters  (gradient,   i,
and back pressure,  P)  were studied as
follows. A 3 x 3 full factorial design  was
used to determine the effects of i and P.
The values for i were  8, 116, and  227,
and for P were 14, 69, and 124 kPa.
  The hydraulic  conductivity of each of
the 8%, 9%, and 10% matrices was  con-
currently  measured using the three
permeameters. To study the initial tempo-
ral effects, median levels of i and P were
used.  After  suitable  equilibrium  was
reached,  i and P were  then varied  in  a
random manner according to that design.
Equilibrium was defined as when the varia-
tion of hydraulic conductivity with time, as
computed from the  slope of a linear re-
gression, could not be proven to be differ-
ent from zero. Because of time constraints,
the full factorial design was  applied  only
to  the 8% and 9% matrices. In addition,
three 8% matrix samples were tested  con-
currently at median levels of  i and P  over
* Mention of trade names or commercial products does
 not constitute endorsement or recommendation for

-------
a 29-day period to  obtain precision val-
ues.

Results and Discussion

Bulk Density
  Although a standard compaction proce-
dure was carefully  applied,  there were
overlaps in  bulk density  among the se-
lected matrices, as shown in Table 1.
  The criterion of  selecting  only  those
samples within ±0.5% of the sample mean
of each  matrix  was thus applied to ensure
distinct populations.

Temporal Effect
  The hydraulic conductivity of each ma-
trix  decreased  with time in a manner that
could be described by the following equa-
tion:
         Kx  ^06=A(T+^)B

where T is elapsed time, in days, from the
first day of testing  at T=0.  The  least-
squares values of A and B (referred to as
the power function intercept [initial  value]
and slope,  respectively, and computed
over 80 days  of testing at median  levels
of i and P)  are given in Table  2.
  Two  interesting observations can  be
made: there was a marked difference  in
initial value  between the 8% and 9% ma-
trices but not  between the 9%  and 10%
matrices; and the negative slope  increased
in magnitude with increasing cement con-
tent. The first  observation can be readily
                 explained in terms  of  granular  versus
                 paste-like behavior  resulting from differ-
                 ences in  cement and water contents. The
                 second, which suggests some form of ce-
                 ment hydration reaction, will be discussed
                 in the following section.
                   The decrease in hydraulic conductivity
                 with time, up to two orders of magnitude
                 for the 10% matrix, was the opposite of
                 what was anticipated at the onset of the
                 project. Due to  matrix dissolution, it was
                 expected that the connecting pores in the
                 matrix, and  hence  hydraulic conductivity,
                 would increase. Matrix dissolution did oc-
                 cur quite  significantly. In some cases, more
                 than 1%  of  the  sample weight was  lost
                 over 80 days of testing. Yet, the hydraulic
                 conductivity  decreased with time. The re-
                 sults of the investigation into this phenom-
                 enon  are described below.

                 Mechanisms of the Decrease In
                 Hydraulic Conductivity
                   Visual  inspections of the samples after
                 testing showed the presence of white ma-
                 terials in the dark-colored matrix. SEM
                 examinations  revealed  profuse  fibrous
                 growth, the morphology of which was simi-
                 lar to that  of  ettringite (3CaO.AI2O3.
                 CaSO4.31H  O), rather than to  that  of  a
                 cement  hydration product, such  as  cal-
                 cium  silica-hydrate or calcium  hydroxide.
                 X-ray analyses  of the individual fibres re-
                 vealed the  presence of Ca, Al, S  and
                 traces of  Fe, and a Ca/S ratio of 2.62±0.52.
 Table 1.  Bulk Density of Stabilized Waste Samples

                                    Matrices, cement percent weights
                            8%
                                                 9%
                                                                   10%
 Mean Bulk
 Density, (SD) *
 (gms per cm3)

 Relative Standard
 Deviation (n)+
2.398 (±0.024)
 1.00 (16)
2.480 (±0.043)
1.73(18)
2.527 (±0.363)
0.36 (16)
 * SD = Standard Deviations.
 * = Number of Samples.
 Table 2.  Temporal Effect: Regression Analysis Results *
                                               Matrix
                             8%
                                                  9%
                                                                      10%
 Initial Value, ±se *

 Slope, ±se
82.4 ± 4.4

-0.413 ±025
1.571 ± 044

-0.743 ±0.032
 1.356 ±0.148

 -1.476 ±0.070
 *  The models account for more than 98% of observed variation in hydraulic conductivity.
 * Standard error; significance levels p<0.001.
These  results suggest the  presence of
mainly aluminoferrite trisulphate (Ca/S=2.1 )
and  some aluminoferrite monosulphate
(Ca/S=4.1).
  The formation of fibrous materials was
a result of permeant-matrix reactions. No
such formation was observed in compan-
ion samples that were kept in a humidity
chamber.
  The above phenomenon has been pre-
viously reported. The hydraulic conductiv-
ity of cement pastes could be reduced by
six orders  of magnitude as a  result of
curing under water. The explanation was
given in terms of expansion in volume of
hydrated paste and hydration products fill-
ing the pores and cavities, thereby reduc-
ing and blocking flow channels.

Effect of Matrix
  As previously mentioned, there was a
marked difference  in hydraulic conductiv-
ity between the 8% and 9% matrices. Even
when the effects of time, i and P are not
considered, a statistically  significant differ-
ence (p<0.01) in hydraulic conductivity ex-
ists between them: 10±5 x  10~6 cm.sec'1
versus 0.06±0.03 x 10'6 cm. sec'1. Hydrau-
lic conductivity is  thus very sensitive to
matrix composition.

Effect of Instrument Parameters
   The effects of i and P were investigated
after equilibrium conditions were reached.
By the criterion used, these were reached
after 27, 34, and 59 days of testing for the
8%,  9% and 10% matrices, respectively.
Note however that the criterion is  based
on failure to reject the null hypothesis  of
zero slope. The power of the test and the
Type II error were  not considered. Tem-
poral effect could still be present, and this
would be considered as random errors.
   The variation of  hydraulic conductivity
with gradient and back pressure was mod-
elled according to a second  order polyno-
mial of the form:
                                                                          , x2+bt1 x,2+b22 x22+ bl2 x,x2
 where x, and Xj, are the transformed (range
 from  -1 to +1) i and P, respectively. The
 regression results are shown in Table 3.
   The results show statistically significant
 effects of gradient for the 8% matrix, with
 a positive linear term and a negative qua-
 dratic term; and gradient and back pres-
 sure  for the 9% matrix, with only positive
 linear terms. The equations show that K is
 less sensitive to i and P at high and me-
 dian  levels, and thus is the  region where
 hydraulic conductivity should be measured
 to minimize variability.
   The gradient was varied by a factor of
 25, and the back pressure by  8. Yet the

-------
  Table 3.  Regression Analysis Results
                                     Coefficient
 Matrix
8%

9%

13.01*
(0.8) S
0.06"
(.005)
2.33 +
(0.7)
0.023 *
(.005)
0.18
(0.7)
0.015"
(.0005)
-4.3 +
(1.1)
-0.013
(.008)
"& bi2 n *
-1.2
(1.1)
0.003
(.008)
-0.3
(0.9)
-0.001
(.006)
22
26
    The models account for 68% and 58% of the total variance for the 8% and 9% matrices
    respectively.
 n+ Number of data points.
 s  Standard error of the coefficient.
 "  Significant effect at p<0.01.
 variation in hydraulic conductivity was less
 than 4-fold for the 9%  matrix,  and less
 than 3-fold for the 8% matrix. Therefore,
 as a first approximation and only with re-
 spect to the effects of i and P, accelerated
 laboratory conditions produced  results
 close to what might be found in the field.

 Distribution and Precision
   Log-transformation  has been suggested
 as a way to normalize the data and obtain
 constant variance. The triplicate measure-
 ment results  for the 8%  matrix were log-
 transformed.  The  null hypothesis of nor-
 mal distribution could not be rejected for
 the transformed data. The precision val-
 ues are shown in Table 4.
   For comparison, a  precision of x/+ 7.3
 for four replicates was reported in a previ-
 ous study on solidified/stabilized  waste.
 The improvement could  be attributed to
 sample preparation, sample  acceptance
 criterion and measurements at higher lev-
 els of i and P.

 Correlation with Porosity
   For the 8%, 9%, and 10% matrices, the
 hydraulic conductivity and porosity E could
 be correlated by:

        K=2.83x  10-29 10725E
 with p<0.0001  and ^=0.89. It should be
 emphasized, however, that only one waste/
 treatment combination was used.

 Saturation
   Saturation is defined as the ratio of free
 water to pore volume. Changes in satura-
 tion as a result of testing are shown in
 Table 5.
   There was a marked increase in satura-
 tion, which  in  part was effected  in  the
 beginning of the  testing when vacuum
 saturation was applied.  In the field, how-
 ever, the treated waste would not be satu-
 rated.  The  effect  of  not saturating  the
 sample, coupled with  applying low levels
 of gradient and back pressure, still needs
 to be investigated.
Table 5.  Permeant Saturation Before and
         After Testing
                     Matrix
Testing
8%
         9%
                    10%
Before

After
35%

96%
42%

93%
52%

94%
Table 4.  Results of Log-Transformation for the 8% Matrix

                                            Test Day

Log Transformation          1          8
                                              15
                                                        22
                                                                   29
Mean
(SD)*
Precision *: xA
-3.94
(.18)
3.80
-4.31
(.25)
4.4
-4.48
(.23)
4.25
                                                        -4.61
                                                        (.26)
                                                        4.48
                         -4.78
                         (.24)
                         4.36
* Standard Deviation.
+ 95% confidence limits for mean K for three samples.
 Effect of Gradient: Comparison
 with Soil/Clay Liners
   High  gradients applied  to  soil/clay
 samples have been reported to cause con-
 solidation  and the consequent  reduction
 in hydraulic conductivity. This seems rea-
 sonable considering that clay, with a typi-
 cal  unconfined compressive  strength
 (UCS) of 100 kPa, would be subjected to
 a pressure of up to 450  kPa (for 15 cm
 sample length at a gradient of 300).  For
 comparison, the samples used in this study
 had a range of UCS from 3000 to 6000
 kPa, and were subjected only to a maxi-
 mum pressure of 340  kPa.  This may ex-
 plain the difference in  the effect of gradi-
 ent  between soil/clay and solidified/stabi-
 lized waste.

 Conclusions and
 Recommendations

 Conclusions
 1. Hydraulic conductivity was sensitive to
   matrix composition.  Significantly differ-
   ent hydraulic conductivities were mea-
   sured between  samples  differing only
   by 1% in cement content. The ability to
   distinguish  such samples was attrib-
   uted to the institution of a strict quality
   control criterion  on  the basis of bulk
   density. The corollary is that  variance
   resulting from sample preparation can
   be minimized by  using that criterion.
2. Hydraulic conductivity decreased with
   elapsed time during testing. A  power
   function in the form  of y=axb describes
   the relationship.  The  decrease could
   be explained  by  long-term cement hy-
   dration reactions forming  ettringite  in
   the permeant-conducting pores. Al-
   though matrix dissolution  occurred,  no
   effect  was observed over the testing
   period of up to 80 days.
3.  The effects of gradient and back pres-
   sure on hydraulic conductivity were the
  following:
   •  Gradient  was  the most significant
     parameter,  and its correlation with
     hydraulic  conductivity was  positive,
     the opposite to that for soil and clay
     liners. This was  attributed  to  the
     higher unconfined  compressive
     strength and the corresponding lesser
     degree of sample consolidation.
  •  Medium and high  levels of gradient
     and  back  pressure were the  less
     sensitive   region  for  hydraulic
     conductivity measurements.  Falling-
     head  permeameters, in which low
     levels of these parameters are used,
     are  thus operated  in  the  more
     sensitive region.

-------
  •  Over the entire region of the chosen
    experimental levels,  the measured
    hydraulic conductivities varied by  a
    factor of four or less. Values obtained
    in the laboratory are thus reasonable
    estimates  of those in the field
    conditions, provided that compaction
    and curing conditions are similar.
4. An exponential relationship between hy-
  draulic conductivity and sample poros-
  ity was shown to be statistically signifi-
  cant.  How  such a relationship varies
  with different matrices was not investi-
  gated.
Recommendations for
Regulatory Test Development
1. Bulk density should be used as a qual-
  ity control criterion to reduce variance
  resulting from sample preparation.
2. To improve  precision, temporal effects
  should be taken into account and mea-
  surements carried out at high levels of
  gradient and back pressure.
3. To estimate maximum  hydraulic con-
  ductivity,  measurements should be
  made as soon as the sample is cured.
Recommendations for Future
Work
1. Other  common  waste/treatment sys-
   tems should  be  studied to  investigate
   permeant-matrix interactions and the ef-
   fect of instrument parameters.
2. Saturation effects should be studied to
   predict field hydraulic conductivity.

  The full report was submitted in fulfill-
ment of CR#814860-01-1 by Alberta Envi-
ronmental Centre under the sponsorship
of the U.S.  Environmental  Protection
Agency.
                                                                    6ll.S. GOVERNMENT PUNTING OFFICE: 1993 • 75WJ7I/W026

-------
D. J. Conrad, S. A. Shumborski, L. Z. Florence, and A. J. Liem are with Alberta
  Environmental Centre, Vegreville, Alberta, Canada JOB 4LO.
C. I.  Mashni is the EPA Project Officer (see below).
The complete report, entitled "Parameters Affecting the Measurement of
    Hydraulic Conductivity for Solidified/Stabilized Wastes," (Order No. PB93-
    199 396/AS; Cost: $19.50, subject to change) will be available only from:
      National Technical Information Service
      5285 Port Royal Road
      Springfield,  VA 22161
      Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
      Risk Reduction Engineering Laboratory
      U.S. Environmental Protection Agency
      Cincinnati, Ohio 45268
  United States
  Environmental Protection Agency
  Center for Environmental Research Information
  Cincinnati, OH 45268

  Official Business
  Penalty for Private Use
  $300

  EPA/600/SR-93/099
     BULK RATE
POSTAGES FEES PAID
         EPA
   PERMIT No. G-35

-------