ELSEVIER

Dielectric Amplification in
Cement Pastes
S.J. Ford, J.-H. Hwang, I.D. Shane, R.A. Olson, G.M. Moss, H.M. Jennings, and
T.O. Mason

Dielectric amplification (dielectric constants >80) is observed in much larger than the dielectric constants of all other
cement pastes at early ages. Standard nonlinear least squares fitting phases, the dielectric behavior of the bulk paste should
routines yield artificially large “capacitances” when constant phase be dominated by the amount of pore water in the sys-
elements are employed. Instead, capacitance vs. frequency analysis tem [4]. It can be assumed that the dielectric response is
provides reliable evidence of dielectric amplification. A physical proportional to the volume fraction of porosity (4) in
model system consisting of a polycarbonate box with electrodes at saturated pastes. This is not borne out experimentally,
each end, divided into two compartments by a polycarbonate barrier however, as the reported values for the bulk dielectric
with a single hole, andfilled with electrolyte solution, simulates the constant appear to be much higher than that of the pore
impedance response in young cement pastes. The barrier represents solution. Such high values have given rise to several
hydration products whereas the hole represents the constriction be- theories to account for the enhanced dielectric behavior
tween two adjacent capillary pores, i.e., the pore network remains of these materials [4,5,10]. In this paper, we discuss this
percolated. Dielectric amplification is inversely proportional to bar- “dielectric amplification.” The origin of this amplifica-
rier thickness, i.e., it decreases as the barrier (product phase) thick- tion is explained as a combination of analysis errors and
ens. Impedance spectra from real pastes vs. water/cement (w/c) ratio microstructural effects. An explanation of the micro-
and during freezing or solvent exchange (to preferentially reduce the structural contribution is presented, based on simula-
conductivity of the capillay pores) exhibit significant dielectric am- tions of a constricted pore system and experimental
plification, even a&r fveezing or exchange, suggesting that C-S-H data from both frozen cement pastes and pastes whose
gel also has a dielectrically amplified microstructure. ADVANCED pore solution has been exchanged with isoproponal.
CEMENTBASEDMATERIALS1997,5,4148.0 1997 Elsevier Science
Ltd.
KEY WORDS: Dielectric amplification, Impedance spectros- Experimental
copy, Cement paste, Constricted microstructure, Percolation,
Pore system, Dielectric constant. Simulations
As with the electrical conductivity of these materials,
their dielectric constants may be related to the tortuos-
ity of the liquid phase. To test the effect of a constricted
microstructure on the dielectric response, a simplified
everal groups have investigated the electrical
experimental model was used. Containers, as shown in
response of cement-based systems as a nonin-
Figure 1, were constructed from polycarbonate with di-
vasive technique for monitoring the micro-
mensions 25.4 x 25.4 x 100 mm. C-1018 steel plates,
structural evolution of these materials [l-lo]. Recently,
polished to 600 grit, were used as electrodes for the
the relative dielectric constant of cementitious materials
impedance measurements described below, and poly-
(k,) has been the focus of much attention, and values as
carbonate barriers of varying thickness were placed
large as lo5 have been reported for cement pastes at
midway between them. Tap water was used as the sys-
early hydration times [1,4,11]. Cement paste can be de-
tem electrolyte, and the barriers, having the same cross-
scribed as a 3-3-O electrocomposite having intercon-
sectional area as the containers, constricted the current
nected pore fluid and C-S-H gel phases (3D connectiv-
flow through a hole 2.5 mm in diameter that was drilled
ity) and various other disconnected solid phases (OD
through their centers. Epoxy was used between the bar-
connectivity) [12,13]. It has been shown that the highly
riers and container walls to prevent leakage.
conductive, continuous pore phase controls the conduc-
tivity of cement-based materials, being proportional to
the volume fraction of pore fluid ($), its conductivity Cement Pastes
(o,), and connectivity (p) [1,7]. Similarly, because the Cement pastes of various water/cement ratios (w/c)
dielectric constant for the liquid phase is near 80, and is were prepared using ASTM Type I ordinary Portland

0 1997 by Elsevier Science Ltd. ISSN 1065.7355/97/$17.00

655 Avenue of the Americas, New York, NY 10010 PII SlO65-7355(96)00009-O
42 SJ. Ford et al. Advn Cem Bas Mat
1997:5:41-48

addresses numerical fitting artifacts and possible micro-

structural origins of dielectric amplification.

Numerical Fitting
The first contribution to the reported dielectric amplifica-
tion is the error introduced as a result of numerical fitting.
Analysis and deconvolution of impedance spectra are
usually accomplished with the aid of non-linear least
FIGURE 1. Schematic of the polycarbonate containers used to squares (NLLS) fitting programs such as “EQUIVCRT”
model constricted microstructures. 1171.These programs allow the measured spectra to be
fit to a user-defined equivalent circuit which simulates
the impedance response of the system, and is composed
cement (OK), mixed with distilled water for 15 min-
of ideal circuit elements such as resistors, capacitors,
utes in a Hobart planetary mixer. The pastes were then
and inductors. These circuit elements, if chosen cor-
cast into polycarbonate molds, similar to those de-
rectly, represent various electrochemical processes,
scribed above, and stored in 100% relative humidity
such as ionic conduction and dipole polarization occur-
(RH) chambers. Details of the freezing and isopropanol
ring in the system. Because most systems are not ideal,
exchange experiments, including specimen geometries
these NLLS programs often allow non-ideal spectra to
and experimental procedures, are provided elsewhere
be fit with constant phase elements (CPEs). Figure 2
111,141.
shows a numerically simulated, non-ideal response for
the inset circuit combination. This equivalent circuit is a
lnstrumen ta tion common representation for many electrochemical pro-
Impedance spectra were acquired using a Solartron cesses, and consists of a resistor in series with a resistor
1260 frequency response analyzer (FRA) with “Z60” and non-ideal capacitor, or CPE, combined in parallel.
data collection software 1151,over the frequency range 1 The complex plane, or Nyquist plot in Figure 2 shows
Hz to 10 MHz. An oscillating voltage of 25 mV was the characteristically depressed semi-circular arc for
used as the excitation signal. Nulling procedures, de- this circuit. The depression angle, 0, is a measure of the
scribed elsewhere 111, were used to remove inductive CPEs deviation from an ideal capacitor.
and capacitive effects from the leads and measurement Constant phase elements are empirical functions hav-
apparatus, and the corrected spectra were analyzed us- ing an impedance response of the form [17-191:
ing the “EQUIVCRT” software 1161.

Z(CPE) = B(jw)-” (2)

Results and Discussion
or, separated into its real and imaginary components,
Dielectric amplification in cement-based materials is de-
fined herein as the increase in the capacitance of the
system above the values that would be expected if the
volume fraction of pore solution, using the rule of mix-
tures, resulted in the dielectric response of the compos-
ite. The predicted capacitance values should be the
product of the pore fluid capacitance and its volume
fraction, expressed mathematically as:
%EJ- CPE

where C is the capacitance of the system, k,(ps) is the

dielectric constant of the pore solution, A and I are the
specimen area and length respectively, E, is the permit-
tivity of free space, and + is the volume fraction of fluid RI RI + R2
filled porosity in the system. As discussed previously,
capacitance values orders of magnitude higher than Z’
those predicted by Eq. 1 have been reported for cement FIGURE 2. Schematic Nyquist plot for a resistor in parallel
pastes at early hydration times Ill. The present work with a non-ideal capacitor.
Advn Cem Bas Mat Dielectric Amplification in Cement Pastes 43
1997;5:41-48

NLLS fitting procedures are unnecessary for obtaining

Z(CPE) = Bw-” [ cos( y) -jsin (F)] (3) capacitance information from experimental spectra.
Using the proper conversion factors [18], the mea-
where B is a constant, w is the angular frequency, and n sured impedance can directly give the capacitance of a
is a measure of arc depression having values between system, without the need for numerical fitting. The im-
-1 and 1, which is related to the depression angle by the pedance and capacitance values are related through the
equation: equation [18]:

(4) (6)

with 9 in radians. Eq 3 is completely general and can be where Re(C) is the real part of the capacitance, Im(Z) is
used to model all of the basic circuit elements, with the imaginary part of the impedance, o is the angular
variations only in their n-value, as shown in Table 1. frequency, and I Z I is the total impedance given by:
Most often, CPEs are used to represent non-ideal ca-
pacitors to model depressed arcs in the complex imped- I Z I2 = [Re(Z)]’ + [Im(Z)12 (7)
ance plane, as discussed above.
Cement-based materials usually produce semi- This approach allows capacitance data to be directly
circular impedance responses, similar to the depressed obtained. It also highlights where errors in interpreta-
arc illustrated in Figure 2, where the CPE typically has tion can be made.
n-values between 1 and 0.75, indicating predominately Figure 3 shows the capacitance vs. frequency re-
capacitive behavior. This depression of the impedance sponse of the inset equivalent circuit, simulated using
arc has been attributed to non-Debye behavior and a the “EQUIVCRT” software. This equivalent circuit is a
distribution of relaxation times in the system [18,19]. typical, although simplified, representation of cementi-
Regardless of the cause of the depression, this non- tious systems, where the individual elements were as-
ideality can lead to misinterpretations of the system sumed to have the values given in Table 2. The two
capacitance, as will be described below. curves differ only in the n-value assigned to the bulk
When constant phase elements are used to represent CPE, as indicated in the figure, and thus in their devia-
non-ideal capacitors, the inverse of the CPE constant, B tion from ideality. The n = 1.0 curve is the ideal re-
from Eqs 2 and 3, is often the reported parameter. This sponse, and the n = 0.75 curve is a non-ideal response.
“capacitive” constant Q, is defined as: Because real systems are seldom ideal, the latter curve
more closely resembles the response of typical cement-
Q = B-l (5) based systems [l]. At frequencies above approximately
lo4 Hz, these curves diverge owing to the difference in
The constant Q is of interest because, when the n-value the bulk n-values. In this region misinterpretations are
is equal to unity, Q is defined as the true, frequency often made concerning the bulk capacitance of non-
independent, capacitance of the system. Because the ideal materials systems. When n is equal to unity, as in
real term of Eq 3 goes to zero as n approaches one, the the upper curve, Q is by definition the true capacitance
impedance of the CPE becomes identical to that of a of the bulk, and in this simulation would have a value
pure capacitor (see Table 1). Problems arise when Q is of 10 nF (10e8F). When n diverges from one as in the
taken to be a true, single-valued capacitance for a non- lower curve, it can be seen that Q is not an accurate
ideal system, as the actual capacitance of the system measure of the bulk capacitance. In the non-ideal case,
may be orders of magnitude lower than Q. Fortunately, a distinct value for the bulk capacitance cannot be ob-
these misinterpretations are easily avoided because tained because of the frequency dependence of the di-

TABLE 1. Impedance relations for some basic circuit elements, derived from the impedance function of a constant phase element
given in Eq 2.
Circuit Element n-value Z(CPE) Constant
resistor 0 B B=R
capacitor 1 -jB/w B=C-l
inductor -1 jwB B=L
Warburg
diffusional 0.5 B(l - j)(2o)-l/2 B=Q-l
element
44 S.J. Ford et al Advn Cem Bas Mat
1997;5:41-48

1o-5

1o-6

8 lo-‘, lo-’
B
3 lo-* 7 lo-*
it
8 10-9, 1o-y

lo-lo lo-lo
Non-Amplified
lo-” 10-l’ Response
1o-5 10” 10-l 10’ lo3 lo5 10’ lo-l3 1 , ,..., ,.,,,,1 , m
frequency (Hz) 10” 10-l 10’ lo3 lo5 lo7 lo9
FIGURE 3. The capacitive response of the inset equivalent frequency (Hz)
circuit with bulk n-values of 1 and 0.75. The two parallel
resistor-CPE circuit combinations represent bulk and interfa- FIGURE 4. Extended spectrum for the dielectric behavior of a
cial impedance responses. neat OPC paste of w/c = 0.4 at a hrdratfon time of 20 hr (+ =
0.4 + 0.05). The behavior above 10 Hz is taken from Ref 22.

electric response. The bulk capacitance for this simula-

have resonant frequencies in the GHz range, these mea-
tion can be estimated as having a value in the 10-l°F
surements provide information on the dielectric behav-
range, as can be seen in the figure. The Q-value ob-
ior of the pore solution in cement-based materials. Fig-
tained by NLLS fitting of the lower curve (lo-‘) is not a
ure 5 shows the results of dielectric measurements
good approximation of the true bulk capacitance
made at 10 GHz for several porous systems as a func-
(=10-l’), being in error by more than an order of mag-
tion of the volume fraction of porosity they contain
nitude.
Numerical fitting errors cannot completely account for
the observed dielectric behavior at intermediate frequen-
cies (104-lo7 Hz) in cement-based systems. Figure 4 (2.2% Superplasticizer)
OPC w/c = 0.3 (
shows the dielectric response as a function of frequency (2.2% Superplasticizer)
White Cement w/c = 0.5
of a w/c = 0.4 OK paste at 20 hours of hydration.
Various mechanisms are indicated in the extended v OPC w/c = 0.5
spectrum with an electrode interfacial response at lower v OPC w/c = 1.0
frequencies (mHz-Hz), a water dipolar response at the (1.6% Superplasticizer)
highest frequencies (MHz-GHz), and a bulk response 30
in the intermediate frequency range (kHz-MHz) [20]. k-
The highest frequency capacitive response is beyond
the measurement range of most impedance equipment,
20
which are usually limited to below 30 MHz. Recent
reports in the literature have provided this high fre-
quency information, measured using microwave fre-
10
quency analyzers [21-231. Because aqueous electrolytes
OPC w/c = 0.5
OPC w/c = 0.3

TABLE 2. Circuit element values assumed in the capacitance 0 1 . . . , . _ , . .I A , . . , . . White Cement w/c = 0.5

0 0.2 0.4 0.6 0.8 1

vs. frequency simulations shown in Figure 3.
volume fraction porosity
Resistance Q-value n-value
FIGURE 5. Dielectric constant values collected at 10 GHz for
Interface 105 n 10-4 0.95
various porous systems vs. their volume fraction of porosity
Bulk 150 n WS 1.0 and 0.75
(after Brodwin).
A&n Cem Bas Mat Dielectric Amplification in Cement Pastes 45
1997;5:41-48

[24,25]. The GHz frequency capacitance values appear

to depend only on the amount of free (capillary) water
in the system, in validation of Eq 1.
Between the dipolar (GHz) and interfacial (sub-Hz)
plateaus in Figure 4, there is clear indication of an in-
termediate frequency (104-lo7 Hz) plateau. Capaci-
tances in this frequency range fall typically one to two
orders of magnitude higher than can be accounted for
on the basis of the dipolar response of capillary pore a) 0
0 5000 1 lo4 1.5 lo4 2 10’ 2.5 lo4 3 lo4 3.5 IO4
water, based upon Eq 1 and the plot in Figure 5. In all
subsequent plots, the capacitance-frequency behavior is Re (Z)
examined over the 104-lo7 Hz frequency range and
compared with the “non-amplified response,” i.e., the
capacitance based upon a reasonable range of free wa-
ter contents and the relationship in Figure 5.

-=-thin barrier
Dielecfric Amplification
In the computer modeling work of Coverdale et al. [4],
dielectric amplification could be obtained when com-
plete or incomplete C-S-H gel phase barriers blocked
the current flow between two adjacent capillary pores.
In the present work, a physical model which simulated
incomplete blockage between pores was constructed
and tested for dielectric amplification.
b)
d
frequency (Hz)
Simplified Model of a
Constricted Microstructure FIGURE 6. a) Nyquist and b) Capacitance vs. frequency plots
for the simple polycarbonate experiments used to model the
The simplified experimental model of Figure 1 was behavior of other constricted microstructures.
used to simulate the effects of hydration on the electri-
cal properties of cement paste microstructures. As hy-
dration proceeds, constriction of the pore microstruc-
pb A -a k:
ture occurs in cement pastes as C-S-H gel is formed, -B -+- (8)
which consumes free water and constricts passages in Pe a k:
the capillary pore network [26,27]. This constriction of
the capillary pores makes fluid and current flow more where pb and pe are the resistivities of the barrier and
difficult [7], and these impediments are simulated by electrolyte respectively, k: and k$.are the dielectric con-
the polycarbonate barriers in the experimental model. stants of the barrier and electrolyte materials, A is the
Figure 6 shows the Nyquist and capacitance behavior area of the barrier, and a is the area of the hole. There-
of three different polycarbonate configurations; one fore, as long as the ratio of the barrier and hole areas is
without a constricting barrier and two with barriers of within this range, the equivalent circuit in Figure 7a is
different thicknesses but identical hole diameters. reduced to the circuit in Figure 7b, giving rise to two
When a barrier is present, the Nyquist plot shows a arcs in the complex impedance plane. It should be
second arc which implies series behavior. The presence stated, however, that this circuit model is not strictly
of two arcs in the complex plane is observed only when correct. The high frequency arc, to the left of the
the constricting hole in the barrier is within a certain Nyquist plot in Figure 6a, is affected by current spread-
size range and can be understood with the aid of the ing at the point of constriction. In essence, the current
equivalent circuits in Figure 7. When the hole is small, flow near the hole is perturbed so that the sampled
the capacitance of the electrolyte within the hole is “pore” area is less than the geometric area (A), and
much less than that of the barrier itself, and the barrier resistance values for the electrolyte (hater) are larger
capacitance dominates the response. For the polycar- than expected. This effect becomes more pronounced as
bonate model, the size restrictions needed to produce the microstructure becomes more constricted. (Note:
two arcs in the complex plane can be described by the The RC elements of the polycarbonate side walls are
relation: ignored for the moment, but will become important in
46 S.J. Ford et al. Advn Cem Bas Mat
1997;5:41-48

crease in plate separation in a parallel plate capacitor.

The amplified capacitance is the capacitance of the bar-
rier above the expected capacitance of the system with-
out a barrier. This response is similar to that described
for cement paste during hydration, as in Figure 4. As an
analogy, C-S-H barriers can be thought to form in the
pore network at early times, causing an increase in the
bulk capacitance due to the constriction of the pore net-
work and the voltage drop across C-S-H barriers. As
hydration proceeds, these barriers thicken, causing the
capacitance to decrease. At long times, the hydration
reaction slows and the bulk capacitance reaches a pla-
teau.
a)

Experimental Results
To test the validity of the barrier/hole analogy, theo-
retical results from Eq 1 were compared with experi-
mental observations of real cement paste systems. Fig-
ure 8 shows how the capacitance of a neat OK paste of
w/c = 0.4 changes with hydration time. As with the
simple microstructural model presented above, the ca-
pacitance decreases with time, presumably due to the
C warn Cbmier thickening of the C-S-H barriers. An implication of this
microstructural model is that dielectric amplification
FIGURE 7. Equivalent circuit for the single barrier polycar-
bonate model. a) reduces to b) for small holes. Note that Kate, should be more pronounced when the system is more
represents a combination of the two side chambers of Figure open and fewer barriers are present. In other words, the
1 into a single element, likewise for C,,,,,. same voltage is being dropped across shorter distance,
i.e., higher capacitances can be expected. A higher wa-
ter content in a cement paste produces larger capillary
the discussion of cement pastes, especially during freez-
ing and solvent exchange.)
This simple model of cement-based systems does not 1
require the pore system to be depercolated, which was
a major concern in the early modeling work of Cover-
dale et al. [4]. Percolation in any model concerning the
microstructure of cement-based materials is important,
because the pore network for most cementitious sys-
tems is continuous at early hydration times when di-
electric amplification is at a maximum, and for many
systems the pore network is continuous even at long

I
hydration times [27]. It can be seen from Eq 8 that de- 100
percolation nullifies the circuit model of Figure 7b. As . non-amplified response 4 = k(~f.9
the area of the hole goes to zero, the barrier becomes a
series component in the circuit, greatly increasing the
~111111~11111111111
I I111~1111111111111
resistance of the system. It will therefore become im- -neat paste- w/c = 0.4 *grout - solnhlids = 1

1
10: (1 day old) (1 day old)
portant to include the C-S-H gel phase in the equivalent -neat paste w/c = 0.4 -C-neat paste - w/c = 0.7
circuit models of depercolated cementitious systems. (3 days old) (3 days old)
Figure 6b illustrates how the constriction of electrical t. . .‘.,.I_ . ‘...‘I_ . “‘.-
current flow in the experimental models has an ampli- 10’ IO’ lo6 10’
fying effect on the bulk capacitance of the system. When frequency (Hz)
the barrier is thin, the voltage drop across it produces a FIGURE 8. Capacitance vs. frequency plot for various cement-
large, amplified capacitance. As thicker barriers are based systems, comparing hydration times (at fixed w/c ra-
used, the capacitance decreases analogously to an in- tio) and w/c ratios (at fixed hydration time). (+ = 0.4 k 0.05)
Advn Cem Bas Mat Dielectric Amplification in Cement Pastes 47
1997:5:41-48

pores and creates a more open microstructure with

fewer barriers [26,27]. To further test the model, pastes w/c = 0.4
with various w/c ratios were compared. Figure 8 also normal OPC paste
1 day old
shows how the dielectric constant vs. frequency re-
sponses of a w/c = 0.7 OPC paste, and a grout with a
solution:solids ratio of approximately 1.0 [28] compare
with a paste of w/c = 0.4 at the same hydration times.
It can be seen that dielectric amplification is more pro-
nounced at high w/c ratios, as anticipated when larger kI
pores (fewer barriers) are present,
The C-S-H gel phase may itself be a dielectrically J

amplified microstructure, dominating the electrical re- ~1111111111111111111111111111111111111111

sponse of cementitious systems in which the larger cap- k,=k&).4
illary pore network is nearly consumed or is highly 10: non-amplified
tortuous. This is illustrated by the relatively high ca- : responses
k, = k,( isopropanol) @
pacitances evident for cement pastes at long times when
the free water is consumed and the volume fraction of ~IIIIIIIIIIIIHIIIIIIIIIIIIIIIIIIIIIIIIII
capillary porosity is near zero [l]. This additional level IIO'
I . ."'."I . '......I
IO' IO" IO'
of amplification can be seen more clearly in freezing
and solvent exchange experiments.
Figures 9 and 10 show the effects of freezing and FIGURE 10. Dielectric constant vs. frequency plots for a neat
solvent exchange, respectively, on the capacitive behav- OPC paste of w/c = 0.4 at 1 day of hydration, before exchange
ior of cement pastes. During freezing (Figure 9), a pre- and after nearly 10 hours of solvent exchange with isopropa-
cipitous drop occurs in the dielectric constant at ap- nol. (+ = 0.4 * 0.05)
proximately -5”C, followed by a more gradual decline
as the temperature is decreased 1141. The precipitous
The gel pores, being on a much smaller size scale, keep
drop can be explained by the freezing of the free water
the gel water from freezing until lower temperatures
in capillary pores, eliminating the conductive large
are reached. Figure 9 shows that the capacitance of the
pores/constrictions (Rwater, Rfwater + w in Figure 7b).
frozen system is still higher than the capacitance pre-
What remains is the dielectric response of the continu-
dicted by Eq 1. This implies that there is a residual
ous C-S-H product phase, i.e., the side walls in Figure 7.
dielectric amplification associated with C-S-H gel, pos-
sibly of similar origin to the amplification in the capil-
lary pore system, but on a smaller scale.
w/c = 0.4
A similar explanation can be made concerning the
normal OPC paste
1 day old isopropanol exchange of the capillary porosity in ce-
ment pastes. Because isopropanol has a large molecular
size relative to water, it is thought to exchange only the
capillary water in the system, because the gel pores are
smaller than the solvent molecules [ll]. As in freezing,
solvent exchange greatly reduces the conductivity of
kI this capillary pore network (Rwaten Rrwater + ~0in Fig-
ure 7b). Figure 10 shows that the system capacitance
after solvent exchange is still large relative to the non-
1
~II1111111111111111111111~11~11I11111111 amplified response, implying that C-S-H gel is a dielec-
k,=k,(ps).d trically amplified microstructure. This residual level of
i. _ non-amplified amplification can be virtually eliminated upon solvent
: responses exchange with methanol, a much smaller molecule than
k, = k&z).+
isopropanol which is thought to more fully exchange
~1111111111111111111111111111111I1111111 gel pores in addition to capillary pores 111,291.
1 I I I
IO4 IO5 lo6 IO’

frequency
Conclusions
FIGURE 9. Dielectric constant vs. frequency plots for a frozen
and unfrozen neat OK paste of w/c = 0.4 at 1 day of hydra- The impedance spectra of cement pastes at early ages of
tion (4 = 0.4 + 0.05) hydration show evidence of dielectric amplification, i.e.,
48 S.J. Ford et aI. Advn Cem Bas Mat
1997:5:41-48

dielectric constants larger than can be accounted for on 3. Coverdale, R.T.; Christensen, B.J.;Jennings, H.M.; Mason,
the basis of water dipolar response. This is best seen in T.O.; Bentz, D.P.; Garboczi, E.J. 1. Mat. Sci. 1995, 30, 712-
719.
plots of capacitance vs. frequency in the 104-lo7 Hz
4. Coverdale, R.T.; Christensen, B.J.; Mason, T.O.; Jennings,
range. Because of arc depression, non-linear least H.M.; Garboczi, E.J. 1. Mat. Sci. 1994, 29, 4984-4992.
squares fitting with equivalent circuits involving con- 5. Gu, I’.; Xie, I’.; Beaudoin, J.J.; Brousseau, R. Cem. Concr.
stant phase elements can lead to erroneously high “ca- Xes. 1992, 22, 833-840.
pacitance” values. This problem is avoided by using 6. Gu, P.; Xie, P.; Beaudoin, JJ.; Brousseau, R. Cem. Concr.
capacitance vs. frequency plots. Res. 1993, 23, 157-168.
A simple physical model of the “constricted” pore 7. Garboczi, E.J. Cem. Concr. Res. 1990, 20, 591-601.
structure of cement pastes consists of reservoirs (pores) 8. Xu, Z.; Gu, I’.; Xie, I’.; Beaudoin, J.J.Cem. Concr. Res.1993,
23, 1007-1015.
separated by a barrier (C-S-H gel) with a hole to retain
9. Xu, Z.; Gu, P.; Xie, P.; Beaudoin, J.J. Gem. Concr. Res. 1993,
percolation. The resulting capacitance-frequency be- 23, 853-862.
havior is similar to that of cement pastes, i.e. there is 10. Brantervik, K.; Niklasson, G.A. Gem. Concu.Res. 1991, 21,
substantial dielectric amplification in the 104-lo7 Hz 496-508.
range. The model shows that dielectric amplification is 11. Moss, G.M.; Christensen, B.J.;Mason, T.O.; Jennings, H.M.
inversely proportional to the barrier thickness, and that ACBM Vol. 4, 68-75 (1996).
the pore network need not be depercolated, as pre- 12. McLachan, D.S.; Blaszkiewicz, M.; Newnham, R.E. I. Am.
dicted by Coverdale et al. [4]. Ceram. Sot. 1990, 73, 2187-2203.
Dielectric amplification in cement pastes can be ex- 13. Ford, S.J.;Mason, T.O.; Christensen, B.J.; Coverdale, R.T.;
Jennings, H.M.; Garboczi, E.J. I. Mat. Sci. 1995, 30, 1217-
plained on the basis of the physical model. Depending 1224.
on w/c ratio and degree of hydration, capacitance val- 14. Olson, R.A.; Christensen, B.J.; Coverdale, R.T.; Ford, S.J.;
ues up to two orders of magnitude larger than what can Moss, G.M.; Jennings, H.M.; Mason, T.O.; Garboczi, E.J.1.
be accounted for by water dipole relaxation can be ob- Mat. Sci. 1995,30,5078-5086.
tained. Factors which increase pore size (e.g., w/c ratio) 15. Scribner Associates Z6O/ZVIEWfor Windows; Charlottes-
ville, VA 22901, 1994.
or decrease product layer thicknesses (short times of
16. Boukamp, B.A. Equivalent Circuit; University of Twente:
hydration) lead to higher dielectric constants. The di-
Netherlands, 1988.
electric amplification arising from the constricted pore 17. Boukamp, B.A. Solid State lonics 1986, 20, 3144.
network can be reduced by preferentially freezing or 18. Macdonald, J.R. Impedance Spectroscopy: Emphasizing Solid
solvent exchanging the capillary pores. In these cases, Materials and Systems; John Wiley & Sons: New York, 1987.
the residual dielectric amplification (upon freezing or 19. Cole, K.S.; Cole, R.H. I. Cherrz.Phys. 1941, 9, 341-351.
solvent exchange) argues for dielectric amplification in 20. Ford, S.J.; Mason, T.O. Techniques to Assess the Corrosion
the C-S-H gel phase, albeit on a finer scale. Activity of Steel Reinforced Concrete Structures 1995, ASTM
STP 1276.
21. Gu, P.; Beaudoin, J.J. J. Mat. Sci. Lett. 1995, 14, 613-614.
22. Gu, P.; Beaudoin, J.J. I. Mat. Sci. Lett. 1996, 25, 182-184.
Acknowledgments 23. Zhang, X.; Ding, X.Z.; Ong, C.K.; Tan, B.T.G.; Yang, J. I.
Mat. Sci. 1996, 31, 13451352.
This work was supported by the National Science Foundation 24. Chang, J.T. Ph. D. Thesis; Northwestern University, 1995.
through the Science and Technology Center for Advanced Cement- 25. Brodwin, M.E., 1996, manuscript in preparation.
Based Materials under grant. DMR-91-20002. The authors also wish
26. Taylor, H.F.W. Cement Chemistry; Academic Press: San
to gratefully acknowledge Donggy Sohn and Professor M.E. Brodwin
Diego, CA, 1990.
for beneficial advice and discussions.
27. Mindess, S.; Young, J.F. Concrete; Prentice-Hall, Inc.:
Englewood Cliffs, NJ, 1981.
28. Olson, R.A.; Tennis, P.D.; Bonen, D.; Christensen, B.J.;Jen-
References nings, H.M.; Mason, T.O.; Brough, A.R.; Young, J.F.1.Haz-
1. Christensen, B.J.; Coverdale, R.T.; Olson, R.A.; Ford, S.J.; ard. Mat. 1996, in press.
Garboczi, E.J.; Jennings, H.M.; Mason, T.O. 1. Am. Cerum. 29. M&hail, R.S.; Selim, S.A. Adsorption of Organic Vapors in
Sot. 1994, 77, 2789-2804. Relation to the Pore Structure of Hardened Portland Cement
2. Christensen, B.J.; Mason, T.O.; Jennings, H.M. j. Am. Ce- Pastes; Powers, T.C. ed.; National Research Council:
ram. Sot. 1992, 75, 93945. Washington, D.C., 1996; pp. 123-134.