Sunteți pe pagina 1din 11

Cement & Concrete Composites 28 (2006) 349359

www.elsevier.com/locate/cemconcomp

A methodology for determining complex permittivity of construction


materials based on transmission-only coherent,
wide-bandwidth free-space measurements
Oral Buyukozturk *, Tzu-Yang Yu, Jose Alberto Ortega
Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139-4307, USA

Abstract

An integrated methodology for determining the unique combination of complex permittivity based on measured transmission coef-
cient and time dierence of arrival (TDOA) information in free-space measurements is proposed. The methodology consists of an
estimation procedure of the real part of complex permittivity based on TDOA, and a root-searching procedure based on parametric
system identication (SI) together with an error sum of squares (SSE) criterion. Generally, non-unique combinations of dielectric
constant and loss factor are encountered when lossy or low-loss materials are measured and the proposed methodology is aimed at
the determination of unique combinations of dielectric constant and loss factor for such materials. The proposed methodology is
validated by measurements of several materials with known dielectric properties. The estimated complex permittivity values for Teon,
Lexan, Bakelite, and concrete are in good agreement with those reported in the literature. The method has potential for in-situ measure-
ment of dielectric properties for construction materials. Applicability and limitations of the methodology are discussed.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Complex permittivity; Transmission coecient; Radar; Time dierence of arrival; Parametric system identication

1. Introduction structures using radar techniques [20,21]. However, the


utilization cannot be accomplished without the aid of a
Dielectric properties (complex permittivity) of materials reliable methodology for the measurement and determina-
have received increasing attention along with the use of tion of dielectric properties.
electromagnetic (EM) waves (radar/microwave) in the Up to the present, several experimental methods have
investigations of material and structural assessment. been applied to the measurement of dielectric properties
Dielectric properties of a material correlate to other mate- of a material. These methods include parallel plate capa-
rial characteristics and may be used to determine properties citor technique [1], resonator/oscillator technique [2],
such as moisture content, bulk density, bio-content, chem- transmission line technique [36], and free-space technique
ical concentration, and stressstrain relationship. Such [711]. Although each technique has its own features and
knowledge can be utilized for research and application in constraints, among these, the free-space technique appears
food science, medicine, biology, agriculture, chemistry, to be more applicable for in-situ measurements. In free-
electrical devices, defense industry (security), and engineer- space measurements, transmission and reection coe-
ing. In civil engineering, for example, dielectric properties cients (which can be real or complex) of the material under
of construction materials such as concrete are the key test can be measured, depending on the measurement
information to the non-destructive testing (NDT) of civil scheme. However, the use of reection coecients may suf-
fer from surface condition of the material in high frequency
*
Corresponding author. Tel.: +1 617 253 7186; fax: +1 617 253 6044. ranges. Implementation of complex transmission coe-
E-mail address: obuyuk@mit.edu (O. Buyukozturk). cients in determining dielectric properties appears feasible.

0958-9465/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2006.02.004
350 O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359

However, a unique determination of the complex per- tric properties of a material is implemented by the use of a
mittivity may not be directly accomplished with the avail- scalar, eective complex permittivity, ee , to account for EM
ability of complex transmission coecient (real and features (polarization and capacitance, dielectric losses,
imaginary parts). This is because one cannot explicitly and conductivity) of the material:
derive relationships between the real and imaginary parts    
 0 00  r 0 r00 00 r0
of the complex permittivity with the respective real and ee ee  jee e e j e 1
jx x x
imaginary parts of complex transmission coecient. Typi-
cal approach to solve such problems is the use of root- where e0e is the real part of ee and represents the ability of a
searching or optimization techniques, which search for material to store the incident EM energy through wave
the most reasonable (optimal) combination within given propagation, e00e is the imaginary part of ee and represents
range of variables. An error evaluation criterion is usually the degree of EM energy losses in the material, j is the
needed in the object function, which serves as the basis for imaginary number; e* = e 0  je00 is the complex permittivity
determining the optimal solution. For lossless materials, a (F/m), r* = r 0 + jr00 is the complex electric conductivity
unique optimal combination can be expected after the (X/m), and x = 2pf is the angular frequency (rad/s). The
application of root-searching techniques at single fre- dimensionless relative permittivity er is more frequently
quency. However, for lossy or low-loss materials such as used, which is dened as
concrete, multiple combinations of real and imaginary ee e0e  je00e
parts of complex permittivity for the same transmission er e0r  je00r 2
e0 e0
coecient at single frequency are observed [8,1214].
The objective of this paper is to propose an integrated where e0 is the permittivity of free space and
methodology for determining unique combinations of com- e0 = 8.85 1012 F/m. The real part of the relative permit-
plex permittivity using transmission coecient and the time tivity is known as dielectric constant and the imaginary
dierence of arrival (TDOA) information (both obtained part as loss factor. The ratio between the loss factor and
from free-space measurement) for low-loss materials. the dielectric constant is called loss tangent. For dielectric
Experimental measurements are conducted in coherent materials, e00r P 0 and e0r  e00r
condition which suggests the proportionality between e00r
the signal-to-noise ratio (SNR) and the length of signals. tan d 3
e0r
The coherent condition provides a consistent phase from
measurement to measurement. In the experimentation, fre- Dielectric constant and loss tangent are functions of mea-
quency domain measurements are made through a network surement frequency, material homogeneity and anisotropy,
analyzer using frequency bands ranging from 8 GHz to moisture, and temperature in the material. Various mea-
18 GHz. In this paper, rst, denition of dielectric proper- surement techniques are available for the experimental
ties is given followed by a brief review of current measure- determination of dielectric properties. A brief review of
ment techniques for determining these properties. A current techniques is provided in the next section.
detailed description of the proposed methodology is then
provided. Validation of the proposed methodology is 3. Review of current measurement techniques for dielectric
performed by comparing the determined dielectric proper- properties
ties of Teon, Lexan, Bakelite with reported values.
Further application of the method is given by measure- Current measurement techniques for dielectric proper-
ments on Portland cement concrete (PCC) slabs and a glass ties of materials are based on the concept of impedance
ber-reinforced polymer (GFRP) sheet. Experimental in which complex permittivity and complex permeability
set-up as a basis for the measurements is described. Imple- are involved. The material is modeled as a dielectric and
mentation issues and limitations of the methodology are the measurements can be made using dierent techniques
discussed. in order to calculate the dielectric properties of the mate-
rial. These techniques are briey described as follows.
2. Denition of dielectric properties
3.1. Capacitor modelparallel plate capacitor technique
Dielectric properties of materials can be interpreted
both macroscopically and microscopically. From the mac- With this technique, the complex permittivity of materi-
roscopic point of view they are the relationship between the als is measured using a perfect capacitor model. The spec-
applied electric eld strength E (V/m2) and the electric imen is placed between two parallel plates made of perfect
displacement D (C/m2) in the material. Microscopically, conductors, and a uniform electric eld over a large volume
dielectric properties represent the polarization ability of of space is generated [1]. Parallel plate capacitor technique
molecules in the material corresponding to an externally requires the specimen to possess at surfaces on its two
applied electric eld E. In engineering practices, generally, sides contacting the two parallel plates. The technique is
macroscopic descriptions of dielectric properties of materi- more applicable for a laboratory rather than an in-situ
als are used. In this paper, such a characterization of dielec- material characterization.
O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359 351

3.2. Resonator/oscillator modelresonate cavity technique anisotropic materials under various incident angles and
polarizations can be made. The free-space technique can
Resonate cavity technique is suggested by the American be further categorized into reectiontransmission [7],
Society for Testing and Materials (ASTM) [15] as the reection-only [8,11], and transmission-only [9] methods,
ASTM D2520 method. Baker-Jarvis et al. [2] performed depending on the experimental set-up.
measurements of the dielectric properties of low-loss mate- Estimation of dielectric properties based on reection
rials with loss tangent less than 0.005. The method employs data will encounter potential diculties when the radar
closed and open cavity congurations in which resonant operates in high frequencies. For example, multiple reec-
EM responses are measured from the material as a basis tions occurring at boundaries between layers in multi-
for determining the real and imaginary parts of the com- layered systems need to be handled, and the surface
plex permittivity. Resonate cavity technique provides more condition of the material becomes crucial for wave reec-
accurate results than broadband techniques but its results tion in high frequencies. Such diculties can be avoided
are for one frequency, thus requiring a signicant measure- by the use of transmission coecient in the estimation of
ment eort when a wide range of frequency response is to dielectric properties.
be derived. Additionally, with this method one cannot However, as mentioned in the introduction section of
measure sample sizes greater than the cavity of the this paper, the use of transmission coecients cannot
resonator. uniquely determine the complex permittivity of lossy or
low-loss construction materials such as concrete. There-
3.3. Transmission line model fore, for such materials, an improved methodology which
contains an estimation procedure for the real part of com-
Open-ended coaxial dielectric probe technique. Coaxial plex permittivity together with a root-searching procedure
dielectric probe technique is basically a cut-o section of with minimum estimation error in transmission coecients
transmission line, and the material is measured by placing is needed. In what follows, such a methodology is proposed
the probe on its machined at surface. The EM elds at the and described as a basis for predicting complex per-
end of the probe change when the probe contacts the mate- mittivity.
rial, and permittivity can be computed from the measured
reection signal [6]. Coaxial probe technique requires an
4. Methodology
intimate contact between the probe and the specimen to
eliminate the measurement error induced by air gap. Appli-
This methodology consists of two main components: (1)
cation of the technique requires certain conditions with
an estimation procedure of the real part of complex permit-
respect to the surface roughness and thickness of the mate-
tivity based on the TDOA information, and (2) a root-
rial. For example, Arai et al. [4] suggested that the speci-
searching procedure of possible combinations of real and
men surface roughness should be less than 0.5 lm to
imaginary parts of complex permittivity based on paramet-
minimize the air gap error.
ric SI and SSE criterion. In this section, a theoretical repre-
Rectangular waveguide technique. Waveguide technique
sentation of transmission coecient which is derived from
is one of a class of two-port measurement (transmission
EM wave theory is introduced. The estimation procedure
line) techniques. A sample of the material needs to be
for the real part of complex permittivity using TDOA
machined to ll in the contact area of the waveguide, and
along with the use of parametric SI and SSE criterion is
the reection from and transmission through the material
also described. The overview of the methodology is illus-
is measured. The complex relative permittivity er and com-
trated in Fig. 1.
plex relative permeability lr of the specimen is determined
using the formula provided by Nicholson and Ross [3].
Waveguides can only operate in designed frequency bands 4.1. Theoretical representation of transmission coecient
associated with certain wave propagation modes. Several
dierent samples are needed when the measurement is con- The EM wave transmission analysis is based on theo-
ducted over a large frequency range. Inaccuracy in the retical transmission coecients for a two-dimensional
measurement may occur due to the air gap between model for EM uniform plane wave propagation through
the waveguide and the specimen, and the dimensions of a dielectric medium proposed by Kong [16] (Fig. 2). By
the specimen [5]. Waveguide technique can be tedious in solving Maxwells equations and applying energy conser-
terms of sample preparation when the designated fre- vation principles with appropriate dispersion relations
quency band is not available in advance. for the particular case in hand, the expression of the com-
plex transmission coecient T* for transverse electric
3.4. Free-space technique (TE) waves in normal incidence condition is given as
follows:
Free-space technique is non-contact and non-destruc-
4ejk1z k0z d 1
tive. It requires little or no sample preparation. With this T 4
technique, broadband characterization of isotropic or 1 p01 1 p10 1 R01 R10 ej2k1z d 1
352 O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359

Fig. 1. Overview of the proposed methodology.

Network
Analyzer

Region 0 Region 1 Region 2


0, 0 1*, 1* 0, 0

Transmitting Receiving
horn antenna horn antenna

1 2

d0 d1 d0

- z-direction

Fig. 2. The two-dimensional model for EM wave transmission analysis.

p p
where k 0z x e0 l0 , k 1z k 0z lr er , p01 kk1z0z p1 ll0 kk1z0z , tion through a dielectric specimen is modeled as a two-
10 1
1p01 dimensional problem. The use of this model is justied
R01 1p01
, d1 = thickness of the specimen, l0 = permeabil-
by the choice of an appropriate experimental set-up, in
ity of free space (4p 107 H/m), lr = complex relative which uniform plane wave conditions are achieved by sat-
permeability of the specimen, e0 = permittivity of free isfying the far-eld condition. Additionally, the assumption
space (8.85 1012 F/m), er e0r  je00r = complex relative of material homogeneity in the theoretical model leads to
permittivity of the specimen where e0r and e00r are its real the description of dielectric properties for the bulk
and imaginary parts, respectively. The EM wave propaga- material.
O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359 353

4.2. Estimation procedure based on time dierence Experimental measurements were conducted in coherent
of arrival condition where measured complex S-parameter, S 21 ,
equals the complex transmission coecient, T*. Coherent
The TDOA technique is used to estimate the dielectric condition provides a non-distorted phase in the measure-
constant of a low-loss (less conductive) material using ment of S 21 (between antennas 2 (receiver) and 1 (transmit-
experimental measurements of transmission coecients. ter)) such that amplitude attenuation within the specimen
This technique is conceptually based on the same two- does not contribute to the measured transmission coe-
dimensional model for EM wave propagation shown in cient. In coherent condition, the magnitude of T* expressed
Fig. 2. Under the assumptions of normal incidence, rst- in decibel (dB), TdB, is related to S 21 by
peak response, and minor transmission losses, the time  
dierence of arrival of an EM plane wave due to the pres- T dB 10  log S 21  S21 7
ence of the specimen is [10]
where TdB is a real number and S21 is the complex conju-
d p0  gate of S 21 . Since it is observed from Eq. (4) that T* is a
Dt er  1 5
c function of measurement frequency, thickness of the dielec-
where Dt constitutes the additional propagation time (time tric specimen, and the complex permittivity (dielectric con-
dierence of arrival) between the transmitting and receiving stant and loss factor), TdB can also be expressed as
 
antennas when the specimen is present compared to the T dB T dB x; d; e0r ; e00r 8
measurement when the specimen is absent. The estimation
of Dt is achieved by processing the measured transmission The measured TdB, denoted by T m dB , can be calculated by
coecient from frequency-domain to time-domain using substituting the measured S 21 into Eq. (7). Fig. 3 shows
inverse Fourier transformation. Eq. (5) can be used as a the T m dB values as a function of frequency for the measured
tool for assessing the dielectric constant of the specimen materials of Teon, Lexan, Bakelite, GFRP, and concrete.
by estimating its time dierence of arrival using a set of The theoretical/predicted TdB, denoted by T pdB , can be
experimentally measured transmission coecients. The calculated by substituting theoretical T* into Eq. (7), given
expression of dielectric constant can be derived from Eq. possible combinations of e0r and e00r . Parametric SI is carried
(5) as follows: out by generating a set of T pdB from the possible combina-
 2 tions of e0r and e00r . From the possible combinations of e0r and
0 c  Dt e00r , those resulting in minimum dierence between T m
er 1 6 dB
d (measurement) and T pdB (theory) will be the estimates clos-
est to the real values.
The accuracy of the Dt estimation depends on the band-
Estimation error is evaluated using an error sum of
width of the signal being processed and the accurate
squares (SSE) criterion. Each combination of the estimated
measurement of the specimen thickness. Note that Eqs.
e0r and e00r provides a corresponding SSE value, which is
(5) and (6) must be modied when the loss factor is signif-
calculated by
icant. This estimate will be the basis for the identication of
the complex permittivity that characterizes the dielectric   Xn 
 
SSE e0r ; e00r T m xi  T p xi ; e0 ; e00 2 9
material specimen. dB dB r r
i1

4.3. Root-searching procedure based on parametric where n is the number of frequency bands. With the assis-
SI and SSE criterion tance of SSE criterion, an error surface is generated
for various combinations of e0r and e00r , as shown in Fig. 4.
As noted by other research studies [8,12,14], the complex Using the estimated dielectric constant (from TDOA) a
permittivity cannot be explicitly represented in terms of the corresponding error curve containing various combi-
transmission coecient T* or the scattering parameter (S- nations of loss factor can be located from the error sur-
parameter), S 21 , which is the forward transmission gain face, as shown in Fig. 5. The most possible loss factor
measured by the transmitting and the receiving radar is determined by selecting the one with minimum error
antennas. Furthermore, the exact solution for the complex on the curve. Hence, the root-searching procedure is
permittivity is not straightforward due to the multiple roots accomplished.
associated with Eq. (4) for lossy materials [13]. In order to
resolve the problem, a root-searching procedure involving 4.4. Validation and application
the use of parametric SI and SSE criterion is proposed.
Parametric SI refers to the use of a mathematical model In order to evaluate the predictions from the proposed
to characterize the behavior of a system [17], and in this methodology, experimental measurements of transmission
application it is used to provide the theoretical estimation coecients were conducted for several materials whose
of transmission coecients. Thereafter, SSE criterion is dielectric properties are known from the literature. These
introduced for evaluating the optimal combination of com- materials are Teon, Lexan, and Bakelite. In addition,
plex permittivity. Portland cement concrete and glass ber-reinforced
354 O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359

0 0

-1 -1

-2 -2
Tm [dB]

Tm [dB]
-3 -3
dB

dB
-4 -4

-5 -5
Teflon Lexan
Bakelite GFRP
-6 -6
8 10 12 14 16 18 8 10 12 14 16 18
Frequency [GHz] Frequency [GHz]

0
Concrete

-5

-10
Tm [dB]

-15
dB

-20

-25

-30
8 10 12 14 16 18
Frequency [GHz]

Fig. 3. Magnitude of experimental complex transmission coecient (in dB).

polymer (GFRP) specimens were also measured as exam- of the complex transmission coecient. The objective of
ples of construction materials. the experiments was to obtain the complex transmission
coecients for a broad range of frequencies. A schematic
4.4.1. Sample description and experimental conguration of the experimental set-up is shown in Fig. 2.
For the measurements, slab-type material specimens of Transmission measurements were collected from X-
305mm-by-305mm cross-section with varying thicknesses band through Ku-band (818 GHz) with a frequency step
were used. The selected width and height of the specimens of 12.5 MHz. A typical set of measurements consisted of
meet the requirement of radar measurements in the far- amplitudes (dB), phase angles (deg), and their correspond-
eld condition. Table 1 shows the various specimen thick- ing frequency. Raw data was then properly calibrated using
nesses for the materials used. The concrete sample was free-range measurements, which were conducted in the
manufactured with a cement/sand/aggregate mix ratio of absence of the specimen. Far-eld test conditions were
1:2.25:3.2 by weight. The water-to-cement ratio (w/c) was ensured during measurements. The far-eld condition is
0.60. Portland cement of Type I was used. The uniaxial required to ensure that the wave front is approximately
compression strength of the concrete was 24 MPa at 28 plane, which is directly related to the theoretical methodol-
days. The glass ber-reinforced polymer (GFRP) sample ogy for dielectric property characterization. Far-eld
was manufactured by extruding a 305mm-by-305mm conditions also minimize complex wave behavior in near-
square from a unidirectional ber woven sheet, which eld between the horn antennas and the specimen. Consid-
was then saturated with an epoxy thermoset polymer. ering the highest frequency of 18 GHz for the proposed
The sample was cured for 7 days before testing. experiments, it was calculated that the specimen should
The experimental set-up used in this study to measure be placed at least 10.8 cm away from the horn antennas,
the transmission of EM waves through a dielectric material and the minimum area of the slab specimens should be
involves a network analyzer and a pair of horn antennas. greater than 161.29 cm2 for adequate illumination to satisfy
The network analyzer was a Hewlett Packard Model the far-eld condition. The applied experimental set-up
8510C that was operated in step frequency mode. At each meets these requirements. Direct coupling between the
frequency the network analyzer performs a measurement horn antennas is also eliminated in this experimental
O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359 355

Fig. 4. The estimation error surfaces using parametric SI and SSE criterion: (a) teon, (b) lexan, (c) bakelite, (d) portland cement concrete.

set-up due to the use of the network analyzer. The ed as the most appropriate value. The estimated loss
unwanted coupling between antennas is measured in free- factors thus found are given in Table 3 for the test
range and stored in the phase angle information collected materials.
by the network analyzer. This information is then used in
the calibration of all consequent measurements [18]. 4.5. Discussion

4.4.2. Estimation of the dielectric constant using TDOA From the results, it is found that a unique combination
The TDOA technique was used to estimate the dielectric of real and imaginary parts of the complex permittivity can
constant of materials under investigation from both X- and be found through the proposed methodology for low-loss
Ku-bands. After performing inverse Fourier transforma- materials, such as Teon, Bakelite, Lexan, and dried/hard-
tion of the experimental transmission coecient data, the ened concrete in this study. The non-uniqueness problem
time dierences of arrival (Dt) for the materials considered of real and imaginary parts of complex permittivity has
in this study were calculated. These results are tabulated in been resolved, and appropriate values were predicted using
Table 2. Using the time dierence of arrival information, the integrated methodology. The method has potential for
application of Eq. (6) yielded estimates for dielectric con- in-situ measurement of dielectric properties for construc-
stants, which are also tabulated in Table 2. For the case tion materials.
of GFRP the TDOA technique could not provide reliable The estimated results of complex permittivity using the
results due to the very small thickness of the specimen. proposed methodology are summarized in Table 3. The
estimated dielectric constant values of Teon, Lexan,
4.4.3. Root-searching results of loss factor Bakelite, and concrete are close to the reported values or
Following the previously described procedure, the loss within the ranges reported in the literature. It is empha-
factor can be found by extracting an error curve using sized that the values found using the proposed methodol-
TDOA (shown in Fig. 5) from the error surface generated ogy for dielectric constants and loss factors are not
by parametric SI and SSE criterion. The loss factor corre- expected to exactly comparable with the reported values
sponding to the minimum SSE on the error curve is identi- in the literature because of the dierences in frequency
356 O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359

7 7

6.5
6.6

6
6.2
log (SSE)

log (SSE)
5.5

5.8
5

5.4
4.5

4 5
0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
(a) Loss Factor (b) Loss Factor

2.2 2.6

2.15

2.1 2.4
log (SSE)

log (SSE)

2.05

2 2.2

1.95

1.9 2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(c) Loss Factor (d) Loss Factor

Fig. 5. Results of root-searching procedure for identication of imaginary part of complex permittivity: (a) teon, (b) lexan, (c) bakelite, (d) portland
cement concrete.

Table 1 Table 3
Specimen thickness Characterization of complex permittivity for the test materials
Material Thickness (mm) Material Estimated Estimated Dielectric
dielectric loss constant from
Teon 6
Lexan 6 constant factor literature
Bakelite 6 Teon 1.79 0.04 2.0 [19]
Portland cement concrete 50 Lexan 2.48 0.04 2.77 [23]
GFRP 1.5 Bakelite 5.95 0.28 5.0 [22]
Concrete 5.69 0.62 4.447.22 [20]

Table 2 could not be made because of the lack of information on


Dielectric constant estimation based on TDOA for 818 GHz frequency loss factors in the literature. Also, loss factor values are
range
generally more sensitive to measurement frequency and
Material Dt (ns) Estimated dielectric experimental parameters.
constant
Eect of selected frequency bands in TDOA. TDOA esti-
Teon 0.007 1.79 mates for the dielectric property characterization of the
Lexan 0.012 2.48
Bakelite 0.030 5.95
materials presented in this study were obtained using the
Concrete 0.235 5.69 entire set of measurements for the X- and Ku-bands. For
practical applications of this methodology, the question
arises as to what would be the optimum frequency range
ranges specimen preparation, and measurement tech- to provide a sound estimate for the dielectric constant. Cal-
niques. Note that in Table 3 the loss factor comparison culations were conducted in 1 GHz, 2 GHz, 3.3 GHz,
O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359 357

Table 4
Estimates for dielectric constant using TDOA for dierent frequency ranges
Material Frequency range (GHz)
89 910 1011 1112 1213 1314 1415 1516 1617 1718
Teon 1.42 5.80 1.89 1.89 2.37 1.18 1.32 2.28 2.28 2.34
Lexan 6.05 6.96 2.25 2.53 2.06 4.53 3.61 2.41 3.23
Bakelite 0.96 14.75 5.38 8.86 8.17 5.45 4.69 9.76
Concrete 4.80 6.03 6.64 5.30 5.01 4.66 8.04 6.76 5.99 4.47

5 GHz, and 10 GHz intervals covering the range from (2) Accuracy of the use of simplied wave velocity
8 GHz to 18 GHz to study the eect of frequency band- From EM wave theory it is known that the wave
width. Table 4 presents a summary of dierent estimates (phase) velocity of EM waves is a function of dielec-
of dielectric constant using TDOA that were calculated in tric properties of materials. For example, for lossy
1 GHz bandwidth. The convergence of these estimate materials, the phase velocity within the medium is
curves is evaluated by the variance shown in Fig. 6. It is [16]
observed that estimates calculated from narrow frequency
" r !#1=2
bandwidths may provide poor estimates for dielectric con- x 1 1 r2
stant. Accurate estimates using the TDOA technique vp p  1 2 21
kR le 2 xe
require the use of wide frequency bandwidths for uniquely
identifying the dielectric constant. Reliable estimation was Theoretical vp for lossy materials 10
achieved when frequency bandwidth used exceeded
3.3 GHz in our measurements. The accuracy of such esti- where kR is the real part of the complex wavenumber,
mation is critical to the subsequent determination of the k = kR + j kI, and l is the magnetic permeability (H/
loss factor. m). The rst order expansion of vp provides its
Limitations of the use of TDOA. approximation as

1
(1) Eective thickness of specimenFrom the results for 1 1  r 2
GFRP, it is observed that the thickness of the sample vp p  1
le 8 xe
aects the performance of TDOA technique. Thus, a
Approximated vp for lossy materials 11
minimum thickness limitation might apply to the
accurate characterization of dielectric constant of
construction materials using the proposed methodol- For lossless materials r = 0 and the wave velocity
ogy. Our preliminary investigations indicate that becomes
measurements of specimens of 6 mm or larger would 1 c
provide acceptable results. vp p p
12
le l0r e0r

25
where l0r is the relative magnetic permeability. For
Teflon non-magnetic materials such as concrete, vp becomes
Lexan pc0 which is the wave velocity representation used in
er
Bakelite
20 the TDOA estimation. Eq. (12) is used in the TDOA
Concrete
procedure described in this paper because of its sim-
plicity, and it is considered as a convenient approxi-
15
mation of Eq. (11) when r/xe  1. To study the
Variance

accuracy of this approximation, a quantitative study


10
was performed. Wave velocities using Eqs. (10)(12)
are calculated with respect to dierent values of r/
xe ranging from 0 to 20, shown in Fig. 7. It is ob-
5 served that the dierences between Eqs. (10) and
(12) or between Eqs. (11) and (12) may be substantial
when r/xe is greater than 1. In the range of investiga-
0 tion where r/xe < 0.1, the dierence between either
1 2 3 4 5 6 7 8 9 10 Eqs. (10) and (12) or Eqs. (11) and (12) is insignicant
Frequency Bandwidth as demonstrated in Fig. 7(b). Hence, the use of Eq.
Fig. 6. Convergence of estimates of dielectric constant at dierence (12) in our methodology for low-loss materials is
frequency bandwidths using TDOA. justied.
358 O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359

1.001
Theoretical vp in lossy media
1
Approximated vp in lossy media
Normalized Phase Velocity

Normalized Phase Velocity


1
vp in lossless media
0.8
0.999

0.6
0.998

0.4
0.997

0.2 Theoretical vp in lossy media


0.996
Approximated vp in lossy media
vp in lossless media
0 0.995
0 5 10 15 20 0 0.02 0.04 0.06 0.08 0.1
(a) / ( ) (b) / ( )

Fig. 7. Normalized phase velocity vs. loss factor: (a) r=xe 020, (b) r=xe 00:1.

5. Conclusion under the supervision of Dennis Blejer. We gratefully


acknowledge his eorts and contributions to the research
Non-unique combinations of dielectric constant and loss reported in this paper.
factor are encountered when lossy materials are measured.
In order to resolve this issue, an integrated methodology References
for determining the unique combination of complex per-
mittivity based on measured transmission coecient and [1] Al-Qadi IL, Hazin OA, Su W, Riad SM. Dielectric properties of
time dierence of arrival (TDOA) information is proposed. Portland cement concrete at low radio frequencies. ASCE J Mater
Civil Eng 1995;7(3):1928.
The methodology consists of a root-searching procedure [2] Baker-Jarvis J, Geyer RG, Grosvenor Jr JH, Janezic MD, Jones CA,
based on parametric systems identication and an estima- Riddle B, et al. Dielectric characterization of low-loss materialsa
tion of the real part of complex permittivity based on comparison of techniques. IEEE Trans Dielectr Electr Insulat 1998;
TDOA. Studies performed on selected specimens of Teon, 5(4):5717.
Lexan, Bakelite, and Portland cement concrete show good [3] Nicholson AM, Ross GF. Measurement of the intrinsic properties of
materials by time domain techniques. IEEE Trans Instrum Measure
correlation with previously reported values. One limitation 1970;IM-19:37782.
of the proposed method relates to the determination of [4] Arai M, Binner JPG, Cross TE. Estimating errors due to sample
dielectric constant for very thin specimens such as the surface roughness in microwave complex permittivity measurements
GFRP sheet with a thickness of 1.5 mm as used in this obtained using a coaxial probe. Electron Lett 1995;31(2):1157.
study. Our preliminary investigations indicate that the [5] Deshpande MD, Reddy CJ, Tiemsin PI, Cravey R. A new approach
to estimate complex permittivity of dielectric materials at microwave
measurement of specimens with a thickness of 6 mm or lar- frequencies using waveguide measurements. IEEE Trans Microwave
ger provides satisfactory results. This is not considered to Theory Techn 1997;45(3):35965.
be a signicant limitation in measuring properties of con- [6] Gershon DL, Calame JP, Carmel Y, Antonsen TM, Hutcheon RM.
struction materials since typical dimensions involved in Open-ended coaxial probe for high-temperature and broad-band
dielectric measurements. IEEE Trans Microwave Theory Techn 1999;
construction applications are greater than such a thickness.
47(9):16408.
It is anticipated that the proposed methodology will be [7] Maurens M, Priou A, Brunier P, Aussudre S, Lopez M, Combes P.
applicable to the in-situ measurements of dielectric proper- Free-space microwave measurement technique for composite miner-
ties in construction applications for eective non-destruc- als. In: Priou A, editor. Progress in electromagnetics research (PIER)
tive evaluation of materials and structures. The developed 06, 1992. p. 34585.
method has practical advantage in application to variety [8] Ghodgaonkar DK, Varadan VV, Varadan VK. A free-space method
for measurement of dielectric constants and loss tangents at micro-
of materials. However, the results reported in this paper wave frequencies. IEEE Trans Instrum Measure 1989;37(3):78993.
are preliminary and further validation will be necessary [9] Aurand JF. Measurement of transient electromagnetic propagation
as more measurements become available. through concrete and sand. Sandia Report 96-2254, Sandia National
Laboratory, New Mexico, 1995.
[10] Stone WC. Non-line-of-sight (NLS) construction metrology. NIST
Acknowledgements Construction Automation Program Report No. 1, National Institute
of Standards and Technology, Maryland, 1996.
The support for this work was partially provided by the [11] Seo IS, Chin WS, Lee DG. Characterization of electromagnetic
National Science Foundation through Grant CMS- properties of polymeric composite materials with free space method.
0324607, and by the Massachusetts Institute of Technology Compos Struct 2004;66:53342.
[12] Zhang J, Huang Y. Extraction of dielectric properties of building
(MIT) Lincoln Laboratory through Grant ACC-376 (Ad- materials from free-space time-domain measurement. In: Proceedings
vanced Concepts Committee). Wide-bandwidth radar mea- IEEE 1999 high frequency postgraduate student colloquium, 1999. p.
surements were performed at MIT Lincoln Laboratory 12732.
O. Buyukozturk et al. / Cement & Concrete Composites 28 (2006) 349359 359

[13] Jose KA, Varadan VV, Hollinger RD, Tellakula A, Varadan VK. [18] Cullen AL. A new free-wave method for ferrite measurement at
Non contact broadband microwave material characterization at low millimeter wavelengths. Radio Sci 1987;22(7):116870.
and high temperatures. In: Nyugen C, editor. Proceedings of SPIE in [19] van Krevelen DW. Properties of polymers, correlations with chemical
subsurface sensing technologies and applications II, 2000. p. 32431. structure. Amsterdam, The Netherlands: Elsevier; 1972.
[14] Millard SG, Davis J, Huang Y, Bungey JH. A wide band system for [20] Rhim HC, Buyukozturk O. Electromagnetic properties of concrete at
measuring dielectric properties of concrete. In: Koppenjan S, Lee H, microwave frequency range. ACI Mater J 1998;95(3):26271.
editors. Proceedings of the ninth international conference on ground [21] Rhim HC, Buyukozturk O. Wideband microwave imaging of con-
penetrating radar, 2002. p. 41621. crete for nondestructive testing. ASCE J Struct Eng 2000;126(12):
[15] ASTM Subcommittee D09.12. American Society for Testing and 14517.
Materials Standards, vol. 10.02. West Conshohocken, PA: ASTM [22] Goeckel DL, Webb KJ, Gallagher NC. Massively parallel iterative
International; 2001. determination of stratied dielectric parameters from scattered-eld
[16] Kong JA. Electromagnetic wave theory. Cambridge, MA: EMW measurements. J Opt Soc Amer A 1993;10(5):1093100.
Publishing; 2000. [23] Power Corporation. Lexclad Insulation Systems Brochure #PC-025,
[17] Venkatesh SR, Dahleh MA. On system identication of complex Severn, MD, 1996.
systems from nite data. IEEE Trans Automat Control 2001;46(2):
23557.