This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

Design and Application of the CSRR Based

Planar Sensor for Non-Invasive Measurement of
Complex Permittivity
M. Arif Hussain Ansari, Abhishek Kumar Jha, Student Member, IEEE and M. J. Akhtar, Senior Member, IEEE

 developed in the past for the permittivity measurement of

Abstract— A novel microwave non-invasive planar sensor material under test (MUT). These methods can be classified
based on the complementary split ring resonator (CSRR) is into the free space method, the transmission line method and
proposed for accurate measurement of the complex permittivity the resonant method. The free space method usually employs
of materials. The CSRR is etched in the ground plane of the
the highly directive lens antenna placed on either one or both
planar microstrip line. Two CSRRs of rectangular and circular
cross-sections are chosen for the sensitivity analysis where the sides of the MUT. The antennas are connected to the vector
later is found to possess higher sensitivity, and hence appears to network analyzer (VNA) to measure the attenuation and phase
be more appropriate for the sensor design. At resonance, the constant in order to characterize the sample [6]. The
electric field induced along the plane of CSRR, is found to be measurement of the complex permittivity over a wide band of
quite sensitive for the characterization of specimen kept in frequency can be achieved using this method, which can also
contact with the sensor. A numerical model is developed here for
be used when the sample can only be accessed from one side
the calculation of the complex permittivity as a function of
resonant frequency and the quality factor data using the due to some technical difficulties [7]. However, this method
electromagnetic simulator, the CST. For practical applications, a usually requires the larger samples (laterally extended enough
detailed air gap analysis is carried out to take into account the to neglect the diffraction effect at the edges), which is the
effect of any air gap present between the test sample and the major limitation for testing of thin and electrically small
CSRR. The designed sensor is fabricated and tested, and samples.
accordingly the numerically established relations are
The transmission-line based method is commonly being
experimentally verified for various reference samples e.g., Teflon,
PVC, Plexiglas, Polyethylene, Rubber and wood etc. used to characterize the dielectric properties of materials by
Experimentally, it is found that the permittivity measurement placing the sample inside a section of the transmission line
using the proposed sensor is possible with a typical error of 3%. such as waveguide, coaxial airline, microstrip-line etc. The
complex permittivity of MUT is determined in terms of the
Index Terms— Complementary split ring resonator (CSRR), measured reflection and transmission coefficients [8]. This
complex permittivity, non-Invasive measurement, planar sensor, method is more cost effective than the free-space method.
resonant frequency.
However, the accuracy of the transmission-reflection approach
is not very good especially for low loss samples, and the
I. INTRODUCTION
sample preparation is also quite often a challenging task [9].
ermittivity is one of the most important parameters of
P materials in the context of RF and microwave engineering.
The response of any microwave planar circuit is greatly
The resonant cavity perturbation is usually the most
accurate technique for low loss materials. The change in the
measured resonant frequency and quality factor gives the
influenced by the dielectric properties of materials used in the
information about the dielectric properties of materials [10]. In
substrate. For modeling of sophisticated microwave devices, a
this method, traditionally the cavity resonators have been used
prior knowledge of the complex permittivity is necessary.
for determination of the dielectric properties of the MUT.
Therefore, an accurate determination of the complex
permittivity of materials is an important task in the field of RF Recently, the resonant sensors based on the planar microstrip-
and microwave engineering. The evident applications of line loaded with complementary split ring resonator (CSRR)
accurate complex permittivity measurement have also been have been reported to determine the dielectric constant of the
explored in various areas like food [1]-[2], healthcare [3], MUT [4], [11]-[13]. The design of resonant sensors based on
defense [4] and agriculture [5]. the planar technology has certainly many advantages such as
the low cost, portability, being non-invasive and ease in
Several microwave characterization techniques have been sample preparation. However, most of these papers mostly
consider the determination of only the real part of the
M. A. H. Ansari is with the Materials Science Programme, Indian Institute permittivity without considering the effect of air gap between
of Technology Kanpur, Uttar Pradesh 208016, India (e-mail
[email protected]). the test specimen and the planar structure. The determination
A. K. Jha is with the Department of Electrical Engineering, Indian Institute of loss factor is recently considered in one of the papers [13],
of Technology Kanpur, Uttar Pradesh 208016, India (e-mail: but the air gap effects between the test specimen and the
[email protected]).
M. J. Akhtar is with the Department of Electrical Engineering and the CSRR unit cell have still not been considered. Moreover, all
Materials Science Programme, Indian Institute of Technology Kanpur, Uttar these papers simply consider the CSRR shape to be
Pradesh 208016, India (e-mail: [email protected])

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

rectangular for measuring the dielectric properties of ∆𝜇 is the change in complex permeability while 𝜀0 and 𝜇0
materials. represent the free space permittivity and permeability,
In this paper, the geometry of CSRR is chosen only after respectively. The symbols 𝑬𝟎 , 𝑯𝟎 are the electric and the
performing the detailed sensitivity analysis where it is found magnetic fields of the empty cavity while 𝑬𝟏 , 𝑯𝟏 represent the
that the circular CSRR provides better sensitivity in electric and the magnetic fields, respectively under loaded
comparison with the rectangular CSRR having the same unit condition. For electrically small samples, the electric and the
area. This finding is also validated with the help of magnetic fields inside the resonator, before and after loading
electromagnetic analytical expressions. Additionally, the air the sample is assumed to be unchanged. For dielectric
gap effect between the test sample and the sensor, which is a materials, the change in complex permeability (∆𝜇) may be
common and inevitable problem in these types of assumed to be zero. Also, the electric and magnetic field
measurements, is critically studied and numerically modeled energy stored in the resonant structure must be equal at the
here. New sets of equations are accordingly proposed for the resonant frequency. Under these conditions, the expression (1)
complex permittivity calculation including the air gap effect, is simplified as
which are then verified for various practical cases. To the best ∆𝑓𝑟 ∫𝑣 ∆𝜀 𝑬𝟏 ∙ 𝑬𝟎 𝑑𝑣
of authors’ knowledge, the procedure of selection of CSRR = 𝑠
(2)
𝑓𝑟 𝟐 ∫𝑣 𝜀0 |𝑬𝟎 |2 𝑑𝑣
unit cell geometry based on the sensitivity analysis by 𝑐

comparing various CSRR cells having same unit area have not where, 𝑣𝑠 represent the sample volume. The above
been considered earlier in literature. Similarly, the effect of air expressions (1) and (2) are mostly being used in case of the
gap between the CSRR based sensor and the test specimen has waveguide cavities, where the electromagnetic field
not been taken into account in earlier papers. It is mainly due expressions with and without material perturbation can be
to these reasons that the aforementioned effects are properly easily derived. However, it is difficult to use this concept for
studied and considered in the present paper, which is certainly the electrically small resonators, such as those based on the
a step forward in the accurate complex permittivity CSRR structures. It is rather more convenient to design such
determination of test samples using the CSRR based sensors. kind of resonators with the numerical methods so that the
The proposed technique is based on the design and effects of various parameters can be accurately observed and
development of a microstrip based CSRR resonant sensor, taken into account. Many research groups have proposed
which shows the characteristics akin to a stop band filter. For several geometries of CSRR in the recent past [11], [19].
testing procedures, the MUT is placed on the ground plane of However, in material characterization, the sensitivity of the
the microstrip line such that it covers the complete area of the planar sensor is of prime concern, which is further associated
with the electric and the magnetic field intensity present across
CSRR. The proposed sensor is designed and simulated using
the planar architecture. Therefore, in this paper, firstly two
the numerical electromagnetic solver, the CST MWS [14],
extensively used planar architectures viz., the circular and the
which facilitates the development of a numerical model for
rectangular CSRRs, are compared for the sensitivity analysis
microwave characterization of materials. The accuracy of the using the CST MWS.
developed method is verified using the data available in the
literature [15]-[17]. The proposed sensor is fabricated on a III. COMPARISON OF CIRCULAR AND RECTANGULAR CSRRS
FR4 substrate and is found to be working in the frequency
range of 1.7 to 2.7 GHz. The resonant frequency and The configuration and the design parameters of circular and
magnitude of the transmission coefficient of the dielectric rectangular CSRRs are shown in Fig. 1(a) and 1(b),
samples measured using the vector network analyzer (VNA) respectively. The equivalent circuit of both the CSRRs may be
are used in the derived numerical model to calculate the represented using the topology shown in Fig. 1(c). The
complex permittivity of the test specimen. Additionally, a symbols Lr and Cr represent the inductance and capacitance of
detailed analysis is performed to determine the error in CSRR, respectively, L is the line inductance; Cc is the
measurement due to possible air gap present between the coupling capacitance between microstrip line and CSRR. The
CSRR and the MUT. subscript i in this figure refers to c for the circular CSRR and r
for rectangular CSRR. For the equivalent circuit model shown
II. THEORY in Fig. 1(c), the resonant frequency may be given as [20]

For conventional microwave resonator based technique, the 1

general expressions relating the change in resonant frequency 𝑓𝑟 = (3)
2𝜋 √𝐿𝑟 (𝐶𝑐 +𝐶𝑟 )
with permittivity and permeability of the test sample are given
as [18] For a fair comparison between both the CSRRs, the unit cell
area as well as the design parameters a, b, c, d, g of the
∆𝑓𝑟 ∫𝒗 (∆𝜀 𝑬𝟏 ∙ 𝑬𝟎 + ∆𝜇 𝑯𝟏 ∙ 𝑯𝟎 )𝑑𝑣 circular and the rectangular CSRR are taken as identical
= 𝒄
(1)
𝑓𝑟 ∫𝒗 (𝜀0 |𝑬𝟎 |𝟐 + 𝜇0 |𝑯𝟎 |𝟐 )𝑑𝑣 throughout this manuscript. The FR4 substrate of thickness 0.8
𝒄
mm, and of cross-sectional dimension 40 mm x 26 mm, is
where, 𝑣𝑐 represent the cavity volumes, ∆𝑓𝑟 is the change in used to model the proposed sensor. The CSRR design
resonant frequency, ∆𝜀 is the change in complex permittivity, parameters taken in the simulation are a = 0.39 mm, b = 0.22

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

mm, c = 0.38 mm, g = 0.22 mm and d = 7 mm (Fig. 1). A. Sensitivity Analysis of Circular and Rectangular CSRRs

The quasi-static electrically small resonators are usually

designed with a loop and a gap separating the loop in two
parts. The electrically small structure resonates due to
inductance induced by circulating current in the loop and the
effective capacitance developed across the gap between the
loops. The change in capacitance of CSRR usually depends on
the change in the permittivity of the MUT, whereas the
inductance of the CSRR is considered to be unchanged in case
of dielectric materials. In this paper, two CSRRs (rectangular
(a) (b) and circular as shown in Fig. 1) having equal unit cell area, are
taken for the sensitivity analysis and a detailed discussion is
given in next few sections.

1) The resonant frequency analysis

The circular and rectangular CSRRs having same unit cell

are separately excited with the help of microstrip line. Since
the CSRRs are usually excited by the normal component of
the electric field, hence they are modeled in the ground plane
(c) of the microstrip line (Fig. 2). It is to be noted that the sample
Fig 1. A typical design of (a) circular and (b) rectangular CSRRs with the under test in the present situation is not required to cover the
same unit cell area, and (c) equivalent circuit model. Black area represents whole sensor area. However, it should cover the full area of
conductor part and white area represents etched out part.
the CSRR unit cell for the effective perturbation of the electric
At first, both the sensors are modeled in the CST, and the field as shown in Fig.2 (b). In both cases, the electric field
simulation is carried out to obtain the two port scattering associated with the microstrip usually provides the coupling
parameters in the specified frequency band. The lumped required for the excitation of the CSRRs. The designed
parameters of both the CSRRs, excited by the microstrip line, sensors are simulated here using the time domain solver of
are calculated by considering the transmission notch CST-MWS, where two port scattering parameters including
frequency, blotch impedance and the reflection coefficient the transmission coefficient data are calculated. According to
using the Smith chart plot [20]. The calculated lumped the equivalent circuit model of the proposed sensor (Fig. 1(c)),
parameters are presented in Table I. the resonant frequency is considered as the transmission-stop
frequency of the unloaded CSRR (fo). The proposed sensor is
TABLE I: EXTRACTED CIRCUIT PARAMETERS FOR BOTH CSRRS then loaded with the material under test, where the dielectric
constant of the sample is varied in a wide range from 1 to 10.
Lumped Circular CSRR Rectangular CSRR The resonant frequency (fr) corresponding to each sample is
Parameters calculated, which is also plotted with the dielectric constant
L (nH) 4.2141 4.7115 variation as shown in Fig. 3.
Cc (PF) 0.7737 0.8069
Cr (PF) 1.2817 1.3960
Lr (nH) 1.7547 2.1362

It is interesting to note from Table I that the values of

lumped parameters associated with the circular CSRR are
somewhat smaller than the rectangular CSRR. This basically
means that as per (3), the resonant frequency of the circular
CSRR will be slightly higher than the rectangular CSRR. This
can also be justified by the fact that the resonant frequency of
such resonators actually decrease with increasing value of
circumference of the CSRRs [21]. Since, for same unit cell
area, the circumference of the circle is smaller than the
rectangle, the resonant frequency corresponding to the circular
CSRR would be larger than rectangular one. After getting the
required circuit parameters, the electromagnetic field intensity
present at the plane of CSRRs is compared such that the Fig. 2. A typical CSRR based planar sensor (a) Top view with transparent
distinction in terms of sensitivity can be evaluated. A detailed substrate, (b) Perspective view of simulated model with MUT placed over
ground plane, (c) Side view showing the thickess of substrate, ground plane
sensitivity analysis is presented in the following section. and MUT

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

numerator of (5) would be larger as compare to rectangular

CSRR. Similarly, the average energy dissipation per cycle for
the rectangular CSRR will be more than that of the circular
CSRR because of its larger circumference. The energy
dissipation is directly proportional to the tangential component
of the magnetic field which basically means that the
denominator of (5) will be larger in case of the rectangular as
compared to the circular CSRR. Hence, by cumulative effect
of both the numerator and the denominator terms of (5), the
overall quality factor of the circular CSRR based sensor will
be larger than that of the rectangular based CSRR structure
having the same width and unit cell area.
To validate the above discussion, a similar approach like
resonant frequency analysis is adopted here to calculate the
Fig. 3. Response of sensor for circular and rectangular CSRRs due to change quality factor of the circular and rectangular CSRR from the
in dielectric permittivity of the sample. The sensitivity of sensor (relative
change in resonance frequency with the dielectric constant) due to circular simulated transmission coefficient. The proposed sensors are
CSRR is larger than the rectangular CSRR. simulated for various values of loss tangent data ranging from
0 to 0.1, by considering the dielectric constant to be constant
For better understanding, the change in resonant frequency
at a typical value of 𝜺′𝒓 =3. The quality factor corresponding to
(∆f = fo-fr) is also plotted in the same graph. It may be noted all values of loss tangent is plotted in Fig. 4. Additionally, for
from Fig. 3 that the resonant frequency of the circular CSRR better understanding, differential change in the quality-factor
is about 350 MHz greater than resonant frequency of the (represented in terms of the transmission coefficient) is also
rectangular CSRR, which again validates the observations plotted in the same figure.
made in earlier section and data given in Table I. However, it
may be interesting to note that the change in resonant
frequency for the circular CSRR under loaded condition is
also greater than that of the rectangular CSRR. In other words,
for a given unit cell area, the circular CSRR appears to be a
good choice to achieve high sensitivity for a given dielectric
sample as compared to that of the rectangular CSRR.
2) The Quality factor analysis

The quality factor, Q of a generalized resonator may be

given as [22]
𝑊
𝑄 = 𝜔0 𝑃 (4)
𝐿

where, W is the total stored energy, PL is the average energy

dissipated per cycle, and 𝜔0 represents the angular resonant Fig. 4. Response of sensor in terms of quality factor and change in insertion
frequency. The above equation can also be written in terms of loss.
the H-field as follows From Fig. 4, it may be noted that the change in quality
2 ∰|𝐻|2 𝑑𝑣 factor w.r.t change in the loss tangent of MUT in both the
𝑄= (5) cases (circular and rectangular) are similar. However, as
𝛿 ∯|𝐻𝑡 |2 𝑑𝑠
explained earlier, it is quite interesting to note that the quality
where, Ht is the tangential component of H-field. In order to factor corresponding to the circular CSRR is greater than the
differentiate two CSRRs in terms of the quality factor, the rectangular CSRR for all values of loss tangent.
magnetic fields appearing in numerator and denominator of
(5) are calculated individually here for both the cases starting IV. NUMERICAL MODELING
from first principle. From definition of the induced magnetic From the discussion in earlier sections, it may be concluded
field due to a current carrying ring of circular and rectangular that for a given sample, the circular CSRR provides better
shape having same cross section, the ratio of magnetic fields resolution as compared to that of the rectangular CSRR. It is
appearing in the numerator of (5) may be derived as [23] mainly due to this reason that the circular CSRR is used in the
𝐻𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑟 𝜋 proposed work to design the planar resonant microwave
𝐻𝑟𝑒𝑐𝑡𝑎𝑛𝑔𝑢𝑙𝑎𝑟
= 2√2
(6) sensor as shown in Fig. 2. Now, in order to characterize the
test specimen using the proposed resonant sensor, a numerical
From above equation, it may clearly be seen that the total model is required which usually maps the measured
magnetic field for the circular CSRR corresponding to parameters (e.g., the resonant frequency and the quality factor)
to the complex permittivity of the test sample. For numerical

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

simulation of the proposed sensor, the electromagnetic solver It is interesting to note that the result obtained using the full
CST MWS [14] is used. The time domain solver of CST wave field analysis (Fig. 6) does verify the relation (3)
MWS is used to optimize various parameters of the designed obtained from the equivalent circuit model. From (3), it may
sensor. The proposed sensor is excited by means of a be observed that the inverse square of the resonant frequency
waveguide port, and the transmitted electromagnetic wave is is directly proportional to the real permittivity, i.e. 𝑓𝑟−2 ∝ 𝜀𝑟′ .
recorded by means of another waveguide port, each having a It may be noted here that the values of 𝐿𝑟 and 𝐶𝑐 are
50Ω port impedance. The hexahedral meshing with perfect considered to be constant due to the fixed values of the overall
boundary approximation is adopted for the finite integration length of CSRR and the dielectric constant of substrate,
technique (FIT). A perspective view of the proposed model respectively. In Fig. 6, it may also be observed that the slope
simulated using the time domain solver of CST is shown in of plotted curve depends on the thickness of MUT. However,
Fig. 2(b) along with the waveguide ports and the MUT. it is noticed that the slope of the curve remains almost constant
The simulated resonant frequency and the quality factor of for the sample thickness (ts) greater than 5 mm. This particular
the proposed sensor in the unloaded condition are found to be behavior may be observed from the two plotted lines
2.65 GHz and 80, respectively. After loading the sensor with corresponding to the sample thickness of 5mm and 10mm,
the MUT, a change in the resonant frequency as well as in the where both the curve overlaps each other.
magnitude of S21 (dB) is observed as shown in Fig. 5. This is
achieved by varying the dielectric properties of the MUT.
During the numerical simulation, various dielectric materials
with the dielectric constant ranging from 𝜀𝑟′ = 1 to 10 and 𝑡𝑎𝑛𝛿
varying from 0 to 0.1 are simulated to obtain the transmission
coefficients. The values of the resonant frequency and the
quality factor are extracted from the recorded transmission
coefficients, and are then used to derive the numerical model
with the help of the curve fitting technique.
For the curve fitting of the simulated scattering data
obtained using the CST Studio, the curve fitting tools of the
commercially available software origin pro 9.1 are directly
employed. The curve fitting procedure generates the
imperative equations for the complex permittivity
determination in terms of the resonant frequency and the
insertion loss/quality factor. The equations are generated as
per the obtained sets of data to be fitted, and the particular Fig. 6. Plot of 𝑓𝑟−2 and 𝜀𝑟′ for different thickness of sample.
curve profile is chosen according to the least error between the
chosen profile and the sets of numerically obtained data. Therefore, in order to incorporate all the above effect, the
dielectric constant of the sample is mathematically expressed
below in terms of the family of straight lines, where the
independent parameter are taken as the resonant frequency (fr)
and the sample thickness (ts).
𝑓𝑟−2 − 0.11685
𝜀𝑟′ = (7)
0.0202−0.0168×0.2883𝑡𝑠
The above expression is established using the curve fitting
technique, which provides a numerical model of the proposed
sensor to calculate the real permittivity of sample having finite
thickness in terms of the measured resonant frequency.
B. Determination of imaginary permittivity
After establishing the numerical expression (7) for
calculation of the dielectric constant of test sample, a similar
Fig. 5. Variation of S21 (dB) magnitude of sensor with loss tangent value analysis is performed to establish a numerical expression for
ranging from 0 to 0.1 for 𝜀𝑟′ =2 and 4.
calculating the loss tangent of the test sample. At first, in this
section, the effect of sample thickness on the loss tangent
A. Determination of the real permittivity calculation is analyzed. For this analysis, the sample thickness
At first, the effect of sample loading is observed in terms of of MUT is varied in a range of 2 to 10 mm, while loss tangent
the resonant frequency of the proposed sensor. The inverse is varied in the range of 0 to 0.1 keeping the dielectric constant
square of resonant frequency, extracted from the simulated fixed as 3 as shown in Fig. 7. From this figure, it may be noted
transmission coefficient data, is plotted with the corresponding that the variation of inverse of the quality factor with loss
real permittivity (𝜀𝑟′ ) of MUT as shown in Fig. 6. It can be tangent data is almost linear, and slope of the line increases
observed from Fig. 6 that the variation of 𝑓𝑟−2 with 𝜀𝑟′ is linear. with the sample thickness. Hence, it can be postulated from

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

Fig. 7 that an increase in sample thickness increases the The relationship between the loss tangent, 𝑡𝑎𝑛𝛿 and the Q-
overall sensitivity of the measurement using the proposed factor can be given as follows [22]
sensor.
1 𝜀𝑟′′
𝑄𝑀𝑈𝑇 = = (9)
𝑡𝑎𝑛𝛿 𝜀𝑟′

The symbol QMUT represents the quality factor of the

proposed sensor after loading the MUT, which may be
calculated using the S21 (dB) magnitude corresponding to the
resonance frequency of the loaded CSRR sensor [21],
𝑆21 (dB)
𝑄𝑀𝑈𝑇 = 𝑄𝑈 (1 − 10 20 ) (10)
The symbol QU appearing in above expression is the Q-
factor of the CSRR sensor under unloaded condition. The
imaginary part of the complex permittivity is thus calculated
using (9) and (10).
Meanwhile it is observed from Fig. 5 that the magnitude of
S21 is changing with the variation in 𝑡𝑎𝑛𝛿 at a constant 𝜀𝑟′ .
This behavior is again studied in detail as shown in Fig. 8.
−1
Fig. 7. Relation of Q-1 with 𝑡𝑎𝑛𝛿 for different values of sample thickness. From Fig. 8, it is found that the variation of 𝑄𝑀𝑈𝑇 with 𝑡𝑎𝑛𝛿
is linear whereas the slope of plotted straight line depends on
The above explained behavior may be expressed in terms of
the real part of complex permittivity of the MUT. Therefore,
following numerical expression which is derived from Fig. 7
to quantify the loss tangent of the MUT, which depends on the
−1
𝑄𝑀𝑈𝑇 = (0.03424 − 0.01665 × 0.61094𝑡𝑠 )𝑡𝑎𝑛𝛿 + 0.01346 (8) loaded quality factor as well as the relative permittivity of the
MUT, a curve fitting tool is used to deduce the numerical
However, it may also be observed from Fig. 7 that when the
model as given below
sample thickness is increased beyond 8 mm, the variation in
quality factor is almost independent of the loss tangent values. −1
𝑄𝑀𝑈𝑇 − (9.048×10−5 𝜀𝑟′ + 0.01321)
Hence, for simplicity of the approach, our further study of the 𝑡𝑎𝑛𝛿 = ′ (11)
loss tangent throughout this paper is subjected to the sample [−0.149(1−𝑒𝑥𝑝0.067𝜀𝑟 )]
thickness greater than the threshold value of 8 mm. Now to
determine the loss tangent in terms of the Q factor, a number After determining the real permittivity from (7) and loss
of test samples corresponding to the dielectric constant in the tangent from (11), the imaginary part of the complex
range of 2 to 10 are taken. The loss tangent value permittivity can be calculated using (9).
corresponding to each dielectric sample is varied in the range
of 0 to 0.1, and the quality factor for each case is calculated V. MEASUREMENT AND RESULT
from the simulated transmission coefficient data. The inverse
The sensor is fabricated using planar technology comprising
of quality factor and the corresponding loss tangent are plotted
of standard photolithography technique on commercially
in Fig. 8.
available 0.8 mm FR4 substrate with a conductive copper film
coating of 35 μm. The standard photolithography technique
develops any geometric pattern from a photo-mask with the
help of light-sensitive chemical "photoresist" on the substrate.
A series of chemical treatments is required to engrave the
exposure pattern into the desired conductive film pattern upon
the substrate material underneath the photo resist. A pair of
50Ω SMA connectors is then mounted on the fabricated sensor
through the mechanical welding for the measurement of
reflection and transmission data. The measurement setup is
shown in Fig 9(a). The fabricated planar sensor length is 40
mm and its width is 26 mm. The top and bottom view of the
fabricated sensor are shown in Fig. 9(b) & 9(c), respectively.
The sensor is connected with the vector network analyzer
(VNA) through coaxial SMA connectors. All the
measurements were performed using the Agilent N5230C
Network Analyzer. The short-open-load-through (SOLT)
calibration was performed using the Agilent 85052D
Fig. 8. Linear relationship between inverse of Q-factor and 𝑡𝑎𝑛𝛿 for different CALIBRATION KIT. The intermediate frequency bandwidth
values of 𝜀𝑟′ . (IFBW) of the VNA is set to 100 Hz and the input power is
taken as 0 dBm. The number of sweep point is taken as 1601.

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

TABLE II: MEASURED COMPLEX PERMITTIVITY USING PROPOSED CSRR BASED PLANAR SENSOR TECHNIQUE

Method→ Reference values of the specimen [15]-[17] Measured data Percentage error
MUT↓ 𝜺′𝒓 𝜺′′
𝒓 𝜺′𝒓 𝒕𝒂𝒏𝜹 𝜺′′
𝒓 %𝜺′𝒓 %𝜺′′
𝒓
Teflon 2.10 0.00058 2.14 0.00025 0.00053 1.90 8.62
Poly-ethylene 2.26 0.00070 2.30 0.00034 0.00075 1.76 7.14
Plexiglas 2.60 ---- 2.70 0.01012 0.02733 3.84 ----
PVC 2.65 0.02650 2.69 0.00891 0.02396 1.50 9.58
Rubber 3.00 ---- 3.05 0.00192 0.00585 1.67 ----
Wood 2.30 0.06900 2.37 0.03201 0.07586 3.04 9.94

number of factors such as the slight geometry mismatch of

CSRRs between the fabricated and the simulated model, and
possible air gap between the sensor and the sample. The
geometry mismatch can be reduced by using sophisticated
instruments having high accuracy in photolithography and
sensor fabrication. A fraction of error present due to air gap
can be minimized by tightening the sample and the sensor
together with high tensile clip; while the error due to the
roughness of the sample-surface can be numerically modeled
which is beyond the scope of this paper. Alternatively the
effect of air gap can also be taken into account with the help of
some rigorous analysis as explained in the next section.
Fig.9. (a) Measurement setup, (b) Top view and (c) Bottom view of the
proposed sensor indicating the etched CSRR.

After calibration, a number of available materials are

characterized using the proposed numerical model. The
transmission coefficient data for all the cases are recorded
using VNA. The resonant frequency and the S21 (dB)
magnitude are calculated from the measured transmission
coefficient data. Fig. 10 shows the measured S21 (dB) plots
for various dielectric samples. The recorded resonant
frequencies for Teflon, Poly-ethylene, Plexiglas, PVC, Rubber
and wood are 2.495, 2.47, 2.410, 2.411, 2.365 and 2.459 GHz,
while their insertion losses are 19.55, 19.22, 17.35, 17.61,
16.01 and 13.48 dB, respectively. The measured resonant
frequency and the quality factor for each specimen are then
fitted into the developed numerical model and the
corresponding dielectric constant and loss tangent are
calculated using (7) and (10), respectively and shown in
columns 4 and 5 of Table II. Fig. 10. Measured S21 (dB) for various dielectric samples.
The measured results are also compared with the standard
data available in the literature [15]-[17] which are given in VI. AIR GAP ANALYSIS
columns 2 and 3 of Table II. The measured permittivity values
The consideration of the air gap between the sensor and the
of these samples using the fabricated CSRR sensor are given
sample-surface is quite important for the accurate
in 4th to 6th columns of Table II. For the accuracy
measurement of complex permittivity using the CSRR based
measurement of the proposed sensor, the percentage error is
sensors. An approximate model of the air gap effect is
also calculated and given in columns 7 and 8 of Table II. It
developed here which is based on the prior information of the
can be concluded from Table II that the results of the complex
air gap present between the sensor and the sample-surface.
permittivity measurement using the proposed sensor is in good
The numerical model is developed by changing the air gap in
agreement with the standard values available in literature. It is
practically permissible range. For this analysis, the air gap is
also to be noted from Table II that the real and imaginary parts
varied from 0 to 100 μm, and the resonant frequency of the
of the complex permittivity measurement are possible with a
proposed sensor is recorded in each case for various dielectric
typical error of 3% and 9%, respectively.
materials having relative permittivity ranging from 1 to 10.
The error obtained in the complex permittivity measurement
The realization of air gap is done using the polypropylene
presented in columns 7 and 8 of Table II can be attributed to
strips. A minimum air gap of 30 μm is realized using one layer

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

of polypropylene strip placed on sides of the CSRR, which is produces same results as obtained using (7).
effectively considered as a 30 μm air gap between the MUT
and the sensor. Similarly, two and three layers of
polypropylene strips are considered for the 60 μm and 90 μm
air gap, respectively. The plot of 𝑓𝑟−2 vs. 𝜀𝑟′ for this situation is
shown in Fig. 11 where it can be observed that the slope of the
curve decreases with an increase in air gap indicating decrease
in sensitivity.

Fig. 12. Comparison of real permittivity values obtained using the proposed
sensor from the air-gap model w.r.t. the ideal condition.

VII. CONCLUSION
In this paper, an attractive CSRR based planar sensor has
been presented for microwave characterization of the
dielectric materials with a single step measurement. The
Fig. 11. Plot of 𝑓𝑟−2 with real permittivity at different air gap ranging from 0 proposed sensor is operating in the frequency band ranging
to 100 μm. from 1.7 to 2.7 GHz. A numerical model has been developed
Based on the data presented in Fig. 11, a numerical model is for determining the relative permittivity and the loss factor of
derived here using the curve fitting technique for the real the material under test. The measurement has been performed
permittivity in term of the resonant frequency and the air gap for various standard samples and it has been found that the
ta as given in (12). measured results are in close agreement with the published
data. The importance of the proposed method can be
𝑓𝑟−2 − 0.11685 appreciated from the fact that the dielectric constant
𝜀𝑟′ (𝑚𝑜𝑑) = (−4.72182×10−5 )𝑡𝑎 +0.02045
(12) measurement is possible with a typical error of 3%. Further, a
numerical model has also been proposed for the accurate
It may be noted that now we have two numerical formulae measurement of the permittivity of materials having tolerable
(7) and (12) for calculation of the relative permittivity of air gap present between the sensor and the sample-surface
dielectric materials using the proposed CSRR based planar which makes the proposed method non-invasive. The
sensor, where (7) is developed for ideal case i.e. no air gap and proposed method is an ideal technique for microwave
(12) is presented here for the practical case which includes the characterization of the substrates commonly being used in
effect of air gap. The developed air gap model is tested with microwave planar circuits, as it is non-destructive and an
various dielectric samples having different amount of economical method to measure the complex permittivity.
intentionally developed air gap. The effect of newly derived
air gap model is studied by comparing the results obtained REFERENCES
using ideal case. The relative permittivity of MUT in both the
[1] M Shaji and M J Akhtar, “Microwave coplanar sensor system for
cases is calculated by putting the measured value of the detecting contamination in food products,” IEEE MTT-S International
resonant frequency in (7) and (12). For this analysis, the air Microwave and RF Conference (IMaRC), pp. 1-4, Dec. 2013.
gap is varied from 0 to 90 μm, however, the thickness of all [2] A K Jha and M J Akhtar, “Automated RF measurement system for
the samples are taken constant. For a fair comparison between detecting adulteration in edible fluids,” IEEE Applied Electromagnetics
both the proposed models, the results obtained from the Conference (AEMC), pp. 1-2, 18-20 Dec. 2013.
measured data are plotted in Fig. 12. [3] N. Celik, R. Gagarin, G. Huang, M. F. Iskander and B. Berg.
“Microwave stethoscope: development and benchmarking of a vital
Form above figure, it may be noted that the results obtained signs sensor using computer controlled phantoms and human studies,”
from (12) are almost independent of the air gap values as IEEE Trans. Biomed. Eng., vol. 61, no. 8, pp. 2341-2349, Aug. 2014.
compared to the results obtained from (7). It should be noted [4] M. S. Boybay and O. M. Ramahi., “Non-destructive thickness
measurement using quasi-static resonators,” IEEE Microw. Wireless
that the average variation in the results obtained using (7) is Compon. Lett., vol. 23, no. 4, pp. 217-219, Apr. 2013.
17% lesser than the results obtained from (12) which is a quite [5] F. H. Wee, P. J. Soh, A. H. M. Suhaizal, H. Nornikman and A. A. M.
significant deviation for a small air gap of 0.09 mm. It may be Ezanuddin, “Free space measurement technique on dielectric properties
interesting to note that in ideal case, the developed air gap of agricultural residues at microwave frequencies,” In Microwave and
Optoelectronics IEEE MTT-S International conference., pp. 183-187, 3-
model (12) of the proposed CSRR based planar sensor 6 Nov. 2009.

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2015.2469683, IEEE Sensors Journal

[6] D. K. Ghodgaonkar, V. V. Varadan and V. K. Varadan, “Free-space Abhishek Kumar Jha (S’14) received the
measurement of complex permittivity and complex permeability of B.E and M.Tech (Hons.) degrees in
magnetic materials at microwave frequencies,” IEEE Trans. Instrum. electronics and communication
Meas., vol. 39, no. 2, pp. 387-394, Apr. 1990. engineering from the University of
[7] Z. Akhter and M J Akhtar, “Time domain microwave technique for Burdwan, West Bengal, India, in 2009 and
dielectric imaging of multi-layered media,” Journal of Electromagnetic 2011, respectively. He is currently
Waves and Applications, vol. 29, no. 3, pp. 386-401, Jan. 2015. pursuing the Ph.D. degree in electrical
[8] M. Shete, M. Shaji and M. J. Akhtar, “Design of a coplanar sensor for engineering with the Indian Institute of
RF characterization of thin dielectric samples” IEEE Sensors J., vol.13, Technology Kanpur (IITK), Kanpur, India.
no.12, pp. 4706-4715, Dec. 2013. His research interests encompass the
numerical analysis and design of
[9] P. M. Narayanan, “Microstrip transmission line method for broadband microwave circuits and waveguide
permittivity measurement of dielectric substrates,” IEEE Trans. Microw. components, development of microwave
Theory Tech., vol. 62, no. 11, Nov. 2014.
sensors for nondestructive testing of dielectric and magnetic properties of
[10] A. K. Jha and M. J. Akhtar, “A generalized rectangular cavity approach materials, synthesis of advance composites for microwave absorbers and
for determination of complex permittivity of materials,” IEEE Trans. microwave heating applications etc. He has authored/coauthored
Instrum. Meas., vol. 63, no. 11, pp. 2632-2641, Nov. 2014. approximately 20 papers in peer reviewed international journals and
[11] M. S. Boybay and O. M. Ramahi, “Material characterization using conference proceedings.
complementary split-ring resonators” IEEE Trans. Instrum. Meas., vol. Mr. Jha founded and chaired the IEEE Microwave Theory and Techniques
61, no. 11, pp. 3039-3046, Nov. 2012. Society Student Branch Chapter IITK, Uttar Pradesh Section, India. He was a
[12] C. Lee and C. Yang, “Thickness and permittivity measurement in multi- Project Trainee with the Microwave Tube Research and Development Centre,
layered dielectric structures using complementary split ring resonators,” Defense Research and Development Organization, New Delhi, India, where
IEEE Sensors J., vol. 14, no. 3, pp. 695-700, March 2014. he was involved in medium power injection locked magnetron. He was an
[13] Chieh-Sen Lee and Chin-Lung Yang, “Complementary Split-Ring Assistant Professor with the Electronics & Communication Engineering
Resonators for Measuring Dielectric Constants and Loss Tangents,” Department, Seacom Engineering College, Howrah, India, from 2011 to 2012.
Microwave and Wireless Components Letters, IEEE, vol.24, no.8, Mr. Jha is the recipient of highly prestigious Graduate Fellowship Award in
pp.563-565, Aug. 2014. recognition of his academic achievement and excellence bestowed by IEEE
Microwave Theory and Techniques Society for 2015. He is also the recipient
[14] (2012), CST-Computer Simulation Technology [Online]. Available: of the University Gold Medal for being first in the first class of B.E and
http://www.cst.com. M.Tech degrees in 2009 and 2011, respectively.
[15] “Rfcafe\electronics\dielectric constant, strength and loss,” (2014)
[online].Available:http://www.rfcafe.com/references/electrical/dielectric M. Jaleel Akhtar (S’99–M’03–SM’09)
-constants-strengths.htm received the Ph.D/Dr. Ing. degree in
[16] “Engineeringtools\electricals\relativepermittivty-dielectricconstant,” Electrical Engineering from the Otto-von-
(2014) [online]. Available: http://www.engineeringtoolbox.com/relative- Guericke University of Magdeburg,
permittivity-d_1660.html Magdeburg, Germany in the year 2003.
[17] Z. Awang, F. A. M. Zaki, N. H. Baba, A. S. Zoolfakar and R. A. Bakar, From 1994 to 1997, Dr. Akhtar was a
“A free space method for complex permittivity measurment of bulk and Scientist with the Central Electronics
thin film dielectric at microwave frequencies,” Progress In Engineering Research Institute (CEERI),
Electromagnetics research B, vol. 51, pp. 307-328, 2013. Pilani, India, where he was mainly
involved with the design and development
[18] C. L. Feng, C. K. Ong, C. P. Neo, V. V. Varadan and V. K. Varadan,
“Microwave electronics: measurement and materials characterization,” of high power microwave tubes. From
John Wiley & Sons, Mar 2004. August 2003 to March 2009, he was a
Postdoctoral Research Scientist and a Project Leader with the Institute for
[19] M. D. Sindreu, J. Naqui, J. Bonache and F. Martin, “Split rings for Pulsed Power and Microwave Technology (IHM) at the Karlsruhe Institute of
metamaterial and microwave circuit design: a review of recent Technology (KIT), Karlsruhe, Germany, where he was involved with a
developments.” International Journal of RF and Microwave Computer number of projects in the field of microwave material processing. In March
Aided Engineering., vol 22, no. 4 pp 439-458, Jul 2012. 2009, he joined the Department of Electrical Engineering at the Indian
[20] A. Ebrahimi, W. Withayachumnankul, S. F. Al-Sarawi and D. Abbott, Institute of Technology Kanpur, India, where he is currently as Associate
“Dual-mode behavior of the complementary electric-LC resonators Professor.
loaded on transmission line: analysis and applications,” J. Appl. Phys., He is the author of two books, two book chapters, and has
vol. 116, no. 8, 083705, 2014. authored/coauthored more than 100 papers in various peer-reviewed
[21] C. Kai and L. H. Hsieh, “Microwave ring circuits and related international journals and conference proceedings. He also holds one patent
structures,” 2nd ed. John Wiley & Sons, 2004. on coplanar based RF Sensors. His current research interests include
[22] D. M. Pozar, “Microwave engineering,” 3rd ed. Danvers, MA Wiley, microwave imaging and nondestructive testing, RF sensors, electromagnetic
2005. modeling and testing of artificial dielectrics and metamaterials, UWB
antennas for imaging, and design of RF filters and components using the
[23] J. Walker, R. Resnick and D. Halliday, “Fundamentals of physics”,
Wiley, 2008. electromagnetic inverse scattering.
Dr. Akhtar is a Fellow of the Institution of Electronics and
Telecommunication Engineers, New Delhi, India, and a Life Member of the
M. Arif Hussain Ansari received his B.Tech Indian Physics Association and the Indo-French Technical Association. He is
degree in electronics and communication the recipient of the CST University publication award 2009 from the CST AG,
engineering from Magadh University Bodh Germany. He is serving as Chair of the IEEE MTT-S Uttar Pradesh Chapter,
Gaya, Bihar, India in 2012 and M.Tech degree and Vice-Chair of the IEEE Uttar Pradesh Section.
in Materials Science Programme from Indian
Institute of Technology Kanpur, India in 2015.
Mr. Ansari is a research associated at
Microwave Imaging and Material Testing
Laboratory, Department of Electrical
Engineering, Indian Institute of Technology
Kanpur.
His current research interests include microwave planar sensors for testing of
materials.

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.