Sensors and Actuators A 239 (2016) 153–160

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

journal homepage: www.elsevier.com/locate/sna

A new way to find dielectric properties of liquid sample using the

quartz crystal resonator (QCR)
Luis Armando Carvajal Ahumada a,b,c,∗ , Nuria Peña Pérez d ,
Oscar Leonardo Herrera Sandoval c,e , Francisco del Pozo Guerrero a,b ,
José Javier Serrano Olmedo a,b
a
Center for Biomedical Technology (CTB), Technical University of Madrid (UPM), 28040 Madrid, Spain
b
Networking Research Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Spain
c
Research and Technological Development Centre of Electrical, Electronic and ICT Industry (CIDEI), Bogotá, Colombia
d
Carlos III University, Madrid, Spain
e
Central University. Bogotá, Colombia

a r t i c l e i n f o a b s t r a c t

Article history: The main objective of this article is to demonstrate by performing experimental measurements how the
Received 1 September 2015 equivalent capacitance C0 changes when a fluid sample such as water is in contact with the crystal and to
Received in revised form 7 December 2015 relate this change with the relative permittivity of the fluid. These measurements were compared with
Accepted 11 January 2016
simulations of traditional models like Butterworth–Van Dyke (BVD). To obtain the change of C0 when the
Available online 14 January 2016
crystal is in contact with water, the relation between the series resonance frequency (fs ) and frequency
at minimum impedance with zero phase (fr ) is used. To know these values, a new way of finding principal
Keywords:
parameters in quartz crystal resonator (QCR) sensors, such as series resonant frequency (fs ), half band
QCR
Susceptance half width ( ) and maximum peak of conductance (Gm ), is proposed; both for an unperturbed crystal and
Conductance a crystal loaded with a liquid. The method consists in measuring the current (I) that flows through the
Resonance frequency crystal and the voltage (V) between electrodes at frequency values near to resonance (sweep frequency).
Impedance analysis Additionally, the susceptance |B| of the crystal is also measured, multiplying the 90 degrees shifted current
of the crystal and its voltage, using a mixer. The DC component of this operation is proportional to the
susceptance of the crystal. With the magnitudes of the admittance and susceptance, the real value of the
conductance |G| is obtained for each frequency value in the sweep. The conductance and susceptance
curves were fitted with a summation of Gaussian and sine functions respectively with the minor RMSE
possible.
The proposed method has been compared with simulations done in COMSOL Multiphysics in order to
verify the experimental results with simulation data. MATLAB curve fitting toolbox was used to fit the
experimental curves.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction different investigators who have obtained theoretical models that

TSM sensors and particularly the quartz crystal resonator (QCR)
sensors are considered highly efficient sensing systems not only
because of their low manufacturing cost but also because of their
high accuracy, bio-functionalization capacity, sensitivity and relia-
bility in measuring the deposited sample, either gaseous or liquid
media. So, QCR have been widely studied for several decades by
∗ Corresponding author at: Center for Biomedical Technology (CTB), Technical
University of Madrid (UPM), 28040 Madrid, Spain.
E-mail addresses: luis.carvajal@ctb.upm.es, la.carvajal@alumnos.upm.es
(L.A. Carvajal Ahumada).
explain the behavior of the sensor, especially in the frequency
ranges close to the resonance frequency [1–5]. A model that stands
out for its use and ease of understanding is that proposed by
Butterworth–Van Dyke (BVD) [2] which is derived from the trans-
mission network model proposed by Mason [1,2,6]. The BVD model
is especially used in many publications to support the results
because of the electrical representation that provides of the physi-
cal behavior of the sensor [3,7–10].
However, the results obtained from the performance of the sen-
sor during our experimentation and from the existing simulation
models, based on the equations of BVD basic electric model, do not
always correspond so well when the sensor is exposed to fluids like
water. This is due to the fact that small changes between series res-

http://dx.doi.org/10.1016/j.sna.2016.01.021
0924-4247/© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.
0/).
154 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160

onance frequency (fs ) and frequency at minimum impedance with

zero phase (fr ) are not taken into account in the basic model.
Within the BVD model, the static capacitance C0 models losses
caused by external electrical components to the sensor and the
parasitic capacitance generated by the electrodes on the crystal
[9,10]. This capacitance is to be significant for values far from the
resonance frequency and negligible when the crystal is operating
at values close to the resonance frequency [7]. Furthermore, this
value is considered static and dependent only on the physical char-
acteristics of the crystal, so in many experiments is eliminated
externally [9,10]. However in practice, the results of the exper-
iments expose that this capacitance value depends also on the
properties of sample, so it is important to take it into account.
This article aims to demonstrate how the capacitance C0 changes Fig. 1. Change in morphology of the conductance curve for bare crystal and in
when a fluid sample is in contact with the crystal by experimen- contact with a liquid.
tal measures and comparison with simulations models developed
using COMSOL, and how this C0 change generates additional infor-
mation about the sample. The BVD model, when the change in C0
is taken into account, is closer to experimental results.

2. Theory and methods

Fig. 2. Schematic of BVD (Butterworth–Van Dyke) model for quartz crystal res-
onator.
The quartz crystal resonator (QCR) is traditionally considered to
be a mass sensitive sensor that generates response (changes its res-
onant frequency) to different thin film samples or liquids in contact is useful to determine the viscoelastic properties of the fluid. Both
with its surface. For a thin film mass of the order of nanograms, the values are related through the quality factor Q:
crystal response will be of the order of Hertz according to Sauerbrey Fs
Q = (5)
Equations [4]: 2
2 According to theoretical models, when the QCR electrode is in
2f
f = − 0 m (1) contact with a sample, the morphology of the conductance curve
A  q q changes as it is shown in Fig. 1, shifting it to the left on the
frequency axis, increasing its bandwidth and decreasing its mag-
n
 
q ⁄ nitude. The electrical model BVD explains this phenomenon by
f0 = q (2)
2hq adding impedance that represents the characteristics of the sample.
For liquids in contact with the crystal, the equivalent impedance is
where q and q are the specific density and the shear modulus the series of an inductance and a resistance [10–12] (See Fig. 2). The
in quartz, respectively, f0 is the fundamental resonant frequency values of these impedances are related with the characteristics of
in quartz, related to its thickness hq , m is the thin film mass the samples according with the Eqs. (6)–(11) [1,2,7,10,12]:
deposited A is the piezoelectrically active crystal area and n is the
overtone number. 8e26 2 A
Cq = (6)
In contact with liquids, the crystal is capable of giving informa- (N )2 C 66 hq
tion about the density-viscosity product ()1/2 [1,11] of the fluid
1
by changing its resonant frequency and quality Q-factor according Lq = (7)
ωs 2 C q
with Kanazawa equations [5]:
 q
√ Rq = (8)
L L C 66 C q
f = − nf 0 3/2 (3)
q q  
ωs L q 2L L
 Lw =
N
(9)
2L ωs C 66 q
ı= (4)  
ωL
ωs L q 2ωs L L
Rw = (10)
N C 66 q
where L and L are the density and viscosity in fluid respectively.
Eq. (4) shows the decay characteristic length (ı) as proportional to ε22 Ae
the ratio viscosity to density of the liquid and as inversely propor- C0 = (11)
hq
tional to the angular frequency (ω) [12]. For 10 MHz AT cut quartz
crystal this length is: ı = 178 nm. where, C66 is the piezoelectrically stiffened elastic constant for
Additional to Sauerbrey and Kanazawa equations, the quartz loss-less quartz (2947 × 1010 N/m2 ), e26 is the piezoelectric stress
crystal comportment can be seen as an electrical resonator circuit constant for quartz (953 × 103 A s/m2 ), ε22 is the quartz permittiv-
[2]. There are some electrical models like Mason or BVD that explain ity (3.982 × 10−11 A2 s4 /kg m−3 ), A is the active electrode area (m2 ),
the crystal behavior near to the resonance frequency. hq is the quartz crystal thickness (m), q effective viscosity of quartz
Fig. 1 shows the equivalent conductance comportment (Geq ) for (3.5 × 10−4 kg/m s), ωs is the series resonance frequency (2␲Fs ). For
unperturbed crystal (kept in air) and for the top face of the crys- the liquid, L and L are the density of liquid sample (kg/m3 ) and
tal immersed in liquid. It is important to note that the maximum the viscosity of the liquid sample (kg/m s) respectively [3].
value takes place at resonance frequency [3,13]. The bandwidth The model shown in Fig. 2 explains the behavior of QCR In con-
shift ( ) gives essential additional information, for example, it tact with a sample of liquid. In this theoretical model, the maximum
 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160 155

Fig. 3. Circuit schematic sensor system containing an oscillator (10 MHz), a mixer chip (AD835), a toroidal transformer used as current sensor and conditioning stage for
signals.

value of the conductance matches with the zero susceptance value
independently of the properties of the sample. In other words, at
resonance, the behavior of quartz crystal is purely resistive. Even,
C0 is considered negligible so is compensated externally with other
capacitor and according with Eq. (11) its value only depends of
quartz crystal parameters.
However, in the obtained experimental results, the zero-
crossing point of the susceptance curve and the maximum peak
and secondary coils (1:1). The voltage signal (V2 ) related with the
current through the crystal (I1 ) and the voltage used to stimulate
the crystal (V1 ), are multiplied in a second mixer AD835 according
to Fig. 3. The signal resulting is filtered using an active filter with
LT1057 Amplifier to obtain the DC component. The DC component
is related with the susceptance of quartz crystal according with the
following equations:

of the conductance mismatch when the crystal is in contact with V1 = Vp cos ω1 t + 1 (12)
liquid. This phenomenon occurs because C0 value also depends on

I1 = Ip cos ω2 t + 2 (13)
the properties of the sample. In other words, in the electrical model
BVD, the C0 parameter is not negligible and may even provide addi-

V2 = kIp cos ω2 t + 2 + ⁄2 (14)

tional information about the sample, more specifically, about the
electrical permittivity of the liquid.
The proposed methodology to characterize the behavior of QCR
consists of two stages: the first one involves the design and con-
struction of a circuit to stimulate and acquire the peak values of
voltage, current and susceptance signals in the quartz crystal. The
second stage involves processing data so as to obtain the resonance
where k is a proportional constant due to the relation between
the coils in the toroidal transformer, ω1 and ω2 are the angular
frequencies of the signals and  1 and  2 are the phases of the signals.
Based on the theory of the mixer of signals [15], the result of the
multiplication can be written as:


k
frequency (fs ) value, the frequency at minimum impedance with V1 V2 = Vp Ip cos (ω1 + ω2 ) t + 1 + 2 + ⁄2
2
zero phase (fr ) value, the half band half width ( ) value, maximum
peak of conductance (Gm ) related to the value of the motional resis-
tance (Rq = 1/Gmax ) and the C0 value related with sample dielectric
properties.
2.1. Stage 1—acquisition data
+cos (ω1 − ω2 ) t + 1 − 2 − ⁄2
As ω1 = ω2 because the signal of voltage on the crystal and the
signal of current through the crystal have the same frequency, the
Eq. (15) can be rewritten as:

The proposed circuit obtains the voltage and current values for k 


V1 V2 = Vp Ip cos 2ω1 t + 1 + 2 + ⁄2 + cos 1 − 2 − ⁄2 (16)
several frequencies in a configurable bandwidth near to the crystal 2
resonance frequency (fc ). This circuit is a variation of Nakamoto and Then, by using a low pass filter, the high frequency component
Kobayashi proposal [14]. is removed:
The circuit consists of an oscillator chip that generates a square
k 

signal of 10 MHz of frequency; its signal is filtered with a pas- V1 V2 = Vp Ip cos 1 − 2 − ⁄2 (17)
sive band-pass filter to obtain the fundamental frequency of the 2
signal. The filtered signal (sinusoidal wave) is multiplied with a The following trigonometric identity is used to get the final
sinusoidal DAQ signal using a mixer (AD835AN – Analog Devices). equation:
The resulting signal is filtered and used to stimulate the quartz crys-

tal. The frequency of the resulting signal is controlled changing the Sin (A) = Cos A − ⁄2 (18)
frequency of DAQ signal.
k 

The quartz crystal is placed in series with a toroidal-core trans- V1 V2 = Vp Ip sin 1 − 2 (19)
2
former used as current sensor. The current is detected using the
toroidal transformer. The secondary coil of the transformer gives Since the relation in the toroidal transformer is 1:1, the constant
a voltage proportional to the current in the primary which is the k ≈ 1 so:
crystal current. The voltage is shifted 90 degrees with respect to the vp vp 

Mixer output (DC component)= sin 1 − 2 ≈ |B| (20)
original current signal. The number of turns is the same for primary 2
156 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160

It is important to note that the signal obtained is equivalent

to the susceptance because the constant term (VI/2) is modulated
by the sine of the difference in the phase of the signals. In con-
sequence, the signal obtained reproduces the behavior capacitive
and inductive of the piezoelectric crystal in frequency. The zero-
crossing occurs when the voltage on the crystal and the current
through the crystal are in phase. When 0 <  1 −  2 < ␲, the behavior
of the crystal is inductive and ␲ <  1 −  2 < 2␲ the behavior of the
crystal is capacitive.
On the other hand, a high-speed amplifier (AD8010ANZ-Analog
Devices), is used together with a rectifier system and passive fil-
ters to obtain the peak of the signals of voltage (V out) and current
(I out) with a DAQ acquisition system.
Fig. 4. Adjustment of the conductance and susceptance curves in Gaussian and sum
The values of the current, voltage and susceptance signals are
of sine functions respectively.
acquired in real time from a computer using the DAQ and Labview.
Reading speed depends on the change in frequency sweep; the tests
were performed with changes of 1 Hz.
The 10 MHz quartz crystal purchased from International Crystal
Manufacturing, For DAQ generator and DAQ acquisition system the
multifunction DAQ of National Instruments NI USB-6351 was used.

2.2. Stage 2—processing data

After acquiring voltage, current and susceptance data for each

frequency value in Labview, these are exported to Matlab where
an analysis of the behavior of the crystal is performed. The conduc-
tance magnitude curve |G| is calculated using Eq. (21). The baseline
in conductance and susceptance are removed to fit them [3].


Fig. 5. Curves for bare crystal. Magnitude of conductance |G| (green), magnitude of
susceptance |B| (blue), admittance phase curve (red).
|G| = |Y|2 − |B|2 (21)

The conductance and susceptance data are adjusted to a sum-

mation of Gaussian functions and to a summation of sine functions
respectively. In both cases, adjustments are made with the least
mean square error (RMSE) possible. The fitting is achieved using
the Matlab Curve Fitting tool.
Summation of Gaussian functions:

2  
2  

f −b 1 f −b 2 f −bn 2
− c1 − c2 − cn
G (f ) = a1 e + a2 e + . . . + an e

Summation of sine functions:

B (f ) = a1 sin (b1 × F + c 1 ) + a2 sin (b2 × F + c 2 )

+ . . . + an sin (bn × F + c n )
The coefficients of the conductance and susceptance curves after
fitting them are registered but not analyzed since the curves param-
eters are different for each quartz crystal and the coefficients do
not give additional information. When the crystal is in contact
with liquid, the coefficients also change because the morphology
of the curves changes too. The number of terms in the summation
of Gaussian functions and in the summation of sine function can
also change, depending on the quality factor of the quartz crystal
and the liquid sample properties. However, is important to adjust
the curves for minimizing the noise in the acquired signals and
to obtain the first derivative of conductance curve. The parameter
used to select the appropriate fitting function for conductance and
susceptance curves is the RMSE (Root Mean Squared Error).
Fig. 4 shows the fitting of the conductance and susceptance
curves. In general, the fitting of the conductance curve is more accu-
rate than the fitting of the susceptance curve. However, in the zone
of interest (Frequency range between the maximum and minimum
value of the susceptance curve), both fitted curves are really similar
to the original data. After adjusting the susceptance and conduc-
tance curves, the admittance phase is obtained as reference to shift
Fig. 6. Curves for crystal in contact with water. Magnitude of conductance |G|
(green), magnitude of susceptance |B| (blue), admittance phase curve (red), first
derivative of |G| curve (Turquoise).
the susceptance across the frequency-axis. The admittance phase
is obtained with the following equation:
 |B| 
= tan−1 (22)
|G|
The same acquisition and processing steps were realized for the
unperturbed crystal and for the crystal loaded with water.
Fig. 5 shows that the crossing between susceptance and admit-
tance phase curves corresponds with the conductance peak for the
unperturbed crystal. In other words, the series resonance frequency
and frequency at minimum impedance with zero phase are the
same for the unperturbed crystal (the vertical line drawn in black
was set manually to show the phenomenon)
Fig. 6 shows how the water sample shifts the susceptance curve
to the right so the crossing between the susceptance and the admit-
tance phase does not match with the peak of the conductance. This
phenomenon occurs because the water sample generates an incre-
ment of C0 which is usually negligible. In other words, the series
 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160 157

Fig. 7. Mesh of crystal geometry in COMSOL multiphysics. The enlarged image corresponds to the mesh in thickness axis.

resonance frequency and the frequency at minimum impedance

with zero phase mismatch. To demonstrate this phenomenon, the
first derivative of the conductance curve has been calculated and
its zero-crossing point is compared with the zero crossing of the
susceptance curve and admittance phase. Additionally, two verti-
cal black lines were placed to graphically show the difference in the
values at zero crossing.
To calculate the increase in C0 , the relationship between the
series resonance frequency (intercept of the first derivative of the
conductance curve with the frequency axis) and the frequency at
minimum impedance with zero phase (intercept of the susceptance
curve with the frequency/horizontal axis) is used [16].
 1

fr = fs 1+ (23)
2QM
Where M is the crystal figure-of-merit and Q is the crystal quality Fig. 8. The total displacement in X axis (vibration axis) at resonance frequency for
factor: quartz crystal. As expected, the major vibration is in electrode center and decreases
it with distance to periphery.
1
M= (24)
ws Rq C 0
1 a voltage between the electrodes generates a shear deformation of
Q = (25) the crystal making it sensitive to the contact with a deposited mass
ws Rq C q
or liquid.
Solving the Eqs. (23)–(25) for C0 when the crystal is in contact Since the relationship between the electrodes thickness
with water: <100 nm and the crystal thickness 167 ␮m is really small, it is nec-

 
2 Lq + Lw essary to make a very fine mesh to obtain good results, but this
∗ fr
C0 =
2 −1 (26) implies more computational recourses [22]. For this reason, it is
Rq + R w fs
better not to create the electrodes with thickness into the geom-
etry and assume they have the same thickness as the total crystal
Eq. (26) shows the total capacitance C0 *, it is the sum of capaci-
area but with a different equivalent density using the following
tance due to the electrodes and external cables (traditional C0 ) and
equation:
the change generated by water sample. To know the change in C0 *,
the component due to the electrodes shown in Eq. (11) is subtracted   2
dau × hau
Eq. (27). The values with subscript (q) refer to the quartz crystal and d electrodes = d q × 1+ (28)
values with subscript (w) refer to water as sample. d q × hq

Cw = C0∗ − C0 (27) where, delectrodes is the equivalent density of electrodes, dq is the

2.3. Simulation models
The finite element analysis (FEA) method is a useful technique
that can be used to accurately model the characteristics of com-
plex structures, as well as to optimize the design parameters for
microsystems [17,18]. The software package COMSOL Multiphysics
allows coupling electric and mechanic structures and it is used to
simulate the Quartz Crystal behavior in a 3D model [19,20] and the
BVD electric model
2.3.1. 3D model
The 10 MHz Quartz Crystal has been modeled with two elec-
trodes, the crystal diameter is 13.66 mm, the electrode diameter is
5.1 mm and the crystal Thickness is 167 ␮m. Since the AT cut Quartz
Crystal forms 35.35 degrees with the Z-axis [21], the application of
density of quartz crystal (2651 kg/m3 ), dau is the density of gold
(1930 kg/m3 ), hq is the crystal thickness, hau is the electrode thick-
ness. Solving the equation, the equivalent density of the zone of
electrodes is 2674 kg/m3 [22].
The applied voltage was 1 V and a damping coefficient of 2% was
used [21].
Considering the QCR is a Thickness Shear Mode (TSM) sensor, the
mesh in the thickness direction is a principal variable to obtain good
simulation accuracy. The Fig. 7 shows the mesh of the crystal. The
fundamental frequency value was stable when the number of mesh
layers in the thickness was more than 18 [20]. The fluid is modeled
like a cylinder over the electrode of the crystal and it is meshed the
same way. Fig. 8 shows the total displacement in X axis, the major
vibration is in electrode center and decreases it with distance to
periphery. In consequence, the major sensibility is in the center of
electrode.
158 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160

Table 1 Table 3
Measures for unperturbed crystal. Measures for crystal in contact with water.

Experiment Rq (ohm) Fs (Hz) Rho (Hz) Experiment Rq (ohm) Fs (Hz) Rho (Hz) Cw (pF)

Mean 14.15 9985755.9 365.5 Mean 202.2 9983738 2904.8 7.9

Std. desv 0.4 7.4 3.9 Std. desv 1.9 14.2 97.1 0.8

2.3.2. BVD model The resonance frequency of the unperturbed crys-

The electric model library in COMSOL is used for modeling the tal is Fs = 9985755.9 ± 7.4 Hz. The motional resistance is
electric BVD model for quartz crystal. Solving the Eqs. (6)–(11) for a
10 MHz crystal with thickness of 167 ␮m, the following values are
obtained:
䊉F s = 9984729 Hz
䊉C q = 3.159 × 10−14 F
䊉Lq = 8.043 × 10−3 H
䊉Rq = 0.378 ˝
䊉C 0 = 4.88 × 10−12 F
䊉Lw = 3.254 × 10−6 H
䊉Rw = 204.140 ˝
The equations used for calculating the resonance frequency and
the half band half width using equivalent electrical parameters are:
1
Fs =  (29)
2 Lq C q
Rq
 =
4Lq
For measures with water as sample, Lq and Rq are changed
by (Lq + Lw ) and (Rq + Rw ) respectively, according to Kirchhoff laws
[17].
Initially, the static capacitance C0 was calculated using Eq. (11)
to simulate the BVD model. In a second step, the theoretical value
of C0 was changed by the obtained experimental value (C0 *) in the
BVD2 model and the simulation was performed again with this
modified parameter to compare both results.
3. Results and discussion
3.1. Unperturbed crystal
Following the procedure described in acquisition and process-
ing data sections the principal variables of interest are obtained in
experimental measures with an unperturbed crystal. Ten measure-
ments were performed at intervals of 10 min with the intention
of validating the reproducibility of the data and each data read-
ing took 15 min for covering all frequency intervals near to the
resonance frequency (10,000 point). The results of the mean and
standard deviation for the ten measures are shown in Table 1.
Rq = 14.15 ± 0.4  and the value of the half band half width
 = 365.5 ± 3.9 Hz.
According to Cassiède et al. [3], the dissipative processes occur-
ring in a quartz resonator appear to be the sum of the internal
friction inside the crystal and the resistive losses due to the exper-
imental assembly, including electrodes and electrical connexions.
For this reason, a “BVD model 2” has been calculated with the same
parameters that BVD model, except its Rq value. The new Rq value is
obtained from the experimental results because the experimental
Rq contains the internal losses of the crystal and the external losses
caused by the acquisition data electronic circuit.
Additionally, the goal of taking into account a second BVD mod-
ified model is to show that with some experimental measures it is
possible refine the model and generate simulation data closer to
reality.
Table 2 summarizes the comparison between the experimental
data results with the 3D model, the BVD model and a second “BVD
model 2” for an unperturbed crystal.
The 3D model and the BVD model results have close values. The
3D model has a better Quality factor because it does not contem-
plate external losses caused by the electronic oscillator and the
acquisition system. The BVD model was calculated using theoret-
ical equations and its results are very similar to the 3D model.
As expected, the BVD model 2 is closer to the experimental data
(30) than the other models because those models do not contemplate
external losses caused by the acquisition data electronic circuit (the
resonance frequency does not depend on the motional resistance
value but the quality factor and the half band half width depend on
such value).
3.2. Crystal in contact with water
After calculating the variables of interest for the unperturbed
crystal, a sample of water Mili-Q was deposited on the electrode
of the crystal. Following the same procedure than the used for
the unperturbed crystal; voltage, current and susceptance data for
water samples were acquired.
In the same way than for the unperturbed crystal, a mea-
surement was taken every 10 min,; each data reading took
30 min,. In this case, the time for data reading is greater because
the frequency interval increases to get all curves since the
morphology of the curves changes, losing amplitude and increasing
its bandwidth. In consequence, it is necessary to take 20,000 points.
The procedure was performed ten times; the results are show in
Table 3.
The resonance frequency of the crystal with a water sam-
ple is fs = 9983738 ± 14.2 Hz. The equivalent motional resistance

Table 2
Results for bare crystal. BVD model 2 has the same BVD model parameters to except Rq correction.

BARE crystal

Indicator Experimental 3D model BVD model BVD model 2

Motional resistance () 14.15 ± 0,4 0.19 0,38 14.15
Frequency (Hz) 9985755.9 ± 7.4 9850699 9984729 9984729
Half band half width (Hz) 365.5 ± 3.9 2.3 3.7 140
Quality factor 13661.90 2141456 1334856 35653.1
 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160 159

Table 4
Results for crystal in contact with water. BVD model 2 has the same BVD model parameters to except Rq and C0 * corrections.

Water sample

Indicator Experimental 3D model BVD model BVD model 2

Motional resistance equivalent () 202.2 ± 1.9 208 204.5 218
Frequency (Hz) 9983738 ± 14.2 9848654 9982712 9982712
Half band half width (Hz) 2904 ± 97.1 1150 2022 2160
Quality factor 1719 4282 2468 2310

Table 5
Comparative between simulation models and experimental results for F and  .

3D model BVD model BVD model 2 Experimental results %Error 3D model %Error BVD model %Error BVD model2

F −2045 −2017 −2017 −2017 1.37 0.2 0.1

 +1148 +2018 +2020 +2539 31.24 20.5 20,4

Rq + RL = 202.2 ± 1.9  and the value of the half band half width Rho-
 = 2904.8 ± 97.1 Hz. The capacitance obtained due to the water
sample is Cw = 7.9 ± 0.8 pF.
The capacitance obtained due to the water sample (Cw ) is
very similar to the value got by Rodahl et al. in 1996 [23].
The value obtained by Rodahl was Cw = 7.7pF for non-conductive
water.
Using the equation proposed by Rodahl et al. (Cw = 0.095
r pF),
the relative dielectric constant for the water sample obtained is
r = 83.47. This value is very close to the official value for Milli-Q
water which is 81.
Table 4 shows the data of the 3D model and BVD model simu-
lation comparing them with the experimental results. In this case,
like for the Unperturbed Crystal, a “BVD model 2” is calculated by
changing the value of the theoretical Motional Resistance by the
value of the Motional Resistance obtained experimentally. Addi-
tionally, the C0 value is replaced too by C0 *, which is obtained from
the measures.
All variables of interest are very similar and save coherence with
the expectative. The multiphysics simulation generates better Q-
factor because it does not contain external losses, BVD model 2
with Rq and C0 * corrections are closer to the experimental results.
To calculate F and  the procedure described in the theory
section is used.
The results presented in Table 5, are obtained by subtracting the
series resonance frequencies (fs ) and the half band half width ( )
for the unperturbed crystal and for the crystal in contact with a
liquid, for the simulation model and experimental results respec-
tively.
F = Fcontactwithliquid − Funperturbedcrystal
 =  contactwithliquid −  unperturbedcrystal
The results of the 3D model differ from the other results because
the model does not consider any external loss, for example: the
external electrical circuit or the crystal manufacturing or its assem-
bly in the sensing system.
The models BVD, BVD 2 and the experimental results are very
close in F. This is consistent because the adjustments made in BVD
model 2 do not affect the value of the resonant frequency and, in
general, the BVD model gives a good representation of the behavior
of the crystal. The half band half width value in BVD model 2 is a
little closer to experimental result than the value in the BVD model.
However the difference is less significant than for the values shown
in Tables 2 and 4.
Finally, the value of the viscosity of the water was obtained
with the aim of validating the procedure described above and
the values obtained in Tables 2, 4 and 5. The utilized values
were: n = 1, f0 = 9985755.9 Hz, L = 1000 kg/m3 , L = 2651 kg/m3 ,
Gq = 2.93 × 1010 N/m2 , F = 2017 Hz.
Through the Kanazawa Eq. (3), the viscosity of water obtained is
L = 9.98E–4 Pa s. The official value of the water viscosity is 1E3 Pa s
measured with the viscometer Fungilab (alpha series). Is impor-
tant to highlight that the quantity of fluid used to calculate the
viscosity using the viscometer was 20 mL, while the quantity of
fluid deposited on the QCR was only 20 ␮l. The relative error in the
measure was: 0.2%.
4. Conclusions
The model BVD is appropriated to obtain values to viscosity in
samples with known density. However, if one considers that the C0
value also depends on the fluid sample, it is possible with the QCR
used in liquids, to know additional properties of samples, differ-
ent to viscosity or density, characterizing not only the conductance
curve but also the susceptance and phase curves of the crystal.
The relation between the series resonance frequency and the
frequency at minimum impedance with zero phase when the crys-
tal is in contact with a fluid sample is directly related with the
change in capacitance C0 . The change in the capacitance C0 is asso-
ciated with the dielectric properties of sample.
The proposed system allows knowing the variables of inter-
est to characterize the unperturbed crystal and the crystal loaded
with liquids. It also allows relating quantitatively the mismatch
between series resonance frequency and zero phase frequency with
the dielectric effect of the non-conductive water. Additionally, the
required quantities of the sample for the proposed system are in
order of microliters, while the traditional methods need milliliters
of fluid to calculate the viscosity (viscometer).
In future works, it is expected to improve the sensor design to
characterize viscosity and dielectric properties in different biolog-
ical fluids. Additionally, it is aimed to perform tests with biological
cellcultures to obtain results of interest in the determination of
mechanical and electrical properties of cells during its growth.
Acknowledgements
The authors thank Research and Technological Development
Centre of electrical, electronical and ICT industry (CIDEI) for its par-
ticipation with ideas and developing of prototype system. We also
thank participants of Bioinstrumentation and Nanomedicine group
in Center for Biomedical Technology (CTB) for its recommenda-
tions and loan of equipment. We thank COLCIENCIAS for supporting
this work through the grant No. 0375 – 2013 with number project
6570577636375 and Universidad Central (Colombia) for support-
ing this work through the Research Cluster on Converging Sciences
and Technologies (NBIC).
160 L.A. Carvajal Ahumada et al. / Sensors and Actuators A 239 (2016) 153–160
References
[1] R.W. Cernosek, S.J. Martin, A.R. Hillman, H.L. Bandey, Comparison of
lumped-element and transmission-line models for thickness-shear-mode
quartz resonator sensors, IEEE Tans. Frequency Control Symposium, 1997,
Proceedings of the 1997 IEEE International 45 (1998) 1399–1407, http://dx.
doi.org/10.1109/58.726468.
[2] V.E. Granstaff, S.J. Martin, Characterization of a thickness-shear mode quartz
resonator with multiple nonpiezoelectric layers, J. Appl. Phys. 75 (1994),
http://dx.doi.org/10.1063/1.356410, 1319.
[3] M. Cassiède, J.H. Paillol, J. Pauly, J.-L. Daridon, Electrical behaviour of AT-cut
quartz crystal resonators as a function of overtone number, Sens. Actuators A
Phys. 159 (2010) 174–183, http://dx.doi.org/10.1016/j.sna.2010.03.028.
[4] G. Sauerbrey, Verwendung von Schwingquarzen zur Wägung dünner
Schichten und zur Mikrowägung, Zeitschrift Für Phys. 155 (1959) 206–222,
http://dx.doi.org/10.1007/BF01337937.
[5] K. Keiji Kanazawa, J.G. Gordon, The oscillation frequency of a quartz resonator
in contact with liquid, Anal. Chim. Acta 175 (1985) 99–105, http://dx.doi.org/
10.1016/S0003-2670(00) 82721-X.
[6] S. Sherrit, Use of piezoelectric resonators for the characterization of
mechanical properties of polymers, Proceedings of SPIE—The International
Society for Optical Engineering (2002) 262–276, http://dx.doi.org/10.1117/12.
475175.
[7] G.H. Kim, G. a, S. Rand, V. Letcher, Impedance characterization of a
piezoelectric immunosensor. Part I: antibody coating and buffer solution,
Biosens. Bioelectron. 18 (2003) 83–89 http://www.ncbi.nlm.nih.gov/pubmed/
12445448.
[8] C. Barnes, Development of quartz crystal oscillators for under-liquid sensing,
Sens. Actuators A Phys. 29 (1991) 59–69, http://dx.doi.org/10.1016/0924-
4247(91)80032-K.
[9] K.C. Vavra, G. Yu, M. Josowicz, J. Janata, Magnetic quartz crystal microbalance:
magneto-acoustic parameters, J. Appl. Phys. 110 (2011) 01395, http://dx.doi.
org/10.1063/1.3602998.
[10] N.-J. Cho, J.N. D’Amour, J. Stalgren, W. Knoll, K. Kanazawa, C.W. Frank, Quartz
resonator signatures under Newtonian liquid loading for initial instrument
check, J. Colloid Interface Sci. 315 (2007) 248–254, http://dx.doi.org/10.1016/
j.jcis.2007.06.020.
[11] B. Jakoby, G. Art, J. Bastemeijer, Novel analog readout electronics for
microacoustic thickness shear-mode sensors, Sens. J. IEEE 5 (2005)
1106–1111, http://dx.doi.org/10.1109/JSEN.2005.844330.
[12] G.N.M. Ferreira, A.-C. da-Silva, B. Tomé, Acoustic wave biosensors: physical
models and biological applications of quartz crystal microbalance, Trends
Biotechnol. 27 (2009) 689–697, http://dx.doi.org/10.1016/j.tibtech.2009.09.
003.
[13] M. Cassiède, J.-L. Daridon, J.H. Paillol, J. Pauly, Characterization of the
behaviour of a quartz crystal resonator fully immersed in a Newtonian liquid
by impedance analysis, Sens. Actuators A Phys. 167 (2011) 317–326, http://
dx.doi.org/10.1016/j.sna.2011.03.002.
[14] T. Nakamoto, T. Kobayashi, Development of circuit for measuring both Q
variation and resonant frequency shift of quartz crystal microbalance,
Frequency Control Symposium, 1997., Proceedings of the 1997 IEEE
International 41 (1994) 806–811, http://dx.doi.org/10.1109/58.330261.
[15] P. Bretcho, R. Ludwig, RF Circuit Design—Theory and Applications, Prentice
Hall, 2000, pp. 201–253.
[16] STATEK Corporation, The quartz crystal model and its frequencies,TECHNICAL
NOTE 32, Rev. A, pag, 1-9. www.statek.com/pdf/tn32.pdf.
[17] F. Lu, H.P. Lee, P. Lu, S.P. Lim, Finite element analysis of interference for the
laterally coupled quartz crystal microbalances, Sens. Actuators A Phys. 119
(2005) 90–99, http://dx.doi.org/10.1016/j.sna.2004.09.013.
[18] P. Castro, P. Resa, L. Elvira, Apparent negative mass in QCM sensors due to
punctual rigid loading, IOP Conf. Ser.: Mater. Sci. Eng. 42 (2012) 01206, http://
dx.doi.org/10.1088/1757-899X/42/1/012046.
[19] F. Liu, F. Li, A.N. Nordin, I. Voiculescu, A novel cell-based hybrid acoustic wave
biosensor with impedimetric sensing capabilities, Sensors (Basel) 13 (2013)
3039–3055, http://dx.doi.org/10.3390/s130303039.
[20] J. Ji, H. Oigawa, H.H. Chen, M. Zhao, T. Ueda, Size optimization for high
frequency quartz resonator using finite element vibration analysis, 2011, IEEE
Sens. Proc. (2011) 1652–1655, http://dx.doi.org/10.1109/ICSENS.2011.
6126974.
[21] S.K.R.S. Sankaranarayanan, R. Singh, V. Bhethanabotla, Computational design
of quartz crystal nanobalance for uniform sensitivity distribution, in: 2010
IEEE Sensors, IEEE, 2010, http://dx.doi.org/10.1109/ICSENS.2010.5690037.
[22] H. Sun, P. Lu, P. Zhang, H. Chen, Dynamic analysis of AT-cut quartz resonators
with ANSYS, Sensors, 2004. Proceedings of IEEE (2004) 95–98 (accessed 15.1.
15) http://ieeexplore.ieee.org/xpls/abs all.jsp?arnumber=1426108.
[23] M. Rodahl, F. Höök, B. Kasemo, QCM operation in liquids: an explanation of
measured variations in frequency and Q factor with liquid conductivity, Anal.
Chem. 68 (1996) 2219–2227, http://dx.doi.org/10.1021/ac951203m.
Biographies
Luis Armando Carvajal Ahumada was born in Bogotá,
Colombia on August 8, 1985. He received his degree in
electronic engineer from National University of Colombia,
Colombia, in 2009 and his Master degree in Telemedicine
and Bioinstrumentation from Technical University of
Madrid, Spain, in 2012. He is currently a Ph.D student in
the laboratory of Bioinstrumentation and Nanomedicine
in Center for Biomedical Technology at Technical Univer-
sity of Madrid. His work is centred on the development
of new ways to find useful information of biological
and non-biological fluids deposited in a quartz crystal
microbalance and the improved the system sensibility.
Nuria Peña Pérez was born in Madrid, Spain on Nove-
meber 7, 1994. She started her Bachelor in Biomedical
Engineering in 2012 at the University Carlos III de Madrid.
She was granted with a scholarship to study her third year
at the University of California, Irvine. She is a member of
the TBP engineering honor society and has been granted
two years with the Excellency award given by the gov-
ernment of Madrid. She is currently collaborating at the
laboratory of Bioinstrumentation and Nanomedicine in
Center for Biomedical Technology at Technical University
of Madrid and doing an internship at the Hospital Gregorio
Marañón in Madrid.
Oscar Leonardo Herrera Sandoval is physicist at the
National University of Colombia, Electronic Engineering
from the University Antonio Nariño, Master in Elec-
tronic Engineering with emphasis in Bioengineering at the
University of the Andes, Doctor Nanotechnology of the
University of Genova in Italy and the Technical Univer-
sity of Darmstadt in Germany and Post doctorate from the
Technical University of Darmstadt in Germany. He is Dean
of the Faculty of Engineering of the Central University and
Associate Professor of the Central University. He is cur-
rent President of the Colombian Chapter of the Society for
Engineering in Medicine and Biology of the IEEE (Institute
of Electrical and Electronics Engineers).
Francisco del Pozo Guerrero was born in Madrid, Spain,
in 1945. He received a degree in telecommunications
engineering in 1969 and the Ph.D. degree in 1979, both
from the Universidad Politécnica of Madrid (UPM), Spain.
He was a Research Fellow in the Physics Department,
Imperial College, London, U.K.; Neurobiology Department,
University of Yale, New Haven, CT; Department of Labo-
ratory Medicine and Pathology, University of Minnesota,
Minneapolis; and Biophysics and Physiology Department,
University of New York. He was a Founder (1985) and
Director of the Bioengineering and Telemedicine Depart-
ment, UPM. He directed more than 400 scientific papers
and 20 Ph.D. theses. He was a founder and coordinator of
M.S. programs and a Ph.D. program on biomedical engineering.
Jose Javier Serrano Olmedo got his PhD. in Telecomu-
nication Engineering in 1996 at the Engineering School
on Telecommunication (technical University of Madrid)
where he teaches on Bioinstrumentation and Biosen-
sors. He has worked on several research fields including
superconductor materials characterization and sensor
networking. Now is the head of the Laboratory of
Bioinstrumentation and Nanomedicine of the Center for
Biomedical Technology, of which he is the Deputy Direc-
tor, at the same mentioned university. On of his main fields
of research is now the application of nanotechnologies to
the development of new technologies to help new thera-
pies against cancer.