International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

A non-specific microwave sensor for water quality monitoring

G. Gennarelli, F. Soldovieri
Institute for the Electromagnetic Sensing of the Environment, National Research Council of Italy,
Via Diocleziano, 328-80124 Napoli, Italy.
E-mail: gennarelli.g@irea.cnr.it

ABSTRACT
This paper deals with a microwave measurement system for the real-time monitoring of water
quality. The sensor is a compact cavity resonator tuned at 1.91 GHz which has been designed by a
Finite Element Method code to maximize the Q-factor. The sensing principle relies on the changes of
the device response in terms of the transmitted power, resonance frequency, and 3-dB bandwidth.
These variations are caused by the change of the permittivity and conductivity of the water sample
under test induced by the presence of pollutants. Experiments are carried out in laboratory conditions
on water solutions contaminated with different substances. As shown, the system is capable to reliably
detect the presence of nitrates and sulfites dissolved in water in concentrations above 10 mg/L. The
proposed set-up is portable and thus particularly tailored for its deployment on-field. Moreover, it has
reduced cost since it does not necessitate expensive instrumentation such as network analyzers.

Keywords: microwave sensor, nitrates, sulfites, water pollution analysis.

1. INTRODUCTION
Water pollution concerns any chemical, physical or biological variation in the quality of water which
exceeds some reference values and may cause health problems. The most evident ones are the result of
microbial (bacteriological, viral, protozoan or other biological) contamination which usually arises in
presence of untreated waste and sewage discharges. Nevertheless, an appreciable number of serious
health problems may also occur as a result of chemical contamination. Chemical pollutants are usually
classified into organic and inorganic materials. They may be products expelled from factories, e.g.
industrial solvents, heavy metals, plastics, fertilizers and pesticides which are harmful to humans and
also to all plants and animals in water.
Nowadays, the quality of water intended for human consumption is usually high when it leaves the
purification plant, but it can deteriorate on its way to the consumers for the presence of chemical
substances or because of stagnation in reservoir which may facilitate the growth of algae and bacteria.
As a result, water monitoring is of primary concern for most health and security authorities.
In this framework, the European Directive 98/93/EC [1] has set quality standards of drinking water on
the basis of the guidelines from the World Health Organization (the most recent guidelines can be
found in [2]). In particular, it has fixed the maximum allowed concentrations of several chemical and
microbiological substances based on their maximum daily consumption over long period of times,
their nature and their potential toxicity.
Water quality sensors exploiting different technologies (e.g. chemical, optical, acoustic, etc.) have
been presented in the literature [3]-[9]. Moreover, many commercial solutions are currently available
on the market and commonly adopted to analyze specific parameters of interest such as PH,
conductivity, free and residual chlorine concentration, dissolved oxygen, turbidity, etc. Sometimes
these features can be even monitored by a single multi parameter sonde which provides much more
flexibility at the expense of a major complexity and cost.

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

However, real-time monitoring of drinking water can involve the presence of several sensors deployed
along the distribution network. In this scenario, it may be convenient to use non-specific devices,
possibly low-cost, which are sensitive to several impurities instead of many specific sensors each
sensitive to a single substance. For instance, to reduce the complexity and the cost of the overall
monitoring system, it may be sufficient at an early stage to measure only the variations, with respect to
some nominal conditions, of some typical sensor parameters. Only when significant variations are
recorded, detailed analyses with specific sensors are carried out.
The aim of this paper is to demonstrate the feasibility to detect water contaminants by exploiting
microwave technology. The sensing device here considered is a rectangular cavity resonator working
at 1.91 GHz in the fundamental TE101 mode. The basic idea is to introduce a water sample inside the
cavity by means of a low loss capillary. As well-known, the presence of pollutants in water alters the
complex permittivity of the solution (i.e. the dielectric permittivity and conductivity) [9] thus
modifying the cavity response in terms transmitted power, resonance frequency and 3-dB bandwidth.
As a consequence, it is possible to detect the pollutants by monitoring these parameters in real-time. It
must be stressed that the sensor is non-specific, i.e. it is not capable to discriminate the type of
contaminants and their respective concentrations without a-priori information on the solution
composition and the availability of proper calibration curves. This point is easily understood since the
same permittivity and/or conductivity variations could be for instance induced by different substances
or different combinations of substances.
Cavity perturbation methods have been widely investigated in the past years for characterizing the
dielectric behavior of materials at microwave frequencies [10]. Moreover, their application water
pollution analysis has been considered in [12] and [13]. Unlike these works, here we propose a novel
cavity design and portable measurement system suitable for on field applications. Moreover, the main
contribution is the appraisal of the achievable sensing performance in terms sensitivity and lowest
detectable concentrations.
The paper is organized as follows. Section 2 provides the description of the sensor, its working
principle and design parameters. The measurement system and the characterization of several polluted
waters are presented in Section 3. Conclusions follow in Section 4.

2. THE CAVITY SENSOR

A drawing of the rectangular cavity-based sensor also comprising the elements for signal coupling is
depicted in Fig. 1. The structure is made of aluminum and has dimensions a , b , and l along the x, y,
and z axis, respectively. Such dimensions are fixed to a = l = 11.065 cm and b = 5.45 thus giving a
fundamental resonant TE101 mode at the frequency 1.9158 GHz. The structure is coupled to the
external circuit via coaxial probes connected to 50 Ohm cables. The probes are inserted in the cavity
via holes drilled in the bottom wall. Their geometrical features have been optimized by using a Finite
Element (FEM) code in order to maximize the Q-factor, i.e., frequency selectivity. Based on this
optimization, a resonant frequency of 1.915 GHz and a Q-factor of 3800 have been experimentally
determined. These values are in good agreement with the simulation results (errors less than 1 %).
More details on the design of the cavity and the characterization of its operation are available in a
recent paper [13] dealing with the usage of this device as concentration sensor for binary liquid
mixtures.
As shown in Fig. 1, a cylindrical capillary with radius rc passes through the cavity at its center where
the electric field is maximum to allow the water flow. The water sample to be analyzed is modeled as
a dielectric material characterized by the complex relative permittivity = ' j '' = '(1 j tan ) ,
where ' and '' are the real and imaginary parts and tan is the loss tangent.
According to resonant perturbation theory [10], the presence of a water solution inside the capillary
gives rise to a shift of the resonant frequency and an increase of the 3-dB bandwidth (reduction of the
Q-factor) with respect to the empty cavity case. Of course, the amount of these perturbations depends
on sample volume as well as on its dielectric properties, i.e. its chemical composition. Based on the
same principle, a variation of the electromagnetic properties of water caused by contaminants is

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

Fig. 1 Sketch of the rectangular cavity and its coupling elements.

expected to produce a change of the resonance frequency and of the 3-dB bandwidth. Moreover, also
the reflected and transmitted powers at the ports of the cavity (ports 1 and 2 in Fig. 1) vary.
In order to get insight into the physical behavior of the sensor when it is filled with water samples,
numerical simulations are carried out with the FEM method. An example showing the transmission
coefficient S21 of the cavity is reported in Fig. 2. The effect of the capillary radius as well as of the
complex permittivity is investigated (see caption of Fig. 2 for further details). Once ' and '' ( tan )
are fixed, increasing rc yields a higher resonant frequency shift and a broadening of the 3-dB
bandwidth. Moreover, increasing the sample radius yields a higher sensitivity with respect to complex
permittivity changes (see in particular the curves relevant to rc = 1.5 mm). Figure 2 also reveals that
the transmission peak of the S21 coefficient lowers with the sample volume. In a similar way, the
reflection coefficient at the input port (not shown) is strongly influenced by the sample volume and the
mismatch increases with rc . Accordingly, depending on the type of liquid to be characterized, a
limitation on the maximum sample volume is expected in practice. Note also that the accuracy of the
resonant frequency evaluation is directly related to the Q-factor and to the signal-to-noise ratio on
measured data [14].
15
rc = 0.5 mm
20
25
rc = 1 mm

S21 (dB)

30
35
rc = 1.5 mm

40
45
50
55
60

1.7

1.75

1.8
f (GHz)

1.85

1.9

Fig. 2 Simulated transmission coefficient for different sample radii and dielectric constants. The solid
curves refer to '=74 whereas the dashed lines are related to '=75 . The dielectric loss tangent is equal to
0.05.

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

3. EXPERIMENTAL ANALYSIS
In this Section, the results of laboratory tests carried out in controlled conditions at IREA-CNR are
reported. A photo of the measurement set-up is shown in Fig. 3. The cavity-based sensor is fed at the
input port by a synthesized RF signal generator (mod. SSG-4000LH manufactured by Mini Circuits)
and the power transmitted at the output port is measured by a power meter (mod. PWR-4RMS
produced by Mini Circuits). Both devices are programmable and communicate via USB with a laptop
which represents the control unit. The RF generator transmits CW signals over the frequency range
[1815, 1840] MHz with a step of 200 KHz and acquires each resonance curve in about 20 seconds.
Then, the resonance frequency and the 3-dB bandwidth are extracted from measured curve by a
Lorentzian fitting method [14]. A software has been developed to easily configure measurement
parameters, acquire the data, process them and display the results in real-time.
The water solutions are introduced in the cavity via a thermoplastic capillary of inner radius equal to 1
mm and thickness of 1 mm. In particular, the capillary is connected to an external pump via tube
fittings so that a continuous water flow is created and no air bubbles are formed in the capillary. The
temperature of the liquid under test is kept constant at 25C using a thermostatic bath (mod. Lauda
M12). As will be shown, temperature variations represent a key point that must be taken into account.
As a matter of fact, they alter the electromagnetic properties of the solution under test thus lowering
the sensor capability to detect pollutants.
The water contaminants considered for the experiments are potassium-nitrate (KNO3) and potassiumsulfite (K2S2O5). Nitrates are generally not extremely toxic but their reduction to nitrites can lead to a
generalized lack of oxygen in organ tissue and a dangerous condition called methemoglobinemia [2].
Their allowed maximum concentration is fixed by law to 50 mg/L. As regards the potassium-sulfite
(K2S2O5), it is commonly employed as an additive for food conservation and it can cause allergies to
sensitive people. However, no guidelines values are available for it.
Several water-KNO3 and water-K2S2O5 samples are prepared with growing solute concentrations in
the range [0, 500] mg/L. For each concentration value, ten measurements are collected in order
perform an estimate of measurement errors. The sensor response is evaluated only in terms of
transmitted power at resonant frequency and 3-dB bandwidth since the resonance frequency changes
are less detectable for the considered solutions. This point is clearly highlighted by the results shown
in Fig. 4 concerning the measured power versus frequency in the case of water-KNO3 solutions. Note
that increasing the nitrate concentration mainly lowers the peak power and increases the bandwidth
due to the higher conduction losses.
Starting from the measured curves, the average (mean value) variation of the transmitted power at the
resonance frequency and of the 3-dB bandwidth are computed and plotted versus the KNO3
concentration in Figs. 5 a and b, respectively. As can be seen, these quantities well agree with a linear
model. Moreover, the sensitivity to peak power variation is 0.0018 (dBm/(mg/L)), whereas that related
to the 3-dB bandwidth variation is equal to 3.8 (KHz/(mg/L)).

Fig. 3 A photograph of the measurement system.

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

27

28

Transmitted power (dBm)

29

30

31

KNO3=0 mg/L
KNO3=100 mg/L
KNO3=200 mg/L

32

KNO3=300 mg/L
KNO3=400 mg/L

33

34
1810

KNO3=500 mg/L

1815

1820

1825
1830
f (MHz)

1835

1840

1845

0.7

1.4

0.6

1.2

Change of 3dB bandwidth (MHz)

Change of peak power (dBm)

Fig. 4 Transmitted power versus frequency for different water-KNO3 solutions. Only one measurement is
shown for each concentration value to improve results readability.

0.5
0.4
0.3
0.2
0.1

1
0.8
0.6
0.4
0.2

Linear fit
Data
0
0

100

200
300
concentration (mg/L)

400

Linear fit
Data
0
0

500

100

200
300
concentration (mg/L)

400

500

(a)
(b)
Fig. 5 Measurement results (mean values and error bars) and the linear fit curve for water-KNO3
solutions. (a) Variation (modulus) of the peak power. (b) Variation of the 3-dB bandwidth.

After, we analyze the water-K2S2O5 samples for the same concentrations of solute. The results
obtained in this case, plotted in Figs. 6 a and b, reveal similar trends as for the water-KNO3 solutions,
and the main difference is a reduction of the sensor sensitivities (0.001 dBm/(mg/L) and 2.1
KHz/(mg/L)).
Linear fit
Data

Linear fit
Data
1.2
Change of 3dB bandwidth (MHz)

Change of peak power (dBm)

0.6

0.5

0.4

0.3

0.2

0.1

0
0

0.8

0.6

0.4

0.2

100

200
300
concentration (mg/L)

400

500

0
0

100

200
300
concentration (mg/L)

400

500

(a)
(b)
Fig. 6 Measurement results (mean values and error bars) and the linear fit curve for water-K2S2O5
solutions. (a) Variation (modulus) of the peak power. (b) Variation of the 3-dB bandwidth.

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

Based on the above results, we perform an estimation of the Limit of Detection (LoD), i.e., the
minimum detectable concentration for a given pollutant. To this end, the standard deviations of the
sensor response have been evaluated on a water sample. The sensor performance achieved is
summarized in Tab. 1 and, as can be noted, they confirm the feasibility to detect pollutant
concentrations in excess of 10 mg/L.

Table 1 Sensor performance at 25 C

Solution

Power sensitivity Bandwidth sensitivity Power LoD Bandwidth LoD

(dB/(mg/L))
(KHz/(mg/L))
(mg/L)
(mg/L)

WaterKNO3

0.0013

2.6

0.001

2.1

WaterK2S2O5

Finally, we investigate the effects on the sensor response caused by temperature variations
intentionally induced with the thermostatic bath. This analysis is accomplished to estimate the
temperature sensitivities which may aid to compensate unwanted temperature changes arising when
the sensor is needed to operate on field. In particular, the temperature of a water solution is
progressively increased by steps of 1 C in the interval [20, 25] C. The variations of the transmitted
power, resonance frequency and 3-dB bandwidth versus temperature are plotted in Figs. 7 a, b, and c.
In this case, increasing the temperature lowers the permittivity and conductivity of the solution with a
resulting increase of the peak power, of the resonant frequency, and reduction of the bandwidth.
Moreover, the trends are linear and the sensitivity of power variation is equal to 0.2 dBm/C, the
sensitivity of resonant frequency is 400 KHz/C, and that relevant to bandwidth is equal to 380
KHz/C.
27

1828

21

Linear fit
Data

Linear fit
Data

27.5

20.5
3dB bandwidth (MHz)

Resonance frequency (MHz)

Peak power (dBm)

1827.5

1827

1826.5

1826

28

20

19.5

19
Linear fit
Data

20

21

22

23
T (C)

24

25

1825.5
20

21

22

23
T (C)

24

25

18.5
20

21

22

23

24

25

T (C)

(a)
(b)
(c)
Fig. 7 Measurement results (mean values and error bars) and the linear fit curve for water versus
temperature in the range [20, 25]C. (a) Transmitted peak power. (b) Resonance frequency (c) 3-dB
bandwidth.

4. CONCLUSIONS
A non-specific microwave cavity sensor has been proposed as a useful tool for real-time water
pollution analysis. The measurement system here considered is portable and has a relative low cost
with respect to more expensive solutions based on network analyzers. Experimental tests have
confirmed that the sensor is capable to detect pollutants such as nitrates and sulfites dissolved in water
at concentrations above 10 mg/L regardless of the parameter measured. These results turn out to be
very promising and suggest that, if temperature variations of the liquid under test are monitored and
compensated, it is possible to detect the presence of various pollutants of practical interest.

International Water Technology Journal, IWTJ

Vol. I - Issue 3, December 2011

ACKNOWLEDGMENTS
This research has received funding from the Italian Ministero dello Sviluppo Economico,
under Grant Agreement "Industria 2015 - New Technologies for the Made In Italy", No. MIOl
00223 (project ACQUASENSE).
REFERENCES
[1]
[2]
[3]
[4]
[5]

[6]
[7]

[8]

[9]

[10]
[11]
[12]
[13]
[14]

Council directive 98/83/EC of 3 November 1998 on the quality of water intended for human
consumption, L 330/32 EN, Official Journal of the European Communities.
Guidelines for drinking-water quality 4th edition, World Health Organization, 2011.
J. Wang, R. K. Bhada, J. Lu, D. MacDonald, Remote electrochemical sensor for monitoring
TNT in natural waters, Analytica Chimica Acta, vol. 361, pp. 85-91, 1998.
J. W. Gardner, H. W. Shin, E. L. Hines, C. S. Dow, An electronic nose system for monitoring
the quality of potable water, Sensors and Actuators B, vol. 69, pp. 336341, 2000.
W. Bourgeois, P. Hogben, A. Pike, R. M. Stuetz, Development of a sensor array based
measurement system for continuous monitoring of water and wastewater, Sensors and Actuators
B, vol. 88, pp. 312-319, 2003.
E. M. Nolan and S. J. Lippard, A Turn-On fluorescent sensor for the selective detection of
mercuric ion in aqueous media, J. Am. Chem. Soc., vol. 125, pp. 14270-14271, 2003.
E. S. Forzani, H. Zhang, W. Chen, and N. Tao, Detection of heavy metal ions in drinking water
using a high-resolution differential surface plasmon resonance sensor, Environ. Sci. Technol.,
vol. 39, pp. 1257-1262, 2005.
M. I. S. Verssimo, M. T. S. R. Gomes, The quality of our drinking water: aluminium
determination with an acoustic wave sensor, Analytica Chimica Acta, vol. 617, pp. 162166,
2008.
O. Korostynska, A. Mason, A. Al-Shammaa, Monitoring of nitrates and phosphates in
wastewater: current technologies and further challenges, International Journal on Smart Sensing
and Intelligent Systems, vol. 5, no. 1, pp. 149-176, 2012.
L. Chen, C. Ong, C. Neo, V. V. Varadan, and V. K. Varadan, Microwave electronics:
Measurement and materials characterization, John Wiley and Sons, New-York, 2004.
H. J. C. Blume, Measurement of dielectric properties and determination of microwave
emissivity of polluted waters, IEEE Trans. Instrum. Measur., vol. 29, no. 4, pp. 289-291, 1980.
U. Raveendranath, and K. T. Mathew, Microwave technique for water pollution study, Journal
Microwave Power and Electromagnetic Energy, vol. 30, no. 3, pp. 188-194, 1995.
G. Gennarelli, S. Romeo, M. S. Scarfi, F. Soldovieri, A microwave resonant sensor for
concentration measurements of liquid solutions, IEEE Sensors Journal, vol. PP, no.99, pp.1, 0.
R. Inoue et al., Data analysis of the extraction of dielectric properties from insulating substrates
utilizing the evanescent perturbation method, IEEE Trans. Microwave Theory Tech., vol. 54, no.
2, pp. 522-532, 2006.