Abstract—A novel sensor for environmental pollution moni-toring is presented in this paper. A capacitive measuring system,together with suitable signal conditioning electronics, has beendeveloped for performing real-time estimations of heavy metalsconcentration in air particulate. The proposed measuring systemis biologically inspired from the observation of lichens that arewidely used as bio–monitors for the environmental pollution.A synthetic replica of such bio–chemical elements has beendeveloped regarding their behavior in the trapping of atmosphericparticulate. A capacitive probe is then applied to this substratein order to perform conversion of trace element concentrationto an electric signal. The proposed prototype has been char-acterized through a large set of measurements performed inlaboratory conditions with several pollutants as found in the realenvironment: features such as element selectivity and sensitivityto pollutant concentration have been highlighted. A procedure forperforming both the pollutant classification and the estimation ofits concentration has also been developed.
Index Terms—Bio–geochemical sensors, capacitive sensors, en-vironmental monitoring, heavy metals sensors.
I. INTRODUCTION
T HE need for a continuous monitoring of trace elementsconcentration in the environment is continuously growing
both for understanding the long-term physical-chemical be-havior of the ecosystem and for verifying that legislation relatedto the discharge of these elements is respected. In particular,heavy metals, even in low concentration, may represent asevere hazard to the normal functioning of the ecosystem asthey are not biodegradable but involved in bio–geochemicalcycles and then distributed under different physical-chemicalforms. Any variation in the speciation of an element will affectits bio–availability, its rate of transport into the sediments, andits overall mobility. Due to the complexity of the ecosystemand the large number of possible artifacts in the trace metalanalysis, such as contamination, adsorption change of specia-tion due to coagulation, or microbial activity, the developmentof novel analytical measuring systems, capable of performingin situ real time monitoring of specific forms with minimumperturbation of the media, are greatly necessary.
The design of such tools is still a challenge for researchers in-volved in the development of sensory systems for environmentalpollution monitoring since techniques that combine high sensi-tivity, speciation capability, integrity of samples, and unattended
Manuscript received January 12, 2001; revised May 27, 2003.The author is with the Dipartimento di Ingegneria Elettrica Elet-
tronica dei e Sistemi, University of Catania, Catania, Italy (e-mail:[email protected])
Digital Object Identifier 10.1109/TIM.2003.817157
operation are prerequisite. Specific technical criteria, such asminiaturization and robustness of the equipment, rapidity ofdata acquisition, and transmission other than low energy con-sumption, are also important factors.
Capacitive techniques are potentially able to satisfy criteriasuch the ones reported above and thus the development of suit-able measurement systems, capable to perform estimations ofheavy metals concentration through the observation of changesin the dielectric properties of a given substrate, are greatly de-sirable.
The leading idea of this paper is derived from the observa-tion of the behavior of particular biomonitors such as lichens[1], [2] when used for the estimation of trace elements concen-tration in the environment. Biomonitoring is an experimentalmethod to measure the response of organisms to air pollution.The underlying assumption is that the presence of an element inplants reflects the content of that element in the environment inwhich the plant itself grows. The adoption of biomonitors hasrevealed to be suitable mainly for the permanent and commonoccurrence in the field, the ease of sampling, and the degree oftrace element accumulation. Among others, lichens have beenused as biomonitors [1]–[3] mainly due to their resistance tostresses deriving from excessive heat and dryness and to highpollutant tolerance levels.
In this paper, a synthetic replica of lichens has been developedin particular for what regards their behavior in the trapping ofparticulate from the environment. The measurement approach isbased on the characterization of changes in the dielectric prop-erties of such synthetic substrates induced by the presence ofthe trace elements whose concentration has to be monitored.The choice made here to adopt a synthetic prototype is partic-ularly suitable in order to ensure repeatability of measurementconditions, however, the extension of the measurement systemdeveloped in this work to hybrid configurations, where naturalbiomonitors are tied to electrical measurement systems, can beseen.
An analog measurement system has been developed and itsmetrological performances have been characterized. Several ex-perimental results will be reported and commented that show thesuitability of the results obtained in terms of sensitivity to traceelement concentration. Moreover, studies on the sensor selec-tivity features have been performed and the results obtained willbe shown.
Heavy metals pollution monitoring problems, typicallyarising in the case of thermo-electrical power plants, have beenconsidered here; therefore, trace elements as lead, vanadium,zinc, cadmium, and nickel have been taken into account.However, these elements can be found as pollutants also in
Fig. 1. Schematic representation of a lichen cross section. The uptakemechanism takes place in the medullar layer. Typical cross-section thickness isin the order of 100�m.
several other processes where internal combustion engines areinvolved, thus several other fields of potential applications forthe proposed measurement system can be envisaged.
II. THEORETICAL BACKGROUNDS
It is well known that metal contents in vegetation may changesignificantly according to the type of growing substrate and toenvironmental conditions (moisture, air quality) [1], [4]. Someplants are able to accumulate heavy metals without toxic effectsand so it is possible to make use of them to monitor the incre-ment of concentration in a period; these plants are called bio–ac-cumulators for this reason.
This work draws its inspiration from the observation of theheavy metals trapping and accumulation mechanisms in lichens.Lichens are biological elements highly sensitive to pollutingsubstances and, therefore, widely used as air quality indicators[1], [2], [4] They are symbiotic organisms resulting from theassociation between algae and fungi as shown in Fig. 1. Thereare several ways in which lichens may take chemical elementsfrom their surroundings, however one significant method con-sists of mechanical trapping of particulate inside the medullarlayers [4].
Traditional measurement strategies are based on chemicalanalysis of a small piece of one lichen suitably exposed to theenvironmental actions. These measurements have produceda large set of interesting results in the field of environmentalpollution monitoring furtherly assessing the suitability oflichens as biomonitors [1]–[3].
However, biomonitors, and in particular, lichens can be con-sidered as medium whose characteristics change with the con-centration of pollutants. This fact allows for considering syn-thetic particulate collectors and to relate the particulate concen-tration to the electric impedance of these collectors.
The samples used for the experimental measurement areporous synthetic filters here used to emulate lichen uptakemechanism of particulate. Even if the direct use of lichenswere practicable, the necessity to have repeatable measurementconditions suggested the adoption of such a synthetic strategy.A scanning electron microscope (SEM) photograph of thestructure of the adopted porous substrate, here called “pad,” isshown in Fig. 2. It can be observed that the porosity scale is of
Fig. 2. SEM photograph of a cross section of the “pad” used as a syntheticporous substrate for the estimation of heavy metal concentration.
Fig. 3. Typical electric field lines for a capacitive proximity sensor.
the order of few tenths of micrometers as for the medullar layerin lichens [4]. In order to perform our experimental laboratorycharacterization of heavy metals concentration estimations,the substrates were first saturated with aqueous solutions ofheavy metals diluted in different concentrations and then, afterdrying, they were included in a capacitive measuring system.
The measurement strategy utilized in this paper is basedon the assumption that the dielectric properties of the porousmedium change with the presence of elemental particles. Ca-pacitive sensors have been therefore developed for performingthe estimation of the dielectric characteristics of the consideredmaterial such to relate these characteristics to the concentrationand to the nature of some heavy metals pollutants. In order tobetter proceed in the development of the system, an analyticmodel has been developed.
The measurement system comprises the sensors and condi-tioning circuit. The operating principle of the adopted sensorsis the same of typical capacitive proximity pick-up [5] as shownin Fig. 3. The target object is placed in the region of the elec-tric flux and thus causes a change in the resulting capacitancevalue with respect to free space. While for a proximity sensorthis change is related to the distance between the sensing elec-trode and the target object, in the case considered in this paper,the distance is held constant and, therefore, changes observedin the sensor output can be related to variations in the dielectricproperties of the target material.
Several methodologies have been proposed in literature inorder both to study the behavior of electric field in heteroge-neous materials and to investigate their electromagnetic proper-ties in terms of conductivity, permittivity, as well as magneticpermeability. In particular, theGrain Consolidation Model[6]and theEffective medium theory[7] appear to be suitable. Thebasis of these approaches is that the electric field in heteroge-neous materials can be roughly expressed as a weighted com-bination of the contributions due to each component. Thus, for
1476 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 52, NO. 5, OCTOBER 2003
example, in the case of a two phases medium, such as oil im-pregnated cellulose, it is possible to consider the electric fields
as follows:
(1)
where and represent the electric field in cellulose and oil,respectively, and in the cellulose, whileand in the oil.
Such approaches can be extended to our case; in particular,the capacitive contribution of the substances captured on thetotal capacitance of the sensor can be considered in series withthat of the substrate, that is
(2)
where is the total capacitance, is the capaci-tance due to the pad, and is the capacitance dueto the kept substances.
The geometrical parameters of the capacitive sensor are main-tained constant thus the capacitance variations are related tochanges in the permittivity of the region subjected to the elec-tric field generated by the measurement system. The analysis ofchanges in the electric behavior of the sample is performed byvarying the frequency of the exciting signal, thus the frequencydependence of permittivity must be represented as [7], [8]
(3)
moreover the conductivity can be expressed as
(4)
In the case considered in this paper, the capacitance dependsalso on the concentrationof the pollutant; therefore, we get
(5)
where represents the proportionality coefficient between ca-pacitance and permittivity related to geometrical quantities.
The total capacitance can be therefore considered as follows:
(6)
The measurement of both the capacitance and the conductivityof the sample permits estimation of both the real part and theimaginary part of the impedance associated with the measure-ment system, sensor plus sample, according to the model of losscapacitor [5], [7].
III. SENSORPROTOTYPESDEVELOPED
The measurement system is composed by two sensors andsuitable signal conditioning circuits. A differential measure-ment strategy has been here adopted; in fact, two capacitivesensors are used to realize each measuring probe: one is appliedto a known and fixed substrate while the other one is used withthe active porous substrate.
In Fig. 4, a schematic representation of the measurementsystem developed is reported: a capacitance to voltage con-verter circuit is adopted that, once excited with a sinusoidalsignal , produces the output voltage that is a function of
Fig. 4. Schematic representation of the measurement system developed.
the difference between the sensor capacitances; a phase shiftdetector circuit is finally used to obtain a voltage that istherefore proportional also to the time delay between theand the signals.
Two sensor topologies have been realized as shown in Fig. 5:one cylindrical, with a single electrode, and one with twocoplanar electrodes, with semicircular shape. In the case ofthe planar sensor, a guard ring is also used in order to reducefringe capacitances. Both the lateral surface in the cylindricalsensor and the ring in the planar sensor are held at the groundpotential.
In both cases, the porous substrate is placed at a fixed distanceclose to the electrodes; therefore, a change of the substrate com-position in terms of element concentration and, therefore, of itspermittivity causes electric field variations. These variations arethen detected and, after calibration, they can be related to thepresence of a given trace element into the pad and to its concen-tration.
To convert the sensor capacitance variations the bridge circuitwith operational amplifiers shown in Fig. 6, in the case of thetwo electrodes, planar sensor has been used: this circuit acts asa capacitance to voltage converter [9].
The two sensors are considered as made by the combinationof capacitive and resistive impedances; in the circuit, the admit-tances and are then considered.
Given the significant differences between the two capacitivesensors proposed, it should be highlighted that, for the circuitshown in Fig. 6, when the cylindrical sensors are considered, thecircuit is modified by connecting the node “” to the ground andthe voltage source the node “ .” The cylindrical capacitivesensor is then connected with the reference electrode to the node“ ” and the sensing electrode to the amplifier input.
The two sensor configurations investigated gave similar re-sults; however, the coplanar device showed slightly better per-formances especially in terms of noise immunity. Therefore, inSection IV, this latter sensor is considered.
Differential measurement strategies are, in this case, used inorder to filter out undesired effects related to modifying andinterfering inputs. The use of alternate voltage signals avoidsthe possibility of static charge migration that could affect thecharge distribution on the porous substrate.
Referring to the circuit shown in Fig. 6, the output voltagesand can be expressed as
(7)
where and are the admittancesof the reference and of the active probes, respectively. Due tothe reactance of and , all of the terms in (7) are frequencydependent. Because is the active sensor, the actual outputsignal will be .
Fig. 5. Capacitive sensors developed. The “cylindrical single electrode” on the left and the “two electrodes planar sensor” on the right. The porous substrate,“pad,” and the signal conditioning circuits (in between the sensors) are also shown.
Fig. 6. Capacitance to voltage converter circuit adopted in the proposedmeasuring system.
A circuit for phase delay estimation has been realized byusing a suitable combination of Schmitt triggers, inverting andsumming amplifiers. This circuit is shown in Fig. 7 together withthe corresponding time domain waveforms. The RMS values ofoutput voltages and are proportional to the time delayfrom of and , respectively, and, therefore, to their phaseshift.
The real and imaginary part of and can be also obtainedas follows:
(8)
where and represent the phase shift between the inputsignal and the output signals and , respectively. Theselatter quantities are derived with simple calculation fromand .
IV. EXPERIMENTAL RESULTS
In order to characterize the changes in the dielectric proper-ties of the active capacitive sensor, in the following peak-to-peakamplitude of the output voltage together with the timedelay between and will be taken into account.
An extensive measurement program has been undertaken inwhich five types of heavy metals have been considered: lead,cadmium, vanadium, zinc, and nickel. Trace elements concen-trations ranging between 10 and 200 [ppm] have been taken intoaccount. These values are consistent with the results obtainedby using lichens as biomonitors and presented in literature astypical of heavy metals concentration in nature resulting fromenvironmental pollution related, for example, to the presence ofthermo-electric power plants [2], [3].
One “pad,” with a concentration of 100 ppm of Vanadium,was taken as reference and placed in the fixed capacitor of thecircuit as . For every concentration and for each element, re-peated measurements at several frequencies spanning the range500 Hz–1000 kHz were performed.
The output voltage , which shows greater sensitivity tochanges in the variable admittance, was considered: the be-havior of this voltage signal is directly related to the sample con-centration. From the experimental data the frequencies of majorinterest lie in the range 5–500 kHz, and, thus, it is the frequencyrange in which the signal conditioning circuit shows the greatestsensitivity.
Several measurements have been carried out in order to char-acterize three main cases.
1478 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 52, NO. 5, OCTOBER 2003
Fig. 7. Circuit schematic and waveforms of the phase detector. The RMS value of the output signalV is proportional to the time delay betweenV andV andtherefore their phase difference.
Fig. 8. Experimental results in the case of single pollutant. The output voltage and the time delay betweenV andV are reported for several trace elements atdifferent concentrations.
• One single polluting substance is present in the pad. Inthis case, both of the estimation of concentration and theidentification of the nature of the substance have been per-formed.
• Two polluting substances A and B, with concentrations, are present in the pad. In this case, the goal
was to perform sensitivity analysis and, in particular, toidentify the largest value of ratio that does not af-fect the estimate of the trace element A.
• Two polluting substances A and B, with arepresent in the pad. In this case, measurements were per-formed at different frequencies in order to investigate the
possibility of both discriminating between elements andestimating their concentration.
A. Single Polluting Species
Results obtained in the assumption of a single metal haveshowed the possibility to distinguish among the selected classes:cadmium, lead, nickel, vanadium, and Zinc.
For all of the trace elements observed, the output voltageamplitude increases as expected with the increment of bothconcentration and exciting signal frequency. Measurements ofoutput voltage amplitude and time delay allow discriminating
Fig. 9. Example of identification procedure. Taking into account the time delay, the ambiguity deriving from the reading of the output voltage (points showncircled) is solved.
among the various elements and estimating their concen-trations. In Fig. 8, typical experimental results are shown.Repeated measures are reported in the case ofat 5 kHz inorder to illustrate the typical repeatability. In particular, thepeak-to-peak value of the output voltage is reportedtogether with a simple linear interpolation of the experimentaldata. In the case of time delay betweenand , the linearinterpolation is directly shown.
The measurement procedure consists of the following steps:measurement of and as output voltage peak-to-peak am-plitude and time delay, respectively; combined use of the char-acterization graphs as shown in the example is reported in Fig. 9.In this way, it is possible both to estimate the element concen-tration and to classify its chemical type.
In the example of Fig. 9, the analysis of (1.21 V) givesthe ambiguity between “lead at 70 ppm,” “nickel at 150 ppm,”and “vanadium at 190 ppm.” The simultaneous acquisition ofthe time delay s between and s allows the cor-
Fig. 10. Sensitivity analysis results. Changes in the output voltage inpresence of a “contaminant” element (Cd); the reading remains valid until theconcentration of Cadmium is higher than 15%.
1480 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 52, NO. 5, OCTOBER 2003
Fig. 11. Output voltage for mixed samples of lead (Pb) and vanadium (V) at 5–10 kHz (bottom) and 50–100 kHz (top), it is possible to observe as each point ofthe grid behaves differently with frequency.
rect identification of the sample as polluted by “nickel with aconcentration of 150 ppm.”
As a remark, it should be highlighted that the pollutant iden-tification procedure discussed above could be further improvedif a more complex interpolation procedure, instead of the simplelinear function, were to be applied to the experimental data. Thiswould reduce the large uncertainty that arise from the obser-vation of Fig. 8 and would be of great importance if a largernumber of polluting elements would be considered. However,in what is shown here, the combined use of and infor-mation are sufficient both to unambiguously recognize the pol-lutant species, among those considered, and to estimate its con-centration.
B. Sensitivity Analysis
In this case, two polluting species A and B, with concentra-tions , are simultaneously present in the substrate. Thesample should be classified as to contain the element A with
concentration in spite of the presence of the interfering ele-ment B. The goal of this sensitivity analysis is to determine thelargest value of the ratio that results in a correct classi-fication of the sample.
The experimental sensitivity analysis was performed bychoosing the element A to be Nickel in 100 ppm, while, aselement B, Cadmium had concentrations ranging from 5 to25 ppm. The results of this sensitivity analysis are shown inFig. 10. It is possible to observe that the output voltage doesnot change appreciably, in a large range of frequencies, untilthe concentration of Cadmium in the samples is higher than15% of the concentration of the main element.
C. Two Polluting Species
In this case the samples are sodden with solutions of waterand two metals, lead and vanadium, in various concentrationsso to have a grid of nine samples. Starting on the hypothesis thatthere are two polluting species in the pads, the output voltage has
been gathered for several frequency values of the input signalin order to verify the possibility of performing identification ofthe elements contained in the pad.
This problem is of significantly greater complexity. The goalis to show how the output of the measurement system changesin different situations.
The collected measures are presented in Fig. 11 in the case oflead and vanadium mixed into the same substrate with differentconcentrations; as can be observed, the points in the grid showdifferent behaviors with frequency. Therefore, a measurementstrategy based on the observation of the output of the proposedmeasurement system at different frequencies can permit a suit-able classification of the samples.
V. CONCLUSION
In this paper, a novel capacitive measurement system forreal-time heavy metals pollution monitoring has been pre-sented. A methodology to measure the concentration of suchsubstances has been developed here that draws inspirationfrom the observation of the uptake mechanisms of lichens.Two sensor prototypes have been realized and a large set ofexperimental measurements has been presented in order toshow the suitability of the proposed approach.
The capacitive measuring system proposed allows for esti-mating the concentration of trace elements in the environment;moreover, the selectivity of the system has been exploited.
The sensor is based on the estimation of the dielectric prop-erties of a given substrate that are a function of both the kindof trace element and of its concentration. A sinusoidal voltageinput signal has been used; by taking into account, at several fre-quency values, both the output voltage amplitude and the phaseshift, a characterization of the system has been performed toallow for estimating both the pollutant concentration and its na-ture.
The measuring system proposed appears therefore as aninteresting alternative to traditional biomonitors. It allows forreal-time monitoring and consists of an inexpensive, robust,and simple device that can be easily reproduced and adopted inlarge scale in environmental monitoring.
ACKNOWLEDGMENT
The author would like to thank Prof. M. Triscari for his pre-cious and friendly collaboration.
REFERENCES
[1] J. E. Sloof, “Lichens as quantitative biomonitors for atmospheric traceelement deposition,”Atmos. Environ., vol. 29, no. 1, 1995.
[2] G. Dongarrà, D. Ottonello, G. Sabatino, and M. Triscari, “Preliminarydata on heavy metal content in lichens from north eastern sicily,”Min-eral Petrogr. Acta, vol. 37, 1994.
[3] M. Rossbach, R. Jayasekera, G. Kniewald, and N. H. Thang, “Largescale air monitoring: Lichen vs. air particulate matter analysis,” inTheScience of the Total Environment. New York: Elsevier, 1999, vol. 232,pp. 59–66.
[4] E. Nieboer, D. H. S. Richardson, and F. D. Tomassini, “Mineral uptakeand release by lichens: An overview,”The Bryologist, vol. 81, 1978.
[5] L. K. Baxter,Capacitive Sensors Design and Applications. New York:IEEE Press, 1997.
[6] B. Nettelblad and B. E. Mellander, “The electric field in porous materialscalculated with the grain consolidation model,”IEEE Trans. Dielect.Elect. Insul., vol. 3, Feb. 1996.
[7] I. J. Youngs, “Dielectric measurements and analysis for the design ofconductor/insulator artificial dielectrics,”Proc. Inst. Elect. Eng. Sci.Meas. Technol, vol. 14, no. 4, 2000.
[8] P. T. Moseley and A. J. Crocker,Sensor Materials: IOP, 1996.[9] K. Watanabe, H. Sakai, O. Ogawa, K. Mochizuki, and T. Masuda, “High
accuracy signal processing of differential pressure transducers,” inIn-strum. Meas. Soc. Newslett., 1997.
Salvatore Baglio(S’91–M’94–SM’03) was born inCatania, Italy, in 1965. He received the Laurea andPh.D. degrees from the University of Catania in 1990and in 1994, respectively.
He was a Lecturer of automatic control theory withthe University of Messina, Messina, Italy, and elec-tronic measurement systems, University of Catania.He was also a Consultant for the STMicroelectronicsin the field of soft computing methodologies for non-linear and chaotic circuits and systems. Since 1996,he has been with the Dipartimento di Ingegneria Elet-
trica Elettronica e dei Sistemi, University of Catania, where he is now AssociateProfessor of electronic instrumentation and measurement. He teaches coursesin measurement theory and sensors and transducers. He is the coauthor of morethan 120 scientific publications, among which includes papers published in in-ternational journals or presented at international conferences, chapters in books,and U.S. patents. His research interests include nonlinear and chaotic dynam-ical systems, soft computing methodologies for instrumentation and measuringsystems, smart sensors, microsensors, and microsystems.
