192

IEEE SENSORS JOURNAL, VOL. 7, NO. 2, FEBRUARY 2007

Hydrocarbon and Fluorocarbon Monitoring by MIS Sensors Using an Ni Catalytic Thermodestructor Vladimir I. Filippov, Werner Moritz, Alexander A. Terentjev, Alexey A. Vasiliev, and Sergey S. Yakimov

Abstract—An increase in the number of gases detectable by sensors based on Pd−SiO2 −Si (MIS) and Pt−LaF3 −Si3 N4 − SiO2 −Si (MEIS) structures was achieved by the application of an external catalyst element (CE). It was shown that as a result of the decomposition of hydrocarbon and fluorocarbon molecules on a Ni coil (CE), the products detectable by metal– insulator–semiconductor (MIS) and metal-electrolyte–insulator– semiconductor (MEIS) sensors are formed. The simultaneous catalytic oxidation of hydrocarbons and their thermal decomposition result in an optimum CE temperature of about 1050 K for propane. The kinetics of the thermal decomposition of gases on Ni were investigated. The activation energy of the reaction for C3 H8 and the enthalpy in the case of CF3 −CCl were estimated. Index Terms—Fluorocarbons, hydrocarbons, metal– insulator–semiconductor (MIS) and metal-electrolyte–insulator– semiconductor (MEIS) sensors, thermal decomposition.

I. I NTRODUCTION

G

AS sensors based on metal–insulator–semiconductors (MISs) and structures with an additional undergate layer of solid electrolyte LaF3 [metal-electrolyte–insulator– semiconductor (MEIS)] have high sensitivity to hydrogen, fluorine, and hydrogen fluoride. An extension of the number of detectable gases can be achieved by means of a considerable increase in sensor working temperature. Application of widebandgap semiconductor silicon carbide (SiC) instead of silicon gives the possibility of increasing the working temperature of the sensor to ∼1000 K. This leads to the registration of hydrocarbons by MIS sensors [1] and chlorofluorocarbons (Freon) by sensors with LaF3 layers [2]. Unfortunately, a high working temperature results in a decrease in sensor lifetime, which is due to oxidation of the ohmic contact. Application of an external catalyst element (CE) gives the possibility of avoiding sensor degradation at high temperature. The high-temperature catalytic decomposition of gas molecules leads to gaseous products, which can be detected by sensors working at the usual temperature [3]. Manuscript received March 2, 2006; revised June 28, 2006. This work was supported by ISTC under Grant 2503. The associate editor coordinating the review of this paper and approving it for publication was Dr. Andre Bossche. V. I. Filippov, A. A. Terentjev, and S. S. Yakimov are with the Russian Research Center, Kurchatov Institute, Institute of Molecular Physics, 123182 Moscow, Russia (e-mail: [email protected]; [email protected]; [email protected]). W. Moritz is with the Institut fur Chemie, Humboldt-Universitt zu Berlin, 12489 Berlin, Germany (e-mail: [email protected]; http://www. chemie.hu-berlin.de/wmoritz/index.html). A. A. Vasiliev is with the Russian Research Center, Kurchatov Institute, Institute of Molecular Physics, 123182 Moscow, Russia, and also with the University Rovira i Virgili, DEEEA, 43007 Tarragona, Spain (e-mail: [email protected]; [email protected]; [email protected]). Digital Object Identifier 10.1109/JSEN.2006.882778

The possibility of using Ni as a material for the external thermodestructor was investigated in this paper. The high catalytic activity of nickel and its stable operation in fluorine-containing atmosphere are the reasons for our interest on this metal [4]. II. E XPERIMENT Sensors based on Pt−LaF3 −Si3 N4 −SiO2 −Si (MEIS) and Pd−SiO2 −Si (MIS) structures were used. The Pt−LaF3 − Si3 N4 −SiO2 −Si structures were prepared using an n−Si/ SiO2 (20 nm)/Si3 N4 (80 nm) wafer with a donor concentration of 2.5 · 1014 cm−3 . A backside ohmic aluminum contact was fabricated by vacuum evaporation of metal. LaF3 (150 nm) layers were deposited on the insulator by thermal evaporation in a high-vacuum system. DC cathode sputtering in argon plasma through a metal shadow mask was used to deposit the Pt gate with an area of 3 mm2 . The thickness of the Pt gate of the structures was increased up to 400 nm to avoid the influence of the dipole moment of the polar molecules [5], [6], which can be generated during the reaction thermal decomposition. The aging of the MEIS sensors, which work at room temperature, was observed. A method using a short heating pulse [7] was developed to maintain stable sensor response for a long time. The energy necessary for this sensor activation is only 10−3 J per day. The Pd−SiO2 −Si-based sensors were manufactured as hybrid circuits. A gas-sensitive element (MIS capacitor) and temperature sensor (a silicon diode) were installed on a thin insulator wafer with resistive heater. The thermal sensor was used in the feedback circuit of the electronic controller. This controller stabilized the preset temperature of the gas-sensitive structure in a temperature range of 300–460 K with an accuracy of 0.2 K. Palladium gates (about 100 nm thick and 0.3 mm2 in area) were formed by thermal laser-beam evaporation (laser ablation). The MIS sensor aging was not observed in all experiments (more than some hundreds of measurements). The sensors were characterized by the high-frequency capacitance–voltage method using a Hewlett Packard 4284A type LCR meter. A computer-controlled system for the determination of the bias voltage shift at constant capacitance was used. The stationary value of bias voltage shift ∆U and the initial rate of bias voltage shift d∆U/dt were used as a sensor signal correlated with gas concentration. Gas mixtures of Freon, fluorine, propane, or hydrogen in synthetic dry air were used in this paper. The gas was diluted with synthetic air using computer-operated mass-flow controllers to obtain desirable concentrations. The gas mixture passed first through the reactor (a quartz tube with a diameter of 0.4 cm

1530-437X/$25.00 © 2006 IEEE

FILIPPOV et al.: FLUOROCARBON MONITORING BY MIS SENSORS USING Ni CATALYTIC THERMODESTRUCTOR

Fig. 1. Pt−LaF3 −Si3 N4 −SiO2 −Si sensor responses to 5000-ppm F134 in air at different temperatures of CE.

193

Fig. 2. Pt−LaF3 −Si3 N4 −SiO2 −Si sensor responses versus Ni destructor temperature for 1000-ppm F133 and F113 in air. Gas flow rate: −2 cm3 /s.

whose volume was equal to about 1.2 cm3 ) with a CE (Ni wire coil) and then through a measurement cell with the sensors. The reactor and the measurement cell were connected by a 20-cmlong Teflon tube. An increase in tube length of up to 100 cm did not influence the sensor response and only leaded to increasing the “transport” time. The temperature of the thermodestructor was controlled by the measurement of the Ni coil resistance. The gas flow in the range of 1–6 cm3 /s was stabilized by a mass-flow-controller-based installation. III. R ESULTS AND D ISCUSSION A. Thermal Decomposition of Freons on a Ni Destructor The following fluorochlorocarbons with and without hydrogen were investigated: CF3 −CH2 Cl (F133), CF3 −CH2 F (F134) and CF3 −CCl (F113). It can be assumed that hydrogen, fluorine, and hydrogen fluoride will be the products of the thermal decomposition of these gases [8]. The sensors based on the MEIS structure can detect H2 [9], F2 , and HF [10] at room temperature. The sensor signal for hydrogen is opposite in sign to that for fluorine and hydrogen fluoride. Furthermore, the sensitivity to HF and F2 is different. This could be used for preliminary conclusions about decomposition products. More complete information about the products of the Freon thermal decomposition on Ni can be obtained from mass-spectrometer measurements, but it is beyond the scope of this paper. The MEIS sensor response curves for 5000-ppm F134 in air that passed through the reactor at different temperatures of the Ni thermodestructor are shown on Fig. 1. For all the Freons used in the experiments, both with and without hydrogen, positive values of the sensor response (∆U > 0) were observed. It means that the main products of the Freon decomposition reaction on Ni that cause the sensor signal are fluorine or hydrogen fluoride. More detailed experiments were carried out with F133 and F113. The bias voltage shifts ∆U of the MEIS sensor, which depend on the Ni destructor temperature, for a concentration of these two gases of equal to 1000 ppm in air are presented in Fig. 2. The sensitivity of the MEIS sensor for these two Freon types is shown in Fig. 3. The temperature of the Ni destructor was

Fig. 3. Sensitivity of the MEIS sensor to F133 (TNi = 823 K) and F113 (TNi = 623 K). Gas flow rate: −2 cm3 /s.

623 K for F113 and 823 K for F133 for the same gas flow rate of about 2 cm3 /s. A linear dependence of the bias voltage shift on logarithm of the Freon concentration was found for both gases. This corresponds to the Nernst equation ∆E =

RT C2 ln ZF C1

(1)

where ∆E is the change in electrical potential caused by a change in the gas concentration from C1 to C2 , R is the gas constant, F is the Faraday number, and Z is the number of electrons transferred in the reaction. The estimation of the sensitivity from Fig. 3 gives with the 10% accuracy 136 and 280 mV per decade of F133 and F113 concentration, respectively. As it will be shown later in the text, the reaction of the F113 thermal decomposition is in equilibrium, and the Nernst equation is simple to understand. For F133, the analysis of the results presented in Fig. 4 lead to another reaction behavior. The reaction order of the decomposition was not investigated in detail, but the linear relation between the concentration and the reaction time is an argument for zero-order kinetics. The sensitivity of the MEIS sensors to F2 was measured in this paper and was about 120 mV/dec of the fluorine

194

IEEE SENSORS JOURNAL, VOL. 7, NO. 2, FEBRUARY 2007

Fig. 4. MEIS sensor responses as a function of time of the thermodecomposition reaction for 2000-ppm F133 (with a Ni temperature of 823 K) and F113 (with a Ni temperature of 623 K).

concentration. The sensitivity to hydrogen fluoride is only 55 mV/dec [10]. The comparison of these sensitivity data leads to the conclusion that fluorine is the main product of decomposition. For F133, one molecule of F2 is produced from Freon. There is no explanation for the high value of sensitivity for F113 of 280 mV/dec. The detailed investigation of the thermodecomposition of Freon will be done in future. The sensor signal of ∆U = 150 mV was observed at a concentration of 20-ppm fluorine in air. Thus, the partial pressure of the active component (calculated for F2 ) corresponds to 2% of the Freon pressure in the middle range of the Ni temperature. For the practical application of the external CE in the measuring sensor systems, the estimation of the value of the thermodecomposition reaction time necessary for the formation of detectable concentration of the product is important. The sensor signal as a function of the reaction time (t = reactor volume/gas flow rate) is shown in Fig. 4 for F113 and F133. In the case of F133, an increase in the reaction time did lead to an increase in the sensor signal. Taking into account that the change of the sensor signal is proportional to the logarithm of the concentrations, it easy to show that for F133, dependence between the concentration of the detected components and the reaction time is linear. Decreasing the sensor signal from 150 mV at a reaction time of 0.63 s to 70 mV at a reaction time of 0.16 s is a result of decreasing the concentration by a factor of 3.9 (log C1 /C2 = (150 − 70)/136). This value is very close to the ratio of the reaction times. For F133, the conditions are leading to dynamic control. In contrast to this, for F113, the equilibrium of the thermodecomposition reaction was observed, as proven by the nearly constant values in Fig. 4. The enthalpy ∆H of this reaction was estimated using the van Hoff equation, i.e., d ln K ∆H =− d(1/T ) R

(2)

where K is the equilibrium constant. Because of only 2% decomposition, the Freon concentration can be regarded as constant, and the change in K is proportional to the change in fluorine concentration. From the Nernst equa-

Fig. 5. MEIS sensor signal at 1000-ppm F113 in air as a function of the Ni destructor temperature.

Fig. 6. MIS sensor signal at 200-ppm H2 in air as a function of the Ni destructor temperature. Tsens = 403 K. Gas flow rate: 1.4 cm3 /s.

tion, ln(C) is ∆U divided by the sensitivity. The sensor signal ∆U versus the inverse Ni temperature is shown in Fig. 5. The value of the reaction enthalpy calculated from this dependence was equal to 40 ± 2 kJ/mol. B. Thermal Decomposition of Propane on Ni Destructor The interaction of propane C3 H8 with a metal catalyst in the presence of oxygen leads to the decomposition and oxidation of hydrocarbon. The oxidation results in the formation of water and carbon dioxide. The sensitivity of the MIS sensor to water vapor is rather small [11] and completely absent to carbon dioxide. As a result of a competition between the decomposition and oxidation processes of propane, the optimum temperature of the CE, at which the concentration of products registered by a sensor is maximum, is observed. To estimate the working range of the CE temperature and for calibration of the MIS sensor, the influence of the Ni destructor temperature on the initial rate of the bias voltage shift d∆U/dt under the action of 200-ppm hydrogen in air was analyzed (Fig. 6). The intensive hydrogen oxidation with an activation energy of about 18 kJ/mol starts at a CE temperature that is higher than 900 K.

FILIPPOV et al.: FLUOROCARBON MONITORING BY MIS SENSORS USING Ni CATALYTIC THERMODESTRUCTOR

Fig. 7. MIS sensor signals |∆U | (dotted line) and d|∆U |/dt (solid line) versus CE temperature. Propane concentration: −2000 ppm in air. Gas flow rate: 1.4 cm3 /s.

Fig. 8.

MIS sensor signal versus sensor working temperature.

The Pd−SiO2 −Si sensor signals ∆U and d∆U/dt versus the CE temperature for 2000-ppm propane in air after passing through the reactor are shown in Fig. 7. The maximum sensor signal corresponding to 200 ppm of hydrogen in air is observed at a CE temperature of about 1050 K. The working temperature of the sensor was 403 K. To determine the main product of the propane decomposition on Ni, the sensor signal as a function of working temperature of the gas-sensitive structure was investigated. The value of the activation energy ∆Eact was determined from
 ∆Eact d∆U = const · exp − . (3) dt kT For hydrogen interaction with the Pd−SiO2 −Si sensor, the activation energy is in the range of 0.25–0.3 eV [12]. The estimation of the value of ∆Eact from the temperature dependence d∆U/dt plotted in the Arrhenius coordinate for the product of the thermodecomposition of 2000-ppm propane in air at a CE temperature of 1000 K gives the value of 0.3 eV (Fig. 8). This result leads to the assumption that the main detected product of the C3 H8 thermodecomposition on Ni is hydrogen, with the decomposition efficiency being about 10% under the experiment conditions (a CE temperature of about 1050 K and a gas flow rate of 1.4 cm3 /s).

195

Fig. 9. MIS sensor signal d|∆U |/dt versus Ni destructor temperature in Arrhenius coordinates.

Fig. 10. MIS sensor signal on 2000-ppm propane in air at the impulse-heating Ni destructor from 873 K to 993 K. Gas flow rate: 1.4 cm3 /s.

The initial rate of the bias voltage shift d∆U/dt is proportional to the hydrogen concentration in air [13]. It was shown that at a CE temperature of about 1000 K, the sensor signal d∆U/dt and therefore the concentration of the detected gas (hydrogen) is proportional to the reaction time in the range of 0.2–1 s. The estimation of the activation energy of the propane thermal decomposition on Ni taken from the temperature dependence d∆U/dt plotted in Arrhenius coordinate gives the value of 320 ± 20 kJ/mol (Fig. 9). The typical values of the activation energy of the dissociation for different bonds of hydrocarbons are in the range of 350–460 kJ/mol [14]. The value of the reaction time for the formation of a detectable concentration of the thermodecomposition products is rather low. It gives the possibility of using the method of impulse CE heating. The result of the experiment with 2000-ppm propane is shown in Fig. 10. There is no sensor signal at a Ni destructor temperature of about 873 K. An impulse increase in a CE temperature of up to 993 K leads to the appearance of the MIS sensor response. The duration of the temperature impulse in the experiment was 20 s. The bias voltage shift ∆U did not reach the stationary value, but the concentration could be estimated from the initial rate of the bias voltage shift d∆U/dt. Practically, the pulse duration could be decreased in time, as is evident from the experiment The next impulse of the

196

IEEE SENSORS JOURNAL, VOL. 7, NO. 2, FEBRUARY 2007

temperature after signal relaxation (about 60 s) gives the same sensor signal. The application of the impulse-heating method allows elimination of the influence of gas tubes and decrease in power consumption.

[13] I. Lundström, M. Shivaraman, and C. Svensson, “Chemical reactions on palladium surfaces studied with Pd-MOS structures,” Surf. Sci., vol. 64, no. 2, pp. 497–519, May 1977. [14] G. Centi, F. Triéro, J. Ebner, and V. Franchetti, “Mechanistic aspects of maleic anhydride synthesis from C4 hydrocarbons over phosphorus vanadium oxide,” Chem. Rev., vol. 88, no. 1, pp. 55–80, 1988.

IV. C ONCLUSION The properties of Ni as a material for the external catalytic element were investigated. It was demonstrated that the MEIS and MIS sensors at their usual working temperature could detect the products of the thermodecomposition of Freon and propane, respectively. The nature of the reactions, which occur on the surface of the Ni catalyst, was investigated. It was shown that the kinetics of Freon thermodecomposition on Ni CE depends on the structure of the Freon molecule. The reaction of propane oxidation on Ni at a CE temperature of more than 1050 K caused the decrease in detectable product concentration. It is the reason for the maximum in the temperature dependence of the MIS sensor signal. The application of the CE impulse-heating method was demonstrated for propane thermodecomposition. Pure gases were used in the experiments, and no changing of the Ni properties was observed. However, for practical application (ambient atmospheric air) of the Ni thermodestructor, poisoning of the catalyst could occur. It imposes certain restrictions on the opportunities of the considered method and on possible fields of application of sensors with catalytically active gates.

Vladimir I. Filippov received the degree from Moscow Physical Engineering Institute, Moscow, Russia, in 1972, and the Ph.D. degree in 1994 from Russian Reseach Center (RRC) Kurchatov Institute, Moscow. He joined the Igor Vasilievich Kurchatov Institute of Atomic Energy, Moscow, where he investigated the influence of radiation on the electrical properties of the semiconductors. Since 1986, he has been working in the field of semiconductor sensors with the Institute of Molecular Physics of RRC Kurchatov Institute. His current research interest includes the experimental and theoretical research of MIS chemical sensors and heterogeneous catalysis.

R EFERENCES

Alexander A. Terentjev received the degree from Moscow Physical Engineering Institute, Moscow, Russia, in 1964. He was with I.V. Kurchatov Institute of Atomic Energy, Moscow, where he worked on the problems of atomic particle physics. He is currently with the Russian Research Center, Kurchatov Institute, Moscow, where he works on solid-state gas sensors.

[1] A. Arbab, A. Spetz, and I. Lundström, “Gas sensors for high temperature operation based on metal oxide silicon carbide (MOSiC) devices,” Sens. Actuators B, Chem., vol. 15/16, no. 1–3, pp. 19–23, Aug. 1993. [2] W. Moritz, V. I. Filippov, A. A. Vasiliev et al., “Semiconductor sensors for the detection of fluorocarbons, fluorine and hydrogen fluoride,” Anal. Chim. Acta, vol. 393, no. 1, pp. 49–57, Jun. 1999. [3] W. Hornik, “A novel structure for detecting organic vapours and hydrocarbons based on Pd-MOS sensor,” Sens. Actuators B, Chem., vol. 1, no. 1–6, pp. 35–39, Jan. 1990. [4] A. A. Vasiliev, V. N. Bezmelnitsyn, V. F. Sinianski, and B. B. Chaivanov, “Rate constants for heterogeneous dissociation of fluorine in a temperature range of 700–900 K on a nickel surface,” J. Fluorine Chem., vol. 95, no. 1/2, pp. 153–159, Jun. 1999. [5] V. I. Filippov, A. A. Terentjev, and S. S. Yakimov, “Electrode structure effect on the selectivity of gas sensors,” Sens. Actuators B, Chem., vol. 28, no. 1, pp. 55–58, Jul. 1995. [6] A. Spetz, M. Armgarth, and I. Lundström, “Hydrogen and ammonia response of metal–silicon dioxide–silicon structures with thin platinum gates,” J. Appl. Phys., vol. 64, no. 3, pp. 1274–1283, Aug. 1988. [7] W. Moritz, U. Roth, M. Heyde, K. Rademann, M. Reichling, and J. Hartmann, “Submicrosecond range surface heating and temperature measurement for efficient sensor reactivation,” Thin Solid Films, vol. 391, no. 1, pp. 143–148, Jul. 2001. [8] E. Kemnitz, A. Kohne, and E. Lieske, “Heterogeneous catalysed decomposition reactions of dichlorodifluoromethane in the presence of water on γ-alumina,” J. Fluorine Chem., vol. 81, no. 2, pp. 197–204, Mar. 1997. [9] V. I. Filippov, A. A. Vasiliev, J. Szeponik, and W. Moritz, “Room temperature hydrogen sensitivity of MIS-structure based on Pt/LaF3 interface,” IEEE Sensor J., vol. 6, no. 5, pp. 1250-1255, Oct. 2006. [10] W. Moritz, V. I. Filippov, A. A. Vasiliev et al., “Monitoring of HF and F2 using a field-effect sensor,” Sens. Actuators B, Chem., vol 24, no. 1–3, pp. 194–196, Mar. 1995. [11] V. Chaplanov, V. I. Filippov, and A. A. Terentjev, “Sensitivity of Pd/SiO2/Si sensor to humidity,” Sens. Actuators B, Chem., vol. 5, no. 1–4, pp. 187–188, Aug. 1991. [12] J. Janata and R. J. Huber, Eds., Solid State Chemical Sensors. New York: Academic, 1985, p. 38.

Werner Moritz was born in Berlin, Germany, in 1952. He received the Ph.D. degree from HumboldtUniversity of Berlin in 1981, working on the investigation of noble metal electrodes. In 1989, he received the habilitation degree for his study on chemical semiconductor sensors. He is currently working as the Leader of the Chemical Sensors Group with the HumboldtUniversity of Berlin. His research interest is in chemical sensors, thin films, and impedance imaging.

Alexey A. Vasiliev received the diploma degree from Moscow Institute of Physics and Technology, Moscow, Russia, in 1980, the Ph.D. degree for the “Study of the kinetics of low-temperature reactions of atomic fluorine by ESR method” from RRC Kurchatov Institute, Moscow, in 1986, and the Dr. of Science degree (habilitation) in solid-state microelectronics, for the investigation of “Physical and chemical principles of design of gas sensors based on metal oxide semiconductors and MIS structures with a solid electrolyte layer” in 2004 from Moscow Institute of Power Engineering. He has recently been working with the sensor group of the University Rovira i Virgili, Tarragona, Spain, and with the Russian Research Center Kurchatov Institute, Moscow. His research interests are related to the study of the kinetics and mechanisms of heterogeneous processes related to chemical sensing with design and prototyping of gas sensors and instruments based on these sensors.

Sergey S. Yakimov received the degree from Moscow University, Moscow, Russia, in 1962 and the Ph.D. degree in 1978. He is currently a Doctor of Science and a Professor of experimental physics with the Russian Research Center, Kurchatov Institute, Moscow. His current research interests include solid-state physics, Mössbauer spectroscopy, and semiconductor sensors.