Analytica Chimica Acta 437 (2001) 183–190 Re-activation of an all solid state oxygen sensor W. Moritz a,∗ , S. Krause b , U. Roth a , D. Klimm c , A. Lippitz d a Walther Nernst Institute of Physical and Theoretical Chemistry, Humboldt-University Berlin, Bunsenstrasse 1, 10117 Berlin, Germany b Department of Chemistry, University of Sheffield, Sheffield S3 7HF, UK c Institute of Crystal Growth, Berlin, Germany d Federal Institute for Materials Research and Testing, Berlin, Germany Received 28 November 2000; received in revised form 8 March 2001; accepted 14 March 2001 Abstract The silicon based semiconductor structure Si/SiO2 /Si3 N4 /LaF3 /Pt can be used as a potentiometric oxygen sensor working at room temperature. A thermal re-activation can be applied to overcome the earlier disadvantage of an increase in response time with continuous use. Using the Pt gate electrode as a resistive heater, very short electrical high-power pulses can be applied. A heating time as short as 300 ns was sufficient for the re-activation of the sensor. This way, only the sensitive thin layer system LaF3 /Pt was heated, and the whole sensor was at room temperature immediately after heating. Impedance spectroscopy, X-ray photoelectron spectroscopy (XPS) and quadruple mass spectrometric (QMS)–thermogravimetry (TG) were used to investigate the mechanism of deterioration in dynamic sensor behaviour and re-activation. The formation of hydrated carbonate and the desorption of CO2 and H2 O have been shown to be the causes. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Chemical sensor; Thermal treatment; Activation; LaF3 ; Pt 1. Introduction There are several oxygen sensors using different principles on the market. However, they either need high temperature operation or liquid components [1–3]. An all-solid state sensor operated at room temperature was suggested by Miura et al. [4] using the reaction at the LaF3 /Pt interface. A semiconductor sensor using the field effect in silicon was developed based on the same sensitive interface in a structure Si/LaF3 /Pt or Si/SiO2 /Si3 N4 /LaF3 /Pt [5]. A sensitivity according to the Nernst equation of 58 mV per decade of oxygen partial pressure p(O2 ) was determined. A ∗ Corresponding author. Tel.: +49-302-093-5566; fax: +49-302-093-5559. E-mail address: [email protected] (W. Moritz). detailed investigation of the sensor behaviour is given in [6]. X-ray photoelectron spectroscopy (XPS) measurements, thermodynamic data and kinetic investigations lead to a suggestion for the sensor mechanism [7]. In a first step, oxygen adsorbs at surface sites A∗ at the three phase boundary LaF3 /Pt/gas (1). This is followed by the rate-determining step, a one-electron reduction of oxygen in the presence of water (2). The well-known mobility of OH− on fluoride sites in the LaF3 lattice (3) stabilises the OH− produced in reaction (2) shifting the equilibrium of reaction (2) to the right. O2 + A∗ O2 (A∗ ) (1) O2 (A∗ ) + H2 O + e− HO2 • (A∗ ) + OH− (2) OH− + (F∗ ) OH− (F∗ ) (3) 0003-2670/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 3 - 2 6 7 0 ( 0 1 ) 0 0 9 9 3 - X 184 W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 sensor surface to a constant temperature on this time scale [8]. In this paper, nanosecond range heating is used for sensor re-activation. The activation energy of the process was determined and several methods such as impedance spectroscopy, thermogravimetry (TG) and XPS were used to investigate the mechanism of sensor ageing and re-activation in more detail. Fig. 1. Schematic of the sensor structure. Although, the oxygen sensor showed several promising properties such as Nernst sensitivity and a fast response, a considerable decrease in response rate was observed within a few days. A solution to this problem was achieved by a simple thermal treatment. Typically the sensor structures were heated on a hot plate for 10 s at 300◦ C and cooled own for 5 min [7]. This re-activation procedure can be repeated without limitations, and therefore, a lifetime of the sensors of at least 8 months was shown. The influence of the ambient gas was investigated in [7] showing that the ageing of the sensor was faster with increasing humidity and oxygen concentration while in vacuum no ageing was observed [6]. An advantageous solution to the thermal treatment was realised by using the Pt gate electrode of the structure shown in Fig. 1 as a heating resistance. The re-activation was shown to be a fast process, and therefore, electrical heating pulses as short as 100 s enabled the sensor re-activation. It is an advantage of short heating pulses applied to the Pt layer as a part of the oxygen sensitive LaF3 /Pt thin layer system, that mainly this region of the sensor is at high temperature while the much thicker Si substrate remains nearly at room temperature. As a consequence the cooling of the sensor is very fast and measurements can be continued immediately. XPS and AFM investigations have been carried out to explain the mechanism of the ageing and re-activation, but only the Pt/La surface atomic ratio showed a correlation to the alternating ageing and activation processes. This was interpreted by reversible changes in the geometry of the three-phase boundary LaF3 /Pt/gas due to surface diffusion of Pt caused by the strong adsorption of species from the gas phase [7]. Recently, we developed a method for the measurement of the surface temperature in the nanosecond range and heating the 2. Experimental The sensor structures were prepared by thermal evaporation of 240 nm LaF3 on a n-Si/SiO2 (60 nm)/Si3 N4 (80 nm) substrate (5 mm × 10 mm) (donor concentration 2.5 × 1014 cm−3 ; IMS Dresden, Germany). The dc cathode sputtering in an argon plasma through a metal mask was used to deposit 60 nm thick Pt films in the geometry shown in Fig. 1. The small path in the centre was heated and had dimensions of 0.5 mm × 1.5 mm. The large areas on both sides were used as contact pads. The high temperature contact adhesive EPO-TEK P-1011 (Polytec) was used to provide a stable contact between the Pt in the enlarged area and a copper strip. A contact resistance of less than 0.1 was achieved. High current POGO contacts of Everett Charles technologies were used for the current input via the copper strips. A four-point resistance measurement was achieved by the measurement of the voltage drop in the path region at two needle shaped POGO contacts with a distance of 1 mm. Therefore, the resistance was determined directly using the heating current and the effective voltage. High power pulses were generated using a combination of the computer controlled high power supply BERTAN 606C-15P, the pulse generator PHILIPS PM 5712 and the fast high voltage transistor switch HTS 31 of BEHLKE. The voltage and current were recorded using the digital oscilloscope Le Croy 9450 A. This Equipment reaches pulse times down to 10 ns and a heating power up to 100 kW (s range). Photocurrent/voltage measurements were used for determining the oxygen sensitivity. Only the area of the small path of Pt film shown in Fig. 1 was illuminated ensuring sensitivity measurements only resulting from the area heated before. An IR-laser diode with λ = 850 nm was modulated at 10 kHz. The photocurrent was detected using a lock-in amplifier W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 SR 830 DSP from Stanford Research Systems. The oxygen sensitivity was determined in oxygen/nitrogen mixtures in the concentration range from 0.3 to 100% oxygen with a relative humidity ranging from 5 to 80%. A Zahner IM5d system was used for impedance spectroscopy. The XPS-spectra were obtained by employing a VG ESCALAB 200X electron spectrometer using Mg K␣ X-ray radiation (15 kV, 300 W). The electron analyser was used in the constant analyser energy mode (CAE 10). The pressure in the analysis chamber was in the 10−7 Pa range. For the analyser calibration, the method proposed by Anthony and Seah [9] was applied. Binding energy data were referenced to the aliphatic C 1s peak at 285.0 eV [10]. Simultaneous differential thermoanalytic (DTA), TG and quadruple mass spectrometric (QMS) measurements were performed using a NETZSCH STA 409C thermoanalytic device with sample carrier HIGH RT 2 (Pt-Pt90/Rh10 thermocouples). The samples (LaF3 powder or LaF3 /Pt layers on Al foil) were placed in corundum crucibles without lid. The heating with rates of typically 5 K/min in the temperature range from 20◦ C ≤ T ≤ 350◦ C in argon was performed using a graphite furnace. A vitreous carbon skimmer coupling with intermediate vacuum managed the gas flow from the sample room to the QMS to maintain a constant pressure of 10−5 mbar within the QMS. 3. Results and discussion 3.1. Sensor re-activation For the re-activation of the oxygen sensor, thermal treatments using an external heater and in situ heating by electrical high power pulses applied to the gate electrode of the sensor were performed. The electrical heating and the surface temperature measurement at the sensor structure shown in Fig. 1 were described in the experimental part and in more detail in [8]. It was proven that the resistance change of the Pt film could be used for a fast measurement of the surface temperature down to the nanosecond range. For fundamental investigations, e.g. the determination of activation energy of the re-activation process, electrical heating at a constant voltage is disadvantageous in that there is 185 Fig. 2. Surface temperature measurement at the sensor structure while electrical heating is applied, maximum voltage 293 V, maximum current 8.4 A. no constant temperature but a continuous increase in temperature. Therefore, it was necessary to develop a specific heating power profile, which lead to a constant temperature at a defined point in time during the heating period. An example for sensor heating and temperature measurement is given in Fig. 2. The temperature increased for 700 ns in this experiment and then remained constant in the narrow range between 703 and 708 K for a period of 310 ns. Even for this very short time a re-activation of the sensor was successfully proven. Energy of only 1.8×10−3 J was sufficient for the thermal reactivation, because only the thin sensitive layers, LaF3 and Pt, were heated while the thick Si-substrate remained at room temperature [8]. This low energy consumption is especially advantageous for battery powered sensor systems. The dynamic response of a sensor after reactivation for 320 ns is compared to the behaviour before activation in Fig. 3. The change of the equilibrium potential with the oxygen concentration is given in Fig. 4 for a reactivated sensor. The sensitivity was determined to be 58.1 ± 0.9 mV per decade of oxygen partial pressure p(O2 ). The sensor response time was proven to be independent of the oxygen concentration. For the control and the understanding of the mechanism of the re-activation process in the thin layer system, knowledge of the activation energy is important. However, the activation energy cannot be 186 W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 Fig. 3. Response of the oxygen sensor to changes in the oxygen concentration (as indicated in the graph): (a) after electrical heating (320 ns constant temperature range of 520 K); (b) 25 days after last reactivation. determined at the final state of activation but should be determined during the process of re-activation. Therefore, the process was divided into several steps, each leading only to a partial improvement in the sensor response dynamics. An example for such a set of measurements is given in Fig. 5 (only three curves are given for simplification). Sensors aged for a long period were used for this experiment. Before activation, there was nearly no response to a change in gas concentration. After heating the sensor at 120◦ C for 1 min, a response to the change in gas concentration was observed, but it was rather slow. A prolongation Fig. 4. Sensitivity of the oxygen sensor after reactivation (30% relative humidity, oxygen/nitrogen mixtures). Fig. 5. Response of the oxygen sensor to changes in the oxygen concentration (as indicated in the graph) after thermal treatment at 120◦ C for 1, 3 and 9 min. of heating time lead to the improvements shown. From the response curves a response time t50 (the time necessary for 50% of potential change) was determined. In a next step these t50 values were plotted against activation time. From this plot we interpolated to an activation time that would lead to t50 = 20 s, which was defined as the reference sensor state. For activation on an external heater, the temperature was kept constant and the heating time varied. However, for thermal treatment with electrical heating pulses, the total time of the single pulses was kept constant and the heating voltage (and hence, the temperature) was increased from pulse to pulse in order to simplify the experiment. In this case, the response time–activation temperature relationships were interpolated to the temperature required to achieve t50 = 20 s in order to obtain the activation times at different temperatures. The resulting activation time and temperature values are given in Fig. 6. Since the initial and the final states of the sensor were identical for all experiments, the reciprocal of the activation time corresponded to a rate constant and the slope in Fig. 6 leads to an activation energy according to the Arrhenius equation. The activation time varied by nearly 10 orders of magnitude in our experiments, while the temperature was in the relatively narrow range of 120–250◦ C only. As a consequence a very high value of the activation energy E A = 262 ± 28 kJ/mol was determined. Due to the high value of the activation energy, re-activation can be performed at times even shorter than 300 ns W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 Fig. 6. Dependence of the logarithm of the reciprocal re-activation time (method of determination see text) on the reciprocal of the re-activation temperature. at only marginally higher temperatures. Since no further improvement of the reactivation procedure was expected, this time range was not investigated. 3.2. Impedance spectroscopy Impedance spectroscopy is a valuable tool for characterisation of electrochemical processes. A detailed investigation of the impedance behaviour of the LaF3 /Pt interface used as an oxygen sensor was given in [6]. A single crystal of LaF3 was used in order to eliminate the dominating capacitance of the insulator in the semiconductor sensor. It was shown that the results obtained with a LaF3 single crystal were transferable to those obtained with a polycrystalline layer of LaF3 . The interface LaF3 /Pt could be described by a typical combination of a capacitance and a resistance in parallel. The capacitance was determined to be 86 ± 21 F/cm2 for a polished crystal, which was a typical value for a double layer capacitance. Therefore, the resistance in parallel could be understood as the charge transfer resistance of the potential forming process. The response rate of the oxygen sensor was not influenced by the oxygen concentration [7]. In agreement with this, the charge transfer resistance was also found to be independent of the oxygen concentration. The results of the investigation into the long-term behaviour and the re-activation are presented in Fig. 7. 187 Fig. 7. Response time t90 (upper part) and exchange resistance R (lower part) determined at a single crystal LaF3 /Pt interface at different times after preparation (䊏) and after re-activation (䊐). The response time t90 increased with time and decreased again after re-activation. A similar behaviour was observed for the charge transfer resistance R. It can, therefore, be concluded that re-activation and ageing phenomena of the sensor can be attributed to changes in the rate of an electrochemical charge transfer process. 3.3. Differential TG The re-activation of the oxygen sensor might be due to desorption of gases from the surface or from the bulk of the thin layer system. TG combined with QMS proved to be a suitable tool to examine this hypothesis. Investigations at Si/SiO2 /Si3 N4 /LaF3 /Pt structures did not show any reasonable result. This might be due to the unfavourable mass ratio Si (500 m) to LaF3 (240 nm) and Pt (60 nm). Therefore, in a next set of experiments, the LaF3 and Pt layers were deposited onto an Al foil of 0.8 m thickness. An area of 84.3 cm2 of this multi-layer system was investigated. Small but unambiguous signals were observed for mass numbers 1, 18 and 44 corresponding to H, H2 O and CO2 , respectively. In Fig. 8 the temperature dependence for M = 44 is given. A maximum was found at 128◦ C. For M = 1 and 18, the maximum is in the same temperature range. For a heating rate of 5 K/min as used in this experiment, the thin layer system under investigation was in the temperature range 120– 140◦ C for several minutes. This concurred with the 188 W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 Fig. 8. QMS signals for M = 44 (CO2 , dashed lines) and M = 18 (H2 O, solid lines): (1) LaF3 (240 nm)/Pt (60 nm) thin layer system prepared on Al foil (0.8 m) (outer scales and curves with dots); (2) LaF3 powder (inner scales and curves without dots). observation that sensor re-activation took several minutes in this temperature range. Therefore, the gas desorption observed could be correlated to the re-activation process. Gas adsorption (or a chemical reaction) can occur at either the LaF3 or the Pt. Therefore, LaF3 powder was also investigated. The result is shown in Fig. 8 for M = 44 and 18. The width of the signal was greater here, but it appeared at the same temperature as for the thin layer system. The QMS peaks shown in Fig. 8 were accompanied by a drop of the TG signal of about 0.1%. Hence, for the thin layer system, the source of desorption is the LaF3 . A difference between the powder and the thin layer system was, that for the powder the ratio of desorbed H2 O to desorbed CO2 was about 6:1 while it was 1:1 for the thin layer system. This variation seemed to be caused by the presence of a Pt layer, which was shown to change the surface chemistry, e.g. by production of OH− (reaction (2)). Using the thermogravimetric data, a mass loss of 6× 10−8 g/cm2 at 128◦ C was calculated for the thin layer system. This represents the sum of the H2 O and the CO2 desorption signal. With respect to the LaF3 layer this mass loss represents 0.05 mass% of the 240 nm thick layer, i.e. it probably originates from a very thin surface film. 3.4. Photoelectron spectroscopy XPS of the LaF3 /Pt thin layer system was used for investigating the re-activation process. Apart from the main elements (La, F and Pt), oxygen and carbon were found. In the La 4d spectrum (Fig. 9a), the peak visible at 104.0 eV could be interpreted as a mixed-phase of LaF3 and lanthanum bound to oxygen. This oxygen-containing phase was located near the surface, because with increasing take-off-angle Fig. 9. Photoelectron spectra of the sensor surface; La 4d signal: (a) take-off-angel θ = 15◦ ; (b) take-off-angel θ = 60◦ ; (c) difference between (a) and (b). W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 Fig. 10. Photoelectron spectra of the sensor surface; C 1s signal: (a) surface exposed to air for several days; (b) measurement directly after thermal re-activation at 300◦ C. an increase of the intensity was observed. (Fig. 9b). This effect was expressed better in the difference spectrum (Fig. 9c). However, changes in this signal could not be related to the change in sensor kinetics. Therefore, the C 1s signal was studied in more detail. For sensors aged for several days, a signal at a binding energy of 289 eV was found which disappeared during the course of measurement (Fig. 10a). This signal was not observed for sensors measured directly after re-activation (Fig. 10b). The binding energy measured can be related to carbonate ions, which decomposed during the measurement [11]. The heating of the sample during the measurement (estimated to be near 70◦ C) and the additional influence of vacuum and X-rays can explain the fact that the carbonate signal was only observed in the beginning of the measurement. 4. Conclusions It was shown that the re-activation of an oxygen sensor based on the LaF3 /Pt interface reaction can be accomplished in very short times down to the nanosecond range. The advantage of very short electrical heating pulses is that the sensor can be used directly after activation because only the sensitive surface region is heated. Furthermore, energy consumption is reduced, which is advantageous for battery powered applications. 189 The very high value of 262 kJ/mol for the activation energy determined for the re-activation process is the reason for the possibility of ultra-fast activation at moderate temperatures. Furthermore, it improves the understanding of the mechanism of the reactivation process. The results of impedance spectroscopy reveal that an electrochemical charge transfer process at the LaF3 /Pt interface determines the sensor behaviour. This process slows down with continuous use of the sensor and is accelerated again after thermal treatment. A change in the geometry of the three-phase boundary as suggested in [7] can influence the rate of an electrochemical process. This hypothesis was connected with a diffusion of Pt atoms on the surface. But activation energies determined for surface diffusion of Pt are near to 25 kJ/mol [12,13] and even for clusters such as Pt7 the maximum value given is 113 kJ/mol [14]. These much lower activation energies indicate that the surface diffusion of Pt and the changes in geometry of the three-phase boundary are not responsible for ageing and re-activation of the sensor. TG and XPS data indicate that CO2 and H2 O desorption are connected with the re-activation process. Although, the oxygen concentration was shown to influence the sensor ageing, no change in the surface oxygen concentration related to the re-activation was observed. Hence, the influence of oxygen may be due to an increase in surface pH according to reaction (2), which is promoting the production of surface carbonate. For Lanthanum carbonates, different hydrated carbonates as well as oxide and hydroxide carbonates are known [15]. The decomposition of the carbonates was observed at temperatures higher than 500◦ C. On the other hand, a loss of CO2 was observed simultaneously with the dehydration of hydrated carbonates. This process was observed at temperatures between 127 and 160◦ C [16]. For the activation energy of the decomposition of the lanthanum carbonate, values between 209 and 345 kJ/mol were found [17]. At the surface of the hygroscopic LaF3 , the formation and decomposition of a carbonate may differ from the behaviour of pure phases, but nevertheless it is noteworthy that the very high value of 262 kJ/mol for the activation energy of the re-activation process corresponds to the values for the decomposition of the carbonate. Therefore, hydrated carbonate incorporation in the surface of the LaF3 is assumed to be the reason for a 190 W. Moritz et al. / Analytica Chimica Acta 437 (2001) 183–190 decrease in the rate of the potential determining reaction and the increase in response time of the sensor. The poisoning of the surface as it is known for several catalysts was only avoided in vacuum while even in very pure gases a small amount of CO2 was sufficient to cause the sensor to age. Because of the electrical re-activation procedure developed this poisoning is not a relevant disadvantage of the sensor anymore. Acknowledgements The financial support of DFG is gratefully acknowledged. References [1] T. Takeuchi, Sens. Actuators 14 (1988) 109–124. [2] C. Nylander, J. Phys. E.: Sci. Instrum. 18 (1985) 736–750. [3] W. Göpel, J. Hesse, J.N. Zemel (Eds.), Sensors — A Comprehensive Survey, Vol. 2/3, VCH, Weinheim, 1991. [4] N. Miura, J. Hisamoto, N. Yamazoe, Sens. Actuators 16 (1989) 301–310. [5] S. Krause, W. Moritz, I. Grohmann, Sens. Actuators B 9 (1992) 191. [6] S. Krause, Dissertation, Humboldt-University Berlin, Berlin, 1994. [7] W. Moritz, S. Krause, I. Grohmann, Sens. Actuators B 18 (1994) 148. [8] W. Moritz, U. Roth, M. Heyde, K. Rademann, M. Reichling, J. Hartmann, Thin Solid Films, in press. [9] M.T. Anthony, M.P. Seah, Surf. Interf. Anal. 6 (1984) 95. [10] G. Beamson, D. Briggs, THE SCIENTA ESCA300 Database, Wiley, New York, 1992. [11] NIST X-Ray Photoelectron Spectroscopy Database, Version 2.0. [12] M. Bott, M.T. Hohage, M.M. Michely, G. Comsa, Phys. Rev. Lett. 76 (1996) 1304. [13] F. Maca, M. Kotrla, O.S. Trushin, Vaccum 54 (1999) 113. [14] K. Kyuno, G. Ehrlich, Surf. Sci. 437 (1999) 29. [15] H. Bergmann (Ed.), Gmelin Handbook of Inorganic and Organometallic Chemistry, Rare Earth Elements, Vol. C12b, Springer, Berlin, 1994. [16] H. Bergmann (Ed.), Gmelin Handbook of Inorganic and Organometallic Chemistry, Rare Earth Elements, Vol. C12b, Springer, Berlin, 1994, p. 84. [17] H. Bergmann (Ed.), Gmelin Handbook of Inorganic and Organometallic Chemistry, Rare Earth Elements, Vol. C12b, Springer, Berlin, 1994, pp. 39–40.
© Copyright 2026 Paperzz