Chemical imaging by direct methods

Thin Solid Films 436 (2003) 25–33 Chemical imaging by direct methods J. Mizsei* Department of Electron Devices, Budapest University of Technology an Economics, Goldmann Gy. ter 3, Budapest 1521, Hungary Abstract Chemical imaging is a very old principle of analysis. It has been used for many millions of years in biological tasting and smelling systems. Artificial versions of chemical imaging are traditional paper chromatography and the pH indicator strip used in chemical analysis, which directly result in real (visible) chemical images. In a recent article the indirect and direct chemical imaging methods are introduced, especially the evaluation methods for gas sensitive surfaces by direct pixelizing of the surface and interface potential changes. The scanned light pulse technique (SLPT) and the scanning vibrating capacitor are very effective tools for the chemical mapping of the surfaces. These methods are sensitive for the interface or surface potential shifts (adsorption induced work function shifts) respectively, that depend on the composition of adsorbed layers. Advantages and disadvantages, some technical limits, theory, practice and some results are discussed, too. 䊚 2003 Elsevier Science B.V. All rights reserved. Keywords: Gas sensors; Chemical picture; Olfactory sensors; Semiconductor sensors; Vibrating capacitor; Work function; Sensor array 1. Introduction Chemical imaging has been used for many millions of years in biological tasting and smelling systems: different receptors are separated and located in different places. We sense the sweet, bitter and sour at different areas of our tongue. Another example is the traditional paper chromatography in chemical analysis, which results in real chemical images. However, the terminology is a bit confused in the recent literature, as expressions such as ‘chemical pictures’, ‘chemical images’, ‘chemical mapping’ appear also in articles dealing with chemical analysis of solid states. The composition of solid surfaces can be characterised by many different methods, and the results are often interpreted in the form of chemical images. A well known example of this is the hardening of iron: the colour of the surface is characteristic of the heat treatment temperature. A similar effect has been observed during the heat treatment of Pd thin films w1x. Fig. 1a shows a photograph taken of a glass substrate, covered with a cathode sputtered Pd thin layer after asymmetrical heating, i.e. the temperature was only approximately 150 8C at the bottom edge and 450 8C at the top edge *Corresponding author. Tel.: q361-463-2715; fax: q361-4632973. E-mail address: [email protected] (J. Mizsei). of the substrate, respectively. Four different regions (amorphous, crystalline, partly oxidised and fully oxidised) can be distinguished by the colours from the bottom upwards in the photograph and also by the XRD patterns, measured from the corresponding four regions. Fig. 1b shows the lateral distribution of contact potential difference with respect to a graphite reference electrode measured from the layer in Fig. 1a by scanning vibrating capacitor. Values of the contact potential differences in Fig. 1b are in volts. Highest values of the work function are on the oxidised area at the top of the sample (PdO is a p-type semiconductor), while the lowest values are at the crystallised palladium region. The map also reveals the four different regions shown in the photograph in Fig. 1a. Our point of view connected with olfactory pictures is somewhat related to the previously discussed chemical pictures, but the purpose is different, namely, chemical analysis of the environment of the solid state. The composition of adsorbed layers depends on the environment around the solid state sensor surface and as it is shown later, some tools for surface mapping are useful also for making (direct) olfactory pictures. Selective gas analysis usually needs a multi-sensor system (sensor array) composed of sensors having different sensing characteristics w2,3x. Different sensor materials, temperatures and different activation (doping 0040-6090/03/$ - see front matter 䊚 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0040-6090(03)00522-4 26 J. Mizsei / Thin Solid Films 436 (2003) 25–33 of sensing layers) are the most common methods for building multi-sensor systems. If the temperature or the composition of the sensor, changes continuously then it can be considered as a distributed chemical sensor w4,5x. In many cases the chemical information is transformed into surface potential change. There are many different methods for reading out the information from the highly integrated surface potential sensors w6x. The multi-sensor system, together with a computer hardware and software (pattern recognition) yields an electronic nose w7–9x. Chemical images can be composed from the output signals of a multi-sensor system (discrete sensor array, based on four resistive type semiconductor gas sensors, w10x) by converting the different sensor responses to pixel elements (see Fig. 2, for very simple examples). Logarithmic ratio of resistances (ln RyRg values) have been used for the conversion, these numbers are related to the work function (chemical potential), changes in kTyq units at the surface w11x. Based on this example, the indirect method for generating an olfactory picture is the pixelizing of the sensor features. This way an olfactory video camera has been developed for a gasyodour flow visualisation system w12x. Practically the sensor array response is a data file, which can be converted into a picture. If the pattern recognition software determines the results, this data conversion is not very important, consequently, the olfactory picture may remain virtual. However, the direct method means the direct pixelizing of some sensor features. In this case, the response of the sensor array is a real picture. The simplest examples for direct chemical imaging are traditional paper chromatography and the pH indicator strip in the chemical analysis, which directly result in real (visible) chemical images. Recently there are many different methods for direct chemical imaging. A calorimetric sensor array has been constructed w13x and compared to other smell imaging Fig. 1. Results of surface analysis: (a) chemical picture from the sputtered and heat treated Pd surface (background photo behind the XRD diagrams), (b) work function map of the previously analysed Pd surface. The sample size is 30=40 mm. J. Mizsei / Thin Solid Films 436 (2003) 25–33 27 Fig. 1 (Continued). procedures w14x. Of course, the real and visible chemical picture can be converted back into electrical signal by a CCD camera w15x. The most advanced gene analysis chip technology is based on direct chemical imaging, too w16x. The recent article reports about the ‘state of art’ of the direct chemical imaging methods, especially the evaluation methods for gas sensitive surfaces by direct pixelizing of the surface and interface potential changes. ¨ and The basis of this job was worked out by Lundstrom co-workers, their pioneering works had started at the Pd gated FET hydrogen sensor w17x and continued through the last 30 years with the catalytic field-effect device w18x to olfactory imaging w19x. 2. Theoretical background Adsorption usually results in work function shifts on the catalytic surfaces. These shifts have strong effects Fig. 2. Olfactory (chemical) picture by indirect methode: 2=2 pixel pictures composed from responses of a discrete gas sensor array (Pd and Ag activated SnO2 resistive sensors). The logarithmic ratio of resistances is calculated and printed from the results of Ref. w10x, scale: blacks0, whites5. 28 J. Mizsei / Thin Solid Films 436 (2003) 25–33 Fig. 3. Band diagram of a vibrating capacitor-sensor(activated semiconductor)-insulator-semiconductor model system. Vibrating capacitor yields the contact potential difference (cpd, W12 or W13), MOS and SPV measurements yield the change in W4, and absolute value of W56 (potential barrier). on potential distribution in layered sensor structures, especially in the case of structures containing semiconductors. As the Debye length can be compared to the characteristic dimensions of the semiconductor layer or crystal, the space charge layer (surface potential barrier) in semiconductors are very sensitive on surface condition w20x. Fig. 3. shows the energy band diagram of a reference electrode-sensor (activated semiconductor)–insulator– semiconductor model system. This system contains all layers and interfaces which are important in the semiconductor gas sensor technology. Of course, it is not necessary using all these layers in the realised semiconductor gas sensor structures. Interface potential changes can be measured on these gas sensitive sensor (activated semiconductor)–insulator–semiconductor system by light excitation (surface photovoltaic method, SPV). In this case, the adsorption induced change in the metalyoxide or sensor layery oxide interface work function shift can be determined by measuring the photocapacitive current, generated by light pulses in the depletion layer. A good review of this method is given in w21x, but the original idea is a bit older w22x. In this system, the SPV gives the same information as C-V methods: the photocapacitive current is proportional with the W56 silicon interface potential barrier, which is related to the W4 work function at the oxide-sensor interface (see in the Fig. 3). If the semiconductor gas sensor layer is fully depleted, then W4 is affected by W3, too. This method needs transparent or semi-transparent reference electrode and conducting layers as adsorbing materials (special construction is needed), extrinsic semiconductor substrate (temperature range is limited), as well as good quality of Si–SiO2 interface (technological limit). The recent improvements of the SPV or rather SLPT w23x (scanning light pulse technique or LAPS, light addressable potentiometric sensor w6x) step over some of these limits. If the substrate is thin enough with low optical absorption coefficient, then the light excitation can be applied on the backside, too w24x. In this case the transparency of the reference electrode and gas sensor layer is not important. The temperature can be higher than the intrinsic temperature of silicon, using wide gap semiconductor substrate, for example SiC. The x–y resolution depends on the size of the optical excitation (spot dimensions and penetration depth), and is limited also by the diffusion length in the substrate. A detailed analysis of the LAPS spatial resolution is presented in w25x. Another surface voltage sensing device is the vibrating capacitor, which is old, however, very effective tool J. Mizsei / Thin Solid Films 436 (2003) 25–33 for the surface investigation. It can be considered as a combined mechanical-electrical excitation. Similarly to SPV, this method is also suitable for imaging the lateral distribution of the surface potential. The basis of this method is the compensation of the electric field (work function difference) between a vibrating electrode and the surface to be investigated w22x. At the compensated state the vibrating capacitance, induced alternating current is near zero, and the compensating voltage is equal to the work function difference (see in the Fig. 3, W12 or W13). The vibrating capacitor can be used together with light excitation for determining the absolute value of the surface potential barrier in the semiconductor substrate w21x. The main advantage of the vibrating capacitor method is that a very simple technology can be used for sample preparation, because it is not necessary to fabricate the whole semiconductor substrate based structure, containing gas permeable and transparent electrodes, only adsorbing surface layers are indispensable. Moreover, there are no strict limitations concerning the substrate, as well as the sheet resistance of the previously mentioned surface layer. Conducting layers and semiconductor films with sheet resistance values up to 1010y 1011 ohms are equally acceptable. About the applicable temperature range, it can be stated that there is no theoretical temperature limit for the surface to be investigated by scanning Kelvin probe. Of course, the practical realisation of the vibrating reference electrode limits the temperature; i.e. some component of the vibrating capacitor can be oxidised or decomposed at the higher temperature. The relatively low scanning rate limits the number of receptor materials. Thus, the theoretically existing advantages of the cpd mapping method can be exploited, using heat resistant materials in the vibrating capacitor head and improving the scanning rate. 3. Experimental methods and results As discussed above, the selective chemical analysis needs a sensor array composed from devices having different characteristics. There is an other way to produce a sensor array, which is the use of several different sensing and adsorbing materials (material gradient) on an asymmetrically heated substrate (temperature gradient), as has been demonstrated in some earlier work on olfactory pictures w18,19,26x. This kind of distributed sensor array, fits better to direct pixelizing methods by SPV (SLPT, LAPS) or scanning vibrating capacitor. The composition and temperature gradient yield a two dimensional arrangement (Figs. 4–6). The regularity of the temperature distribution and the adsorbing material arrangement are not essential conditions, the only requirement is the reproducible heterogenity of surface. However, the orthogonally arranged material composi- 29 tion and temperature gradient helps to distinguish the different receptor areas w28x, keeping them as far from each other as possible (Fig. 5). The general meaning of the ‘reproducible heterogeneity of the surface’, includes also the material gradient in two different directions with w28x or without temperature gradient w23x. Fig. 6. shows the experimental arrangement (a) and photograph (b) of the artificial olfctory ‘epithelium’, together with scanning vibrating capacitor head. The plastic xyy slipping box covers the integrated and distributed sensor system for adjusting the environmental gas composition. The total temperature difference is 100 8C among the ceramic substrate, this is the temperature gradient in the y direction. The artificial olfactory ‘epithelia’ are composed from Pd–Ag–Au–Pt–V–Pt– SnO2 and Pt–Pd–Cu–Fe–Ni–Pt stripes, this is the material composition gradient in the x direction. The structures and scan dimensions of artificial olfactory epithelia are shown in Fig. 5. Thin metal or semiconductor layers (Pd–Ag–Au–Pt–V–Pt–SnO2 and Pt–Pd–Cu–Fe–Ni–Pt) cover the ceramic plates. There are platinum heating resistor strips at the backside of ceramic substrata. All the thin layers have been produced by RF cathode sputtering. A 4-mm-diameter graphite (C) reference electrode scanned the surfaces, results are included also in this figure. 4. Discussion The SPV (Fig. 4) and vibrating capacitor scanned (Fig. 5) olfactory images are quite characteristic: the different material segments (stripes) can be clearly distinguished in most cases (see Fig. 5). There are remarkable deviations in the work function shifts over areas having different temperatures and surface materials. The work function shifts are the highest for H2 gas over the hot Pd and Pt covered areas, and there is no big difference among the Cu, Fe, Ni, Au and Ag covered areas. However, the saturated C2H5OH vapour results in a higher contrast between the Au and Ag covered areas (Fig. 5d). The above shown model systems are equivalent to 36 or 30 different ‘discrete’ sensors at least, i.e. six or five kinds of different materials over six different (4 mm diameter) areas at significantly different temperatures (regarding 15 K as a significant temperature difference), see Fig. 5. These ‘discrete sensor equivalent numbers’ can be improved by using higher temperature difference and a broader selection of chemically active adsorbing (receptor) materials. This would result in higher gradients; i.e. higher resolution mapping and lower heat conductivity substrate are necessary for higher temperature gradient. In the case of limited resolution and temperature gradient, the higher requirements can be fulfilled, using higher surface areas for the more receptor materials and high temperature differences. 30 J. Mizsei / Thin Solid Films 436 (2003) 25–33 Fig. 4. Experimental arrangements for SLPT: (a) with double material gradients w23x, (b) with material and temperature gradients w27x. Olfactory pictures are also presented: (a) for 250 ppm H2 flow, (b) for 400 ppm acetone and air, respectively. The doubled Pt stripes are redundant; this redundancy is an appropriate tool for checking the reproducibility. The material composition changes like a step function in the recently discussed arrangement, but it can be continuously changing as well. The cpd (contact potential difference) maps characterize the changes in the whole system, i.e. the work function shift on the reference electrode surface is also included in all chemical pictures. Thus, the effect of the adsorbed materials on the reference electrode does not disturb the effectiveness of this chemical imaging method. The scanning speed of the recent experimental arrangement is not very high, approximately 1 pys. These high resolution pictures contain 3000 pixels, therefore, this method is slow, it seems suitable for the evaluation of sensor materials, but may not be very practical for gas sensing. Using lower resolution, for example the same pixels as the above-mentioned ‘discrete sensor equivalent numbers’, the picture can be obtained within several minutes. The sensitivity of the SPV and cpd mapping is in the similar range, as the conventionally used MOS gas sensors, because the adsorption induced work function shift is the primary effect in both cases. 5. Conclusions Chemical imaging generally means the treatment of these sensor features as elements (pixels) of a picture. The hyperspace of chemical sensor features (response J. Mizsei / Thin Solid Films 436 (2003) 25–33 31 Fig. 5. The structures and scan dimensions of artificial olfactory epithelia and the 4 mm diameter graphite (C) reference electrode for the scanning vibrating capacitor surface mapping: (a) the Pd–Ag–Au–Pt–V–Pt–SnO2 thin film stripes, (b) the adsorption induced shifts in the work function differences (at 460–360 K top-bottom of the Pd–Ag–Au–Pt–V–Pt–SnO2 strips for 1% H2 -air mixture), (c) section among the top line (surface voltage shifts for 1% H2 –air mixture vs. pixels), (d) response for the saturated C2 H5 OH vapour, (e) section among the top line (surface voltage shifts vs. pixels for C2H5OH vapour), (f) the Pt–Pd–Cu–Fe–Ni–Pt thin film stripes, (g) the adsorption induced shifts in the work function differences (at 460–360 K top-bottom of the Pt–Pd–Cu–Fe–Ni–Pt strips for 1% H2 –air mixture), (h) response for the saturated CHCl3 vapour (at 460–360 K top-bottom). 32 J. Mizsei / Thin Solid Films 436 (2003) 25–33 Fig. 6. The experimental arrangement (a) and photograph (b) of the artificial olfactory ‘epithelium’ together with scanning vibrating capacitor head. signal, temperature, surface doping etc.) has a huge dimensionality w2x. The recently reviewed direct chemical imaging methods, use only three sensor features (work function shifts of the different surface segments, different materials and different temperatures). Some of these images may still have a certain degree of redundancy, as the ‘discrete sensor equivalent numbers’ are lower than the pixel numbers of the pictures. The main advantages of direct chemical imaging methods are that the pictures may contain very large amount of information and they can easily be processed by image processing software (image arithmetic, pattern recognition, Fourier transformation, etc.). Further advantages are its simplicity, as well as clearness and suggestiveness. The system having material and temperature gradients can serve as the basis (artificial olfactory ‘epithelium’) of electronic noses. Moreover, these two-dimensional arrangements can help in selecting the best gas sensitive materials, too, as many different materials and temperatures can be tested and compared in one picture. Evaluation of the picture may help to build better and better receptor systems, which may result in quick evolution of the artificial olfactory ‘epithelium’ technique in the near future. Acknowledgments This study was financed by the Academy of Finland (Grant No. 37778) and supported by the Hungarian Science Foundation (OTKA No. T34739), and the Hungarian Ministry of Education (FKFP 0064y1999, Pr. No. 502-121). References w1x J. Mizsei, J. Voutilainen, S. Saukko, V. Lantto, Thin Solid Films 391 (2001) 209. w2x W. Gopel, ¨ Sensors Actuators B65 (2000) 70. w3x U. Weimar, W.Gopel, ¨ Chemical Imaging: Trends in Multiparameter Sensor Systems, Proceedings of XI European Conference on Solid-State Transducers, 2, 1997, 527. w4x T. Eklov, ¨ H. Sundgren, I Lundstrom, ¨ Sensors Actuators B45 (1997) 71. w5x T. Eklov, ¨ ¨ I. Lundstrom, Sensors Actuators B57 (1999) 274–282. w6x M. George, W.J. Parak, H.E. Gaub, Sensors Actuators B69 (2000) 266. w7x J.W. Gardener, H.W. Shin, E.L. Hines, Sensors Actuators B70 (2000) 19. w8x H.T. Nagle, R.G. Osuna, S.S. Schiffmann, IEEE Spectrum 9 (1998) 22. w9x R.R. Das, K.K. Shukla, R. Dwivedi, A.R. Srivastrava, Microelectron. J. 30 (1999) 793. J. Mizsei / Thin Solid Films 436 (2003) 25–33 w10x J. Mizsei, Sensors Actuators B18–19 (1994) 264. w11x J. Mizsei, Sensors Actuators B28 (1995) 129. w12x H. Ishida, T. Tokuhiro, T. Nakamoto, T. Moriizumi, Sensors Actuators B83 (2002) 256. w13x N.A. Rakow, K.S. Suslick, Nature 406 (2000) 710. w14x I. Lundstrom, ¨ Nature 406 (2000) 682. w15x R. Tantra, J. Cooper, Sensors Actuators B82 (2002) 233. w16x K. Takahashii, K. Seio, M. Sekine, O. Hino, M. Eshashi, Sensors Actuators B83 (2002) 67. w17x I. Lundstrom, ¨ Sensors Actuators 1 (1981) 403, continued in 2 (1981y82) 105. w18x I. Lundstrom, ¨ C. Svensson, A. Spetz, H. Sundgren, F. Winquist, Sensors Actuators B13–14 (1993) 16. w19x I. Lundstrom, ¨ R. Erlandsson, U. Frykman, E. Hedborg, A. Spetz, H. Sundgren, S. Welin, F. Winquist, Nature 352 (1991) 47. w20x w21x w22x w23x 33 J. Mizsei, Electron Technol. 33 (2000) 236. L. Kronik, Y. Shapira, Surf. Sci. Rep. 37 (1999) 1. W.H. Brattain, J. Bardeen, Bell Syst. Techn. J. 32 (1953) 1. M. Lofdahl, M. Eriksson, I. Lundstrom, ¨ ¨ Sensors Acuators B70 (2000) 77. w24x E.A. De Vasconcelos, H. Uchida, W.Y. Zhang, T. Katsube, Jap. J. Appl. Phys. 38 (1999) 2893. w25x M. George, W.J. Parak, I. Gerhardt, W. Moritz, F. Kaesen, H. Geiger, I. Eisele, H.E. Gaub, Sensors Actuators A83 (2000) 187. w26x J. Mizsei, Sensors Actuators B48 (1998) 300. w27x Q. Zhang, P. Wang, J. Li, X. Gao, Biosens. Bioelectron. 15 (2000) 249. w28x J. Mizsei, S. Ress, Sensors Actuators B83 (2002) 164.