Materials Science and Engineering B69 – 70 (2000) 210 – 214 www.elsevier.com/locate/mseb
NO2 monitoring at room temperature by a porous silicon gas sensor Luca Boarino a,*, C. Baratto b, F. Geobaldo c, G. Amato a, E. Comini b, A.M. Rossi a, G. Faglia b, G. Le´rondel a, G. Sberveglieri b a
Thin Film Laboratory, Istituto Elettrotecnico Nazionale Galileo Ferraris, Strada delle Cacce 91, 10135 Turin, Italy b Gas Sensor Laboratory, Uni6ersity of Brescia, Via Valotti 9, 25123 Brescia, Italy c Department of Chemistry and Materials Science, Politecnico of Torino, Corso Duca degli Abruzzi 24, 10128 Turin, Italy
Abstract A study on reactivity of p+ porous silicon layers (PSL) to different gas atmosphere has been carried out. Substrate doping was 5–15 mV cm and 0.5 V cm, porosity ranged from 30 to 75% and the thickness of the porous layers was 20 – 30 mm. Three different processes to insure good electrical contact are proposed and discussed. PSL were kept at constant bias and current variations due to interaction with different concentrations of NO2 were monitored at constant relative humidity (R.H.). Measurements were performed at room temperature (R.T.) and at atmospheric pressure. Concentrations as low as 1 ppm were tested, but the high sensitivity of the sensor makes possible to test lower values. The recovery time of the sensor is of the order of one minute. Response to interfering gases (methanol, humidity, CO, CH4, NO, NO2) has been examined also. In-situ FTIR spectroscopy in NO2 atmosphere shows a fully reversible free-carrier detrapping in the IR region, confirming the validity of the models proposed in the recent past for electrical conduction in mesoporous silicon. © 2000 Elsevier Science S.A. All rights reserved. Keywords: Porous silicon; Gas sensor; NO2; Electrical conductivity; FTIR
1. Introduction Nitrogen oxides generated by combustion are dangerous to health (particularly NO2) and their alarm levels in urban environment are limited by Italian legislation to 0.2 ppm for NO2 and 0.32 ppm for NO. So, there is a strong demand for NOx cheap and reliable sensors. Porous silicon (PS) is a quite new material in the field of gas sensors, except for its well-known sensitivity to humidity [1]; it is very interesting for its high surface to volume ratio and reactivity to the environment. Many features of this material have been used for gas detection, such as work function [2], refractive index [3], photoluminescence and conductivity variation [4], the latest being the easiest way to realize a gas sensor. Transport in PS has been studied by several authors in the last years to achieve higher electroluminescence efficiency. Describing electronic transport phenomena * Corresponding author. Tel.: +39-11-3919627; fax: + 39-113919621. E-mail address:
[email protected] (L. Boarino)
in PS, a comprehensive picture is the so-called ‘pea pod’ model, [5] which includes two different effects. The first is conduction among silicon crystallites attributed to simple transport [6], tunnelling [7] with great similarity to granular metals [5], and the second is a conduction pathway in the amorphous Six Hy Oz matrix in which crystallites are embedded. This last mechanism has been suggested to be band conduction [8], activated hopping in band tails [9], activated deep states hopping [10], Pool Frenkel process [11] and activated hopping in a fractal network [12]. In the case of p+ mesoporous silicon, with typical structures of 10 nm, quantum confinement is not evident, but strong changes of resistivity in the presence of polar liquids, vapours and gases has been observed. [13] In this work first results of electrical and FTIR characterisation of p+ mesoporous silicon in the presence of NO2 are presented, offering a clear confirmation of the proposed conduction mechanisms [14], [15] and opening new perspectives of investigation of molecular dynamics at the surface of PSL. Three different technological processes are also proposed for the prototyping of the gas sensor.
0921-5107/00/$ - see front matter © 2000 Elsevier Science S.A. All rights reserved. PII: S 0 9 2 1 - 5 1 0 7 ( 9 9 ) 0 0 2 6 7 - 6
L. Boarino et al. / Materials Science and Engineering B69–70 (2000) 210–214
211
Fig. 1. Electrical contact structure realised on p+ PSL for gas sensor prototype. A patterned deposition of Ti – W oxide is performed by sputtering, then an adhesion layer of Ti–W, and Pt stripes are deposited to carry the contacts out from the porous silicon area for bonding on c-Si.
2. Experimental
2.1. Preparation PSL have been prepared from p+- and p 9 -type, respectively, 5–15 mV cm and 0.5 V cm in HF 25%, deionised water 25% and ethanol 50%, in order to obtain porosity ranging from 30 to 80% porosity, on the base of gravimetry. According to previous investigations [16] etch stops have been introduced during anodisation to get good homogeneity and in-depth morphology. Typical thickness ranges from 15 to 30 mm. During the first characterisation trials, two gold pads (50 nm thick) were sputtered on the surface of PS in a planar configuration. Gold wires were stuck to them with silver paste. In the successive work, three different processes have been applied to the samples, to realize a reproducible contact on PSL. The first is a patterned deposition of Ti – W oxide, an adhesion layer (Ti – W) and Pt contact from the porous silicon area to the external c-Si. On this region bonding by a parallel-gap microwelder is possible (Fig. 1). Sometimes, in the presence of defects and powders on c-Si, the formation of pinholes in the oxide layer could short circuit the Pt contacts with the substrate. Another process uses a ‘comb’ structure obtained by ‘lift-off’ of gold on negative photoresist, and successive anodisation in HF solution. With an appropriate chromium adhesion layer, the metal contact resists in solution for 3 min. The underetching can be complete, depending on anodisation parameters and lateral structure dimensions, as shown in Fig. 2. The third process takes advantage of selectivity to doping of anodisation [17], so using a 0.5 V cm wafer (p 9 ) with a 0.5-mm phosphorus implanted layer (3.5× 1012 cm − 2, 180 KeV). Photolithography allows the structuring, then plasma etching (1 mm deep) is performed, before anodisation. Thanks to doping selectivity, a complete underetching of all the structures (up to 300 mm and more) is obtained, as results from Fig. 3. A further metal mask step can be then applied, depositing a gold layer by thermal evaporation or sputtering. A double step resist
Fig. 2. Contacting process for p+ PSL using a ‘comb’ structure obtained by ‘lift-off’ of gold on negative photoresist, and successive anodisation in HF solution. With an appropriate adhesion layer of Cr, the metal contact resists in solution for 3 min. As showed by SEM cross-section, The underetching can be complete, depending on anodisation parameters and lateral structure dimensions.
Fig. 3. Process for p 9 (0.5 V cm), with a 0.5 mm phosphorus implanted layer (3.5 × 1012 cm − 2, 180 KeV). Photolithography allows the structuring, then plasma etching (1 mm deep) is performed, before anodisation. Thanks to doping selectivity, a complete underetching of all the structures (up to 300 mm and more) is obtained, as result from the SEM cross-section.
L. Boarino et al. / Materials Science and Engineering B69–70 (2000) 210–214
212
The membranes have been deposited on p-type double polished substrates and immediately characterised.
3. Results
3.1. Electrical conducti6ity
Fig. 4. Dynamic response of samples with porosity ranging from 38 to 75% to low concentrations of NO2. Measurements were performed at R.T. and constant R.H. 20%. Table 1 Relative response DG/G of PS samples of different porosity to the listed NO2 concentrations o (%)
DG/G of 3 ppm
DG/G of 5 ppm
DG/G of 10 ppm
38 43 53 62 75
B0.1 1 0.3 29.3 3.5
B0.1 1.9 1.7 84.3 45
0.7 3.5 4.6 197 164
Fig. 5. Relative response of 75% porosity sample to 2, 5 and 10 ppm of NO is, respectively, 0.3 and 1.5. Measurements were performed at R.T. and 20% R.H.
process can also be used in alternative to metal mask, with a first exposure and lift-off of gold layer and subsequent re-aligning for implanted layer stripping. For FTIR characterisation, free-standing PSL have been produced with 62% porosity, different thickness (20, 40, 60 mm) using a current density of 600 mA applied for 20 and 10 s of 800 mA to obtain a good quality surface finishing on the back of the membrane.
Measurements were performed at room temperature (R.T.) in a stainless steel chamber kept at 20°C and at atmospheric pressure. The PS change of conductance in presence of different gases in humid air was monitored by a d.c. volt amperometric technique at constant bias V= 5 V [18]. Gas traces in dry air coming from certified bottles are diluted with air to obtain the desired concentration into the test chamber at constant relative humidity (R.H.). A molybdenum oxide converter performs the conversion of NO2 into NO (conversion efficiency was equal to 98%). The PS response to NO2 has been examined in relation to porosity. Fig. 4 shows the dynamic response of PS samples of different porosity o (38, 43, 53, 62 and 75%) to NO2. Table 1 summarises the relative conductance variation DG/G for all the samples shown in Fig. 4. The lowest concentration examined was 1 ppm and the relative response was 1.6 for the 60% sample. An extrapolation from the sensitivity curve shows that it is possible to reach alarm level concentration in urban environment. The relative response is quite low for PS of o B 53% but shows a pronounced increase at greater porosity. High specific surface available for gas adsorption is a key factor to obtain an efficient chemical sensor. For PS surface to volume ratio is a function of the porosity, o, reaching the maximum value for o= 55%. The suggested correlation between the specific surface and sensor response could justify the slight decrease of response for the 75% sample as well. The dynamic rate of the sample at 75% in porosity is very slow. After a fast increase of sensor conductivity following NO2 introduction, that takes place in less than a minute, a slower process occurs, such that steady state is not reached after 30 min (Fig. 4). Fig. 5 reports the dynamic response for the 75% sample to 5 and 10 ppm of NO. Since NO is a reducing species the sign of conductivity change is opposite compared with NO2, a well known oxidising gas. Moreover, the relative change of resistance in presence of NO is much lower than the relative change of conductance in presence of NO2 under the same concentration. The relative response of p+ PS to interfering species as CO (up to 1000 ppm) and CH4 (up to 5000 ppm) is negligible and anyhow low for alcohol such as methanol at concentrations up to 800 ppm. Instead, as well known from literature [1], humidity influences strongly the sensor response of p+ PS to-
L. Boarino et al. / Materials Science and Engineering B69–70 (2000) 210–214
wards NO2. For example, when the R.H. changes from 20 to 40%, sensor response to 5 ppm of NO2 decreases from 45 to 10.7.
3.2. FTIR spectroscopy FT-IR spectroscopy has been used extensively in the characterisation of the environment of hydrogen in porous silicon because of its high sensitivity to surface molecular structures [19]. On the contrary, to the best of our knowledge, nobody has investigated the interaction of gaseous species by means of FT-IR spectroscopy. It is known that electrical conductivity of porous silicon is modified strongly (enhanced) depending upon the presence or the absence at the surface of adsorbed species like NO2, NO, O2 and other. In order to evaluate the phenomena occurring at the PS surface when it interacts with such gases we have decided to study the modification occurring at the PS surface when it comes in contact at R.T. with NO2. In Fig. 6 we report our preliminary results. The FT-IR spectra, showed in the 650 – 3800 cm − 1 range, have been recorded in the transmission mode by means of Bruker IFS 55 FT-IR spectrophotometer equipped with a MCT cryodetector, working at 2 cm − 1 resolution. The spectra have been collected on the same free-standing sample. Spectrum 1 is related to sample outgassed under dynamic vacuum at R.T. and 1.33× 10 − 1 Pa. This spectrum shows the common features well documented in literature, characteristic of porous silicon [19], i.e. the absorption bands attributed to the vibrational modes of the SiHx species. Moreover close to these absorptions,
213
other bands assigned to the vibrational modes of oxidised species are also evident. In particular: (i) in the 1000–1250 range cm − 1 a couple of bands assigned to the stretching vibrations of Si–O species; (ii) in the 2150–2300 cm − 1 range a couple of bands attributed to the Si–H stretching modes of OSi–Hx species; (iii) at 3748 cm − 1 the characteristic absorption band of the SiO–H stretching vibration. Spectrum 2 has been obtained after dosage of some 1 Torr of pure NO2. We note that: (i) a featureless and intense absorption is formed, extending on the entire spectral range and markedly evident at nB 3000 cm − 1; (ii) the absorption bands related to the various Si–Hx, OSi–Hx and SiO– H modes are not influenced significantly; (iii) no new discrete bands related to molecular species in the adsorbed state (i.e. adsorbed at the PS surface) are formed. As far as point (i) is concerned, the observed featureless absorption is characteristic of a loss of transparency of the sample. This phenomenon is well known in the case of semiconductor or non-stoichiometric oxides (mainly because of the presence of oxygen vacancies). In these systems the loss of transparency in the mid-IR (nB 3000 cm − 1, very broad and featureless like that observed in Fig. 6) is assigned to the absorption due to free electrons in the conduction band and/ or electronic transition, mainly from V+ (or from 0 similar shallow levels in the band gap) to the conduction band [20]. On these bases we infer that this broad absorption are attributed to electrons populating the conduction band. Finally, spectrum 3 has been recorded after outgassing the sample, at R.T., under dynamic vacuum (1× 10 − 3 Torr). It is evident that removing NO2 from the gas phase leads to the almost total restoration of initial conditions: the broad and featureless absorption disappears. This means that conduction band is no more populated. As concluding remark we can comment that: (i) NO2 admission markedly modifies the spectral shape; (ii) at R.T is not possible to distinguish any spectral features arising from molecular species adsorbed or interacting with the PS surface; (iii) removing NO2 almost totally restores the initial transparency; (iv) admission of air (room pressure) on the outgassed samples (after NO2) completely restores the initial condition (spectrum not shown).
4. Discussion
Fig. 6. FTIR spectra of free-standing p+ PSL: spectrum 1 is related to sample outgassed under dynamic vacuum at R.T. and 1.33 ×10 − 1 Pa. Spectrum 2 has been obtained after dosage of some 1 Torr of pure NO2, spectrum 3 has been recorded after outgassing the sample, at R.T., under dynamic vacuum. The featureless and intense absorption extending on the entire spectral range is markedly evident at nB 3000 cm − 1, is evident in presence of NO2.
The simple models proposed by Lehmann and Stievenard, explaining the resistivity variations of p+ PSL as a surface effect due to the screening of traps at the silicon-oxide interface by polar species has been confirmed substantially by our measurements; NO2 is a polar gas, and the sensitivity of the PSL reaches a maximum in correspondence of the higher specific sur-
214
L. Boarino et al. / Materials Science and Engineering B69–70 (2000) 210–214
face (around 60% porosity) but a dependence on the structure dimensions is also expected. Further work must be devoted to the correlation among crystallite size and response of sensor. Moreover, the loss of free carrier absorption in the IR region of the spectrum mentioned by Beale and predicted by Tsu and Babic for quantum confined structures cannot be due to preferential etching of dopants, as proved by SIMS measurements [14]. DLTS spectroscopy indicates a free carrier capture by surface traps or dangling bonds with activation energy of 0.44 – 0.53 eV [21]. By FTIR measurements of p+ free-standing PSL in NO2, a strong release of free-carrier in the IR has been observed, as a result of electrostatic interaction (screening) between the polar gas and the interface defects. The phenomenon is completely reversible evacuating the gas and exposing the PSL to air, and could explain the strong electrical conductivity variations in presence of low NO2 concentrations. The interaction between polar gases and mesoporous silicon layers has been proved, by CCl4 and thermal treatments [15], to be through the Si – H bonds and not directly with the dangling bonds. Also from our FTIR measurements there is no evidence of direct adsorption of gas on interface states. Anyway polar momentum of NO2 is not so high if compared with other polar species, so the strong reactivity of PSL to this gas must be investigated in detail. Different mechanisms can be present in the case of interaction with NO, but more experimental studies are needed, in this case.
5. Conclusions These results show the applicability of porous silicon to the development of a low cost, low power consumption sensor for air quality. The effect of interfering gases has been checked: no response towards CO (1000 ppm) and CH4 (5000 ppm) have been checked and low response to methanol and NO is shown. Variations in electrical conductivity with 1 ppm of NO2 have been observed at R.T. As well known, humidity influences the performances of the PS gas sensor: in order to obtain a reproducible device for NO2 sensing, it must be normalised for R.H., using a dedicated R.H. sensor.
.
The sensor response could be optimised by varying the porosity of the layer, since the electrical response is essentially conditioned by specific surface of material. In situ FTIR measurements in NO2 show a reversible free-carrier detrapping in the IR region of the spectra. More work in this field will be carried out to obtain a deeper insight of the molecular dynamics at the surface of mesoporous silicon in presence of liquids and gases. Acknowledgements This work was supported by Progetto SUD INFM. References [1] J.J. Mares, J. Kristofik, E. Hulicius, Thin Solid Films 255 (1995) 272. [2] D. Bilenko, O. Belobrovaja, E. Jarkova, O. Coldobanova, I. Mysenko, E. Khasina, Sensors and Actuators A 62 (1997) 621. [3] S. Zangooie, R. Bjorklund, H. Arwin, Sensors and Actuators B 43 (1997) 168. [4] I. Schechter, M. Ben-Chorin, A. Kux, Anal. Chem. 67 (1995) 3727. [5] I. Balberg, Philos. Mag. B80 (4) (2000) in press. [6] R. Tsu, D. Babic, Appl. Phys. Lett. 64 (1994) 1806. [7] A. Diligenti, A. Nannini, G. Pennelli, F. Pieri, Appl. Phys Lett. 68 (1996) 687. [8] H. Koyama, N. Koshida, J. Lumin. 57 (1993) 293. [9] J. Kocka, I. Pelant, A. Fejfar, J. Non Cryst. Solids 198-200 (1996) 857. [10] M. Ben-Chorin, F. Moller, F. Koch, J. Lumin. 57 (1994) 159. [11] M. Ben-Chorin, F. Moller, F. Koch, Phys. Rev. B 49 (1993) 2981. [12] M. Ben-Chorin, F. Moller, F. Koch, W. Schirmacher, M. Eberhard, Phys. Rev. B 49 (1995) 2199. [13] M. Ben-Chorin, A. Kux, I. Schechter, Appl. Phys. Lett. 64 (1994) 481. [14] V. Lehmann, F. Hoffmann, F. Mo¨ller, U. Gru¨ning, Thin Solid Films 255 (1995) 20. [15] D. Stievenard, D. Deresmes, Appl. Phys. Lett. 67 (11) (1995) 1570. [16] M.G. Berger, M. Tho¨enissen, W. Theiss, H. Mu¨nder, in: G. Amato, C. Delerue, H.J. von Bardeleben (Eds.), Optical and Structural Properties of Porous Silicon Nanostructures, Gordon and Breach, Amsterdam, 1997, p. 557. [17] K. Imai, H. Unno, IEEE Trans. Electron. Dev. 31 (1984) 297. [18] G. Sberveglieri, L.E. Depero, S. Groppelli, P. Nelli, Sensors and Actuators B 26-27 (1995) 89 – 92. [19] Y. Ogata, H. Niki, T. Sakka, M. Iwasaki, J. Electrochem. Soc. 142 (1995) 1595. [20] A. Chiorino, G. Ghiotti, F. Prinetto, M.C. Carotta, G. Martinelli, M. Merli, Sensor and Actuators B 44 (1997) 474. [21] L.Z. Yu, C.R. Wie, Sensor and Actuators A 39 (1993) 253.
