Localized and In-Situ Integration of Different Nanowire Materials for Electronic Nose Applications ^†

Abstract

A new method for the site-selective synthesis of nanowires has been developed to enable the material growth with specific morphology and different compositions on one single chip. Based on a modification of the chemical vapor deposition method, the growth of nanowires on top of micromembranes can be easily tuned and represents a simple and adjustable fabrication process for the direct integration of different nanowire-based resistive multifunctional devices. This proof-of-concept is exemplified by the deposition of SnO₂, WO₃ and Ge nanowires on the membranes of one single chip and their gas sensing responses towards different concentrations of CO, NO₂ and humidity diluted in synthetic air are evaluated. The principal component analysis of the collected data allows gas identification and, thus, the system is suitable for environmental monitoring.

1. Introduction

Nowadays there is an increasing concern about the presence of toxic and potentially harmful gases in the atmosphere, both for outdoor and indoor environments. To monitor them, gas sensor systems are required. Among the different sensors, solid state gas sensors are an excellent choice due to their low cost and relatively high sensitivity. Among these sensors, those based on semiconducting metal oxide nanowires with remarkable performances in terms of sensitivity and response time [1].

However, metal oxides usually lack in selectivity, being only able to distinguish between oxidizing or reducing gases due to the different responses. In the case of a mixture with both reducing and oxidising gases, the gas concentration of each of the components cannot be distinguished. A solution to this problem is the use of pattern recognition across sensor arrays with different sensing configurations (either using different sensing materials or operation temperatures), giving rise to a called e-nose configuration. This work, recently published [2], demonstrates for the first time that site-selective growth of different materials in form of nanowires for sensing applications is possible on a single chip in a defined geometry, with little interference on the growth parameters caused by the prior deposition of other nanostructured material. The use of chemical vapor deposition techniques is compatible as a CMOS post-process and, therefore, the impact of preparing gas sensors on any kind of chips for multifunctional devices is intriguing.

The devices presented here have the same gas sensing behavior as devices made on separate chips and CVD chambers. Their responses to atmospheres containing known harmful gases such as CO and NO₂, and at different levels of relative humidity, have been measured and the discrimination between all the three analytes in the environment has been achieved by the well-known PCA representation.

2. Materials and Methods

The NW growth was performed using as platform bulk micromachined substrates containing 4 electrically separated Si₃N₄/SiO₂ micromembranes each one containing an embedded resistive poly- Si heater and Pt interdigitated electrodes on top that allow the electrical access to the active sensing layer [3]. Operating temperature has been controlled through the voltage applied to the poly- Si heater in the respective membranes. The chip containing the micromembranes is mounted on a TO-8 holder and is wire-bonded to them. Two of the membranes have been sputter-coated with ~1 nm Au layer to allow a seed mediated VLS-type growth, while the other two have been shadow-masked to avoid Au deposition. Ge NWs have been grown in a low-pressure micro-CVD reactor via VLS using gold as growth seed similar to our previous published procedure [4]. SnO₂ NWs have been grown on a gold coated membrane using the same procedures described in literature [5]. Tungsten oxide NWs have been grown in a home-built aerosol assisted CVD (AACVD) reactor using 15 mL methanol as solvent for the W(CO)₆ with a concentration of 7.58 mmol/L and carrier gas flow rates of 50–200 sccm of welding argon, while keeping the substrate temperatures at ~360 °C. After the growth, the complete device was three times soaked in ethanol for 2 min to remove potential contamination due to the AACVD process and potential remains of carbonyls from the precursor. Finally, the material was annealed in air for 2 h at 400 °C to oxidize the NWs to WO₃, using the resistive heater for this purpose.

The response of the NW-based gas sensors towards different gases has been recorded using a home- made stainless-steel chamber of 8.6 mL volume connected to a Gometrics MGP2 gas mixer with 4 Bronkhorst Mass-Flow Controllers. Electrical measurements and flowing gas concentrations were controlled using a self-developed Labview software. Keithley 2602A dual Source Measure Units and various Keithley 2280S-32-.6 Source Measurement DC Supply units were used for the electrical control and measurement. The gas measurements have been done using a constant flow of 200 mL/min. The three NW-based gas sensors were characterized towards concentrations from 10 to 50 ppm of carbon monoxide (CO) which is very close to the legal limit for 8 h exposure time-weighted average (TWA) of 8.6 ppm CO [6], from 1 to 5 ppm of nitrogen dioxide (NO₂) and from 0 to 80% of relative humidity (RH). The measurements were repeated at different temperatures to optimize the working temperature, except for the case of Ge NWs based sensors. Ge NWs were maintained at 100 °C to ensure a constant thickness of the GeOx layer, important to ensure a stable behavior as gas sensor.

3. Results

3.1. Nanowire Growth

The growth of the different NWs has been achieved successfully and can be individually addressed on a single chip. Figure 1 shows scanning electron microscope (SEM) image of one chip with the 4 membranes with the 3 different NWs grown on top.

3.2. Gas Sensing

The site-selectively grown SnO₂ NWs meshes showed responses to CO (50 ppm) at 400 °C and to NO₂ (5 ppm) at 300 °C of up to 12% and 120%, respectively, accompanied by a response time of 40 s for CO and 10 min for NO₂. The studied range is slightly higher than the upper time-weighted average exposure limit in the EU (0.1 ppm) for air quality standards in urban areas and one day exposure times [7]; however, in specific areas such as underground parking garages [8] or ice arenas [9] the values can be higher. However, responses of 56% towards 0.2 ppm NO₂, as illustrated in Figure 2a, demonstrate that these values can be easily detected with the here presented devices. We have recently demonstrated that Ge NW-based resistors fabricated on microhotplates can be successfully used as gas sensors [4]. The dynamic response is rather slow but reliable and reproducible. The sensitivity against CO and NO₂ was observed to be 0.8% for 50 ppm and 11% for 1 ppm, respectively. The measured responses of tungsten oxide NW meshes, highly temperature dependent, towards NO₂ are in the range of more than 300% for concentrations of ≥3 ppm in synthetic air at a temperature of 350 °C. However, the response towards CO is moderate, below 10%.

In real ambient conditions complex gas mixtures need to be measured, which introduces difficulties to deduce the contribution of the individual species because the responses obtained are not simply the sum of the response to the individual gases. In this case, the different gases can compete for the same adsorption sites and, consequently, react otherwise as for one single gas. The principal component analysis (PCA) method helps classifying different gas species using clustering, and helps, visually, to easily identify the ability of a sensor array to distinguish in the gas mixture between the individual gases. In our PCA, the results of the SnO₂, WO₃ and Ge NW-based gas sensors, working at 300, 250 and 100 °C respectively, towards CO (10 to 50 ppm), NO₂ (1 to 5 ppm) and RH (from 0 to 80%) reveal 3 different clusters corresponding to each gas species, as shown in Figure 2b. The existence of different separated clusters in the PCA representation proves the capacity of the three- sensor system to distinguish between the studied gases in the measured concentration range, which can be harmful for human health. The next generation of such NW devices is expected to be even more efficient in discriminating between different gases, when the processes will be further optimized incorporating a fourth NW material and an additional site-selective surface decoration with metal or metal oxide particles.

4. Conclusions

The here presented proof-of-concept, based on the localized preparation of NW-based sensors by different CVD approaches, shows clearly that the methodology can be successfully applied. Three different materials are demonstrated to be grown on different membranes of one single chip in a defined geometry and with little interference in the growth parameters caused by the prior deposition of other NW material. The sensors here show responses like those of individual devices prepared exclusively on single micromembranes and can discriminate between all the three analytes studied (CO, NO₂ and RH) in the environment, demonstrated by cluster separation in the well-known PCA representation.