Carbon nanotubes (CNTs) are nearly one-dimensional molecules obtained by rolling up one or more sheets of single-atom-thick carbon, called graphene. Since their discovery in 1991, many scientists have explored the extraordinary physical properties of these materials. The first paper showing the potential of CNT-based transistors for gas sensing applications was published in 2000.1 In this paper, Kong et al. at Stanford University observed that a single CNT (used as a transistor channel between two gold electrodes) interacted with gas molecules, which altered the current flowing through the channel as a function of the substrate (gate) voltage.
The use of a single CNT as a transistor channel presents some significant challenges. First, current fabrication techniques produce both metallic and semiconducting nanotubes. Only the semiconducting ones will work as a transistor channel, so there is a risk of making a useless device if a metallic CNT ends up between the electrodes. Second, it is quite laborious to identify the position of one single CNT using precision observation methods such as atomic force microscopy (AFM). From an industrial point of view, this is not a suitable solution for the mass production of sensors. Third, it is very difficult to obtain reproducible devices, since the transistor electrical characteristics are dependent on a single nanotube, whose physical characteristics are hard to control at such small dimensions. Therefore, we have focused our work on transistors using CNT mats as channels.2–4 We deposit these multi-nanotube networks with a dynamic spray-gun deposition technique compatible with large surfaces (see Figure 1).
Figure 1. (left) The deposition machine developed at Thales Research and Technology. (right) A visualisation of the deposition process.
The novelty of our deposition method is its compatibility with large surfaces and flexible substrates. Our technique can fabricate high-performance transistors with reproducible characteristics by exploiting the percolation effect in networks of nanotubes. The percolation effect is a statistical phenomenon that occurs above a specific density of CNTs per area (called the threshold percolation density), where it becomes highly probable to have at least one chain of interconnected CNTs linking the two electrodes (like a network of sticks sprinkled over a floor). We have developed a method for making dense CNT mats that is suitable for industrial transfer. Specifically, our technique employs a dynamic deposition machine that sprays micro-drops of CNT solutions on heated substrates. The heating of the substrate is necessary to avoid the so-called “coffee-ring effect”, which is a tendency of particles in a slowly-drying drop to migrate and deposit on the borders. In our patented technique,5 we heat the substrate to the boiling point temperature of the solvent, so that the drops instantaneously evaporate at impact with the substrate. Since the nanoparticles do not have time to migrate to the borders, we can dramatically improve the uniformity of the deposited mats.
Figure 2. Scheme of an array of carbon-nanotube-based transistors for a gas-sensing application. We achieve an electronic fingerprint of the gas by using the relative change of current, IDS, for each transistor after exposure to specific gas molecules. This change is represented in the graphs on the right, where the dependence on the voltage between gate and source, VGS, is shown before (solid lines) and after (dashed lines) exposure.
Figure 3. a) Sensor chip containing 16 carbon-nanotube-based transistors with electrodes made from 4 different metals (4 transistors for each metal). b) Chip dimensions compared to a euro coin. c) Chip mounted on the ceramic dual in-line packaging used for tests.
Figure 4. Change in IDS as a function of time for concentrations of NO2 between 100ppb and 10ppm. The voltages between gate and source and between drain and source were, respectively, VGS=- 16V and VDS=1.6V. Au: Gold, Pt: Platinum, Pd: Palladium, Ti: Titanium.
Figure 5. Change in IDS as a function of time for concentrations of NH3 between 10ppb and 10ppm (VGS=- 16V and VDS=1.6V).
Figure 6. Change in IDS as a function of time for concentrations of di-methyl-methyl-phosphonate (DMMP) between 1ppm and 20ppm (VGS=- 16V and VDS=1.6V).
Our goal is to use our CNT mat fabrication technique to make selective gas sensors that perform a sort of electronic fingerprinting of each gas. This is done by combining CNT-based transistors with different metal electrodes to obtain metal/nanotube junctions with specific characteristics. Each gas interacts in a specific way with each junction, causing the current through the nanotube mat to change in a way corresponding to that particular gas. We expose several different transistors to a gas sample at the same time so that the fingerprint can be recorded instantaneously. The concept6 is shown in Figure 2.
Using our spray-gun deposition method, we fabricated a sensor chip with 16 CNT-based transistors (see Figure 3). The electrodes consisted of four different metals (four transistors for each metal): gold (Au), platinum (Pt), palladium (Pd) and titanium (Ti). We tested the chip by simultaneously recording the changes of the transfer characteristics of the different transistors when exposed to three different gases: nitrogen dioxide (NO2), ammonia (NH3) and a sarin gas simulant called di-methyl-methyl-phosphonate (DMMP). To evaluate the sensitivity limits of the chip, we exposed the transistors to concentrations of NO2 and NH3 that varied from 100 parts per billion (ppb) to 10 parts per million (ppm). We recorded the change of the current through each transistor between the drain and source, IDS (see Figure 4 for NO2 and Figure 5 for NH3). A single exposure cycle consisted of 300 seconds of exposure to the particular gas species followed by 600 seconds of air exposure.
For NO2, we observed a clear change in the IDS of the transistors with platinum and gold electrodes for concentrations as low as 100ppb, and with all the metals at 200ppb. In the case of NH3, platinum and gold showed a change in IDS for 100ppb, but in order to see a clear change for all of the transistors, we needed a concentration of 1ppm. The signal for NH3 is smaller than that for NO2 because the NH3 molecules create a weaker electric dipole at the interface between the metal and the CNTs. Therefore, the change of the Schottky barrier height and the corresponding change in the current are both lower in the case of NH3. These results are preliminary, considering that our chip has not been functionalized (chemically modified with, for example, specific polymers or biological molecules to enhance the sensitivity) and the measurements have been performed using air at ambient pressure as the carrier gas.
We also performed another set of measurements exposing the same chip to DMMP (see Figure 6). The results demonstrate a sensitivity down to 1.6ppm, which is the lowest concentration achievable with our test bench.
In this work, we demonstrated that NO2, NH3, and DMMP gases have unique interactions with transistors fabricated using different metals as electrodes. Based on these results, we believe that our technology could lead to simple, relatively low-cost devices to perform selective sensing. Moreover, the deposition technique developed is extremely versatile and could be used for other applications such as fabrication of bolometers, free-standing membranes, replacements for transparent and conductive indium-tin-oxide layers in organic LEDs, and light and cheap ultracapacitors on flexible substrates. Our future work will focus on these applications.7
Paolo Bondavalli, Gilles Feugnet
Thales Research & Technology
Paolo Bondavalli is head of the nanomaterials team for the joint team at Thales and CNRS (National Center for Scientific Research). He was awarded his PhD from INSA (National Institute of Applied Sciences) in Lyon in 2000 with a thesis dealing with micro-opto-electro-mechanical systems and subsequently his diploma to supervise PhD Thesis at Paris Sud in 2011. His research interests deal with nanomaterials, supercapacitors, sensors, and spintronics.
Gilles Feugnet is an engineer, whose research interests include optics and sensors.
1. J. Kong, N.R. Franklyn, C. Zhou, M.G. Chapline, S. Peng, K. Cho, H. Dai, Nanotube molecular wires as chemical sensors, Science 287, p. 622-625, 2000.
2. P. Bondavalli, P. Legagneux, D. Pribat, CNTFET based gas sensors : State of the art and critical review, Sensors and Actuators B 140, p. 1 304-31, 2009.
3. P. Bondavalli, L. Gorintin, P. Legagneux, D. Pribat, J.P. Simonato, L. Cailler, CNTFET Gas Sensors using SWCNT Mats : Method for Low-cost Fabrication, Solution to Improve Selectivity, Influence of Humidity (and Methods to Reduce it), Experimental Results Using Interfering Agents, MRS Proc. 1204, p. 1204-K07, 2011.
4. P. Bondavalli, L. Gorintin, P. Legagneux, CNTs for gas sensing, Applications of Carbon Nanotubes, Pan Stanford, 2011.
5. P. Bondavalli, L. Gorintin, P. Legagneux, P. Ponard, Process for depositing nanomaterials using a dynamic air-brush technique, Patent No. WO2012049428, 2012.
6. P. Bondavalli, P. Legagneux, D. Pribat, P. Lebarny, J. Nagle, Conductive nanowire or nanotubes based transistors network and corresponding electronics device for detecting analytes, Patent No. WO2006128828, 2006.
7. P. Bondavalli, D. Pribat, C. Delfaure, L. Gorintin, J.P. Schnell, L. Baraton, P. Legagneux, Non faradic carbon nanotubes based supercapacitors: state of the art, Euro. Phys. J. - Appl. Phys. (accepted), 2012.