- Email: [email protected]
Highly sensitive and selective carbon nanotube-based gas sensor arrays functionalized with different metallic nanoparticles
Accepted Manuscript Title: Highly Sensitive and Selective Carbon nanotube-Based Gas Sensor Arrays Functionalized with Different Metallic Nanoparticles Author: Ahmed Abdelhalim Maximilian Winkler Florin Loghin Christopher Zeiser Paolo Lugli Alaa Abdellah PII: DOI: Reference:
S0925-4005(15)30051-4 http://dx.doi.org/doi:10.1016/j.snb.2015.06.138 SNB 18715
To appear in:
Sensors and Actuators B
Received date: Revised date: Accepted date:
4-2-2015 28-6-2015 30-6-2015
Please cite this article as: A. Abdelhalim, M. Winkler, F. Loghin, C. Zeiser, P. Lugli, A. Abdellah, Highly Sensitive and Selective Carbon nanotube-Based Gas Sensor Arrays Functionalized with Different Metallic Nanoparticles, Sensors and Actuators B: Chemical (2015), http://dx.doi.org/10.1016/j.snb.2015.06.138 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Manuscript Click here to view linked References
Highly Sensitive and Selective Carbon nanotube-Based Gas Sensor Arrays Functionalized with Different Metallic Nanoparticles
ip t
Ahmed Abdelhalim, Maximilian Winkler, Florin Loghin, Christopher Zeiser, Paolo Lugli and Alaa Abdellah
cr
Institute for Nanoelectronics, Technische Universität München, Arcisstraße 21, 80333 Munich,
us
Germany.
Abstract
an
We present the fabrication and characterization of selective CNT-based gas sensor, which is
M
capable of discriminating between four test gases (NH3, ethanol, CO and CO2). Different metals are utilized in order to alter the response of each sensing element towards the different test gases.
d
We discuss as well the different possibilities for metal combinations in the sensor arrays. The
te
optimum combination between the metallic nanoparticles is obtained through an efficient and quick framework. We study the effect of the thickness of the CNT thin-film on the
Ac ce p
reproducibility and the response of the sensor array. Furthermore, we investigate the response of the fabricated CNT sensor array towards a mixture of two gases. Such kind of characterization was performed by exposing the CNT sensor array to a mixture of (NH3, ethanol) and (NH3, CO) at different ratios for each mixture. We show that the response towards a mixture of two gases lies in the plane between the data points of each gas separately.
1 Page 1 of 26
Introduction Since their discovery by Ijima in 1991 [1] , Carbon nanotube (CNT) networks have attracted a broad community of researchers. Their impressive structural, mechanical and electronic
ip t
properties have been exploited in wide range of applications in science and engineering [2]–[5].
cr
The inherently reduced dimensionality and, hence, high surface-to-volume ratio, which, together with outstanding gas adsorption capabilities, renders CNTs ideal candidate for environmental
us
sensing applications [6], [7]. In general, CNTs could be classified into single-walled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs). Devices incorporating
an
both kinds of CNTs have been repeatedly proposed as sensors for a variety of gases [8]–[13].
M
The main aspect that is still lacking in CNT gas sensor is their ability to differentiate between different test gases. Adding such a feature to the outstanding properties of CNTs render them as
d
the ideal candidate for low cost, low weight and high performance gas sensing application.
te
One common approach to solve such an issue is by depending on a sensor array in detecting the different gases rather than a single gas sensor.
Ac ce p
Sensor array is a commonly used technique to enhance the selectivity of gas sensors in general. This technique was applied in metal oxide semiconducting [14], [15] as well as polymer-based gas sensors [16] In principle; sensor array is a device that is composed of several sensing elements. Each sensing element has a certain functionalization that provides a specific interaction with the gas molecules in varying degrees. The combined response of all the sensing elements results in a unique pattern for each specific test gas. Pattern recognition techniques are required to discriminate between the different test gases. Several techniques have been reported to perform the pattern recognition process such as: discriminant function analysis [17], [18] least partial square regression (PLS) [17], [19] and principle component analysis (PCA) [17], [20]–[22]. Star et al [23] were among the first to demonstrate the design, fabrication and testing for sensor arrays 2 Page 2 of 26
based on metal decorated CNT layers. CNTs were grown by CVD on Si/SiO2. They fabricated FETs utilizing a CNT network as a channel decorated with different metal nanoparticles. Metal decoration was done by thermal and electron-beam evaporation. They were able to distinguish
ip t
between CO, H2, H2S, NH3 and NO2. Conductance response showed dependency on the type of the metal NP decorating the CNT. Lu et al [24] showed the fabrication of CNT based gas sensor
cr
array that consists of 32 sensing element. The sensing elements were pristine CNT, CNT loaded
us
with metal clusters and CNT coated with polymers. Combining different kinds of functionalization enabled the fabricated sensor array to discriminate between different gases and
an
vapors such as: NO2, HCN, HCl, Cl2, acetone and benzene. All sensors were detected in the ppm concentration levels in dry air. CNT based gas sensor arrays are capable as well to be employed
M
in the detection of lung cancer bio-markers. Peng et al [20] were able to detect simulated patterns
d
of lung cancer biomarkers by CNT coated with non-polymeric organic materials. The sensor
te
array showed excellent discrimination between VOCs found in the breath of patients with lung cancer relative to healthy controls. Guerin et al. [25] showed the possibility of discriminating
Ac ce p
between different gas analytes (ethanol, NH3, toulene and H2) by changing the metal electrodes on top of the CNT channel. They used Pt, Pd and Au as metal electrodes where each metal creates an analyte-specific pattern and enables discriminative detection. In that context, Bondavali et al. [26] discussed the working mechanism of CNT FETs that have different contact metals under the exposure of DiMethyl-Methyl-Phosphonate (DMMP). They attributed the change in the transfer characteristics to the interaction between the gas and the metal. In this article, we present the design, fabrication and characterization of a selective CNT-based gas sensor array functionalized with different types of metallic NPs. Such sensor array is capable to generate unique patterns for each test gas. The performance of the sensor array is tested under the exposure of NH3, ethanol, CO and CO2.The data analysis is performed by the mathematical 3 Page 3 of 26
tool principle component analysis. During the process of developing such sensor array, we investigate the relation between the stability of the film and its resistance from one side and the achieved sensitivity of the sensor from the other side. We developed a quick and efficient
ip t
strategy framework for fabricating an optimum CNT-based gas sensor array. We investigate as well the response of the fabricated sensor arrays upon exposure to a mixture of two different
cr
gases at the same time.
us
Materials & Methods Design and Fabrication
an
The basic layout of the sensor array consists of four sensing elements, with an active area of 3 x 3
M
mm each. Si wafers with 200 nm thermally grown SiO2 is used as substrate. An inter-digitated electrode structure of 5 nm thick Cr layer and 40 nm Au layer is evaporated on top of the
d
substrate. The inter-digitated structure has a spacing of 100 µm between the electrodes. The four
te
sensors are connected to a common ground pad and another pad for each sensor which make them five contacts in total. Note that the area of the whole sensor array (1 x 1 cm) was kept as the
Ac ce p
area of the single sensor layout introduced previously [27]. In order to disperse CNTs in an aqueous solution, a high molecular weight cellulose derivative, sodium carboxymethyl cellulose (CMC), is used. This kind of surfactant has been reported previously as an excellent agent for dispersing CNTs in water [28]. Further discussion about the effect of different kinds of surfactants on the prepared CNT solution can be found in [29]. CMC is dissolved in distilled water to make 0.5 wt% aqueous solution. The solution is stirred overnight in order to uniformly dissolve the surfactant in water. After this, 5 mg of CNTs are added in 10 g of 0.5 wt% CMC-aqueous solution to make 0.05 wt% CMC based aqueous CNT solution. Sonication of the CMC based solution is performed for 20 minutes using a probe sonicator at
4 Page 4 of 26
50% power (~48 Watt). Solution is centrifuged for 90 minutes at 15000 rpm and 80% of the centrifuged solution is taken from top to be used for the deposition. For depositing the CNT layers we used dynamic spray system. An automated spray system which
ip t
consists of an industrial air atomizing spray valve (Nordson EFD, USA) in combination with an overhead motion platform (Precision Valve and Automation, USA) is used. Air atomizing along
cr
with ultrasonic spray nozzles provide a fine degree of atomization [30]. Several parameters have
us
to be adjusted in the spraying process in order to achieve the desired characteristics of the CNT layer. These parameters are material flow rate, atomizing gas pressure, nozzle-to-sample distance,
an
substrate temperature and motion speed. The most significant dimension in air-assisted nozzles is the diameter of the orifice. For the used setup, the diameter of the nozzle is 0.3 mm. We used the
M
same approach for spray deposition as we introduced previously [29], [31], [32], in which we
d
spray in the intermediate regime while heating up the substrate to the proper temperature. Such
te
adjustment speeds up the drying process of the droplets arriving at the substrate and consequently results in uniform, homogenous and reproducible layers. Figure 1 shows a schematic for the
Ac ce p
layout design of the sensor array as well as the different spray parameters. In order to remove the surfactant, a post treatment process is performed in diluted HNO3 overnight. Functionalization of the CNT layer was performed via thermal evaporation process. PET mask that has one opening with the same size of the single sensor is used. This mask is rotated three times for the different functionalization loads on each sensor in the array. The evaporation rate for all metals was kept at the lowest possible value (0.1 Å/s). Slow evaporation rate for the nanoparticles (NPs) enhances and enables clustering of the metal on the CNT surface rather than forming a continuous layer. Figure 2 shows an SEM image for a CNT thin-film functionalized with 0.5 nm load of Au.
5 Page 5 of 26
Characterization Procedure The complete sensor module is composed of the CNT sensor, mounted on a carrier glass together with a Peltier heating element for temperature control and a Pt100 sensor for in-situ temperature
ip t
monitoring. The heating element is located beneath the glass carrier, while the Pt100 sensor is placed next to the CNT sensor on top of the glass carrier. This arrangement allows precise
Ac ce p
te
d
M
an
us
cr
determination of the actual temperature reached by the CNT sensor. Sensor performance was
Figure 1 A schematic for the sensor array layout as well as the spray setup and its different parameters.
6 Page 6 of 26
ip t cr us
Figure 2 An SEM image of a spray deposited CNT thin-film onto Si substrate functionalized with 0.5 nm of Au along with a zoom-in acquired within the same area
an
investigated by exposure to different concentrations of the test gases (NH3, CO2, CO and ethanol). The desired concentration is achieved by dilution of the test gas with a carrier gas while
M
maintaining the total gas flux constant. Pure nitrogen is used as the carrier gas for NH3, CO and ethanol while air is used as the carrier gas for CO2. Additionally, we investigated the performance
d
of the sensor array upon the exposure to a mixture of two gases. In such case, we only mixed
te
between the gases that have N2 as its carrier gas. The exposure cycle for each gas was 100 s and it
Ac ce p
was performed at room temperature. For consistent, systematic and reliable results, the four sensors of the sensor array should be characterized simultaneously. In order to do so, the sensors read-out was performed by a Digital Multi-Meter (Keithley 2707) with an embedded multiplexer module, which consists of 20 different channels. Such a setup, enabled us to multiplex between the four sensors each second and recording the resistances simultaneously. Data Analysis
One of the most common methods for evaluating and analyzing the data generated by characterizing the response of the sensor array under the exposure of different gases is the Principle Component Analysis (PCA). PCA is a relatively simple mathematical method to reduce complex datasets to a lower dimension. The basic idea of such method is to express a given 7 Page 7 of 26
dataset through another basis, which then allows the projection of the data to a lower dimensionality [33]. In analyzing the data measured from the CNT sensor arrays fabricated in this article, we built a
ip t
Matlab code, which perform the PCA. Briefly, the code reads the text files generated after measuring the response towards the four test gases, and then built the sensitivity matrix. This
cr
matrix consists of 4 columns and 16 rows. Each column represents the response of one of the
us
sensors in the array towards the four test gases with the four different concentrations for each gas, hence the 16 rows. The PCA is performed on that matrix after normalizing the data and
an
subtracting its mean. Development Framework
M
The main issue that we faced while fabricating the sensor array was the choice of the metal for
d
each sensor as well its corresponding load. In order to exploit such a problem in a systematic and
te
consistent manner, we developed a framework strategy for the fabrication of the optimum CNT sensor array. It can be described in three main steps. Step I: Optimum CNT Thin-film Thickness
Step II: Simulation of Virtual Arrays
Step III: Fabrication of the Optimum Array
Ac ce p
In the first step, we try to identify the optimum CNT thin-film thickness that results in stable and reproducible results. In that context, we investigate the response of the sensor array at different film thicknesses and at different metal functionalization as well. Moreover, we functionalize each sensor in the array with the same metal but at different loads. For example if we choose Au for the first array, we functionalize three sensors in the array with loads 1.0 nm, 1.5 nm and 2.5 nm. The fourth sensor in the array was always kept bare (no functionalization) as a reference. The same procedure is applied for Pd, Ag, Ti and Cr. The load of the metals is chosen based on the 8 Page 8 of 26
knowledge we reported previously in [27]. This phase enabled us to have a general idea about the combined response of each metal towards the different gases. Additionally, it helped us to collect the response of large amount of sensors with different metal functionalization at different loads.
ip t
The collected data were the corner stone for the starting of step II. There are an infinite number of possible combinations for the different metal functionalization and the different loads that can be
cr
deposited on each individual sensor in the sensor array. In order to decrease such complexity, in
us
step II, we simulated the response of different combination, which seemed to be the optimum solution to decrease the development time. Different combinations of metal at different loads
an
were used to build virtual arrays. We developed a Matlab code to simulate the results of these different combinations. The choice of metals at different loads was based on the modification that
M
such functionalization can introduce to the film resistance, moreover whether such metal at that
d
specific load enhances or degrades the response towards one or more test gas. The optimum case
te
is having a certain combination, where each individual sensor behaves completely different towards each one of the four test gases. After different simulation iterations, we were able to
Ac ce p
reach the optimum combination of different metals at specific loads that can select between the four test gases. Step III is the fabrication and characterization of such combination as well as comparing the measured data to the simulating results.
Results and Discussion
Optimum CNT Thin-Film Thickness
The sensor arrays fabricated utilize CNT thin-films with rather low tube densities. At such densities, the film operates near its percolation threshold, thereby reducing the number of metallic pathways in the network and amplifying the contribution of the semiconducting pathways. Such film density results in the highest possible sensor sensitivity towards the four test gases. The resistance of the CNT layer at such density was ~ 8 kΩ. We obtained promising results in terms 9 Page 9 of 26
of distinction between the different test gases when the CNT network is functionalized with different loads of Au and Cr (fig S1). Nevertheless the obtained results were not reproducible. One possible explanation for such a behavior could be that the sensors before the
ip t
functionalization process were not identical, in terms of the resistance of the CNT layer. In order to investigate the effect of such variation in the resistance on the PCA results, we characterized
cr
two bare sensor arrays. The results obtained are shown in fig. 3 and the corresponding resistances
us
of the individual sensor in each array are shown in table 1. We can observe two main points from
Ac ce p
te
d
M
an
the shown results:
Figure 3 PCA plots for two bare CNT-based gas sensor arrays sprayed at percolation threshold
10 Page 10 of 26
Both sensor arrays show separation in the PCA plot between the different gases, which should not be the case as the sensing element in each array are supposed to be „identical‟. PCA plots obtained from each array are not identical, which mainly due to the difference
M
an
us
cr
range of resistances of the individual sensors in each array.
ip t
Table 1: Resistance variation of bare CNT-based gas sensor arrays deposited at percolation threshold
d
The variation in the resistance of the CNT thin-film is one common feature when the layers are
te
deposited at percolation threshold. At such low density the deposited layers are highly uniform
Ac ce p
but the resistance is highly sensitive to the number of deposited tubes. Depositing thicker film could omit such a problem but it would result immediately in decreasing the sensitivity of the sensor towards all test gases because of the domination of the metallic tube in that case. This implies that we have to trade-off the sensitivity in order to stabilize the resistance of the film. In order to achieve the optimum thickness for the CNT layer, we prepared samples at different film thicknesses besides the percolation threshold. The results obtained from all the fabricated samples are shown in fig. 4. The plot represents the relation between the resistance of the individual sensors in the different arrays and their corresponding normalized response (the mean normalized response of the four test gases). We can distinguish between three different film thicknesses, percolation threshold (blue circles), intermediate thickness (red squares) and high 11 Page 11 of 26
thickness (black rhombus). At the percolation threshold, a large variation can be observed in the resistance of the individual sensors as well as the corresponding normalized response. On the other hand, at the high film thickness the resistance and the normalized response are highly
ip t
stable. Nevertheless, the obtained normalized response is very low towards all test gases, which in some cases lower than the noise level. The intermediate film thickness is the one more suitable
cr
for the sensor array fabrication. At such thickness the resistance of the film is relatively stable
te
d
M
an
us
and the normalized response is sufficient to detect the different gases.
Ac ce p
Figure 4 Relation between the resistance and the normalized response for the bare CNT sensing element in all the fabricated sensor arrays, which are sprayed at different thicknesses: percolation threshold (circle), intermediate thickness (square) and high thickness (rhombus)
By achieving the optimum CNT film thickness, we fabricated different sensor arrays functionalized with same metal such as Au, Pd, Cr, Al, Ag, Ti but at different loads. Some of these metals were capable to achieve partial selection between the test gases (data not shown here). The main aim of such step was to collect the responses of the sensing element functionalized with different metals at different loads which was the corner stone in implementing step II in the development framework. Simulation of Virtual Sensor Array
12 Page 12 of 26
Each sensor array fabricated till now consists of four sensors, three are functionalized with the same metal at different loads and the fourth sensor was always kept bare. Effectively, these sensor arrays consisted only of two types of sensors, as we assume that each type of metal
ip t
functionalization influences the CNT sensor in similar manner, although the functionalization load might be different. In this step, we use the data we collected for the entire previously
cr
measured sensors in order to build a virtual sensor array to achieve the optimum combination of
us
metal functionalization and their corresponding loads that can separate and select between the four test gases. To carry on such a step, we implemented a Matlab program that simulates the
an
PCA result of the virtual arrays to deliver in a quick and efficient way the most promising combination. The input file for this program is the responses of all the individual sensors in the
M
fabricated sensor arrays based on one metal functionalization. The developed program allows
d
selecting sensors from different arrays to be combined in one virtual array and simulate the result
te
of such combination. This mimic the behavior we expect from the real array. The schematic in
Ac ce p
fig. 5 shows the working principle of such program.
Figure 5 A schematic shows the working principle of the developed Matlab program that allow simulating virtual arrays.
PCA plots in the previous sections were always two-dimensional because in all the discussed cases the third principle component almost had no influence on the outcome result. Applying one 13 Page 13 of 26
type of functionalization results in sensors with correlated results in the PCA plot, hence most of the information and the variance within the data are only the first and the second PCs. By combining different metals, the third component will have bigger influence, so the best
ip t
presentation of data is in 3D plots. Since sensor arrays functionalized with Au, Pd and Cr delivered the most adequate and diverse results in terms of stability as well as partial gas
cr
separation in the PCA plot, we chose these three metals for the different combination in the
us
virtual sensor array. After several simulation iterations and building virtual sensor arrays with only two or three sensors as well, we found that the best combination consists of an array of four
an
sensors: first sensor is bare, the second sensor is functionalized with Au (1.5 nm), the third sensor is functionalized with Pd (0.2 nm) and the fourth sensor is functionalized with Cr (1.0 nm). The
M
superiority in the normalized response of Au functionalization towards NH3 and for Pd towards
d
CO2 with Cr functionalization resulted in CNT-based gas sensor array that can isolate NH3 in one
te
plane, separate between CO, CO2 and ethanol in the other two planes as shown in fig. 6. The data are presented in two different angles to give a better perspective of the 3D plots. The simulated
Ac ce p
sensor array showed a strong influence of the second and the third principle components, which resulted in better distinction between the different gases. Realization of Optimum CNT-Based Gas Sensor Array In this step, we spray deposited CNT solution, which has 90% semiconducting tubes instead of the intrinsic CNT solution (67% semiconducting tubes) that we used in all the previous experiments. Using a semiconducting solution enabled us to deposit significantly thick CNT layers far beyond the percolation threshold and obtaining stable film resistances similar to the values obtained previously (~1 k). The higher network density guarantees higher stability in the resistance, thus boost the reproducibility of the sensor. For further confirmation on the stability of the used semiconducting solution, we repeated the previously performed procedure of fabricating 14 Page 14 of 26
ip t cr us an M d te Ac ce p
Figure 6 PCA plots for the simulated sensor array functionalized with 1.5 nm of Au, 0.2 nm of Pd, 1.0 nm of Cr and the fourth sensor is bare. The data are presented in two different angles to give better perspective of the 3D plot.
one bare sensor array and evaluating its performance in terms of the resistance stability, normalized response and PCA plots. The obtained results are shown in fig. S2 After successfully simulating the most promising combination of metal functionalization, which consists of Au (1.5 nm), Pd (0.2 nm) and Cr (1.0 nm) and a bare sensor, we fabricated and characterized such sensor array. Figure 7 depicts the measured resistance over time for the four test gases in this study, NH3, ethanol, CO and CO2. After each exposure cycle the sensors were heated up to 80 °C (active recovery) in order to regain their original resistance. Further discussion 15 Page 15 of 26
on the difference between active recovery and passive recovery was previously reported by our group in [6]. The minimum shift in the baseline reflects the stability of the sensors‟ resistances
Ac ce p
te
d
M
an
us
cr
ip t
after the successive measurements.
Figure 7 Plot of the measured sensors resistance over time. Each graph is divided into four segments, based on four-exposure/recovery cycle at different gas concentrations. The exposure interval is indicated by the area between closely adjacent dotted lines marking the start and the end. The different color indicates the response of each single sensor in the array. The four test gases shown here are (a) NH3, (b) Ethanol, (c) CO, (d) and CO2.
16 Page 16 of 26
We further investigated the relation between the resistance of the individual sensors during the successive measurements towards the different test gases and their normalized response. In order to perform such kind of characterization, we monitored the resistance of the individual sensors in
ip t
the array at the moment before gas exposure at each concentration of the different test gases, we are showing here only the resistance before the highest concentration, since we assume that most
cr
of the changes that occur during the measurement is reflected in the last exposure cycle. Taking a
us
closer look at the results obtained in fig. 8, we can observe a stable resistance (~1 kΩ) for the Au, Pd and bare sensors during the successive measurements. The resistance of the Cr sensor (~3.5
an
kΩ) was almost stable for the first three measurements and then dropped to almost half of its value at the last measurement.
M
It‟s worth mentioning here that Au and Pd functionalization results in decreasing the resistance of
d
the CNT thin-film. On the other hand, Cr functionalization results in increasing the resistance of
te
the film. This implies that the interaction between the CNT layer and the Cr NPs is entirely different than its interaction with Au and Pd. This topic is still under further investigation by our
Ac ce p
group. The PCA plot showed a clear separation of NH3 data points in one plane, while ethanol, CO and CO2 remain relatively in another plane. Nevertheless, ethanol is clearly separated from CO and CO2, which are relatively clustered together. By comparing this result to the results obtained from the simulated sensor array in the last section, we can clearly see the same pattern and the same behavior for the different data points. Such results are remarkable due to the fact that the data points used in the simulated version are extracted from measurements performed in different time periods and at slightly different ambient conditions.
17 Page 17 of 26
ip t cr us an M d te
Ac ce p
Figure 8 (a) The resistance of the individual sensors and their normalized response towards the four test gases (NH3, ethanol, CO and CO2) for the CNT-based gas sensor array, which is functionalized with Au 1.5 nm, Cr 1.0 nm and Pd 0.2 nm. In (b) & (c) is the PCA plots for such sensor array in 2D and 3D respectively.
This is a strong evidence for the reproducibility of the sensors fabricated at this resistance range. Note that in the previous discussions, we concentrated mainly on the stability of the resistance as a crucial issue for the outcome of the sensor array. It is important to mention that the outcome result is highly sensitive as well to any minor error in the lithography process, which is the first step in the fabrication process. In general, the result of the sensor array is highly sensitive to any minor change in the process steps as well as the response of the individual sensor. Regarding the sensing mechanism for the CNT-based gas sensors, there is no complete theory that fully explains the working principles [25]. The debate is whether the CNT layer or metal- CNT interface plays the main role in the sensitivity of the gas sensor. We do not favor the theory where the sensitive 18 Page 18 of 26
part is the metal/CNT interface since we reported previously CNT-based gas sensors that have relatively high sensitivity based on CNT electrodes instead of metal electrodes [6]. In general, metallic functionalization enhances the response of CNT–based gas sensors. Such an
ip t
effect is mainly attributed to the catalytic activity of the deposited NPs, which boost the gas sensitivity of the CNT film. It is worth mentioning here that the catalytic activity of Au NPs
cr
diminishes considerably when the particle size exceeds a certain limit [34]. Due to the fact that
us
post treatment process for the CNT layer is performed in diluted HNO3, defect sites and oxidation are introduced on the surface [35] which allows binding the NPs to the CNT films via the
an
process-induced carboxylic groups. Upon gas exposure, in case of NH3, the gas molecules coordinate with the formed NPs [36]. Our hypothesis is that the metallic NPs allow binding more
M
ammonia molecules to the same defect site instead of only one molecule per pristine CNTs. This
d
implies that a stronger charge transfer to the CNT network takes place in the presence of
te
nanoparticles. Another effect that takes place when depositing NPs on top of the CNT film is the formation of a „non-traditional Schottky barrier‟ at the interface between the semiconducting
Ac ce p
tubes and the metal NPs [37], [38]. Such Schottky barriers could be further modulated upon gas exposure, thus leading to an additional change in sensitivity of the sensors. Since in our case the work function of the functionalized film changes only slightly with coverage as shown in fig. S3, we cannot quantitatively assess such hypothesis. Characterization Towards Gas Mixture Another key aspect for the evaluation of the CNT-based gas sensor array performance is investigating its response towards mixture of two test gases simultaneously. Such characterization procedure is a major step towards realizing CNT-based gas sensor array in wide range of applications where the detected gas molecules consists of different concentration and from different gases rather than a single pure gas. In order to perform such a step, we examined 19 Page 19 of 26
the response of the optimum CNT-based gas sensor array, described in the last section, towards different mixtures of NH3 and CO as well as NH3 and ethanol. We limited this type of characterization only to the three test gases, which has pure N2 as a carrier
ip t
gas (NH3, CO and ethanol) to obtain systematic and consistent results,. Moreover, the total concentrations were always 10 ppm, 25 ppm, 50 ppm and 100 ppm independent from the ratios
cr
of the two gases. Before characterizing the sensor array towards the mixture, we measured first
us
its response towards each gas separately. We chose ratios of 1:1, 1:2 and 2:1 between the NH3 and ethanol as well as NH3 and CO for such characterization procedure. Figure 9a shows the
an
PCA plot for the CNT sensor array towards NH3 and CO as well as the different mixtures between them. The response towards equal ratios of the two gases lies almost in the middle
M
between the data points for each gas separately. By increasing the ratio of NH3 w.r.t CO, its data
d
points tends to move towards the NH3 data points and similarly vice versa. Investigating the
te
response of the sensor array towards different mixtures between NH3 and ethanol reveals similar results as shown in figure 9b. By achieving a separation between any two gases, the response
Ac ce p
towards a mixture of those two gases should lie in the plane between the data points of each gas separately. Such preliminary results indicate the possibility of identifying not only the type of the different gases in the mixture but their concentration as well
20 Page 20 of 26
ip t cr us an M d te
Ac ce p
Figure 9 PCA plots for a CNT-based gas sensor array functionalized with 1.5 nm of Au, 1.0 nm of Cr, 0.2 nm of Pd and a bare sensor under the exposure of (a) NH 3, CO and three mixture between the two gases and (b) NH3, ethanol and three mixtures between the two gases. In both cases the ratios of the mixture were 1:1, 1:2 and 2:1
Conclusion
We reported the complete development of a selective CNT-based gas sensor array functionalized with different metallic NPs. We introduced the principle component analysis method for analyzing the response of the sensor array. Reaching the selective CNT sensor array with optimum combinations of the metallic NPs was done through a developed framework strategy. We adopted the concept of building and simulating CNT virtual sensor arrays depending on the data collected from sensors with different functionalization at different load. Such step played an important role in decreasing the development time. In order to achieve a selective pattern in the 21 Page 21 of 26
PCA plot, we modified the spraying parameter to obtain a rather thicker layer of the CNT network. Such modification resulted in a small sacrifice in the sensitivity of the sensors. Nevertheless, the achieved sensitivity remains among the highest reported in literature. Taking
ip t
into consideration the short exposure interval used which is only 100 s for all test gases. The best combination of the metallic functionalization was Au at a load of 1.5 nm, Cr at a load of 1.0 nm
cr
and Pd at a load of 0.2 nm. The fourth sensor was kept bare. Such combination resulted in a clear
us
distinction between the four test gases (NH3, ethanol, CO and CO2) at an acceptable response and a stable film resistance. We took the characterization of CNT sensor array one step further where
an
we investigated its response towards a mixture of two gases simultaneously. The result showed that the data points of the mixture lie on the plane between the data points of the each gas
M
separately. The lifetime of the functionalized CNT – based gas sensors remains questionable as
d
further investigation should be held regarding the effect on the sensor performance when the
te
nanoparticle are exposed to the environment for a long time, such exposure might cause oxidation
Ac ce p
to the metallic particle which in turn might degrade the response of the sensor.
References [1]
S. IIjima, “Helical microtubules of graphitic carbon,” Nature, 354 (1991) 56–58.
[2]
Q. Cao and J. a. Rogers, “Ultrathin Films of Single--Walled Carbon Nanotubes for Electronics and Sensors: A Review of Fundamental and Applied Aspects,” Adv. Mater., 21 (1) (2009) 29–53.
[3]
L. Hu, D. S. Hecht, and G. Gruener, “Carbon nanotube thin films: fabrication, properties, and applications.,” Chem. Rev., 110 (10) (2010) 5790–844.
[4]
T. Zhang, S. Mubeen, N. V Myung, and M. a Deshusses, “Recent progress in carbon nanotube-based gas sensors.,” Nanotechnology, 19 (33) (2008) p. 332001.
[5]
D. Zhang, K. Ryu, X. Liu, E. Polikarpov, J. Ly, M. E. Tompson, and C. Zhou, “Transparent, conductive, and flexible carbon nanotube films and their application in organic light-emitting diodes,” Nano Lett., 6 (9) (2006) 1880–6. 22 Page 22 of 26
A. Abdellah, A. Abdelhalim, M. Horn, G. Scarpa, and P. Lugli, “Scalable Spray Deposition Process for High-Performance Carbon Nanotube Gas Sensors,” IEEE Trans. Nanotechnol., 12 (2) (2013) 174–18.
[7]
A. Abdellah, A. Abdelhalim, F. Loghin, P. Köhler, Z. Ahmad, G. Scarpa, and P. Lugli, “Flexible Carbon Nanotube Based Gas Sensors Fabricated by Large-Scale Spray Deposition,” IEEE Sensors Journal, 13 (10) (2013) 4014-4021.
[8]
J. Kong, “Nanotube Molecular Wires as Chemical Sensors,” Science, 287 (5453) (2000) 622–625.
[9]
Y. Wang, Z. Zhou, Z. Yang, X. Chen, D. Xu, and Y. Zhang, “Gas sensors based on deposited single-walled carbon nanotube networks for DMMP detection.,” Nanotechnology, 20 (34) (2009) 345502.
us
cr
ip t
[6]
an
[10] O. K. Varghese, P. D. Kichambre, D. Gong, K. G. Ong, E. C. Dickey, and C. a. Grimes, “Gas sensing characteristics of multi--wall carbon nanotubes,” Sensors Actuators B Chem., 81 (1) (2001) 32–41.
d
M
[11] Y. Wang, N. Hu, Z. Zhou, D. Xu, Z. Wang, Z. Yang, H. Wei, E. S.-W. Kong, and Y. Zhang, “Single-walled carbon nanotube/cobalt phthalocyanine derivative hybrid material: preparation, characterization and its gas sensing properties,” J. Mater. Chem., 21(11) (2011) 3779.
te
[12] Y.-T. Jang, S.-I. Moon, J.-H. Ahn, Y.-H. Lee, and B.-K. Ju, “A simple approach in fabricating chemical sensor using laterally grown multi-walled carbon nanotubes,” Sensors Actuators B Chem., 99 (1) (2004) 118–122.
Ac ce p
[13] M. Arab, F. Berger, F. Picaud, C. Ramseyer, J. Glory, and M. Mayne-L‟Hermite, “Direct growth of the multi-walled carbon nanotubes as a tool to detect ammonia at room temperature,” Chem. Phys. Lett., 433 (1-3) (2006) 175–181. [14] A. A. Tomchenko, G. P. Harmer, B. T. Marquis, and J. W. Allen, “Semiconducting metal oxide sensor array for the selective detection of combustion gases,” Sensors Actuators B Chem., 93 (1-3) (2003) 126–134. [15] B. T. Marquis and J. F. Vetelino, “A semiconducting metal oxide sensor array for the detection of NOx and NH3,” Sensors Actuators B Chem., 77 (1-2) (2001) 100–110. [16] F. Zee and J. W. Judy, “Micromachined polymer--based chemical gas sensor array,” Sensors Actuators B Chem., 72 (2) (2001) 120–128. [17] J. W. Gardner and P. N. Bartlett, Sensors and Sensory Systems for an Electronic Nose. Dordrecht: Springer Netherlands, 1992. [18] J. W. Gardner, H. V. Shurmer, and T. T. Tan, “Application of an electronic nose to the discrimination of coffees,” Sensors Actuators B Chem., 6 (1-3) (1992) 71–75. 23 Page 23 of 26
[19] T. C. Pearce, J. W. Gardner, S. Friel, P. N. Bartlett, and N. Blair, “Electronic nose for monitoring the flavour of beers,” Analyst, 118 (4) (2003) 371.
ip t
[20] G. Peng, E. Trock, and H. Haick, “Detecting simulated patterns of lung cancer biomarkers by random network of single-walled carbon nanotubes coated with nonpolymeric organic materials.,” Nano Lett., 8 (11) (2008) 3631–5.
cr
[21] H. Abe, T. Yoshimura, S. Kanaya, Y. Takahashi, Y. Miyashita, and S.-I. Sasaki, “Automated odor--sensing system based on plural semiconductor gas sensors and computerized pattern recognition techniques,” Anal. Chim. Acta, 194 (1987) 1–9.
us
[22] T. Nakamoto, A. Fukuda, and T. Moriizumi, “Perfume and flavour identification by odoursensing system using quartz--resonator sensor array and neural-network pattern recognition,” Sensors Actuators B Chem., 10 (2) (1993) 85–90.
an
[23] A. Star, V. Joshi, S. Skarupo, D. Thomas, and J.-C. P. Gabriel, “Gas sensor array based on metal-decorated carbon nanotubes.,” J. Phys. Chem. B, 110 (42) (2006) 21014–20.
M
[24] Y. Lu, C. Partridge, M. Meyyappan, and J. Li, “A carbon nanotube sensor array for sensitive gas discrimination using principal component analysis,” J. Electroanal. Chem., 593 (2006) 105–110.
te
d
[25] H. Guerin, H. Le Poche, R. Pohle, E. Buitrago, M. Fernández-Bolaños Badía, J. Dijon, and A. M. Ionescu, “Carbon nanotube gas sensor array for multiplex analyte discrimination,” Sensors Actuators B Chem., 207 (2015) 833–842.
Ac ce p
[26] P. Bondavalli, P. Legagneux, D. Pribat, A. Balan, and S. Nazeer, “Gas fingerprinting using carbon nanotubes transistor arrays,” Sensors and Actuators B Chem., 140 (2009) 304-312. [27] A. Abdelhalim, A. Abdellah, G. Scarpa, and P. Lugli, “Metallic nanoparticles functionalizing carbon nanotube networks for gas sensing applications,” Nanotechnology, 25 (5) (2014) 055208. [28] T. Takahashi, K. Tsunoda, H. Yajima, and T. Ishii, “Dispersion and Purification of SingleWall Carbon Nanotubes Using Carboxymethylcellulose,” Jpn. J. Appl. Phys., 43 (6A) (2007) 3636–3639. [29] A. Abdelhalim, A. Abdellah, G. Scarpa, and P. Lugli, “Fabrication of carbon nanotube thin films on flexible substrates by spray deposition and transfer printing,” Carbon, 61 (2013) 72-79. [30] R. C. Tenent, T. M. Barnes, J. D. Bergeson, A. J. Ferguson, B. To, L. M. Gedvilas, M. J. Heben, and J. L. Blackburn, “Ultrasmooth, Large-Area, High-Uniformity, Conductive Transparent Single--Walled--Carbon--Nanotube Films for Photovoltaics Produced by Ultrasonic Spraying,” Adv. Mater., 21 (31) (2009) 3210–3216.
24 Page 24 of 26
[31] A. Abdellah, B. Fabel, P. Lugli, and G. Scarpa, “Spray deposition of organic semiconducting thin--films: Towards the fabrication of arbitrary shaped organic electronic devices,” Org. Electron., 11 (6) (2010) 1031–1038.
ip t
[32] A. Falco, L. Cinà, G. Scarpa, P. Lugli, and A. Abdellah, “Fully-sprayed and flexible organic photodiodes with transparent carbon nanotube electrodes.,” ACS Appl. Mater. Interfaces, 6 (13) (2014) 10593–601.
cr
[33] P. C. Jurs, G. A. Bakken, and H. E. Mcclelland, “Computational Methods for the Analysis of Chemical Sensor Array Data from Volatile Analytes,” Chem. Rev., 100 (7) (2000) 2649–2678.
us
[34] A. Corma and H. Garcia, “Supported gold nanoparticles as catalysts for organic reactions.,” Chem. Soc. Rev., 37 (9) (2008) 2096–126.
an
[35] V. Lordi, N. Yao, and J. Wei, “Method for Supporting Platinum on Single--Walled Carbon Nanotubes for a Selective Hydrogenation Catalyst,” Chem. Mater., 13 (3) (2001) 733– 737.
M
[36] A. Bilic, J. R. Reimers, N. S. Hush, and J. Hafner, “Adsorption of ammonia on the gold (111) surface,” J. Chem. Phys., 116 (20) (2002) 8981.
te
d
[37] D. R. Kauffman, D. C. Sorescu, D. P. Schofield, B. L. Allen, K. D. Jordan, and A. Star, “Understanding the Sensor Response of Metal--decorated Carbon Nanotubes,” ACS Nano, 10 (3) (2010) 958–963.
Ac ce p
[38] D. R. Kauffman and A. Star, “Chemically Induced Potential Barriers at the Carbon Nanotube Metal Nanoparticle Interface,” Nano Lett., 7 (7) (2007) 1863-1868.
25 Page 25 of 26
*Author Biographies
ip t
Ahmed Abdelhalim received the B.Sc. degree in electronics engineering from the German University in Cairo, Cairo, Egypt in 2009, and the M.Sc. degree in communication and electronics engineering from the Technische Universität München, Munich, Germany, in 2011, where he is currently working towards the Ph.D. degree at the Institute for Nanoelectronics. His current research interests include fabrication and characterization of Carbon nanotubes chemical gas sensors based on spray deposition technique as well as different functionalization methods to enhance the sensor performance. Maximilian Winkler
an
us
cr
Florin Loghin received the B.Sc. degree in Electrical Engineering from the Technische Universität München, Munich, Germany in 2012, and the M.Sc. degree in electrical engineering from the Technical University of Munich, Munich, Germany, in 2015, where he is currently working toward the Ph.D. degree at the Institute for Nanoelectronics. His research interests include fabrication and characterization of carbon nanotubes based field effect transistors as well as developing spray deposition techniques for various organic materials. Christopher Zeiser
Ac ce p
te
d
M
Paolo Lugli received the Laurea degree in Physics from the University of Modena, Italy, in 1979. In 1981, he joined Colorado State University, Fort Collins, CO, USA, where he received the Mas- ters of Science in Electrical Engineering in 1982, and the Ph.D. in 1985. In 1985, he joined the Physics Department with the University of Modena, Reggio Emilia, Italy, as Research Associate. From 1988 to 1993 he was an Associate Professor of “Solid State Physics” at the “Engineering Faculty” with the University of Rome “Tor Vergata”. In 1993, he became Full Professor of “Optoelectronics” with the same University. In 2002, he joined the Technical University of Munich, where he was appointed as a head of the newly created Institute for Nanoelectronics. His current research interests include nanoimprint lithography, the modeling, fabrication and characterization of organic devices for electronics and optoelectronics applications, the design of circuits and architectures for nanostructures and nanodevices, the numerical simulation of microwave semiconductor devices, and the theoretical study of transport processes in nanostructures. He is author of more than 250 scientific papers and co-author of the books “The Monte Carlo Modelling for Semiconductor Device Simulations”(Springer, 1989) and “High Speed Optical Communications” (Kluver Academic, 1999). In 2004, he served as General Chairman of the IEEE International Conference on Nanotechnology, Munich, Germany. He is the member of the German National Academy of Science and Engineering. Alaa Abdellah received the B.Sc. degree in Electronics and Communications Engineering from the Cairo University, Cairo, Egypt, in 2005, and the M.Sc. degree in High-Frequency Engineering and Optoelectronics from the Technische Universität München, Munich, Germany, in 2007. He is currently a Member of the scientific staff with the Institute for Nanoelectronics, Technische Universität München, Munich, Germany, where he obtained the Ph.D. degree in 2012. His current research interests include the area of organic and nanostructured electronics, focusing on large-area thin-film devices utilizing solution-processable organic semiconductors, and nanomaterials.
Page 26 of 26
S0925-4005(15)30051-4 http://dx.doi.org/doi:10.1016/j.snb.2015.06.138 SNB 18715
To appear in:
Sensors and Actuators B
Received date: Revised date: Accepted date:
4-2-2015 28-6-2015 30-6-2015
Please cite this article as: A. Abdelhalim, M. Winkler, F. Loghin, C. Zeiser, P. Lugli, A. Abdellah, Highly Sensitive and Selective Carbon nanotube-Based Gas Sensor Arrays Functionalized with Different Metallic Nanoparticles, Sensors and Actuators B: Chemical (2015), http://dx.doi.org/10.1016/j.snb.2015.06.138 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Manuscript Click here to view linked References
Highly Sensitive and Selective Carbon nanotube-Based Gas Sensor Arrays Functionalized with Different Metallic Nanoparticles
ip t
Ahmed Abdelhalim, Maximilian Winkler, Florin Loghin, Christopher Zeiser, Paolo Lugli and Alaa Abdellah
cr
Institute for Nanoelectronics, Technische Universität München, Arcisstraße 21, 80333 Munich,
us
Germany.
Abstract
an
We present the fabrication and characterization of selective CNT-based gas sensor, which is
M
capable of discriminating between four test gases (NH3, ethanol, CO and CO2). Different metals are utilized in order to alter the response of each sensing element towards the different test gases.
d
We discuss as well the different possibilities for metal combinations in the sensor arrays. The
te
optimum combination between the metallic nanoparticles is obtained through an efficient and quick framework. We study the effect of the thickness of the CNT thin-film on the
Ac ce p
reproducibility and the response of the sensor array. Furthermore, we investigate the response of the fabricated CNT sensor array towards a mixture of two gases. Such kind of characterization was performed by exposing the CNT sensor array to a mixture of (NH3, ethanol) and (NH3, CO) at different ratios for each mixture. We show that the response towards a mixture of two gases lies in the plane between the data points of each gas separately.
1 Page 1 of 26
Introduction Since their discovery by Ijima in 1991 [1] , Carbon nanotube (CNT) networks have attracted a broad community of researchers. Their impressive structural, mechanical and electronic
ip t
properties have been exploited in wide range of applications in science and engineering [2]–[5].
cr
The inherently reduced dimensionality and, hence, high surface-to-volume ratio, which, together with outstanding gas adsorption capabilities, renders CNTs ideal candidate for environmental
us
sensing applications [6], [7]. In general, CNTs could be classified into single-walled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs). Devices incorporating
an
both kinds of CNTs have been repeatedly proposed as sensors for a variety of gases [8]–[13].
M
The main aspect that is still lacking in CNT gas sensor is their ability to differentiate between different test gases. Adding such a feature to the outstanding properties of CNTs render them as
d
the ideal candidate for low cost, low weight and high performance gas sensing application.
te
One common approach to solve such an issue is by depending on a sensor array in detecting the different gases rather than a single gas sensor.
Ac ce p
Sensor array is a commonly used technique to enhance the selectivity of gas sensors in general. This technique was applied in metal oxide semiconducting [14], [15] as well as polymer-based gas sensors [16] In principle; sensor array is a device that is composed of several sensing elements. Each sensing element has a certain functionalization that provides a specific interaction with the gas molecules in varying degrees. The combined response of all the sensing elements results in a unique pattern for each specific test gas. Pattern recognition techniques are required to discriminate between the different test gases. Several techniques have been reported to perform the pattern recognition process such as: discriminant function analysis [17], [18] least partial square regression (PLS) [17], [19] and principle component analysis (PCA) [17], [20]–[22]. Star et al [23] were among the first to demonstrate the design, fabrication and testing for sensor arrays 2 Page 2 of 26
based on metal decorated CNT layers. CNTs were grown by CVD on Si/SiO2. They fabricated FETs utilizing a CNT network as a channel decorated with different metal nanoparticles. Metal decoration was done by thermal and electron-beam evaporation. They were able to distinguish
ip t
between CO, H2, H2S, NH3 and NO2. Conductance response showed dependency on the type of the metal NP decorating the CNT. Lu et al [24] showed the fabrication of CNT based gas sensor
cr
array that consists of 32 sensing element. The sensing elements were pristine CNT, CNT loaded
us
with metal clusters and CNT coated with polymers. Combining different kinds of functionalization enabled the fabricated sensor array to discriminate between different gases and
an
vapors such as: NO2, HCN, HCl, Cl2, acetone and benzene. All sensors were detected in the ppm concentration levels in dry air. CNT based gas sensor arrays are capable as well to be employed
M
in the detection of lung cancer bio-markers. Peng et al [20] were able to detect simulated patterns
d
of lung cancer biomarkers by CNT coated with non-polymeric organic materials. The sensor
te
array showed excellent discrimination between VOCs found in the breath of patients with lung cancer relative to healthy controls. Guerin et al. [25] showed the possibility of discriminating
Ac ce p
between different gas analytes (ethanol, NH3, toulene and H2) by changing the metal electrodes on top of the CNT channel. They used Pt, Pd and Au as metal electrodes where each metal creates an analyte-specific pattern and enables discriminative detection. In that context, Bondavali et al. [26] discussed the working mechanism of CNT FETs that have different contact metals under the exposure of DiMethyl-Methyl-Phosphonate (DMMP). They attributed the change in the transfer characteristics to the interaction between the gas and the metal. In this article, we present the design, fabrication and characterization of a selective CNT-based gas sensor array functionalized with different types of metallic NPs. Such sensor array is capable to generate unique patterns for each test gas. The performance of the sensor array is tested under the exposure of NH3, ethanol, CO and CO2.The data analysis is performed by the mathematical 3 Page 3 of 26
tool principle component analysis. During the process of developing such sensor array, we investigate the relation between the stability of the film and its resistance from one side and the achieved sensitivity of the sensor from the other side. We developed a quick and efficient
ip t
strategy framework for fabricating an optimum CNT-based gas sensor array. We investigate as well the response of the fabricated sensor arrays upon exposure to a mixture of two different
cr
gases at the same time.
us
Materials & Methods Design and Fabrication
an
The basic layout of the sensor array consists of four sensing elements, with an active area of 3 x 3
M
mm each. Si wafers with 200 nm thermally grown SiO2 is used as substrate. An inter-digitated electrode structure of 5 nm thick Cr layer and 40 nm Au layer is evaporated on top of the
d
substrate. The inter-digitated structure has a spacing of 100 µm between the electrodes. The four
te
sensors are connected to a common ground pad and another pad for each sensor which make them five contacts in total. Note that the area of the whole sensor array (1 x 1 cm) was kept as the
Ac ce p
area of the single sensor layout introduced previously [27]. In order to disperse CNTs in an aqueous solution, a high molecular weight cellulose derivative, sodium carboxymethyl cellulose (CMC), is used. This kind of surfactant has been reported previously as an excellent agent for dispersing CNTs in water [28]. Further discussion about the effect of different kinds of surfactants on the prepared CNT solution can be found in [29]. CMC is dissolved in distilled water to make 0.5 wt% aqueous solution. The solution is stirred overnight in order to uniformly dissolve the surfactant in water. After this, 5 mg of CNTs are added in 10 g of 0.5 wt% CMC-aqueous solution to make 0.05 wt% CMC based aqueous CNT solution. Sonication of the CMC based solution is performed for 20 minutes using a probe sonicator at
4 Page 4 of 26
50% power (~48 Watt). Solution is centrifuged for 90 minutes at 15000 rpm and 80% of the centrifuged solution is taken from top to be used for the deposition. For depositing the CNT layers we used dynamic spray system. An automated spray system which
ip t
consists of an industrial air atomizing spray valve (Nordson EFD, USA) in combination with an overhead motion platform (Precision Valve and Automation, USA) is used. Air atomizing along
cr
with ultrasonic spray nozzles provide a fine degree of atomization [30]. Several parameters have
us
to be adjusted in the spraying process in order to achieve the desired characteristics of the CNT layer. These parameters are material flow rate, atomizing gas pressure, nozzle-to-sample distance,
an
substrate temperature and motion speed. The most significant dimension in air-assisted nozzles is the diameter of the orifice. For the used setup, the diameter of the nozzle is 0.3 mm. We used the
M
same approach for spray deposition as we introduced previously [29], [31], [32], in which we
d
spray in the intermediate regime while heating up the substrate to the proper temperature. Such
te
adjustment speeds up the drying process of the droplets arriving at the substrate and consequently results in uniform, homogenous and reproducible layers. Figure 1 shows a schematic for the
Ac ce p
layout design of the sensor array as well as the different spray parameters. In order to remove the surfactant, a post treatment process is performed in diluted HNO3 overnight. Functionalization of the CNT layer was performed via thermal evaporation process. PET mask that has one opening with the same size of the single sensor is used. This mask is rotated three times for the different functionalization loads on each sensor in the array. The evaporation rate for all metals was kept at the lowest possible value (0.1 Å/s). Slow evaporation rate for the nanoparticles (NPs) enhances and enables clustering of the metal on the CNT surface rather than forming a continuous layer. Figure 2 shows an SEM image for a CNT thin-film functionalized with 0.5 nm load of Au.
5 Page 5 of 26
Characterization Procedure The complete sensor module is composed of the CNT sensor, mounted on a carrier glass together with a Peltier heating element for temperature control and a Pt100 sensor for in-situ temperature
ip t
monitoring. The heating element is located beneath the glass carrier, while the Pt100 sensor is placed next to the CNT sensor on top of the glass carrier. This arrangement allows precise
Ac ce p
te
d
M
an
us
cr
determination of the actual temperature reached by the CNT sensor. Sensor performance was
Figure 1 A schematic for the sensor array layout as well as the spray setup and its different parameters.
6 Page 6 of 26
ip t cr us
Figure 2 An SEM image of a spray deposited CNT thin-film onto Si substrate functionalized with 0.5 nm of Au along with a zoom-in acquired within the same area
an
investigated by exposure to different concentrations of the test gases (NH3, CO2, CO and ethanol). The desired concentration is achieved by dilution of the test gas with a carrier gas while
M
maintaining the total gas flux constant. Pure nitrogen is used as the carrier gas for NH3, CO and ethanol while air is used as the carrier gas for CO2. Additionally, we investigated the performance
d
of the sensor array upon the exposure to a mixture of two gases. In such case, we only mixed
te
between the gases that have N2 as its carrier gas. The exposure cycle for each gas was 100 s and it
Ac ce p
was performed at room temperature. For consistent, systematic and reliable results, the four sensors of the sensor array should be characterized simultaneously. In order to do so, the sensors read-out was performed by a Digital Multi-Meter (Keithley 2707) with an embedded multiplexer module, which consists of 20 different channels. Such a setup, enabled us to multiplex between the four sensors each second and recording the resistances simultaneously. Data Analysis
One of the most common methods for evaluating and analyzing the data generated by characterizing the response of the sensor array under the exposure of different gases is the Principle Component Analysis (PCA). PCA is a relatively simple mathematical method to reduce complex datasets to a lower dimension. The basic idea of such method is to express a given 7 Page 7 of 26
dataset through another basis, which then allows the projection of the data to a lower dimensionality [33]. In analyzing the data measured from the CNT sensor arrays fabricated in this article, we built a
ip t
Matlab code, which perform the PCA. Briefly, the code reads the text files generated after measuring the response towards the four test gases, and then built the sensitivity matrix. This
cr
matrix consists of 4 columns and 16 rows. Each column represents the response of one of the
us
sensors in the array towards the four test gases with the four different concentrations for each gas, hence the 16 rows. The PCA is performed on that matrix after normalizing the data and
an
subtracting its mean. Development Framework
M
The main issue that we faced while fabricating the sensor array was the choice of the metal for
d
each sensor as well its corresponding load. In order to exploit such a problem in a systematic and
te
consistent manner, we developed a framework strategy for the fabrication of the optimum CNT sensor array. It can be described in three main steps. Step I: Optimum CNT Thin-film Thickness
Step II: Simulation of Virtual Arrays
Step III: Fabrication of the Optimum Array
Ac ce p
In the first step, we try to identify the optimum CNT thin-film thickness that results in stable and reproducible results. In that context, we investigate the response of the sensor array at different film thicknesses and at different metal functionalization as well. Moreover, we functionalize each sensor in the array with the same metal but at different loads. For example if we choose Au for the first array, we functionalize three sensors in the array with loads 1.0 nm, 1.5 nm and 2.5 nm. The fourth sensor in the array was always kept bare (no functionalization) as a reference. The same procedure is applied for Pd, Ag, Ti and Cr. The load of the metals is chosen based on the 8 Page 8 of 26
knowledge we reported previously in [27]. This phase enabled us to have a general idea about the combined response of each metal towards the different gases. Additionally, it helped us to collect the response of large amount of sensors with different metal functionalization at different loads.
ip t
The collected data were the corner stone for the starting of step II. There are an infinite number of possible combinations for the different metal functionalization and the different loads that can be
cr
deposited on each individual sensor in the sensor array. In order to decrease such complexity, in
us
step II, we simulated the response of different combination, which seemed to be the optimum solution to decrease the development time. Different combinations of metal at different loads
an
were used to build virtual arrays. We developed a Matlab code to simulate the results of these different combinations. The choice of metals at different loads was based on the modification that
M
such functionalization can introduce to the film resistance, moreover whether such metal at that
d
specific load enhances or degrades the response towards one or more test gas. The optimum case
te
is having a certain combination, where each individual sensor behaves completely different towards each one of the four test gases. After different simulation iterations, we were able to
Ac ce p
reach the optimum combination of different metals at specific loads that can select between the four test gases. Step III is the fabrication and characterization of such combination as well as comparing the measured data to the simulating results.
Results and Discussion
Optimum CNT Thin-Film Thickness
The sensor arrays fabricated utilize CNT thin-films with rather low tube densities. At such densities, the film operates near its percolation threshold, thereby reducing the number of metallic pathways in the network and amplifying the contribution of the semiconducting pathways. Such film density results in the highest possible sensor sensitivity towards the four test gases. The resistance of the CNT layer at such density was ~ 8 kΩ. We obtained promising results in terms 9 Page 9 of 26
of distinction between the different test gases when the CNT network is functionalized with different loads of Au and Cr (fig S1). Nevertheless the obtained results were not reproducible. One possible explanation for such a behavior could be that the sensors before the
ip t
functionalization process were not identical, in terms of the resistance of the CNT layer. In order to investigate the effect of such variation in the resistance on the PCA results, we characterized
cr
two bare sensor arrays. The results obtained are shown in fig. 3 and the corresponding resistances
us
of the individual sensor in each array are shown in table 1. We can observe two main points from
Ac ce p
te
d
M
an
the shown results:
Figure 3 PCA plots for two bare CNT-based gas sensor arrays sprayed at percolation threshold
10 Page 10 of 26
Both sensor arrays show separation in the PCA plot between the different gases, which should not be the case as the sensing element in each array are supposed to be „identical‟. PCA plots obtained from each array are not identical, which mainly due to the difference
M
an
us
cr
range of resistances of the individual sensors in each array.
ip t
Table 1: Resistance variation of bare CNT-based gas sensor arrays deposited at percolation threshold
d
The variation in the resistance of the CNT thin-film is one common feature when the layers are
te
deposited at percolation threshold. At such low density the deposited layers are highly uniform
Ac ce p
but the resistance is highly sensitive to the number of deposited tubes. Depositing thicker film could omit such a problem but it would result immediately in decreasing the sensitivity of the sensor towards all test gases because of the domination of the metallic tube in that case. This implies that we have to trade-off the sensitivity in order to stabilize the resistance of the film. In order to achieve the optimum thickness for the CNT layer, we prepared samples at different film thicknesses besides the percolation threshold. The results obtained from all the fabricated samples are shown in fig. 4. The plot represents the relation between the resistance of the individual sensors in the different arrays and their corresponding normalized response (the mean normalized response of the four test gases). We can distinguish between three different film thicknesses, percolation threshold (blue circles), intermediate thickness (red squares) and high 11 Page 11 of 26
thickness (black rhombus). At the percolation threshold, a large variation can be observed in the resistance of the individual sensors as well as the corresponding normalized response. On the other hand, at the high film thickness the resistance and the normalized response are highly
ip t
stable. Nevertheless, the obtained normalized response is very low towards all test gases, which in some cases lower than the noise level. The intermediate film thickness is the one more suitable
cr
for the sensor array fabrication. At such thickness the resistance of the film is relatively stable
te
d
M
an
us
and the normalized response is sufficient to detect the different gases.
Ac ce p
Figure 4 Relation between the resistance and the normalized response for the bare CNT sensing element in all the fabricated sensor arrays, which are sprayed at different thicknesses: percolation threshold (circle), intermediate thickness (square) and high thickness (rhombus)
By achieving the optimum CNT film thickness, we fabricated different sensor arrays functionalized with same metal such as Au, Pd, Cr, Al, Ag, Ti but at different loads. Some of these metals were capable to achieve partial selection between the test gases (data not shown here). The main aim of such step was to collect the responses of the sensing element functionalized with different metals at different loads which was the corner stone in implementing step II in the development framework. Simulation of Virtual Sensor Array
12 Page 12 of 26
Each sensor array fabricated till now consists of four sensors, three are functionalized with the same metal at different loads and the fourth sensor was always kept bare. Effectively, these sensor arrays consisted only of two types of sensors, as we assume that each type of metal
ip t
functionalization influences the CNT sensor in similar manner, although the functionalization load might be different. In this step, we use the data we collected for the entire previously
cr
measured sensors in order to build a virtual sensor array to achieve the optimum combination of
us
metal functionalization and their corresponding loads that can separate and select between the four test gases. To carry on such a step, we implemented a Matlab program that simulates the
an
PCA result of the virtual arrays to deliver in a quick and efficient way the most promising combination. The input file for this program is the responses of all the individual sensors in the
M
fabricated sensor arrays based on one metal functionalization. The developed program allows
d
selecting sensors from different arrays to be combined in one virtual array and simulate the result
te
of such combination. This mimic the behavior we expect from the real array. The schematic in
Ac ce p
fig. 5 shows the working principle of such program.
Figure 5 A schematic shows the working principle of the developed Matlab program that allow simulating virtual arrays.
PCA plots in the previous sections were always two-dimensional because in all the discussed cases the third principle component almost had no influence on the outcome result. Applying one 13 Page 13 of 26
type of functionalization results in sensors with correlated results in the PCA plot, hence most of the information and the variance within the data are only the first and the second PCs. By combining different metals, the third component will have bigger influence, so the best
ip t
presentation of data is in 3D plots. Since sensor arrays functionalized with Au, Pd and Cr delivered the most adequate and diverse results in terms of stability as well as partial gas
cr
separation in the PCA plot, we chose these three metals for the different combination in the
us
virtual sensor array. After several simulation iterations and building virtual sensor arrays with only two or three sensors as well, we found that the best combination consists of an array of four
an
sensors: first sensor is bare, the second sensor is functionalized with Au (1.5 nm), the third sensor is functionalized with Pd (0.2 nm) and the fourth sensor is functionalized with Cr (1.0 nm). The
M
superiority in the normalized response of Au functionalization towards NH3 and for Pd towards
d
CO2 with Cr functionalization resulted in CNT-based gas sensor array that can isolate NH3 in one
te
plane, separate between CO, CO2 and ethanol in the other two planes as shown in fig. 6. The data are presented in two different angles to give a better perspective of the 3D plots. The simulated
Ac ce p
sensor array showed a strong influence of the second and the third principle components, which resulted in better distinction between the different gases. Realization of Optimum CNT-Based Gas Sensor Array In this step, we spray deposited CNT solution, which has 90% semiconducting tubes instead of the intrinsic CNT solution (67% semiconducting tubes) that we used in all the previous experiments. Using a semiconducting solution enabled us to deposit significantly thick CNT layers far beyond the percolation threshold and obtaining stable film resistances similar to the values obtained previously (~1 k). The higher network density guarantees higher stability in the resistance, thus boost the reproducibility of the sensor. For further confirmation on the stability of the used semiconducting solution, we repeated the previously performed procedure of fabricating 14 Page 14 of 26
ip t cr us an M d te Ac ce p
Figure 6 PCA plots for the simulated sensor array functionalized with 1.5 nm of Au, 0.2 nm of Pd, 1.0 nm of Cr and the fourth sensor is bare. The data are presented in two different angles to give better perspective of the 3D plot.
one bare sensor array and evaluating its performance in terms of the resistance stability, normalized response and PCA plots. The obtained results are shown in fig. S2 After successfully simulating the most promising combination of metal functionalization, which consists of Au (1.5 nm), Pd (0.2 nm) and Cr (1.0 nm) and a bare sensor, we fabricated and characterized such sensor array. Figure 7 depicts the measured resistance over time for the four test gases in this study, NH3, ethanol, CO and CO2. After each exposure cycle the sensors were heated up to 80 °C (active recovery) in order to regain their original resistance. Further discussion 15 Page 15 of 26
on the difference between active recovery and passive recovery was previously reported by our group in [6]. The minimum shift in the baseline reflects the stability of the sensors‟ resistances
Ac ce p
te
d
M
an
us
cr
ip t
after the successive measurements.
Figure 7 Plot of the measured sensors resistance over time. Each graph is divided into four segments, based on four-exposure/recovery cycle at different gas concentrations. The exposure interval is indicated by the area between closely adjacent dotted lines marking the start and the end. The different color indicates the response of each single sensor in the array. The four test gases shown here are (a) NH3, (b) Ethanol, (c) CO, (d) and CO2.
16 Page 16 of 26
We further investigated the relation between the resistance of the individual sensors during the successive measurements towards the different test gases and their normalized response. In order to perform such kind of characterization, we monitored the resistance of the individual sensors in
ip t
the array at the moment before gas exposure at each concentration of the different test gases, we are showing here only the resistance before the highest concentration, since we assume that most
cr
of the changes that occur during the measurement is reflected in the last exposure cycle. Taking a
us
closer look at the results obtained in fig. 8, we can observe a stable resistance (~1 kΩ) for the Au, Pd and bare sensors during the successive measurements. The resistance of the Cr sensor (~3.5
an
kΩ) was almost stable for the first three measurements and then dropped to almost half of its value at the last measurement.
M
It‟s worth mentioning here that Au and Pd functionalization results in decreasing the resistance of
d
the CNT thin-film. On the other hand, Cr functionalization results in increasing the resistance of
te
the film. This implies that the interaction between the CNT layer and the Cr NPs is entirely different than its interaction with Au and Pd. This topic is still under further investigation by our
Ac ce p
group. The PCA plot showed a clear separation of NH3 data points in one plane, while ethanol, CO and CO2 remain relatively in another plane. Nevertheless, ethanol is clearly separated from CO and CO2, which are relatively clustered together. By comparing this result to the results obtained from the simulated sensor array in the last section, we can clearly see the same pattern and the same behavior for the different data points. Such results are remarkable due to the fact that the data points used in the simulated version are extracted from measurements performed in different time periods and at slightly different ambient conditions.
17 Page 17 of 26
ip t cr us an M d te
Ac ce p
Figure 8 (a) The resistance of the individual sensors and their normalized response towards the four test gases (NH3, ethanol, CO and CO2) for the CNT-based gas sensor array, which is functionalized with Au 1.5 nm, Cr 1.0 nm and Pd 0.2 nm. In (b) & (c) is the PCA plots for such sensor array in 2D and 3D respectively.
This is a strong evidence for the reproducibility of the sensors fabricated at this resistance range. Note that in the previous discussions, we concentrated mainly on the stability of the resistance as a crucial issue for the outcome of the sensor array. It is important to mention that the outcome result is highly sensitive as well to any minor error in the lithography process, which is the first step in the fabrication process. In general, the result of the sensor array is highly sensitive to any minor change in the process steps as well as the response of the individual sensor. Regarding the sensing mechanism for the CNT-based gas sensors, there is no complete theory that fully explains the working principles [25]. The debate is whether the CNT layer or metal- CNT interface plays the main role in the sensitivity of the gas sensor. We do not favor the theory where the sensitive 18 Page 18 of 26
part is the metal/CNT interface since we reported previously CNT-based gas sensors that have relatively high sensitivity based on CNT electrodes instead of metal electrodes [6]. In general, metallic functionalization enhances the response of CNT–based gas sensors. Such an
ip t
effect is mainly attributed to the catalytic activity of the deposited NPs, which boost the gas sensitivity of the CNT film. It is worth mentioning here that the catalytic activity of Au NPs
cr
diminishes considerably when the particle size exceeds a certain limit [34]. Due to the fact that
us
post treatment process for the CNT layer is performed in diluted HNO3, defect sites and oxidation are introduced on the surface [35] which allows binding the NPs to the CNT films via the
an
process-induced carboxylic groups. Upon gas exposure, in case of NH3, the gas molecules coordinate with the formed NPs [36]. Our hypothesis is that the metallic NPs allow binding more
M
ammonia molecules to the same defect site instead of only one molecule per pristine CNTs. This
d
implies that a stronger charge transfer to the CNT network takes place in the presence of
te
nanoparticles. Another effect that takes place when depositing NPs on top of the CNT film is the formation of a „non-traditional Schottky barrier‟ at the interface between the semiconducting
Ac ce p
tubes and the metal NPs [37], [38]. Such Schottky barriers could be further modulated upon gas exposure, thus leading to an additional change in sensitivity of the sensors. Since in our case the work function of the functionalized film changes only slightly with coverage as shown in fig. S3, we cannot quantitatively assess such hypothesis. Characterization Towards Gas Mixture Another key aspect for the evaluation of the CNT-based gas sensor array performance is investigating its response towards mixture of two test gases simultaneously. Such characterization procedure is a major step towards realizing CNT-based gas sensor array in wide range of applications where the detected gas molecules consists of different concentration and from different gases rather than a single pure gas. In order to perform such a step, we examined 19 Page 19 of 26
the response of the optimum CNT-based gas sensor array, described in the last section, towards different mixtures of NH3 and CO as well as NH3 and ethanol. We limited this type of characterization only to the three test gases, which has pure N2 as a carrier
ip t
gas (NH3, CO and ethanol) to obtain systematic and consistent results,. Moreover, the total concentrations were always 10 ppm, 25 ppm, 50 ppm and 100 ppm independent from the ratios
cr
of the two gases. Before characterizing the sensor array towards the mixture, we measured first
us
its response towards each gas separately. We chose ratios of 1:1, 1:2 and 2:1 between the NH3 and ethanol as well as NH3 and CO for such characterization procedure. Figure 9a shows the
an
PCA plot for the CNT sensor array towards NH3 and CO as well as the different mixtures between them. The response towards equal ratios of the two gases lies almost in the middle
M
between the data points for each gas separately. By increasing the ratio of NH3 w.r.t CO, its data
d
points tends to move towards the NH3 data points and similarly vice versa. Investigating the
te
response of the sensor array towards different mixtures between NH3 and ethanol reveals similar results as shown in figure 9b. By achieving a separation between any two gases, the response
Ac ce p
towards a mixture of those two gases should lie in the plane between the data points of each gas separately. Such preliminary results indicate the possibility of identifying not only the type of the different gases in the mixture but their concentration as well
20 Page 20 of 26
ip t cr us an M d te
Ac ce p
Figure 9 PCA plots for a CNT-based gas sensor array functionalized with 1.5 nm of Au, 1.0 nm of Cr, 0.2 nm of Pd and a bare sensor under the exposure of (a) NH 3, CO and three mixture between the two gases and (b) NH3, ethanol and three mixtures between the two gases. In both cases the ratios of the mixture were 1:1, 1:2 and 2:1
Conclusion
We reported the complete development of a selective CNT-based gas sensor array functionalized with different metallic NPs. We introduced the principle component analysis method for analyzing the response of the sensor array. Reaching the selective CNT sensor array with optimum combinations of the metallic NPs was done through a developed framework strategy. We adopted the concept of building and simulating CNT virtual sensor arrays depending on the data collected from sensors with different functionalization at different load. Such step played an important role in decreasing the development time. In order to achieve a selective pattern in the 21 Page 21 of 26
PCA plot, we modified the spraying parameter to obtain a rather thicker layer of the CNT network. Such modification resulted in a small sacrifice in the sensitivity of the sensors. Nevertheless, the achieved sensitivity remains among the highest reported in literature. Taking
ip t
into consideration the short exposure interval used which is only 100 s for all test gases. The best combination of the metallic functionalization was Au at a load of 1.5 nm, Cr at a load of 1.0 nm
cr
and Pd at a load of 0.2 nm. The fourth sensor was kept bare. Such combination resulted in a clear
us
distinction between the four test gases (NH3, ethanol, CO and CO2) at an acceptable response and a stable film resistance. We took the characterization of CNT sensor array one step further where
an
we investigated its response towards a mixture of two gases simultaneously. The result showed that the data points of the mixture lie on the plane between the data points of the each gas
M
separately. The lifetime of the functionalized CNT – based gas sensors remains questionable as
d
further investigation should be held regarding the effect on the sensor performance when the
te
nanoparticle are exposed to the environment for a long time, such exposure might cause oxidation
Ac ce p
to the metallic particle which in turn might degrade the response of the sensor.
References [1]
S. IIjima, “Helical microtubules of graphitic carbon,” Nature, 354 (1991) 56–58.
[2]
Q. Cao and J. a. Rogers, “Ultrathin Films of Single--Walled Carbon Nanotubes for Electronics and Sensors: A Review of Fundamental and Applied Aspects,” Adv. Mater., 21 (1) (2009) 29–53.
[3]
L. Hu, D. S. Hecht, and G. Gruener, “Carbon nanotube thin films: fabrication, properties, and applications.,” Chem. Rev., 110 (10) (2010) 5790–844.
[4]
T. Zhang, S. Mubeen, N. V Myung, and M. a Deshusses, “Recent progress in carbon nanotube-based gas sensors.,” Nanotechnology, 19 (33) (2008) p. 332001.
[5]
D. Zhang, K. Ryu, X. Liu, E. Polikarpov, J. Ly, M. E. Tompson, and C. Zhou, “Transparent, conductive, and flexible carbon nanotube films and their application in organic light-emitting diodes,” Nano Lett., 6 (9) (2006) 1880–6. 22 Page 22 of 26
A. Abdellah, A. Abdelhalim, M. Horn, G. Scarpa, and P. Lugli, “Scalable Spray Deposition Process for High-Performance Carbon Nanotube Gas Sensors,” IEEE Trans. Nanotechnol., 12 (2) (2013) 174–18.
[7]
A. Abdellah, A. Abdelhalim, F. Loghin, P. Köhler, Z. Ahmad, G. Scarpa, and P. Lugli, “Flexible Carbon Nanotube Based Gas Sensors Fabricated by Large-Scale Spray Deposition,” IEEE Sensors Journal, 13 (10) (2013) 4014-4021.
[8]
J. Kong, “Nanotube Molecular Wires as Chemical Sensors,” Science, 287 (5453) (2000) 622–625.
[9]
Y. Wang, Z. Zhou, Z. Yang, X. Chen, D. Xu, and Y. Zhang, “Gas sensors based on deposited single-walled carbon nanotube networks for DMMP detection.,” Nanotechnology, 20 (34) (2009) 345502.
us
cr
ip t
[6]
an
[10] O. K. Varghese, P. D. Kichambre, D. Gong, K. G. Ong, E. C. Dickey, and C. a. Grimes, “Gas sensing characteristics of multi--wall carbon nanotubes,” Sensors Actuators B Chem., 81 (1) (2001) 32–41.
d
M
[11] Y. Wang, N. Hu, Z. Zhou, D. Xu, Z. Wang, Z. Yang, H. Wei, E. S.-W. Kong, and Y. Zhang, “Single-walled carbon nanotube/cobalt phthalocyanine derivative hybrid material: preparation, characterization and its gas sensing properties,” J. Mater. Chem., 21(11) (2011) 3779.
te
[12] Y.-T. Jang, S.-I. Moon, J.-H. Ahn, Y.-H. Lee, and B.-K. Ju, “A simple approach in fabricating chemical sensor using laterally grown multi-walled carbon nanotubes,” Sensors Actuators B Chem., 99 (1) (2004) 118–122.
Ac ce p
[13] M. Arab, F. Berger, F. Picaud, C. Ramseyer, J. Glory, and M. Mayne-L‟Hermite, “Direct growth of the multi-walled carbon nanotubes as a tool to detect ammonia at room temperature,” Chem. Phys. Lett., 433 (1-3) (2006) 175–181. [14] A. A. Tomchenko, G. P. Harmer, B. T. Marquis, and J. W. Allen, “Semiconducting metal oxide sensor array for the selective detection of combustion gases,” Sensors Actuators B Chem., 93 (1-3) (2003) 126–134. [15] B. T. Marquis and J. F. Vetelino, “A semiconducting metal oxide sensor array for the detection of NOx and NH3,” Sensors Actuators B Chem., 77 (1-2) (2001) 100–110. [16] F. Zee and J. W. Judy, “Micromachined polymer--based chemical gas sensor array,” Sensors Actuators B Chem., 72 (2) (2001) 120–128. [17] J. W. Gardner and P. N. Bartlett, Sensors and Sensory Systems for an Electronic Nose. Dordrecht: Springer Netherlands, 1992. [18] J. W. Gardner, H. V. Shurmer, and T. T. Tan, “Application of an electronic nose to the discrimination of coffees,” Sensors Actuators B Chem., 6 (1-3) (1992) 71–75. 23 Page 23 of 26
[19] T. C. Pearce, J. W. Gardner, S. Friel, P. N. Bartlett, and N. Blair, “Electronic nose for monitoring the flavour of beers,” Analyst, 118 (4) (2003) 371.
ip t
[20] G. Peng, E. Trock, and H. Haick, “Detecting simulated patterns of lung cancer biomarkers by random network of single-walled carbon nanotubes coated with nonpolymeric organic materials.,” Nano Lett., 8 (11) (2008) 3631–5.
cr
[21] H. Abe, T. Yoshimura, S. Kanaya, Y. Takahashi, Y. Miyashita, and S.-I. Sasaki, “Automated odor--sensing system based on plural semiconductor gas sensors and computerized pattern recognition techniques,” Anal. Chim. Acta, 194 (1987) 1–9.
us
[22] T. Nakamoto, A. Fukuda, and T. Moriizumi, “Perfume and flavour identification by odoursensing system using quartz--resonator sensor array and neural-network pattern recognition,” Sensors Actuators B Chem., 10 (2) (1993) 85–90.
an
[23] A. Star, V. Joshi, S. Skarupo, D. Thomas, and J.-C. P. Gabriel, “Gas sensor array based on metal-decorated carbon nanotubes.,” J. Phys. Chem. B, 110 (42) (2006) 21014–20.
M
[24] Y. Lu, C. Partridge, M. Meyyappan, and J. Li, “A carbon nanotube sensor array for sensitive gas discrimination using principal component analysis,” J. Electroanal. Chem., 593 (2006) 105–110.
te
d
[25] H. Guerin, H. Le Poche, R. Pohle, E. Buitrago, M. Fernández-Bolaños Badía, J. Dijon, and A. M. Ionescu, “Carbon nanotube gas sensor array for multiplex analyte discrimination,” Sensors Actuators B Chem., 207 (2015) 833–842.
Ac ce p
[26] P. Bondavalli, P. Legagneux, D. Pribat, A. Balan, and S. Nazeer, “Gas fingerprinting using carbon nanotubes transistor arrays,” Sensors and Actuators B Chem., 140 (2009) 304-312. [27] A. Abdelhalim, A. Abdellah, G. Scarpa, and P. Lugli, “Metallic nanoparticles functionalizing carbon nanotube networks for gas sensing applications,” Nanotechnology, 25 (5) (2014) 055208. [28] T. Takahashi, K. Tsunoda, H. Yajima, and T. Ishii, “Dispersion and Purification of SingleWall Carbon Nanotubes Using Carboxymethylcellulose,” Jpn. J. Appl. Phys., 43 (6A) (2007) 3636–3639. [29] A. Abdelhalim, A. Abdellah, G. Scarpa, and P. Lugli, “Fabrication of carbon nanotube thin films on flexible substrates by spray deposition and transfer printing,” Carbon, 61 (2013) 72-79. [30] R. C. Tenent, T. M. Barnes, J. D. Bergeson, A. J. Ferguson, B. To, L. M. Gedvilas, M. J. Heben, and J. L. Blackburn, “Ultrasmooth, Large-Area, High-Uniformity, Conductive Transparent Single--Walled--Carbon--Nanotube Films for Photovoltaics Produced by Ultrasonic Spraying,” Adv. Mater., 21 (31) (2009) 3210–3216.
24 Page 24 of 26
[31] A. Abdellah, B. Fabel, P. Lugli, and G. Scarpa, “Spray deposition of organic semiconducting thin--films: Towards the fabrication of arbitrary shaped organic electronic devices,” Org. Electron., 11 (6) (2010) 1031–1038.
ip t
[32] A. Falco, L. Cinà, G. Scarpa, P. Lugli, and A. Abdellah, “Fully-sprayed and flexible organic photodiodes with transparent carbon nanotube electrodes.,” ACS Appl. Mater. Interfaces, 6 (13) (2014) 10593–601.
cr
[33] P. C. Jurs, G. A. Bakken, and H. E. Mcclelland, “Computational Methods for the Analysis of Chemical Sensor Array Data from Volatile Analytes,” Chem. Rev., 100 (7) (2000) 2649–2678.
us
[34] A. Corma and H. Garcia, “Supported gold nanoparticles as catalysts for organic reactions.,” Chem. Soc. Rev., 37 (9) (2008) 2096–126.
an
[35] V. Lordi, N. Yao, and J. Wei, “Method for Supporting Platinum on Single--Walled Carbon Nanotubes for a Selective Hydrogenation Catalyst,” Chem. Mater., 13 (3) (2001) 733– 737.
M
[36] A. Bilic, J. R. Reimers, N. S. Hush, and J. Hafner, “Adsorption of ammonia on the gold (111) surface,” J. Chem. Phys., 116 (20) (2002) 8981.
te
d
[37] D. R. Kauffman, D. C. Sorescu, D. P. Schofield, B. L. Allen, K. D. Jordan, and A. Star, “Understanding the Sensor Response of Metal--decorated Carbon Nanotubes,” ACS Nano, 10 (3) (2010) 958–963.
Ac ce p
[38] D. R. Kauffman and A. Star, “Chemically Induced Potential Barriers at the Carbon Nanotube Metal Nanoparticle Interface,” Nano Lett., 7 (7) (2007) 1863-1868.
25 Page 25 of 26
*Author Biographies
ip t
Ahmed Abdelhalim received the B.Sc. degree in electronics engineering from the German University in Cairo, Cairo, Egypt in 2009, and the M.Sc. degree in communication and electronics engineering from the Technische Universität München, Munich, Germany, in 2011, where he is currently working towards the Ph.D. degree at the Institute for Nanoelectronics. His current research interests include fabrication and characterization of Carbon nanotubes chemical gas sensors based on spray deposition technique as well as different functionalization methods to enhance the sensor performance. Maximilian Winkler
an
us
cr
Florin Loghin received the B.Sc. degree in Electrical Engineering from the Technische Universität München, Munich, Germany in 2012, and the M.Sc. degree in electrical engineering from the Technical University of Munich, Munich, Germany, in 2015, where he is currently working toward the Ph.D. degree at the Institute for Nanoelectronics. His research interests include fabrication and characterization of carbon nanotubes based field effect transistors as well as developing spray deposition techniques for various organic materials. Christopher Zeiser
Ac ce p
te
d
M
Paolo Lugli received the Laurea degree in Physics from the University of Modena, Italy, in 1979. In 1981, he joined Colorado State University, Fort Collins, CO, USA, where he received the Mas- ters of Science in Electrical Engineering in 1982, and the Ph.D. in 1985. In 1985, he joined the Physics Department with the University of Modena, Reggio Emilia, Italy, as Research Associate. From 1988 to 1993 he was an Associate Professor of “Solid State Physics” at the “Engineering Faculty” with the University of Rome “Tor Vergata”. In 1993, he became Full Professor of “Optoelectronics” with the same University. In 2002, he joined the Technical University of Munich, where he was appointed as a head of the newly created Institute for Nanoelectronics. His current research interests include nanoimprint lithography, the modeling, fabrication and characterization of organic devices for electronics and optoelectronics applications, the design of circuits and architectures for nanostructures and nanodevices, the numerical simulation of microwave semiconductor devices, and the theoretical study of transport processes in nanostructures. He is author of more than 250 scientific papers and co-author of the books “The Monte Carlo Modelling for Semiconductor Device Simulations”(Springer, 1989) and “High Speed Optical Communications” (Kluver Academic, 1999). In 2004, he served as General Chairman of the IEEE International Conference on Nanotechnology, Munich, Germany. He is the member of the German National Academy of Science and Engineering. Alaa Abdellah received the B.Sc. degree in Electronics and Communications Engineering from the Cairo University, Cairo, Egypt, in 2005, and the M.Sc. degree in High-Frequency Engineering and Optoelectronics from the Technische Universität München, Munich, Germany, in 2007. He is currently a Member of the scientific staff with the Institute for Nanoelectronics, Technische Universität München, Munich, Germany, where he obtained the Ph.D. degree in 2012. His current research interests include the area of organic and nanostructured electronics, focusing on large-area thin-film devices utilizing solution-processable organic semiconductors, and nanomaterials.
Page 26 of 26