Toward Highly Sensitive and Energy Efficient Ammonia Gas Detection

Nov 30, 2017 - Fabrication and comparative analysis of the gas sensing devices based on individualized single-walled carbon nanotubes of four differen...
0 downloads 7 Views 2MB Size
Article Cite This: ACS Sens. 2018, 3, 79−86

pubs.acs.org/acssensors

Toward Highly Sensitive and Energy Efficient Ammonia Gas Detection with Modified Single-Walled Carbon Nanotubes at Room Temperature Luis Antonio Panes-Ruiz,† Mehrdad Shaygan,†,‡ Yangxi Fu,† Ye Liu,† Vyacheslav Khavrus,†,§ Steffen Oswald,§ Thomas Gemming,§ Larysa Baraban,†,∥ Viktor Bezugly,*,†,§,∥ and Gianaurelio Cuniberti†,∥ †

Institute for Materials Science and Max Bergmann Center of Biomaterials and ∥Center for Advancing Electronics Dresden (CfAED), TU Dresden, 01062 Dresden, Germany ‡ Advanced Microelectronic Center Aachen (AMICA), AMO GmbH, Otto-Blumenthal-Str. 25, 52074 Aachen, Germany § IFW Dresden, P.O. Box 270016, D-01171 Dresden, Germany S Supporting Information *

ABSTRACT: Fabrication and comparative analysis of the gas sensing devices based on individualized single-walled carbon nanotubes of four different types (pristine, boron doped, nitrogen doped, and semiconducting ones) for detection of low concentrations of ammonia is presented. The comparison of the detection performance of different devices, in terms of resistance change under exposure to ammonia at low concentrations combined with the detailed analysis of chemical bonding of dopant atoms to nanotube walls sheds light on the interaction of NH3 with carbon nanotubes. Furthermore, chemoresistive measurements showed that the use of semiconducting nanotubes as conducting channels leads to the highest sensitivity of devices compared to the other materials. Electrical characterization and analysis of the structure of fabricated devices showed a close relation between amount and quality of the distribution of deposited nanotubes and their sensing properties. All measurements were performed at room temperature, and the power consumption of gas sensing devices was as low as 0.6 μW. Finally, the route toward an optimal fabrication of nanotube-based sensors for the reliable, energy-efficient sub-ppm ammonia detection is proposed, which matches the pave of advent of future applications. KEYWORDS: gas sensor, ammonia detection, doped carbon nanotube, semiconducting carbon nanotube, adsorption mechanism

T

ammonium gas sensors should represent a part of the industrial process, being (a) miniaturized and easy integrated into the available facility and (b) contributing to its energy efficiency via, e.g., low power consumption to be used in networks of numerous independently working sensors. Further, such sensors for permanent remote monitoring of environment must be able to detect low concentrations of ammonia. Integration of nanomaterials as active elements of sensors is an efficient approach to achieving an extremely high sensitivity and low power consumption. Due to their particular nanoscale features like reduced dimensionality and extraordinarily high surface-to-volume ratio, nanomaterial-based sensors can potentially offer low detection limits, a considerably increased speed of response as well as reduced production costs.13−17 Several nanomaterial classes like nanostructured metal oxides, metallic nanoparticles, metal complexes, organic polymers, and carbon-

he development of highly efficient gas sensing technologies is the focus of research worldwide due to their application in different areas. In industry, gas sensors are used to monitor volatile organic compounds (VOCs) like ethanol, ethylene, etc., in food quality control1 and to prevent potential accidents due to liquefied petroleum gas (LPG),2 methane,3,4 or H2 leaks;5 in environmental studies, to preserve air quality by monitoring potentially harmful gases like formaldehyde6 or CO2;7 and in biomedical applications, for the diagnosis of specific diseases.8 A new noninvasive and potentially inexpensive route for diagnosis relies on the analysis of exhaled breath samples to detect some specific VOCs and/or gases characteristic to certain diseases like different types of cancer,9 liver cirrhosis, kidney failure,10 and diseases caused by Helicobacter pylori.11 The last three diseases can be identified by abnormal breath ammonia levels with NH3 concentrations below 1 ppm. This particular molecule not only plays a role as an indicator of diseases, but it is also the most common gas found in different industrial processes and one of the most dangerous already being harmful at concentrations above 25 ppm.12 Thus, © 2017 American Chemical Society

Received: May 29, 2017 Accepted: November 30, 2017 Published: November 30, 2017 79

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors

We have exploited a simple and robust method for fabricating gas sensor devices in which direct interaction of individualized SWCNTs with ammonia can be achieved. Gas sensor devices based on pristine single-walled carbon nanotubes (pSWCNTs), boron-doped nanotubes (B-SWCNTs), nitrogendoped nanotubes (N-SWCNTs), and semiconducting nanotubes (sc-SWCNTs) were fabricated and systematically studied at the same conditions. All nanotubes were dispersed without surfactants and the doping was realized via introduction of N or B atoms into the host carbon lattice of nanotubes, not via onsurface molecular dopants, that simplifies interpretation of results on the ammonia−nanotube interaction. Finally, all measurements were performed at room temperature, and the issue of energy efficiency of SWCNT-based ammonia sensors was addressed.

based nanomaterials have already been extensively studied in the literature as active elements of gas sensors.18 For instance, nanostructured metal oxide based gas sensors have demonstrated an efficient detection of low concentrations of carbon monoxide and several other gases and organic vapors; however, their high operational temperatures (200 °C and above)19 together with complicated and expensive fabrication techniques often limit their application. Another example is sensors based on organic polymers, which have relatively low production costs, work at room temperature, and are very sensitive to certain analytes like acetone or methanol vapors. However, they have shown lack of selectivity, a rather low stability regarding temperature changes, and reduced lifetime.18 Carbon nanomaterials have emerged as a promising alternative to resolve the aforementioned constraints of nanomaterial-based chemiresistor gas sensors. In contrast to polycrystalline materials or polymer films, high quality crystal lattices of carbon nanomaterials have shown high carrier mobility, ballistic charge transport, and low current noise, necessary to ensure good transduction properties.20 Moreover, single-walled carbon nanotubes (SWCNTs) are quasi-onedimensional objects with ultimately high surface-to-volume and length-to-diameter ratios. Their typical diameters are 0.8 to 1.6 nm that is comparable to sizes of some analyte gas molecules. This means that a single molecule being adsorbed at the sidewall of a nanotube has the ability to locally influence electronic structure of the latter that can be measured as the change of electric current through this nanotube. Therefore, extremely low concentration of analyte gases can be potentially detected with SWCNTs. Finally, carbon nanomaterials like carbon nanotubes and graphene possess an outstanding structural stability that guarantees their long-term operation capabilities. SWCNT-based chemiresistor sensors for the detection of ammonia gas have been proposed and studied in recent years. Since the seminal paper on the detection of down to 0.1% of ammonia in Ar with a single semiconducting nanotube,21 a number of publications have appeared focusing on the optimal modification and integration of carbon nanotubes in chemical sensors to achieve lower detection limits. Recently, thin-film pristine SWCNT based chemiresistor sensor has been shown to detect sub-ppm concentrations of ammonia in air at ambient conditions (room temperature and a fixed humidity).22 Still, there are some contradictive opinions on how sensitivity of SWCNT-based sensors toward ammonia can be increased by means of SWCNT doping. Also, there is no full understanding about the interaction of ammonia with nanotubes. It is argued23−26 that NH3 molecules do not attach to the sidewalls of pristine SWCNTs but rather to defects or add-atoms only. Some studies indicated that nitrogen-doped nanotubes (NSWCNTs) may increase sensitivity to ammonia,26,27 and another one argued that boron-doped nanotubes (B-SWCNTs) must have higher sensitivity because of strong binding of NH3 molecules to B atoms and charge transfer.28 However, these studies suffered from a limited number of considered atomic configurations as well as computational limitations. On the other hand, a field-effect response can be expected21 if semiconducting-only nanotubes (sc-SWCNTs) are used for ammonia detection. In this paper, we address two main questions: what is the nature of interaction of NH3 molecules with SWCNTs and which modification of pristine carbon nanotubes is optimal for the most efficient detection of ammonia at low concentrations.



EXPERIMENTAL SECTION

Doping of Carbon Nanotubes. Pristine SWCNTs were purchased from OCSiAl company (TUBALL) as a mixture of metallic and semiconducting nanotubes. Boron- and nitrogen-doped SWCNTs were synthesized using these TUBALL nanotubes as a starting material by high-temperature substitution reaction and hydrothermal treatment, respectively.29,30 (Detailed description in SI.) The obtained samples had 2.5% of boron in B-SWCNTs and 2.3% of nitrogen atoms in N-SWCNTs. Sc-SWCNTs were purchased from Sigma-Aldrich as a sorted out mixture containing 98% of semiconducting SWCNTs according to producer specifications. Device Fabrication. SWCNTs of each type were dispersed in Nmethyl-2-pyrrolidone (NMP) purchased from Sigma-Aldrich in a proportion 1 mg/mL, sonicated for 2 h, and centrifuged for 2 h (14000 rpm) to remove agglomerations. Then, dispersion dilutions of 2% v/v in NMP were prepared and drop-casted onto the interdigitated electrode area of fabricated devices (for details on device fabrication steps see SI). A schematic view of the sensing device is shown in Figure 1a. The devices were heated to 50 °C for 5 min and then to 100 °C for 10 min on a hot plate. In order to fabricate devices with a reduced amount of deposited SWCNTs the last heating step was shortened to 5 min. The remaining NMP dispersion was removed from electrodes with the air gun. For some gas sensing devices based on sc-SWCNTs, the nanomaterial was drop-casted onto the Si wafer before the lithography process to demonstrate that sensing response is not dependent on placing SWCNTs under or onto metallic electrodes. Characterization. X-ray photoelectron spectroscopy (XPS) measurements were performed to determine the chemical constitution of SWCNT samples. A PHI 5600 CI (Physical Electronics) system with a hemispherical energy analyzer was used. Low-voltage scanning electron microscopy (LV-SEM) was used for investigation of SWCNT distribution between interdigitated electrodes (Figure 1b). The sensing devices were extensively characterized with a probe station setup before (to exclude shortcuts or leakage in the comb structure) and after deposition of SWCNTs (to get the IV-characteristics of the devices before gas exposure). More details on characterization can be found in SI. Ammonia Exposure Experiments. The gas sensing devices were exposed to 1.5, 2.5, 5, 10, and 20 ppm of ammonia at room temperature in a specially designed gas exposure setup31 (see Figure 2). The exposures to ammonia were separated by 15 min of recovery time under pure N2 flow (2000 sccm/min). A constant voltage of 0.1 V was applied between contact pads and the current was read (one time per second) using Keithley 2602 source meter.



RESULTS AND DISCUSSION Content of Dopants. It was suggested that the sensitivity of SWCNTs to ammonia at low concentration can be improved by inserting dopant atoms such as nitrogen26,27 or boron28 into the carbon structure of the sidewalls. Here, the local 80

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors

quaternary-N (substitution of C atom with N atom in the hexagonal carbon structure), 1.1 at. % to pyrrolic-N, and 0.3 at. % to pyridinic-N (see Figure 3c). Quaternary-N is donating one electron to the delocalized π-electron system of nanotubes (ndoping), whereas the last two types are electronically neutral defects. The remaining 0.3 at. % of nitrogen atoms correspond to NO2 molecules absorbed on the SWCNT surface after the hydrothermal treatment which are finally removed during the sensor fabrication process. In B-doped SWCNT sample, the total amount of oxygen bound to carbon atoms was increased up to 9.4 at. % since B2O3 was used in high-temperature treatment of nanotubes. As can be seen in Figure 3d and Figure 3e, oxygen atoms are bound to carbon, boron, or to carbon and boron atoms. The total amount of bound B atoms is 2.5 at. % with the main contribution of B4C bonding configuration (1.6 at. %) corresponding to a local formation of boron carbide structure. Boron atoms which are bound to 3 neighbor C atoms corresponding to a substitution of a single C atom with B atom in the hexagonal structure of pristine SWCNTs (graphitic B) and forming electron acceptor state are present at 0.1 at. % only. The other B atoms are bound to C together with oxygen atoms (0.6 at. %) or are present on the surface of SWCNTs as rests of B2O3 (0.2 at. %) (see Figure 3e). Unsorted p-SWCNTs samples contain a mixture of semiconducting and metallic nanotubes with typical ratio of 2 to 1. Electronic properties of metallic nanotubes are only slightly affected by doping.32 Therefore, doping is mainly influencing electronic properties of semiconducting SWCNTs that is discussed below. The analysis of the XPS data suggests that p-SWCNT sample has bound oxygen atoms that leads to the shift of Fermi level below the top of the valence band and ptype conductivity.33 The amount of electron-donating quaternary-N in N-SWCNT sample is too low (0.6 at. %) to lead to a shift of Fermi level to the conduction band taking into account the increase of oxygen content by 4.1 at. %. The difference of N-SWCNT sample to the pristine one is mainly in the presence of new chemically active sites on nanotube sidewalls. In contrast, Fermi level in semiconducting nanotubes of the BSWCNT sample is expected to be shifted further away from the top of the valence band compared to those in the pristine sample because of both additional O atoms and graphitic boron. Consequently, the amount of free charge carriers (holes) is raised that leads to an increase of intrinsic conductivity in B-doped SWCNTs. Sensing Response. At first, we assessed the influence of doping on the sensitivity of the SWCNT-based chemiresistortype gas sensors to ammonia. In Figure 4 the normalized sensing response ΔR/R0 to successive exposures to NH3 with concentrations of 1.5, 2.5, 5, 10, and 20 ppm under room temperature is presented. After each exposure, a few-minutes flushing with pure nitrogen was performed. Figure 4a presents the response of the p-SWCNT-based sensor, whereas Figure 4b shows the response of sensors with N-doped and B-doped SWCNTs as active elements. The three sensors have similar resistance values (250 Ω, 186 Ω, and 160 Ω, respectively) that suggests a similar distribution of SWCNTs between electrodes in all three devices. The normalized sensing response is used to compare sensitivity of different devices. It is defined as the relative change of device resistance and is calculated using the following formula34

Figure 1. (a) Schematic view of the CNT-based gas sensing device with interdigitated electrodes and SWCNTs on Si/SiO2 substrate. (b) SEM image of interdigitated electrode area showing SWCNTs bridging two neighboring gold microelectrodes.

Figure 2. Sketch of gas detection setup. The required concentration of ammonia in N2 is set by mixing pure N2 gas with the 0.5% NH3 in N2 gas mixture that is controlled with mass flow controllers (MFC).

arrangement of atoms, or bonding situation of dopant atoms, may play the key role in sensing effects. In order to characterize the content and configuration of the dopant atoms in BSWCNTs and N-SWCNTs samples, XPS studies were performed. The p-SWCNTs have 2.1 at. % of bound oxygen and no traces of B or N atoms. According to the XPS spectra analysis, the main bonding situations (2.0 at. %) for the incorporation of O atoms in this sample are C−O−C, C−OH, or C−O−OH with a little contribution (0.1 at. %) of CO and O−CO (see Figure 3a). After treatment in nitric acid and partial substitution of excess O atoms with N during the hydrothermal treatment, the total amount of oxygen in NSWCNT sample was increased (compared to p-SWCNT) up to 6.2 at. % (Figure 3b). The total content of nitrogen in NSWCNTs is 2.3 at. % where 0.6 at. % corresponds to 81

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors

Figure 3. XPS data on bonding of O atoms in p-SWCNTs (a), in N-doped SWCNTs (b), and in B-doped SWCNTs (d), N atoms in N-doped SWCNTs (c), and B atoms in B-doped SWCNTs (e). Black lines represent the total XPS signal whereas contributions from different atom arrangements are depicted with colored lines.

Figure 4. Sensing response ΔR/R0 under exposure to different ammonia concentrations of 1.5, 2.5, 5, 10, and 20 ppm: (a) device based on pristine SWCNTs and (b) devices based on N-SWCNTs (purple) and B-SWCNTs (green). An incomplete recovery under pure N2 flow was present in all devices. Dotted lines delimit exposure times with the corresponding concentrations.

response (%) =

strong and weak adsorption of NH3 molecules to SWCNT sidewalls. As can be seen in Figure 4 almost no desorption during N2 flushing after 1.5 and 2.5 ppm ammonia exposure takes place. This can be attributed to a strong binding of NH3 molecules (chemisorption) to the defects or functional groups24,38,39 that happens when ammonia starts to interact with nanotubes. A negligible recovery was observed even after 12 h under continuous flow of dry nitrogen. At higher ammonia concentrations, more molecules are physisorbed at nanotube sidewalls that can be seen from a stronger decay of the sensing response ΔR/R0 suggesting a more efficient removal of weakly attached NH3 molecules during N2 flushing. Comparing sensing performance of p-SWCNT based device with N-SWCNT and B-SWCNT ones (Figure 4) it is apparent that their sensing responses to ammonia are quite similar to slightly higher sensitivity of the pristine nanotubes and slightly lower sensitivity of the doped nanotubes. Introducing the dopant atoms (either nitrogen or boron) into the host structure of nanotube walls does not lead to a noticeable improvement in sensing response of SWCNTs to ammonia. It has to be

I − I (t ) ΔR(t ) × 100 = 0 × 100 R0 I (t )

where ΔR(t) is the difference in resistance before and during NH3 exposure, R0 and I0 are the values of resistance and current before NH3 exposure, and I(t) the current at a fixed voltage which is monitored during gas exposure experiments. In all tested devices, a decrease of current (an increase of resistance) upon ammonia exposure was observed for all ammonia concentrations. This is in agreement with the previous findings that adsorption of NH3 molecules at SWCNTs leads to the hole depletion.20,21 No effect of the interaction of ammonia with metallic leads changing the current in the device is expected here as the influence on the lead resistance can be detected under exposure to much higher ammonia concentrations and for much longer exposure times.35 In our experiments, no additional treatment like heating36 or exposure to UV light37 was conducted to force desorption of the attached molecules and to reach initial resistance values during flushing with N2 gas. This allowed us to distinguish 82

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors

exposure that was experimentally demonstrated recently.40 In p-SWCNT samples approximately 1/3 of nanotubes are metallic. Electric current in these nanotubes is not affected by interaction with NH3 molecules so strongly as in semiconducting nanotubes. Therefore, the response of p-SWCNTbased sensors is lower than that of sc-SWCNT-based ones. On the other hand, it is generally expected that in doped semiconducting nanotubes (2/3 of nanotubes in samples BSWCNTs and N-SWCNTs) the effect at the nanotube/lead interface is reduced by doping compared to undoped semiconducting nanotubes in p-SWCNT sample. This could be considered as the main reason for the reduced response of B-SWCNT and N-SWCNT based sensors to ammonia compared to the p-SWCNT ones. Further increase in sensitivity of SWCNT-based devices can be achieved by avoiding formation of SWCNT bundles during device fabrication. Blue line in Figure 5 depicts sensing response of a sc-SWCNT-based sensor with the resistance of 15 kΩ. In this device SWCNTs are efficiently individualized compared to the device with resistance of 268 Ω, that was achieved by depositing a smaller amount of material (see Figure S2 in SI). Despite a higher resistance, the device with smaller amount of SWCNTs shows higher sensitivity to ammonia gas exposure. In this case, the surface of SWCNTs can be more efficiently used for the interaction with NH3 molecules (higher surface-to-volume ratio) compared with bundled nanotubes. By contrast, a substantial drop of sensitivity was found when the amount of deposited nanotubes is enlarged, and consequently, more nanotube bundles are formed (see Figure S4 and Table S1 in SI). Thus, it can be concluded that in SWCNT-based sensors where individual nanotubes bridge electrodes a higher sensitivity can be achieved compared to thin-film-SWCNT based ones as the formation of SWCNT bundles cannot be avoided in the latter case. Finally, 3.2% sensing response to 1.5 ppm ammonia concentration after 10 min of exposure was achieved as the highest sensitivity of SWCNT-based devices within current study when individualized semiconducting SWCNTs are used as active elements of chemiresistor-type sensors. Sensing experiments for ammonia concentrations ranging from 100 to 1000 ppb were performed for a series of sensing devices based on different types of SWCNTs. In Figure 6 sensing responses of five sc-SWCNT based sensors with

emphasized again that the source material for the doped SWCNTs was the same as in the pristine SWCNT sample, and all sensing devices were tested at the same conditions. All samples have C−O and C−OH groups on sidewalls of nanotubes (see Figure 3). It can be argued that these functional groups played a role of adsorption centers for NH3 molecules taking into account the similarity of the behavior of all sensors. No noticeable improvement of sensing response of doped SWCNTs could be detected here, which was expected from chemisorption of NH3 molecules at pyridinic-N atoms27 in the N-SWCNT sample or at B atoms28 in the B-SWCNT sample. Recently, it was argued36 that pyridinic-N atoms in SWCNTs could improve sensitivity to ammonia by 2.3 times compared to pristine SWCNTs. However, in their case a relatively thick (up to 150 nm) nanotube films between distant electrodes (5 to 100 μm interelectrode distance) were investigated. Therefore, the influence of N doping on the nanotube/nanotube junction resistance in the SWCNT network and its modulation by adsorbed NH3 molecules could be the main contribution to the sensing response.40 Moreover, the presence of a noticeable amount of Fe catalyst particles in samples of the discussed work36 might have an additional influence on the sensor performance. In our case, SWCNTs form single-nanotube thin layer and bridge microelectrodes (see Figure 1b), they have no surfactant molecules on sidewalls or noticeable amount of catalyst particles. Therefore, we conclude that the presence of N atoms in all three possible configurations (see Figure 3c) does not improve sensing response of SWCNTs. Semiconducting Nanotubes. A higher response of the sensing devices to ammonia was achieved with sc-SWCNTs as the active material. Figure 5 shows the behavior of sc-SWCNT-

Figure 5. Sensing response ΔR/R0 of two devices based on semiconducting SWCNTs under exposure to different ammonia concentrations of 1.5, 2.5, 5, 10, and 20 ppm. The devices had initial resistance of 268 Ω (red) and 15 kΩ (high-resistance device, blue line).

based sensors under similar conditions as the sensors discussed above. The red line corresponds to the device with resistance 268 Ω, which is close to 250 Ω for p-SWCNT-based device. However, the response of semiconducting SWCNTs is noticeably higher than in the case of pristine ones for all concentrations (see data in Table S1 in SI). The adsorbed NH3 molecules, which have quite a large dipole moment of 1.42 D, affect electronic structure of semiconducting nanotubes. The highest resistance has the SWCNT/electrode contact region where Schottky barrier is formed in case of semiconducting nanotube. The modulation of the Schottky barrier with NH3 molecules leads to the largest contribution to the change in the current in SWCNT-based sensing device under analyte gas

Figure 6. Sensing response ΔR/R0 of five devices based on scSWCNTs under exposure to different ammonia concentrations of 100, 400, 700, and 1000 ppb. The initial resistance values R0 for the devices are shown in legends. Colored dashed lines connecting data point are drawn to guide an eye; error bars depict signal noise at the data point. Horizontal dashed line is drawn at three times the noise averaged over all data points. 83

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors

sensor was 0.6 μW, i.e., approximately 5 orders of magnitude lower. Future Trends. When comparing obtained results to existing literature, it is worth noticing that our sensors based on p-SWCNTs performed similarly to those recently presented by Rigoni et al., where surfactants were utilized to disperse nanotubes in water.22 These sensors showed 1.6% to 1.9% response to 1.5 ppm of ammonia, which is comparable to our 0.7% response of p-SWCNT based device. The discrepancy may be attributed to the effect of surfactants on the sidewalls of nanotubes which could facilitate adsorption of NH3 molecules. Finally, the sensors were shown to be able to detect ammonia in concentrations down to 20 ppb at room temperature and in a controlled humidity. It can be expected that our devices based on sc-SWCNTs, which have higher sensitivity, may be used for the detection of ammonia in the sub-ppm range. A further increase in sensitivity can be achieved not by doping with N or B atoms as was proposed in the past, but rather with oxygen plasma treatment as has been demonstrated recently.44 This way an ultrasensitive device able to reliably detect ammonia in the whole ppb range of concentrations can be fabricated. This makes the SWCNT based ammonia sensors attractive as components of a single-chip sensor array for the analysis of gas mixtures45 with a little fraction of ammonia. Such arrays can be potentially used for noninvasive and fast medical diagnostics such as exhaled gas mapping that can be employed for the detection of particular disease fingerprints (liver cirrhosis, kidney failure,10 or diseases caused by Helicobacter pylori11). The sensor fabrication technology used in current work is rather simple and can be easily scaled up. The main component is semiconducting SWCNTs which should be long enough to bridge electrodes. The dispersion of sc-SWCNTs can be deposited onto the interdigitated electrode area in a controlled manner using, e.g., a nanoplotter. The amount of nanotubes per sensing device is extremely small. The sensor with the highest sensitivity had the lowest power consumption, 0.6 μW, that suggests it for mobile applications or a remote environment monitoring. However, for the full recovery of the sensor after each sensing cycle, a local heater or a UV LED has to be used for a short time to force NH3 desorption. Then, power consumption of such an additional component should be taken into account. The main goal of this work was to systematically study the influence of SWCNT modification like doping with nitrogen and boron or selecting semiconducting-only nanotubes on the sensitivity of SWCNT-based sensors to ammonia. As the main route toward fabricating energy-efficient highly sensitive ammonia sensors with SWCNTs as sensing element is now determined, presenting calibration curves, evaluating the limit of detection, and determining an optimal sensor recovery protocol are the subjects of the future work.

different initial device resistances are presented. As can be seen all devices have qualitatively similar response properties. The horizontal dashed line depicts the value of three times average noise. Sensing responses at 100 ppb of four out of five sensors lie above this dashed line, and it can be concluded that the limit of detection of sc-SWCNT based sensors is 100 ppb. Remarkably, in the sub-ppm range of concentrations an improved sensitivity of high-resistance sensing device is not so pronounced as in the ppm rage of concentrations. Sensing response of p-SWCNT, B-doped, and N-doped SWCNT based devices in the sub-ppm range of concentrations is roughly an order of magnitude lower than that of sc-SWCNT (see Figure S3 of Supporting Information) that is similar to ppm range (cf. Figure 4 and Figure 5). In our experiments ammonia gas was diluted in pure N2; thus, the gas mixture consisted just of two gases. In real samples (air) further gases are present which are mainly O2, H2O, Ar, CO2, and CO. As our sensing devices were prepared and stored at ambient conditions and, therefore, nanotubes were already exposed to these gases for a relatively long time, the presence of these gases is not expected to affect ammonia sensing. Ar and CO2 are neutral gases with zero dipole moment; CO also has no influence on pristine nanotubes.41 It has been recently demonstrated that sensing response of SWCNT-based ammonia sensors is slightly increased with increasing relative humidity of the gas mixture.42 Therefore, humidity favors sensing response to ammonia; however, for applications where relative humidity is noticeably varying a SWCNT-based ammonia sensor should be calibrated and supplemented with a standard humidity sensor. As mentioned above, a minor oxygen content of 2.1 at. % is present even in pristine nanotubes (see Figure 3a) that leads to p-type conductivity in semiconducting nanotubes. Exposure to an increased concentration of oxygen can only lead to a slight p-doping which would be manifested by a little decrease of device resistivity because of the reduction of Schottky barrier. In order to demonstrate how our SWCNT-based sensors respond to the presence of p-doping gases, we exposed our sensing devices to NO gas at low concentrations (also diluted in pure N2). In this case, device resistivity decreases and the sensing response shows negative values (see Figure S5 in SI). Therefore, other gases present in real samples have either no effect on nanotubes or may lead to a little additional p-doping of nanotubes. This, however, does not reduce the sensitivity of our devices to ammonia, which acts as an n-type dopant being adsorbed at nanotube sidewalls. Still, a systematic cross-sensitivity study and calibration of SWCNT-based sensing devices for ammoniacontaining gas mixtures typical for target applications should be performed that is out of the scope of current work. Commercially available sensors used in industry for monitoring ammonia leakage, ventilation control, or in agriculture (e.g., sensors from Figaro Engineering Inc.43) are based on metal oxide sensing elements which operate at elevated temperatures and consume typically from 50 mW to 1 W. Reducing the size of sensing elements by employing metaloxide nanowires does not lead to a noticeable reduction of sensor power consumption since these sensors should also work at elevated optimal temperatures (ranging from 200 to 600 °C depending on material). For example, power consumption of ammonia sensor based on SnO2 nanowires working at 530 °C was 46 mW.17 In contrast, all ammonia sensing experiments in our work were performed at room temperature, and power consumption of the sc-SWCNT-based



CONCLUSIONS In this work, gas-sensing devices based on semiconducting SWCNTs, pristine SWCNTs, B-doped SWCNTs, and Ndoped SWCNTs for ammonia detection purposes were successfully fabricated using a drop casting approach and characterized under NH3 exposure. All sensors tested in this work exhibited a measurable response to 1.5, 2.5, 5, and 20 ppm of ammonia. The highest response was achieved by sensors based on sc-SWCNTs: 3.2% to 1.5 ppm for an exposure time of 10 min and almost 5% to 10 ppm for 5 min exposure. The sensors based on sc-SWCNTs are shown to 84

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

ACS Sensors



reliably detect NH3 in the sub-ppm range of concentrations down to 100 ppb after few minutes of exposure. Such sensitivity is relevant for diagnosis of certain diseases such as live cirrhosis, kidney failure, and diseases caused by Helicobacter pylori. The sensor consumes 0.6 μW, is compact, and works at room temperature that suggests its application in networks of numerous sensors (also mobile) for permanent remote monitoring of environment. Devices based on p-SWCNTs, B-SWCNTs, and N-SWCNTs performed qualitatively similarly, however, showing lower sensing responses to all ammonia concentrations. Thus, no significant improvement in response to NH3 was achieved by using doped SWCNTs. It is concluded that NH3 molecules interact with carboxyl groups on the SWCNT sidewalls rather than with dopant atoms, and therefore, the sensitivity of SWCNT-based ammonia sensors can be increased rather by additional oxidation of nanotubes than by doping with nitrogen or boron atoms. Based on SEM and electrical characterizations, it was found that the deposition of nanomaterial is a crucial step in the process of fabricating devices highly sensitive to low ammonia concentrations. It was observed that increasing the amount of SWCNTs between the IDE leads to a decrease in ammonia sensitivity primarily due to formation of SWCNT bundles and reducing this way the surface-to-volume ratio. Finally, a route toward an optimal fabrication of SWCNT-based sensors for the efficient subppm of NH3 detection was proposed.



REFERENCES

(1) Manzoli, A.; Steffens, C.; Paschoalin, R. T.; Correa, A. a.; Alves, W. F.; Leite, F. L.; Herrmann, P. S. P. Low-Cost Gas Sensors Produced by the Graphite Line-Patterning Technique Applied to Monitoring Banana Ripeness. Sensors 2011, 11, 6425−6434. (2) Chaisitsak, S. Nanocrystalline SnO2:F Thin Films for Liquid Petroleum Gas Sensors. Sensors 2011, 11, 7127−7140. (3) Cordos, E.; Ferenczi, L.; Cadar, S.; Costiug, S.; Pitl, G.; Aciu, A.; Ghita, A. Methane and Carbon Monoxide Gas Detection System Based on Semiconductor Sensor. 2006 IEEE Int. Conf. Autom. Qual. Testing, Robot. 2006, 2, 208−211. (4) Sekhar, P. K.; Kysar, J.; Brosha, E. L.; Kreller, C. R. Development and Testing of an Electrochemical Methane Sensor. Sens. Actuators, B 2016, 228, 162−167. (5) Caucheteur, C.; Debliquy, M.; Lahem, D.; Mégret, P. Catalytic Fiber Bragg Grating Sensor for Hydrogen Leak Detection in Air. IEEE Photonics Technol. Lett. 2008, 20, 96−98. (6) Xiao, G.; Zhang, Z.; Weber, J.; Ding, H.; McIntosh, H.; Desrosiers, D.; Nong, G.; Won, D.; Dunford, J.; Tunney, J.; et al. Trace Amount Formaldehyde Gas Detection for Indoor Air Quality Monitoring. IEEE Int. Instrum. Meas. Technol. Conf. 2011, 1−4. (7) Kwon, J.; Ahn, G.; Kim, G.; Kim, J. C.; Kim, H. A Study on NDIR-Based CO2 Sensor to Apply Remote Air Quality Monitoring System. ICROS-SICE 2009, 8, 1683−1687. (8) Valera, J. L.; Togores, B.; Cosio, B. G. Use of the Electronic Nose for Diagnosing Respiratory Diseases. Arch. Bronconeumol. 2012, 48, 187−188. (9) Haick, H.; Broza, Y. Y.; Mochalski, P.; Ruzsanyi, V.; Amann, A. Assessment, Origin, and Implementation of Breath Volatile Cancer Markers. Chem. Soc. Rev. 2014, 43, 1423−1449. (10) Narasimhan, L. R.; Goodman, W.; Patel, C. K. Correlation of Breath Ammonia with Blood Urea Nitrogen and Creatinine during Hemodialysis. Proc. Natl. Acad. Sci. U. S. A. 2001, 98, 4617−4621. (11) Kearney, D. J.; Hubbard, T.; Putnam, D. Breath Ammonia Measurement in Helicobacter Pylori Infection. Dig. Dis. Sci. 2002, 47, 2523−2530. (12) Spector, O.; Jacobson, E. Advanced Ammonia (NH3) Monitoring System for Industrial Applications. Proc. SPIE 1999, 3853, 233−238. (13) Cui, Y.; Wei, Q.; Park, H.; Lieber, C. Nanowire Nanosensors for Highly Sensitive and Selective Detection of Biological and Chemical Species. Science 2001, 293, 1289−1292. (14) Zörgiebel, F. M.; Pregl, S.; Römhildt, L.; Opitz, J.; Weber, W.; Mikolajick, T.; Baraban, L.; Cuniberti, G. Schottky Barrier-Based Silicon Nanowire pH Sensor with Live Sensitivity Control. Nano Res. 2014, 7, 263−271. (15) Karnaushenko, D.; Ibarlucea, B.; Lee, S.; Lin, G.; Baraban, L.; Pregl, S.; Melzer, M.; Makarov, D.; Weber, W. M.; Mikolajick, T.; et al. Light Weight and Flexible High-Performance Diagnostic Platform. Adv. Healthcare Mater. 2015, 4, 1517−1525. (16) Schütt, J.; Ibarlucea, B.; Illing, R.; Zörgiebel, F.; Pregl, S.; Nozaki, D.; Weber, W. M.; Mikolajick, T.; Baraban, L.; Cuniberti, G. Compact Nanowire Sensors Probe Microdroplets. Nano Lett. 2016, 16, 4991−5000. (17) Sama, J.; Barth, S.; Jimenez-Diaz, R.; Prades, J. D.; Casals, O.; Gracia, I.; Cane, C.; Romano-Rodriguez, A. Locally Grown SnO2 NWs as Low Power Ammonia Sensor. Procedia Eng. 2015, 120, 215−219. (18) Jiménez-Cadena, G.; Riu, J.; Rius, F. X. Gas Sensors Based on Nanostructured Materials. Analyst 2007, 132, 1083−1099. (19) Kim, H.-J.; Lee, J.-H. Highly Sensitive and Selective Gas Sensors Using P-Type Oxide Semiconductors: Overview. Sens. Actuators, B 2014, 192, 607−627. (20) Llobet, E. Gas Sensors Using Carbon Nanomaterials: A Review. Sens. Actuators, B 2013, 179, 32−45. (21) Kong, J. Nanotube Molecular Wires as Chemical Sensors. Science 2000, 287, 622−625. (22) Rigoni, F.; Tognolini, S.; Borghetti, P.; Drera, G.; Pagliara, S.; Goldoni, A.; Sangaletti, L. Enhancing the Sensitivity of Chemiresistor Gas Sensors Based on Pristine Carbon Nanotubes to Detect Low-ppb

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acssensors.7b00358. Nanomaterial synthesis, nanomaterial dispersions, sensor fabrication, gas detection setup, device characterization, sensing response to ammonia exposure, and sensing response to nitrogen monoxide exposure (PDF)



Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Larysa Baraban: 0000-0003-1010-2791 Viktor Bezugly: 0000-0001-9301-4302 Gianaurelio Cuniberti: 0000-0002-6574-7848 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS L.A.P.-R. thanks Dr. Cindy Schmädicke for the introduction to the gas exposure setup. Authors thank Jochen Förster, Gero Wiemann, and Uwe Gutsche for the technical assistance. This work is partly supported by the European Union (ERDF) via the FP7 projects “CARbon nanoTube phOtONic devices on silicon” (CARTOON) and “Nano-carbons for versatile power supply modules” (NanoCaTe). We also acknowledge the support by the German Research Foundation (DFG) within the Cluster of Excellence “Center for Advancing Electronics Dresden”, by the German Academic Exchange Service (DAAD) and China Scholarship Council (CSC). 85

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86

Article

ACS Sensors Ammonia Concentrations in the Environment. Analyst 2013, 138, 7392−7399. (23) Peng, S.; Cho, K. Chemical Control of Nanotube Electronics. Nanotechnology 2000, 11, 57−60. (24) Feng, X.; Irle, S.; Witek, H.; Morokuma, K.; Vidic, R.; Borguet, E. Sensitivity of Ammonia Interaction with Single-Walled Carbon Nanotube Bundles to the Presence of Defect Sites and Functionalities. J. Am. Chem. Soc. 2005, 127, 10533−10538. (25) Andzelm, J.; Govind, N.; Maiti, A. Nanotube-Based Gas Sensors - Role of Structural Defects. Chem. Phys. Lett. 2006, 421, 58−62. (26) Rocha, A. R.; Rossi, M.; Fazzio, A.; Da Silva, A. J. R. Designing Real Nanotube-Based Gas Sensors. Phys. Rev. Lett. 2008, 100, 3−6. (27) Villalpando-Páez, F.; Romero, A. H.; Muñoz-Sandoval, E.; Martínez, L. M.; Terrones, H.; Terrones, M. Fabrication of Vapor and Gas Sensors Using Films of Aligned CNx Nanotubes. Chem. Phys. Lett. 2004, 386, 137−143. (28) Bai, L.; Zhou, Z. Computational Study of B- or N-Doped SingleWalled Carbon Nanotubes as NH3 and NO2 Sensors. Carbon 2007, 45, 2105−2110. (29) Borowiak-Palen, E.; Pichler, T.; Graff, a.; Kalenczuk, R. J.; Knupfer, M.; Fink, J. Synthesis and Electronic Properties of B-Doped Single Wall Carbon Nanotubes. Carbon 2004, 42, 1123−1126. (30) Zhang, H.; Kuila, T.; Kim, N. H.; Yu, D. S.; Lee, J. H. Simultaneous Reduction, Exfoliation, and Nitrogen Doping of Graphene Oxide via a Hydrothermal Reaction for Energy Storage Electrode Materials. Carbon 2014, 69, 66−78. (31) Baraban, L.; Cuniberti, G.; Schmädicke, C.; Voigt, J. Messsystem Zum Testen von Mindestens Zwei Gassensoren. German Patent No DE 102014218205 b3, 2014. (32) Krstić, V.; Rikken, G. L. J. a.; Bernier, P.; Roth, S.; Glerup, M. Nitrogen Doping of Metallic Single-Walled Carbon Nanotubes:n-Type Conduction and Dipole Scattering. Europhys. Lett. 2007, 77, 37001. (33) Collins, P. G.; Bradley, K.; Ishigami, M.; Zettl, A. Extreme Oxygen Sensitivity of Electronic Properties Extreme Oxygen Sensitivity of Electronic Properties of Carbon Nanotubes. Science 2000, 287, 1801−1805. (34) Battie, Y.; Gorintin, L.; Ducloux, O.; Thobois, P.; Bondavalli, P.; Feugnet, G.; Loiseau, A. Thickness Dependent Sensing Mechanism in Sorted Semi-Conducting Single Walled Nanotube Based Sensors. Analyst 2012, 137, 2151. (35) Dube, I.; Jiménez, D.; Fedorov, G.; Boyd, A.; Gayduchenko, I.; Paranjape, M.; Barbara, P. Understanding the Electrical Response and Sensing Mechanism of Carbon-Nanotube-Based Gas Sensors. Carbon 2015, 87, 330−337. (36) Battie, Y.; Ducloux, O.; Thobois, P.; Susi, T.; Kauppinen, E. I.; Loiseau, A. Selective Differential Ammonia Gas Sensor Based on NDoped SWCNT Films. Phys. Status Solidi B 2011, 248, 2462−2466. (37) Chen, G.; Paronyan, T. M.; Pigos, E. M.; Harutyunyan, A. R. Enhanced Gas Sensing in Pristine Carbon Nanotubes under Continuous Ultraviolet Light Illumination. Sci. Rep. 2012, 2, 343. (38) Robinson, J. A.; Snow, E. S.; Bǎdescu, Ş. C.; Reinecke, T. L.; Perkins, F. K. Role of Defects in Single-Walled Carbon Nanotube Chemical Sensors. Nano Lett. 2006, 6, 1747−1751. (39) Khavrus, V. O.; Vinzelberg, H.; Schumann, J.; Leonhardt, A.; Oswald, S.; Buechner, B. On the Potential of Long Carbon Nanotube Forest for Sensing Gases and Vapors. Phys. E 2011, 43, 1199−1207. (40) Boyd, A.; Dube, I.; Fedorov, G.; Paranjape, M.; Barbara, P. Gas Sensing Mechanism of Carbon Nanotubes: From Single Tubes to High-Density Networks. Carbon 2014, 69, 417−423. (41) Hannon, A.; Lu, Y.; Li, J.; Meyyappan, M. Room Temperature Carbon Nanotube Based Sensor for Carbon Monoxide Detection. J. Sens. Sens. Syst. 2014, 3, 349−354. (42) Rigoni, F.; Freddi, S.; Pagliara, S.; Drera, G.; Sangaletti, L.; Suisse, J.-M.; Bouvet, M.; Malovichko, A. M.; Emelianov, A. V.; Bobrinetskiy, I. I. Humidity-Enhanced Sub-Ppm Sensitivity to Ammonia of Covalently Functionalized Single-Wall Carbon Nanotube Bundle Layers. Nanotechnology 2017, 28, 255502. (43) Figaro Engineering Inc.; http://www.figaro.co.jp/en/.

(44) Ham, S. W.; Hong, H. P.; Kim, J. W.; Kim, J. H.; Kim, K. B.; Park, C. W.; Min, N. K. Comparison of Gas Sensors Based on Oxygen Plasma-Treated Carbon Nanotube Network Films with Different Semiconducting Contents. J. Electron. Mater. 2015, 44, 1344−1350. (45) Lu, Y. J.; Meyyappan, M.; Li, J. Fabrication of CarbonNanotube-Based Sensor Array and Interference Study. J. Mater. Res. 2011, 26, 2017−2023.

86

DOI: 10.1021/acssensors.7b00358 ACS Sens. 2018, 3, 79−86