Abdullah Abdulhameed*ab,
Izhal Abdul Halinb,
Mohd Nazim Mohtar
bc and
Mohd Nizar Hamidonbc
aDepartment of Electronic Engineering, Faculty of Engineering, Hadhramout University, Mukalla, Yemen. E-mail: gs50667@student.upm.edu.my
bDepartment of Electrical and Electronic Engineering, Faculty of Engineering, Universiti Putra Malaysia, Serdang, 43400, Malaysia
cInstitute of Advanced Technology (ITMA), Universiti Putra Malaysia, Serdang, 43400, Malaysia
First published on 10th December 2021
The dielectrophoresis (DEP) method is used to fabricate sensor devices by assembling and aligning carbon nanotubes (CNTs) across electrode structures. The challenges of the method increase as the gap width between the electrodes increases. In this work, a novel DEP setup is proposed to reduce the resistance mismatch in manufacturing carbon nanotube-based sensors. The proposed setup utilizes hot airflow and thermal annealing to fabricate long-aligned multi-walled carbon nanotube (MWCNT) bridges across transparent electrodes with a gap width up to 75 μm. The best alignment results were obtained at airflow velocities between 1.5 m s−1 and 2.5 m s−1. The minimum variation in the resistance of the aligned bridges was 1.81% observed at a MWCNT concentration of 0.005 wt% and deposition time of 10 min. Long MWCNT bridges have many contact points that link MWCNTs to each other, making the contact resistance a robust indicator of the variation in the ambient temperature. The characteristics of the MWCNT bridges as a temperature sensor, including the response, sensitivity, and recovery, were investigated.
The remarkable optical, mechanical, and electrical properties of CNTs allowed them to be used in various applications.7,8 For example, in gas sensor applications, CNTs can detect gas molecules at low concentrations, such as the detection of ammonia and hydrogen.9,10 In biosensor applications, CNTs have the ability to sense heated bacteria and COVID19 virus.11,12 CNTs are also used to sense environmental data such as temperature and humidity.13 However, ambient temperature is the most frequently measured variable in daily life, monitor systems, and industrial applications.14
The high stability and low power consumption made CNTs suitable to work as temperature sensors.15 Moreover, CNTs can operate in a wide temperature range without being physically affected. Several sensing mechanisms were reported to theoretically explain the change in the CNT resistance following the change in the ambient temperature.16 Most of the studies reported negative temperature coefficients (TCs), where the sensor resistance decreases as the temperature increases.17,18 On the other hand, positive TCs were also reported, where the sensor resistance increases with the temperature increase.19 Kuo et al. related the change in the resistance to the nature of the CNTs.19 They concluded that the resistance is directly proportional to the temperature in the case of CNTs with metallic nature and inversely in the case of CNTs with semiconducting nature. The same study showed that the gap width between the sensor electrodes significantly affects the sensor response.
The sensing material in temperature sensors is either solo CNTs or a composite of CNTs and other materials. In the first case, limited studies addressed sensors that use only CNTs (either single-walled carbon nanotubes (SWCNTs) or multi-walled carbon nanotubes (MWCNTs)) to detect temperature variation.20,21 In the second case, the sensing material was manufactured by mixing CNTs with other materials such as epoxy resin,22 poly(vinylidene fluoride) (PVDF),23 graphene oxide,24 silicon,25 and PEDOT:PSS.26 In both cases, CNTs are preferred to be aligned in order to achieve their best performance.
The inability to produce aligned CNTs across the sensor electrodes is the major limitation in manufacturing CNT-based sensors. Using random CNT networks produces devices with anisotropic electrical properties with considerable variation in their resistance.27 Fabrication of temperature sensors using methods such as screen printing,28 ink printing,29 additive printing,30 doctor blade,31 spinning process,32 chemical vapor deposition (CVD),33 and vacuum filtration34 consume large amounts of CNTs, making them costly. Moreover, the repeatability of the manufacturing process produces sensors with different resistance due to the low alignment quality, which causes a mismatch in the sensor performance.35 Consequently, this would lead to necessitating individual calibration of each sensor.36
Dielectrophoresis (DEP) is one of the sensors' fabrication methods that is used to overcome the isotropic properties of CNTs by assembling them onto the device in an aligned form.37 DEP does not require high specific setups compared with the mentioned methods. However, DEP is restricted to the device geometry where there are difficulties in maintaining the alignment quality across gaps with large-scale dimensions due to the poor controllability of the method.38 Improving the DEP setup while maintaining its simplicity is expected to preserve the alignment quality and thus reduce the resistance variation that might occur due to the distortion in the aligned CNTs.
In this article, a novel DEP setup is proposed to fabricate CNT-based temperature sensors. The setup utilizes airflow with a specific velocity and temperature to assist the alignment of long MWCNT bridges across transparent electrodes. The shear force generated by the airflow is expected to produce a torque on the tubes to maintain their alignment quality. In addition to the alignment, airflow simultaneously spreads the solution droplet and helps in the drying process. The new setup is expected to minimize the mismatch in the resistance of the fabricated temperature sensors. The article also investigates the characteristics and performance of the temperature sensors in terms of sensitivity, response, and recovery.
Sensor code | Sensing layers (samples) | Functional group |
---|---|---|
SC1 | MWCNTs | OH |
SC2 | MWCNTs | COOH |
SC3 | MWCNTs | OH + COOH |
SC4 | MWCNTs + CAC | OH + COOH |
The reproducibility of the AA-DEP method was studied by measuring the variation in the resistance of a batch of sensors (five sensors in each case) fabricated at identical conditions. The variation in the sensor resistance (mean value and standard deviation) was investigated at different fabrication conditions, such as different MWCNT concentrations, deposition times, and sensing materials. Fig. 2b shows the variation in the sensors' resistance as a function of MWCNT concentration in the suspension. The variations in the resistance of the sensors were 4.89%, 6.06%, and 1.81% at MWCNT concentrations of 0.002, 0.003, and 0.005 wt%, respectively. The minimum resistance mismatch was observed at the highest concentration.
Fig. 2c shows the variation in the sensors' resistance as a function of deposition time. The variations in the resistance were 12.39%, 8.05%, and 4.89% at a deposition time of 3, 5, and 10 minutes. The minimum resistance mismatch was observed at the longest deposition time. The AA-DEP's ability to reduce the resistance mismatch was also tested for sensors fabricated using MWCNTs with different functional groups. Fig. 2d shows that the variations in the resistance were 4.89%, 2.914%, and 5.284% for MWCNTs functionalized with COOH + OH, COOH, and OH, respectively. In conclusion, the AA-DEP setup successfully reduced the resistance mismatch by avoiding the distortion and deformation that might occur during the drying process. The minimum resistance variation was observed at a concentration of 0.005%, and the worst resistance mismatch was observed at a deposition time of 3 minutes (resistance variation in basic alignment setups could exceed 10%).42
Fig. 3a shows an example of the fabricated transparent sensor using the AA-DEP setup. The thin dark layer across the ITO electrodes is the deposited MWCNTs. The darkness (density/thickness) of the layer and its resistance were controlled by controlling the fabrication parameters, especially the AC signal amplitude and duration (deposition time). Further details regarding how to control the density of the aligned MWCNTs can be found elsewhere.43 Fig. 3b shows that the density of the deposited MWCNTs is directly proportional to the concentration of the MWCNTs in the medium. Thus, mediums with high concentration MWCNTs can be used to deposit a high-density MWCNT layer. At constant concentration, the density can also be controlled by the deposition time, as shown in Fig. 3c. The devices shown in Fig. 3d have the same density even though different types of functional groups were attached to the MWCNTs, which indicates that the functional groups on the MWCNT walls have less significant effects on the density compared with the deposition time and concentration. Although the density of the MWCNTs looks similar in Fig. 3d, the three devices have different resistance (refer to Fig. 2d).
![]() | ||
Fig. 3 Deposited MWCNTs across ITO electrodes. (a) The fabricated sensor. (b) Deposition of MWCNTs at different concentrations (time: 10 min, sample: SC3). (c) Deposition of MWCNTs at different deposition times (concentration: 0.002 wt%, sample: SC3). (d) Deposition of MWCNTs functionalized with different functional groups (time: 10 min, concentration: 0.002 wt%) [scale bar: 50 μm, Optical microscope: Moticam, BA310E]. Large-scale and high-resolution images can be found in the ESI (Fig. S8 and S9†). |
![]() | (1) |
The TCR equation is equivalent to the slope of the fitting line of the experimental data over the initial sensor resistance. Thus, the sensor with a higher absolute TCR value indicates strong sensitivity. The experimental result showed that the sensitivity of the aligned MWCNT bridges depends on several factors, such as the density of the aligned bridges and the type of the functional group attached to the MWCNT walls.
Fig. 4a shows the response of MWCNT-based temperature sensors that differ from each other based on the functional group attached to their walls. The sensors that were fabricated using sample SC2 showed the strongest sensitivity with a TCR of −0.228% °C−1, while the worst sensitivity was observed in sensors fabricated using sample SC3 with TCR of −0.07% °C−1. The negative signal means the sensor resistance decreases as the temperature increases. Four reasons explain the difference in the sensitivity of the sensors: sample electrical conductivity, sample solubility in DMF, sample treatments, and the contact resistance (MWCNT–MWCNT/MWCNT–ITO). In terms of electrical conductivity, MWCNTs have different electrical conductivities based on the functional groups attached to their walls.45 For example, the sensor fabricated using sample SC1 has a TCR value higher than the sensor fabricated using sample SC3 (|−0.115| > |0.07|% °C−1) because the electrical conductivity of sample SC1 was higher than sample SC3. In terms of solubility, MWCNTs functionalized with COOH showed better solubility in DMF, which results in better alignment thus strong sensitivity. MWCNTs that were treated or functionalized with more than one functional group have low quality due to the harsh treatment, which explains why sensor SC3 shows the worst sensitivity. Finally, the MWCNT–MWCNT and MWCNT–ITO contact resistance plays a critical role in the sensors' behavior. However, the contact resistance also depends on the quality of the MWCNTs and their dispersity in mediums.46
Fig. 4b shows the response of three sensors fabricated under the same conditions using sample SC2 with different MWCNT concentrations. The range of the operating resistance of the sensors increased with the increase of the MWCNT concentration in the suspension. AA-DEP assembled more MWCNTs from the suspension with high MWCNT concentration, resulting in high-density MWCNT bridges with low resistance. Although the slope of the linear fitting lines was convergent, the TCR of the sensors was different at different concentrations (−0.12725% °C−1 at 0.003 wt% and −0.19572% °C−1 at 0.004 wt%).
Fig. 4c shows the performance of sensors fabricated using a mixture of MWCNTs and CAC (SC4). The slope of the fitting lines was steeper than the fitting lines of the sensors that were fabricated using MWCNTs only (Fig. 4b). However, the TCR value was −0.20984% °C−1 at a concentration of 0.0015 wt% and −0.16504% °C−1 at a concentration of 0.0010 wt%. Although the concentration of SC4 was much lower than the concentration of SC1–SC3, the TCR value was higher, which indicates that the spherical conductive carbon particles played a role in reducing the contact resistance between the tubes.
The sensors' behavior after annealing was entirely different in contrast to the behavior before annealing. The sensor resistance increased with the increase of the temperature. Fig. 5 shows the response of the sensor after exposing it to three cycles of hot airflow. The airflow temperature at the back of the glass substrate was ∼60 °C. In sensor SC2, the resistance dropped from 543.8 Ω (the initial resistance after the alignment and before annealing) to 529.5 Ω in the first 200 seconds. However, the sensor resistance recovered to only 530.4 Ω after 200 seconds from switching off the hot air gun. The sensor minimum and maximum resistance continued decreasing in the following cycles (Fig. 5a). The behavior of the sensor was stable after the annealing process, where the variation in the minimum and maximum resistance was very small. The sensor resistance increased from 443.6 Ω to 447.41 Ω. The response cycles of SC3 after annealing were similar to those of SC2, with a significant decrease in the minimum and maximum resistance values in each cycle (Fig. 5b). However, the minimum and maximum resistance values in the first cycle were 528.8 Ω and 531.5 Ω, respectively.
Fig. 5c presents the sensitivity after annealing the temperature sensors. The sensor that was fabricated using COOH–MWCNT (SC2) showed better sensitivity than the other sensors, while the sensor that was fabricated with COOH–OH–MWCNT (SC3) showed the worst sensitivity. The results were consistent with the TCR obtained before the thermal annealing. The inverse relationship (positive TCR) may be due to the changes in the MWCNT metallic/semiconducting ratio within the sensing material after the annealing process.19 In addition, the annealing can reduce the oxygen content of the functionalized MWCNTs, and it is known that raw MWCNTs generally show a positive TCR, while functionalized MWCNTs show a negative TCR. The remains of the solvent inside and between the tubes before the thermal annealing might affect the temperature measurements. Fig. 5d shows that the sensitivity of the temperature sensors was improved by assembling CAC particles along with the MWCNTs. The conductive carbon particles work as junctions that link the tubes to each other and form full bridges across the gap. The sensors also showed strong repeatability, where the response cycles of the two sensors were almost identical.
The response and recovery times of the temperature sensors were also investigated. Fig. 6a presents the sensing cycle of sensor SC4 exposed to hot airflow with a temperature of ∼60 °C. The sensor resistance reached 90% of its maximum resistance in 60 seconds. However, the sensor required more than 3 minutes to recover to 7.7% of its initial resistance.
![]() | ||
Fig. 6 Characterization of sensor SC4 (fabrication parameters: 20 V, 10 min, 0.005 wt%). (a) Response and recovery time. (b) The response cycle of the sensor with reduced sensing areas. |
The resistance mismatch was investigated at different sensing areas of sensor SC4 to confirm that the AA-DEP method assembled and aligned the MWCNTs across the ITO electrodes uniformly. The total gap area of the fabricated IDE electrodes was 6 mm × 50 μm = 300000 μm2. The initial resistance of the MWCNT bridges that coated this area was 187.09 Ω. The area was then reduced to 285
000 μm2 by wiping 5% of the aligned bridges. The new initial resistance of the sensor increased to 199.25 Ω. However, the sensor's sensitivity was identical and did not change when the sensing area was reduced. The sensing area was further reduced by 5% to result in a sensing area of 270
000 μm2. The new initial resistance was 209.56 Ω, and the sensor response was also similar to the response of the total area (Fig. 6b). We conclude that the AA-DEP method showed good reproducibility where the sensor's performance was identical although after reducing the sensing area.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1ra08250g |
This journal is © The Royal Society of Chemistry 2021 |