Single Wall Carbon Nanotubes Based Cryogenic Temperature Sensor Platforms

We present an investigation consisting of single walled carbon nanotubes (SWCNTs) based cryogenic temperature sensors, capable of measuring temperatures in the range of 2–77 K. Carbon nanotubes (CNTs) due to their extremely small size, superior thermal and electrical properties have suggested that it is possible to create devices that will meet necessary requirements for miniaturization and better performance, by comparison to temperature sensors currently available on the market. Starting from SWCNTs, as starting material, a resistive structure was designed. Employing dropcast method, the carbon nanotubes were deposited over pairs of gold electrodes and in between the structure electrodes from a solution. The procedure was followed by an alignment process between the electrodes using a dielectrophoretic method. Two sensor structures were tested in cryogenic field down to 2 K, and the resistance was measured using a standard four-point method. The measurement results suggest that, at temperatures below 20 K, the temperature coefficient of resistance average for sensor 1 is 1.473%/K and for sensor 2 is 0.365%/K. From the experimental data, it can be concluded that the dependence of electrical resistance versus temperature can be approximated by an exponential equation and, correspondingly, a set of coefficients are calculated. It is further concluded that the proposed approach described here offers several advantages, which can be employed in the fabrication of a microsensors for cryogenic applications.


Introduction
Temperature is one of the most important parameters to measure, especially when we refer to the cryogenic field. Cryogenics is the science of attaining very low temperatures and observing its effect on materials, being widely used in various domains such as aerospace, nuclear, medical, mechanical, chemical, and electrical technologies, or for studying nature. The units used for temperature measurements are Kelvin (K) or Celsius ( • C). The relation between them is defined by the Internationally Temperature Scale of 1990 (ITS-90) as: K = • C + 273. 16. ITS-90 defines the thermodynamic temperature scale from 0.65 K to the highest temperature measurable in terms of the Planck radiation law using monochromatic radiation [1]. It specifies several temperature

Materials and Methods-Sensors Preparation
In a similar manner as described in [14], commercially produced CNTs (single-walled, purchased from Sigma Aldrich, Munich, Germany, 60% purity) were mixed together with isopropyl alcohol (proportion 1/10). The resulted solution was rigorously mixed using a sonication bath with controlled inside temperature of 30 • C to avoid alcohol evaporation effect. After 60 min, the solution was considered as being completely homogenized and then was filtered and dried.
Before deposition of the platinum nanoclusters on the above-mentioned SWCNTs, the nanotubes suffered a functionalization and drying process. For the deposition process, a solution of H 2 PtCl 6 , 1% concentration, was obtained by dissolving chloroplatinic acid hexahydrate in distilled water. One milliliter of H 2 PtCl 6 solution was extracted and introduced over SWCNTs together with high purity Hydrogen (5.0). A special bubbling device was used to limit the gas flow and the dimension of bubbles, as shown schematically in Figure 1. For the following experiments, two sensor structures were obtained by employing a dropcast technique of a small quantity of solution (20 µ L) between the electrodes. The first structure (here called Sensor 1) consists of two copper electrodes on a ceramic substrate with a distance of 1.25 mm between electrodes. The second structure (here called Sensor 2) consists of two gold electrodes on a Sitall substrate (crystalline glass-ceramic with ultra-low coefficient of thermal expansion) with a distance of 0.25 mm between electrodes. To align the SWCNTs and to form multilayers of nanotubes between the two electrodes [19,20], the structures that we prepared were subjected to a dielectrophoresis process by applying 1 VDC potential, for 1 min, over the electrodes. During this time, dry air at 60 °C was blown over the structures. Further, the sensors were subjected to a high temperature treatment (200 °C for 60 min) for achieving structure stability of the sensitive active layer.

Characterization Methods
The two sensor structures prepared in this investigation were tested at very low temperature in a refrigeration unit Quantum Design Physical Property Measurement System (PPMS) with EverCool-II ® Cryogen-Free Cooling Technology, model P935A (referred as PPMS, Quantum Design, Inc., GmbH, Darmstadt, Germany). The PPMS provides a flexible automated workstation that can perform a variety of experiments requiring precise thermal control in the range 1.9 K to 400 K. The temperature measurement accuracy is ± 0.5%, and temperature stability is ≤0.2% for temperatures ≤10 K and ≤0.02% for temperatures >10 K.
To measure the resistance by the four-wire method, the structures (Sensor 1 and Sensor 2) were mounted on a standard PPMS sample puck (see Figure 2). Then, the sample puck was introduced in the PPMS sample chamber, the cold region of the refrigeration unit under investigation, which is constructed of copper to provide uniform temperature region. The sample puck is connected to the PPMS acquisition system using 12-pin connector placed on the bottom of the sample chamber. In a four-wire resistance measurement general configuration, the current is passed through a sample via two current leads, and two separate voltage leads measure the potential difference across the sample. To maintain a minimum value of the current adsorbed by the voltage leads, the internal voltmeter used has a very high impedance. For the following experiments, two sensor structures were obtained by employing a dropcast technique of a small quantity of solution (20 µL) between the electrodes. The first structure (here called Sensor 1) consists of two copper electrodes on a ceramic substrate with a distance of 1.25 mm between electrodes. The second structure (here called Sensor 2) consists of two gold electrodes on a Sitall substrate (crystalline glass-ceramic with ultra-low coefficient of thermal expansion) with a distance of 0.25 mm between electrodes. To align the SWCNTs and to form multilayers of nanotubes between the two electrodes [19,20], the structures that we prepared were subjected to a dielectrophoresis process by applying 1 VDC potential, for 1 min, over the electrodes. During this time, dry air at 60 • C was blown over the structures. Further, the sensors were subjected to a high temperature treatment (200 • C for 60 min) for achieving structure stability of the sensitive active layer.

Characterization Methods
The two sensor structures prepared in this investigation were tested at very low temperature in a refrigeration unit Quantum Design Physical Property Measurement System (PPMS) with EverCool-II ® Cryogen-Free Cooling Technology, model P935A (referred as PPMS, Quantum Design, Inc., GmbH, Darmstadt, Germany). The PPMS provides a flexible automated workstation that can perform a variety of experiments requiring precise thermal control in the range 1.9 K to 400 K. The temperature measurement accuracy is ±0.5%, and temperature stability is ≤0.2% for temperatures ≤10 K and ≤0.02% for temperatures >10 K.
To measure the resistance by the four-wire method, the structures (Sensor 1 and Sensor 2) were mounted on a standard PPMS sample puck (see Figure 2). Then, the sample puck was introduced in the PPMS sample chamber, the cold region of the refrigeration unit under investigation, which is constructed of copper to provide uniform temperature region. The sample puck is connected to the PPMS acquisition system using 12-pin connector placed on the bottom of the sample chamber. In a four-wire resistance measurement general configuration, the current is passed through a sample via two current leads, and two separate voltage leads measure the potential difference across the sample. To maintain a minimum value of the current adsorbed by the voltage leads, the internal voltmeter used has a very high impedance. Scanning electron microscopy (SEM) image, as shown in Figure 3, reveals the alignment, between the electrodes, of the carbon nanotubes with Pt nanoparticles deposited on them. The role of the Pt is to facilitate the adherence of the SWCNTs between gold/copper and between individual SWCNTs, knowing that the length of nanotubes is much shorter than the distance between electrodes. Strong adherence is expected to be obtained, which is formed during the drying process.

Analysis and Interpretation of Results
Both sensors have been cooled several times from room temperature (approximately 300 K) down to approximately 2 K, until resistance reached a constant value with error less than 1%. This procedure can be considered as an ageing process. The resulting variations of resistance with temperature for the studied structures, for two different run tests, initial (Test 1) and after five runs (Test 5), during heating/cooling cycles, are shown in Figure 4. Scanning electron microscopy (SEM) image, as shown in Figure 3, reveals the alignment, between the electrodes, of the carbon nanotubes with Pt nanoparticles deposited on them. The role of the Pt is to facilitate the adherence of the SWCNTs between gold/copper and between individual SWCNTs, knowing that the length of nanotubes is much shorter than the distance between electrodes. Strong adherence is expected to be obtained, which is formed during the drying process. Scanning electron microscopy (SEM) image, as shown in Figure 3, reveals the alignment, between the electrodes, of the carbon nanotubes with Pt nanoparticles deposited on them. The role of the Pt is to facilitate the adherence of the SWCNTs between gold/copper and between individual SWCNTs, knowing that the length of nanotubes is much shorter than the distance between electrodes. Strong adherence is expected to be obtained, which is formed during the drying process.

Analysis and Interpretation of Results
Both sensors have been cooled several times from room temperature (approximately 300 K) down to approximately 2 K, until resistance reached a constant value with error less than 1%. This procedure can be considered as an ageing process. The resulting variations of resistance with temperature for the studied structures, for two different run tests, initial (Test 1) and after five runs (Test 5), during heating/cooling cycles, are shown in Figure 4.

Analysis and Interpretation of Results
Both sensors have been cooled several times from room temperature (approximately 300 K) down to approximately 2 K, until resistance reached a constant value with error less than 1%. This procedure can be considered as an ageing process. The resulting variations of resistance with temperature for the studied structures, for two different run tests, initial (Test 1) and after five runs (Test 5), during heating/cooling cycles, are shown in Figure 4.  The nanotubes based structures showed sensitivity and fast response at temperature variations, having the initial resistance, after the ageing, approximately 7.69 kΩ for Sensor 1 and 55.06 kΩ for Sensor 2. The mechanism of conductivity in SWCNTs structures can be explained by the percolation theory [21,22] and can be calculated as:

 
, where L is a characteristic length, which depends of the concentration of sites, Z is the resistance of the path with the lowest average resistance. With a decrease in temperature, the charge concentration (electrons and holes) decreases, resulting in a reduced number of charge carriers available for recombination, and, statistically, the conductivity of the CNTs decreases, whereas the resistivity of the CNTs increases [23].
Using a nonlinear fitting algorithm, from the Origin software package, we investigated several mathematical expressions capable of describing the experimental data. For both, Sensor 1 and Sensor 2, the experimental data can be well approximated by the following equation:   R (Root Mean Square) for Sensor 1 and Sensor 2, corresponding to Test 1 and Test 5, respectively, are listed in Table 1. The differences between the experimental results and fitted data, Test 1 and Test 5, for Sensor 1 and Sensor 2, are plotted in Figure 5. The nanotubes based structures showed sensitivity and fast response at temperature variations, having the initial resistance, after the ageing, approximately 7.69 kΩ for Sensor 1 and 55.06 kΩ for Sensor 2. The mechanism of conductivity in SWCNTs structures can be explained by the percolation theory [21,22] and can be calculated as: σ = 1 LZ , where L is a characteristic length, which depends of the concentration of sites, Z is the resistance of the path with the lowest average resistance. With a decrease in temperature, the charge concentration (electrons and holes) decreases, resulting in a reduced number of charge carriers available for recombination, and, statistically, the conductivity of the CNTs decreases, whereas the resistivity of the CNTs increases [23].
Using a nonlinear fitting algorithm, from the Origin software package, we investigated several mathematical expressions capable of describing the experimental data. For both, Sensor 1 and Sensor 2, the experimental data can be well approximated by the following equation: The value of the coefficients a 0 , a 1 , a 2 , a 3 , b 1 , b 2 , b 3 and R 2 (Root Mean Square) for Sensor 1 and Sensor 2, corresponding to Test 1 and Test 5, respectively, are listed in Table 1. The differences between the experimental results and fitted data, Test 1 and Test 5, for Sensor 1 and Sensor 2, are plotted in Figure 5. In Figure 5, it is shown that the differences between the fitted data and experimental data are very small, suggesting that the empirical equation estimates are almost 100% comparable with the experimental results, especially after undergoing the ageing process ( 2 R = 1).
The characteristics Resistance vs. Temperature are almost linear in temperature domain from 300 K to 77 K, suggesting an ohmic contact, this being confirmed by [6,7,14,24,25]. Further, our attention was focused on the sensor characterization after ageing process, on the temperature domain below liquid nitrogen temperature (77 K). As the temperature moved down from 77 K to approximately 2 K, the sensor structure resistance increased from 18.3 kΩ to 623 kΩ for Sensor 1, and increased from 153 kΩ to 979 kΩ for Sensor 2, having an allure that suggests a semiconductor behavior (p-type semiconductor) [14,26,27].
One of the parameters used to characterize a temperature sensor is the Temperature Coefficient of Resistance (TCR), defined as the relative change of a physical property (resistance) that is associated with a given change in temperature where R is the resistivity of the sensor structure at temperature T and 0 R is the resistivity at a reference temperature 0 0 C T  [28]. The plotted results for TCR are presented in Figure 6. In Figure 6, it can be further seen that Sensor 1 has a bigger change in resistance than Sensor 2, especially at temperatures below 20 K, where the TCR average for Sensor 1 is 1.473%/K and for Sensor 2 is 0.365%/K. The CNTs resulted films used in the structures of Sensor 1 and Sensor 2, and the contacts between them together with the terminal electrodes were not affected or destroyed by In Figure 5, it is shown that the differences between the fitted data and experimental data are very small, suggesting that the empirical equation estimates are almost 100% comparable with the experimental results, especially after undergoing the ageing process (R 2 = 1).
The R-T characteristics are almost linear in temperature domain from 300 K to 77 K, suggesting an ohmic contact, this being confirmed by [6,7,14,24,25]. Further, our attention was focused on the sensor characterization after ageing process, on the temperature domain below liquid nitrogen temperature (77 K). As the temperature moved down from 77 K to approximately 2 K, the sensor structure resistance increased from 18.3 kΩ to 623 kΩ for Sensor 1, and increased from 153 kΩ to 979 kΩ for Sensor 2, having an allure that suggests a semiconductor behavior (p-type semiconductor) [14,26,27].
One of the parameters used to characterize a temperature sensor is the Temperature Coefficient of Resistance (TCR), defined as the relative change of a physical property (resistance) that is associated with a given change in temperature where R is the resistivity of the sensor structure at temperature T and R 0 is the resistivity at a reference temperature T 0 = 0 • C [28]. The plotted results for TCR are presented in Figure 6. In Figure 5, it is shown that the differences between the fitted data and experimental data are very small, suggesting that the empirical equation estimates are almost 100% comparable with the experimental results, especially after undergoing the ageing process ( The R-T characteristics are almost linear in temperature domain from 300 K to 77 K, suggesting an ohmic contact, this being confirmed by [6,7,14,24,25]. Further, our attention was focused on the sensor characterization after ageing process, on the temperature domain below liquid nitrogen temperature (77 K). As the temperature moved down from 77 K to approximately 2 K, the sensor structure resistance increased from 18.3 kΩ to 623 kΩ for Sensor 1, and increased from 153 kΩ to 979 kΩ for Sensor 2, having an allure that suggests a semiconductor behavior (p-type semiconductor) [14,26,27].
One of the parameters used to characterize a temperature sensor is the Temperature Coefficient of Resistance (TCR), defined as the relative change of a physical property (resistance) that is associated with a given change in temperature where R is the resistivity of the sensor structure at temperature T and 0 R is the resistivity at a reference temperature 0 0 C T  [28]. The plotted results for TCR are presented in Figure 6.  In Figure 6, it can be further seen that Sensor 1 has a bigger change in resistance than Sensor 2, especially at temperatures below 20 K, where the TCR average for Sensor 1 is 1.473%/K and for Sensor 2 is 0.365%/K. The CNTs resulted films used in the structures of Sensor 1 and Sensor 2, and the contacts between them together with the terminal electrodes were not affected or destroyed by temperature variations occurring because of possible mismatches at the device interfaces [6]. The sensors, which were tested in the domain 2-300 K with slow heating-cooling cycles, proved to present very good stability. To intensely investigate the R-T curves, we plotted the dimensionless sensitivity (as shown in Figure 7) and absolute temperature resolution (Figure 8) for each structure.
The dimensionless sensitivity S D = T R dR dT gives the relative temperature sensitivity of the sensor at temperature T. S D ranges from 0.2 to 6 for most common cryogenic temperature sensors, depending on temperature and sensor type [29]. A large specific sensitivity allows the resolution of small temperatures relative to the temperature measured [4], but the temperature range is dependent and, in some situations, is limited by the resistance measurement system accuracy.
Sensors 2017, 17, x FOR PEER REVIEW 7 of 10 temperature variations occurring because of possible mismatches at the device interfaces [6]. The sensors, which were tested in the domain 2-300 K with slow heating-cooling cycles, proved to present very good stability. To intensely investigate the Resistance vs. Temperature curves, we plotted the dimensionless sensitivity (as shown in Figure 7) and absolute temperature resolution ( Figure 8) for each structure.
The dimensionless sensitivity gives the relative temperature sensitivity of the sensor at temperature T . D S ranges from 0.2 to 6 for most common cryogenic temperature sensors, depending on temperature and sensor type [29]. A large specific sensitivity allows the resolution of small temperatures relative to the temperature measured [4], but the temperature range is dependent and, in some situations, is limited by the resistance measurement system accuracy. The absolute temperature resolution is the smallest temperature change that can be detected, and depends on the sensor characteristics and on the measurement system resolution according to the following expression [4]: where T  is the temperature resolution of a sensor measuring a temperature  The absolute temperature resolution is the smallest temperature change that can be detected, and depends on the sensor characteristics and on the measurement system resolution according to the following expression [4]: where ε T is the temperature resolution of a sensor measuring a temperature T, ε V = 19.07 nV is the measurement system resolution, I = 1 mA is excitation current and dR dT is the sensor sensitivity. From graphics in Figure 7, it is evident that the effect of temperature on the sensor resistance is about the same for both sensor structures. In the domain analyzed here, viz. 2-77 K, Sensor 2 is more sensitive to temperature change than the Sensor 1 due to the nanotube density between the structures' electrodes and due to the gap dimension between Cu electrodes (Sensor 1) and gold electrodes (Sensor 2).
Additionally, we studied the influence of the magnetic field on the Sensor 1 in the temperature From graphics in Figure 7, it is evident that the effect of temperature on the sensor resistance is about the same for both sensor structures. In the domain analyzed here, viz. 2-77 K, Sensor 2 is more sensitive to temperature change than the Sensor 1 due to the nanotube density between the structures' electrodes and due to the gap dimension between Cu electrodes (Sensor 1) and gold electrodes (Sensor 2).
Additionally, we studied the influence of the magnetic field on the Sensor 1 in the temperature range 2-77 K. Figure 9 shows the temperature dependence of the resistance under the influence of a magnetic field of B = 2 T (Tesla), compared with the characteristic measured at zero field. We conclude that the sensor is sensitive to the magnetic field with steeper drop in resistance, especially at temperatures below 15 K. From graphics in Figure 7, it is evident that the effect of temperature on the sensor resistance is about the same for both sensor structures. In the domain analyzed here, viz. 2-77 K, Sensor 2 is more sensitive to temperature change than the Sensor 1 due to the nanotube density between the structures' electrodes and due to the gap dimension between Cu electrodes (Sensor 1) and gold electrodes (Sensor 2).
Additionally, we studied the influence of the magnetic field on the Sensor 1 in the temperature range 2-77 K. Figure 9 shows the temperature dependence of the resistance under the influence of a magnetic field of B = 2 T (Tesla), compared with the characteristic measured at zero field. We conclude that the sensor is sensitive to the magnetic field with steeper drop in resistance, especially at temperatures below 15 K.

Discussion and Conclusions
We have investigated resistance vs. temperature characteristic for two different temperature sensors with SWCNTs as sensing element. The sensors were obtained by dropcasting the nanotubes from a solution between two electrodes of a support structure, followed by a dielectrophoresis alignment process and high temperature annealing. Using this methodology, stable layers of aligned SWCNTs were obtained. For temperature range under investigation, i.e., from 300 K to 2 K, we have established an empirical equation for the sensors characteristics R(T) in the form of an exponential

Discussion and Conclusions
We have investigated resistance vs. temperature characteristic for two different temperature sensors with SWCNTs as sensing element. The sensors were obtained by dropcasting the nanotubes from a solution between two electrodes of a support structure, followed by a dielectrophoresis alignment process and high temperature annealing. Using this methodology, stable layers of aligned SWCNTs were obtained. For temperature range under investigation, i.e., from 300 K to 2 K, we have established an empirical equation for the sensors characteristics R(T) in the form of an exponential equation, consisting of sum of three exponential terms and one constant. This is indicative of a semiconductor behavior for the SWCNTs, and fit the existing data. The differences between experiments and fitted data are very small, thus confirming a very good mathematical approximation of the experimental results. In addition, the structures' specific sensitivity, thermal coefficient of resistance and absolute temperature resolution are presented. These experimental results show a good reproducibility over the entire temperature range. We highlighted that the SWCNTs based structures were very sensitive, especially in the temperature range 2-77 K. The R(T) characteristic obtained here proves that CNT-based sensors can be successfully used for cryogenic temperature measurements and we expect that it will be intensely studied for temperature measurements for future application.

Conflicts of Interest:
The authors declare no conflict of interest.