Enhanced thermal sensitivity in single metal thermocouple: significance of thickness-engineering of the metal layers

Single metal thermocouples (SMTs) have recently been developed with a new design concept of width-engineering of metal segments. In such designs, two segments of different micro-width are formed to obtain different levels of Seebeck effects. The variations in the Seebeck effect achieved from dissimilar segment width are small. In addition, the fabrication of such micro-width patterns requires special fabrication facilities such as photolithography or electron-beam lithography. In this paper, an alternative method is presented that has the potential to give high thermal sensing SMTs and requires no sophisticated facilities to fabricate. The method is based on thickness-engineering instead of width-engineering, and thus devices can be obtained from commonly available thin film deposition techniques. Constructing better thermal sensing SMTs is possible with this approach as thickness can be easily and conveniently varied down to nanoscale range which is necessary to achieve significant changes in the Seebeck effects from effectively utilizing size effects. As a result, a high thermal sensing bismuth based-SMT has been fabricated with a sensitivity of as high as 31 μV K−1, one of the highest values reported for SMTs. It is straightforward, more convenient over width-engineering approach and thus SMTs can be easily developed.


Introduction
Thermocouples, a universal type of thermometer, are typically made up of two dissimilar metals [1,2]. These devices work on the principle of the Seebeck effect and therefore use of two metals of dissimilar Seebeck coefficients is a basic requirement to observe a voltage signal from these sensors [1]. When two dissimilar metals/conductors of different Seebeck coefficients (S 1 and S 2 ) are joined to form a thermocouple junction, an open circuit voltage signal can be observed by heating or cooling the junction. The magnitude of this voltage (ΔV ) is directly proportional to (i) temperature difference (ΔT) between the junction and open ends of the metals and (ii) difference between the Seebeck coefficients (S 1 −S 2 ) of metals, which can be expressed as [1]: Where S1 and S2 are the Seebeck coefficients of metal 1 and 2, respectively. Alternatively, researchers have come up with a new concept of 'single metal thermocouple' (SMT) in which a single metal layer with two segment patterns of different widths are used, as illustrated in figure 1(a) [1][2][3][4]. In these width-engineered patterns with at least one of the segments of narrow width, the Seebeck coefficient can be modified with respect to the wider segment to obtain a net voltage from the sensor. As they need micrometer wide patterns, the construction of the width-engineered thermocouples requires special fabrication techniques such as photolithography or electron-beam lithography [1,2,5,6]. Further, due to the limitations of reducing width down to sub 100 nm range, the thermoelectric changes between the wide and narrow segment are also limited. As a result, most of the reported width-engineered SMTs have low thermal sensitivity of about 1-5 μV K −1 . Here, an alternative and straightforward method to fabricate SMTs has been developed that overcomes the above difficulties, enabling the construction of SMTs with better thermal sensitivity. The method is based on Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. thickness variation of the metal patterns instead of width, as the thickness of the patterns can be easily and conveniently varied or controlled in the nanoscale range with common thin film deposition techniques ( figure 1(b)). As a proof of concept, bismuth-based thickness-engineered SMT has been designed and developed that shows a high thermal sensitivity of 31 μV K −1 , one of the highest values reported in SMTs.

Experimental details 2.1. Thin film fabrication
For initial thermoelectric testing, bismuth (Bi) thin films of different thickness (t∼500 nm, 200 nm, 100 nm, 50 nm, 20 nm) on glass substrates have been deposited by the sputter coating method (Model: Quorum Q150T S coater) using a bismuth target (99.9% purity, PI-KEM Ltd). For every deposition run, the sputtering chamber was evacuated to 10 −4 mbar and films were deposited at room temperature under an argon atmosphere (10 −2 mbar). A sputter current of 20 mA was used for all the films and thickness was controlled using the film thickness monitor accessory of the instrument.

Fabrication of single metal thermocouple (SMT)
The SMT device consisting of two Bi layer strips/segments of t∼20 nm and 500 nm was patterned on a microscope glass slide (size: 75 mm × 25 mm). To fabricate this device as illustrated in figure 1(b), first, a thin metal layer strip (t∼20 nm, width: 8 mm) was deposited and then an L-shaped thick metal layer (t∼500 nm, width: 8 mm) using a mask. Another SMT device having two junctions was fabricated by patterning a thin metal layer (t∼20 nm) between the two thick metal layers (t∼500 nm).

Measurements and characterisation
The Seebeck coefficient measurements on Bi thin films of different thickness were performed using a standard method with a homemade setup [7]. The thermoelectric voltage (ΔV ) produced from the thin film samples have been recorded at different temperature gradients (ΔT) and the Seebeck coefficients were estimated using the expression, S=−(ΔV/ΔT) [8,9]. The electrical resistivity of the films were obtained from the standard fourprobe method. The output characteristics of the SMT sensors were performed by heating junction to different temperatures and recording the output voltage signal. The junction heating was achieved from a small resistive heater (size: 5 mm × 5 mm, thickness: 1 mm, resistance: 100 Ω) [10] by supplying DC power inputs. This heater and a K-type thermocouple were fixed on a thin copper plate (thickness∼0.5 mm) of similar size as the heater using high-temperature glue; the resulting heating element was then placed on the junction area of the SMT. Good physical contact between the heating element and junction was maintained by a pressing arrangement with a spring clamp set. All the thermoelectric voltages of the individual films and the output voltage signal of SMT sensors were measured with micro-voltmeters (Model: HMC 8012 DMM). The DC voltage accuracy for this model of multimeter is 0.015% of reading +0.002% of range. So, for the most precise range of 400.000 mV, the best accuracy would be about 8 μV. The surface morphology of the junction area of a sensor was obtained by Scanning Electron Microscopy (SEM) using Zeiss Evo LS25 SEM.

Results and discussion
The proposed thickness-engineered design of a single metal thermocouple (SMT) based on different thickness metal layers is schematically displayed in figure 1(b). Bi metal has been chosen for the purpose due to its more prominent thermoelectric changes with respect to the changes in the material dimension and also it is one of the metals to display substantial thermoelectric effects, as compared to any other metal [11,12]. The changes in the thermoelectric effects have been confirmed by measuring the Seebeck coefficient (S) of the individual Bi films of different thickness, the data is shown in figure 2(a). The observed room temperature S value of a 500 nm thick film is −52 μV K −1 , which has seen a decreasing trend with respect to reduction of film thickness; −46 μV K −1 , −39 μV K −1 , −21 μV K −1 , and −15 μV K −1 are the S values of the films having thickness 200 nm, 100 nm, 50 nm, and 20 nm, respectively. The reduction in the Seebeck coefficient in thin films as compared to their bulk values has been observed in most of the metal films due to the increased electron scattering at the surface and theoretically described by the Fuchs-Sondheimer model [1,13,14]. The variation of resistivity as a function of film thickness is shown in figure S1 (available online at stacks.iop.org/ERX/3/035015/mmedia) (supplementary file); the resistivity of the thick film (t∼500 nm) is found to be 5.4 × 10 −4 Ω cm which is slightly more than its bulk resistivity and consistent with previously reported values for Bi thin film [12,15]. With the decrease in film thickness, it is found to increase as shown in figure S1 (supplementary file). This increase is attributed to the fact that charge carriers in thin films undergo scattering at film surface which is more pronounced in the case of very thin films.
The dependence of absolute values of S reduction, |ΔS|, on the films of thickness, t∼200 nm, 100 nm, 50 nm, and 20 nm with respect to the first film (t∼500 nm) are shown in figure 2(b) with an observed maximum |ΔS| for the film with t∼20 nm. The change of S from −52 μV K −1 to −15 μV K −1 demonstrates a very high relative reduction of Seebeck coefficient |ΔS/S t=500 nm | of about 70% which is higher than values achieved in width controlled segments. For illustration, a maximum relative reduction of Seebeck coefficients of about 30%-35% was reported in Ta and Ti metals by decreasing the segment width from 100 μm to 3 μm [14]. But in the case of Bi, the reduction of segment width to 3 μm has resulted in a less than 6% reduction [14]. These data indicate that the thickness-engineered sensors can have better output signal or thermal sensitivity in contrast to the width-engineered sensors.
The changes in the Seebeck coefficient with respect to material dimension can be understood through standard expressions, such as the Mott-Jones equation for S (in metals) which is expressed as [14,16] p s = -  The variations to the S from the second term in the above equation are not expected as the term (dlnA/dE) E=Ef can change in the case of atomic-scale material dimensions hence the second term is approximately considered constant [14]. Therefore, the size effects on the S are in general dominated by the first term (related to λ). It is well-known that surface scattering is an additional mechanism that occurs in thin films in addition to the phonon, impurity, and defect scattering mechanisms [13]. The surface scattering can influence the S and other transport properties when the thickness is comparable to the mean free path (λ) of the charge carriers (for most metals, the range of λ is around 40-100 nm) [13,14]. Therefore, the bulk S values can be changed by reducing at least one of the dimensions to nanoscale and, significant changes in S can be expected in thin films of nanoscale thickness as compared to microscale width-engineered patterns. The photograph of SMT displayed in figure 3 has two segments of t=500 nm and 20 nm as the maximum difference in the Seebeck coefficient |ΔS| can be obtained from this combination. The surface morphology of the films and the junction region are also given in figure 3 which show the morphological differences between the thick and thin film. The open-circuit voltage (ΔV ) as a function of temperature gradient (ΔT) is shown in figure 3(d); a linear increase in the ΔV with respect to the increase in ΔT has been observed with a slope value of 31 μV K −1 , one of the highest sensitivity values from the SMTs reported till now. The observed sensitivity is in near agreement with the estimated ΔS=37 μV K −1 from S values of the individual segments of t=500 nm and 20 nm. The output signal data points are well aligned with linear fittings with a good R 2 value of 0.995 which suggests that the sensor has a nearly constant sensitivity factor and can be useful for thermal sensing applications around room temperature. These measurements were carried out at a room temperature of ∼21±2°C and the devices can be used up to ∼80°C without any problems. The small decrease in the observed sensitivity of the thermocouple as compared to the estimated ΔS might be associated with the uncertainties in the junction heating temperatures. Previous work on a Pd thin film SMT with dual strips of widths 30 μm-200 nm has reported a maximum sensitivity of 2.2 μV K −1 and for a Ni film based SMT with strips widths 100 μm-3 μm, a maximum sensitivity of 1.1 μV K −1 has been reported [2,5]. Recent studies suggest that the high thermal sensitivity values as compared to metal-based devices can be observed in carbon or graphene-based devices [4,17]. The sensing capability and reliability of the Bi SMT have been studied under a continuous and cyclic operation by touching the junction with a finger wearing a nitrile glove for 10 s and recording the generated signal from the SMT. As shown by figure 4(a), a sharp increase in the voltage signal is observed for each touching cycle and the observed maximum voltage is consistent. The observed voltage signal is then converted into temperature by using the thermal sensitivity value of 31 μV K −1 in order to realize the temperature change before and after touching the sensor; the data of which is plotted in figure 4(b). The observed peak values from the 10 repetitive touching cycles are found to be in the range of 6.70°C-6.96°C with an estimated average value of 6.85 (±0.08)°C (the data scatter range ±0.08°C was estimated from ten peak temperature values of the heating cycles). Further, the reliability of this observed temperature change from the Bi SMT was confirmed by a similar procedure with a standard K-type thermocouple sensor; the rise in temperature was recorded by touching it with a gloved finger for a similar time period of 10 s. The recorded variation in the temperature is shown in figure S2 (supplementary file); the temperature reading has changed from 21.3°C (room temperature) to 27.8°C before removing the finger after 10 s which indicates a net change of 6.5°C. These measurements confirm that the SMT is giving reliable information, the small difference between the temperature data from the SMT and standard K-type thermocouple, which is about 0.3°C, could be due to the variations in the pressure applied by the finger or the pressing time. Further, the thermal response time of the SMT is very short which is evident from the sharp rise of the voltage signal observed immediately after touching the sensor; a magnified single curve has been shown in figure S3 (supplementary file).
The sensing reliability of the SMT under heating cycles was also studied by heating with a standard resistive heater fixed on the junction; figure 5 shows data of two heating cycles. The heater was applied with a fixed dc power for ∼1 min and turned off, the output voltage variation of the SMT was recorded; a similar procedure was repeated to record sensing data for the second heating cycle. The sensor output voltage was then converted into temperature using its thermal sensitivity value (see figure 5(b)). The temperature change of ∼15.5°C was observed from the SMT sensor whereas the temperature of the standard thermocouple attached to the heater The dissimilar Seebeck effects in Bi film segments of different thickness have also been observed and confirmed with a design of two junction (J1 and J2) pattern. For this, a thin film segment (t∼20 nm) was deposited between two thick film segments (t∼500 nm) as shown in figure 6. When one of the junctions was heated using a small resistive heater placed on one of the junctions, there was a voltage generation from the pattern and also, there was a change in the polarity of the output voltage when the hot junction was changed from J1 to J2 or vice-versa; the output voltage response is presented in figure 6(b). This confirms that the change in the direction of net flow of charges through the pattern made up of Bi films of different thickness due to different levels of Seebeck effects in the segments. Overall, the present study demonstrates that the changes in thermoelectric powers of Bi metal are small in microscale and thus significant changes can be observed by going down to nanoscale which can be easily and effectively achieved by varying thickness instead of the width of the segments.
In addition, studies on Bi thin films and nanowires have reported a significantly decreased thermal conductivity in contrast to its bulk value [18][19][20][21]. With good Seebeck coefficient and good electrical conductivity added with reduced thermal conductivity can make this element as a promising n-type material [22,23] and can be used in combination with good p-type thermoelectric materials for low temperature heat conversion applications [24][25][26][27][28][29].

Conclusions
The present work introduces an alternative method to achieve higher thermal sensitivity from the single metal thermocouples (SMT). More prominent changes in the Seebeck coefficient can be observed by changing the metal film thickness instead of width and, as a proof of concept, high thermal sensitivity of 31 μV K −1 has been achieved in thickness-engineered Bi-based SMT constructed from a combination of 500 nm and 20 nm thick segments. The fabrication of thickness-engineered SMTs is simple and does not require any special facilities, which makes the approach more preferable and economical.