Fabrication and Thermoelectric Characterization of Transition Metal Silicide-Based Composite Thermocouples

Metal silicide-based thermocouples were fabricated by screen printing thick films of the powder compositions onto alumina tapes followed by lamination and sintering processes. The legs of the embedded thermocouples were composed of composite compositions consisting of MoSi2, WSi2, ZrSi2, or TaSi2 with an additional 10 vol % Al2O3 to form a silicide–oxide composite. The structural and high-temperature thermoelectric properties of the composite thermocouples were examined using X-ray diffraction, scanning electron microscopy and a typical hot–cold junction measurement technique. MoSi2-Al2O3 and WSi2-Al2O3 composites exhibited higher intrinsic Seebeck coefficients (22.2–30.0 µV/K) at high-temperature gradients, which were calculated from the thermoelectric data of composite//Pt thermocouples. The composite thermocouples generated a thermoelectric voltage up to 16.0 mV at high-temperature gradients. The MoSi2-Al2O3//TaSi2-Al2O3 thermocouple displayed a better performance at high temperatures. The Seebeck coefficients of composite thermocouples were found to range between 20.9 and 73.0 µV/K at a temperature gradient of 1000 °C. There was a significant difference between the calculated and measured Seebeck coefficients of these thermocouples, which indicated the significant influence of secondary silicide phases (e.g., Mo5Si3, Ta5Si3) and possible local compositional changes on the overall thermoelectric response. The thermoelectric performance, high sensitivity, and cost efficiency of metal silicide–alumina ceramic composite thermocouples showed promise for high-temperature and harsh-environment sensing applications.


Introduction
Accurate, real-time, and reliable temperature measurements are crucial for safer and more efficient operation in many industrial applications, such as power generation, coal gasification, and metal and glass manufacturing [1,2]. Many of these systems operate at temperatures reaching 1500-1650 • C, with possible extreme physical conditions such as highly redox atmospheres, and corrosive liquids and gases, which present additional challenges [3,4]. Therefore, there is an increasing demand for advanced sensing materials that are capable of measuring temperatures under harsh environments.
Noble metal and metal alloy thermocouples (e.g., Pt-Rh//Pt, chromel//alumel) have been extensively used in high-temperature sensing over the years. However, these thermocouples are proven to be incapable of withstanding increasingly harsh conditions due to some limitations, such as selective oxidation in air or oxygen atmospheres (e.g., oxidation of Rh at~600-800 • C), investigate their microstructures after sintering. For high-temperature thermoelectric characterization of the as-fabricated composite//Pt and composite//composite thermocouples embedded in alumina preforms, an atmospheric controlled high-temperature furnace was used. Thermoelectric measurements were conducted at temperatures up to 700-1000 °C with increasing thermal gradient by 2 °C/min under argon atmosphere. All electrical contacts were made by using platinum wires and paste at the cold junction. During the measurements, K-and S-type thermocouples were used to record cold and hot junction temperatures with a digital connection from a National Instruments (NI; Austin, TX, USA) thermocouple reader to the computer through LabVIEW software. Thermoelectric voltage (mV) data were acquired using a digital multimeter (Keithley 2100, Tektronix, Beaverton, OR, USA) controlled by the LabVIEW software during the measurements. The thermoelectric voltage output was recorded as point-values due to the stabilized thermal gradient. Seebeck coefficients (S; µ V/K) were then calculated by applying the polynomial fitting to the thermoelectric voltage-temperature gradient (mV-ΔT) curves obtained. It is also important to note that multiple samples were tested for each thermocouple configuration. However, the presented thermoelectric results are representation of one sample for each thermocouple configuration, and thus, they do not represent the statistical average values.

Structural and Thermoelectric Characterization
The phase development and secondary phase formation of the as-sintered metal silicide-alumina composite thick-film thermocouple legs were analyzed by using an X-ray diffraction instrument (XRD; X'Pert Pro Panalytical, Westborough, MA, USA) with a CuK α radiation source. A field-emission scanning electron microscopy (FE-SEM; Hitachi S-4700F, Tokyo, Japan) was used to investigate their microstructures after sintering. For high-temperature thermoelectric characterization of the as-fabricated composite//Pt and composite//composite thermocouples embedded in alumina preforms, an atmospheric controlled high-temperature furnace was used. Thermoelectric measurements were conducted at temperatures up to 700-1000 • C with increasing thermal gradient by 2 • C/min under argon atmosphere. All electrical contacts were made by using platinum wires and paste at the cold junction. During the measurements, K-and S-type thermocouples were used to record cold and hot junction temperatures with a digital connection from a National Instruments (NI; Austin, TX, USA) thermocouple reader to the computer through LabVIEW software. Thermoelectric voltage (mV) data were acquired using a digital multimeter (Keithley 2100, Tektronix, Beaverton, OR, USA) controlled by the LabVIEW software during the measurements. The thermoelectric voltage output was recorded as point-values due to the stabilized thermal gradient. Seebeck coefficients (S; µV/K) were then calculated by applying the polynomial fitting to the thermoelectric voltage-temperature gradient (mV-∆T) curves obtained. It is also important to note that multiple samples were tested for each thermocouple configuration. However, the presented thermoelectric results are representation of one sample for each thermocouple configuration, and thus, they do not represent the statistical average values.

Phase Development and Microstructures of Composites
The XRD phase analysis was completed to investigate the phase development in the various metal silicide-alumina composite systems (thermocouple legs) after sintering at 1500 • C for 2 h. The XRD patterns of the MoSi 2 -Al 2 O 3 , WSi 2 -Al 2 O 3 , TaSi 2 -Al 2 O 3 , and ZrSi 2 -Al 2 O 3 composites all having  volume percentages are presented in Figure 2. The XRD peaks corresponding to the starting metal silicides (MoSi 2 , WSi 2 , TaSi 2 , and ZrSi 2 ) and alumina phases were detected. Secondary silicide phases were also observed after sintering for all the composite systems. The XRD results revealed that secondary 5-3 metal silicide phases (Mo 5 Si 3 , W 5 Si 3 , Ta 5 Si 3 ) were formed during the sintering of the MoSi 2 -Al 2 O 3 , WSi 2 -Al 2 O 3 , and TaSi 2 -Al 2 O 3 composites. However, the XRD pattern of the ZrSi 2 -Al 2 O 3 composite showed a different trend, since zirconium monosilicide (ZrSi) and silicon (Si) phases were identified as secondary phases after sintering. The secondary phase formation was found to be related to the interaction of metal silicides with alumina, as well as their reaction with environmental sources such as residual oxygen entrapped in pores or starting silicide powders [15][16][17][18]. In the case of the ZrSi 2 -Al 2 O 3 system, it may be also correlated to the possible presence of residual zirconium monosilicide (ZrSi) phase in the starting powder and its relatively low melting point (~1500-1520 • C) compared to that of ZrSi 2 (1620 • C) [19]. Since the melting point of ZrSi is close to the sintering temperature, it could result in partial melting during the sintering hold at temperature, and thus, a slight formation of the silicon (Si) phase as detected by XRD. The presence of ZrSi and Si phases along with the ZrSi 2 phase was similarly found in the Zr/Si diffusion couple after high-temperature annealing at 1200 • C [20]. Furthermore, Rietveld refinement results in the authors' previous study [14] demonstrated that~20.2 vol % of a secondary Mo 5 Si 3 phase formed in the [90-10] MoSi 2 -Al 2 O 3 composite after sintering. Studies reported that the presence of the metal silicide phases with 5-3 stoichiometry could be also beneficial for high-temperature applications due to their higher melting points, good thermal stability, high resistance to creep, and sufficient mechanical strength [21][22][23]. It is important to point out that these secondary phases could significantly affect not only the thermoelectric properties of the fabricated ceramic composite thermocouples, but also their stability and mechanical and thermal properties at elevated temperatures. Figure 3 shows the SEM microstructures of the MoSi 2 -Al 2 O 3 , WSi 2 -Al 2 O 3 , TaSi 2 -Al 2 O 3 , and ZrSi 2 -Al 2 O 3 composites all with  volume percentages after sintering at 1500 • C for 2 h. No significant contrast was achieved to differentiate silicide and alumina particles. It is known from the authors' previous studies [15,21] that these metal silicides have random morphology with 4.2-6.4 µm average particle size, whereas alumina particles have equiaxed morphology and an average size of 0.4 µm. Therefore, the agglomeration of large (silicide) and small (alumina) grains was evident in certain regions for all composite microstructures. This could not adversely affect the electrical percolation network due to the presence of a high metal silicide concentration (90 vol %). A certain level of porosity within all composite microstructures can be also seen. In addition, it is important to note that good adhesion was achieved between screen-printed thermocouple legs (composite//Pt and composite//composite) and the alumina substrate after sintering at 1500 • C. No interfacial microcracking was observed within any composite microstructures. It could be directly interrelated with the close thermal expansion coefficient match between these metal silicides (MoSi 2 : 8.2 × 10 −6 K −1 , WSi 2 : 7.9 × 10 −6 K −1 , ZrSi 2 : 8.3 × 10 −6 K −1 , TaSi 2 : 8.9 × 10 −6 K −1 ) and Al 2 O 3 (~8.0 × 10 −6 K −1 ) in Sensors 2018, 18, 3759 6 of 14 a wide temperature region [19]. This is one of the major benefits of using silicide-alumina composites for high-temperature and harsh-environment sensing.
with the ZrSi2 phase was similarly found in the Zr/Si diffusion couple after high-temperature annealing at 1200 °C [20]. Furthermore, Rietveld refinement results in the authors' previous study [14] demonstrated that ~20.2 vol % of a secondary Mo5Si3 phase formed in the  MoSi2-Al2O3 composite after sintering. Studies reported that the presence of the metal silicide phases with 5-3 stoichiometry could be also beneficial for high-temperature applications due to their higher melting points, good thermal stability, high resistance to creep, and sufficient mechanical strength [21][22][23]. It is important to point out that these secondary phases could significantly affect not only the thermoelectric properties of the fabricated ceramic composite thermocouples, but also their stability and mechanical and thermal properties at elevated temperatures.    contrast was achieved to differentiate silicide and alumina particles. It is known from the authors' previous studies [15,21] that these metal silicides have random morphology with 4.2-6.4 µ m average particle size, whereas alumina particles have equiaxed morphology and an average size of 0.4 µ m. Therefore, the agglomeration of large (silicide) and small (alumina) grains was evident in certain regions for all composite microstructures. This could not adversely affect the electrical percolation network due to the presence of a high metal silicide concentration (90 vol %). A certain level of porosity within all composite microstructures can be also seen. In addition, it is important to note that good adhesion was achieved between screen-printed thermocouple legs (composite//Pt and composite//composite) and the alumina substrate after sintering at 1500 °C. No interfacial microcracking was observed within any composite microstructures. It could be directly interrelated with the close thermal expansion coefficient match between these metal silicides (MoSi2: 8.2 × 10 −6 K −1 , WSi2: 7.9 × 10 −6 K −1 , ZrSi2: 8.3 × 10 −6 K −1 , TaSi2: 8.9 × 10 −6 K −1 ) and Al2O3 (~8.0 × 10 −6 K −1 ) in a wide temperature region [19]. This is one of the major benefits of using silicide-alumina composites for high-temperature and harsh-environment sensing.

Thermoelectric Properties of Composite Thermocouples
Prior to thermoelectric characterization of the composite//composite thermocouples, hightemperature thermoelectric properties of the composite//Pt thermocouples were examined to better understand the intrinsic Seebeck coefficient of each composite. Figure 4a displays the thermoelectric voltage (E in mV) of the MoSi2-Al2O3//Pt, WSi2-Al2O3//Pt, TaSi2-Al2O3//Pt, and ZrSi2-Al2O3//Pt thermocouples as a function of temperature gradient (ΔT in °C). It is clear that the thermoelectric

Thermoelectric Properties of Composite Thermocouples
Prior to thermoelectric characterization of the composite//composite thermocouples, high-temperature thermoelectric properties of the composite//Pt thermocouples were examined to better understand the intrinsic Seebeck coefficient of each composite. Figure 4a Figure 4b shows the effective Seebeck coefficients (S) of these composite//Pt thermocouples, which were calculated by applying the second-or third-order polynomial fitting to the thermoelectric voltage-temperature gradient (E-∆T) curves and then using their first derivatives (slopes) as shown in the equations below based on the theory (slope method) [24,25]: Sensors 2018, 18, x 7 of 14 which was significantly lower compared to the other composite//Pt thermocouples. Figure 4b shows the effective Seebeck coefficients (S) of these composite//Pt thermocouples, which were calculated by applying the second-or third-order polynomial fitting to the thermoelectric voltage-temperature gradient (E-ΔT) curves and then using their first derivatives (slopes) as shown in the equations below based on the theory (slope method) [24,25]: The fitting coefficients (A, B, and D), which were achieved via polynomial fitting and then used for Seebeck coefficient calculations, are additionally listed in Table 2. Prior to these calculations, units of fitting coefficients (including mV) were converted to calculate the Seebeck coefficients in µ V/K. The effective Seebeck coefficient of the [90-10] MoSi2-Al2O3//Pt thermocouple increased from 9.6 to 47.3 µ V/K with increasing temperature gradient (27 °C → 700 °C). This thermocouple displayed higher Seebeck coefficients than other composite//Pt thermocouples throughout ΔT range. In  The fitting coefficients (A, B, and D), which were achieved via polynomial fitting and then used for Seebeck coefficient calculations, are additionally listed in Table 2. Prior to these calculations, units of fitting coefficients (including mV) were converted to calculate the Seebeck coefficients in µV/K. The effective Seebeck coefficient of the    The experimental effective Seebeck coefficient data for the composite//Pt thermocouples (S composite//Pt ) were used to obtain the calculated intrinsic Seebeck coefficients of the metal silicide-oxide composites. For these calculations, the measured temperature-dependent Seebeck coefficient of a platinum wire (S Pt ranging from −5.1 to −17.3 µV/K) was used as a reference from the literature [26]. The intrinsic Seebeck coefficients of silicide-oxide composites (S composite ) were calculated using the equation below: Figure 4c presents the calculated intrinsic Seebeck coefficients of the silicide-oxide composites (S composite ), which were calculated as described above. The reference Seebeck coefficient data of platinum (S Pt ) were also shown within the figure. reported values. This could be related to a certain presence of Mo 5 Si 3 and W 5 Si 3 secondary phases within the composite structures after sintering. As discussed earlier, the amount of Mo 5 Si 3 phase within the [90-10] MoSi 2 -Al 2 O 3 composite was determined as 20.2 vol % using the Rietveld method. These results may indicate that the intrinsic Seebeck coefficients of Mo 5 Si 3 and W 5 Si 3 could have positive and negative signs at room temperature, respectively. However, no study was found on the Seebeck coefficients of these 5-3 intermetallic phases, and thus, future studies are needed. Additionally, the thermoelectric output of the MoSi 2 thin-film thermocouple at 500 • C ranged between 19.4 and 64.1 µV/ • C, depending on the heat treatment conditions [6]. On the other hand, the [90-10] ZrSi 2 -Al 2 O 3 composite exhibited a decreasing linear trend (Figure 4c). This decrease was determined to be from 1.3 to −6.7 µV/K with increasing temperature gradient. For the  TaSi 2 -Al 2 O 3 composite, two different regimes were observed. Its intrinsic Seebeck coefficient firstly increased in negative sign from −1.1 to −6.4 µV/K with increasing ∆T from 27 • C to 300 • C, and then, decreased to −1.6 µV/K with increasing temperature gradient to 700 • C. The Seebeck coefficient of TaSi 2 was reported as 25.0 µV/K at room temperature [7], which is highly different from that of the TaSi 2 -Al 2 O 3 composite in this study. This could be due to the formation of Ta 5 Si 3 phase, which may have a negative intrinsic Seebeck coefficient at the same temperature. Similarly, no data were reported on thermoelectric properties of Ta 5 Si 3 and ZrSi secondary phases. As a result, it should be noted that the calculated intrinsic Seebeck coefficients of  silicide-oxide composites relatively differ from the reported thermoelectric data of these silicides, which demonstrates that secondary silicide phases could have significantly different intrinsic thermoelectric properties than their disilicide forms. Based on the calculated values, MoSi 2 -Al 2 O 3 and WSi 2 -Al 2 O 3 composites showed the highest (and also positive) intrinsic Seebeck coefficients at high temperatures. The intrinsic Seebeck coefficients of ZrSi 2 -Al 2 O 3 and TaSi 2 -Al 2 O 3 composites were calculated to be relatively low, and they both showed a negative value at high temperatures.
After studying the thermoelectric properties of composite//Pt thermocouples and silicide-oxide composites, various composite//composite thermocouples were also fabricated and tested. Similar to the composite//Pt thermocouples, all composite//composite thermocouples were made of  vol % silicide-oxide composites. The thermoelectric voltage and effective Seebeck coefficients of composite//composite thermocouples are presented as a function of temperature gradient in Figure 5.  Figure 5b shows the effective Seebeck coefficients of the composite//composite thermocouples, which were similarly calculated by utilizing the second-or fourth-order (Equation 4) polynomial fitting to the thermoelectric voltage-temperature gradient (E-∆T) curves and then using their first derivatives: 4th order → E = A·∆T + B·∆T 2 + D·∆T 3 + E·∆T 4 + C → S = A + 2B·∆T + 3D·∆T 2 + 4E·∆T 3 (4) Lastly, the effective Seebeck coefficients of composite//composite thermocouples were additionally calculated using the previously calculated intrinsic Seebeck coefficients of  vol % silicide-oxide composites, which were previously presented in Figure 4c. The equation is listed below for these calculations based on the Seebeck theory: The composite and related thermocouple compositions were also abbreviated for simplicity (MA: MoSi2-Al2O3, WA: WSi2-Al2O3, TA: TaSi2-Al2O3, ZA: ZrSi2-Al2O3). For example; the Seebeck coefficient of a MoSi2-Al2O3//TaSi2-Al2O3 (SMA//TA) composite thermocouple was calculated by subtracting the calculated intrinsic Seebeck coefficient of the TaSi2-Al2O3 composite (STA) from that of the MoSi2-Al2O3 composite (SMA). The comparison of the calculated and measured effective Seebeck coefficients of composite//composite thermocouples is presented as a function of temperature gradient (27-700 °C) in Figure 6. The measured Seebeck coefficients of MoSi2-Al2O3//ZrSi2-Al2O3 (MA//ZA) and MoSi2-Al2O3//TaSi2-Al2O3 (MA//TA) thermocouples were found to be mostly lower than their calculated Seebeck coefficients. However, at ΔT > 500 °C, the calculated and measured Lastly, the effective Seebeck coefficients of composite//composite thermocouples were additionally calculated using the previously calculated intrinsic Seebeck coefficients of  vol % silicide-oxide composites, which were previously presented in Figure 4c. The equation is listed below for these calculations based on the Seebeck theory: The composite and related thermocouple compositions were also abbreviated for simplicity These results also pointed out that the thermoelectric performance of composite//composite thermocouples may be positively or negatively affected by possible local compositional changes at the junction, which need to be further investigated in detail.
As presented in Figure 6, the calculated and measured Seebeck coefficients of the MoSi 2 -Al 2 O 3 //WSi 2 -Al 2 O 3 (MA//WA) thermocouple were 8.3 and 14.4 µV/K at ∆T = 27 • C, respectively. The effective Seebeck coefficient of a MoSi 2 //WSi 2 thermocouple (without the addition of Al 2 O 3 ) could be estimated as 5.0 µV/K at room temperature from the previously reported data [7]. Therefore, it is evident that its calculated Seebeck coefficient is relatively close to this theoretical estimation, but its measured Seebeck coefficient is substantially higher, indicating the positive influence of secondary 5-3 silicide phases (Mo 5 Si 3 , W 5 Si 3 ) on overall thermoelectric performance. The calculated and measured Seebeck coefficients of the MoSi 2 -Al 2 O 3 //TaSi 2 -Al 2 O 3 (MA//TA) thermocouple were 5.7 and −8.7 µV/K at ∆T = 27 • C, respectively. At the same temperature gradient, the WSi 2 -Al 2 O 3 //TaSi 2 -Al 2 O 3 (WA//TA) thermocouple displayed calculated and measured Seebeck coefficients of −2.7 and 1.2 µV/K, respectively. From the reported data [7], the effective Seebeck coefficients of MoSi 2 //TaSi 2 and WSi 2 //TaSi 2 thermocouples should be near 30.4 and 25.4 µV/K at room temperature, respectively. These theoretical values are significantly higher than the calculated and measured Seebeck coefficients of these composite//composite (MA//TA and WA//TA) thermocouples. This indicates that the intrinsic Seebeck coefficient of Ta 5 Si 3 secondary phase may be greatly lower than that of tantalum disilicide (TaSi 2 ), which could negatively affect the thermoelectric performance of these composite//composite thermocouples. Similar comparisons were not used for the MoSi 2 -Al 2 O 3 //ZrSi 2 -Al 2 O 3 (MA//ZA) thermocouple due to the lack of intrinsic Seebeck coefficient data for ZrSi 2 phase in literature. In brief, all composite//composite thermocouples exhibited sufficiently high sensitivity (ranging from −8.7 to 73.0 µV/K) at the range of 27-1000 • C. In a similar temperature range, it was reported for the mostly used high-temperature thermocouples that S-type (90% Pt-10% Rh//Pt) and R-type (87% Pt-13% Rh//Pt) thermocouples possess 5.5-13.0 µV/K sensitivity [27]. They also showed higher Seebeck coefficients than platinum//palladium (Pt//Pd) thin-film thermocouples, which displayed performances near −12.9 and −14.3 µV/ • C at 900 • C [3]. Furthermore, in the authors' previous study, the thermoelectric characterization of a long MoSi 2 -Al 2 O 3 //WSi 2 -Al 2 O 3 composite thermocouple demonstrated its highly stable thermoelectric response with no drift in voltage output during a 10 h isothermal hold at 1350 • C [14]. It is highly evident that the thermoelectric performance and the sensitivity of these silicide-oxide-based ceramic composite thermocouples are very promising for high-temperature sensing under harsh environment conditions, and also highly advantageous and cost effective in comparison to precious metal-based temperature sensors. values were similar for the MoSi2-Al2O3//TaSi2-Al2O3 thermocouple. On the contrary, the measured Seebeck coefficients of the MoSi2-Al2O3//WSi2-Al2O3 (MA//WA) thermocouple (14.4-18.9 µ V/K) were higher than the calculated values (7.8-8.3 µ V/K). Such variations could be due to the certain presence of secondary phases within the composite systems. These results may indicate that the overall thermoelectric effect of molybdenum-silicide phases (MoSi2, Mo5Si3) with respect to zirconium-and tantalum-silicide phases (ZrSi2, ZrSi, TaSi2, Ta5Si3) may adversely influence the thermoelectric performance of MoSi2-Al2O3//ZrSi2-Al2O3 and MoSi2-Al2O3//TaSi2-Al2O3 thermocouples. However, the higher measured Seebeck coefficients for the MoSi2-Al2O3//WSi2-Al2O3 thermocouple could imply the positive influence of different silicide couples (MoSi2-W5Si3, Mo5Si3-WSi2, Mo5Si3-W5Si3) on its thermoelectric output. On the other hand, the measured and calculated Seebeck coefficients of the WSi2-Al2O3//TaSi2-Al2O3 (WA//TA) thermocouple displayed a very close match throughout the ΔT range. These results also pointed out that the thermoelectric performance of composite//composite thermocouples may be positively or negatively affected by possible local compositional changes at the junction, which need to be further investigated in detail.