Towards Woven Fabrics with Integrated Stainless Steel-Nickel-Carbon Thermopile for Sensing and Cooling Applications

Featured Application

Protective wear such as uniforms that can monitor temperature under ballistic wear or cool their users on demand.

Abstract

Thermocouples can be combined into thermopiles to sense heat differences or achieve localized heating and cooling. However, integrating them into textiles using yarns is not straightforward, and chemical methods face challenges like complex processing, poor scalability, and voltage non-uniformity. This study employs conventional weaving to fabricate textile-based thermocouples and thermopiles for wearable sensing and potential cooling applications, with a focus on protective clothing. Using stainless steel and nickel-coated carbon yarns, we demonstrate a more stable thermocouple than those made with chemical or welded methods, with minimal fabric damage. Four conductive yarns, stainless steel, carbon fiber (CF), and nickel-coated carbon fiber (NiFC), were woven and laser-cut to form thermocouples using three different binding types to connect them. Inox1–NiFC was the most efficient thermocouple, achieving the highest Seebeck coefficient of 21.87 µV/K with Binding 3. Binding 3 also reduced contact resistance by 66% across all configurations. Slightly lower but comparable performance was seen with Inox1–NiFC/Binding 2 (21.83 µV/K) and Inox2–NiFC/Binding 1 (15.79 µV/K). In contrast, FC-based thermocouples showed significantly lower Seebeck values: 5.67 µV/K (Inox2–FC/Binding 2), 5.43 µV/K (Inox1–FC/Binding 3), and 5.06 µV/K (Inox2–FC/Binding 1). A woven thermopile with three junctions made with the optimal binding and thermocouple combination generated an average of 55.54 µV/K and about 500 µV at small temperature differences (4–5 °C), with a linear voltage response suitable for sensing. While thermal sensing proved effective, Peltier cooling needs further optimization. This method offers a stable, low-cost, and scalable platform for textile-integrated thermoelectric systems, with strong potential for use in uniforms and other protective garments.

1. Introduction

Wearable technology as a part of our daily clothing can introduce functions like bio-sensing and thermoregulation without being intrusive. Textile-based thermocouples have opened doors for advanced fabrics for energy harvesting, sensing (Seebeck effect), and thermoregulation (Peltier effect), supporting the development of sustainable, self-powered wearables [1]. Ongoing research in sensor technology is well defined, aiming at seamless integration and the use of materials that can withstand harsh conditions such as extreme temperature and humidity changes, washing, and abrasion. The performance of thermoelectric materials is commonly evaluated using a dimensionless figure of merit, ZT, defined in Equation (1) [2]:

$$\mathrm{ZT}={\displaystyle \frac{S^2\sigma T}{K}}$$

(1)

where S is the Seebeck coefficient (µV/K), σ is the electrical conductivity (S/m), T the absolute temperature (K), and K is the thermal conductivity (W/m K). The key material thermoelectric parameters are the Seebeck coefficient, the electrical conductivity, and thermal conductivity [2]. A ZT > 1 indicates great efficiency in converting heat into electricity, while a ZT ≥ 2–3 is ideal for practical thermoelectric applications [3]. The Seebeck coefficient is a material property expressing the generated voltage per 1 Kelvin temperature gradient if the material is combined with platinum [3]. The Peltier effect is the counterpart of the Seebeck effect, and its strength is directly influenced by the Seebeck coefficient.

There is a limited amount of research for textile-based thermoelectric cooling compared to power generation and sensing [4]. Flexible textile-based thermoelectric coolers can offer the benefits of active cooling with improved skin contact, comfort, compactness, and noiseless operation and should ideally be fully integrated into garments without exterior structures such as heat sinks [5,6,7,8,9]. Cooling efficiency is influenced by both material properties and fabric geometry [2]. High Seebeck coefficients, high electrical conductivity, and low internal and maximized interfacial skin-garment-ambient thermal conductivity will result in better performance. Further optimization includes adjusting thermocouple aspect ratios (yarn length/diameter) based on the thermal gradient, increasing junction density (junctions per m²), and improving separation of hot and cold sides (e.g., by reducing support yarn crimp). Shorter yarns lower thermal resistance, reducing cooling. It is important to minimize Joule heating within the fabric [10]. Applying higher current in cases of good electrical conductivity, low contact resistance, and high junction density will boost Peltier cooling [2]. At low current and low density, parasitic heat transfer outweighs the Peltier effect. Integrated heat sinks can significantly enhance heat dissipation and temperature gradients [10]. A thermal resistance > 117.6 cm² K/W has been determined to predict effective cooling and power generation [10]. Promising new thermoelectric materials are under development [10,11,12,13]. Zhang et al. [14] created flexible semiconducting fibers coated with borosilicate glass, demonstrating high thermoelectric properties and 6.2 °C cooling when woven into textiles. Zheng et al. [15] developed fiber-based thermoelectric textiles using inorganic thermoelectric materials and liquid metal, showing a stable 3.1 °C cooling, excellent stretchability, and flexibility. Newby et al. [6] have recently introduced a thermoelectric cooling garment using seamless knitting technology that operates without a heat sink. The double jersey knit with copper and constantan wires (0.17 mm diameter) as conductive pathways enabled Seebeck and Peltier effects. At 150 mA, it could cool the body by up to 1.0 °C. Jing et al. [10] recently developed an active thermoelectric textile that achieved a stable body surface cooling of 11.8 °C with heat sinks (4.2 °C without heat sinks) at 34 °C ambient temperature, powered by solar radiation and outperforming many existing passive and active systems. It could power a phone at ΔT = 15 °C and generated 6.13 W/m² and 150 mA current at ΔT = 25 °C. The double plain-woven fabric featured pockets with rigid Bi₂Te₃ pillars, copper wire weft electrodes forming TE junctions, and thermally conductive filaments to enhance heat dissipation. However, materials like Bismuth telluride, which have very ZTs near room temperature, have some toxicity concerns and are of low abundance [16,17].

A recent overview, indicating the strong growth of research in thermoelectric fabrics, can be found in [18]. Different conductive materials (metals, conductive polymers), thermocouple geometries and shapes, and fabric integration techniques (embroidery, coating, printing) have been explored [19]. A lot of materials with high ZT have been developed, but the main research challenges seem to be garment integration, quick thermal equilibrium with human skin, mechanical and thermal stability, flexibility, durability, and simple, low-cost, and scalable production [20,21]. Since traditional methods (weaving, knitting, and embroidery) alone have not always provided the optimal contact for thermocouple production, alternatives like conductive glues, soldering, and thermocompression welding have been explored [22]. However, these techniques have their drawbacks. Chemical methods can lead to non-uniformity [3], instability, complex production, and the potential introduction of expensive or toxic materials to the fabric. Soldered fabrics cannot withstand bending, while conductive glues increase stiffness [18].

This work is a continuation of a previous study by Hardianto [3], where chemical methods were used to fabricate textile thermocouples. We revisit conventional weaving to fabricate thermocouples by exploring optimal yarn binding at thermocouple connection points. The aim is to create a wearable, stable thermoelectric device for on-body temperature sensing and potential cooling, which is essential in protective clothing. Weaving has been explored in previous research and provides a fast and low-cost production technique for thermopiles. Combinations of stainless steel, carbon (CF), and nickel-carbon fiber (NiFC) yarns were explored, directly woven and laser-cut at specific points to create thermocouples in a left–right orientation. Through linear and contact resistance measurements, thermocouple and thermopile Seebeck coefficient determination, and thermal measurements, the Seebeck and Peltier effects of the constructed thermopiles were investigated. An optimal binding method was identified that significantly enhanced the Seebeck coefficient by improving the connection between thermocouple components through weaving. Yarn combinations also influenced the performance, though to a lesser extent than the binding technique. The use of Ni-FC over FC improved results due to lower electrical resistance, and changing the type of stainless-steel yarn reduced contact resistance at the thermocouple connections, further boosting performance. Overall, the proposed technique showed a promising Seebeck effect and allows for an upscalable, stable, low-cost thermoelectric fabric production and a simple garment integration. For garment integration, the hot and cold junctions can be separated with a special construction at the seam (illustrated in Figure 1) to move the left side to the bottom of a thicker fabric and the right side to the top, thereby increasing internal thermal resistance and maintaining large temperature gradients. The cooling Peltier effect needs further development, mainly to reduce excess heating at stainless steel–stainless steel contacts.

2. Materials and Methods

2.1. Conductive and Support Yarns

Four different yarns were considered based on their weavability and applicability in a thermocouple and as a continuation of the work in [3]. Their details, based on producer data, are provided in

Table 1. Conductive yarn details.

Code	1—Inox1	2—Inox2	3—FC	4—NiFC
Material	100% Stainless Steel	100% Stainless Steel	100% Carbon	Nickel-coated Carbon
Producer	Bekinox VN12-4x275-100, Bekaert	Bekinox fibers, Bekaert	Tenax EHTA40 E13 3K200 tex 15Z, Toho Tenax	Tenax JHTS40 A23 12K 1420 tex MC, Toho Tenax
Filament number	275	/	3000	12,000
Ply number	4	1	1	1
Filament diameter (µm)	12	/	7	7.5
Yarn diameter (µm)	398	227	383	821.6
Yarn count (tex)	1000	93.7	200	1420
Turns per meter (t/m)	100 S	307 S	15 Z	0

, and optical images are shown in Figure 2. The stainless-steel yarns were used in warp and weft, while the carbon-based fibers can only be used in the weft, as they tend to break under tension.

An orange wool/polyacrylonitrile (50/50) yarn (Wo/PAN) and a white 100% viscose yarn were used as weaving support yarns (Figure 3). The WO/PAN yarn has a count of 70.6 tex and a yarn diameter of 705 µm and is composed of two single yarns, each measuring 35.3 tex. It is a 2-ply yarn, with a ply twist of 212 t/m and a single yarn twist of 401 t/m. The 100% viscose yarn has a count of 75.02 tex and a diameter of 402 µm and is a 2-ply yarn. The ply twist is 230 t/m, while the single yarn twist is 428 t/m. The viscose yarn has a smaller diameter than Wo/PAN yarns despite their similar fineness values because of the higher density of viscose compared to wool and acrylic.

2.2. Weave Bindings

Different bindings were created to connect the different conductive yarns. Fabrics were woven on a manual treadle loom with a foot-controlled harnesses and hand-operated shuttle. Weave patterns were software-programmed with the selection of harnesses PC-controlled, and the rest of the weaving was performed manually. The weaving schemes were designed using DB-WEAVE 5, Brunold Software, Wädenswil, Switzerland.

2.2.1. Binding 1

For Binding 1, warp threads (yarns) were arranged in the following formula: 24A-2B-46A-2B-24A, with 98 threads in total. In the formula, letters represent different yarns: A—nonconductive Wo/PAN yarn and B—conductive yarn (Inox2, FC, or NiFC). Weft threads followed a 24A-1B-5A arrangement. Alternatively, to make one thermocouple, 24A-1B-29A-1C-5A was used in the weft, with B and C representing two different conductive yarns. The layout of this binding is provided in Figure 4.

2.2.2. Binding 2

In Binding 2, represented in Figure 5, the fabric was constructed using a double-face warp pattern with zone-dependent yarn distribution. The warp followed the formula 1A-62B-1C-144B-1C-62B-1A, with 272 warp threads in total, where A represents Wo/PAN, B viscose, and C Inox2 stainless steel. The result is a symmetrical warp layout dominated by viscose and segmented by narrow stainless-steel inserts. The weft sequence consisted of 27A-1C-44A-1B-17A, with A corresponding to Wo/PAN, B to Inox2, and C to either FC or NiFC conductive yarns.

The fabric density of Binding 2 in the warp is 24 yarns/cm. The density of Binding 2 in the warp is 24 yarns/cm, which corresponds to approximately 61 picks per inch (PPI), while Binding 1 had 12 yarns/cm (30 PPI), which is why a finer viscose yarn was used in Binding 2 for the warp construction, while Binding 1 contained only the thicker WO/PAN non-conductive yarn. The difference between the double-face warp in Binding 2 and a plain weave (construction approach used in Binding 1), is shown in Figure 6 (with viscose yarn used for Binding 1 only to help visualize the construction difference), while the resulting fabrics are provided in Figure 7.

2.2.3. Binding 3

For the third binding type, the Binding 2 formula is slightly adapted to 17A-1B-34A-1C-17A, and Inox1 is used instead of Inox2. The result of Binding 3 is shown in Figure 8.

2.3. Thermopile Construction

A thermopile with Binding 3 was created by cutting the stainless-steel warp thread at specific points, similar to [10,23], either manually or using a laser-cutter. The cutting points are indicated with gray tape in Figure 9, while the full conductive path is shown in yellow. Two thermopiles were created with Binding 3: one with NiFC and one with FC.

2.4. Linear and Contact Resistance Measurements

The linear resistance of the conductive yarns, R_yarn, was measured over a length of 20 and 40 cm, with the four-point measurement [3,24], using a Fluke 87V Industrial Multimeter (Fluke Corporation, Everett, WA, USA). Measurements were read directly from the device’s digital display. The yarns are stretched over the given lengths using a clamping device; two outer probes inject current while two inner probes measure the voltage output. From the current–voltage relationship, the linear resistance (Ω/cm) is calculated for each length. The contact resistance between two conductive yarns in the constructed fabrics, R_contact (Ω), was determined using the measurement setup shown in Figure 10 and Ohm’s Law for calculations (Equation (2)), where the measured linear resistance of the yarns is subtracted from the result:

$$R_{contact}={\displaystyle \frac{U_{meas}}{I_{meas}}}-{(R}_{warp yarn}\times L_{warp yarn}+R_{weft yarn}\times L_{weft yarn})$$

(2)

where U_meas (V) represents the measured voltage, I_meas (A) is the measured current, R (Ω) represents measured yarn resistances, and L (cm) represents the length of yarns considering the weave take-up.

2.5. Thermocouple and Thermopile Seebeck Coefficient

The Seebeck coefficient of a thermocouple was determined with hot plate measurements (Isotemp Advanced Hot Plate Stirrer by Fischer Scientific, Waltham, MA, USA) by measuring the generated voltage at different temperature gradients. The zone where the two conductive yarns cross for each fabric binding was placed on the hot plate, with close contact achieved by putting weight on top (as shown in Figure 11). The other yarn ends were placed away from the hot plate, in the air, creating a cold junction, and were connected to a nanovoltmeter (NV724 by Setaram, Lyon, France) (Figure 11, left). The nanovoltmeter was used to obtain the generated voltage, and a thermometer (54 II-B by Fluke, Everett, WA, USA) was used to measure the temperature of the hot and cold junctions. The room temperature was 23 °C, and the hot plate was heated to a maximum of 80 °C; data were recorded every 2 min as the hot plate gradually heated. The temperature difference was read directly from the thermometer. Three repetitions were made for each yarn combination.

To determine the Seebeck coefficient of the thermopile, half of the fabric was placed on the hot plate, while the other half was sticking out (Figure 11, right). The temperature was measured on the hot side of the fabric, which was in direct contact with the hot plate, while the cold side was exposed to air (see Figure 11, right). The resulting electrical potential (voltage) was, again, measured with the nanovoltmeter.

A variant of this test was used to evaluate the sensing power of the final thermopile. In this variant, the sample was placed on the hot plate, and the temperature increased in steps of 10 °C, starting at ambient and rising to 140 °C. In this case, the electrical potential can be measured with a Fluke 87V Industrial Multimeter.

2.6. Measuring the Peltier Cooling Effect

The Peltier effect is the counterpart of the Seebeck effect. To assess the Peltier effect in the thermopile, a direct current of 0.35 mA was applied across the constructed fabric thermopiles containing three thermocouples. Temperature measurements were taken at the top and bottom surfaces of the thermopile relative to the ambient background using contact sensors. Then, the current was gradually increased to identify the threshold at which temperature changes became detectable. To further improve spatial resolution and observe localized temperature variations, an infrared (IR) camera (FLIR-T62101 by FLIR, Täby, Sweden) was also used. The current path and thermocouple locations were marked to aid interpretation of the thermal response.

Embroidery of a highly conductive yarn at the contact points was further used to improve the resistance at contact points, using a PES multifilament/Cu-Ag wire (HI-COND CA 74 by CleverTex, Ústí nad Orlicí, Czech Republic), with a linear resistance of 2.80 Ω/m.

3. Results

The obtained results are presented in the following subsections.

3.1. Electrical Linear Resistance

The conductive yarn resistances obtained from the four point measurements are provided in

Table 2. Electrical linear resistance of the yarns.

Yarn	Length [cm]	Electrical Linear Resistance [Ω/m]	CV * [%]
Inox1	20	8.89	16.2
	40	7.43	6.8
Inox2	20	90.16	17.0
	40	89.00	12.8
FC	20	155.9	6.3
	40	145.8	8.6
NiFC	20	1.50	16.8
	40	1.44	3.6

* CV = Coefficient of variation for the electrical linear resistance measurements.

.

As expected for yarns, on shorter distances, a large variation in linear resistance can often be observed with CV > 15%, which reduces for longer lengths. Inox1 has a much lower electrical resistance than Inox2. As mentioned in the literature, for cooling applications, thermocouples with a low electrical resistance are preferred [2]. Based on this, yarns Inox1 and NiFC are an optimal combination. Though Inox1 showed significant non-uniformity in linear resistance at different lengths, with a t-test between 20 cm and 40 cm yielding a p-value of 0.007, since the overall resistance remains below 10 Ω/m, this variation is unlikely to affect the potential cooling function of constructed fabrics.

3.2. Thermocouple Contact Resistance

Contact resistance is expected to vary depending on the binding type used and the yarn materials in contact. Ideally, it should be so low that it is effectively 0 Ω to ensure consistent thermoelectric performance with maximum electrical efficiency and to avoid Joule heating at contact points. The results are presented in

Table 3. Computed contact resistance depending on the binding type.

Contact	Binding	Resistance [Ω]	CV * [%]	Improvement over Previous Binding
Inox2–Inox2	1	17.44	34.0	/
Inox2–Inox2	2	12.05	12.4	−30.9%
Inox1–Inox1	3	4.07	36.3	−66.2%
Inox2–NiFC	1	5.52	45.4	/
Inox2–NiFC	2	3.63	1.2	−34.2%
Inox1–NiFC	3	1.20	5.7	−66.9%
Inox2–FC	1	24.67	10.7	/
Inox2–FC	2	26.47	3.0	+7.3%
Inox1–FC	3	8.89	6.7	−66.4%

* CV = Coefficient of variation for the contact electrical resistance measurements.

.

As can be observed, changing the binding has a dramatic impact on contact resistance. Switching from Binding 1 to Binding 2 across different yarn combinations results in improvements of up to 34%. However, the Inox2–FC contact does not improve, showing that changing the binding type alone does not guarantee better performance. Using Inox1 instead of Inox2, along with a slight modification in binding (Binding 3), leads to further reductions in contact resistance, exceeding 66% across all three yarn combinations (Inox1–Inox1, Inox1–NiFC, and Inox1–FC). We conclude that using Inox1 instead of Inox2, combined with Binding type 3, yields the lowest contact resistances.

3.3. Thermocouple Seebeck Coefficient

The obtained Seebeck coefficients for different thermocouples (binding and yarn combinations) are provided in

Table 4. Obtained Seebeck coefficients and the reference values found in [3].

Contact	Binding	Seebeck Coef. [μV/K]	Change with Ref.
Inox1–NiFC	Ref. *	25.30	/
Inox2–NiFC	1	15.79	−37.6%
Inox2–NiFC	2	21.83	−13.7%
Inox1–NiFC	3	21.87	−13.6%
Inox1–FC	Ref. *	5.06	/
Inox2–FC	1	4.88	−3.6%
Inox2–FC	2	5.67	+12.1%
Inox1–FC	3	5.43	+7.3%

* Ref. = Reference contact resistance value for the different conductive yarn combinations (Inox1–NiFC or Inox1–FC), when the yarns were placed over each other without direct integration in a fabric.

. The change in Seebeck coefficient is calculated relative to a reference for Inox1–NiFC or an Inox1–FC yarn combination taken from Hardianto’s work [3], where the thermopower (µV/K) of various yarn pairs was measured to identify suitable thermocouples. In his work, the yarns were not interlaced into a fabric but simply crossed at one end on a hot plate to form the hot junction, with the other ends positioned away from the heat source. Additionally, Figure 12 shows the resulting slopes from the generated voltage as a function of the temperature difference between the hot and cold junctions of the thermocouples.

These results clearly show that achieving the reference values depends on minimizing contact resistance through effective binding methods, with Binding 2 and the similar Binding 3 outperforming Binding 1. We can also see that changing from Inox2 to Inox1 has a limited effect on the output (below 5%), especially for sensor applications. Still, choosing yarns with lower contact resistances (Table 3) should help reduce Joule heating when they are used to make a cooling device instead of a sensor. We can conclude that weaving can be used to create a stable thermocouple that will match reference values even though the materials have an interlacing contact.

3.4. Thermopile Efficiency

3.4.1. Thermopile Seebeck Coefficient

In theory, when properly connected, the Seebeck coefficient of a thermopile should be equal to the Seebeck coefficient of a single thermocouple multiplied by the number of thermocouples connected in series. However, literature shows that achieving this is challenging. In particular, chemical methods for thermocouple creation are likely to cause inconsistencies at each junction in the Seebeck coefficient, resulting in degraded thermopile performance, with single junction Seebeck coefficients varying from 3 μV/K to 16.5 μV/K [3]. In our case, in the thermopile, there are extra shifts from warp to weft to maintain the conductive path (Figure 9), apart from the shift at the thermocouple junction, which could lead to degraded Seebeck performance. These additional shifts may disturb the electrical contact, introduce variations in electrical resistance, and affect the consistency of the junctions due to imperfect contact, mechanical stress, or other disturbances. The results are provided in

Table 5. Obtained Seebeck coefficients, converted to single junction values, and differences from the values in Table 4.

Contact	Binding	Thermocouples per Thermopile	Thermopile Seebeck Coef. [μV/K]	Single Junction Seebeck Coef. [μV/K]	Difference
Inox1–NiFC	3	3	55.54	18.51	−15.4%
Inox1–FC	3	3	16.79	5.6	+3.1%

and Figure 13. Here, only three junctions are used, as the start junction was damaged during sample cutting.

In this study, two thermopile constructions were created with three thermocouples (Inox1–NiFC or Inox1–FC) per thermopile and with Binding 3. Thus, the Seebeck coefficients of these thermopiles must be divided by three to obtain the average coefficients of each thermocouple. When comparing these average values to those determined previously (Table 4), it can be observed that the Seebeck coefficient for the Inox1–NiFC thermocouple decreased by 15.4%, while that of the Inox1–FC thermocouple increased by 3.1%, as measured in the thermopile. The experimental results indicate that adding each thermocouple leads to an almost proportional increase in the total Seebeck coefficient. Even if individual thermocouples vary slightly in performance, the total output improves when multiple thermocouples are combined into a thermopile. In Figure 13, we can see that with a small temperature difference, such as one created by human touch and ambient room temperature (4–5 °C), the thermopile can generate around 500 µV output. This suggests that the thermopile design is effective and supports the potential for integrating Seebeck sensor elements into textiles in a reliable and scalable way.

3.4.2. Peltier Cooling Effect

The Peltier cooling effect is directly proportional to the applied current [2]. However, high electrical current can lead to Joule heating. Since the thermopile developed includes only three thermocouples spaced apart, the resulting Peltier effect and any associated heating are expected to be minimal. A current of 0.35 mA was applied to the created thermopile, and the surface temperatures were measured. However, no difference from the background temperature of 21.8 °C could be detected. Next, the current was increased up to a point where differences from the background can be obtained. At 1.45 mA, an increase of 0.1 °C at both the top and bottom of the thermopile, compared to the background temperature, was detected. No cooling effect was detected, which may be due to the limitations in measurement sensitivity. With only three junctions spaced relatively far apart from each other, the resulting temperature differences could be too small to be captured for the available equipment. Still, some differences in the top and bottom of the thermopile would have been expected, so the effect was further investigated with an IR camera. The current was increased further to obtain clear differences from the background temperature, and the result is provided in Figure 14. Here, the hotspots shown were 29.9 °C, while the background was 22.8 °C. In the figure, we have added the path of the current and the thermocouple locations. From this, we can observe two problems. First, the stainless steel–NiCF junctions have a good contact (no hot points), but the stainless-steel Inox1–Inox1 junctions have a problem leading to excess heating (indicated by cyan circles). From Table 3, we know the Inox1–Inox1 contact resistance of 4.07 Ω is almost four times higher than the Inox1–NiCF contact resistance, which seems to be causing this problem. This measured contact resistance near the stainless-steel junction generates enough Joule heating to cancel out any potential cooling effects from the Peltier mechanism. Also, two of the cuts in the stainless steel led to excess heating (indicated by magenta circles). This is likely due to different strands of the stainless-steel yarn being at slightly different voltages at the cut ends, which creates local currents and results in Joule heating. So, the stainless-steel junctions also require a special weave binding to make them stronger, and future research is needed to understand why some cuts lead to excess heating and others do not.

In the literature, several techniques are described to lower contact resistance, one of which is embroidery of a highly conductive yarn at the contact points. We tried this method to improve the thermopile, using Binding 3 and a PES multifilament/Cu-Ag wire, with a linear resistance of 2.80 Ω/m. The result is shown in Figure 15, where reduced Joule heating can be observed and the problem at the cuts in the stainless steel seems to be resolved. However, there are still hot spots at the hot side for all Inox1–Inox1 contact points, and no detectable cooling over the background is observed before Joule heating dominates at the cold side.

3.4.3. Seebeck Sensor Element

The thermopile Inox1–NiFC with Binding 3 was used to verify its application as a sensor element. For this, a hot plate was used from ambient temperature (27 °C) up to 140 °C in steps of 10 °C. The results from three repetitions are shown in Figure 16. An excellent linear progression is observed, with a set temperature difference leading to a fixed voltage output up to a 40 °C difference between the hot and cold sides. After this, scatter starts to occur, while at 140 °C (more than 100 °C over ambient), the linear relation starts to break down. The observed slope corresponds with the thermopile Seebeck coefficient provided in Table 5 and shows how the sample can directly be used for sensing.

4. Discussion

A big challenge that the current research on thermoelectric textiles is facing seems to be producing a low-cost, upscalable, simple thermopile with stable, effective thermoelectric properties and good wearability. In this study, thermocouples placed in series were directly woven and laser-cut at specific points to create thermocouples in a left–right orientation, as in [19]. They can be easily integrated into garments by separating the hot and cold junctions with a special construction at the seam (illustrated in Figure 1), as in [20]. The produced thermoelectric device is 100% textile-based, suitable for wearable applications, and made of commercially available conductive metal yarns directly woven into the fabric. For this reason, it is more advantageous than other techniques like screen printing, wire soldering, electroconductive glues, and interwoven metal wires, which often result in stiff fabrics [18]. The findings show the critical role of material selection and binding techniques in developing efficient thermoelectric sensing and cooling woven textiles. Also, possible reasons for the inefficiency of the Peltier effect were identified in the constructed thermopile. Thermoelectric textile sensors have strong practical potential for protective clothing, especially in military use where a real-time detection of heat stress or fatigue is essential to quickly detect risks and provide valuable data for research and long-term health monitoring. Promising results for a wearable temperature sensor were obtained.

For temperature sensing applications, a higher Seebeck coefficient is proportional to the resulting sensitivity to temperature differences: the higher the amount of voltage generated per Kelvin, the better. Among the tested yarns, Inox1 (stainless steel) and NiFC demonstrated the lowest linear resistances, a low contact resistance, and the highest Seebeck values. The Seebeck coefficients for the NiFC and FC thermocouples aligned closely with the reference values from fabrics made with similar materials but different production techniques [3], indicating that woven structures could give similar results while also allowing for stable thermopile construction. Contact resistance measurements revealed the significant impact of binding methods, with Binding 2 and 3 leading to lower contact resistance. Also, shifting from Inox2 to the lower-resistance Inox1 yarn further decreased the contact resistance. This is in contrast with the research by Radulescu et al. [19], where higher-resistance yarns (silver/stainless steel) achieved better voltage generation than lower-resistance yarns (silver-copper/silver) in the same fabric configuration, which could be related to the different nature of the raw materials. Similar to our study, the researchers used metallic yarns as thermocouples sewn in a left–right orientation, achieving 160 μV at a 65 °C gradient. The lower voltage output compared to our results may be due not only to material choice but also to their smaller yarn diameter and linear density (Table 1). Our fabric thermopile with three junctions achieved an average of 55.54 µV/K. In

Table 6. Comparison of Seebeck coefficients and output voltages of various textile-based thermoelectric devices.

Thermocouple Type	Thermocouple Number	Production Method	Seebeck Coef. (µV/K)	Output Voltage (µV) at ∆T (K)	Ref.
Stainless-steel/NiCF yarn thermocouples	3	Conductive yarns on a woven substrate, woven with left–right orientation.	21.87 µV/K per single junction; 55.54 µV/K per thermopile	500 µV at 4 K; 1250 µV at 15 K	This work
Silver/stainless-steel yarn thermocouples	5	Metallic yarns sewn in a left–right orientation on a cotton textile substrate.	~2.7 μV/K per thermopile (ΔT = 20–60 K)	160 μV at 65 K	[19]
NiFC yarn thermocouples	150	Electrodeposited Ni on carbon fiber; sewn on PES fabric; top–bottom orientation of hot/cold junctions.	11.61 μV/K per junction; 1219 μV/K per thermopile	~40 μV at 4 K; ~180 μV at 15 K	[29]
PEDOT-Cl (p-type) films/carbon fiber yarns (n-type)	2	PEDOT-Cl vapor printed rectangles (45 × 5 mm) on cotton with 5 mm spacing; carbon fibers sewn across legs; silver-coated nylon used at junctions.	16 μV/K per junction	1200 μV at 30 K	[20]
Stainless steel (Inox1)/NiFC yarn thermocouples	1	Nickel-coated carbon fiber and stainless-steel yarns, just crossed over each other (not woven).	25.30 μV/K per junction	Not given	[3]
Silver selenide and polyvinylpyrrolidone films	6	Screen-printed Ag2Se/PVP films (40 × 5 mm) on PI mesh, spaced 10 mm apart and connected with conductive silver adhesive.	56–58.5 μV/K per thermopile (at T = 27–117 °C); 36.0 μV/K per thermopile (at T = 137 °C)	1700 µV at 4 K; 7300 µV at 15 K; 11,100 µV at 20 K; 21,600 µV at 40 K	[27]
CNT/PLA composite nanofiber web thermocouples	4	CNT electrosprayed onto electrospun PLA; four CNT/PLA units connected in series with silver paste on electrospun PLA fabric.	62.9 μV/K per thermopile	Not given	[30]
Indium oxide/indium tin oxide films	Not given	Screen printing on flexible polyimide substrate.	160–180 μV/K per thermopile	Not given	[31]
MWCNT (-p and -n type) conductive yarn thermocouples	1	MWCNT conductive yarns woven with planar weave and polyester support yarns.	138.95 μV/K per single junction	Not given	[2]
PEDOT:PSS silk yarn/ silver-plated polyamide embroidery yarn	8	3D fabric with PEDOT:PSS silk yarns sewn on a nine-layer wool felt, with silver paste printed at thermocouple connections.	~85–115 μV/K per thermocouple (estimated based on T = 20–60 K)	~5800 μV at 60 K; ~1800 μV at 20 K	[32]
CNT yarns	11–12 (per cm2 fabric)	Doped CNT yarns on a 3D spacer fabric	870 µV/K per thermopile	Not given	[33]

, we compared the achieved Seebeck coefficient with those from other recent studies of textile-based thermocouples. While lower than some of the recently developed devices, it remains promising for wearable applications due to its simple, low-cost production, scalability, and linear voltage response to temperature. Landsiedel et al. [25] mention that a thermoelectric coefficient of 3–4 μV/K is sufficient for manufacturing textile-based temperature measurement devices. A range of 160–180 μV/K achieved by recent studies has been considered high thermal sensitivity for textile thermoelectric devices [26]. Liu et al. [27] developed a thermoelectric wearable temperature sensor based on silver selenide and polyvinylpyrrolidone, achieving voltages of 1700 µV and 7300 µV at temperature differences of 4 °C and 15 °C, respectively. In our study, we managed to achieve around 500 and 1250 µV at the same temperature differences, respectively, which is less, yet still promising. Further research should address additional requirements for skin temperature sensors, such as those set by ASTM E1112-00 (2018) [28], which sets electronic thermometer accuracy at 0.1–0.3 °C for reliable patient temperature measurement.

The results showed challenges in achieving effective cooling with the given thermopile design. The Peltier effect was hindered by Joule heating, particularly at specific weave points. This was observed from the distinct hotspots of 29.9 °C appearing on the fabric at higher voltages, highlighting two issues: too high contact resistance at Inox1–Inox1 junctions and inconsistent behavior at cut points at the stainless-steel yarn, likely caused by voltage differences between the individual strands. Further research is needed to understand why some cuts lead to excess heating and others do not. All junctions in a woven fabric require low contact resistances, not only the thermocouple junctions, and cuts in the conductive yarns need to be strictly controlled. Thus, further optimization of the weaving technique is needed to strengthen the stainless-steel junctions and control current flow at cut yarn ends. Approaches include adding embroidered zones at junctions and adding conductive glue/paste/inks at junctions; however, these will, again, increase the cost and complexity of the design. In this study, we tried adding embroidery at the intersections, which resulted in some improvement (reduction of hot spots); still, no cooling was observed. The contact resistance between the connection points of the stainless steel–stainless steel connecting yarns have 4.07 Ω, causing Joule heating that can cancel out any potential cooling effect. Increasing the junction density in our thermopile could promote the Peltier effect. In comparison with Newby [6], we expect no cooling with only three junctions in our sample, since they achieved only 1 degree of cooling with 85 junctions. Considering this, detecting potential cooling would be below the resolution of the IR camera, especially taking into account the influence of emissivity differences in materials on camera readings. A minimum of 25 junctions are needed for a cooling of 0.3 °C to be achieved [6] in a relatively small sample. A woven fabric prototype developed by Chatterjee [2], with MWCNT conductive yarns woven with planar weave and polyester support yarns, demonstrated cooling up to 1.3 °C, with a 3 °C heating on the hot side under 6 mA current, highlighting that increased junction density will significantly enhance cooling. However, they used film tapes instead of yarns to create the prototype. The significantly lower Seebeck values in our configuration may explain the absence of observable Peltier cooling. The Seebeck coefficient of the MWCNT junction reported by Chatterjee was 138.95 μV/K, with a contact resistance of 80.24 Ω. In comparison, our Inox1–NiFC junction showed a Seebeck value of 21.87 μV/K (Table 4), with a contact resistance of 1.20 Ω at thermocouple bindings and 4 Ω at Inox1–Inox1 contacts. While we chose our materials based on previous work by Hardianto [3], the weaving approach presented in this study can be further explored with other materials with higher Seebeck coefficients, such as Bi₂Tu₃. Using thinner conductive yarns with less fraying that can be woven closer together, or investigating further binding variations and producing a larger sample, could show the cooling effect. Improving the stainless-steel yarn is another option, as the shift from Inox2 to Inox1 yarn in the design already resulted in a large improvement, indicating that yarn morphology plays an important role. As previously shown in Figure 1, the garment integration for cooling purposes was planned as illustrated.

5. Conclusions

This paper revisits conventional weaving as a method to fabricate thermocouples and thermopiles by exploring optimal yarn bindings and four conductive metallic yarns. Weaving was used to produce thermopiles, with thermocouples placed in series for sensing and potential cooling applications. This study aimed to provide a scalable, simple approach for developing textile-based thermoelectric devices suitable for on-body use, particularly in protective clothing. Combinations of stainless steel, carbon fiber (CF), and nickel-coated carbon fiber (NiFC) yarns were directly woven and laser-cut to form thermocouples in a left–right orientation. The stainless steel–NiFC pairing demonstrated greater thermocouple stability than the chemical methods from the previous work used as a reference for this research, with minimal fabric damage compared to welding. Optimizing the interlacing between yarn junctions (Binding 3) led to a 66% reduction in contact resistance across all yarn combinations, showing that proper weave construction can significantly improve electrical performance. The fabric thermopile with three junctions achieved an average of 55.54 µV/K. The voltage response showed good linearity, confirming suitability for sensor applications. This demonstrates a promising approach for creating stable, scalable, and low-cost textile-integrated temperature sensors that can be incorporated into garments through a special seam construction separating hot and cold junctions. Although the sensing performance was promising, further development is needed to achieve the Peltier cooling effect, particularly by addressing excess heating at stainless steel–stainless steel contacts. While embroidery showed some improvement in reducing contact resistance, no cooling was observed with IR imaging. Future work should explore higher junction densities, larger samples, and yarns with higher Seebeck coefficients to enhance cooling performance.