Stretchable fabric generates electric power from woven thermoelectric fibers

Wearable thermoelectric generators (TEGs) are sought to harvest electricity from human body heat, but conventional devices are rigid, bulky, or rely on flat 2D architectures that poorly align with the out-of-plane thermal gradient between the body and the environment. Flexible prototypes using out-of-plane gradients have been fragile or require substrates that introduce thermal shunting. The central challenge is to develop a truly wearable textile TEG that: efficiently captures heat along the thermal gradient; integrates stable, high-performance p- and n-type modules (especially n-type); offers high flexibility and conformality; provides many junctions per area for sufficient voltage; scales to large-area unobtrusive fabrics; is stretchable without performance loss; and supports intended textile insulation while facilitating heat transfer through the generator. The study presents a device-engineering solution integrating high-performance fiber-based modules and a woven 3D architecture to address these requirements.

Prior efforts include wave-like polymer substrates to orient thin-film TE materials upright; island-like inorganic or metal TE materials bonded or printed into soft fabrics; helical architectures transforming 2D films into 3D forms; and fiber/yarn-based woven or knitted TE textiles with insulating yarns or spacer fabrics to separate hot and cold sides. These solutions often suffer from limited flexibility due to inorganic materials, reliance on substrates that cause parasitic thermal losses, difficulty in large-scale garment integration, and challenges achieving conformality and stretchability without sacrificing TE performance. Woven or knitted TE textiles reported to date typically remain 2D or require 3D spacer substrates. A truly wearable, woven TEG with substrate-free 3D architecture and stable n-type organic performance had not been reported.

Materials: Carbon nanotube films (CNT films, ~900 μm width) synthesized by floating catalyst CVD were consolidated by water bath and twisted (four films) into CNT fibers (CNTF) ~280 μm diameter consisting of multiwalled CNTs (~11 nm diameter). PEDOT:PSS (PH1000) and oleamine (80–90%) were used as received.

TE fiber/module fabrication: 1) p-hybridization: CNTF was periodically dipped into PEDOT:PSS solution over segments of length (L − 4 mm)/2 (L = repeat length), then rinsed and dried (60 °C, 4 h). 2) n-doping: Using polypropylene masks covering p-hybridized and electrode segments, exposed CNTF regions were n-doped by electrospray of oleamine solution (4.2 g oleamine in 20 mL ethanol) through a 27G needle at 0.1 mL h−1; needle-to-collector distance 7 cm; 24 kV applied; RH ~35%; spinneret width 10 cm; traverse speed 1 cm s−1; duration 1 h; excess was wiped off. This created alternating p and n segments separated by 2 mm undoped conductive electrode segments. 3) Insulation: Doped CNTF was coverspun with acrylic fibers to insulate p/n legs while leaving 2 mm electrode sections exposed for thermal contact. 4) Looping/weaving: TE fibers were bent into loops (modules) with knitting needles (3 or 6 mm diameter), defining arc (electrode) and pillar (p/n legs) regions. Loops were interlocked in an alternating pattern so that the pillar of one loop is overlapped by the arc of the next, allowing loops to stand out-of-plane via elastic bending forces and form a substrate-free 3D textile where p/n legs are in series electrically and parallel thermally.

Characterization: Mechanical stability assessed by resistance change vs bending angle/cycles (<0.5% variation). TE properties measured in air at room temperature: electrical conductivity σ via four-terminal method (Keithley 2450) using silver contacts and σ = l/(R·S) with S from SEM-measured diameter (circular cross-section assumption); Seebeck coefficient α from linear fit of ΔV vs ΔT (R2 > 0.999) using a custom setup (Keithley 2182A). Power output measured with a custom system. Air stability of n-type Seebeck tracked >800 h.

Simulations (FEA, ANSYS): Modeled a loop leg contacting a 313 K hot side with the other end exposed to 298 K air. TE unit with L = 32 mm modeled as two semicircles (radius 3.25 mm) and two straight segments. Thermal conductivities: CNTF 26 W m−1 K−1; wrapping 0.051 W m−1 K−1; convective heat transfer coefficient at wrapping surface 5 W m−2 K−1; no contact thermal resistance between CNTF and wrapping. Simulations compared wrapped vs unwrapped legs for temperature distribution and heat flow.

Device testing: Compared 3D interlocked architecture vs conventional 2D interlock. Measured output voltage vs ΔT (Peltier hot/cold sources). Stretching tests: longitudinal strain up to ~80% and transverse up to ~60%; monitored real-time voltage and ΔTTE; infrared imaging captured standing angle changes. Human-body test: device with 15 units attached to an elbow to monitor voltage during movement. Optimization: Varied repeat length L (32 to 16 mm) and fiber diameter φ (3 to 1.5 mm) at fixed unfilling factor δ = L/φ = 10.6 to study voltage, internal resistance, power, and power density. Normalized power density to ΔT2 for comparison with literature.

• Stable n-type CNTF achieved via oleamine electrospray doping: Seebeck coefficient ≈ −64 μV K−1 after 1 h, stable in air >800 h; power factor ≈ 320 μW m−1 K−2 (despite slight σ decrease due to insulating coating). p-hybridized CNTF (PEDOT:PSS) showed σ increase 820→950 S cm−1, α increase, and power factor 185→330 μW m−1 K−2.
• Electrospray n-doping localized the dopant, preventing cross-contamination of p-segments and yielding ~3× higher output voltage per repeat unit compared with conventional dipping.
• Interlocked-loop 3D architecture (substrate-free) aligned TE legs with out-of-plane heat flow. For identical textiles (15 units), the 3D architecture produced ~24× higher output voltage than the 2D interlock arrangement at the same ΔT.
• Wrapping with acrylic fibers improved thermal performance: FEA showed slightly higher ΔT across TE legs and ~10× higher heat flow through the legs with wrapping versus without, while also providing electrical insulation and better hot/cold contact surfaces.
• Stretchability and conformality: Longitudinal strain >80% with minimal device deformation at the unit level (standing angle reduced from ~40° to ~20% at 80% strain). Output degradation under 80% strain was <3.2% at ΔT = 40 K; small strain (~20%) increased voltage by ~3% due to improved contact. Transverse stretching >60% increased voltage up to ~10% from larger contact area.
• On-body demonstration: A 15-unit device on a human elbow generated ~2.8 mV with ~4% amplitude modulation during bending, demonstrating compatibility with movement.
• Optimization (δ = 10.6): Reducing L from 32 to 16 mm (φ from 3 to 1.5 mm) decreased open-circuit voltage by ~26% but reduced internal resistance from ~107 Ω to ~47 Ω, increasing power and current. At ΔT = 44.4 K: power rose from 4.40 to 4.64 μW; short-circuit current increased from ~450 to ~700 μA. Power density increased 4× from ~14,000 to ~69,000 μW m−2 (≈70 mW m−2).
• Normalized performance: 35 μW m−2 K−2, surpassing previously reported flexible organic TEGs and some inorganic generators cited.
• Practical projection: A TE apparel covering ~40% of an adult’s surface (0.86 m2; ~190,000 couples in series) could deliver ~200 μW and ~20 V open-circuit at ΔT ≈ 2.6 K (26°C ambient), suitable for microwatt wearable body sensors.

The work addresses the core challenge of aligning flexible TE devices with the out-of-plane thermal gradient in wearable scenarios by engineering a substrate-free, interlocked-loop 3D textile architecture. The elastic standing of TE loops ensures thermal resistance matching and efficient heat flow through the active legs while preserving fabric-like flexibility and conformality. The integration of a stable, high-performance n-type CNTF via localized oleamine electrospray doping overcomes a common bottleneck in organic-based TEGs, enabling reliable π-type modules within a single fiber. Wrapping not only prevents electrical shorting but enhances heat exchange, increasing heat flow and ΔT across TE legs. Consequently, the 3D woven textile exhibits strongly improved voltage and power density relative to 2D arrangements, while maintaining biaxial stretchability and stable output under large strains and real body motion. These results demonstrate a viable pathway toward truly wearable thermoelectric textiles capable of continuous power harvesting from body heat, with performance sufficient for low-power sensing applications and scalable integration into garments.

The study demonstrates the first truly wearable, substrate-free 3D thermoelectric textile by weaving alternately interlocked TE loops made from alternately p-/n-doped CNT fibers. An electrospray oleamine process yields air-stable n-type segments with high power factor, while the looped, interlocked architecture aligns the TE legs with the body-to-ambient heat flow and provides intrinsic stretchability and conformality. The textile achieves a peak power density of ~70 mW m−2 at ΔT ≈ 44 K and normalized 35 μW m−2 K−2, outperforming prior flexible organic TEGs, with stable operation under up to ~80% strain and on-body movement. Future improvements include optimizing repeat length L and wrapping thickness to tune thermal resistance and internal resistance, employing ultrathin insulating wraps (e.g., electrospinning) for short L, and adopting higher-efficiency inorganic-based or hybrid fiber materials. The interlocked loop architecture is compatible with textile manufacturing, enabling large-area, unobtrusive energy-harvesting garments for powering microwatt wearable sensors.

• Electrospray n-doping is not compatible with all dopants, potentially limiting material choices for n-type segments.
• Device performance depends sensitively on geometric parameters (repeat length L, fiber diameter φ, and tightness δ = L/φ); overly small φ or long L can reduce standing angle, degrade heat flow, and weaken mechanical robustness, requiring careful optimization.
• Predicted on-body power under small body-to-ambient ΔT is in the microwatt range, suitable for low-power sensors but not higher-power electronics without further scaling or efficiency gains.