Thermoelectric generators operate on the principle of converting temperature gradients into electric voltage. Their appeal in the domain of wearable electronics is immense, promising sustainable, battery-free power sources integrated seamlessly into clothing or affixed to the skin. Thin-film thermoelectric devices, in particular, present an opportunity for comfort and flexibility. However, to date, the field has wrestled with an inherent contradiction: the very thinness that allows for flexibility simultaneously permits heat to escape vertically with ease, equalizing the temperature on both sides of the device and crippling its ability to generate electricity effectively.

Conventional approaches to overcoming this fundamental issue have included bending the thermoelectric films or fabricating complex three-dimensional microstructures, such as pillar-like arrays. While improving temperature differential retention to some extent, these solutions invariably increase device bulk, negating the essential benefits of wearability—namely lightweight, low-profile design and user comfort. Addressing this impasse, the SNU research team embarked on a fundamentally new direction.

The cornerstone of their innovation lies in designing a dual thermal conductivity substrate — a composite formed by integrating copper nanoparticles selectively into specified regions of a stretchable polydimethylsiloxane (PDMS) silicone matrix. This engineering creates discrete zones within a single planar substrate characterized by starkly contrasting levels of thermal conductivity. Unlike traditional substrates where heat dissipates directly upward through a uniform medium, the engineered substrate guides heat laterally along the path of high thermal conductivity created by copper nanoparticle inclusion.

When thin-film thermoelectric semiconductor elements are strategically positioned along the interface between these high and low thermal conductivity regions, the system encourages body heat from the skin to flow horizontally. This produces distinct warm and cool zones across the planar surface, fostering a robust temperature gradient essential for electricity generation in a thin-film configuration that remains perfectly flat.

This pseudo-transverse thermoelectric effect, conceptually inspired by classical transverse thermoelectric phenomena, had not been realized in solution-processed, flexible formats before. By mimicking transverse heat flow structurally within a fully planar and thin device, the team has sidestepped the conventional trade-off between device thickness and thermoelectric efficacy. Crucially, their process is compatible with scalable, all-solution-based inkjet printing techniques, ensuring that the generator can be mass-produced, patterned in diverse shapes, and seamlessly integrated into wearable textiles or skin sensors.

Beyond its elegant physics and materials science ingenuity, this wearable pseudo-transverse thermoelectric generator boasts practical implications for next-generation electronics. Given its low-profile design and mechanical flexibility, it can act as a self-sustaining power source for an array of applications such as smart garments that monitor biometric data, health tracking sensors, and other skin-mounted electronic devices. These capabilities open doors not only to longer-lasting wearables but also potentially to entirely new classes of self-powered, maintenance-free electronics.

Professor Kwak emphasizes that the novelty of their work stems from controlling heat flow within a planar geometry, overcoming a long-standing barrier in wearable thermoelectric technology. The ability to generate electricity without resorting to bulky 3D structuring or device deformation marks a transformative step forward. He envisions their solution as a foundational platform that will link wearable electronics with sustainable, continuous power harvesting directly from the human body.

Co-first authors Dr. Juhyung Park and Dr. Sun Hong Kim were instrumental in realizing this concept, leveraging expertise in organic electronic materials and nanoscale fabrication. Their multidisciplinary approach, from fundamental material design to device-level integration, exemplifies the collaborative spirit crucial to technological breakthroughs. Notably, Dr. Park has continued probing organic electronic applications at KU Leuven, while Dr. Kim advances research in soft electronic nanomaterials at the University of Seoul.

This investigation, recently published in the prestigious journal Science Advances, received funding support from the National Research Foundation of Korea under competitive grants targeting outstanding young scientists and doctoral candidates, alongside institutional backing from the University of Seoul. The work not only pushes the envelope in thermoelectrics but also paves the way for commercially viable, scalable production processes suited to real-world deployment.

The implications of this research reach far beyond wearable health devices. Efficient thermoelectric harvesting of low-grade heat sources represents a critical frontier for sustainable energy management, impacting sectors from the Internet of Things to environmental sensing and beyond. The dual thermal conductivity substrate approach could inspire analogous strategies in other thermal engineering challenges, where directional heat flow control within planar devices is paramount.

In summary, the Seoul National University team’s invention of a pseudo-transverse wearable thermoelectric generator exemplifies how innovative material design and device structuring can overturn entrenched limitations. By enabling robust temperature gradients in an ultra-thin, flexible, and flat device leveraging dual-conductivity substrates, they have established a new paradigm for energy harvesting from body heat. This breakthrough offers a glimpse into a future where self-powered wearable electronics are not a niche aspiration but a common reality, seamlessly blending technology with the human form.

Subject of Research: Not applicable

Article Title: All-solution-processed scalable and wearable organic thermoelectrics by structurally mimicking transverse thermoelectric effects

News Publication Date: 18-Mar-2026

Web References:
10.1126/sciadv.aea9094

Image Credits: © Science Advances, originally published in Science Advances

Keywords

Wearable thermoelectrics, pseudo-transverse thermoelectric generator, dual thermal conductivity substrate, body heat energy harvesting, flexible electronics, thin-film thermoelectric devices, organic electronic materials, solution-processed printing, heat flow engineering, sustainable power sources, skin-mounted sensors, scalable device fabrication

The research team, led by Professor Yousung Jung from the Department of Chemical and Biological Engineering, explored how LLMs could enhance the synthesis of novel materials. Their findings suggest that these models can not only predict synthesizability more accurately than conventional methods but also provide valuable insights into the reasoning behind their assessments. This ‘explainability’ aspect of LLMs could open new avenues for understanding the complex factors influencing material synthesis, which range from thermodynamic stability to chemical interactions within crystal structures.

Currently, predicting the synthesizability of materials predominantly involves evaluating their thermodynamic stability. However, existing models have been criticized for their limited accuracy, often resulting in a significant gap between predicted and actual synthesis success rates. Despite various machine learning approaches being employed to tackle this issue, they typically lack transparency and do not elucidate the basis for their predictions. Professor Jung’s research acknowledges these challenges and showcases how fine-tuning LLMs can transcend these limitations.

The team began their study by fine-tuning a general-purpose LLM using extensive datasets related to inorganic materials. Training the model involved inputting text-based information on inorganic crystal structures to classify synthesizability, predict necessary precursor compounds, and interpret essential factors influencing synthesis. When evaluated, the modified LLM demonstrated superior predictive capabilities compared to traditional bespoke machine learning models, marking a significant advancement in the field.

An impressive aspect of this research is the dual capability of the LLM to predict and rationalize its predictions. By analyzing why specific materials are synthesizable, the study pioneers a pathway to pinpoint the reasons behind the difficulties associated with synthesizing certain hypothetical structures. This critical understanding could not only streamline the material discovery process but also foster a more strategic approach toward developing new materials.

One of the surprising outcomes of the research was the identification of previously unknown correlations that could impact synthesizability. As researchers continue to unravel the complex interplay of factors in material science, these revelations could facilitate the design of more predictable synthesis pathways for complex inorganic compounds. Such advancements hold immense potential for applications across various industries, especially those involving semiconductors and advanced battery technologies.

The potential implications of this research are vast, particularly for the South Korean advanced materials industry. By efficiently predicting and elucidating the synthesizability of materials, this innovative approach can drive competitiveness in industries, such as semiconductors and secondary batteries. Traditional material discovery methods require extensive trial-and-error experimentation; however, the insights provided by LLMs could drastically accelerate the time from idea to production.

In a rapidly evolving technological landscape, maintaining a competitive edge in material design is paramount. The South Korean government has placed a strong emphasis on securing leadership in advanced materials, particularly in sectors crucial for the nation’s future, such as electronics and renewable energy sources. The introduction of LLMs for material synthesis predictions can serve as a game changer, ensuring the country remains at the forefront of innovation.

Commercialization of this research represents an exciting prospect. If successfully implemented, the technology could provide research institutions and companies with powerful tools for identifying new materials and assessing their feasibility for mass production. The research team’s findings could dramatically shorten the timeline necessary for bringing new materials to market, creating opportunities for advancements that were previously considered impractical.

Professor Jung highlighted the significance of their work, noting that the capacity of LLMs to predict and explain synthesizability could transform the material design landscape. As these technologies progress, they are anticipated to deliver increasingly efficient and intuitive methods for novel material creation. The implications of this research extend far beyond academic curiosity; they possess the potential to redefine how industries approach material synthesis and development.

Postdoctoral Researcher Seongmin Kim’s role in the study also underscores the collaborative nature of this research. As the leading author of the paper, Kim aims to extend this work through future studies that integrate machine learning with materials science paradigms. This ongoing journey of exploration signifies a united effort to reshape the possibilities within materials development, promising a future where LLMs could become common tools at the disposal of material scientists globally.

Overall, the advancements made by this research team at SNU represent a monumental leap forward in the integration of artificial intelligence and materials science. By employing LLMs for synthesizability prediction, the scientific community can expect more reliable, explainable, and efficient methodologies in the quest for new materials. These developments promise not only to enhance material creation processes but also to pave the way for innovative breakthroughs that may redefine entire industries.

As interest in this research continues to grow, attention will likely turn towards further developments in machine learning applications within materials science. The journey toward refining the capabilities of LLMs in synthesizability predictions is just beginning, and the future holds substantial prospects for even more significant advancements in materials development.

In summary, the research conducted by Professor Jung and his team signifies a pivotal moment in the intersection of artificial intelligence and chemistry, leading to a new understanding of material synthesizability that could redefine the future of material engineering.

Subject of Research:
Article Title:
News Publication Date:
Web References:
References:
Image Credits:

Keywords