As the need for autonomous and on-site renewable power sources grows, developing efficient energy solutions for distributed sensors, wearable electronics, cooling systems, and other low-power applications has become increasingly critical. Organic thermoelectric generators (TEGs), which convert low-grade heat into electrical energy through the Seebeck effect, offer a promising solution for powering these devices. Organic TEGs possess some advantages over inorganic TEGs in the context of sustainable energy harvesting because the active materials are often solution-processable at room temperature, which enables scalable patterning and printing techniques. Furthermore, these semiconductors are typically derived from Earth-abundant, non-toxic elements, making them environmentally friendly and sustainable. Among organic semiconductors, conducting polymers, particularly PEDOT (Poly(3,4-ethylenedioxythiophene)), emerge as pivotal materials in organic TEGs due to their favorable electrical and thermal properties. Thus, a deep understanding of these polymers is essential for guiding material design and optimizing device performance. In this regard, computational methods represent an important tool in studies of organic thermoelectric materials and devices since they not only provide insights into the electronic and thermal properties of materials on atomic and molecular levels but also allow for the prediction of the device's performance without the need for extensive experimental work.

This thesis employs multi-scale computational modeling to advance the understanding and optimization of organic thermoelectric materials and devices, including: (I) Finite element method modeling to analyze and optimize the micro-TEGs, (II) Ab initio molecular dynamics simulations to investigate charge transport mechanisms in PEDOT conducting polymer, and (III) Machine learning approaches to predict and study the electronic properties of PEDOT thin films.

Part (I) presents that achieving power densities in the range of mW cm−2 at a temperature gradient of 10 K is feasible through geometrical optimization and utilizing advanced organic thermoelectric inks. Particularly, we simulated the PEDOT:PSS/BBL:PEI micro-TEGs and improved device efficiency under varying thermal gradients using COMSOL software.

In part (II), we developed a computational technique based on ab initio molecular dynamics to trace the temporal motion of charge carriers in a single PEDOT chain and in coupled chains with varying degrees of interaction. Subsequently, we used ab initio molecular dynamics to demonstrate that charge transport along the chains is band-like, while transport across chains follows a hopping-like mechanism. The calculated polaron mobility along the chains reached 4 cm2V−1s−1, providing a theoretical upper limit for thiophene-based conducting polymers. Also, we quantified the hopping rate between chains, consistent with Marcus theory, by analyzing polaron jumps.

Part (III) integrates computational modeling with machine learning to explore changes in morphological and transport properties of PEDOT:Tos prepared using different solvents. We employed convolutional neural networks to achieve high accuracy (r2>0.99) in predicting electronic coupling values and significantly accelerated the analysis compared to density functional theory calculations. This approach enabled detailed investigations into how different solvents affect the electronic coupling of PEDOT dimers.

We believe that our findings on organic thermoelectric material and devices provide a comprehensive framework for improving the performance and scalability of organic TEGs and open new avenues for further research.