Skip to content Skip to navigation


PCO Water Splitting


The schematic of thermochemical water splitting process. 

Today’s chemical infrastructure at large scales relies almost exclusively on thermochemical transformations at temperatures ≤ 1100°C, which has led to a search for materials and mechanisms for two-step thermochemical water splitting below this temperature. We have recently found that poly-cation oxides (PCOs) can thermochemically split water to produce hydrogen and oxygen via such a two-step cycle. Well-designed PCOs can get thermally reduced at < 1100°C, avoiding material loss and making reactor design easier, and then produce meaningful amount of H2 with steam-to-H2 conversion of > 0.1%. It is likely that PCOs with complex cation compositions will offer new opportunities for both fundamental investigations in redox thermochemistry as well as scalable hydrogen production using infrastructure-compatible chemical systems. People: Shang Zhai, Jimmy Rojas, Nadia Ahlborg

Wearable Solid-State Refrigeration


Thermal images of the entropy swing material. 

Artificial refrigeration is one of the cornerstones of modern civilization. Air-conditioning provides thermal comfort and high productivity to people in tropical and subtropical areas. The cold supply chain makes fresh produce and heat-sensitive medicine widely available around the world. Although these artificial refrigeration technologies are crucial for us, the vapor-compression refrigeration cycle use hydrofluorocarbons (HFCs) as the refrigerants, which are actually greenhouse gases thousands of times stronger than CO2. Motivated by the urgent need to alleviate climate change, we are developing new kinds of solid-state refrigeration devices that do not rely on refrigerants and compressors. To achieve high temperature span and high efficiency, both large entropy-swing materials and advanced heat transfer control methods are being investigated. We also explore the opportunities of wearable solid-state refrigeration to localize and fully utilize the cooling power onto the human body. People: Po-Chun Hsu



Examples of using deep learning to detect solar installations on satellite imagery. 

Electric grids are undergoing a profound revolution towards a more sustainable system, from centralized to decentralized, with very deep penetration of distributed energy resources (DER) (solar, battery storage, etc) directly connected to distribution grid. However, connecting distributed PV power generation to existing grids is challenging due to the lack of a complete PV location and size information, making grid operation and planning difficult.

Current solar installation profiles in the U.S. are mainly from customer and PV installer’s self-reports, surveys and incentives, which are often incomplete and outdated. In this project, we are developing a powerful tool with deep learning to automatically detect installed solar panels on satellite imagery. With such tool, we aim to construct a comprehensive solar installation profile including the GPS location and size information, and use it to drive grid operation and policy design. People: Jiafan Yu, Zhecheng Wang

Lower Limit of Heat Transfer

Figure 1

Setup to measure sample thermal conductivity. 

More than 90% of current energy budget is related to thermal energy. Its generation, conversion, transportation, and storage are of great importance. Despite the many research on these topics in the past century, especially nano-scale heat transfer during the past few decades, there are still many unsolved problems about the fundamental limit of heat transfer.

In Magic Lab, we are exploring the lower bound of heat transfer. By utilizing the often-neglected wave nature of thermal energy and destructive interference effect, we are developing a ‘perfect’ thermal insulator. The project will give us a better understanding about a new transport regime, breaking the amorphous limit of bulk material thermal conductivity. At the same time, this research will develop a new thermal insulation approach potentially better than materials with super-lattice, porous structure, etc. Currently, we are also trying to improve the ratio between signal and noise in microscopy with SEM. People: Ze Zhang, Joel Martis.