Advancing solar cell technology through materials research

Artificial intelligence is rapidly reshaping the modern world. Data centers that support artificial intelligence require enormous amounts of energy to remain operational – and the demand grows each year.

According to the International Energy Agency, electricity consumption from data centers worldwide is projected to reach 945 terawatt-hours by 2030, over double the 415 terawatt-hour demand reported in 2024.

As the global energy demand increases, the improvement of energy efficiency has become a top priority for researchers.

At Arizona State University, the Feng Yan Lab addresses these challenges through research focused on thin-film solar cells and semiconductor materials designed to improve the harvesting, conversion and storage of renewable solar energy.

Led by Feng Yan, an associate professor in the Materials Science and Engineering program at the School for Engineering of Matter, Transport and Energy, the lab investigates new materials and device architectures to improve the efficiency and performance of next-generation solar technology.

“The AI revolution requires data centers which rely on power plants,” Yan says. “Solar energy can play an important role in powering these systems on Earth and in space.”

Alternatives to traditional solar cells

Traditional solar cells are often composed of silicon, then used in solar panels for energy conversion. Although silicon dominates the current commercial market, other solar materials provide advantages for the development of thin-film cells.

“We investigate different sunlight absorbing materials, such as Cadmium Telluride, Antimony selenide/sulfide and halide perovskite solar technology. We also investigate tandem solar technology by stacking these solar devices,” Yan says. “Each of these materials has a unique performance, crystalline structure and specific thin film deposition.”

One challenge with conventional solar panels is their heavy composition, high material consumption, and high manufacturing costs. Silicon-based panels are bulky, making them difficult to install on certain structures and more expensive to produce. Thin-film solar panels, by comparison, have lower material usage, stronger light absorption, are cost-effective, and offer better flexibility.

However, improving the efficiency and durability of thin-film solar cells remains a challenge for researchers.

Each semiconductor contains a band gap, and the magnitude of the band gap determines what kind of light the material can absorb. In Yan’s lab, there is a particular interest in perovskite solar cells, which contain adjustable band gaps.

“The unique thing about perovskite material is its defect tolerance,” Yan says. “That means the material is able to maintain its performance despite imperfections in the cell compared to other solar technologies, like commercial silicon and Cadmium Telluride, which are very sensitive to the defects.”

Tandem solar cells

To improve light absorption and overall efficiency, Yan’s group also studies the perovskite cells in a different composition; the tandem solar cell.

Sunlight can shine through a single absorbing layer relatively easily. However, additional absorber layers with varying band gaps can maximize light absorption across the solar spectrum. Tandem solar cells stack low and high bandgap solar cells in one device to maximize the sunlight absorption.

“One material can only absorb a certain area of the sunlight spectrum based on its band gap,” Yan says. “The tandem device balances different absorbers with various band gaps, allowing for a broader range of absorption than a single solar cell is capable of.”

Developing these materials, however, requires extensive experimentation.

Traditional solar cell research often relies on trial-and-error testing, repeatedly altering chemical compositions and fabrication conditions to improve device performance.

A recent revolution for Yan’s lab is the integration of machine learning into this process.

“By utilizing machine learning and AI, we can narrow our data down to specific growth windows for the thin-film and achieve the desired device performance in a short period,” Yan says.

The desired thin-film growth window is the range of conditions under which the thin-film material forms the desired structure, such as temperature, film thickness, pressure, etc.

By using machine learning to process experimental data, Yan’s lab can identify patterns and optimize device performance more efficiently.

Learning through collaboration

Because this research combines materials science, physics, chemistry and electrical engineering concepts, collaboration plays a major role within Yan’s lab. “Students may come from a variety of different majors, but they learn from each other while working on the same project,” Yan says.

Through collaborative research, the students gain experience in semiconductor physics, material synthesis, and computational modeling.

For undergraduate students, the lab provides opportunities to bridge the gap between classroom knowledge and hands-on applications.

As for graduate researchers, Yan employs a more exploratory approach to learning.

“We have to think out of the box,” Yan says. “Research is about creating new knowledge based on what is already documented in textbooks and the scientific literature.”