Machine learning unlocks superior performance in light-driven organic crystals

Researchers have developed a machine-learning workflow to optimize the output force of photo-actuated organic crystals. Using LASSO regression to identify key molecular substructures and Bayesian optimization for efficient sampling, they achieved a maximum blocking force of 37.0 mN—73 times more efficient than conventional methods.

These findings, published in Digital Discovery, could help develop remote-controlled actuators for medical devices and robotics, supporting applications such as minimally invasive surgery and precision drug delivery.

Materials that convert external stimuli into mechanical motion, known as actuators, play a crucial role in robotics, medical devices, and other advanced applications. Among them, photomechanical crystals deform in response to light, making them promising for lightweight and remotely controllable actuation. Their performance depends on factors such as molecular structures, crystal properties, and experimental conditions.

A key performance indicator of these materials is the blocking force—the maximum force exerted when deformation is completely restricted. However, achieving high blocking forces remains challenging due to the complex interplay of crystal characteristics and testing conditions. Understanding and optimizing these factors is essential for expanding the potential applications of photomechanical crystals.

In a step toward optimizing the output force of photo-actuated organic crystals, researchers from Waseda University have leveraged machine learning techniques to enhance their performance. The study was led by Associate Professor Takuya Taniguchi from the Center for Data Science, along with Mr. Kazuki Ishizaki and Professor Toru Asahi, both from the Department of Advanced Science and Engineering, Graduate School of Advanced Science and Engineering at Waseda University.

“We noticed that machine learning simplifies the search for optimal molecules and experimental parameters,” says Dr. Taniguchi. “This inspired us to integrate data science techniques with synthetic chemistry, enabling us to rapidly identify new molecular designs and experimental approaches for achieving high-performance results.”

In this study, the team utilized two machine learning techniques: LASSO (least absolute shrinkage and selection operator) regression for molecular design and Bayesian optimization for selecting experimental conditions. The first step led to a material pool of salicylideneamine derivatives, while the second enabled efficient sampling from this pool for real-world force measurements.

As a result, the team successfully maximized the blocking force, achieving up to 3.7 times greater force output compared to previously reported values and accomplishing this at least 73 times more efficiently than conventional trial-and-error methods.

“Our research marks a significant breakthrough in photo-actuated organic crystals by systematically applying machine learning,” says Dr. Taniguchi. “By optimizing both molecular structures and experimental conditions, we have demonstrated the potential to dramatically enhance the performance of light-responsive materials.”

The proposed technology has broad implications for remote-controlled actuators, small-scale robotics, medical devices, and energy-efficient systems. Because photo-actuated crystals respond to light, they enable contactless and remote operation, making them ideal robotic components working in confined or sensitive environments. Their ability to generate force noninvasively with focused light could also be valuable for microsurgical tools and drug delivery mechanisms that require precise, remote actuation.

By leveraging a cleaner energy input—light irradiation—while maximizing mechanical output, these materials hold promise for eco-friendly manufacturing processes and devices aimed at reducing overall energy consumption. “Beyond improving force output, our approach paves the way for more sophisticated, miniaturized devices, from wearable technology to aerospace engineering and remote environmental monitoring,” Dr. Taniguchi adds.

In conclusion, this study highlights the power of a machine learning–driven strategy in accelerating the development of high-performance photo-actuated materials, bringing them one step closer to real-world applications and commercial viability.