Development of new materials with artificial intelligence robots
Ada is a robot driven by artificial algorithms that can help researchers at the University of British Columbia to find new solar cell design solutions. Image source: FRASER PARLANE
Curtis Berlingate is a materials scientist. When working at the University of British Columbia in Canada, he asked graduate students to improve key materials in solar cells to improve their conductivity.
In this process, he found that there are a large number of potential adjustment variables, and different variables can generate tens of millions of possibilities. For example, adding trace metals and other additives can change the heating and drying time.
Berlin Gate and colleagues handed over the work to a single-arm robot with an artificial intelligence algorithm. The robot can mix different solutions and cast it into a thin film before performing heat treatment or subsequent steps.
At a meeting held recently by the American Society for Materials Research, Berlin Gate reported on the latest results of this system: after figuring out the formula and heating conditions, artificial intelligence can create new films for solar cells, which previously required 9 months. The problem solved now takes only 5 days.
In fact, in other fields such as drug development and genetics research, artificial intelligence has been used to design experiments. For example, programming with a DNA synthesizer gives any possibility of DNA assembly.
But for a certain material, it is impossible to process or synthesize it with a single method, which means that the processing flow of the automated system guided by intelligent algorithms will be more complicated. The results of Berlin Gate et al. Mean that similar systems in the field of materials science have been produced. "This is an exciting field." Material scientist Benji Maruyama who works in the US Air Force Research Laboratory commented, "Forming a closed system loop means that the field of materials will innovate at a faster rate."
There are more than 100 elements in the periodic table of elements, and in theory, they can be combined in countless ways, resulting in a considerable amount of materials. This means that there are hundreds of thousands of materials waiting for people to discover, on the other hand, how to select the truly usable parts among them has also become a challenge.
Now artificial intelligence robots can help. Robots can mix dozens of different material formulations (these formulations have subtle differences), and then place the materials produced by different formulations on a single wafer or other materials for processing and testing.
However, Maruyama also said that if the experiment is simply conducted one by one, it can only be regarded as a kind of high-throughput experiment, rather than a breakthrough innovation.
To speed up this process, many research teams use computer modeling to find possible material formulations, and many new materials have been born. But the problem is that the design of these systems often relies on graduate students or experienced scientists in materials science. The system evaluation is given by people with established standards, and it is up to the people to decide whether to conduct experiments. But people cannot always control all steps.
Similar to the Berlin Gate team, mechanical engineer Keith Brown working at Boston University also built a robotic system driven by artificial intelligence.
The research goal of the Brown team is to find a sufficiently strong 3D printed structure. The toughness of the material depends on the details of the structure, which requires both strength and good ductility. These are often not predictable and must be tested experimentally.
As a test case, Brown et al. Used plastic to create a barrel-shaped structure that is about the size of a salt shaker. The research team changed the number, direction, and shape of the pillars on the outer wall of the barrel, but all the variables together may produce about 500,000 combinations.
To find the right structure faster, the Brown team first created 600 different structures with robots and sampled all options. Then, they used artificial intelligence algorithms to measure the optimal design that might be produced in the experiment.
Through experiments and calculations, the relevant program can find out the trend of materials with good toughness, such as the thickness and radius of each pillar, which helps to predict a stronger structure. And all this does not require researchers to keep a close watch. Twenty-four hours after the program started, the researchers obtained a stronger structure than any previous original design.
Whether it is a perovskite solar cell or a 3D printed material, these artificial intelligence-based systems can help researchers find good structures faster, and even have a more profound impact on their field. (Yuan Liu)
Spherical Lens,Optical Glass Sphere Lens,Spherical Meniscus Lens,Spherical Plano-Convex Lens
China Star Optics Technology Co.,Ltd. , https://www.realpoooptics.com