Today, photo voltaic vitality supplies 2% of U.S. energy. However, by 2050, renewables are predicted to be probably the most used vitality supply (surpassing petroleum and different liquids, pure gasoline, and coal) and photo voltaic will overtake wind because the main supply of renewable energy. To attain that time, and to make solar energy extra reasonably priced, photo voltaic applied sciences nonetheless require quite a lot of breakthroughs. One is the power to extra effectively rework photons of sunshine from the Sun into useable vitality.
Organic photovoltaics max out at 15% to 20% effectivity — substantial, however a restrict on photo voltaic vitality’s potential. Lehigh University engineer Ganesh Balasubramanian, like many others, questioned if there have been methods to enhance the design of photo voltaic cells to make them extra environment friendly?
Balasubramanian, an affiliate professor of Mechanical Engineering and Mechanics, research the essential physics of the supplies on the coronary heart of photo voltaic vitality conversion — the natural polymers passing electrons from molecule to molecule to allow them to be saved and harnessed — in addition to the manufacturing processes that produce industrial photo voltaic cells.
Architecture of the OPV bulk-heterojunction construction and the design scope. [Credit: Ganesh Balasubramanian, Joydeep Munshi, Lehigh University]
Using the Frontera supercomputer on the Texas Advanced Computing Center (TACC) — probably the most highly effective on the planet — Balasubramanian and his graduate scholar Joydeep Munshi have been working molecular fashions of natural photo voltaic cell manufacturing processes, and designing a framework to find out the optimum engineering selections. They described the computational effort and related findings within the May issue of IEEE Computing in Science and Engineering.
“When engineers make solar cells, they mix two organic molecules in a solvent and evaporate the solvent to create a mixture which helps with the exciton conversion and electron transport,” Balasubramanian stated. “We mimicked how these cells are created, in particular the bulk heterojunction — the absorption layer of a solar cell. Basically, we’re trying to understand how structure changes correlate with the efficiency of the solar conversion?”
Balasubramanian makes use of what he calls ‘physics-informed machine studying’. His analysis combines coarse-grained simulation — utilizing approximate molecular fashions that signify the natural supplies — and machine studying. Balasubramanian believes the mixture helps stop synthetic intelligence from arising with unrealistic options.
“A lot of research uses machine learning on raw data,” Balasubramanian stated. “But more and more, there’s an interest in using physics-educated machine learning. That’s where I think lies the most benefit. Machine learning per se is simply mathematics. There’s not a lot of real physics involved in it.”
Writing in Computational Materials Science in February 2021, Balasubramanian and Munshi together with Wei Chen (Northwestern University), and TeYu Chien (University of Wyoming) described outcomes from a set of digital experiments on Frontera testing the consequences of assorted design modifications. These included altering the proportion of donor and receptor molecules within the bulk heterojunctions, and the temperature and period of time spent in annealing — a cooling and hardening course of that contributes to the soundness of the product.
They harnessed the info to coach a category of machine studying algorithms referred to as help vector machines to determine parameters within the supplies and manufacturing course of that may generate probably the most vitality conversion effectivity, whereas sustaining structural power and stability. Coupling these strategies collectively, Balasubramanian’s crew was in a position to scale back the time required to achieve an optimum course of by 40%.
“At the end of the day, molecular dynamics is the physical engine. That’s what captures the fundamental physics,” he stated. “Machine learning looks at numbers and patterns, and evolutionary algorithms facilitate the simulations.”
Trade-Offs and Limitations
Like many industrial processes, there are trade-offs concerned in tweaking any aspect of the manufacturing course of. Faster cooling might assist enhance energy effectivity, however it might make the fabric brittle and prone-to-break, as an illustration. Balasubramanian and his crew employed a multi-objective optimization algorithm that balances the advantages and downsides of every change to derive the general optimum manufacturing course of.
Flowchart describing steps in a typical coupled Cuckoo Search-Coarse Grained Molecular Dynamics (CS-CGMD) algorithm. The dashed field represents the augmented machine realized exploration of the areas of curiosity to complement ill-performed nests with newer alternate options throughout every CS optimization era. [Credit: Ganesh Balasubramanian, Joydeep Munshi, Lehigh University]
“When you try to optimize one particular variable, you are looking at the problem linearly,” he stated. “But most of these efforts have multi-pronged challenges that you’re trying to solve simultaneously. There are trade-offs that you need to make, and synergistic roles that you must capture, to come to the right design.”
Balasubramanian’s simulations matched experimental outcomes. They decided that the make-up of the heterojunction and the annealing temperature/timing have the most important results on total effectivity. They additionally discovered what quantity of the supplies within the heterojunction is finest for effectivity.
“There are certain conditions identified in literature which people claim are the best conditions for efficiency for those select molecules and processing behavior,” he stated. “Our simulation were able to validate those and show that other possible criteria would not give you the same performance. We were able to realize the truth, but from the virtual world.”
With an award of extra time on Frontera in 2021-22, Balasubramanian will add additional layers to the machine studying system to make it extra sturdy. He plans so as to add experimental information, in addition to different modalities of laptop fashions, reminiscent of digital construction calculations.
“Heterogeneity in the data will improve the results,” he stated. “We plan to do first principle simulations of materials and then feed that data into the machine learning model, as well as data from coarse-grained simulations.”
Balasubramanian believes that present natural photovoltaics could also be reaching the bounds of their effectivity. “There’s a wall that’s hard to penetrate and that’s the material,” he stated. “These molecules we’ve used can only go so far. The next thing to try is to use our framework with other molecules and advanced materials.”
His crew mined the literature to grasp the options that enhance photo voltaic effectivity after which skilled a machine studying mannequin to determine potential new molecules with excellent cost transport behaviors. They printed their analysis within the Journal of Chemical Information and Modeling. Future work on Frontera will use Balasubramanian’s framework to discover and computationally take a look at these different supplies, assuming they are often produced.
“Once established, we can take realistic molecules that are made in the lab and put them in the framework we’ve created,” he stated. “If we discover new materials that perform well, it will reduce the cost of solar power generation devices and help Mother Earth.”
Balasubramanian’s analysis harnesses the 2 issues that laptop simulations are crucial for, he says. “One is to understand the science that we cannot study with the tools that we have in the real world. And the other is to expedite the science – streamline what we really have to do, which reduces our cost and time to make things and physically test them.”