Four potentially high-impact projects seeking to solve critical energy-efficiency challenges have been awarded more than $240,000 in cumulative funding from UC Santa Barbara’s Institute for Energy Efficiency (IEE), the College of Engineering’s interdisciplinary research center dedicated to cutting-edge science and technologies that support an energy-efficient and sustainable future. Each project aligns with at least one of the institute’s key interdisciplinary thrusts: smart societal infrastructure, computing and communications, and the food-energy-water nexus.

Two of the projects received Investment Group of Santa Barbara (IGSB) Software Impact grants, which support high-impact research of energy-efficient software that is likely to lead to commercialization and positively impact society. The selection committee also awarded $50,000 grants to two projects through the IEE Research Seed Grant Program. Seed grants are intended to help researchers produce preliminary results that can be used to apply for major external funding to expand their projects.

“Nurturing early-stage concepts with modest yet meaningful financial support not only jumpstarts scientific success but also cultivates and continues the culture of collaboration and discovery that thrives within the IEE and the university,” said IEE Director John Bowers, a distinguished professor of electrical and computer engineering, and materials.  

“Whether they are being pursued by junior faculty or highly esteemed researchers, these four projects share key characteristics: they are strong and innovative proposals with significant potential to impact society,” added Mark Abel, the executive director of the IEE. 

The four projects involve a total of five UCSB faculty members from the Departments of Chemical Engineering, Materials, Electrical and Computer Engineering, Computer Science, and Chemistry & Biochemistry.

Turning Waste into Valuable Carbons
Lithium (Li) is the active ingredient in rechargeable batteries that power the majority of today’s smartphones, laptops and electric vehicles. However, lithium’s scarcity and high price have led researchers to see sodium (Na), which shares many of Li’s properties, as a viable alternative. Sodium is nearly 1,200 times more available in the world than lithium, and its abundance makes it more affordable to extract and purify. Materials for Na-ion batteries have already been identified for cathodes, the high-voltage electrode that readily accepts electrons, and electrolytes, the liquid or solid that separates the cathode from anode and conducts ions but not electrons. The deployment of Na-ion batteries, which function through reversible insertion and extraction of Na into/from the electrode material, has been hampered by the lack of materials for anodes, the low-voltage electrode. 

Two UCSB professors, Rachel Segalman and Ram Seshadri, have received an IEE Research Seed Grant to design an alternate anode material for Na-ion batteries and eliminate plastic waste at the same time.

“We come from two separate research fields, but this opportunity reflects the collaborative nature of UCSB and shows how adventurous and supportive the IEE Seed Program is,” said Seshadri, a distinguished professor of materials and chemistry & biochemistry.

The pair will investigate hard carbons, a solid form of carbon that has been pyrolyzed or heated in such a way that it cannot be transformed into graphite and is known to store Na when employed as anodes. Graphitization contributes to the low-charge capacity associated with Na-ion batteries when compared to Li-ion batteries.  Their proposed project unites the polymer expertise of Segalman, the Kramer Professor of Materials and Chemical Engineering, and Seshadri’s expertise in high-temperature processing and vast understanding of anode properties. The first goal of their project is to employ inexpensive precursors and novel processing techniques to make better and more effective carbons.

“Hard carbons can be derived from polymers. So, we are interested in using techniques from my field to pre-organize them and alter their properties. It’s an approach that has not been done before,” explained Segalman. “But it’s extremely difficult to receive significant funding unless you can prove that your hypothesis is valid. This seed grant allows us to do just that.”

After understanding how the pre-treatment of polymers can result in better carbons, their next step will be to convert the materials into hard carbons by using rapid microwave heating, which has already proven its value in generating high-temperature battery materials.

“We believe that these carbons can be optimized to make very effective low-voltage electrodes in a sodium battery,” explained Seshadri

The pair will also investigate if they can reuse polymers or plastics that would otherwise go into waste streams, to make materials that can contribute to sustainability and increase grid-energy storage.

“You could literally use any plastic, such as bottles or fishing nets, because we’re burning it at a high temperature without oxygen,” explained Segalman. “The process doesn’t produce carbon dioxide. Instead, it creates a hard carbon.”

“This project could produce a high-value material that increases the percentage of renewables in the energy pipeline out of something that we don’t need anymore,” said Seshadri.

Empowering a Pie-in-the-Sky Idea
Computer science assistant professor Jonathan Balkind wants to take a page out of the history books about the early days of graphics processing units (GPUs). Back then, researchers figured out how to shoehorn their data and code into low-level computer graphics, known as graphics kernels, in order to exploit the high-performance processing engines inside the GPUs. Once manufacturers understood the GPUs’ potential for high performance and energy efficiency, they generalized their designs, enabling radical improvements in computational energy efficiency for highly parallel workloads.

“The rise of GPUs serves as inspiration for my research group; only we want to make machine learning more efficient by moving data,” said Balkind, who previously received an Early CAREER Award from the National Science Foundation and a Trailblazer Fellowship from the Open Source Hardware Association.

Due to the widespread use of machine learning (ML), specialized processors designed to accelerate ML tasks, called tensor processing units (TPUs), are commonly found in modern computing systems. TPUs specialize in performing matrix multiplication, a fundamental operation in ML that combines the rows and columns of two matrices into a new matrix. Also referred to as accelerators, TPUs perform operations with high efficiency. The problem is that while an accelerator runs, the rest of the computing system, the central processing unit (CPU), remains active and simultaneously performs key functions, such as setting up and composing matrix multiplication kernels, as well as handling the computer’s input/output, and performing calculations and logic, all of which consumes large amounts energy.

“We want to perform more calculations on the TPU rather than relying so much on the CPU,” said Balkind. “TPUs are optimized specifically for machine learning, so there is a tremendous potential to reduce the amount of time and energy required to complete an operation if we are able to broaden the functionality of the accelerators.”

Initial work on a promising solution has already been completed by a master’s student and undergraduate students in Balkind’s research group. The team built a new software program, called the Mullifier, to translate existing high-level programming source code and generate an equivalent lower-level language, which included series of matrix multiplications for the accelerator. To date, their program has handled arithmetic, branching, looping, as well as complex operations like stack and memory management.

“This grant allows us to examine our hardware’s design in order to expand the amount of code that we could support,” explained Balkind. “The more code and pathways between different units that we can incorporate, the faster and more energy efficient the process will be.”

The opportunity to pursue, what he described as a “crazy idea” will allow his research group to investigate if small changes in hardware could have a big impact on energy efficiency.

“One of the reasons why I came to UCSB was because the scientific community encourages people to think outside the box,” he said. “This is a great example of the IEE and this campus empowering students and researchers to pursue pie-in-the-sky research that could potentially impact the world.”

Optimizing for Impact
Large Language Models (LLMs), like ChatGPT, are massive AI foundation models designed to understand and generate human language. Based on neural-network architecture, they are trained on large amounts of text data, which allows them to observe linguistic patterns and execute language-related tasks, including text generation, translation, summarization, and question answering. Despite their increasingly excellent performance, large AI foundation models suffer from a major limitation — the extremely high cost of the foundational step of pre-training a model on a vast and diverse dataset. For instance, Meta’s LLaMA model used 2,048 Nvidia A100 graphics processing units (GPUs) over 21 days. GPUs are used to train LLMs because they can break down and process large datasets simultaneously, which accelerates the training.

“Pre-training an LLM requires a huge data set and massive computing resources to maximize its accuracy, but that is extremely expensive,” said Zheng Zhang, an associate professor of electrical and computer engineering. “Each training run costs a few million dollars, and the practical pre-training can take multiple training runs to complete,"

The sky-high costs mean that pre-training is affordable only for giant tech companies like Google and Amazon, and not for academics. The carbon emissions associated with pre-training have also raised red flags regarding its environmental impacts. Zheng has received an IGSB Software Impact Grant of more than $90,000 to address these shortfalls by developing a novel pre-training framework that requires fewer computing resources, energy, and money, and produces dramatically fewer carbon emissions.

“Our plan focuses on perfecting the algorithm,” explains Zhang, who has previously received an Early CAREER Award from the National Science Foundation, multiple best paper awards, and the Ernest S. Kuh Early Career Award from IEEE’s Council on Electronic Design Automation. “Because pre-training costs are measured by GPU hours, we are working to reduce the amount of time it takes to complete the process.”

Zhang will optimize the algorithm by implementing a low-rank tensor compression model, which reduces the number of parameters needed to represent a neural network. The model breaks down the neural network into a set of smaller tensors or matrices, a process known as factorization, which can lead to computational and storage savings. The seed grant allows Zhang to test his theory and expand on promising work recently completed in his lab. His research group developed CoMERA, which stands for Computing and Memory-Efficient training method via Rank-Adaptive tensor optimization. Preliminary results showed that CoMERA reduced the number of training variables by 50-200 times on medium-sized transformers, greatly decreasing the memory cost in training.

“We have already demonstrated CoMERA’s effectiveness on a single GPU and smaller model,” explained Zhang. “Now, we want to demonstrate that this algorithm could be scalable to large language models on massive GPUs.”

Preliminary results also indicated some areas need improvement, so the grant will allow Zhang’s teams to do the work. They aim to achieve at least 10 times the LLM compression in pre-training and 2.4 times the speed on a single GPU. Once the targeted performance is achieved, Zhang will be able to test the framework on a larger scale by leveraging a previously awarded grant of hundreds of GPU hours from the Department of Energy.

“We are extremely grateful for the opportunity to do some unrestricted research to show the technical and commercial impact of our work,” said Zhang. “We are also thankful to the Department of Energy for providing us with the GPU hours and computing time, which very few academic researchers are granted.”

Zhang believes that decreasing the hours to complete a pre-training run would have significant environmental and economic impacts.

“Greatly reducing pre-training costs would also lower the barrier for small businesses and academic groups to develop large language learning models,” said Zhang. “There is also a significant environmental impact, because if you can reduce the cost and GPU hours by three times, that also means you’re consuming approximately three times less electricity and creating three times fewer carbon emissions.”

Energy-Efficient Membranes for Biosensors
Andrea Carlini, an assistant professor in the Department of Biochemistry and Chemistry, has received an IEE Seed Grant to design a novel selective polymeric membrane, which will enhance the energy efficiency and functionality of biosensors, such as glucose monitoring devices. A grateful Carlini says that the seed funding offers immeasurable opportunity and support.

“This grant gives me a chance to explore a completely unfounded idea and see where the science takes me,” said Carlini. “Association with the IEE also affiliates me with an entirely new group of faculty and resources, all of which will provide tremendous insight and access for my students to be successful.”

Polymers are large molecules made up of hundreds or even millions of repeating subunits called monomers, which are covalently bonded together in a chain-like structure. Carlini describes polymers as beaded strings, where each bead is a monomer that has its own chemical properties. There are two types of polymers: synthetic polymers, which are man-made polymers that are produced through chemical reactions, and natural polymers, which are produced by living organisms through biological processes. Biosynthetic polymers, a marriage of both, offer a wide range of advantageous properties and can be tailored to specific applications through modifications of the monomer’s chemical structure. For example, polymer membranes were developed for biosensors. Carlini’s project stems from a meeting with a local company that designed a glucose monitoring device for diabetics, a global population of 537 million adults. The company is seeking a novel membrane to regulate the diffusion of glucose and oxygen to electrochemical sensors, which also provides more sensitivity and extends the devices functional lifespan. 

“Membranes found in numerous biosensors, like the one involved in my project, are based on polymer strategies developed more than thirty years ago. In my opinion, their architectures are haphazardly mixed together,” explained Carlini. “Since then, however, novel polymerization strategies and fabrication techniques have been developed, a few of which I intend to use in my project to provide precision and reproducibility.”

One of the techniques that she plans to employ is ring-opening metathesis polymerization, or ROMP. Carlini will use ROMP to synthesize various polymer architectures in an effort to increase a membrane’s selective permeability for oxygen over glucose. Researchers in her lab will evaluate the polymer compositions based on their ability to diffuse gasses and small molecules in an effort to overcome the fact that the human body supplies three orders of magnitude more glucose than oxygen, a challenge referred to as “the oxygen-deficit problem.”

“ROMP is awesome,” summarized Carlini. “The technique gives you tremendous control over what type of monomers you can use and their order along the beaded string. This affords me synthetic complexity with minimal effort.”

Carlini will also employ click chemistry using nitrile-imine tetrazole-ene cycloaddition (NITEC) reactions to address another significant problem found in existing membranes; that is, the questionable seal which bonds the glucose-sensing enzymes to the membrane. NITEC is a recently developed technique that is widely used to quickly form new bonds by shining light on it, with the added bonus that each new linkage is fluorescent. Carlini will use the NITEC reaction to bond her polymers to both themselves and the enzyme surfaces, effectively providing a visual display of the polymer membrane seal and any material defects.

“NITEC is a beautiful way to understand a reaction as it happens and make sure that a substance adheres to the device,” explained Carlini. “If the techniques perform as I predict, we could potentially attain reproducible membranes having the desired mechanical and functional properties.”

In addition to improving the semi-permeable membrane of a glucose monitoring device, Carlini’s work could also improve the efficiency of biosensors and be applied to devices that monitor water quality in desalination plants or manage resources on the Smart Grid.

Caption: The UCSB faculty who received 2024 seed funding for their projects include (clockwise from top left) Jonathan Balkind, Andrea Carlini, Rachel Segalman, Ram Seshadri, and Zheng Zhang.