Light-Driven Smart Filters: Thiosemicarbazone Photoswitches for Water Purification

PI: Ivan Aprahamian, Roth Family Distinguished Professor of Chemistry

Project team: Ramanpreet Kaur, PhD student, Chemistry, Aprahamian Group

Abstract: Access to clean water remains a global challenge, particularly because of contamination by environmentally persistent anions such as nitrate, phosphate, sulfate, and halide ions. Current anion remediation technologies, including reverse osmosis and ion exchange, are effective but energy-intensive and costly. They also rely on chemical additives, complex infrastructure, and often generate secondary waste streams. Consequently, the development of low-cost, sustainable, and environmentally benign approaches for anion remediation is essential for protecting both environmental and public health.

To address this challenge, the team will develop visible light-responsive thiosemicarbazone (TSC) photoswitches that can be directly activated by solar irradiation. These smart, light-responsive systems will enable reversible, on-demand anion capture and release. The outcomes of this research will establish the foundation for next-generation, sustainable anion filtration and remediation technologies powered by sunlight, the world's most abundant renewable energy source.
 

MnAl Exchange-Spring Magnets

PI: Ian Baker, Sherman Fairchild Professor of Engineering & Senior Associate Dean, Research and Graduate Programs

Project Team: Sarah P. Slotznick, Assistant Professor of Earth Sciences

Abstract: High-performance rare earth (RE) magnets are increasingly used in applications such as wind turbine generators and motors in both electric and hybrid vehicles since they are the strongest permanent magnets with high demagnetization resistance. However, the use of RE elements is problematic due to the heavy reliance on Chinese RE magnets; the severe environmental degradation associated with RE mining and refining (87% in China); the high cost of RE magnets; and RE price volatility. Thus, there is a need for new magnets that can replace RE magnets. MnAl magnets, which consist of abundant and low-cost elements, can potentially fulfill this need if their magnetic strength can be improved. Here, we attempt to improve MnAl magnets by adding up to 20% of a magnetically stronger, yet easier to demagnetize magnetic material, FeCo, via powder processing followed by compaction. The resulting so-called “exchange-spring” magnet will combine these two materials in a nanolayered microstructure.  Our processing approach can easily be scaled up for commercial applications and thus, if successful, is a clear stepping stone for significant reductions in the need for RE magnets.
 

Biophysical AI for Enzyme Engineering at the Frontiers of the Energy Transition

PI: Emme Burgin, Assistant Professor of Engineering

Project Team: Tony Trent, PhD candidate, Molecular & Cellular Biology; Myra Afzal, PhD student, Thayer School of Engineering; Matt LaCapra, PhD student, Biochemical Engineering, Thayer School of Engineering

Abstract:  Enzymes are proteins used by living things to accomplish all of the rich chemistry that makes life possible. New types of enzymes that have been engineered or designed by humans are central to existing and proposed biotechnologies that are essential for meeting key goals of the climate and energy transition, such as the manufacture of renewable fuels and materials and the decarbonization of existing industries. Fortunately, a massive boost in enzyme engineering capabilities has lately been made possible by the AI revolution taking place in protein science, built on large databases of protein data collected over the preceding decades. However, without new sources of large-scale training data the field has entered a period of diminishing returns. Existing tools especially struggle with enzymes that are very unlike those in their training datasets, or that are operating under conditions or performing functions different from those for which they were naturally evolved, though such enzymes would be especially transformative in enabling new biotechnologies. In this project, we will demonstrate increased performance for protein AI tools on just those sorts of unusual enzymes and functions by taking advantage of data from molecular dynamics simulations, which are molecule-scale physics simulations that have been largely neglected in the protein AI field so far. This will allow our models to take advantage of general biophysical design rules, even far outside the context of naturally evolved enzymes, and make important headway into designing biocatalysts for novel purposes not addressed by nature.
 

Robust verification of soil carbon sequestration as a viable Natural Climate Solution

PI: Josh Landis, Principal Research Scientist

Project Team: Caitlin Hicks Pries, Associate Professor of Biological Sciences; Joseph Tumber-Davila, Assistant Professor of Environmental Studies; Carl Renshaw, Professor of Earth Sciences; and Jordan Fields, Aquatic Resiliency Scientist, Trout Unlimited

Abstract: The storage of carbon in soils is a natural process that has been degraded over time by changing land use, resulting in more carbon being released into the atmosphere. Restoring the ability of soils to capture more carbon is a potential strategy to reduce the impacts of climate change as well as to restore sustainability of food production. How quickly soil can be restored remains poorly understood, however, so we developed a new method to understand how this occurs. We will test this method using soils that have been identified as important by our partnering land stakeholders. We hypothesize that soil carbon storage is linked to the association of carbon with soil minerals and we will determine how quickly this association grows over time under different soil management strategies. These results will empower stakeholders andcarbon markets to value soil carbon storage as a natural climate solution.
 

Leveraging In-Situ Doping in 3D-Printed Catalysts for Enhanced Energy Conversion

PI: Yan Li, Associate Professor of Engineering

Project Team: Mignzhao Li, PhD student, Thayer School of Engineering;  Xiangbei Liu, PhD candidate, Thayer School of Engineering; William Sober, Research Experiences for Undergraduates student, Amherst College; and Ya Tang, PhD candidate, Thayer School of Engineering

Abstract: Zeolite catalysts play a key role in converting waste materials (e.g., biomass and plastics) into jet fuel, a critical resource for transportation. It was reported that doping zeolite catalysts with certain metal dopants can enhance oil production by up to 41.08%. However, traditional doping methods rely on impregnation, where catalysts are soaked in a liquid solution with dopants. This diffusion-driven postprocessing is time-consuming and often leads to non-uniform dopant distribution and structural damage. To address these challenges, an in-situ doping strategy, which provides precise control over dopant uniformity and concentration, will be integrated into our established 3D printing framework in collaboration with the Oak Ridge National Lab. The resulting open-celled zeolites with optimal dopants are expected to achieve 50% higher catalytic activity compared to the commercial counterparts. This rapid and scalable manufacturing approach contributes to sustainable biofuel production while strengthening U.S. energy resilience against supply chain disruptions.
 

A Meta-composite Thermoelectric Device for Efficient Low-Grade Waste Heat Recovery

PI: Yan Li, Associate Professor of Engineering

Project Team: Mignzhao Li, PhD student, Thayer School of Engineering;  Xiangbei Liu, PhD candidate, Thayer School of Engineering; William Sober, Research Experiences for Undergraduates student, Amherst College; and Ya Tang, PhD candidate, Thayer School of Engineering

Abstract: The rapid expansion of data centers generates substantial low-grade waste heat (<200 °C) that is difficult to recover using conventional thermal technologies. Thermoelectric generators (TEGs), which directly convert temperature gradients into electricity, offer a compact, solid-state solution for harvesting this otherwise wasted energy. However, large-scale commercialization of existing TEGs is limited by low power output arising from inefficiencies at both the material and device levels. This project aims to address these challenges by developing a meta-composite thermoelectric device that can potentially double the power output compared to conventional TEGs. The proposed approach also overcomes key challenges in 3D printing brittle thermoelectric materials, enabling more scalable manufacturing and easier integration onto complex surfaces. This project directly supports the Irving Energy Institute’s mission to develop and deploy innovative technologies for sustainable energy solutions, as efficient waste heat recovery from data centers can reduce grid reliance and advance decentralized energy solutions. 
 

From Chip-Level Computing to Global Decarbonization: Multi-Scale Decision Support for Data Centers

PI: Erin Mayfield, Hodgson Family Assistant Professor of Engineering

Project Team: Anthony Rizzo, Assistant Professor of Engineering; Junbo Zhao, Todd M. Cook and Elizabeth Donohoe Cook Associate Professor of Engineering; and Cong Chen, Assistant Professor of Engineering

Abstract: Artificial intelligence is rapidly transforming computing infrastructure and driving unprecedented growth in electricity demand from data centers. As this expansion accelerates, decisions about computing hardware, facility operations, energy infrastructure, and public policy are becoming increasingly interconnected. However, research and governance remain fragmented across disciplines, sectors, and scales, making it difficult to evaluate tradeoffs and develop coordinated solutions.

This award will support a multidisciplinary effort at Dartmouth to convene leaders from industry, electric utilities, government, academia, and community organizations to map the emerging AI–energy decision landscape. Workshop participants will identify the key objectives, decisions, constraints, and uncertainties associated with the rapid growth of AI infrastructure, spanning topics from hardware innovation and workload scheduling to electricity planning, transmission, water use, and community impacts.

The workshop will foster new partnerships across sectors while helping establish a shared framework for understanding how decisions made at one scale influence outcomes at others. By bringing together diverse perspectives, the project will identify critical knowledge gaps, data needs, and priorities for future research at the intersection of artificial intelligence and energy systems.

Insights from the workshop will be synthesized into a perspective article that articulates a cross-disciplinary research agenda for the AI–energy nexus. The resulting framework will help guide future research and decision-making by providing a transparent conceptual map of the technological, economic, environmental, and societal tradeoffs associated with AI infrastructure. The project lays the foundation for evidence-based collaboration among researchers, industry, policymakers, and communities as they navigate one of the defining infrastructure challenges of the coming decade.
 

ATLAS: Arrayed Technology for Localized Atmospheric Sensing

PI: Katherine Mirica, Associate Professor of Chemistry

Project Team: Wei Ouyang, Assistant Professor of Engineering

Abstract: Air pollution from transportation, industrial activities, aging energy infrastructure, and increasingly frequent wildfire events poses significant risks to human health and the environment. However, existing air-quality monitoring networks rely on sparse, centralized stations that cannot capture the large spatial and temporal variations in pollutant exposure experienced by individuals and communities. This project aims to develop ATLAS (Arrayed Technology for Localized Atmospheric Sensing), a new platform for distributed atmospheric monitoring. Based on conductive metal–organic framework (cMOF) sensor arrays integrated with low-power wireless electronics, ATLAS relies on data analysis using machine learning (ML) to achieve distributed air quality monitoring. The project will establish the technical foundation for scalable sensing networks capable of detecting and differentiating multiple gaseous pollutants, including nitrogen oxides, carbon monoxide, sulfur dioxide, hydrogen sulfide, and ammonia under dynamically changing environmental conditions. Specifically, the project will (1) determine the environmental robustness of cMOF-based gas sensors under varying temperature and humidity, (2) develop machine-learning frameworks for multi-analyte discrimination in complex atmospheric mixtures, and (3) integrate optimized sensor arrays into portable, low-power wireless sensing nodes for real-world deployment. Together, these aims will establish the feasibility of environmentally robust, ML-enabled sensor nodes capable of supporting distributed air-quality monitoring networks. This project has the potential to establish a platform for distributed air-quality monitoring capable of providing high-resolution, localized pollution data that can improve public health responses, environmental management, and climate-resilient infrastructure planning.
 

Glacial interventions to slow sea level rise

PI: Yoshihiro Nakayama, Assistant Professor of Engineering

Project Team: Chris Polashenski, Adjunct Professor of Engineering; Shuntaro Hyogo, Visiting Researcher, Thayer School of Engineering; Mathieu Morlighem, Evans Family Distinguished Professor of Earth Sciences; and John Moore, Research Professor, University of Lapland

Abstract: Warm ocean water flowing into Antarctic ice shelves drives ice loss from Antarctica, contributing to global sea-level rise, representing a significant energy and society challenge. To help protect global coastal regions hosting critical energy infrastructure and to buy time for the energy transition, the “curtain” concept is proposed. This concept is designed to block warm water flowing towards Antarctica but remains largely untested. Previous modeling studies rely on extremely simplified assumptions, including the lack of ocean simulations and neglect of coupled ice–ocean interactions, leading to substantial uncertainty. We will quantify the feasibility and economic impact of this concept by conducting simulations under present and future conditions, explicitly resolving ice–ocean coupled processes, with and without curtains. This work will transform a speculative geoengineering idea into a testable framework, positioning Dartmouth and the Irving Institute for Energy and Society at the forefront of climate intervention research.
 

The Patient-to-Power (P2P) Protocol: A Transferable, Surge-Coupled Framework for Critical-Facility Energy Management During Compound Outage Events

PI: Alaa Selim, Postdoctoral Researcher, Thayer School of Engineering & Irving Institute for Energy and Society

Project Team: Katie Lichter, Principal Investigator, Radiation Oncology & Health Policy, Geisel School of Medicine; Junbo Zhao, Todd M. Cook & Elizabeth Donohoe Cook Associate Professor of Engineering

Abstract: Severe winter storms in New England can place hospitals under two pressures at once: more patients may arrive at the emergency department and require admission while power outages reduce access to reliable electricity. The Patient-to-Power (P2P) project is developing an agentic AI framework and interactive dashboard to help urban and rural hospitals prepare for and respond to these events.

P2P brings together information on emergency department activity, patient admissions, hospital operations, weather, power outages, and available backup energy resources. Specialized AI agents will help organize and analyze these different sources of information and support coordinated decision-making through an interactive dashboard. Hospital teams will be able to assess which critical services should remain powered, how generators and batteries should be used, where electricity use can be safely reduced, and which services should be restored first.

The project will initially deploy and evaluate the framework with urban and rural hospitals in New England, including Dartmouth Health settings, using historical winter-storm conditions and simulated outage scenarios. National datasets will also be used to provide broader context and support future transfer of the approach to other healthcare facilities. The goal is to create a practical and transferable platform that helps hospitals maintain essential patient services and make better use of limited energy resources during prolonged power outages.
 

Shining Light on Plastic Waste: Unraveling the Mechanisms of Photoredox-Driven Polymer Upcycling

PI: Mahima Sneha, Assistant Professor of Chemistry

Project Team: Jayda Kellogg, Chemistry PhD candidate; and Matthew Salinas, Chemistry PhD student

Abstract: Plastic is everywhere, but it comes with a hidden cost. Only about 9% of the plastic we throw away is ever recycled. The rest piles up in landfills and breaks into tiny fragments that end up in our oceans, soil, and food. Plastic also wastes an enormous amount of energy, since making it consumes roughly 6% of the world's oil. Polystyrene alone, the material in foam cups and packaging, wastes the energy equivalent of about 10 million barrels of oil every year.

A promising alternative uses ordinary visible light together with a "helper" molecule (a catalyst) to either break polystyrene down into benzoic acid, a useful industrial chemical, or rebuild it into upgraded plastics with new properties. Unlike today's recycling, which needs high heat and yields lower-quality material each cycle, this approach works gently at room temperature and could even run on sunlight. The trouble is that we still don't understand how these reactions actually happen. The crucial steps unfold in less than a trillionth of a second, far too fast to see, so progress has relied largely on trial and error, and the reactions remain inefficient, with fewer than 5% of the light-activated molecules doing useful work.

This project uses ultrafast laser spectroscopy, which lets us directly follow the short-lived molecules by their spectroscopic fingerprints. By tracking each step, what works, what stalls, and why catalysts fail, we can design better catalysts on purpose rather than by guesswork. Our goal is to move this out of the lab and toward real recycling facilities, turning plastic waste into valuable products while cutting oil use, reducing pollution, and easing burdens that fall hardest on underserved communities.
 

Machine Learning for Glacier Intelligence: Data-Driven Tools to Predict Meltwater Dynamics and Inform Water Security Under Climate Change

PI: Aleah Sommers, Lecturer & Research Scientist

Project Team: Mashrekur Rahman, Artificial Intelligence Research Data Science Specialist; and Colin Meyer, Associate Professor of Engineering

Abstract: Warming temperatures and changing precipitation patterns are leading to glacier loss worldwide, impacting sea levels as well as changes in water supply and flood risks. Water flowing underneath glaciers influences how quickly the ice slides by modulating the friction at the bed due to changes in the water pressure. This process of glacier sliding remains a large source of uncertainty in models that predict future sea level changes. Interactions between ice flow and meltwater may also impact local water availability and hazard risk in glaciated regions. Direct observations beneath glaciers are challenging to obtain, but numerical models that represent the relevant physical processes provide insights into the dynamic processes playing out under the ice. We combine the team’s expertise in subglacial hydrology modeling and machine learning (ML) methods to address critical gaps: (1) How can ML capture the evolution of nonlinear meltwater-ice dynamics? (2) Can interpretable ML extract physically meaningful relationships to improve empirical parameterizations in models? (3) Can computationally efficient surrogate models enable rapid scenario exploration for water resource planning or flooding risk assessment? Our overall objective is to develop an ML-powered “Glacier Intelligence” framework integrating physics-informed machine learning, interpretability methods, and LLM-assisted knowledge synthesis to transform how scientists predict and communicate glacier-driven water availability under changing conditions. This project embodies the Irving Institute’s vision of bringing together multidisciplinary researchers and positions Dartmouth at the intersection of AI and climate adaptation, aligned with Irving's pillars on AI/ML and climate resilience.&
 

Solid-State Neutral-Forming Transformers to Accelerate Adoption of Bidirectional EV Charging

PI: Charlie Sullivan, Sue and John Ballard '55 TT'56 Professor of Engineering

Project Team: Allen T. Nguyen, PhD candidate, Electrical Engineering; and Kavi Krsnadas, master's student, Thayer School of Engineering

Abstract: Electric energy storage is critical for accelerating the energy transition.  A growing number batteries installed in electric vehicles (EVs) are often sitting idle, underutilized.  Accessing those batteries for other purposes, generally termed bidirectional charging, is as yet underutilized. This project aims to reduce the cost and improve the efficiency of bidirectional charging hardware, making wider adoption more attractive and feasible.

A challenge that inhibits implementation of bidirectional changing in North America is the “split-phase” 120/240 V configuration used for residential power.  AC power from a typical bidirectional charging system cannot directly support split-phase power, necessitating an additional piece of hardware, a “neutral forming transformer.” These are now constructed using large, low-efficiency, low-frequency autotransformers.  We propose to replace neutral-forming transformers with a new type of low-cost high-efficiency solid-state transformer (SST), reducing costs and improving efficiency for bidirectional charging. 

The new design meets the specific needs of this application using a hybrid-switched capacitor circuit that achieves high efficiency but omits capabilities of other SSTs, such as isolation and regulation, that are not needed here.  By doing only what is needed, we can dramatically reduce complexity, cost and power loss compared to conventional SST designs.

The result will be reduced size and costs for residential installations, helping to accelerate utilization of this growing and underutilized energy storage resource, and facilitating all the benefits of increasing storage on the grid.  The technology developed can also apply to other residential inverters, used with solar or stationary battery energy storage.