UC San Diego

Lithium-ion battery technology has rapidly developed in the last several decades with the most widely used materials based on intercalation chemistry for both anode and cathode. However, the intercalation chemistry is limited by energy density. To overcome this energy density limitation, alloy-based anode materials like Silicon, Germanium, and Tin are being explored for high-demand transportation applications, including heavy-duty vehicles (trucks and trailers), marine vessels (ships and yachts), and aerial vehicles (drones and air taxis). Among them, Silicon stands out because of its low cost and relative abundance within the earth’s crust. However, there are major hurdles before it can be commercialized due to problems related to high-volume expansion and contraction which limit its use. To address these issues, researchers have proposed many strategies and developed techniques to understand the benefits of these strategies. Most of these techniques are either qualitative or semi-quantitative. In this work, we use Titration Gas Chromatography (TGC), a quantitative tool developed in our previous work for Lithium metal anode to be used for Silicon anode. Using the TGC tool, we quantitatively demonstrated that silicon-based anode failure is influenced by trapped Li-Si alloy and Solid Electrolyte Interphase (SEI), challenging the conventional belief that SEI formation alone is responsible. The amount of formation of trapped Li-Si and SEI depended on the binder's mechanical properties (Polyacrylic acid (PAA) vs Carboxymethyl Cellulose (CMC-Na)) and state of charge (N/P ratio control). We extended the TGC analysis to evaluate ADVANO Inc.'s commercial innovation: a spherical Si-C composite material with a distinctive 'jackfruit-like' structure, manufactured from recycled semiconductor and solar silicon waste. This approach to manufacturing is particularly significant given projections of 80 million tons of solar panel waste by 2050. The composite material achieves impressive performance metrics, including ~3mAh/cm2 areal capacity and 80% capacity retention after 100 cycles with LFP cathode, while maintaining cost competitiveness with commercial graphite anodes. The designed jackfruit-like structure reduces the surface area of active material exposed to the electrolyte and maintains a good electronic pathway. Additionally, the carbon matrix acts as a cushion to dissipate the stress due to the Si volume expansion and contraction upon cycling. The Si-C composite anode with a smaller primary particle size had higher capacity retention and average coulombic efficiency upon cycling compared to the Si-C composite anode with a larger primary particle size. The TGC and the EIS results show that the improved cycling stability for smaller primary particle sizes could be attributed to the reduced lithium inventory loss and lower impedance growth resulting from small Si particle size and improved stress dissipation. Another area where large energy consumption happens is in the field of computing, especially due to the emergence of Artificial Intelligence (AI). A large portion of this energy consumed is wasted due to the current architecture of computing known as Von-Neumann Architecture in which data is stored and processed in two separate units instead of one. A proposed solution to overcome this issue is a new architecture called neuromorphic architecture, which is inspired by the human brain. The data is stored and processed in one unit, thereby being energy efficient. In this work, we propose a lithium ion-based transistor consisting of Lithium Titanium Oxide (LTO) as the channel, Lithium Phosphorous Oxynitride (LiPON) as the Lithium-ion source, Copper (Cu) as the gate terminal, and Gold (Au) as the source and drain for neuromorphic computing application. Our RF-sputtered LTO films demonstrated a six-order magnitude increase in conductivity upon lithiation, plateauing at 20%. Both theoretical calculations and experimental methods confirmed the material's insulator-to-conductor transformation. The device exhibited synaptic-like behavior with an asymmetric ratio of 1.425 and Gmax/Gmin ratio of 7.83. When implemented in a deep neural network (DNN) for handwritten digit recognition, it achieved 92.03% accuracy. A hypothesis was proposed to discuss the working mechanism of the device, which revealed the formation of oxygen vacancies, as well as the lithium insertion and de-insertion near/at the interface, which is responsible for the observed transfer characteristics of the device. This understanding can potentially give rise to greater optimization strategies for device performance in neuromorphic applications.