Our research lies at the intersection of electrochemistry and microelectronics, which are two of the most well-researched topics in materials science and two of the most important technologies. We believe that much can be learned by applying ideas, concepts, and methods from one field onto the other.
Thrust 1: Computing with chemistry-- unleashing a new frontier in microelectronics
Moore's Law and the continued downscaling and improvement of transistors has yielded the fastest pace of technological transformation in human history. Today, as Moore's Law is nearing its physical atomic limits, computing demands are rapidly increasing, especially for data-intensive operations like artificial intelligence (see left). Indeed, over 10% of global electricity is consumed for computing, including processing, storing, and processing data. Alternative technological approaches are needed to satiate the increasing global demand for computing.
[Graphic from Mehonic & Kenyon, Nature, 604, 255 (2022)]
We believe that the next frontier of microelectronics will come from utilizing device chemistry. While synthetic chemical approaches like atomic layer deposition are presently used to synthesize and fabricate semiconductors, semiconducting devices today operate primarily through physical processes such as gating, ferroelectricity, and light emission and generation. By leveraging electrochemistry, which enables the reversible control of chemistry using current and voltage, we enable microelectronic devices to change chemical bonds and electron valence, and therefore embed new functionality. Importantly, we also aim to understand how chemical processes affects the operations of existing electronic devices. We are motivated by the human brain, which also utilize chemical and electrochemical processes, and which are many orders of magnitude more energy efficient than the best digital computers today.
Towards this goal, we leverage fundamental understanding of ionic materials, especially motivated by the development of energy transformation technologies like the Li-ion battery. We follow a rich legacy of cross-pollination between solid-state chemistry and microelectronics. For example, the Nobel Laureates credited for the development of the Li-ion battery all first worked on electronic materials and devices: Stan Whittingham was studying superconductivity in transition metal dichalcogenides; Akira Yoshino was investigating conductive polymers; John Goodenough was already a world-renowned leader in magnetic materials.
Project 1: Oxygen Transport in Memristive Devices
The memristor is a promising next-generation nonvolatile memory and computing device that uses electric fields and joule heating to move point defects like oxygen vacancies. This movement of point defects allow for changes in the resistance state. In recent years, there have been substantial research in using memristors not just for high-density information storage, but also for highly-efficient memory-based computing. As shown below, memristive device switch between low and high resistance states with the formation and dissolution of conducting filmaent.
Memristive devices witch between the low and high resistance state upon applying a voltage.
In the low resistance state, or 1, oxygen vacancies form a conducting filament.
In the high-resistance state, or 0, the filament is broken due to the electric field, resulting in much higher resistance.
Our research in this area is more fundamental: instead of aiming to build the best memristive devices, architectures, and systems, we want to understand how such devices work. We are currently asking one of the most fundamental questions: why are memristors nonvolatile? In general, it is believed that memristors are nonvolatile because oxygen vacancies are too slow to diffuse at room temperatures. However, if we compute the diffusion time based on experimental parameters, we expect retention time of about 1 hour. Instead, many researchers have built memristors that retain state for 10 years.
We hypothesize that the existing model that memristors are nonvolatile due to slow oxygen diffusion may not be correct. This model was modified from the drift-diffusion equation in semiconductor device physics. We believe that oxygen transport in memristors have very different properties than electron and holes in silicon. Instead, other interactions like phase separation become much more important.
Project 2: Electrochemical Random-Access Memory
Electrochemical random-access memory is a three-terminal memory technology that utilizes the electrochemical motion of ions to store information states. By combining two mixed ionic and electronic conductors and a solid electrolyte, the ECRAM cell is able to precisely tune the oxygen vacancy concentration in the "switching layer," or channel. This yields the opportunity to store analog information states in a nonvolatile fashion. In addition to oxygen vacancies, we can also use protons and copper ions.
While we envisioned ECRAM as an analog "synapse" for in-memory computing or neuromorphic computing, our recent developments show that this cell is highly promising as logic and memory architectures in extreme environments including high temperatures and radiation environments.
ECRAM cells switch analog resistance states through the electrochemical modulation of oxygen vacancies.
Experimental demonstration of analog resistance states enabled by ECRAM
Nonvolatile Electrochemical Random-Access Memory Under Short Circuit. Advanced Electronic Materials 9, 2200958 (2023; open access)
ECRAM Materials, Devices, Circuits and Architectures: A Perspective Advanced Materials (2022)
Filament‐Free Bulk Resistive Memory Enables Deterministic Analogue Switching, Advanced Materials, 32, 2003984 (2020)
Thrust 2: Heterogeneity and variability in Li-ion batteries
Li-ion batteries are a crucial technology in our transition to a clean energy future. These batteries are comprised of micron-sized active materials that are able to store charge. While such particles appear similar in size, shape, and morphology, they often exhibit very different properties with regards to performance. Various nanoscale probes have shown that battery particles charge heterogeneously and nonuniformly. One example, done by Yiyang during his PhD, is shown on the right. However, neither the origins for the extent of such variability has been clear. The US Department of Energy has identified unraveling this heterogeneity as one of the important Basic Research Needs for energy storage.
While nanoscale probes have revealed much about the structure and chemistry of batteries with exceptional spatial resolution, quantitative measures of electrochemical current and voltage at this length scale has been missing. To achieve this goal, we borrow an approach from microelectronics and treat each battery particle as a "device." Therefore, rather than charging and discharging millions of battery particles at once, we are able to charge and discharge a single battery particle at a time. We have built a platform that isolates the electrochemical currents on a single battery particle. By doing such high-sensitivity, single-particle measurements with high throughput, we can start to understand how and why individual battery particles are different from one another.
Most research on Li-ion batteries measure the average electrochemical properties of many battery particles, constructed into a porous battery electrode like the one above. Image credit: Dr. Werner Bauer at KIT.
We construct a battery made up of a single particle. This enables us to quantify the intrinsic properties, as well as understand how individual particles are different from one another. Image credit: Mr. Jinhong Min at UMich.
Even though our electrodes comprise of only a single particle, we have been able to successfully charge and discharge them. As shown below, we can replicate the standard current-voltage shape of a Li(Ni,Mn,Co)O2 battery particle with a capcity of about 100 pico-amp hours. Moreover, by measuring the capacity of many particles, we found a direct correlation between the measured capacity and the volume estimated from the SEM images. This suggests that our platform is highly robust, and that there is little difference in the capacity of individual battery particles. In the future, we aim to measure the differences in the electrochemical resistance and degradation rate of single battery particles.