This project aims to develop a new class of material that does not exist naturally - high-entropy oxide (HEO) films, and explore its application in memory and neuromorphic computing systems. The team hypothesizes that by mixing a diverse group of oxides together, the increase in entropy can help stabilize the film during growth and prevent nonhomogeneity effects that typically occur in multi-oxide film growth. The ability to control the material composition can in turn lead to desired ?cocktail? effects and optimized resistive switching (RS) properties that cannot be obtained from simple oxide systems. This approach can allow systematic tuning of the material parameters for a broad range of applications including resistive random-access memory (RRAM) and neuromorphic computing systems.

Three well-defined and tightly integrated tasks have been carried out, following the proposed research plan. Task 1: first principles simulation of the HEO systems to reveal how the different material parameters affect the RS process, and how these parameters can be systematically tuned by controlling the composition of the HEO systems; Task 2: growth of the HEO films and characterization of the film composition and microstructure; and Task 3: device fabrication and characterization to experimentally verify systematic tuning of the RS behaviors, and prototype memory device and neuromorphic computing hardware developments.

During the past 4 years, the PIs and their students have made significant progress in the proposed tasks, and obtained the following outcomes:

1)     We performed hybrid density-functional theory (DFT) calculations to obtain the electronic structure of amorphous HEOs with varied composition. We analyzed the effects of stoichiometry on the cation charge state, which explains the experimental results. We also combined high-throughput DFT calculations with linear regression algorithms to investigate the effects of local configurational and structural disorder on the thermodynamics of vacancy formation in HEOs and their influence on the electrical properties. Our results demonstrate that tuning the local chemistry and associated structural distortions by varying alloy composition acts an engineering principle that enables controlled defect formation in multi-component alloys. These results were reported in J. Lee et al. ACS Appl. Mater. Interfaces, 11, 11579-11586 (2019), and S. Chae, et al. NPJ Computational Materials 8, 95 (2022).

2)     We successfully demonstrated the growth of HEO materials composed of Zr, Hf, Nb, Ta, Mo, and W oxides. This multielement oxide material provides uniform distribution and higher concentration of oxygen vacancies, limiting the stochastic behavior in typical RS devices. Memory devices based on these materials manifest the ?cocktail effect?, exhibiting desired retention and gradual conductance modulation. Electrical characterization of these high-entropy-oxide-based memristors demonstrates forming-free operation, low device and cycle variability, gradual conductance modulation, 6-bit operation, and long retention which are promising for neuromorphic applications. These results were reported in M. Ahn, et al., Advanced Electronic Materials 7, 2001258 (2021).

 3)     We built physics-based models that incorporate detailed ionic drift/diffusion, electrical conduction, and thermal generation/dissipation processes for device modeling and optimization. By leveraging the internal dynamics, we showed the devices can natively implement learning rules such as spike-timing dependent plasticity (STDP) with controlled time-constants, and improved RS characteristics. These results were reported in S. Lee et al., ACS Applied Electronic Materials, 2, 3, 701-709 (2020)

 4)     Leveraging the controlled internal dynamics, we demosntrated prototype neuromorphic computing hardware that can efficiently detect correlation in the input data in an unsupervised fashion, and reservoir computing systems that can efficiently perform time-series analysis. These results were published in S. Yoo et al, Advanced Electronic Materials, 2101025 (2022). 

The outcomes have been broadly disseminated to the research community and the general public. The PIs and their students have published 8 papers in prestigious journals, and given invited talks at major internal conferences on findings of this project. 4 Ph.D. students have been systematically trained based on funding from NSF. The findings have also been incorporated in two graduate-level course the PIs developed, and shared with the general public through publications, presentations, and online resources. Several new research directions have been established, and hardware and software infrastructure has been developed during the project.  The findings have also inspired several new students who have since become actively engaged in research in related topics.

Last Modified: 07/25/2022
Modified by: Wei Lu