Public Abstract

DE-SC0021118: Center for 3-Dimensional Ferroelectric Microelectronics Manufacturing (3DFEM2)

Award Status: Active

Institution: The Pennsylvania State University, University Park, PA
UEI: NPM2J7MSCF61
DUNS: 003403953

Most Recent Award Date: 09/19/2024
Number of Support Periods: 5
PM: Holder, Aaron

Current Budget Period: 08/01/2024 - 07/31/2025
Current Project Period: 08/01/2024 - 07/31/2028
PI: Trolier-McKinstry, Susan

Supplement Budget Period: N/A

Public Abstract

A multi-university team including The Pennsylvania State University, Carnegie Mellon University, Georgia Institute of Technology, The University of Notre Dame, University of Maryland, University of Pennsylvania, University of Tennessee – Knoxville, and University of Virginia, in partnership with Sandia National Labs, Brookhaven National Lab, and Oak Ridge National Lab will establish a “Center for 3D Ferroelectric Microelectronics Manufacturing (3DFeM²).” 3DFeM² will enable a radical enhancement in interconnection between memory and logic, along with order of magnitude reduction in the energy cost of computation. This will be realized by developing the fundamental manufacturing science that assists the semiconductor transition to Industry 4.0, a needed step for integrating robust ferroelectric materials that show outstanding properties throughout the device’s lifetime. The proposed 3D integrated ferroelectrics–deposited on geometrically complex surfaces with realistic back-end-of-the-line (BEOL) metallization and dielectrics–will naturally fuse memory and computation and exploit the 3^rd dimension, unleashing unprecedented capabilities for the next generation of computation and artificial intelligence. 3DFeM² is built around two overarching hypotheses intended to facilitate development of advanced manufacturing processes for new ferroelectrics for non-volatile memory: 1. Understanding domain walls, defects, interfaces, and intermediate switching structures of next-generation ferroelectrics at multiple length-scales will enable the community to engineer their coercive field, leakage current, and endurance characteristics and to discover new ferroelectrics. 2. Combining machine learning, in situ monitoring during synthesis and etch, and in operando device measurements will advance manufacturing and expedite co-development of novel materials into 3D integrated systems with high yields, while creating a blueprint for future process innovations. The proposed 3DFeM² will achieve these revolutionary advances using a center-level effort where materials discovery, manufacturing science, and integration science are co-informed. Reciprocal information paths (both top-down and bottom-up) ensure that basic science in all thrusts propels long-term center goals. New materials will be developed in the context of specific non-von Neumann computer architecture property needs, and the role of defects will be understood in terms of functionality and lifetime, while new process flows enabled by machine learning (ML) and artificial intelligence will expand horizons with unprecendented access to monolithic vertical integration. We will train a diverse, new generation of ML-literate scientists. 3DFeM² offers the possibility to revolutionize the manufacturing of information hardware; the challenges are substantial, and the impact warrants an EFRC effort.

A multi-university team including The Pennsylvania State University, Carnegie Mellon University, Georgia Institute of Technology, The University of Notre Dame, University of Maryland, University of Pennsylvania, University of Tennessee – Knoxville, and University of Virginia, in partnership with Sandia National Labs, Brookhaven National Lab, and Oak Ridge National Lab will establish a “Center for 3D Ferroelectric Microelectronics Manufacturing (3DFeM²).”

3DFeM² will enable a radical enhancement in interconnection between memory and logic, along with order of magnitude reduction in the energy cost of computation. This will be realized by developing the fundamental manufacturing science that assists the semiconductor transition to Industry 4.0, a needed step for integrating robust ferroelectric materials that show outstanding properties throughout the device’s lifetime. The proposed 3D integrated ferroelectrics–deposited on geometrically complex surfaces with realistic back-end-of-the-line (BEOL) metallization and dielectrics–will naturally fuse memory and computation and exploit the 3^rd dimension, unleashing unprecedented capabilities for the next generation of computation and artificial intelligence.

3DFeM² is built around two overarching hypotheses intended to facilitate development of advanced manufacturing processes for new ferroelectrics for non-volatile memory:

1. Understanding domain walls, defects, interfaces, and intermediate switching structures of next-generation ferroelectrics at multiple length-scales will enable the community to engineer their coercive field, leakage current, and endurance characteristics and to discover new ferroelectrics.

2. Combining machine learning, in situ monitoring during synthesis and etch, and in operando device measurements will advance manufacturing and expedite co-development of novel materials into 3D integrated systems with high yields, while creating a blueprint for future process innovations.

The proposed 3DFeM² will achieve these revolutionary advances using a center-level effort where materials discovery, manufacturing science, and integration science are co-informed. Reciprocal information paths (both top-down and bottom-up) ensure that basic science in all thrusts propels long-term center goals. New materials will be developed in the context of specific non-von Neumann computer architecture property needs, and the role of defects will be understood in terms of functionality and lifetime, while new process flows enabled by machine learning (ML) and artificial intelligence will expand horizons with unprecendented access to monolithic vertical integration. We will train a diverse, new generation of ML-literate scientists. 3DFeM² offers the possibility to revolutionize the manufacturing of information hardware; the challenges are substantial, and the impact warrants an EFRC effort.