Collaborative Research: VINES: Track 1: AI-Enhanced Reconfigurable and Scalable Terahertz CMOS Transceiver Panel for NextG Communication Systems via Multidimensional Integration

Every year, more phones, cars, and smart devices try to send information through the air at the same time, and the invisible radio waves that carry this information are becoming crowded, which slows communication down. This collaborative project brings together investigators from the University of Washington, Texas A&M University, and the University of Utah to build a new kind of wireless system that uses radio waves vibrating about 300 billion times each second. These waves travel through a part of the electromagnetic spectrum that is wide open and mostly unused, much like a new superhighway next to a congested road, letting large amounts of information move quickly and securely. The team will design tiny computer chips that work together like tiles in a large panel to aim beams of radio waves toward their destination. The system will also use artificial intelligence to adjust itself to bad weather or to attempts by adversaries to listen in on the signal (eavesdropping) or block it (jamming). In terms of broader impacts, this work strengthens United States competitiveness in semiconductors and next-generation wireless communication and benefits smart cities, intelligent transportation, and augmented reality. The project also trains doctoral, undergraduate, and community-college students at the partner universities, creates a publicly available course on terahertz integrated systems, and supports science, technology, engineering, and mathematics outreach for young students. The intellectual merit lies in a research plan, targeting the 300 gigahertz frequency band, that is organized into three thrusts. Thrust one designs a scalable and reconfigurable silicon bipolar complementary metal-oxide-semiconductor (BiCMOS) transceiver panel that performs electronic beam steering through phased-array antennas and provides dynamic polarization control for adaptation to different weather conditions and for added physical-layer security. Transceiver chips across the panel are synchronized by injection locking at half the carrier frequency through printed electrohydrodynamic-jet interconnects. A packaging effort combines extrusion-based dielectric planarization with electrohydrodynamic-jet printing to fabricate dense, low-parasitic die-to-die connections characterized up to 300 gigahertz. Thrust two implements frequency hopping spread spectrum (FHSS) modulation by self-injection locking of silicon voltage-controlled oscillators to high-quality-factor terahertz resonators. This thrust investigates two resonator forms fabricated on high-resistivity silicon wafers: dielectric ring resonators and micromachined waveguide ring cavity resonators. These resonators are coupled to the terahertz chips through chip-to-waveguide coupling methods based on on-chip dielectric waveguide couplers and micromachined probe transitions. Self-injection locking to high-quality-factor resonators improves the phase noise, frequency stability, and efficiency of terahertz circuits. Thrust three develops an artificial intelligence (AI) and machine learning (ML) backbone that applies convolutional neural networks (CNNs), deep neural networks (DNNs), and the K-nearest neighbors (KNN) method. This backbone enables agile beamforming for smart phased arrays, spectrum sensing that predicts frequency hopping patterns for dynamic spectrum sharing, and adaptive data recovery that reconfigures constellation boundaries under changing channel and weather conditions. Evaluation proceeds through electromagnetic and circuit co-simulation followed by staged laboratory demonstration of the transceiver array, benchmarking oscillator phase noise, efficiency, and model accuracy. The outcome is a robust terahertz communication link adaptable to adverse and evolving environments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.