2026 NanoSI Industry Engagement Board (IEB) Meeting

ABSTRACT:

Modern wireless communication systems require radio transmitters that achieve high energy efficiency while supporting wide bandwidth and spectrally efficient modulation schemes. However, conventional power amplifier (PA) architectures often suffer significant efficiency degradation under power back-off, which is common in high peak-to-average power ratio (PAPR) signals.

This talk presents recent advances in load-modulation-based PA architectures that improve efficiency across a wide output power range. Several architectures developed in our work will be introduced, including the Load-Modulated Balanced Amplifier (LMBA) and its variants, the Inverse-Balun Load-Modulated Amplifier (IBMA), and the Load-Modulated Double-Balanced Amplifier (LMDBA). These designs leverage multi-port passive networks and broadband load modulation to maintain high efficiency over wide bandwidths and deep back-off levels.

Design principles, theoretical insights, and experimental prototypes will be discussed, demonstrating the potential of these architectures for energy- and spectrum-efficient wireless transmitters in future communication systems.

ABSTRACT:

This work proposes a new type of PO-based analog computer, named Analog Floquet Solver (AFS), which usurps the computational limitation of the gradient descent energy minimization exploited by PO-based Ising machines. In AFS, each PO is transformed into a Floquet node via coupling to a high-Q resonant device. This coupling permits the POs to passively activate Floquet states that introduce modulations in the POs’ amplitudes and the system’s energy. Such modulations permit the AFS to escape local minima and suppress the deleterious impact of amplitude heterogeneity on the system’s minimization dynamics. These features result in substantial improvements in computational metrics like probability-of-success and time-to-solution compared to conventional PO-based Ising machines.

ABSTRACT:

A 265 nW, high sensitivity (99.0%), high specificity (99.9%) analog computing machine learning (ML) system-on-chip (SoC) for  EEG seizure detection is presented. This ML SoC reduces the power by over 6× by performing all computation in analog before digitization. The SoC’s subsystems are tolerant to process-temperature-voltage (PTV) variations and remain robust across temperature with a 2% degradation in sensitivity for an 80 °C change. This robustness is achieved through circuit-level design techniques, such as constant transconductance biasing. The SoC can be used for both 1D-CNN (Convolutional Neural Network) and SVM (Support Vector Machine) models, making it adaptable to a broad range of machine-learning-based biomedical applications.

ABSTRACT:

Recent advances in metasurfaces have opened new opportunities for controlling thermal radiation and realizing compact mid-infrared light sources. Using bound states in the continuum (BICs), metasurfaces producing coherent and spectrally selective thermal emission are demonstrated. After presenting BIC-supported metasurfaces made of sputter-deposited silicon, a bonding-based fabrication approach integrating monocrystalline silicon metasurfaces with silicon carbide substrates is described. This novel fabrication process significantly reduces optical losses and enables record-high experimentally measured Q-factors for mid-IR thermal emitters. Finally, a tunable architecture combining BIC metasurfaces with MEMS actuation to dynamically control the emission wavelength while maintaining extremely high Q-factors will be introduced. These results highlight a path toward compact, high-coherence mid-IR sources for spectroscopy, sensing, and integrated photonic systems.

ABSTRACT:

TBA

ABSTRACT:

TBA

ABSTRACT:

Our Optical Microresonators (OMRs) convert electrical signals into optical modulation via the piezoelectric effect. Using optical geometry-guided mode matching, a 20 µm OMR achieves more than three orders of magnitude improvement in modulation sensitivity over prior work, while a 15 µm device enables near-lossless fiber coupling for shot-noise-limited readout. We demonstrate a fiber-bonded assembly that enables optical E-field sensing with sensitivity limited by photodetector shot noise.