High-performance stretchable thin-film transistor and design approach to mitigating strain effect: Electronics with the ability to be bent, stretched, compressed, and deformed into arbitrary shapes, can enable a new paradigm of many transformative applications, ranging from human-machine interface to health monitoring and conformal displays. Intrinsically stretchable technology, which leverages elastomeric electronic materials, offers a viable pathway to realizing this paradigm.
Large-area thin-film-transistors (TFTs) play a central role in intrinsically stretchable electronics. Recently, large-area, stretchable carbon nanotube thin-film transistors (CNT-TFTs) have pushed the channel length down to 10 µm, while maintaining filed-effect mobility of 14 cm2/V∙s . This performance enhancement brings the technology from the sub-kilo-Hertz to the mega-Hertz domain, opening a new design space for sensing and display applications.
One remaining critical challenge is the considerable effect of strain deformation on device characteristics, particularly on transistor on-current and transconductance . Moreover, this effect is highly dependent on the direction in which the strain is applied. In practice, with strains being both unpredictable and subject to change over time, this poses a significant hurdle in the circuit and system design for stretchable technologies. In this work, we investigate the impact of strain on device performance and demonstrate a circular transistor design to effectively mitigate the effects of strain on electrical characteristics. The strain-induced change in on-current is significantly reduced from 56% to 3% at 30% strain. The finite-element simulation shows this is achieved by counteracting the performance variations observed across different segments of a 360o transistor channel. [Nature 24] [IEDM 23 (top ranked student paper)] [TED 23 (invited)]
Gigahertz large-aperture phased array: Large-aperture electromagnetic phased arrays can provide directionally controlled radiation signals for use in applications such as communications, imaging and power delivery. However, their deployment is challenging due to the lack of an electronic technology capable of spanning large physical dimensions. Furthermore, applications in areas such as aviation, the Internet of Things and healthcare require conformal devices that can operate on shaped surfaces. Large-area electronics technology could be used to create low-cost, large-scale, flexible electromagnetic phased arrays, but it employs low-temperature processing that limits device- and system-level performance at high frequencies. In this work, we boost TFT fT and fMAX from mega-Hertz to giga-Hertz regime by a self-aligned fabrication technique. Using the advanced TFTs, we create a three-element phased array operating at giga-Hertz, based on a phase-tunable oscillator that overcomes the fT limitation by injection locking. [Nature Electronics 21] [DRC 19 (best student paper)]
Reconfigurable antenna enabled by gigahertz large-area electronics RF switches: Future IoT and 6G envisions a large number of widely and densely distributed sensing devices, accessed using low-power gigahertz radios. The dense and spatial nature of the network requires several critical functionalities for wireless sensing, including direction finding, addressing and transmission, as well as flexible spectrum allocation and polarization control. Reconfigurable antennas are capable of such agility, but limited by assembly of discrete RF components across large apertures typically desired. This work presents the first monolithically-integrable reconfigurable antenna operating at 2.4 GHz. A key enabler is efficient gigahertz RF switches based on large-area electronics by material, device, and circuit co-designs. [JSSC 23] [IEDM 20]
2.4 GHz backscattering tag with beamforming capability: Future 5G/6G and Internet-of-Things (IoT) networks envision dense, large-scale deployment of wireless nodes. Low-maintenance and battery-free operation is essential to its success. Passive backscattering can achieve zero-power wireless communication, by piggybacking information on reflected radiation generated from a base station. However, significant free-space power loss makes the wireless range of reflected signals inadequate for many practical applications. In this work, we take advantage of large-area electronics (LAE), to substantially enhance reflected power and thereby the wireless range. The designed system based on the geometric attributes of a Van Atta array ensures that reflected signals from any direction are always beamformed back in the direction of the incident signal, while the TFT circuits enable wireless data modulation. A 4-element passive backscattering tag, based on the 2.4 GHz LAE RF resonant switches, is demonstrated with beamforming and frequency-modulated wireless communication capability. This work sets the stage for future ubiquitous, monolithically-fabricated, and low-cost IoT sensing nodes that can be seamlessly integrated with everyday objects, enabled by LAE’s large-area and flexible form factors. [Under review]
Large-area image-sensing and feature-compression system: Large-area electronics (LAE) is compatible with broad ranges of materials, enabling the formation of diverse sensors on flexible substrates. However, low-temperature processing results in LAE thin-film transistor (TFT) performance and energy efficiency being orders of magnitude worse than Si ICs. Thus, to process over the diverse and large arrays of flexible sensors, a hybrid-system approach is required, combining LAE with Si IC. Unfortunately, as the number and expanse of sensors have exploded, the physical interfacing between these two technologies now becomes the bottleneck of system power and scalability. This work reduces the number of interfaces, demonstrated through an intelligent system that performs “inference” using data from large number of sensors. We perform compression of data by in-sensor computing in the LAE domain with TFTs. We demonstrate it on image recognition, using an array of photo-sensing pixels which will be further compressed through TFTs, and then followed by a machine-learning classifier. For 10-way classification of 0-9 digit image data (MNIST), the system achieves 93.5% accuracy. [FLEX 16 (best student poster)]