Penn SEM image

The advancement of silicon electronic technology is slowing down as the field becomes more mature and the technology scaling approaches the quantum limit. However, there are many current and emerging impactful open areas of research such as digitally assisted mm-wave and THz self-correcting systems, quantum engineering assisted electronics, non-planar transistor based electronic systems, low power mm-wave and THz signal generation, processing, and imaging, machine learning assisted analog and RF systems, as well as analog and RF assisted machine learning systems. At the same time, the field of silicon electronics is rapidly becoming inter-disciplinary to advance further.


Electronic Photonic Co-Design graphic

For a given silicon electronic technology, the performance of the active and passive devices are mainly limited by the parasitic components (such as undesired capacitance and resistance) and the substrate loss. The active device level limitations make the signal generation and processing rather challenging at sub-mm-wave and THz regimes. Also, parasitic components of the active devices together with the limited quality factor of passive components make the design of low phase noise oscillators and efficient power amplifiers rather challenging. Additionally, the loss and limited bandwidth in the high-frequency electrical interconnects, increase the power consumption of chip-to-chip or board-to-board data links significantly. Quantum engineering and innovative new devices (such as non-planar active and passive devices) may overcome some of the active and passive device level challenges and are important and promising open areas of research. Moreover, the performance of some of the RF, mm-wave, and sub-mm-wave systems that are designed in silicon based technologies may be improved through the concept of optically assisted electronics. The low loss of the optical medium (such as waveguides), as well as high bandwidth available in optical frequencies can be employed to improve the performance of an existing electronic system. Low-loss optical delays and high quality factor resonators available in the photonic integrated circuits can be used to perform optically assisted electrical signal processing. Due to the small size of these photonic resonators, the optically assisted electrical signal processing may be performed in a smaller area compared to an equivalent all-electrical system.

Similarly, use of sophisticated analog, RF, mm-wave, and sub-mm-wave architectures in addition to near-zero incremental cost of transistors in high node technologies can improve the performance of photonic systems. Examples are RF assisted phase noise reduction of semiconductor lasers and electronically controlled laser phased arrays. In the concept of electronic-photonic co-design, combining the advantages of devices, circuits, and architectures in both electronic and photonic domains can profoundly impact both fields resulting in advances in several areas such as communications, signal processing, imaging, and sensing. Under the DARPA DODOS program, we have developed a highly stable optical synthesizer with sub-Hz level resolution across 5THz range. Under this program, we have also demonstrated the first fully integrated Pound Drever Hall (PDH) laser stabilization system in CMOS. Under DARPA MOABB program, we are working towards implementation of 2D optical phased arrays with sub-wavelength element size and spacing. Under NASA Early Stage Innovations program, we are working on implementation of omni-directional free-space near-earth satellite communication links using such optical phased arrays. As an example of photonic assisted electronics, under DARPA photonic assisted integrated ultra-low phase noise clock (Pi-UPNC), we are working towards highly-stable spectrally pure RF signal generation using highly stabilized phase locked lasers.



With the rapid growth of social media, video streaming, and cloud services, we are witnessing an ever-increasing capacity demand from the inter-rack links to the global IP network. This requires the next generation transceivers to be low cost, highly scalable, and energy efficient. All-silicon transceivers have the potential to be mass-produced at low costs. By leveraging the CMOS-compatible silicon photonics and the advanced electronic processes, we are developing optical transceivers for next generation networks. The co-design of the photonic devices and mm-wave electronic circuitry enables realization of high-speed low-power and low-cost integrated transceivers. Under DARPA PIPES program, we are working on implementation of a monolithic electronic-photonic massive link that provides extreme scalability for data communication by direct integration of a frequency comb into silicon-photonics-on-CMOS platform aiming for 100 Tb/s data-rates and beyond.



Machine learning for on-chip neural signal processing-Brain-machine-interface (BMI) plays a significant role in understanding the brain and treating diseases such as epilepsy, Parkinson’s, Alzheimer’s, etc. Next-generation BMI requires an ultra-low power chip to precisely process neural signals and decrease the wireless data bandwidth thus enabling high recording channel-count while preventing tissue damage caused by heat dissipation. A specific processing task performed on extra-cellularly recorded neural signal is spike sorting, which distinguishes the firing activities of multiple neurons based on the difference between their electrical waveforms. By integrating all of the essential blocks on a CMOS chip, including signal conditioning, feature extraction and classification, we are working toward a self-contained spike sorter that bridges between the raw extra-cellular neural signals and the single neuron activities. Featuring a novel classification algorithm and its analog implementation, the spike sorter chip is expected to achieve high accuracy at an ultra-low power consumption.

Rapid in-flow cell analysis-We are developing on-chip sensor arrays for rapid in-flow analysis of cells and nanoscale biological particles. The high-speed processing capabilities of integrated electronic and photonic circuits allow us to parallelize operation of the sensing array elements. This is critical in breaking the tradeoff between sensitivity and throughput hence improving the effectiveness of these biosensors as tools for fundamental biological studies and clinical monitoring of disease. We are exploring heterogeneous integration of these circuits with microfluidics and 2D material sensors to further enhance sensitivity and speed.



Near-zero incremental cost of transistors in advanced silicon processes with operation frequencies approaching THz regime opens up many impactful and exciting research topics that may result in new applications in detection and sensing, communication, and bio-technology. Meanwhile, despite important achievements in battery technology, the recent trends show that the advancements in battery technology have been significantly slower than that of silicon electronic technologies. At the same time, the market demands more functionalities and better performance from new electronic devices making it essential for RF and mm-wave engineers to come up with new components, techniques, and architectures to improve the performance while reducing the power consumption. Under INTEL center for wireless autonomous systems (WAS), we are implementing ultra-low power mm-wave drone-drone communication links capable of rapid beam tracking and alignment.