Superconducting Electronics:
Superconducting electronics has emerged as a promising alternative to conventional CMOS technology due to its exceptional speed and energy efficiency, offering critical advantages for high-performance computing. Going beyond the classical computing paradigm, superconducting electronics is a crucial component to develop a large-scale quantum computer. In current quantum computers, qubits are maintained at extremely low temperatures (tens of milli-Kelvin) to protect against noise, while peripheral components like the control processor and memory operate at room temperature. This requires extensive amounts of wires and interconnects which adds complexity and limits scalability. Realizing the full potential of quantum computing will require cryogenic-compatible peripheral components that can sit alongside qubits. Superconducting electronics not only meets this demand but also offers potential breakthroughs in improving the efficiency, reliability, and design complexity of spacecraft for aerospace applications. Our work focuses on advancing superconducting electronics by harnessing the unique properties of ferroelectric superconducting quantum interference devices (FE-SQUIDs), superconducting memristors, heater cryotrons, Josephson junctions, and Josephson junction FETs.
VLSI & Nanoelectronics:
For decades, periodic downscaling of transistors, dictated by Moore’s law, has fueled progress of electronic technologies. In continuation to that, we have arrived at the very edge of conventional scaling process and approaching towards the atomic dimension. Quantum phenomena, short channel effects and the fundamental Boltzmann limit (stemming from statistical distribution of free carriers) have become major hindrances against continued improvement of device and circuit performance. At this critical stage, conventional CMOS based electronics is transitioning to the post-CMOS era. Several materials and devices are being explored to supplement/replace existing counterparts to meet performance/power targets. These emerging entities have unique characteristics including new opportunities and challenges. Due to the immense design complexity, secluded device or circuit centric design is not ideal for emergent technologies. Circuit driven device design (and vice versa) is mandatory to best utilize the potentials of these novel materials/devices. We work on device-circuit co-design with emerging technologies to facilitate low-power, reliable VLSI design.
Emerging Memory Devices:
The stupendous increase in the need for data storage for technical and non-technical applications has led to a manifold increase in the need for more reliable and denser memory devices/circuits/arrays. Large scale scientific experiments, space exploration, social media, financial services and interconnected smart devices have been creating data in mind-boggling proportion. Keeping up with the demand, innovation in the memory industry has been strongly accelerated. Non-volatile memory devices have garnered attention due to the increased popularity of portable, wearable, implantable devices and the internet of things. We explore the device/circuit/array level design of emerging memory devices including RRAM, MRAM, PCRAM and ferroelectric memories. We also work on designing and optimizing 3D Cross (X) point memory array with special emphasis on the ‘selector’ design.
Beyond CMOS Technologies:
The quest for next-generation low power devices has triggered a worldwide search for device designs that can achieve < 60 mV/Dec subthreshold swing (known as Boltzmann limit). While this is physically impossible for a Silicon-based conventional transistor, several new physical phenomena have been shown/proposed to be able to enable such performance. Our group contributes to the exploration and development of novel steep switching transistor technologies. We develop the design methodology for these devices and predict their unique circuit-level signatures. We also develop circuit compatible models for these novel devices to facilitate design exploration in higher levels of abstractions.
Topological Electronics:
Topological materials offer remarkable variation robustness and low-power switching capabilities, positioning them as promising candidates for post-CMOS technologies. Our research focuses on developing circuit-compatible, physics-informed compact models for these devices, including Quantum Spin Hall Insulator (QSHI)-based topological transistor, which leverage quantum phenomena like Majorana modes and topologically protected Andreev reflection for robust, energy-efficient operation. Quantum Anomalous Hall Effect (QAHE)-based devices also show potential as scalable, non-volatile memory (NVM) solutions due to their topologically protected non-volatile Hall resistivity. We investigate NVM arrays based on QAHE-driven twisted bilayer graphene (tBLG) moiré heterostructures and explore in-memory computing paradigms such as Binary Content Addressable Memory (BCAM) to enhance AI applications. With these compact models, we enable advanced device-circuit co-design and open new avenues for low-power, high-speed memory and in-memory computing systems that are highly robust against variations.
In-Memory Computing:
Efficiently managing the massive and growing volume of data presents a significant challenge for the electronics industry, as conventional computers rely on constant data transfer between memory and processing units. Google estimates that 20-42% of a system’s energy is spent powering the data bus responsible for these data transfers. Moreover, the speed mismatch between memory and processing units further limits overall system throughput. In-memory computing offers a promising solution by performing computations directly within the memory array, reducing the need for frequent data movement. Our research leverages emerging non-volatile memory systems to develop room-temperature in-memory computing platforms. Additionally, we explore cryogenic in-memory computing, which provides a distinct advantage by reducing cooling costs in cryogenic environments. We have developed in-memory computing systems capable of executing Boolean logic operations, bit-serial addition using majority logic, binary multiplication, content-addressable memory, matrix-vector multiplication, and ternary computing.
Atomistic Simulation:
Atomistic simulations of electronic devices offer a detailed, fundamental understanding of device behavior at the atomic level, enabling us to develop highly accurate models that capture the complex interactions within these devices. This approach allows for the precise prediction of device performance under various conditions, guiding innovations in design and materials. Additionally, atomistic simulations of different structures are crucial for analyzing interfacial thermal and electrical resistance, insights that are essential for optimizing heat dissipation and electrical conductivity in next-generation technologies. Together, these studies advance our capability to create devices with enhanced efficiency, reliability, and scalability.
In-Pixel Processing:
In-pixel processing technology enhances the intelligence of image sensors by performing data processing directly within each pixel, reducing the need to transfer large amounts of data to external processors. This approach minimizes latency and power consumption, making it highly beneficial for applications requiring real-time image analysis, such as autonomous vehicles and surveillance systems. By processing data on-sensor, it also enables high-resolution imaging and computationally intensive tasks without sacrificing performance, opening the door to more efficient, compact, and powerful imaging devices. Keeping this in mind, we design compact circuits capable of performing different operations directly within the pixel circuit, maximizing efficiency and functionality.
Artificial Intelligence & Neuromorphic Hardware:
Modern digital processors are excellent for cumbersome computations and to follow strenuous algorithms. However, the human brain- a marvelous bio-processor, has the unique ability to learn and evolve. Due to the popularity of applications like adaptive virtual assistants (Apple-Siri, Amazon-Alexa, Samsung-Bixby etc.), pattern matching, image recognition and unsupervised data mining; brain-inspired/neuromorphic computing is garnering huge interest. From the hardware point of view, the basic step to implement an efficient neuromorphic system is to establish a circuit equivalent of neurological primitives (neuron and synapse) and optimize/co-design them for specific applications. Phase transition materials and magnetic tunnel junctions are suitable to provide probabilistic/stochastic oscillations and therefore can be utilized to design spiking neurons. In addition, X-point array of RRAMs and FEFET with multi-domain gate stack can provide the spike time dependent plasticity (STDP), observed in the biological synapse. The multi-domain nature of FEFET can be harnessed to create multi-level weights for designing analog synapse, which is more similar to their biological counterparts. We explore hardware primitives for artificial neurons and synaptic networks to realize power-efficient neuromorphic system hardware.
Analog Circuits with Emerging technologies:
In analog circuit applications, phase transition materials (PTMs) like vanadium-oxide (VO₂) and Ag-HfO₂-Pt are gaining attention for their abrupt switching and hysteretic behavior, which enable hybrid PTM-transistor assemblies that surpass the Boltzmann limit of 60 mV/decade. The polarity-dependent hysteresis in these devices enhances rectification and stability, making PTM-based diodes ideal for low-power rectifiers and energy-harvesting circuits. The hybrid PTM-FET structure, or Hyper-FET, further improves low-power digital and analog circuit performance in applications such as sense amplifiers and dynamic logic. We focus on designing PTM-based circuits and optimizing their performance through extensive design-space analysis.
Superconducting nanowire is another steep switching hysteretic device that has immense potential in cryogenic electronics, and it offers ultra-fast GHz-range operation with significantly reduced power demands, positioning them as promising candidates for cryogenic oscillators and neuromorphic circuits. Our comprehensive design space exploration of the oscillators highlights material and circuit-level optimization necessary for their reliable performance, marking significant strides in cryogenic computational paradigms.
Other Areas
- Hardware security: physical unclonable functions, true random number generators
- Analog and Peripheral circuits: Low power rectifiers, sense amplifiers
- Nanoelectronic reliability: Negative bias temperature instability, time-dependent dielectric breakdown, hot carrier injection, radiation, etc.