My research interests broadly cover the spectrum of emerging technologies for computing. Here are some of the areas I have worked on, and continue to be interested in.
Spintronics
Much of the technology, and the face of today's world has been shaped by electronics. Transistors are electronic devices that have provided increasing performance at low energy and allowed scaling to nanoscale dimensions. Electronic devices take advantage of the charge of an electron. The charge-based switching in transistors, although very efficient, requires much more energy than the fundamental limit of energy that must be expended during switching (See Landauer's limit). Electrons have another property, the spin, which can be theoretically switched at much lower energies, providing an ultimate low-power state variable. If spins can be controllably switched, then they can enable ultra-low power computing. Spintronics devices aim at taking advantage of both charge and spin properties of electrons in devices such as Magnetic Tunnel Junctions, for various memory and computing operations. But designing circuits with spin properties is fundamentally different than designing switch-based logic circuits. The overhead of additional circuitry involved to drive spins across multiple devices easily exceeds the benefits derived from efficient device physics. It is therefore necessary to develop novel circuit designs and find novel applications to take advantage of the efficient spintronics technology.
Nano-Electro-Mechanical Systems based on Carbon-NanoTubes
While transistors scale down in area from 45nm, their power density does not linearly scale down since threshold voltages cannot be scaled as desired (See Dennard's Scaling). NEMS switches exploit electro-mechanical actuation for ON/OFF action, providing a near-ideal switch that can address power concerns in current computing circuits. When fabricated using a low-mass material such as Carbon Nano Tubes, these switches can be actuated at speeds competitive with those of the electronic transistors. However, for truly fast circuits with NEMS devices, we need about 100x improvements in their speeds. This will likely require a combination of advanced device structures, parallel circuit designs, and new architectural ideas.
Quantum Computing
Our computing problems have already grown to intractable sizes. We can factorize small integers in a matter of seconds (i.e. given a composite integer c, find its prime factors a and b), but as we increase the number of bits in the integer, the time required blows unreasonably. Recently, it took researchers a few months to compute factors of a 768-bit integer using resources equivalent to about 1500 computing years (see this paper). This is just one of the classically intractable problem. Several others exist in the fields of biology, chemistry and physics, where researchers are trying to push the boundaries of our knowledge and our fundamental capabilities. Quantum computing is a novel computing paradigm that can effectively deal with many of these categories of problems. It elegantly handles solution spaces that grow exponentially with the number of bits, processing all of them in parallel. The catch? - only one answer can be obtained at the end! Quantum computing has been a fantasy ever since Richard Feynman conceived its usefulness to simulate quantum-mechanical systems. Many recent promising efforts, and developments in the much-debated D-Wave machine, have raised new excitement in the field. Many research problems must be solved before we can truly start computing with quantum bits.
Mobile Computing
Mobile devices have rapidly grown to be ubiquitous personal computing devices. They have grown in the number of functions they can do, and are now expected to provide many more diverse functionalities. All of these come at the expense of energy, which is only derived from the battery in the device. For powerful mobile computing, we must find new ways of supplying more performance with lower energy requirements.