N a n o m e c h a n i c a l C o m p u t i n g

The impressive trend of packing more transistors continues with advances in foundry technologies. Now at 22 nm node, some of the current processors pack over a billion transistors in a chip. However, three other markers -- clock speed, power consumption and performance -- have flat-lined for almost a decade. Is there an alternative hardware

architecture that can be used to perform the same computation with lower energy?

Almost two centuries ago in 1822, Charles Babbage presented a mechanical computing device that he called the “Difference Engine,” to the Royal Astronomical Society. Before this event, though, the search for mechanical computing devices had already been inherent to attempts to build machines capable of computation. This search has, today, taken on added urgency as we seek to exploit emerging techniques for the manipulation of matter at nanometer length scales. With Boole’s ideas on logic operations with two states, an added dimension to computing, logic elements or gates, has come to dominate modern computation. However, mechanical logic, especially at the very small length scales and in the presence of a noise floor, has proven difficult to realize despite recent experimental efforts. In our group, we have a comprehensive approach to developing nanomechanical computation. It includes mechanical memory element, mechanical logic gates and other signal processing elements.

Conventional Nanomechanical Logic Gate: Practical realization of a nanomechanical logic device, capable of performing fundamental logic operations, is yet to be demonstrated despite a longstanding effort towards scalable mechanical computation. Recently, we have presented a nanomechanical device, operating as a reprogrammable logic gate, performing fundamental logic functions such as AND/OR and NAND/NOR with 100% fidelity. The logic function can be programmed (e.g. from AND to OR), dynamically, by adjusting the resonator’s operating parameters. The device can access one of two stable steady states, according to a specific logic function; this operation is mediated by the noise floor which can be directly adjusted, or dynamically “tuned” via an adjustment of the underlying nonlinearity of the resonator, i.e., it is not necessary to have direct control over the noise floor. The demonstration of this reprogrammable nanomechanical logic gate affords a practical realization of a new generation of mechanical computers.

Reversible Nanomechanical Logic Gate: Irreversible logic, the basis of modern computing, inevitably leads to loss of information and is thus fundamentally bound by the von Neumann-Landauer (VNL) principle: for every bit of information lost during a computation, kTln2 amount of heat is dissipated into the environment. Reversible logic, in contrast, does not entail information loss, and hence is not bound by the VNL limit. It offers the potential for indefinite performance improvements of digital electronics. Bennett's Turing machine first proved that any computation could be performed reversibly and, in the proper limit, without energy cost. This promise of computing for free has spurred Fredkin, Toffoli, Wilczek, Feynman and others to propose reversible logic gates, though very few experimentally-realized reversible logic gates have since been reported.

Recently, we have experimentally demonstrated for the first time the core of a logically reversible, CMOS-compatible, scalable nanoelectromechanical Fredkin gate, a universal logic gate from which any reversible computation can be built. In addition to demonstrating the truth table, we show that the nanomechanical Fredkin gate can be operated as a reversible AND-, OR-, NOT- and FANOUT gate. We find that this device exhibits ultra-low energy cost per logic operation, on the order of 100 kT, similar in magnitude to energies involved in DNA polymerase (20—100 kT) and approaching the VNL limit to within a factor of 100, two orders of magnitude closer than state-of-the-art transistor-based technologies. Reversible nanomechanical logic gates could play a crucial role in developing highly efficient reversible computers, with implications for efficient error correction and quantum computing.