Title: Part 1:  Physics Issues, Part 2:  Computer Science     

Speaker: Erik DeBenedictis* (PI), Jerry Floro*%, Peter Kogge#, Robert Hull%, Craig Lent#, Sarah Murphy*#, Mike Niemier#, Marco Ottavi*, Aaron Prager*#, Greg Snider# (*Sandia, #Notre Dame, %U. Virginia)

Date/Time: Tuesday, August 8, 2006, 10:00 am and 1:00 pm

Location: Part 1:  823 Breezeway (Sandia NM), Part 2:  CSRI Building, Room 90 (Sandia NM)

Brief Abstract: Part 1:  We will explain how the laws of thermodynamics set the lower limit of power dissipation in computers, but that this limit applies to our historic choice of AND-OR-NOT logic primitives. "Reversible" logic primitives and physical implementations (such as QDCA and biological processes) can and do bypass this limitation. The key distinction is that the physical representation of information must be moved through the system as opposed to being turned into heat and regenerated from the power supply after each logic stage.

The talk will discuss experimental physical implementations of QDCA reversible logic, including proof-of-concept metal island experiments through other proposed implementations including molecular-sized devices that should operate at room temperature. We will also describe certain issues in the future design of such devices, such as the need for clocking, crossovers of logic signals, and the need to control parasitic power dissipation to levels below those tracked by CMOS engineering.

The talk will also discuss a series of experiments underway between Sandia and its partners. These experiments are fabricating and will measure self-aligned SiGe quantum dots in a configuration expected to be suitable for operation as a QDCA.

We will discuss simulation tools in existence at partner institutions applicable to QDCA and the issues above.

The objective of the physics talk is to show how special attention to information handling in device physics research will be important to continue "Moore’s Law."

Part 2:  This partnership is focused on physical QDCA implementations that permit "reversible logic" if used appropriately. This talk will discuss a series of unfamiliar ideas in computation that if applied to a suitable physical system could permit performance orders of magnitude above the ultimate limits of CMOS microprocessors.

We learn AND-OR-NOT Boolean logic in primary school and are taught that it is "universal," implying there is no need to learn anything else.
However, the AND and OR functions are not thermodynamically reversible and must generate heat when executed in our universe. Interestingly, there are other logic primitives (called reversible) that are also universal but need not generate heat (example: the Toffoli gate). The essence of this talk is to understand how to build a familiar computer from a completely different logical basis.

The talk will discuss a series of design methods (or architectures) developed by this LDRD over the summer that permit arbitrary functions to be executed by QDCA systems of the type under study where the designer can radically adjust the speed and power of the resulting system.

We will discuss simulation tools in existence at partner institutions applicable to QDCA and the issues above.

The objective of the computer science talk is to describe how the computer science community can work with the physical science community to create devices for future computers that retain the advantages of reversible logic yet offer enough features to permit large and robust computers to be built.

CSRI POC: Erik P. DeBenedictis, (505) 284-4017