A Molecular Alternative to CMOS?

Ruth DeJule, Associate Editor -- Semiconductor International, 9/1/1999

CMOS-based devices have been at the heart of the technological advances in the microelectronics and computer industries. Faced with the demands of Moore's Law and escalating processing costs, potential alternatives to CMOS logic devices are being investigated, such as single-electron transistors, quantum cellular automata and chemically assembled electronic nanocomputers (CAEN). One particularly interesting approach, molecular tunneling devices, implements processes familiar to silicon device process engineers. Based on molecular switches, researchers at Hewlett Packard Labs and the University of California, Los Angeles, have developed logic gates and memory elements that fit between the wires used to address them.

The structure of the molecular switch is simple, consisting of a monolayer of rotaxanes and a tunneling barrier sandwiched between metal electrodes (Figure). When a voltage is applied across the electrodes, current flows by resonant tunneling through the molecular electronic states. These molecular tunneling switches are normally closed in the 'on' configuration, and while they can be electrically opened, once opened they can't be closed again. Therefore, these switches cannot be used for random access memory (RAM) applications, though programmable read-only memory (PROM) applications are possible. The researchers sought three properties in selecting an appropriate molecule: hysteresis in current-voltage characteristic, processability and reversibility. Rotaxanes provided the first two, but not reversibility. Work is underway to identify and synthesize molecules that embody all three characteristics.

Switches in a linear array (a) consist of a single layer of molecules sandwiched between two electrodes (b). Electrons tunnel through the Al2O3 and Ti/rotaxane layers, changing resistance (c).

The manufacturability of these devices is a clear advantage. Processed on any flat substrate such as silicon, an insulating layer is deposited on top. Next, aluminum wires are evaporated using standard ph otolithography processes followed by exposure to air, at which time a thin native oxide, 1 nm to 1.5 nm thick, is formed. A molecular monolayer is then deposited over the entire wafer using the Langmuir- Blodgett process. The wafer is submerged in a trough of water with the molecules afloat. Upon removal, a dense, uniform film of molecules coats the wafer. Using a shadow mask, metal wires are then formed perpendicular to the original set of wires. The devices are the molecules trapped between two wires; their size is defined by the size of the wire. In addition to potential manufacturability, molecular tunneling switches have the advantage of scalability. The devices currently demonstrated have 100 million molecules between the two metal layers, which are rectangles 6 µm u 11 µm in size. Team co-leader Stan Williams anticipates this will be scalable to wires on the order of nanometers in width, with only a few molecules, 20 Å long, between the wires.

The switches were fabricated into a linear array and electronically configured into AND and OR logic gates, and they performed 'magnificently,' according to Williams. These devices are essentially diode logic gates, but with exponential current-voltage characteristics. On/off ratios were greater than 803. This could mean dramatically simplified circuitry with far fewer transistors (by a factor of ~6) required for a particular function. However, Williams cautions, the same properties can create problems. For instance, how do you design circuits with strongly coupled non-linear elements?

Eventually, molecular memory and processors may replace integrated circuits inside personal computers, and they may one day be found in supercomputers the size of a wristwatch or woven into clothing, Williams suggests. Chemical processors may be the answer to the physical and fiscal limitations of conventional silicon technologies.