A Hewlett-Packard research team has advanced efforts to develop practical chips based on molecular electronics by creating the first molecular-electronic memory chip. The experimental device stored only 64 bits but achieved a storage density 10 times that of conventional silicon electronics. In addition, the chip's memory was nonvolatile (preserved with the power off) and was combined with logic elements, both of which provide advantages that are not available with silicon-based random access memories.

During the past three years, several groups have succeeded in producing individual electronic devices, including transistors, from organic molecules such as rotoxane and from carbon nanotubes. But until now, no one had succeeded in making usable multiple-device chips or in developing techniques with the potential for mass production. Conventional lithography cannot produce the nanoscale devices needed, and no one has yet perfected self-organizing techniques that can enable molecules to form the circuits by themselves.

The Hewlett-Packard group, led by senior scientist Yong Chen, accomplished its feat by using a new imprinting process recently developed by Princeton University physicists (see TIP Briefs, October/November 2002, pp. 10-11). The work, which has not yet been published in a technical journal, is described in U.S. Patent 6,432,740. The method creates a mold out of quartz, which is then pressed into a layer of photoresist. Where the mold compresses the resist, it can be etched away. Where the resist is not compressed, it protects the underlying layer. This process allows the use of the same lithography techniques employed by the chip industry, but it yields narrower linewidths.

It takes a couple of days to create the molds using electron-beam and optical techniques, Chen explains. But this does not matter because using the mold to create the chip only takes a few minutes, and potentially could be done in seconds or less. In making the nanocircuits, a layer of wires is first laid down using 40-nm linewidths, which is one-third the width of the finest lines available in commercial chips. On top of this is laid a two-dimensional crystal layer of rotoxane molecules and then another layer of wires perpendicular to the first. The result is a crossbar switch consisting of about 1,000 molecules wherever the upper and lower wires cross. When a potential is applied to two selected wires, the rotoxane molecule switches from a conducting to nonconducting state or back again. The large difference in conductance between the two states, a factor of 10,000, is not fully understood, Chen says. A much smaller potential can then be applied to read the switch.to detect whether it is in a conducting or nonconducting state. Because the molecule remains in one state until switched out of it, the memory is nonvolatile and is unaffected by turning the power off.

Because each switch can work as a nonvolatile memory or as part of a logic device, the circuitry for multiplexing and demultiplexing can be built into the same chip as the memory cells. This is crucial in molecular electronics, in which many cells must feed into a few output wires. We see the next steps as finding better molecules and combining them more effectively with existing silicon circuits, says Chen. He expects the first commercial uses of molecular electronics to be specialized memory chips, which could reach the market in as little as five years. Such a development would eliminate the existing limits to shrinking today¡'s silicon circuits. Eventually, molecular crossbar switches could be based on a single molecule for each switch, allowing trillions of bits to be packed into a square centimeter.