From: John K Clark (jonkc@att.net)
Date: Tue Sep 10 2002 - 08:22:47 MDT
>From today's New York Times:
--------------------------------------------------------------------------------
Hewlett Finds a Process to Make Chips Even Smaller
By JOHN MARKOFF
ALO ALTO, Calif., Sept. 9 - Researchers at Hewlett-Packard laboratories have developed a manufacturing process capable of producing molecular-scale circuits vastly denser than today's most advanced semiconductor chips.
The discovery offers the hope of assembling billions or even trillions of molecular-size switches in an area comfortably smaller than a fingernail and at a cost far lower than today's computer chips. The advance could lead to immensely powerful and inexpensive computers capable of holding entire libraries of music and movies for the consumer, or calculating scientific problems that are now unsolvable.
At the same time, the scientists said they were startled to discover that their ultratiny switches were behaving in ways not yet completely understood. The switches exhibit swings in electrical resistance that vary by a factor of 10,000. The huge shift is useful for determining when the switches are on or off, but cannot be explained by existing theoretical physics.
Although scientists said that commercial use of the new approach was at least five years off, the field known as molecular electronics is drawing attention in Silicon Valley as perhaps the most promising way to surmount the physical size limits that engineers have feared the semiconductor industry may run into before the end of the decade.
Dr. R. Stanley Williams, a physicist who is director of Quantum Science Research at the Hewlett-Packard laboratory, described the research advance today in a lecture at a symposium celebrating the 175th anniversary of the Royal Institute of Technology of Sweden.
"This is a very big step forward," said Dr. James C. Ellenbogen, a physicist at the Mitre Corporation who specializes in the field. He said Hewlett-Packard had conducted "a very systematic industrial effort" and produced "some really good news for the electronics industry."
The Hewlett-Packard researchers said they had been making working circuits for almost a year, but chose to make the announcement now because they received a crucial patent in the field last month. The patent was granted for work done by Yong Chen, a senior Hewlett-Packard scientist who conceived of the new manufacturing approach. Scientists and industry executives said that the technology had made startling strides in the last two years. Earlier this year, I.B.M. said it would be possible to commercialize later this decade a disklike storage technology named Millipede capable of storing trillions of bits of data per square inch.
In addition to corporate and university research groups, there are now as many as a dozen privately financed start-up companies pushing to develop molecular-scale computer technologies that are claimed be denser, faster and cheaper to make than today's most powerful computer chips.
Perhaps most striking in today's announcement was the early success the Hewlett-Packard researchers have had with a manufacturing approach that can be thought of as an extremely high-tech waffle iron.
The researchers have built working 64-bit memories using what they call nano-imprint lithography. First, an array of parallel platinum wires are formed in a master mold that is created using electron beam lithography. Then a single layer of electronically switchable molecules is deposited on the surface of the wires, and finally another grid of wires is molded at right angles to the first, creating a sandwich.
The junctions where the wires intersect are no bigger than a square micron, and no more than 1,000 molecules sit between them. (By contrast, a human hair is 75 microns in diameter.) Yet each can be addressed individually, using previously designed techniques, and it is possible to read and write digital ones and zeros at each wire junction by electronically changing the properties of the molecules.
Ultimately the goal is to scale the wires to ever finer dimensions and use fewer and fewer switching molecules. Dr. Williams said the researchers now believe that the fundamental limits of the technology could be wires that are two nanometers in width - a single atom wide - spaced 10 nanometers apart. Such a device would achieve storage densities of one trillion bits of information per square centimeter. Current semiconductor memories can store about half a billion bits per square centimeter.
The researchers said they were pushing to meet a challenge set out by the Defense Advanced Research Projects Agency, known as Darpa, the Pentagon arm that is financing a number of research teams with the goal of being able to build 16,000-bit memories by 2004.
The Hewlett-Packard researchers said today that they believed that they would be able to scale their manufacturing technique to meet the challenge.
"To achieve the Darpa challenge we need to improve the pitch of our wires by a factor of two and double the number of wires each year," said Dr. Williams. "We're definitely on track to do that."
He said his laboratory had been working with electron beam experts at Lawrence Berkeley Laboratories to refine the capabilities of the electron beams that are used in making the wire molds used to make the circuits.
The use of molecular-scale molds to make arrays of tiny wires was pioneered by a group of Princeton researchers led by Stephen Y. Chou, the director of Princeton's nanostructures laboratory. In the June 20 issue of the journal Nature, the researchers outlined a process for making wires 10 nanometers thick using an ultrafine molding technique.
The molecules themselves are the source of a scientific puzzle for the researchers who are trying to understand why they are seeing such large shifts in electrical resistance in their molecules.
"We do not fully understand the switching phenomenon," said Dr. Williams, adding that the Hewlett-Packard laboratory currently has a group of theoretical physicists thinking about the properties of electrons being conducted through molecules.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:55 MST