Virtual National Laboratory races Moore’s Law, as minute circuitry requires increasingly careful design.
Extreme ultraviolet lithography, a technology being developed by a consortium of U.S. national laboratories and the semiconductor industry, is a strong contender to produce new generations of computer chips with features perhaps as small as 30 nanometers.
The consortium, formed in 1997, consists of the Virtual National Laboratory, which combines the research capacities of the U.S. Department of Energy’s Sandia National Laboratories, Lawrence Livermore National Laboratory and Lawrence Berkeley National Laboratory; and the EUV Limited Liability Company (EUV LLC), consisting of Intel Corporation, Advanced Micro Devices and Motorola. According to Intel officials, the result should be central processing units (CPUs) that are 100 times faster than current CPUs and memory chips that store 1,000 times more data than those used today.
Both the number of transistors on a chip and the clock speed at which the chip can be operated increase as the size of features on the chip diminishes. More precisely, halving the feature size quadruples the number of transistors.
Current chips have feature sizes of about 250 nanometers, have in the vicinity of eight million transistors, and operate at clock speeds of about 500 megahertz. By comparison, Intel scientists believe that microprocessors in 2011 will have a billion transistors, clock speeds of 10 gigahertz, and deliver 100,000 MIPS, or millions of instructions per second.
Today, chips are made by a photographic process using deep ultraviolet light (DUV). Because of fundamental laws of physics, as feature sizes approach the wavelength of light used to make them, sharpness of features tends to decrease. DUV is reaching this point. Consequently, the industry must go to shorter wavelengths if the present rate of progress in microcircuitry—doubling the power of microprocessors every 18 months—is to continue. While designers say that they expect this rate, known as Moore’s Law, to continue for the foreseeable future, the properties of circuits change in such small dimensions that they require increasingly careful design.
Four technologies are under consideration as successors to DUV: proximity X-ray lithography, scattering with angular limitation projection electron-beam lithography (SCALPEL), ion-beam projection lithography, and extreme ultraviolet lithography (EUVL). However, according to EUV LLC researchers, extreme ultraviolet (EUV) has several important advantages. Chief among them, the EUV approach uses light instead of electrons, as do current manufacturing processes; therefore, much of the manufacturing process now used will also work with EUV. For example, photoresists now used with wavelengths of 248 and 193 nanometers can be used with EUV, avoiding the development of new resists.
A resist is a photosensitive substance used to cover silicon wafers from which chips are made. When a desired pattern formed by an exposure mask is projected through reduction optics onto the resist by the manufacturing exposure system, the resist chemically changes only where the light hits it. The altered resist can be chemically removed, leaving the desired pattern of silicon exposed. Further substances can then be deposited on the exposed silicon to form part of the complete circuit. The process is repeated to form the various layers of the completed chip.
EUV LLC expects to have the technology to allow manufacture of chips with feature sizes below 100 nanometers by 2004. However, while EUVL is progressing satisfactorily, according to researchers, the new technology poses problems not encountered before. For example, extreme ultraviolet light, sometimes called soft X-ray, is strongly absorbed by matter. Manufacturing must therefore be done in a vacuum. Further, the light cannot be focused by lenses. Instead, mirrors must be used. This has required development of very precise multilayered mirrors made of molybdenum/beryllium and molybdenum/silicon.
The EUV light reflects from the interfaces between the layers but is absorbed by the layers themselves. The intensity of the reflected signal is a trade-off between constructive interference of light reflected from the various layers and the concomitant absorption. One concern has been how to get the reflectance high enough to work in a production environment. The EUV light is reflected from nine surfaces in the engineering test stand being used to develop the technology. Each reduces the intensity of the source. However, reflectances of 70 percent per mirror have been achieved, which is sufficient for manufacturing.
In ordinary optical lithography, the mask that defines the pattern projected onto the resist is transparent, except for the pattern itself—much like a photographic negative. Because such a mask would absorb an EUV signal, the mask itself is a patterned mirror with the pattern contained in an additional layer of metal. Extreme care is needed in fabrication to achieve the necessary tolerances without causing defects.
Another problem is the generation of a reliable supply of extreme ultraviolet light with wavelengths in the range of 10 to 14 nanometers. Both synchrotrons and electric-discharge sources were considered but rejected for technical reasons. Current work focuses on laser bombardment of various materials. Early work used a gold target, but this produced debris that contaminated the mirrors. The solution now used focuses a high-power, pulsed, solid-state laser on a supersonic beam of xenon gas clusters. There is nothing to produce a deposit. Work is underway to increase reliability and reduce cost to levels acceptable for manufacturing.