Oh hell I'll post the whole Q&A
isonics.com
Q. What is isotopically pure silicon-28?
A. Normal silicon consists of 3 stable isotopes (similar chemical forms with slightly different atomic masses), silicon-28, silicon-29, and silicon-30. Silicon-28 is an ultra-purified form of normal silicon in which the naturally occurring minority isotopes have been removed.
Q. What benefits does this ultra-purification produce?
A. By removing the different sized atoms, the lattice crystal structure becomes more uniform and lattice vibrations (phonons) and electrons can travel with less scattering. In silicon, heat is primarily conducted by phonons, so the thermal conductivity is increased by up to 60% at room temperature.
Q. How can a thin layer of silicon-28 affect the performance of a computer chip?
A. The active transistors in a computer chip are manufactured in the top few microns of a wafer. Today, all advanced microprocessors use epitaxial wafers because the epitaxial layer can be made with fewer defects than a bulk wafer. The epitaxial layer has fewer crystalline defects (dislocations, vacancies, micro-voids, etc.) and a more uniform distribution of dopant atoms because the layer is grown from a gas mixture rather than from a large boule of silicon. By using an isotopically pure silicon layer, the crystal structure is more perfect, and less scattering of the electrons takes place, generating less heat and electromagnetic noise.
Q. How does a thin layer of silicon-28 with higher thermal conductivity affect the temperature in a microprocessor when the heat must diffuse through the bulk silicon wafer, which is much thicker?
A. The average temperature of the chip will probably not be affected. However, the very localized temperatures at the p/n junction of a transistor can be greatly affected. Different parts of an integrated circuit see very different operating conditions. The logic core of a microprocessor can be operating almost continuously, while the temporary storage areas are accessed less frequently. This leads to areas of a chip generating more heat than other areas. These “hot spots” determine the maximum operating conditions of a circuit design. The higher conductivity silicon-28 layer spreads the heat from these active areas better than natural silicon, thus reducing the maximum temperatures.
Q. How do lower chip temperatures help performance?
A. Since heat is generated by the switching on and off of transistors, using a high thermal conductivity layer that reduces the maximum temperatures generated allows the switching speed to be increased to generate the same amount of heat.
Q. What manufacturing advantages does isotopically pure silicon offer?
A. During the fabrication of integrated circuits, many different masks are used to define the individual areas of a transistor (sources, drains, and channels in CMOS devices). The current flow through these areas is designed to be a constant value, based on the size of the source/drain/channel. Normal manufacturing variations in mask alignments, ion implantation dose, oxidation and annealing temperatures, etc., causes a statistical distribution of the contact resistance and sizes of these areas; some areas are larger than the average, and some are smaller. The smaller areas still have to carry the same current as the larger ones, so the current density is higher and the heat generated is larger than designed. This is one reason why some chips from a wafer are rated at slower clock speeds than others. The lower rated chips can have hot spots that may cause them to fail if run at full rated clock speed. Isotopically pure silicon would minimize these “hot spots” and allow more high speed chips to be obtained from each wafer. Chips that already operate as designed would run cooler and be more reliable.
Q. How does doping silicon change the thermal conductivity?
A. Doping silicon with other atoms will decrease the thermal conductivity. Since normal doping levels are from 1015 to 1018 atoms/cc, or from 0.1 parts per million to 100 parts per million atoms, the effect of dopant atoms will be much less than the effect of silicon isotope atoms (80,000 parts per million). Highly doped silicon-28 will have higher thermal conductivity than highly doped natural silicon.
Q. What other advantages does isotopically pure silicon offer?
A. As the size of transistors is decreased, the amount of material in any given volume decreases. For example, one of the controlling factors in CMOS processing is the gate oxide thickness, which controls the voltage necessary to cause current flow in a transistor. With transistor designs approaching 0.1 micron (100 nanometers) the gate oxide thickness will approach 20 angstroms (2 nanometers). This is equivalent to 5 atomic layers of silicon dioxide. Removing the larger silicon-29 and silicon-30 atoms will allow more uniform oxide layers and higher quality transistors.
Q. What is the cost of silicon-28 epitaxial wafers?
A. Currently the cost is several times the price of a natural silicon epitaxial wafer. As with most things in the electronics area, the cost will come down as the volume produced increases. Once full production is achieved, the price premium could be no more than 50% over natural silicon epitaxial wafers. This equates to a cost per microprocessor of less than 25 cents.
Q. How can a higher cost wafer save chip manufacturers money?
A. By increasing the yields of higher speed chips, which have a higher selling price, the total revenue generated by each wafer is increased. Since microprocessors sell for from about $100 to several thousand dollars each, two or three more high speed chips (out of 300 or more) can pay for the wafer.
Q. What new fab equipment or process steps are required to use silicon-28 wafers?
A. None. Except for the improved properties, they are identical to normal silicon wafers.
Q. Can bulk silicon-28 wafers be made economically?
A. Yes. With our new U.S. production facility large amounts of silicon isotopes will be available at reasonable cost. The first 200 mm diameter bulk wafers should be available for testing by the end of year 2000. The cost of these wafers will be high initially, but the isotope separation process is essentially a chemical process and can be scaled up easily to multi-ton quantities per year. Once large scale production is established, 200 mm diameter bulk silicon-28 wafers could cost less than $10 per microprocessor.
Q. What extra benefits do bulk silicon-28 wafers offer to justify their higher price?
A. Thermal modeling has indicated that the peak temperature in the logic core of a microprocessor can be reduced by about 30 degrees centigrade by the use of silicon-28 bulk wafers. This large temperature decrease will allow heat sinks and fans to continue to be used instead of more expensive cooling solutions like thermoelectric or cryogenic coolers as the clock speeds of microprocessors are increased to several gigahertz in the next few years. |
