“Illustrated Semiconductor”: Moore’s Law was considered to be reaching its limit around 2000. When will the miniaturization of semiconductors last?

Sharing is Caring

Moore’s Law—How long will the miniaturization of semiconductors last?

  In 1965, Moore, one of the founders of Intel, reviewed the number of transistors on a single chip IC in the past five years and found that the number of transistors would double to the original number every year. As a result, Moore published related articles in the magazine, predicting that this trend should continue in the future.   This is the famous “Moore’s Law“. When Moore published this article, about 64 transistors had accumulated on a single wafer. He predicted that in 1975, 10 years later, a wafer could accumulate about 65,000 components.

Figure 3-20 is a schematic diagram of the transition of the number of transistors in a single DRAM chip. When Moore discovered this “law” (1965), the number of transistors was indeed growing at a rate of doubling a year. After that, however, it slowed to a doubling of two years. So later Moore himself revised Moore’s Law to “doubling in two years (24 months)”. Graphical Semiconductor - 130     In addition, if the size of the chip remains the same, but the number of components on the chip is expected to increase, then each component must be made smaller than the original, and the line width of the circuit must be narrowed accordingly. Figure 3-20 shows that the line width (process) has a trend of miniaturization. The earliest 1kb DRAM was born in 1970, with a line width of 10μm; compared with this, the current line width is only 20nm. In Figure 3-19, the number of transistors in the MPU also follows Moore’s law and increases with time.   This Moore’s Law is a rule of thumb with no theoretical basis, but in the following 40 years, the number of transistors mounted on each chip has followed this “law” to increase over time, making Moore’s Law an important issue in semiconductor technology and the industry. guidance.   However, in recent years, many people believe that Moore’s Law will fail soon, because the miniaturization of the process (line width) is close to the limit.   In 2020, the smallest process in commercial products is 5nm, which is only 10 times the lattice constant of silicon crystal (about 0.5nm).

Semiconductor elements are composed of crystals, so it is impossible to make them smaller than the lattice constant.   After trying to compare with tiny objects, it can be found that the semiconductor elements at the beginning were about the size of bacteria, and the current semiconductor elements are as small as viruses or DNA.   When photolithography is used to engrave circuit patterns on silicon, the width of the circuit is also limited by the wavelength of light.

In addition, after miniaturization, the separation of elements is also a problem. If the gate oxide film becomes too thin, leakage current will occur, which will cause great trouble. Although some problems can be solved technically, the cost is too high to be commercialized, and there are layers of obstacles to other solutions.   In fact, Moore’s Law was considered to be reaching its limit around 2000, but then there were technological breakthroughs that made Moore’s Law sustainable.

For example, High-k insulator technology that increases gate capacitance while maintaining the thickness of the oxide film; Low-k dielectric film technology that reduces wiring capacitance; technology that applies stress to the channel part to increase the actual electron mobility ; During exposure, control the phase of light with a mask, so that light can engrave a fine structure with a length shorter than the wavelength of light; expose in liquid to shorten the actual wavelength of light, etc. A variety of technologies have been put into application one after another.

The FinFET technology introduced from the 16nm generation is a very important innovation in recent years. In the past, the planar MOSFET has become a three-dimensional structure and has been miniaturized.   However, as the miniaturized products get closer to the lattice constant, it becomes very difficult technically to further accumulate components.

Now, however, researchers have also discovered techniques that may overcome these difficulties. How long will Moore’s Law last? This battle between technologists and the limits of physics will continue.

Sharing is Caring