Lithography and the Future of Moore’s LawThe definition of “Moore’s Law” has come to refer to almost anything related to the semiconductor industry that when plotted on semi-log paper approximates a straight line. I hesitate to review it's origins and by doing so restrict it's definition. However, today I will review the history and past performance relative to predictions and show where the advances have come from. I will leave the future performance up to you. Certainly continuing on the same slope doesn’t get any easier. It presents a difficult challenge to the industry.The original paper that postulated the first version of the “law” was an article I wrote for the 35th anniversary issue of Electronics Magazine in 1965. My assignment was to predict what was going to happen in the semi-conductor components industry over the next ten years — to 1975. In 1965 the integrated circuit was only a few years old and in many cases was not very well accepted. There was still a large contingent in the user community who wanted to design their own circuits and who considered the job of the semiconductor industry to be to supply them with transistors and diodes so they could get on with their jobs. I was trying to emphasize the fact that integrated circuits really did have an important role to play.Let’s start with two figures from that original paper. Figure 1 shows my estimate of the cost of integrated circuits divided by the number of components, a component being a transistor, resistor, diode or capacitor, in an integrated structure at various times. In 1962 the mini-mum cost per component occurred for circuits containing about ten components. For more complex circuitry costs skyrocketed because yields collapsed. With time, as processing improved, the minimum moved down and to higher complexity. When I wrote this article in 1965 my estimate was that the minimum cost per component was achieved with several tens of components in a circuit, and I predicted that the minimum would continue to go down as we improved out processing capability.Next I looked at how complex integrated circuits should minimize cost per component and reasoned that the most complex circuit at any time would not be much more complex than this minimum, because of the steepness of the curve beyond the minimum. The avail-able data I chose started with the first planar transistor, which had been introduced in 1959. It was really the first transistor representative of the technology used for practical integrated circuits. It is represented by the first point, two raised to the zero power, or one component.Gordon E. Moore, Co-founder Intel Corporation 1995 SPIE SpeechFigure 1 Relative manufacturing cost per component vs. components in the circuit estimated for various times.Reprinted with permission. Gordon E. Moore, Lithography and the Future of Moore's Law, Proc. SPIE Vol. 2437, May 19952Adding points for integrated circuits starting with the early “Micrologic” chips introduced by Fairchild, I had points up to the 50-60 component circuit plotted for 1965 as shown in Figure 2. On a semi-log plot these points fell close to a straight line that doubled the complexity every year up until 1965. To make my prediction, I just extrapolated this line an-other decade in time and predicted a thousand-fold increase in the number of components in at the most complex circuits available commercially. The cheapest component in 1975 should be one of some 64,000 in a complex integrated circuit. I did not expect much precision in this estimate. I was just trying to get across the idea this was a technology that had a future and that it could be expected to contribute quite a bit in the long run.Many of you were not in the industry when the devices represented by the first few points in this plot were intro-duced. I have reproduced photomicrographs of the first planar transistor and the first commercially-available integrated circuit in Figures 3 and 4. I am particularly fond of the transistor, since it is one of the very few products that I designed myself that actually went into production.The unusual diameter of 764 microns was chosen because we were working in English units and that is thirty thou-sandths of an inch, or 30 mils. The minimum feature size is the three mil metal line making the circular base contact. Metal-to-metal spacing is five mils to allow the 2.5 mil alignment tolerance we needed.Interestingly enough at the time the idea for the planar transistor was conceived by Jean Hoerni in the early days of Fairchild Semiconductor, it had to sit untried for a couple of years, because we did not have the technology to do four aligned mask layers. In fact, we were developing the technol-ogy to do two aligned oxide-masked diffusions plus a mesa etching step for transistors. The original step and repeat cam-era that Bob Noyce designed using matched 16 mm movie camera lenses had only three lenses, so it could only step a three-mask set. We had to wait until the first mesa transistors were in production before we could go back and figure out how to make a four mask set to actually try the planar idea.The first integrated circuit on the graph is one of the first planar integrated circuits produced. It included four transis-tors and six resistors. It has always bothered me that the picture of this important device that got preserved was of the ugly chip shown in Figure 4. The circuit had six bonding pads around the circumference of a circle for mounting in an 8-leaded version of the old TO-5 outline transistor can. In this case only six of the eight possible connections were required. We did not think we could make eight wire bonds with Figure 2 The original “Moore’s Law” plot from Electronics April 1965. Figure 3 Photomicrograph of the first commercial planar transistor.Figure 4 Photomicrograph of one of the first planar integrated circuits produced by Fairchild Semiconductor in the early 1960’s.3reasonable yield, so for these first integrated circuits we etched a round die that let us utilize blobs of conducting epoxy to make contact to the package pins. For the die in the picture, the etching clearly got away from the etcher.How good were my predictions? What really happened?Figure 5 adds the points for several of the most complex in-tegrated circuits available commercially from 1965 to 1975. It was taken from an update of the industry’s progress that I