In a lecture that dips into both the anatomy and history of the semiconductor, Grant Willson offers some provocative thoughts on whether industry can continue improving on this most useful of inventions.

He describes how steady advances in miniaturization enabled the astonishing progress of microchips over the past 40 years. Today, says Willson, you can “buy a transistor for less than the cost of a single written character in your local newspaper.” When he began at IBM in the 1970s, the silicon wafers produced were only 1 ¼ inches in diameter; now “they’re bigger than pizzas.”

Willson delves into the technological changes that both enabled printing on circuits to grow smaller, and the final product to grow larger. He details the original process of photolithography, involving designing a circuit pattern, then using a $25 million printer with a focused electron beam to reproduce the pattern on special glass, called a mask. It’s the mask’s pattern, etched onto a silicon wafer that forms the basis of the microchip. Layer after layer of these patterns get laid down on a single chip.

The machines behind these processes cost tens of millions of dollars. Just the lens for focusing laser light onto the wafer through the mask has 40 optical elements and weighs as much as a car. Over time, explains Wilson, “People try to make bigger lenses to make smaller structures. The bigger the lens, the shorter the wavelength of light.” In the ‘70s, recounts Wilson, machines were printing hundreds of miles of stuff the size of a bacterium. Today, with the help of chemical catalysts, printing has been reduced to less than 100 nanometers in diameter.

But there’s a problem in reaching the next generation of super-small, mass-produced chips, believes Willson. Major manufacturers are investing hundreds of millions to figure out the right method to enable light to burn ever more Lilliputian lines on chips. “Even if they succeed in building this tool, they will lose. Chemistry will defeat them in the end, and the machine will never work.” According to Willson, the chemical catalyst diffuses and there’s blurring of lines that should be sharp. Furthermore, a single machine of this type would cost $80 million, says Willson, putting production costs ludicrously high.

So has the march of improvement in semiconductor technology ended? Willson sees hope yet, in the form of Step and Flash Imprint lithography (S-FIL), a new approach to high resolution patterning, which can replicate shapes as small as 10 nanometers and at reasonable cost.