When Japan’s Hiroo Kinoshita projected the first EUV images in the mid-1980s, no one wanted to believe him. Only after a second attempt and a difficult discussion with colleagues in 1989 did the starting gun ring out for EUV lithography as we now know it.
Viewed historically, EUV lithography can be considered a descendant of soft x-ray proximity lithography (SXPL), the first lithographic technique to use x-ray radiation. In SXPL, a mask is held just above the wafer and illuminated, so that the mask structures are transferred at a one-to-one scale. This technique came into use in the early 1980s and would continue to be researched for many years, side by side with EUV lithography. It wasn’t until the turn of the century that science and industry gave up their attempts to make SXPL work for commercial chip production.
The terms soft x-ray and extreme ultraviolet aren’t well defined, but indicate roughly the same wavelength range of a few to a few dozen nanometres. The term EUV was introduced to distinguish between SXPL, which uses a mask as large as the wafer itself, and projection lithography in the same wavelength range, which places the mask much farther from the wafer and prints the mask structures field by field.
In the early 1980s, projection lithography was the standard in the semiconductor industry. Chip manufacturers used the g-line (436 nanometres) emitted by mercury lamps for production, and i-line (365-nanometre) lithography was being readied. The industry was lightyears away from using extremely short wavelengths, but after the umpteenth failed SXPL experiment, Hiroo Kinoshita started to wonder if x-ray light and projection weren’t a better lithographic combination.
Movers and shakers
The problem that Kinoshita, working at the ‘Japanese Bell Labs’ NTT, kept running into with SXPL was the extreme precision with which the mask must be created. Because the structures are imaged at actual scale, every flaw ends up right on the wafer. And even though the industry was using smaller wafers at that time, an error-free mask was simply impossible. Every ‘typo’ made by the electron bundle writing the mask and every wisp of mechanical pressure could mean a ruined chip.
Device lifecycle management for fleets of IoT devices
Microchip gives insight on device management, what exactly is it, how to implement it and how to roll over the device management during the roll out phase when the products are in the field. Read more. .
Though IBM and other heavyweights fully supported SXPL, the mask problem was ultimately its downfall. By the end of the 1990s, the technique had advanced far enough to create working chips, but not far enough for leading-edge chips. Every time mask fabrication was sufficiently refined to produce advanced chips using optical lithography, chip manufacturers had already moved a step further down the road. SXPL’s development simply wasn’t fast enough to keep pace with the semiconductor industry.
Kinoshita is to thank for the fact we now have an alternative x-ray technology – EUV – at our disposal. He was the first to start experimenting with EUV projection lithography. But just like all scientists, he stood on the shoulders of others, and as so often happens, inventions originally intended for different applications turned out to be valuable in other fields.
In the early 1980s, astronomers became interested in multilayer mirrors, because they could be used to construct x-ray telescopes. The first image recorded using EUV light dates from 1981 and was captured by James Underwood at Caltech and Troy Berbee Jr. at Stanford. Using a 76-layer tungsten-carbon mirror, they were able to image a pattern containing five lines per millimetre on the x-ray-sensitive detector.
Astronomy and EUV lithography continued to cross-pollinate for many years. Microscopics and diagnostics also joined in the exchange. Many concepts, materials and components turned out to be broadly interchangeable, and thus the technologies were often developed side by side. Lawrence Livermore National Laboratory was one of EUV technology’s major movers and shakers in the pre-industrial phase, and is to this day an ASML partner.
Language barriers
Kinoshita tried doing lithography using the same type of tungsten-carbon mirror Underwood and Berbee had used. But he had incredible difficulty creating the delicate optics. What’s more, a maximum reflectance of 2 per cent meant the correct alignment of the optical components was absolutely essential. But Kinoshita had to ‘eyeball’ it – with an optical microscope – and hope to luck into an acceptable alignment.
Nonetheless, he succeeded in 1985 in burning a 4-micrometre pattern into a resist using 11-nanometre light (see figures 1 and 2). When he presented his results at a conference a year later, however, his colleagues were sceptical. They weren’t convinced that EUV light was responsible for the projected image. If so little EUV light was being reflected, then even the tiniest amount of ‘stray light’ might have fallen onto the substrate, they reasoned.
Meanwhile, a space flight laboratory in California had developed an EUV mirror with much better reflectance based on molybdenum and silicon – the same type still being used in EUV scanners today. That recipe gave Kinoshita a significant leg up, and in 1989 he convinced his colleagues with images of 0.5-micrometre structures. Researchers at Bell Labs, with whom he engaged in extended discussion after his presentation, were able to take things a factor of ten smaller a year later.
Despite the language barriers that plagued that 1989 discussion, Kinoshita and one of his conversational partners at the time, Obert Wood, view the exchange as the dawn of EUV lithography as we now know it, they jointly write in the comprehensive reference work EUV Lithography, edited by Vivek Bakshi.