1984 · Reduction Projection Stepper · Work 874 of 1,871
An optical exposure machine in a clean room of a chip factory in Japan under amber yellow light, an open machine two metres high with its covers removed, at the top a tall cream-white column of lamp housing and optical blocks, below it a white lens cylinder on a heavy black base frame, in front a tower of fifteen flat white cassette slots stacked one above another, beside it a wafer table of bare aluminium parts, red and yellow cables, small red levers, a hand in a white glove sliding a flat cassette into one slot, further machines and a worker in a white clean-room suit seen from behind soft in the background.
What it was used for
A stepper does not expose the whole wafer at once but one field after another, reducing the mask image as it goes. Kato's chronology describes the GCA DSW 4800 of 1978 as the first successful wafer stepper: g-line, 10x reduction, a Zeiss lens of numerical aperture 0.28, field 10 by 10 millimetres; Mack's list gives its list price as 450,000 dollars. By chiphistory.org a 10x stepper managed 11 wafers of 100 millimetres per hour at about 600,000 dollars, against 40 wafers per hour for a projection aligner.
What was new
The stepper was slower and dearer and prevailed anyway because it was cleaner: chiphistory.org gives about 1 defect per square inch for producers who mastered repeating defects, against about 4 for projection, and about one fifth of the cost per 64K DRAM. By Kato, it was only the 256K DRAM generation that made the stepper mainstream; he records over 1,000 steppers shipped worldwide in 1984, close to 600 of them in Japan, and Nikon's first i-line stepper, the NSR-1010i3. The barb: the successful stepper had been built by GCA; the firm's share of the Japanese stepper market fell from 68 to about 45 percent between 1981 and 1983, by Kato's figures, while American suppliers had still held 90 percent of the whole lithography tool market in 1980.
What the sources leave open: This entry describes the type of tool, not a particular 1984 machine; the documented data belong partly to 1978 (DSW 4800) and partly to 1984 (unit numbers, NSR-1010i3), each dated in the text. chiphistory.org attributes Mostek's failed all-stepper ramp to repeating defects from an optical path that was not kept clean; that is that page's judgement.