English is currently available.
Nanotechnology Engineering DISCUSSION
I am reading about chip fabrication and cannot reconcile two facts. Deep-ultraviolet scanners use light with a wavelength of 193 nm, and I was taught that optics cannot resolve details much smaller than the wavelength. Yet these tools were used to make transistors with features of a few tens of nanometres long before EUV arrived.
What actually sets the smallest printable feature, how do immersion and multiple patterning fit in, and what does EUV at 13.5 nm change?
The resolution of a projection system is given by the Rayleigh criterion: CD = k1 × λ / NA, where CD is the half-pitch of the finest repeating lines, λ the wavelength, NA the numerical aperture of the lens and k1 a process factor. For a single exposure of dense lines, k1 cannot go below 0.25; that is the hard optical limit, because at least two diffracted orders have to pass through the lens to form an image of the pattern.
So the limit is not the wavelength itself but λ / (4 × NA). With λ = 193 nm and a dry lens of NA = 0.93, the smallest half-pitch is 0.25 × 193 / 0.93 = 52 nm, already about a quarter of the wavelength.
Immersion raises the NA. NA = n × sin θ, and in air it cannot exceed 1. Filling the gap between the last lens element and the wafer with ultra-pure water, which has a refractive index of about 1.44 at 193 nm, allows NA up to about 1.35. The limit becomes 0.25 × 193 / 1.35 = 36 nm half-pitch.
Getting k1 down from a comfortable 0.5 or more toward 0.25 took a set of techniques: off-axis illumination matched to the pattern, phase-shifting masks, and optical proximity correction, where the mask shapes are deliberately distorted so that the blurred image comes out as intended. The mask pattern no longer looks like the printed one.
Below about 36 nm half-pitch, the same 193 nm tools are used more than once per layer. In litho-etch-litho-etch, two interleaved patterns are exposed and etched separately. In self-aligned double patterning, lines are printed at a relaxed pitch, a thin film is deposited on their sidewalls, and the original lines are removed, so the sidewall spacers remain at half the original pitch: an 80 nm pitch becomes 40 nm. Doing it twice gives a quarter of the pitch.
The Rayleigh limit applies to the pitch of one exposure, not to the width of an isolated line. A single line can be made narrower by over-exposure or by trimming in the etch step, which is how gate lengths were well below the half-pitch for years.
EUV resets the equation with λ = 13.5 nm. With NA = 0.33 and a practical k1 of about 0.4, CD = 0.4 × 13.5 / 0.33 = 16 nm half-pitch in one exposure, replacing three or four immersion exposures on the critical layers. The cost is everything else: EUV is absorbed by all materials including air, so the optics are multilayer mirrors in vacuum, the mask is reflective, and the source has to deliver enough power through a chain of mirrors that each absorb a good part of the light.
Depth of focus shrinks as well, as k2 × λ / NA², so raising NA buys resolution at the price of a thinner usable focus range and thinner resist. One more caution: node names such as 7 nm or 5 nm are labels and do not correspond to a physical half-pitch on the chip.
Back to Nanotechnology Engineering community
