It is generally considered that excimer lasers for lithography tools require high running costs. Equipment with a relatively low running cost is much sought after. Because replacement of each module of a laser unit costs tens of millions of yen, many users even in the gigantic semiconductor industry worry about subsidizing the expense for their running cost. Also, since the laser unit used in combination with the lithography tool is one of the critical factors affecting wafer throughput, it must operate in a stable manner. Therefore, it has long been said that “Low Running Cost and Stable Operation” are musts for a laser unit.
Performance Means Running Cost and Stable Operation Only?
Does the performance of the laser unit actually affect such key performance indices for wafer processing as resolution and depth of focus (DOF)? The answer is, “Yes, it affects (or may affect) them greatly.” Typically, the spectrum, wavelength, and energy affect, respectively, the resolution, depth of focus, and uniformity.
So, why has this issue not yet been further discussed? The reason is that the lithography tool has been running with a sufficient performance margin; that is, for the target resolution, NA (Numerical Aperture) and k1 (k1 = Design rule x NA of Lithography Tool / Exposure Wavelength of Laser) are big enough.
Today, however, laser performance can no longer be ignored in the quest to achieve 90-nm node lithography by a KrF laser and 65-nm node lithography by an ArF laser.
The Influence of Spectrum
Let me explain the importance of the laser by giving a familiar example. Spectrum may be listed as one of the most important performance indices for the laser unit. We’ll discuss the details of spectrum later; the important point is that fluctuation of the spectrum directly affects the critical dimension (CD).
It is widely known that control over the gate size is the most important aspect in the formation of logic patterns. The gate is formed by isolated lines. In the ArF lithography process, the variation of the CD of each isolated line per 1 pm of spectrum is approx. 3 nm to form a 90-nm (k1 = 0.4) pattern. Also, it has been confirmed that if k1 is lowered to 0.35, it causes a variation of 8 nm per 1 pm. Taking into the consideration that the spectrum variation for the actual ArF laser is approx. 0.5 pm, the laser unit alone causes the CD to vary by 4 nm. Even in the KrF lithography process, the spectrum causes the same effect on the CD (though not to the degree of ArF lithography). KrF lithography causes a variation of 4 nm per 1 pm on a 110-nm (k1=0.35) pattern.
This is a typical example for understanding that the influence of the laser can no longer be ignored in current lithography processes that require tight CD control within a few percent.