Challenges loom as the semiconductor industry struggles to keep pace with Moore's Law. Many of these challenges are lithography related -- the 157-nm and extreme-UV (EUV) generations are confronted with a host of materials and source issues. To satisfy the continual need for shrinking critical dimensions, the incumbent technology, 193-nm lithography, must push to ever higher resolutions. Enter 193-nm immersion lithography.
Immersion lithography is based on an idea conceived of more than 125 years ago.1 Although the technical possibilities of the approach have been known for several decades, only recently has the industry seriously pursued it. In the search for a high-refractive-index fluid compatible with the tools and materials used in integrated circuit (IC) processing, water has surfaced as the enabling media for this "new" technology.2-4 The optical properties of water are such that its index increases significantly in the wavelength region below 250 nm. At 193 nm, the excitation wavelength of the argon-fluoride (ArF) excimer laser, the refractive index of water is 1.437 while its absorption remains low at 0.05 cm-1.5,6
Until recently, 193-nm lithography was expected to reach its resolution limit with the 65-nm IC device generation, or node. Once chip designs reached this point, the semiconductor roadmap called for the use of 157-nm and EUV lithography, but the aforementioned problems cast doubt on their availability for this near-term need. Alternatively, by replacing the air space in the image plane of a lithographic imaging lens with water, we can enhance the resolution at 193 nm by the increase in refractive power, which is 44%. This shift extends current optical lithography to sub-45-nm device generations. Diving In
Progress in microlithography is based on pushing the extremes of the optical resolution limit, at which high contrast photoresist images are produced with very little diffraction information. Using incoherent illumination for a particular wavelength λ and numerical aperture (NA = n sinθ), we can express the pitch, or the minimum periodic separation between device features, as:
Minimum feature pitch = 0.5λ/NA
= 0.5λ/(n sinθ) (1)
where n is refractive index and θ is the half acceptance angle of the lens (see figure 1). The semiconductor industry has used aggressive optical resolution-enhancement techniques to meet the needs of the IC manufacturers, driving wavelengths deep into the UV spectral region and designing systems with NAs close to 1.0. Optical limitations are primarily tied to wavelength, though NA values above 1.0 are not physically allowed in air-media imaging systems in which the refractive index is 1.0. There is a way around the NA condition, however. If we replace the air space after the last imaging element of a projection lens with a higher-index medium, NA values above 1.0 become possible.
Figure 1. Numerical aperture (NA) in a projection imaging system is defined by the half acceptance angle θ and the refractive index n of the medium as NA=n sinθ.
In a conventional "dry" imaging system, a wavefront propagates from a lens system into air and then into a photoresist film. The largest angle allowed in the photoresist film is inversely related to its refractive index, as calculated using Snell's law and assuming a 90° half angle in air. If we could increase this photoresist angle, we could couple additional diffraction energy into the film, directly leading to an increase in system resolution. We can accomplish this by taking the air out of the equation.
Consider the propagation of a wavefront through three media with refractive indices n1 (air), n2 (an immersion fluid), and n3 (a photoresist film) where n3 > n2 > n1 (see figure 2). In this scenario, a wave train enters into the last optical element of a lens system, designed as a glass hemisphere in direct contact with the immersion fluid. The outer spherical surface allows angles in the glass to be equivalent to those in air, avoiding refractive effects by forcing the air-glass interface always to be at normal incidence. By matching the refractive index of the spherical element to the immersion fluid, we can also eliminate refractive and reflective effects at their interface.
Figure 2. In immersion lithography, the optical wavefront passes from air (n1=1.0) through a spherical lens (n3), an index-matching immersion fluid (n2), and into the photoresist film (n3); note, n3 > n2 > n1. The NA for the photoresist, immersion fluid, and glass is defined as ni sinθi for i=1,2,3.
As the wave train passes into the photoresist film, angles are limited only by the differences between refractive indices, allowing numerical apertures (ni sinθi) as large as refractive index values of the corresponding media. Ideally, the glass, photoresist, and immersion fluid would have identically large refractive indices, singling out no medium as the limit to NA. Although water has a refractive index larger than that of air, it is currently the limiting medium when used with photoresists with npr ≈ 1.70. Our group is currently working to identify additives to water that would push its refractive index toward that of the photoresist without significantly increasing its absorption.7
By using water as an imaging medium for 193-nm lithography, we can extend the resolution limits of optical lithography. The attractiveness goes beyond mere resolution improvement, however. In contrast to the next-generation technologies, the source, photoresist, photomask, and optical materials infrastructure for 193-nm lithography are already in place, streamlining the development process and minimizing development cost. The experience gained at 193 nm can also be applied to a shorter-wavelength technology like 157-nm lithography, assuming sufficient time to solve associated technical challenges. Certain issues do, however, become a concern in the transition from conventional dry optical lithography to immersion lithography. Several optical and fluid-mechanical concerns have recently been explored. Challenges
The resolution potential indicated in equation 1, combined with the optical depth of focus (DOF), provides an indication of the capabilities of an optical system. Substrate planarization during the IC process limits the need for substantial DOF, which has made very high NA values feasible. We can apply conventional Rayleigh DOF definitions to immersion lithography by modifying the familiar paraxial form of DOF to account for the impact of the media refractive index:
DOF = ±k2 (λ/(n sin2θ)) (2)
where k2 is a process-dependent factor, determining the allowable image blur from defocus, generally near a value of 0.5. Although DOF does decrease with refractive index, the loss is linear, as opposed to the quadratic loss resulting from an increase in sinθ. This can be considered an attribute to immersion lithography.
Issues relating to other high-NA effects may be more of a concern, however. For the most part, optical microlithography has been carried out with unpolarized or circularly polarized radiation. Differences between orthogonal states of linear polarization (TE and TM) are generally small for NA values below 0.70, at which half acceptance angles in a photoresist film (n = 1.70) may be near 25°. In the case of immersion lithography, however, half angles approaching 90° become feasible, though 70° may be more of a practical limit. At these angles, image contrast from the TM polarization state drops toward zero, severely limiting the ability of unpolarized imaging to yield useful results. Consequently, polarization control throughout the entire optical lithography system becomes a concern, including the source, the illuminator, the photomask, the objective lens, and into the photoresist and any underlying substrate materials. Immersion lithography and polarization control will most likely be tied together for next-generation lithography systems as we pursue the full potential of each technology.
The choice of water as an immersion fluid for optical microlithography reduces many process compatibility issues. The photoresists used in IC manufacturing are polymeric in nature and inherently hydrophobic, while fused silica optics are hydrophilic. Water lets us take advantage of the surface tension differences between these two media, allowing the photoresist-coated semiconductor substrate to leave an exposure tool dry while confining water as an optical component and retaining it in the exposure system.
Several groups have devised workable approaches for water introduction and containment. A bath-immersion configuration involves submerging the entire image substrate in water, along with the last element of the imaging lens. Alternatively, a shower configuration flows a stream of water a few centimeters wide continually across the image field throughout the exposure until the substrate is removed. Both approaches ensure dynamic flow of the water, allowing replenishment and consistency of the immersion fluid. The third method, a static one, involves forming a small meniscus in the sub-millimeter gap between the lens and the image plane. Although the static approach has proven useful for R&D, the dynamic shower method is becoming most attractive for application to manufacturing systems.8
Microbubbles have presented an initial problem to immersion lithography. Water is saturated with air at room temperature, and microbubbles on the order of 1 µm in diameter or more escaping during exposure could degrade image quality. The problem is a solvable one, however, as stepper manufacturers can essentially eliminate scattering effects in the far field and remove near-field imaging effects by degassing the water as part of its purification process prior to introduction to an exposure tool.9Theory to Reality
The implementation of water-immersion lithography is underway on several fronts. Development of systems involving selective interference of two or more beams of UV radiation in water has provided proof of feasibility.7,8,10,11 Using a system with a 193-nm ArF excimer laser and an equivalent NA of 1.07, our group has imaged 45-nm features in a 100-nm photoresist film using TE-polarized illumination (see figure 3).7 We have achieved 38-nm resolution with this system at an NA value of 1.26. Early projection steppers and scanners with NA values as large as 1.05 are making their way into the field. Full-field commercial tools should be available late in 2004.8,11,12
Figure 3. Using TE-polarized illumination from an ArF laser, a test immersion lithography system with a 1.07 NA imaged these 45-nm features in 100-nm-thick photoresist.
Although the industry has yet to achieve consensus regarding the timing of lithography technologies, it is generally agreed that sub-65-nm devices are good targets for 193-nm water-immersion lithography.13 At the 45-nm generation, slated for insertion in 2010, dry 193-nm lithography is virtually impossible to use, while 157-nm technology retains technical material challenges. By adopting immersion lithography, the industry can address the 45-nm generation and the 32-nm generation that follows in 2012 using 193-nm technology. Beyond that time, 157-nm lithography may have reached maturity to enable its insertion as an immersion lithography approach. This, of course, requires the identification of a suitable immersion fluid, as the transparency of water is lost below 185 nm. This is still a challenge with today's materials but is certainly a reasonable goal for the future. oe
1. E. Abbe (1878).
2. B. Smith, et al., SRC Program Review 2000, July 2000.
3. M. Switkes and M. Rothschild, J.Vac. Sci. Technol. Vol. B 19, p. 2353 (2001).
4. B. Lin, J. Microlith., Microfab., Microsyst. 1, p. 7 (2002).
5. J. Burnett and S. Kaplan, J. Microlith., Microfab., Microsyst. 3 p. 67 (2004).
6. B. Smith, et al., Proc. SPIE 5040, p. 679 (2003).
7. B. Smith, et al., Proc. SPIE 5377, p. 68 (2004).
8. B. Streefkerk, et al., Proc. SPIE 5377, p. 285 (2004).
9. Y. Fan, et al., Proc. SPIE 5377, p. 477 (2004).
10. A.K. Raub, S. Brueck, Proc. SPIE 5377, p. 306 (2004).
11. S. Owa, H. Nagasaka, and Y. Ishii, Proc. SPIE 5377, p. 264 (2004).
12. T. Honda, et al., Proc. SPIE 5377, p. 319 (2004).
Shutterbug Champions Semiconductor Industry
Most shutterbugs live for the "click" of the camera. But for an adolescent Bruce Smith, the allure of the darkroom and the understandable mysteries of chemistry and developing photographs shaped a career that would one day enable new generations of computers.
Microlithography, or nanolithography as it is increasingly called today, is the method by which micrometer and nanometer features are patterned on silicon wafers to create microprocessors. Like the root of the word, microlithography's origins are firmly planted in the optical and chemical physics of the printing industry. "All through high school, I had an interest in physics, chemistry, optics, and photographic science," says Smith. "I started classes in photographic science at [Rochester Institute of Technology (RIT; Rochester, NY)] while I was still in high school and figured I would end up working for someone like Eastman Kodak. That was in the '70s and '80s when the semiconductor industry really started growing."
Eastman Kodak lost out when Smith graduated from RIT and headed to Gould AMI Semiconductor (Santa Clara, CA), and then to Digital Equipment Corp. in 1986, but upstate New York certainly did not. By 1988, Smith had married and returned to RIT and the warmth of family on both sides of the wedding aisle. Now, with a wife and three children at his side, Smith is tackling the challenges of immersion 193-nm microlithography and looking forward to conquering the next technical hurdle facing the semiconductor industry.
"We've been able to push optics, as we've known it for the past 200 years, into lithographic applications, but at some point we will not be able to push it any further into the subwavelength regime. At that point, we're going to have to use [extreme UV], or a variation on electron beam lithography, and there are still tremendous technical challenges that need to be addressed to get the device generation out beyond 2007," Smith explains.
Bruce Smith is the associate dean of engineering and Intel Professor of Microelectronic Engineering at the Rochester Institute of Technology, Rochester, NY.