Extreme ultraviolet lithography
From Wikipedia, the free encyclopedia
Extreme Ultraviolet Lithography (also known as EUV or EUVL) is a next-generation lithography technology capable of creating nanometer-scale patterns for use in semiconductor manufacturing.
Contents |
[edit] Background of EUV Development
There are several factors limiting the continued use of photolithography. The primary limitation is that the shortest wavelength of light used for lithography (currently 193 nm) cannot be decreased much further, or else it will begin to be absorbed by the glass lenses. To stave off this limitation, manufacturers have increased the numerical aperture of the system. However, that reduces depth of focus at a given resolution limit and is more sensitive to vectorial interference effects. Immersion lithography, where water or another liquid replaces the air between the final lens of the exposure tool and the substrate, allows a further increase of the possible numerical aparture and is expected to be used to extend the life of 193 nm lithography. To operate at a lower numerical aperture and a higher resolution, a massive decrease in the wavelength of light is required. With the choice of 13.5 nm radiation, EUV lithography is based upon this significant wavelength reduction. As of 2006, 13.5 nm is still considered the most likely lithography wavelength to succeed 193 nm.
[edit] Technical issues
While EUV offers a shorter wavelength than the current deep-ultraviolet (DUV) wavelengths, other technical issues limit its implementation, including:(a) the use of defect-free, smooth, reflective mirrors and photomasks; (b) the lack of a EUV source that is bright yet does not damage the optics; (c) scattering, or flare, and (d) photoresist limitations.
[edit] EUV mirrors
As stated, EUV radiation at 13.5 nm is readily absorbed by any material, including gases, so the system must function in ultra high vacuum (UHV). In addition, very expensive reflective optics must be used instead of the refractive optics (lens systems) traditionally used in DUV photolithography tools (steppers). Hence masks and optics both make use of Mo/Si multilayers which act to reflect light by means of interlayer interference, aka Bragg diffraction. A pressing issue in the development of EUV is the smoothness of the reflective mirrors. These mirrors will be coated with 81 layers of reflective material, giving a reflectivity at normal incidence of around 70% (the rest is absorbed).
If EUV is to be used beyond the 32nm manufacturing node, the mirrors will have to be smooth and clean to within 1 atom(link). However, these multilayers degrade in performance due to defects embedded in the multilayers, contamination, or damage of the top surface[1]. Embedded defects primarily disrupt the phase of the reflected light while surface defects primarily absorb the light. Defects in the multilayer are required to be less than 25 nm in size, yet this level of detection is not available with current inspection technology[2]. Defects of even smaller size located in the middle of the multilayer are still printable since the multilayer performance is disrupted most significantly at that location[3]. Yet, defects at this location would be virtually impossible to detect by optical or SEM-based methods prior to insertion in an EUV optical system, due to the fact that they are buried. An inspection microscope using the EUV wavelength is the only choice, but this increases the overall cost of EUV mask manufacturing, since it is a technology that must be introduced only for a limited application. Only EUV multilayers of a certain thickness can be inspected by such a tool.
[edit] Oblique illumination effects
A necessary feature of reflective optics is oblique incidence illumination. This imposes a stringent requirement on the flatness of the object (mask). The relationship for this is given by: image placement error = flatness error * tan(angle)/M, where M is the image demagnification[4]. For example, for a 6 degree angle of incidence, and an image demagnification of 4, it is required for the mask substrate to be flat within +/-38 nm to prevent an image displacement of +/-1 nm.
The finite source size results in a spread of illumination angles at the object (mask) plane. This results in an inherent asymmetry of the diffraction pattern, which ultimately limits the numerical aperture of a conventional projection system to at most around 0.25[5,6]. A recent study of this inherent asymmetry has shown that features of different widths shift by different amounts[7].
The maximum angle of incidence for adequate pattern fidelity is recommended to be 5 degrees[8]. This would allow less than 10% pattern fidelity error for 22 nm spaces.
[edit] EUV source
A conventional projection system with a numerical aperture of 0.25 is expected to contain at least two condenser multilayer mirrors, six projection multilayer mirrors, and a multilayer object (mask)[9]. With a total of nine multilayer reflective surfaces, each reflecting roughly 70%, the net throughput is about (0.7)^9 or 4%. Since the optics already absorbs 96% of the available EUV light, the ideal EUV source will need to be sufficiently bright. As of 2005, EUV source development has focused on plasmas generated by laser or discharge pulses. However, the mirror responsible for collecting the light is directly exposed to the plasma and is therefore vulnerable to damage from the high-energy ions[10]. In particular, Xenon ion sputtering is the main damage mechanism[11]. This damage associated with the high-energy process of generating EUV radiation has precluded the successful implementation of practical EUV light sources for lithography(see also this link).
[edit] EUV scattering
Scattering from imperfections in the optics is known as "flare" and can manifest itself as a non-zero background light intensity level. Flare worsens with shorter wavelengths, due to larger scattering. Flare is reduced with better quality optics. Alternatively, with a good characterization of the flare, one can attempt to compensate for it in the layout design, similar to optical proximity correction.
[edit] Ionization effects
For electromagnetic radiation, shorter wavelength means higher energy. A wavelength of 13.5 nm corresponds to an energy of 92 eV. This energy easily exceeds the bond energies and ionization potentials of all materials (typically several eV). When EUV radiation is absorbed, photoelectrons (electrons directly released by photons) and secondary electrons are generated by ionization, much like what happens when X-rays or electron beams are absorbed by matter[12]. These secondary electrons have energies of a few to tens of eV and travel up to 50 nm inside photoresist[13,14].
Secondary electron emission from the photoresist surface results in positive charging, which will tend to draw secondary electrons emitted elsewhere toward the surface[15]. The electrons emitted from the surface are eventually drawn back by the net positive charge of the sample, and penetrate the photoresist. As a result the EUV photoresist image contrast is degraded relative to the expected optical performance[16]. In addition, EUV secondary electron-catalyzed contamination [17] is expected to form upon the photoresist.
Ambient gases in the lithography chamber may be used for purging and contamination reduction. These gases are also ionized by EUV radiation, leading to re-emission of EUV light. This light is also considered as resulting from scattering. Hence, light absorbed by the ambient contributes to flare. Ionization of ambient gases also leads to plasma generation in the vicinity of the multilayer, resulting in damage to the multilayer and inadvertent exposure of the sample [18].
[edit] Photoresist
EUV photoresists exhibit properties of both optical photoresists and electron-beam resists, since EUV photons generate secondary electrons. The resolution is limited by the effective point spread function, rather than the diffraction limit due to the wavelength. Secondary electrons can travel as much as 50 nm[13,14,19], and there may also be long-distance backscattering effects in thin photoresist, due to secondary electrons from underlying layers. In particular, electrons with sufficient energy to dissociate a C-C bond (3.6 eV) exhibit a mean free path of around 20 nm[20]. Between scattering events, the electrons continue to travel at high velocity. Recent EUV photoresist images are shown and discussed here. Line edge roughness becomes a significant concern for small linewidths and thin photoresist films.
[edit] Mask Cost
Mask cost is a key consideration for EUV lithography especially when comparing this technique with alternatives. A recent estimate puts the cost at about the same level as an advanced optical photomask with aggressive optical proximity correction. The main contributors to the cost are the blank preparation and the inspection cost.
At the 22 nm node, it is likely that alternating phase-shift masks will be required[21]. As a result the cost would have to increase significantly. In addition, since this is a two-mask approach (see photomask), the cost will increase further and the resolution advantage of EUV over other techniques will be lost, as two simple masks (0.35 micrometre or larger) are sufficient for indefinite resolution extension via sidewall spacer approach[22].
A prohibitive factor for use of EUV masks is the sensitivity to nanometer-thin defects under and in the multilayer. Such defects can only be effectively detected by an EUV microscope[23]. The large drop in wavelength and conversion from transmission to reflection mode severely enhances topography defect sensitivity (down to 2 nm[23]) beyond the capability of the current DUV mask infrastructure.
[edit] Throughput and total cost
The per wafer cost of any lithography technology depends on throughput as well as equipment and tooling costs. EUVL's vacuum requirements, inefficient use of light and lack of high brightness sources significantly reduce the throughput that EUV steppers can achieve. At its current throughput of 5-10 WPH, EUV is substantially more expensive on a per wafer basis than current (2006) technologies. However, this cost equation needs to be updated for 32 nm half-pitch or smaller. Extending 193 nm immersion lithography to 32 nm half-pitch will require very expensive masks, high refractive index immersion fluids, or the use of double patterning techniques. Serial double patterning on the same lithography tool restricts the availability of that tool compared to conventional patterning. EUVL will need to boost throughput by at least several times from current levels to be the more cost effective approach at the 32-nm half-pitch technology node. The higher doses required by photoresists could be the overriding cost factor, independent of wavelength[24]. On the other hand, at this node, other patterning technologies will likely come into cost considerations, such as nanoimprint lithography or maskless lithography.
[edit] Outlook for EUV Lithography: Past, Present, and Future
Much time and effort has been invested by many companies into developing EUV lithography. In particular, EUV development put special emphasis and attention on photomask blank quality and defects[25]. This has actually helped improve the quality of photomasks for conventional photolithography. In this way, regardless of the deployment outcome of EUV lithography, the benefits of EUV lithography development are already being realized.
As of 2006, it has been nine years since EUV development officially started with the formation of the EUV Limited Liability Corporation in 1997 by Intel, Motorola and AMD. Originally, EUV was slated for 100 nm production in 2004, and expected to scale for several generations down to 30 nm[26]. However, technical difficulties have continually pushed out the rollout, more so than the 193 nm wavelength[27]. With the current adoption of immersion lithography, it now appears that EUV has been pushed out again beyond the 32 nm node[28]. Researchers at IBM recently demonstrated 29.9nm lines created with high-index immersion lithography[28]. It appears unlikely that EUV can pattern at 25 nm half-pitch or below due to optical reasons[5,6,8,9] as well as photoresist limitations. Hence, like 157 nm technology, it is questionable whether it should be inserted at all if it requires many infrastructure changes for only one node. Most importantly, it will be integration issues, such as multilayer defects, contamination and secondary electrons, that will prevent EUV from being implemented in manufacturing.
[edit] References
- Y. Katukani et. al., Proc. SPIE 5751, pp.1077-1083 (2005).
- Aaron Hand (2006). Industry Takes Further Steps Into Advanced Lithography. Semiconductor International. Retrieved on 2006-03-03.
- M. Lam, Ph.D. dissertation, U. of California, Berkeley, sec. 7.3 (2005). (Link: http://www.eecs.berkeley.edu/Pubs/TechRpts/2005/EECS-2005-28.html)
- S. E. Gianoulakis et. al., Proc. SPIE 3676, pp. 598-605 (1999).
- K. Otaki, Jpn. J. Appl. Phys. 39, pp. 6819-6826 (2000).
- M. Sugawara et. al., Proc. SPIE 5751, pp. 721-732 (2005).
- M. Sugawara et. al., Jap. J. Appl. Phys, vol. 44, pp. 8409-8421 (2005).
- M. Sugawara et. al., J. Vac. Sci. Tech. B, vol. 21, pp.2701-2705 (2003).
- F. T. Chen, Proc. SPIE 5037, pp. 347-357 (2003).
- H. Komori et. al., Proc. SPIE 5374, pp. 839-846 (2004).
- B. A. M. Hansson et. al., Proc. SPIE 4688, pp. 102-109 (2002).
- B. L . Henke et. al., J. Appl. Phys. 48, pp. 1852-1866 (1977).
- C. R. K. Marrian et. al., J. Vac. Sci. & Tech. B 10, pp. 2877-2881 (1992).
- D. J. D. Carter et. al., J. Vac. Sci. & Tech. B 15, pp. 2509-2513 (1997).
- I. A. Glavatskikh et. al., J. Appl. Phys. 89, pp. 440-448 (2001).
- J. Cain et. al., Proc. SPIE 5751, pp.1101-1109 (2005).
- J. Hollenshead and L. Klebanoff, J. Vac. Sci. & Tech. B 24, pp. 64-82 (2006).
- M. H. L. van der Velden et. al., J. Appl. Phys. 100, 073303 (2006).
- H. Seiler, J. Appl. Phys. 54, R1 (1983).
- D. Attwood, Soft X-rays and Extreme Ultraviolet Radiation, (Cambridge University Press, 1999), p.7.
- P-Y. Yan, Proc. SPIE 4889, pp. 1099-1105 (2002).
- Y-K Choi et. al., J. Phys. Chem. B, vol. 107, pp. 3340-3343 (2003).
- K. Hamamoto et. al., Proc. SPIE vol. 6151, pp. 361-7 (2006).
- U. Okoroanyanwu and J. H. Lammers, Future Fab International Issue 17 (2004).
- J. G. Maltabes, Future Fab International Issue 18 (2005).
- C. W. Gwyn et. al., J. Vac. Sci. & Tech. B 16, pp. 3142-3149 (1998).
- K. Lewotsky, "Making the Jump," OE Magazine, March 2001.
- Mark LaPedus (2006). Intel pushes out EUV lithography. EETimes. Retrieved on 2006-02-22.
- John Markoff (2006). I.B.M. Researchers Find a Way to Keep Moore's Law on Pace. The New York Times. Retrieved on 2006-02-22.
