Home
About Future-Fab
Download New Issue
Contact Us
Volume Archives
Editorial Panel

FUTURE FAB ARCHIVES


Resist Road to the 22 nm Technology Node
(6/21/2004) Future Fab Intl. Issue 17
By Uzodinma Okoroanyanwu, Advanced Micro Devices
Jeroen H. Lammers, Philips Research
Print Article Print this paper
Send As Email Send as email

The resist road to the 22 nm technology node is going to be a difficult one to ply. Conventional resist technology will be resolution-limited at the 22 nm design rule. Chemical amplification chemistry runs into “diffusional limits” with high-activation energy resist systems that are not extendible to this node, while lowactivation energy ones may be extendible to this node, but at the cost of significant outgassing and poor line-edge roughness (LER).

While the chemical amplification concept has served the semiconductor industry very well since its invention in the early 1980s[2-4] in terms of high sensitivity and resolution, it is becoming apparent that by 2016, when the 22 nm technology node is expected to be in production, the very attributes of this concept that made resists based on it become the workhorses of the industry for the last 25 years will unfortunately become resolutionlimiting due to uncontrollable diffusion, image spreading, or resist blur. A growing body of experimental evidence suggests that chemically amplified resists have an intrinsic bias that limits resolution,[5-12] which given aggressive scaling of gate length, critical dimension control will present a very difficult challenge at the 22 nm node.

Welcome to Future Fab Vol. 15

There are many stumbling blocks on the road to the 22 nm technology node. The lithography-related specifications of the 2003 International Roadmap for Semiconductors (ITRS)[1] for this node calls for a DRAM half-pitch of 22 nm, printed gate length of 13 nm, resist thickness of 40-80 nm, line-edge (LER) roughness (3-sigma) of 1 nm, and critical dimension (CD) control (3-sigma) of 1 nm. These specifications call for near-atomic-scale resolution – something that is impossible with the current resist design concepts and imaging mechanism. These stumbling blocks are intimately tied to the nature of resist chemical amplification imaging mechanism, thin film confinement effects, and polymer molecular properties. Acting either separately or in concert, these stumbling blocks are resolution-limiting, with disastrous consequences such as poor CD control, LER, and pattern collapse. How these three resist properties and issues limit resolution is discussed in more detail in the following sections.

Resolution Limits Due to Chemical Amplification

While the chemical amplification concept has served the semiconductor industry very well since its invention in the early 1980s[2-4] in terms of high sensitivity and resolution, it is becoming apparent that by 2016, when the 22 nm technology node is expected to be in production, the very attributes of this concept that made resists based on it become the workhorses of the industry for the last 25 years will unfortunately become resolutionlimiting due to uncontrollable diffusion, image spreading, or resist blur. A growing body of experimental evidence suggests that chemically amplified resists have an intrinsic bias that limits resolution,[5-12] which given aggressive scaling of gate length, critical dimension control will present a very difficult challenge at the 22 nm node.

Empirical results suggest that chemical amplification starts to limit resolution in high activation energy resist systems like those based on tert-butyloxycarbonyl-protected (t- BOC) resist systems as the pitch approaches 80 nm.[5-7] For instance, the blur at full width half maximum (FWHM) of state-of-theart DUV 248 nm and 193 nm environmentally stable, chemically amplified photoresists are 100 nm and 60 nm, respectively. State-of-the-art high-activation energy resist blur (at FWHM) is roughly 50-60 nm, while that of a state-of-the-art low-activation energy resist is less than 20 nm. Currently, FWHM blur as much as 50-90 nm have been observed at the 180 nm pitch. Lithographic patterning at the 22 nm node will require blur that should not exceed 15 nm, in order to capture the high-resolution aerial image. Blur value less than 15 nm may be difficult to achieve at high throughput, especially with weak EUV sources. Resist blur increases with resist sensitivity (amplification factor), and this will inevitably limit throughput at high resolution. Changing exposure wavelength does not address any of these issues.

The design concept of chemical amplification is based on generation of a chemically stable catalytic species – commonly referred to as a photoacid and designated as a proton H+, as illustrated in Figure 1 for a resist system comprising a copolymer: poly(4-hydroxystyrene- co-4-t butyloxycarbonyloxystyrene) – of the photoresist film that are irradiated. Rather than using exposure energy to directly cause a solubility switch, chemically amplified resists use exposure energy only to generate a catalytic species. The photogenerated catalyst then initiates a chain reaction or promotes a cascade of solubility-switching reactions in the exposed regions of the photoresist. The apparent quantum efficiency for the solubility switching reaction in such a system is the product of the quantum efficiency for catalyst generation and the catalytic chain length. Catalytic chain lengths in the many hundreds are common, so in effect the quantum efficiency can be greater than one. One photochemical conversion can cause several chemical reactions, and thus the exposure can be said to have been “chemically amplified.” In this manner, lithographic imaging can be accomplished with very low exposure doses, saving time and money in manufacturing.[11-12]


Figure 1. Catalytic deprotection reaction of a photoresist system based on copolymer: poly(4-hydroxystyrene-co-4-t-butyloxycarbonyloxystyrene).

While chemically amplified resists have many advantages, they have one major limitation: It is possible for the catalyst generated in exposed regions to diffuse into unexposed regions, causing blurring of the latent image (Figure 2). The migration of the photoacid effectively results in a bias between the distribution of exposure energy and the final distribution of exposure photoproducts. This bias depends upon the photoresist chemistry and the processing conditions, but is largely independent of the exposure conditions and therefore cannot be eliminated by improving the exposure process.

Figure 2. Exposure process of a chemically amplified resist, illustrating photoacid diffusion in the presence of a non-diffusing quencher.

It is, however, possible to reduce catalyst migration by increasing the size of the catalyst or by reducing the temperature of the post exposure bake (PEB) step. Another approach for mitigating the problem of feature dimension bias due to migration of acid catalyst is to incorporate base additives into the photoresist formulation. These approaches for reducing acid catalyst migration also generally reduce the catalytic efficiency of each photoproduct and therefore increase the total dose of exposure energy that is required to pattern the film. Increases in required exposure dosage effectively reduce the throughput of the expensive exposure tools and can seriously reduce the profitability of the manufacturing process.[11]

The fundamental equation governing the temperature dependence of diffusion as a rate process is that given by Gladstone, Laidler, and Eyring (1941)[13] in Equation 1, where e is the exponential constant, ë is the length of the diffusive jump, k is the Boltzmann’s constant, h is Planck’s constant, T is the absolute temperature, .S–: is the entropy of activation of diffusion, and R is the universal gas constant.

Photoacid diffusion in chemically amplified resists is temperature-dependent.[14] Within normal processing conditions, the higher the PEB temperature, the faster the rate of the photoacid diffusion is and the longer the diffusion length is. Although the PEB temperature can be used to modulate photoacid diffusion and consequently resist sensitivity, it nevertheless involves a tradeoff with LER and CD uniformity. While highactivation energy resists based on t-butyl ester and butoxycarbonyl protecting groups typically need high PEB temperature to be deprotected, low-activation energy resists based on acetal and ketal protecting groups typically need low PEB temperature, oftentimes room temperature while within the exposure chamber, to be deprotected. Because of their high PEB temperatures, the high-activation energy resists have much higher blur values and lower LER than their low-activation energy counterparts, which suffer more from high LER and outgassing. So in principle, chemical amplification concept is extendible to the 22 nm node with low activation energy, low blur, poor LER, and bad outgassing resists of the acetal and ketal platforms. High-activation energy resists in their current form are not extendible to this node because of their high blur.

Reduction of Resist Contrast Due to Diffusion – Linear Theory

The photoresist polymer experiences the average acid concentration c during PEB, leading to deprotection. Useful for line/space structures is the well-known gaussian response of the diffusion equation, to the initial concentration c (x,y,t = 0) = ä (x), where x and y are spatial coordinates, t is time, and ä the Dirac impulse. The average concentration field is then given by Equation 2, with D denoting the diffusion coefficient, and illustrated in Figure 3.

Figure 3. Shape of the average acid concentration due to diffusion, the term between brackets in equation 2.

 

Applying the superposition principle, one can calculate the response to an initially sinusoidal modulation of the photoacid concentration with pitch p, the Modulation Transfer Function (MTF) as MTFdiff = p2(1– exp(–4ð2 Dt/p2)/4ð2 Dt), shown in Figure 4. For example, suppose that one wants to know which diffusion length is allowed if one accepts a relative 30 percent reduction of the resist image due to diffusion, then the diffusion length Ld= 2 Dt may be as large as 0.2 times the pitch.

Figure 4. Modulation Transfer Function due to diffusion MTFdiff vs. pitch (p)/diffusion length (Ld)

In the continuum theory, one can shrink the pitch as long as one can shrink the diffusion length. A pitch of 45 nm requires a diffusion length of about 7 nm or less. The blur value (FWHM) is about 2.35 times this value. Even by tuning the size of the anion, the PEB temperature, and the coating thickness, blur value of this magnitude is going to be nearly impossible to achieve with high-activation energy resist systems, but seems achievable with low-activation energy resist systems.

Scaling of LER With Diffusion Length

Shot noise will be a significant contributor to LER at the 22 nm node. Poisson statistics stipulate that the relative variation of the number of absorbed photons N in a volume V scales as 1/ N. For a given number density, the larger the volume, which is the diffusion sphere, and is proportional to Ld 3, the relative variation scales as (Ld)-3/2. We conjecture that it is by increasing the diffusion length that one obtains lower LER numbers for a given dose at large pitches.[15] However, if the diffusion length becomes so large, or the pitch so small, that blurring due to diffusion becomes significant, then this will limit the ultimate resolution. Figure 5 shows the scaling relationship between LER scaling factor as a function of the ratio of diffusion length and pitch. Combining the Poisson statistics of the absorbed photons with the MTF results in the following observation: Increasing the diffusion length up to 0.2 times the pitch will reduce the LER, while beyond that length one has practically reached a plateau, such that there is no more gain in LER to be made, only further loss in exposure latitude. Preliminary indications[15] are that taking quencher statistics, or quantum yield into account does not significantly change the picture described by linear theory.


Figure 5. Line edge roughness scaling factor versus Ld/p ( Scaling factor: ((pLd)-3/2 /MTF(Ld/p), and asymptote (dashed) (pLd)-3/2) ).

 

When Does Chemical Amplification Cease?

The assumption that shot noise of the absorbed photons is a dominant factor for LER implicitly assumes that there is a large chemical amplification factor, such that the deprotection profile of the photopolymer is governed by continuum theory, and fluctuations are due to the statistics of the photon-absorption process only. We present an argument for why chemical amplification will cease to be possible around the 22 nm node.

Suppose that the amplification factor is approaching unity, and that 50 percent of the photoacid generator (PAG) (volume ~0.71nm3/PAG) is converted into acid, which then deprotects 50 percent of the protecting groups of the polymer (for t- BOC systems, volume ~0.22nm3/t-BOC group). This typical unit cell size, consisting of at least two PAGs and two protecting groups, will be of the order of 1.8nm3. To store the information of the location of the line edge, one needs at least 1,000 unit cells, which correspond to a cubic pixel of (10 nm)3 or a rectangular slab at the edge of 1 nm thickness, 60 nm height, and 30 nm length. Poisson statistics stipulate that the 3-sigma variation for an average number of 1,000 is 10 percent. So given this pixel size, the 3-sigma fluctuation on the location of the line edge is 1 nm. Now, the pixel size is set by the acid diffusion length, 10 nm in this example, which should be smaller than 0.2 times the pitch, from the argument presented earlier. Thus, one arrives at a crude estimate that for about 50 nm pitch one can no longer have chemical amplification if one wants to keep LER of the resist under control. Adding quencher does not dramatically change this picture.

Resolution Limits Due to Confinement Effects in Resists

Resist technologies that will enable the extreme ultraviolet lithography (EUV) that will be used to pattern the 22 nm technology node will need to boost CD control with appropriate LER of resist features to levels that seem unrealizable today. Given the significant attenuation of EUV radiation in organic materials, ultrathin resist (UTR) (thickness ¡Ü 100 nm) imaging appears to be the only viable option for patterning at the 22 nm technology node. The 2003 ITRS specifications call for a resist thickness of 40-80 nm for the 22 nm technology node. However, the stability of UTR films presents very difficult challenges due primarily to interfacial and confinement effects. In addition, UTR imaging will require materials (resists and underlayers) with superb etch selectivity and etch processes that lead to negligible roughness transfer between layer stacks. Practical considerations often impose the choice of a resist and a substrate that are not fully compatible, resulting in films that are unstable or metastable with finite relaxation time. While thick films may be stable or metastable due to gravity,[16,17] for thin films intermolecular and surface forces dominate.[18] UTR films are susceptible to both spontaneous thin-film instabilities due to Londonvan der Waals interactions, and defects caused by substrate imperfections (topographical features like pits or mounds).[19,20]

Because UTR films have higher surfaceto- volume ratio than their thicker counterparts, they are highly interfacial. Each macromolecular component of a UTR system is close to an interface, either the solid substrate or the free surface. Consequently, in UTR films, the material properties may be quite different from those of the bulk. In particular, the glass transition temperature (Tg) may be depressed or elevated by as much as 40ºC relative to the bulk values, depending upon the film thickness and the chemical nature of the solid substrate upon which the film is deposited. This may have consequences on the viscoelastic response of the film during subsequent thermal annealing. The interfacial properties of such polymerinorganic interfaces oftentimes determine their lithographic performance and consequently the performance of devices fabricated with them.

Experimental results suggest that material properties of UTR films can differ in significant ways from their bulk counterparts. For example, physical properties such as the degree of crystallinity[21] and diffusion coefficient[14] can not only be different in the ultrathin film state, but these properties become increasingly sensitive to film thickness, substrate surface energy, and local ordering into non-homogeneous and structured phases. Of particular interest, because of its influence on the viscoelastic behavior of the spin-coated film, is the effect of film thickness on Tg (see Figure 6). A variety of experimental methods including spectroscopic ellipsometry,[22- 24] X-ray reflectometry,[25] positron annihilation lifetime spectroscopy,[25] and Brillouin scattering[27] have shown that Tg depends on film thickness and on the chemical nature of the polymer-substrate interaction. For non-interacting substrates, e.g., hydrophobic polymer on a hydrophobic substrate, the Tg value can be depressed by as much as 40ºC.[28] By contrast, for moderately strongly interacting substrates, e.g., an oxide surface interacting with a polymer capable of hydrogen bonding,[26] the Tg value can be elevated by a comparable amount.[23- 24,29]


Figure 6. Glass transition temperature of resist polymer, and resist polymer with photoacid generator (from Shipley Co.) as a function of film thickness.[23,24]

 

Practical consequences of Tg modification in polymer films include significant changes of dissolution, diffusional and etching characteristics, mechanical creep behavior, and adhesion. For example, dissolution rate measurements of spincoated films of poly (3-methyl-4-hydroxy styrene), a common matrix resin in chemically amplified resists used in deep UV lithography, showed dramatic reduction in the diffusion coefficient of the developing base solution in the region close to the native silicon oxide surface of the solid substrate.[30] This was attributed to enhancement of the hydrogen-bonding network due to chain orientation effects. Figure 7 shows a plot of effective diffusion coefficient of perfluorooctane sulfonate photoacid as a function of film thickness of partially protected poly (4-t-butyloxycarbonyloxstyrene).[ 14] The profile shows asymptotic behavior at ~600A, below which diffusion slows remarkably, probably due to interfacial and confinement effects.


Figure 7. Effective diffusion coefficient of perfluorooctane sulfonate photoacid as a function of film thickness in partially protected poly(4-t-butyloxycarbonyloxstyrene). (Adapted from reference 14.)

 

Resist additives such as leveling agents, plasticizers, bases, PAGs, and the like can segregate and partition themselves at different layers of the film, driven by interfacial effects, which are magnified as the film thickness decreases.[31,32] As a result, UTR films tend to have higher LER than their thicker counterparts. They also tend to have higher susceptibility to poisoning effects and unexposed film thickness loss than their thick resist counterparts.[31,32] It is widely appreciated that understanding and controlling the causes of defects associated with UTR films are major challenges for the near-term development of UTR process technology.

Resolution Limits Due to Resist Polymer Molecular Properties

A 1 nm CD and LER tolerance of resist features as specified in the 2003 ITRS is smaller than the dimensions of the polymer molecules in today’s resists,[33] given a typical carboncarbon bond length of 0.13-0.15 nm.[34] A printed gate length of 13 nm is only a small multiple of the radius of gyration, rg (rg ~0.6×MW0.5, where MW is molecular weight) of typical amorphous polymers used in resists today. Therefore, the ITRS specifications call for near-atomic-scale resolution, which is realistically untenable with the present resist architectures. Resolution will in effect be limited, and LER will be negatively impacted, at the 22 nm node as the length scale of resist polymer material properties (polymerization length distribution, polymer end-to-end distance, and radius of gyration, etc.) approach the printed gate length in resists.

Resist Outlook for the 22 nm Technology Node

Image blur caused by acid diffusion into unexposed areas of resists can only be limited, but not eliminated, if high sensitivity due to chemical amplification is to be maintained. This is because the mechanisms responsible for high sensitivity in chemically amplified resists are intimately tied to the processes that lead to image blurring. This makes it extremely difficult, if not impossible, to have a highly sensitive, highresolution chemically amplified resist based on the conventional resist design concepts. Therefore, novel resist design concepts, incorporating different polymer architectures and photoacid generators that can achieve high sensitivity, should be explored. Such alternative resist design concepts could incorporate photo-catalysts with welldefined limited lifetimes, whose diffusion length at nominal processing temperature could be targeted to match the resolution requirements of the 22 nm design rule. This will ensure that at the barest minimum, the acid can only diffuse to a manageable distance that should fall within the geometry of the 22 nm design rule.

Low-activation energy acetal and ketalbased chemically amplified resists could be extendible to the 22 nm technology node, if ways are found to mitigate their outgassing and LER problems. It is of course possible for exposure tools to be designed such that outgassed species from resists are prevented from reaching the optical elements. It is also possible to smoothen the walls of patterned resist features using standard post-lithographic processing techniques like UV-hardening, electron beam irradiation, thermal flow, and surfactant rinse. Also, non-chemically amplified resists are, in principle, extendible to the 22 nm technology node, with highexposure doses and low throughput, if only the exposure tools can support that.

Conclusions

Conventional resist technology will be resolution limited at the 22 nm design rule. Chemical amplification chemistry runs into “diffusional limits” with high-activation energy resist systems that are not extendible to this node, while low-activation energy ones may be extendible this node, but at the cost of significant outgassing and poor LER. Non-chemically amplified resists are also extendible to this node, but at the cost of throughput and high exposure doses. For this node, photoresists will no longer be chemically amplified, at least as we know them today, for the following reason: Starting from molecular dimensions, we derive a typical minimum pixel size of (10 nm)3 in order to contain at least 1,000 sensitizers and switchable groups. With the pitch being at least five times the maximum diffusion length of 10 nm, the minimum pitch for chemically amplified resists is 50 nm. So for the 32 nm half-pitch node there will be hardly any chemical amplification; for the 22 nm there will be no chemical amplification at all. Thin film instabilities, confinement effects as manifested in the degradation of thermophysical properties, will become significant at the 40-80 nm specified resist thickness.

 

References

1. International Roadmap for Semiconductors, SEMATECH: Austin, TX (2003)

2. J.M.J. Frechet, H. Ito, C.G. Willson, Proc. Microcuit. Eng., 260 (1982)

3. H. Ito, C.G. Willson, Polym. Eng. Sci., 23, 1012 (1983)

4. H. Ito, C.G. Willson, J.M.J. Frechet, U.S. Patent 4,491,628 (1985)

5. C.P. Umbach, A.N. Broers, C.G. Willson, R. Koch, and R.B. Laibowitz, J. Vac. Sci. Technol. B 6, 319- 322 (1988)

6. C.P. Umbach, A.N. Broers, C.G. Willson, R. Koch, and R.B. Laibowitz, IBM J. Res. Dev. 32, 454-461 (1988)

7. W. Hinsberg, F. Houle, M.I. Sanchez, J. Hoffinagle, G. Wallraff, D.R. Medeiros, G. Gallatin, J. Cobb, Proc. SPIE 5039, 1-14 (2003)

8. J. Hoffinagle, W. Hinsberg, M.I. Sanchez, F. Houle, J. Vac. Sci. Technol. B 17, 3306 (1999)

9. M. Swittkes, M. Rothschild, Proc. SPIE 4691, 459 (2002)

10. M. Swittkes, M. Rothschild, J. Vac. Sci. Technol. B 19, 2533 (2001)

11. G.M. Schmid, M.D. Stewart, C.-Y. Wang, B.D. Vogt, V.M. Prabhu, E.K. Lin, C.G. Willson Proc. SPIE 5376 (2004), in press

12. M.D. Stewart, Ph.D. Thesis, The University of Texas at Austin (2003)

13. S. Glasstone, K.J. Laidler, H. Eyring, The Theory of Rate Processes, McGraw-Hill: New York, p. 524 (1941)

14. D.L. Goldfarb, M. Angelopoulos, J. Vac. Sci. Technol. B 19(6), 2699 (2001)

15. R.L. Brainard, P. Trefonas, J.H. Lammers, C.A. Cutler, J.F. Mackevich, A.Trefonas, S.A. Robertson, Proc. SPIE, vol. 5374 (2004), in press

16. P.-G. de Gennes, Rev. Mod. Phys. 57, 827 (1985)

17. F. Brochard, C. Redon, F.C. R. Rondelez, Acad. Sci., Ser 2, 306, 1143 (1988)

18. J.N. Israelachvili, Intermolecular and Surface Forces, Academic Press: London (1985)

19. T.G. Stange, R. Mathews, D.F. Evans, W.A. Hendrickson, Langmuir 8, 920 (1992).

20. S. Kalliadasis, C. Bielarz, G.M. Homsy, Submitted to Phys. Fluids (2000).

21. (a) M.M. Despotopoulou, R.D. Miller, J.F. Rabolt, C.W. Frank, J. Polym. Sci.: Pt.B: Polym. Phys. 34, 2335 (1996). (b) C.W. Frank, Private Communication.

22. J.L. Keddie, R.A.L. Jones, R.A. Coury, Europhys. Lett. 27, 49 (1994).

23. J.N. D’Amour, C.W. Frank, U. Okoroanyanwu, Proc. SPIE, 4690, pp. 936-942, 2002.

24. J. D’Amour, C.W. Frank, U. Okoroanyanwu, Advances in Resist Technology & Processing XX, Proc. SPIE 5039, pp.996-1007 (2003).

25. J.H. van Zanten, W.E. Wallace, W.L. Wu, Phys. Rev. E. 53, R2053 (1996).

26. G.B. DeMaggio, W.E. Frieze, D.W. Gidley, M. Zhu, H.A. Hristov, A.F. Yee, Phys. Rev. Lett. 78, 1524 (1997).

27. D.B. Hall, J.C. Hooker, J.M. Torkelson, Macromolecules 30, 667 (1997).

28. O. Prucker, S. Christian, H. Bock, J. Ruehe, C.W. Frank, W. Knoll, Macromol. Chem. Phys. 199, 1435 (1998).

29. C.W. Frank, V. Rao, M.M. Despotopoulou, R.F.W. Pease, W.D. Hinsberg, R.D. Miller, J.F. Rabolt, Science 273, 912 (1996).L.B. Rothman, J. Electrochem. Soc. 127, 2116 (1980).

30. L.B. Rothman, J. Electrochem. Soc. 127, 2116 (1980)

31. U. Okoroanyanwu, J. Vac. Sci. Technol. B 18(6), 3381-3387(2000).

32. U. Okoroanyanwu, Future Fab International, vol. 10, p. 157-163 (2001).

33. F. Billmeyer, “Textbook of Polymer Science,” Second ed., Wiley-Interscience, New York, pp 154-157 (1984)

34. G. Castellan, “Physical Chemistry,” Addison- Wesley, Reading, MA, Second ed., p. 578 (1971)

 

 

 
 
Search



Published By:
?¡Àcheap #167
38 Miller Ave.
Mill Valley, CA 94941
www.mazikmedia.com
converse@mazikmedia.com
Disclaimer | Privacy Policy