Controlling CD and process window limits for implant patterning
10/01/2006
As design rules have shrunk, the overlay of implant layers has become a problem [1]. Existing topography on the substrate caused changes in the resist thickness, which led to varying resist CDs and limitations to the processing window [2]. Along with these problems came three main solutions: dyed photoresist and photoresist with either top or bottom anti-reflective coating (TARC or BARC). Each gives a definite improvement, but comes at the cost of greater implementation complexity and more processing steps. This article examines these options.
One problem that must be overcome in patterning implants is the reflectivity from the surface of the wafer; another is the process window limitation imposed by topography and the substrate’s varying reflectivity. For some technologies, a dyed photoresist or a photoresist with a TARC provides adequate control. For more advanced technologies that require even better reflectivity control, a photosensitive ArF BARC should be considered.
Three potential solutions exist to mitigate the problems of reflectivity and process window limitations. These solutions use 1) a dyed photoresist, 2) a photoresist with a TARC, or 3) a wet-developable bottom anti-reflective coating BARC with a photoresist.
Exploring solutions
First, the advantages and disadvantages of using a dyed resist will be discussed. Then, use of a TARC with a resist system will be examined and simulations presented. Finally, performance of a wet-developable BARC and photoresist will be compared with the other technologies. This comparison will include discussions of simulations and SEM pictures.
Dyed resist. The easiest solution to the current problems of patterning KrF and ArF implants is a dyed resist. The dye in the resist provides for reflectivity control without adding additional layers to the litho stack. Therefore, process development and implementation is simple and straightforward. In addition to this, the CD over topography is improved compared to a nondyed resist. A main disadvantage for a dyed resist is the sloping sidewall angles, which cause problems during ion implantation as they interfere with the travel of ions being implanted into the substrate. Thus, the sidewall slope can slow, or even stop, ions that are needed to dope the substrate. This can change the desired implant profile and the device’s electrical properties.
TARC. The next approach is to apply a TARC. This solution is fairly easy to implement from a processing standpoint. As an additional layer, TARCs are coated directly on top of the resist at a specific thickness. Then, after the exposure step, TARCs are developed away in the same step as the exposed resist. Implementation of a TARC from an optical perspective is simple as well. The refractive index (n) of the TARC and the specific film thickness work together to destructively interfere with light reflected from the resist. This flattens the reflectivity swing curve from the resist, as shown in Fig. 1.
 Figure 2. Substrate reflectivity. The orange line indicates the TARC’s inability to control light reflected off of the substrate. |
The limitations of a TARC and resist system arise from where a TARC sits in the litho stack. A TARC can control light only at the TARC
esist interface. Light reflected off the substrate or other structures coated by the resist cannot be controlled, as shown in Fig. 2. Therefore, a TARC cannot control substrate reflectivity, reflective notching, or changes in reflectivity from varying resist thicknesses.
 Figure 3. CD swing curves for the three systems. The blue line indicates that the BARC flattens the swing curve the most. |
Wet-developable BARC. The best solution for controlling all types of reflectivity and providing optimum CD control for implant layers is to use a wet-developable BARC with photoresist. Previous wet BARC technologies relied on polyamic acids as binder polymers and were thermosetting systems. They developed isotropically and offered a narrow bake window [3-5]. The ratio of BARC thickness to resist thickness using KrF systems and tight process control allowed this technology to succeed. But for ArF technology, the BARC-to-resist ratio changes to ~1:4. If the BARC represents 20% of the film stack, an isotropic antireflectant film would undercut and might lift the photoresist lines during development. A new approach was taken with the wet ArF BARC technology whereby it was made photosensitive to allow the BARC to develop anisotropically. A BARC significantly reduces the reflectivity as compared to a TARC, as shown in Figs. 2 and 3. Figure 3 compares Prolith simulations of a resist-only system, TARC-resist system, and a BARC-resist system for 100nm 1:1 lines. The swing curve with the addition of a TARC is better than for a resist alone, but a BARC and resist flatten out the amplitude of the swing the most.
 Figure 4. a) Topography coverage and b) lithography results of the BARC-resist system. |
Simulations are helpful for gaining an understanding of reflectivity control for flat substrates, but the true benefit of using a BARC over topography cannot be simulated; it must be tested. Figure 4 shows how a BARC and resist system coats 180nm dense lines with a 150nm step. The space between the BARC and photoresist is caused by the BARC shrinking during SEM exposure. Good compatibility and very good adhesion was observed for the ArF 193nm wet BARC as well. As shown in Fig. 4a, the BARC helps to planarize the topography and provides a surface with less variation for the resist to coat. Figure 4b shows the lithography results of patterning semi-dense lines perpendicular to the existing topography. The measurements show that the CD in the space was 115.4nm, and at the top of the topo line, the CD was 104.2nm. Over this extreme topography, the CD variation was <10%. Good adhesion and clearing of the BARC in exposed areas can also be seen.
To achieve this level of reflectivity and development control takes some additional processing work. Like a TARC, the addition of a BARC requires another coating and baking step and since the BARC is photosensitive, the post-exposure bake (PEB) of the resist also serves as the PEB of the BARC. Thus the resist and BARC process must be designed and tuned together.
Conclusion
After comparing the available technologies for patterning implants, it is clear that the engineer has choices in developing a process. A dyed photoresist covers existing topography well and thus provides adequate CD control. In contrast, due to its light-absorbing nature, a dyed photoresist can have sloping sidewalls, which reduce the control of the implant area. A photoresist with a TARC is a bilayer stack and thus requires two coating steps. The TARC can be applied to an existing process, but because the TARC is on the top of the litho stack, it cannot completely control reflectivity, has no effect on reflective notching, and provides less than optimum CD control.
A developable BARC plus a photoresist provides the best CD control and therefore the best process. This process must first be optimized, as improper processing can result in undercutting, footing, or post-develop residue. This wet-developable BARC technology mitigates implant-patterning problems and meets predetermined product specifications.
Acknowledgment
Prolith is a registered trademark of KLA-Tencor Corp.
References
1. D.C. OweYang, H. Chen, R.M. Deng, B.C. Ho, “Process Improvements by Applying 193nm Lithography to 90nm Logic Implant Layers,” Proceedings of SPIE, Vol. 5038, pp. 1095-1105, 2003.
2. X. Shao, A. Guerrero, Y. Gu, “Taking the Wet-Developable Route to Applying BARC in Implant Layers,” Solid State Technology, Vol. 47, No. 6, pp. 61-64, 2004.
3. Y.H. Lim, Y.K. Kim, J.S. Choi, J.G. Lee, “Process Optimization of Developer Soluble Organic BARC and its Characteristics in CMOS Devices,” Proceedings of SPIE, Vol. 5753, pp. 690-698, 2005.
4. C. Cox, D. Dippel, C. Ghelli, P. Valerio, B. Simmons, A. Guerrero, “Developer Soluble Organic BARCs for KrF Lithography,” Proceedings of SPIE, Vol. 5039, pp. 878-882, 2003.
5. I. Guilmeau, A. Guerrero, V. Blain, S. Kremer, V. Vachellerie, D. Lenoble, et al., “Evaluation of Wet-Developable KrF Organic BARC to Improve CD Uniformity for Implant Application,” Proceedings of SPIE, Vol. 5376, pp. 461-470, 2004.
Carlton Washburn received his BS in mechanical engineering from the U. of Missouri-Rolla and his BS in physics from Illinois College. He is an applications engineer at Brewer Science, 2401 Brewer Drive, Rolla, MO 65401; ph 573/364-0300. e-mail [email protected].
Ramil Mercado received his PhD in chemistry from the U. of Massachusetts and is a scientist at Brewer Science.
Douglas Guerrero received his PhD in organic chemistry from the U. of Oklahoma and is a scientist at Brewer Science.
Jim Meador received his BS from the U. of Illinois with a major in chemistry and his PhD in organic chemistry from Cornell U. He is a senior research scientist at Brewer Science.