Optimizing lithographic stack materials when using hyper-NA exposure tools

J. Macie, D. Miranda; Z. Zhu, B. Smith, Brewer Science

Executive overview

As lithography pushes past 32nm resolution, the need to optimize stack materials, including resist, bottom anti-reflective coating (BARC), and substrate has never been greater for IC manufacturers. Material and process complexity has increased dramatically with the introduction of exposure tools having a high numerical aperture (NA >1.2). With thinner resists, double patterning, dual BARCs, and tri- and quad-layer schemes in 193nm immersion technology, the potential combinations are almost endless. To reduce costly iterative experimentation and time in test and integration, process engineers rely on simulation software. A new simulation tool enabling engineers to more accurately model complex systems and make more informed decisions when selecting the best material solutions has been developed.

In prior technology nodes, a single-layer BARC provided adequate adhesion, planarization, and etch resistance, and minimized interference due to reflections from the substrate. Process engineers chose the anti-reflective layer with the lowest reflectivity possible to achieve 90° profiles. In general, the BARC with the lowest reflectivity sufficed to eliminate UV standing waves in the photoresist and provide straight profiles and wider process margins.

In terms of accuracy, conventional BARC simulation gives a single value for the overall reflectivity regardless of illumination coherency. This is not an adequate characterization method under incoherent illumination conditions because standing waves from different incident angles are phase shifted with respect to each other, which results in a single smooth standing wave. Additionally, a single reflectivity value does not account for the pattern-related details of the image. Because diffraction angles of the aerial image are pattern-dependent, it follows that reflectivity should be pattern-dependent, and therefore distributed in the aerial image plane.

Hyper-NA tools pose new challenges

In the new regime of hyper-NA exposure tools, minimum reflectance no longer guarantees optimal lithography results. With continued shrinking of feature sizes, chemical interactions become much more pronounced. These effects can exceed the accuracy currently afforded by modern optical simulation tools. Two chemically distinct layered substrates with the same optical parameters can give very different lithographic results. In most cases, optimizing solely for minimum reflectivity does not give the best resist profile and process window possible. Conventional simulation is therefore limited in its ability to predict good stack design.

The OptiStack tool was developed to examine the optical interference of the substrate reflection and predict processing conditions to compensate for chemical interactions using improved optical UV distribution at the foot of the resist. In addition to simulating reflectivity, the modeling simulates the diffracted image intensity at the top and foot of the resist, and accounts for the various diffracted angles created by the pitch of the pattern. The foot exposure (FE) parameter is the ratio of UV intensities near the resist bottom (averaged in a diffusion region) to the average intensity at the line edge. FE is closely related to the optical reflectivity phase shift. Experimental validation shows that a small amount of substrate reflection with a controlled optical phase shift can dramatically improve the CD profile and processing window in different stack designs.


Figure 1: Single-layer BARC simulation and experimental results: a) effective reflectivity, b) UV distribution in resist, c) cross-section SEM pictures and stack design.

For a single BARC, the simulation and experimental results are shown in Figure 1. The blue curve is the effective reflectivity, and the red curve is FE. Three BARC thickness points on the curves around the first minimum reflectivity were used for lithography testing. Fig. 1b shows the UV distribution in resist across the line pattern and along the vertical direction. The peaks of the standing wave are shifted from each other depending on the optical phase shift. The measurements are taken for different exposure doses and depths of focus. From these results, we can make the preliminary conclusion that minimum reflectivity does not give the widest process window. The widest processing window was measured where there is enough FE level even though the effective reflectivity is as high as 2.5%. Fig. 1c also shows that low FE can predict footing problems.

Dual BARCs are being designed to address issues associated with hyper-NA tools including control of reflectivity without reflectivity swing, reduction of reflectivity bias with varying topography and varying pitch, and the x- and y-polarization that occurs at high reflection angles. When the new tool is used in conjunction with a dual BARC system, such as the ARC 100 coating series, a tailored solution can be optimized to meet the demands of the user’s specific application and stack.

Case studies

To illustrate the subsequent data, Figure 2 shows the anatomy of the simulation contour plot. The x- and y-axes represent the lower and upper BARC thicknesses, respectively. On the plots of reflectivity, the darkest blue areas indicate the lowest reflectivity (≤0.5%), and on the FE plots, the higher intensity into the red region indicates a higher FE. The ideal target area is to have as low reflectivity as possible with an FE close to 1.00.


Figure 2: OptiStack simulation results for dual BARC platform.

To demonstrate the capability of the simulation versus actual litho performance, the following evaluation was performed. Using the simulation tool, with an NA = 1.35, targeting a 40nm 1:1 line/space pattern with a total dual BARC stack thickness of ≤ 80nm, ARC 113 and ARC 145 coatings were selected for the dual BARC platforms. These materials represent the upper and lower BARC layers with k values of 0.13 and 0.45, respectively.

Figures 3a and 3b illustrate the target area for the dual BARC stack, where the optimum percent reflectivity (R%) and FE will be found. In this scenario, an optimal BARC thickness for top and bottom layers based on the optical properties of the stack are defined within the triangles for reflectivity and FE. Based on these parameters, the simulation indicates that top and bottom layer thicknesses of 40nm each offer the most robust area for optimal lithography performance.


Figure 3: Dual BARC reflectivity and FE plots.

These simulations were validated through lithography performed at IMEC, where the target parameters were tested as closely as possible to simulated conditions for evaluation of the dual BARC platform’s performance. The results of this testing are shown in Figure 4.


Figure 4: ARC 113/ARC 145 dual BARC stack results from IMEC.

Tri- and quad-layer materials are required for pattern transfer due to the use of a very thin resist layer to meet resolution requirements and to improve process windows. Silicon hardmask and spin-on carbon layer technologies also lower cost and improve throughput as compared to hardmasks applied using chemical vapor deposition (CVD). Silicon-based hardmasks have very high resistance against oxygen plasma etching. Figure 5a illustrates a tri-layer application where a very thin hardmask is sufficient to open a thick spin-on carbon-based underlayer, and therefore the photoresist, or imaging layer, can be very thin. A significant challenge with a thin spin-on silicon hardmask is its interaction with the underlayer, which strongly affects lithographic performance. The UV distribution can be controlled through stack thicknesses, and n and k values specifically, using different compositions of silicon hardmask materials and different resist thicknesses. As shown in Fig. 5b, the simulation tool predicted the improved profile and wider process window with the higher %R and FE as compared to the condition of minimum reflectance alone.


Figure 5: Trilayer a) simulation and process, b) reflectivity and FE simulations.

Conclusion

To more accurately model complex material stacks for hyper-NA exposure tools, a simulation tool was developed to optimize reflectivity and FE parameters to increase the process window and optimize resist profiles. Creative integration of materials and process development will reduce feature sizes and experimental time and costs, and increase process margins.

Acknowledgments

The authors thank R&D personnel of Brewer Science, Inc., Rolla, MO; Resist Test Center (RTC) at International SEMATECH, Albany, New York; IMEC at Leuven, BE; and JSR Microelectronics for providing material for this article. OptiStack is a trademark and ARC is a registered trademark of Brewer Science, Inc.


Jan Macie received her PhD from the U. of Texas at Dallas in 1996 and is a technical product manager at Brewer Science, 2401 Brewer Dr., Rolla, MO 65401 USA; 573-364-0300; [email protected].

Dominic Miranda received his BS from Iowa Wesleyan College and is a product manager at Brewer Science.

Zhimin Zhu received his PhD from U. Catholique de Louvain, Belgium, and is a scientist at Brewer Science.

Brian Smith received his PhD in electrical engineering from the U. of Texas at Dallas in 1997 and is currently the R&D process development manager at Brewer Science.

POST A COMMENT

Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.