AVS2014 Session PS-TuA: Advanced BEOL/Interconnect Etching

To meet the needs of the device scaling trend, patterning technologies for critical dimension control less than 20 nm is required. For 1X nm pattern formation beyond the conventional optical lithography limit, it is necessary to use double (or multiple) patterning process which increases the process cost. Directed-self assembly (DSA) of block copolymer is one of the most attractive candidates for 1X nm pattern formation process and 12.5 nm hp patterns were formed using polystyrene-block-poly(methyl methacrylate) (PS-b-PMMA) [1]. DSA lithography (DSAL) process using PS-b-PMMA needs selective removal of PMMA to PS, which is called “development process”. A wet development process was applied successfully for contact hole shrink process [2]. Although this method can remove PMMA selectively, pattern collapse will occur for line and space pattern resulting from surface tension of developer solvent.

On the other hand, although a dry development process is expected to solve this problem, selective removal of PMMA is difficult because PMMA is a similar organic polymer to PS. Dry development processes using Ar plasma and O₂ plasma have been reported and their selectivities were 3.9 and 1.7, respectively [3]. However in our case, higher selectivity is needed for etching the underlayer with PS as mask.

In this study, we focused on differences of material composition of PS and PMMA. Our results concluded that the control of ion energy and design of gas chemistry were key factors for the selective etch.

The aromatic group in PS is more durable than the carbonyl group in PMMA for ion bombardment. A selectivity of around 8 was achieved by the control of ion energy in xenon plasma.

PMMA has more oxygen in the film than in PS, so we designed the gas chemistry to realize the selective PMMA etch by using this difference of the oxygen content. We studied carbon containing gas plasma, because carbon radical will deposit on PS which does not contain oxygen. On the other hand, carbon radical will react with the oxygen in the PMMA to make volatile CO_x, therefore selective PMMA etch to PS can be realized. A selectivity of over 20 was realized using CO gas plasma [4].

DSA lithography dry development was successfully realized by controlling ion energy and designing the etching gas chemistry based on the difference of material composition of PS and PMMA.

References

[1] C. Bencher et al., Proc. SPIE 7970, 79700F (2011)

[2] Y. Seino et al., Proc. SPIE 8323, 83230Y (2012)

[3] M. Satake et al., Proc. SPIE 8685, 86850T (2013)

[4] H. Yamamoto et al., Japanese Journal of Applied Physics 53, 03DD03 (2014)

The reduction of device dimension at the sub-15nm technological node requires the use of double patterning for contact etching. Line and space patterns are defined first in a thin TiN layer. Then, a trilayer stack with Si-containing anti reflection coating (SiARC) and organic planarizing layer (OPL) is used to define open areas. The mask is defined by the intersections of both hard mask of TiN and OPL patterns and is used to etch contacts into silicon oxide (TEOS).

The OPL mask must conserve straight profiles during the different etching steps and TiN is exposed to the plasma during silicon oxide etching. The OPL can be etched by different plasmas (N₂/H_2, O₂/SO_2, O₂/CO₂) that may induce a passivation layer on the sidewalls via different passivation elements such has CN, CS, CO. Such passivation layers, as well as the presence of TiN during contact etching, can interfere with the SiO₂ etching process and change the final pattern profile. In this study, we investigated the OPL mask etching in COS/O₂ plasma in comparison with N₂/H₂ plasma.

The XPS analyses are performed into Theta300 angle resolved XPS system from Thermo Scientific, ellipsometry measurements are done with spectra FX 200th multiwavelength ellipsometer from KLA-Tencor and OES spectra are recorded with a SD1024 spectrograph detector from SpectraView.

Electronic microscopy observations of OPL patterns etched in N₂/H₂ or in COS/O₂ with various COS/O₂ ratios show that straight profiles without undercut can be obtained. After SiO₂ etching using a fluorocarbon-based plasma, we observe strong profiles variations in the SiO₂ depending on the OPL etch process. After the COS/O₂ OPL open, during the first seconds of the SiO₂ etching, strong emission lines originating from CS species are observed by OES. In addition, EDX analyses after COS/O₂ OPL etching reveal a large amount of sulfur on OPL sidewalls and TiN surface. TiN mask profile is also degraded during the over etch of the OPL and Ti residues are redeposited on all the surfaces. To precise the interaction mechanisms, XPS analyses are performed on TiN, OPL and TEOS after exposure to the various OPL etching processes. We evidenced that the contact profile is influenced by both the process used during OPL opening and the presence of TiN on the wafer. Degradation of masks profiles leads to Ti or S containing residues formation which tends to block the SiO₂ etching. These effects can be reduced by an increase of COS/O₂ ratio during OPL etching.

Sub-22nm logic technology requires contact level etch to meet aggressive critical dimension (CD) shrinks as ArF immersion pattering has mostly reached its resolution limit. Utilization of elliptical contacts brings new constraints to CD shrink. Controlling the 2-D aspect ratio of oval contacts is critical to both device performances and yield. One challenge is that conventional plasma etch shrink methods can induce more shrinkage in the major (Y axis) direction than the minor (X axis), which can cause line-end shortening and feature tip-to-tip spacing control problems.

This paper presents a unique dry etch process that yields a Y to X shrink ratio range ≤1 concurrent with a 50% CD reduction from lithography. By utilizing a direct current superposition (DCS) technology, along with CxHyFz chemistry to cure a negative tone developed photoresist (NTD), this method creates a controllable in-situ hydrocarbon deposition, which is mainly responsible for the Y to X shrink ratio range ≤1. This is not seen with conventional fluorocarbon etch based shrink where Y/X shrink ratio is typically >1, as a result of the larger collection angle for gas phase deposition along the major axis. The Y/X shrink ratio range can be modulated through process condition such as gas ratio, pressure, time, etc. After the controllable hydrocarbon deposition, multiple mask defining transferring steps can be executed anisotropically to complete the pattern transfer. Reported is the structural characterization pre and post etch detailing shrink ratio control. In addition, a mechanistic model will be proposed based on optical emission spectroscopy (OES), thin film compositional analysis, and mass spectrum data.

One of the key parameters in semiconductor mass production control is Line Edge Roughness (LER) / Line width roughness (LWR) owing to its direct contribution to gate length variation, edge placement error, line resistance variation and others. Due to the resolution-line edge/width roughness-sensitivity (RLS) trade-off for photoresists (PR), photolithography has reached its limit to further improve PR LER/LWR for advanced technology nodes (1xnm and beyond). Thus post lithography roughness reduction treatments have become critical in meeting the ITRS targets for LER/LWR. Vacuum Ultra Violet (VUV) treatment has been showing promising results with photochemical modification based smoothening of PR surface using gases like Ar, H₂, HBr etc. which can produce VUV radiation. While several studies have been published over the past decade for improving PR LER/LWR using VUV, most do so by analyzing the results post-VUV-treatment at resist level (before etch transfer). To get good LER for final pattern, pattern transferring from PR is also critical.

CCP chambers, by design, have advantages to achieve good LER/LWR due to the relatively low plasma density and high deposition/etch radical ratio. CCPs with a wide gap are able to well decouple top and bottom RF powers so that we have either low plasma density with high ion energy incident onto the wafer surface or vice versa. In this presentation, post etch LER/LWR data will be discussed with various approaches, such as different treatment duration, different gas combination, ratio plus DC superposition, PR margin, and different plasma parameter settings etc. We will demonstrate that in-order to meet the ITRS targets for LER/LWR for 1x nm and beyond requires co-optimization of the resist roughness with resist profile; thickness; and the subsequent pattern transfer process onto underlying stack.