ALD2019 Session ALE1-MoA: Energy-enhanced ALE

[1] Lam Research Corporation, September 2016. [Online]: https://investor.lamresearch.com/news-releases/news-release-details/lam-research-introduces-dielectric-atomic-layer-etching. >

[2] K. Kanarik, S. Tan, and R. A. Gottscho, J. Phys. Chem. Lett. 2018, 9, pp. 4814−4821>

[3] Michael Koltonski, Wenbing Yang, Craig Huffman, Mohand Brouri and Samantha Tan, “Opportunities and Challenges Utilizing Atomic Layer Etch for Lead Edge Technology Metal Line Widths”, SPIE Advanced Lithography Conference, San Jose, CA, Feb 24-28, 2019>

Our research group developed a process for atomic layer etching of an SiO₂ film using alternating nanometer-thick fluorocarbon film deposition and O₂ plasma irradiation [1]. This process allows the atomic scale etching of SiO­₂ with high reproducibility because of removing extra carbon on surface and cleaning chamber walls by O₂ plasma. The ALE process could have benefits for etching SiO₂ selective to Si₃N₄ if we actively control the chemistry in the mixture region between Si-compounds and fluorocarbon, and suppress the oxidation of Si₃N₄ by O₂ plasma.

In this time, the ALE process was performed to a SiO₂ and Si₃N₄ deposited by ALD process in a capacitively coupled plasma (CCP) reactor. A 100-MHz electrical power of 100 W was applied to the upper electrode at a pressures of 2.0 Pa. The wafer temperature was set at 20°C. For the deposition process, C₄F₈/Ar plasma was used to form a fluorocarbon film. Figure 1 shows the C 1s spectra of a SiO₂ and Si₃N₄ after the deposition processes. The C 1s spectrum of SiO­₂ after the deposition process exhibits C-C, C-CF_x, CF, CF₂, and CF₃ peaks. On the other hands, the spectrum of Si₃N₄ shows an increased in the fraction of the C-C bond and exhibits the Si-C bond. Higher fraction of the C-C and forming Si-C and C-N bonds in the mixture regions lead to etching SiO­₂ selective to Si­₃N₄. Higher F/C ratio in fluorocarbon film is required for etching SiC compared to SiO₂ and Si₃N₄ because of etching products of SiF₄ with very little SiF and SiF₂[2]. Moreover, higher bond energies of CN, which are C-N of 305 kJ/mol, C=N of 615 kJ/mol and C≡N of 887 kJ/mol, could suppress oxidation of SiN by control of ion energy and wafer temperature. If the oxidation by O₂ plasma is suppressed, our ALE by alternating fluorocarbon deposition and O₂ plasma could apply to industry process for next-generation devices due to high controllability and reproducibility. We analyze the depth profiles of atomic concentrations in the mixture regions for a SiO₂ and Si₃N₄ by angle resolved X-ray photoelectron spectroscopy (XPS) to control chemistry of the mixture regions by some knobs such as ion energy and surface temperature.

[1] T. Tsutsumi et al., J. Vac. Sci. Technol. A 35, 01A103 (2017)

[2] H. Winters et al., J. Vac. Sci. Technol. B 1, 927 (1983)

Single digit nanometer semiconductor manufacturing is increasingly demanding atomic scale process controllability to further decrease critical dimensions and pitches. High etching precision and material selectivity become essential in the atomic scale era. Plasma based atomic layer etching (ALE) shows promise to attain atomic etch precision, enhancing energy control and reaction chemistry control.

Here we study a Fluorocarbon(FC)-based ALE process for controlling the etching of silicon dioxide at the atomic level. Figure 1 shows the schematic of atomic layer etching process using Ar plasma and CHF₃ gas. In this technique, an Ar plasma is maintained continuously through the process, below the energy threshold for SiO₂ sputtering. A fluorocarbon chemistry is then introduced via CHF₃ pulsing to provide the reactant absorption. Subsequently, once the gas pulse has concluded, bias power is introduced to the Ar plasma, to provide enough energy to initiate reaction of the FC with the SiO₂. In ideal ALE, each of the steps is fully self-limiting for over exposure to increase uniformity on the microscale (wafer) and atomic scale.

With the goal of achieving self-limiting FC-based ALE, we investigated the etch step using low energy Ar ion bombardment. By carefully tailoring the energy of ion bombardment, it is possible to control the etching depth to approach a self-limiting behavior. The impact of various process parameters on the etch performance is established. We demonstrated that the SiO₂ amount etched per cycle (EPC) is strongly affected by the forward bias plasma power, as well as the substrate temperature (Figure 1(a)). The substrate temperature has been shown to play an especially significant role, at -10 °C the contributions to chemical etching coming from fluorine and fluorocarbon compounds from chamber walls are minimized and a quasi-self-limiting behavior ALE is observed.

Figure 1(b)-(f) showed the Cr features after being etched for 60 ALE cycles with the optimal ALE self-limiting conditions. Feature trenches vary from 20-200 nm and were defined using metal lift-off. Overall, using the cyclic CHF₃/Ar ALE at -10 °C, we reduced geometric loading effects during etching and reached aspect ratio independent etching, with great potential for significant improvement in future etching performances.

The impact of continuous-wave plasmas in realizing a pattern transfer process with a Ar/fluorocarbon composition on photoresist etching behavior and surface roughness development has been extensively studied.¹ However, the characteristics of photoresists under atomic layer etching (ALE) processes have not been well established. Specifically, the structure and morphology of the photoresist layer is dependent on the interplay between energetic ion bombardment and the diffusion of reactive species at the surface. For evaluating these photoresist properties, we used an ALE process with an Ar carrier gas and a fluorocarbon (FC) precursor gas, for example C₄F_8.²

For sample characterization, we utilized a combination of real-time, in situ ellipsometry and post-process surface roughness and surface chemistry analysis using atomic force microscopy (AFM) and x-ray photoelectron spectroscopy, respectively. The AFM characterization provided information on both the surface roughness magnitude as well as the distribution via a power spectral density analysis. Both an industry-standard 193 nm photoresist and an extreme ultraviolet (EUV) photoresist were evaluated.

Based on the ellipsometric characterization, we find that the 193 nm photoresist initially develops a surface dense amorphous carbon (DAC) layer from Ar ion bombardment of both the native photoresist and deposited FC, which contributes additional carbon to the DAC layer, increasing its thickness. Upon the FC deposition step in the ALE process, the refractive index of the DAC layer decreases due to fluorine diffusing into the layer structure. Corresponding AFM analysis shows a reduction in the surface roughness. Once the DAC layer becomes saturated with fluorine, a discrete FC layer forms on the surface. Subsequently, in the etching step, the discrete FC layer is removed, the DAC layer recovers its thickness, and the cycle repeats.

The authors gratefully acknowledge Eike Beyer for assistance with the experimental setup, R. L. Bruce, S. Engelmann, and E. A. Joseph for supplying the EUV materials, and financial support of this work by the National Science Foundation (NSF CMMI-1449309) and Semiconductor Research Corporation (2017-NM-2726).

¹ S. Engelmann et al., J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 27, 1165 (2009).

² D. Metzler et al., J. Vac. Sci. Technol. A 32, 020603 (2014).

Semiconductor industry has followed Moore’s law through years. Nowadays, industry and researchers are trying to reach nanoscale dimensions to continue the trend “beyond Moore’s Law”.

In this framework appeared atomic layer etching (ALE) where self-limited etching is performed, for instance, on SiO₂, using C₄F₈/Ar plasma [1]. In this process, CF_x monolayers are deposited on the surface before etching.

In order to perform ALE cycles without reactor wall contamination by CF_x species, ALE process is proposed at low substrate temperature. In our experimental protocol, fluorocarbon gas such as C₄F₈, is physisorbed at the sample surface in a first step. Then, in a second step, etching is performed by Ar plasma.

The experiments were carried out in an inductively coupled plasma reactor. Tests were performed on three different samples (SiO₂, Si₃N₄, and a-Si) that were glued on a carrier wafer. Then, they were cooled down to very low temperature by liquid nitrogen. In-situ spectroscopic ellipsometry was used to follow the layer thickness evolution of the central sample. The other samples were characterized ex-situ after the experiment. Surface roughness evolution before and after etching was checked by performing atomic force microscopy (AFM).

Finally, Langmuir probe and Quadrupole Mass Spectrometry (QMS) were used to better understand the involved mechanisms.

The aim of the first tests was to prove that etching occurs only at low temperature when using fluorocarbon gas flow. For that, the same etching test was performed at -120°C and few degrees above. At -110°C, etching is very limited, which shows that physisorption is not as significant as at -120°C. However, at ‑120°C, a few monolayers of C₄F₈ can be physisorbed and etching with Ar plasma has been observed thereafter.

Then, the main goals of this research was to reach self-limited etching regime and get an identical etching rate through cycles. To this end, the influence of various parameters of the process were examined. Different step times and pressures were studied to understand physisorption mechanisms. Temperature effect on species residence time was also evaluated using in-situ ellipsometry and QMS. In order to limit physical sputtering by ion bombardment and reach self-limited etching, different plasma parameters , such as self-bias, ICP power and step time, were studied. The etch per cycle is typically between 0.2 and 0.5 nm depending on the process parameters. The different steps of deposition, purge and etching can be clearly identified by ellipsometry.

Acknowledgments

The authors gratefully thank S.Tahara for his helpful discussions.

1. Metzler et al., J. Vac. Sci. Technol. A 32(2), 020603-1 2014