AVS2012 Session PS2-TuA: Plasma Surface Interaction during Plasma Etching

There is great interest in establishing directional etching methods capable of atomic scale resolution during fabrication of highly scaled electronic devices. We report a new concept to achieve controlled, self-limited etching of extremely thin layers of material using a polymer as a special case. The work was performed in a capacitively coupled plasma reactor. The polymer material is a 248nm photoresist. A complete process cycle consists of: O₂ exposure of the polymer material, exhaust of O₂ from the chamber, low energy Ar⁺ ion bombardment of the surface using Ar plasma to remove the oxygen-bonded carbon species, and Ar exhaust. This sequence is repeated up to 20 times to investigate reproducibility of each cycle and time dependent behavior. Controlled etching is based on deposition of a thin reactive layer during Ar⁺ sputtering from a polymer-coated electrode (polyimide-related material) within the chamber. The polyimide-related film deposition balances etching during the Ar⁺ ion bombardment step once the reactive layer has been removed, and enables control of the etching depth. Ar⁺ ion bombardment energies were selected so that once the oxygen-bonded carbon material and physisorbed layer had been removed, net etching ceased. If the ion energy is too small, deposition of sputtered polyimide-related material dominates over etching. When applying a too high selfbias voltage, significant pristine polymer etching takes place and a self-limited process cannot be achieved. Using real-time ellipsometric monitoring, we demonstrate strongly time-dependent etching rates. Starting with a high etch rate of ≈19 nm/min for 2 sec, it decreases within the next 5 sec to 0 nm/min, and therefore shows self-limitation. During the Ar etching step, the O₂ modified deposited reactive layer along with 0.13 nm unmodified polymer can be removed, while concurrent deposition prevents net etching of the unmodified polymer and enables achievement of self-limited etching cycles. This etch is believed to be d irectional enabled by the energetic ion bombardment of the surface. Subsequently, the reactive surface is modified by O₂ adsorption during the O₂ exposure step. Molecular oxygen does not spontaneously react with carbon-based polymers at room temperature, but can be adsorbed on an activated polymer surface to form a very thin layer of oxidized carbon material over unmodified polymer. A thickness increase of 0.6 nm per cycle is observed via real-time ellipsometry. Additional XPS studies allow the investigation of the surface material composition for further insight on the reactive layer deposited during the Ar⁺ etching step and the adsorbed layer during O₂ exposure.

Development of multi-gate field effect transistors (FETs) such as fin-type FETs (finFETs), which can suppress short channel effects, have been considered as a leading approach to continue to follow Moore’s law after the current planer MOSFET technologies reach the size limitations. In finFETs, the Si vertical walls are typically designed to function as gate channels and therefore damages at the vertical walls possibly caused by ion bombardment during the gate etching processes must be minimized. During such etching processes, ions may impinge upon the vertical surfaces but their angles of incidence should be nearly grazing angles. Energetic ions at large oblique angles of incidence against the surface are often considered to be less harmful than those with normal angle of incidence if their incident energies are the same. However, a recent study [1] performed by mass- selected ion beam system has shown that H⁺ ion injection at 60 degree from the surface normal can form a deep damage later near the substrate surface. In addition, it is known that simultaneously injection of H⁺ ions and O atoms cause enhanced surface oxidation on the Si substrate [2]. In this study, we have used molecular dynamics (MD) simulations to study damages caused by ion bombardment with oblique angles of incidence. It has been found in MD simulations that light ions such as H⁺ can indeed cause a deep damaged layer near the substrate surface even at large angle of incidence whereas heavier ions such as Br⁺ ions cause less damages under the same conditions.

[1] T. Ito, K. Karahashi, K. Mizotani, M. Isobe, S.-Y. Kang, M. Honda, and S. Hamaguchi, Jpn. J. Appl. Phys. (2012) in press.

[2] T. Ito, K. Karahashi, M. Fukasawa, T. Tatsumi and S. Hamaguchi, Jpn. J. Appl. Phys. 50 (2011) 08KD02.

Silicon nitride (SiN) is a chemical compound widely used in semiconductor devices or their manufacturing processes as, e.g., gate spacers, stress liners, or hard masks for reactive ion etching (RIE) processes. As to RIE processes of SiN, it has been known that the content of hydrogen (H) in a hydrofluorocarbon (HFC) plasma strongly affects the SiN etching rates. The goal of the present study is to clarify the reaction mechanism of SiN etching. Our earlier studies based on molecular dynamics (MD) simulations and ion beam etching experiments suggested that hydrogen in HFC plasma tends to reduce the thicknesses of a carbon-rich surface layer deposited on SiN during an HFC plasma etching process by forming volatile hydrocarbon species. In the present study, we examine to what extent hydrogen promotes the formation of volatile species containing Si and/or N, which would directly increase sputtering yields of SiN. Our beam experiments have indicated that, under carbon deposition conditions (with 1keV CF⁺ ion incidence), the hydrogen content in a SiN film has little effect on the deposition rate of a carbon rich film on SiN. In other words, hydrogen typically contained in the bulk of SiN is not sufficient (probably in quantity) to promote the formation of volatile species on the surface under such conditions to the extent that the deposition rate could be reduced. However, our MD simulation results have also indicated that, under direct interaction of a SiN substrate with HFC ion beams, Si containing volatile species can be indeed desorbed from the surface. The results suggest that, in the presence of a large amount of hydrogen, hydrogen termination of Si and/or nitrogen bonds in the SiN substrate surface region increases the etching rate of SiN.

Profile anomalies and surface roughness are critical issues to be resolved in plasma etching of nanometer-scale microelectronic devices, which in turn requires a better understanding of the effects of ion incident energy and angle on surface reaction kinetics. In addition, incident neutral radicals also affect the surface reaction kinetics during plasma etching, and thus the etching characteristics are varied by the neutral-to-ion flux ratio. This paper presents a classical molecular dynamics (MD) simulation of Si etching by energetic Cl⁺ and Br⁺ ion beams and low-energy neutral Cl and Br radicals, using an improved Stillinger-Weber interatomic potential model for Si/Cl and Si/Br systems. Emphasis is placed on a systematic understanding of plasma-surface interactions of Si/Cl and Si/Br systems, which are widely used in manufacturing microelectronic devices.

In the MD simulation, the substrate has a Si(100) surface, which is a square 32.58 Å on a side and contains 72 Si atoms in a monolayer (ML). The simulation cell initially contains 1440 Si atoms (20 ML) in a depth of 26.0 Å, where Si atoms in the bottom layer are fixed during simulation, while periodical boundaries are imposed in the horizontal direction. Energetic Cl⁺ and Br⁺ ions are injected toward the substrate surface from randomly selected horizontal locations above the target, and neutral Cl and Br atoms are introduced onto the surface prior to every ion incidence. The neutral-to-ion flux ratio was varied in the range Γ_n/Γ_i = 0–100. The ion incident energy was in the range E_i = 20–300 eV, and the ion incident angle was in the range θ_i = 0–90˚. The kinetic energy of neutral atoms was taken to be 1 eV.

Numerical results indicated that in etching by beam only (Γ_n/Γ_i = 0), the surface reaction kinetics exhibit a characteristic of the ion-enhanced etching at lower ion energies, where the etching yield is maximum at normal incidence (θ_i = 0˚), while a characteristics of the physical sputtering at higher energies, where the yield is maximum at off-normal incidence (θ_i = 60–70˚). Similar inclinations for the ion incident angle were obtained with increasing Γ_n/Γ_i, although the thickness of the surface reaction layer increases a little and the etch yield increases significantly with increasing Γ_n/Γ_i. These imply that impinging ions disarrange the surface Si lattice and weaken the binding force of Si atoms on the top surface, and incident neutral radicals etch surface Si atoms whose binding force with the neighbor Si atoms becomes lower than that of Si atoms in the original diamond lattice. These effects are clearly observed for Si/Br system, as compared to for Si/Cl system.

The importance UV and VUV photon fluxes during plasma materials processing has been recognized through the damage these fluxes may cause in devices being fabricated and in adversely affecting the permittivity of low-k dielectrics such as SiOCH through demethylation. Recently, synergistic effects between ions and photons have been observed in the roughening of photoresist [1] and in sub-threshold etching of silicon in Cl containing gas mixtures.[2] The growing awareness of the importance of UV and VUV photon fluxes in low pressure plasma materials processing, and the possibility of there being synergies with ion fluxes, motivates development of methods to separately optimize UV/VUV and ion fluxes. For example, one may wish to maximize or minimize the overlap in time between the UV/VUV and ion fluxes depending on the particular process. In this talk, we report on results from a computational investigation of low pressure inductively and capacitively coupled plasmas with the goal of determining the degree to which UV/VUV and ion fluxes can be separately controlled. The model used in this investigation is a 2-dimensional plasma hydrodynamics model with radiation transport. Two strategies are being investigated. The first is pulsed plasmas which rely on the different time scales for production and transport of photons and ions to the substrate during the pulsed period to provide for some degree of separate control of the fluxes. The second is semi-remote plasma sources which rely on isolation of the photon sources and transport of ions to control the ratio of photon and ion fluxes to the substrate.

[1] T-Y. Chug, et al., Plasma Proc. Polymer 8, 1068 (2011)

[2] H. Shin et al., J. Vac. Sci. Technol A 30, 021306 (2012)

* Work supported by the Semiconductor Research Corp. and DOE Office of Fusion Energy Sciences