AVS2009 Session PS1-ThA: Fundamentals of Plasma-Surface Interactions II

Structures in the chip industry are approaching the 32 nm half pitch, which requires radiation in the VUV and EUV range. Cracking of hydrocarbon impurities in the vacuum by the radiation causes C growth on the VUV and EUV optics. Each nm of deposited carbon reduces the reflectivity of the optics by 1%. Fast removal of these contamination layers without damage to the underlying optics is essential for the next generation of lithography devices.

Etching with a low energetic plasma jet can be used to selectively remove coatings such as hydrogenated amorphous carbon (a-C:H) without damage to the underlying structure. Real time, in situ spectroscopic ellipsometry measurements indicate that the highest etch rates are obtained for an Ar/H₂ plasma, rather than for a pure Ar or H₂ plasma.

Even though the etch rate of a-C:H thin films is dependent on both temperature and roughness, the highest roughness in absolute values is attained by the plasma with the lowest etch rate. At low temperatures, the etch rate deviates from an Arrhenius relation, while the activation energy is similar for both the H₂ and Ar/H₂ plasma at higher temperatures.

The two orders of magnitude higher etch rate for the Ar/H₂ plasma is due to chemical sputtering, which is a synergistic effect between atomic H and Ar⁺ ions with an ion energy below the threshold of 58 eV for physical sputtering. Chemical sputtering has been observed by Hopf et al. for energies above 20 eV and an H to Ar⁺ flux ratio over 100 [1]. In our plasma, however, the Ar⁺ ion energy is only a few eV’s and the estimated H to Ar⁺ ratio is lower than 5.

The etch products, released from the surface, consist mainly of CH₄ and C₂H_y, as shown by residual gas analysis. Time resolved optical emission spectra of the Ar/H₂ plasma, from a few mm’s in front of an a-C:H sample, indicates also the presence of C₂ and CH radicals. The CH radical is formed in the plasma phase through charge transfer between Ar⁺ ions and these larger hydrocarbons, and dissociative recombination. Similar plasma chemical processes occur during the remote plasma deposition of a-C:H films. However, in contrast with deposition, the CH rotational temperature shows an overpopulation in the higher excited states, indicating that the (internal state of the) parent molecule is different for an etch plasma than for a deposition plasma.

Spatially resolved optical emission measurements are Abel inverted, by means of the numerical Barr method. While there is some CH production throughout the entire plasma jet, the highest CH production occurs in front of the a-C:H sample.

Films deposited from fluorocarbon (FC) plasmas exhibit low dielectric constants desirable for interlayers in ultra-large scale integrated circuits (ULSIs). The processing of ULSIs has involved the use of small monomer (CF₄, C₂F₆) FC precursors as an avenue for plasma-enhanced chemical vapor deposition (PECVD). To gain a broader understanding of both the FC plasma system and plasma-surface interactions, we have explored gas phase diagnostics and species-surface reactivity under varying plasma parameters. This presentation will reflect upon data obtained from optical emission spectroscopy (OES) concerning the role of excited state species present in FC plasmas. OES data show that during FC plasma treatment of Si and ZrO₂ wafers, CF₂* concentrations increase independent of feed gas and substrate type. The films deposited from such treatments do, indeed, consist of FC moieties and thus plasma-surface interactions are clearly influential in the overall process. We have studied the interaction of FC plasma species at the interface of depositing films using the imaging of radicals interacting with surfaces (IRIS) technique. IRIS data show that scatter probabilities for the CF₂ radical are greater than unity, indicating that CF₂ is produced from films at the surface during FC plasma processing of silicon. Furthermore, we have used quadrupole mass spectrometry to investigate mean ion energies of CF₂⁺ in FC plasmas and have discovered that ion energies increase with increasing applied rf power. We have previously demonstrated that IRIS scatter coefficients for CF₂ produced from larger precursors (C₃F₈ and C₄F₈) correlate directly with ion energy. Thus, we will explore the role of this radical during processing of Si and ZrO₂ with small FC precursors as monitored by IRIS studies and compare these results with the respective ion energies for CF₂⁺ in these systems.

Plasma and surface diagnostics of Cl₂/O₂ mixed-gas inductively coupled plasmas are reported. Using trace rare gas optical emission spectroscopy (TRG-OES) and Langmuir probe analysis, electron temperatures (T_e) and number densities for Cl atoms (n_Cl), electrons (n_e), and positive ions were measured as a function of percent O₂ in the feed gas and position in the plasma chamber. Adsorbates on, and products desorbing from a rotating anodized aluminum substrate exposed to the plasma were detected with an Auger electron spectrometer and a quadrupole mass spectrometer. T_e and n_e increased with increasing percent O₂ in the plasma, while n_Cl fell off with O₂ addition in a manner reflecting simple dilution. Cl atom recombination probabilities (γ_Cl) were measured and were found to be a nearly constant 0.036±0.007 over the range of Cl₂/O₂ mixing ratios and Cl coverage. Large yields of ClO and ClO₂ were found to desorb from the surface during exposure to the plasma, ascribed predominantly to Langmuir-Hinshelwood reactions between adsorbed O and Cl. In addition, the transient surface composition of an anodized aluminum surface was determined as the gas was switched from Cl₂ to O₂ and vice versa. When the surface was first conditioned in an O₂ plasma and then exposed to Cl₂ plasmas, a rapid uptake of Cl was found in the first tens of seconds, followed by a slow approach to a steady state value within ~5 minutes of plasma exposure. Conversely, when the surface was exposed to a Cl₂ plasma for a long time and then switched to an O₂ plasma, the anodized aluminum surface underwent a rapid de-chlorination in the first few seconds and then a slow approach to steady state over ~3 minutes. The buildup and decay of Cl coverage is well described by a stretched exponential function, reflecting a range of binding sites for Cl. Throughout these treatments, the coverages of Si (from erosion of the quartz discharge tube) and O was nearly constant.

For the next technological generations of integrated circuits, the traditional challenges faced by etch plasmas (profile control, selectivity, critical dimensions, uniformity, defects, ...) become more and more difficult, intensified by the use of new materials, the limitations of lithography, and the recent introduction of new device structures and integration schemes. Chemical plasma composition can be changed by modifying the gas mixture, ion flux can be partly controlled by source power, and ion energy can be chosen thanks to the bias voltage applied to the substrate. However, these control parameters are not always sufficient to reach all required etching characteristics and new control parameters are needed. Pulsing the plasma source power or the substrate bias offers new operating parameters (pulse frequency, duty cycle). The main advantages of a pulsed etching process are the improvement of etch selectivity and the reduction of charge-up damages and defects by reducing the electron activity and controlling the dissociation of radicals in the plasma.

Studies are being conducted on the etching characteristics of silicon and silicon dioxide in a 300 mm industrial inductively coupled plasma etching chamber having pulsed plasma discharge capability from Applied Materials. The reactor has been modified to be connected to an Angle-Resolved X Ray Photoelectron spectroscopy analyzer by a robotized vacuum chamber. Hence after an etching process, XPS spectra were recorded as function of take-off angle and the integrated intensities of the core-level peaks were used to obtain chlorine concentration and chemical state information from different depths of the sample, thereby permitting non-destructive characterization of chlorine profile in thin silicon oxide films. Material etch rates were measured in real time by in situ multi-wavelength ellipsometry.

When the plasma is pulsed, two parameters can be adjusted, namely the frequency of the pulse and the duty cycle. While the frequency has only a small influence on the etch rates in the investigated frequency range, our results demonstrate that a low duty cycle clearly modifies etch rate and can considerably improve the etch selectivity between silicon and silicon oxide. When a thin silicon gate oxide layer is exposed to very low energy etching conditions, a first step of chlorine incorporation is observed before etching. Preferential accumulation near the SiO₂/Si interface is observed and chlorine is shown to bond to both silicon and oxygen in multiple distinct chemical states.

An atomic-scale understanding of interactions between chemically reactive plasmas and surfaces is required to establish nanometer-scale processing technologies. Various numerical studies based on molecular dynamics (MD) simulation have been reported so far, but these were limited to simulations of the simple blanket etching to estimate microscopic etching properties [1,2]. Here, we first report a fully atomistic silicon feature profile simulation using classical MD simulations. The potential form can be found in our previous papers [2,3,4]. F, Cl, and Br beams with a translational energy of 100 eV were used as reactive species. The surface area of Si (100) substrates was about 163 × 22 Ų, where 3840 silicon atoms were initially located in the structure of diamond lattice. M ask patterns were introduced in the direction parallel to the short axis with periodic condition, in order to reproduce the trench etching feature. Then, the area without mask was 50 × 22 Ų. By using our new atomistic profile evolution simulation, we investigated halogen plasma-surface interactions at sidewalls and bottom surfaces of nanometer-scale Si trench in detail. It was found that specific feature profiles with different gaseous species appear not only at the sub-micrometer-scale but also at the nanometer-scale etching, and the difference of surface reaction layer formation strongly affects the feature profile evolution during etching. For instance, fluorine beam etching showed that fluoride layer is formed on the entire surfaces containing sidewalls and bottom surfaces, thus giving isotropic etching. Chloride layer was thicker than fluoride and bromide layers, to give feature profiles of sidewall tapering. Bromide layer on bottom surfaces was thinnest among the three, and so the etching rate was lowest. So, it was cleared that the surface reaction layer formation strongly affects the feature profile evolution during etching. Our approach is essential as a reference for macroscopic or empirical profile simulation, where simulation sizes have been reduced recently. We also show some comparison between MD-based profile simulation and our empirical profile simulation (atomic-scale cellular model [5]).

[1] H. Ohta et al., J. Vac. Sci. Technol. A 19, 2373 (2001).

[2] T. Nagaoka et al., J. Appl. Phys. 105, 023302 (2009).

[3] H. Ohta and S. Hamaguchi, J. Chem. Phys. 115, 6679 (2001).

[4] H. Ohta et al., J. Appl. Phys. 48, 020225 (2009).

[5] Y. Osano and K. Ono, J. Vac. Sci. Technol. B 26, 1425 (2008).

The effects of ion bombardment on polymer surfaces can be profound, with implications for all plasma-based pattern transfer processes that involve the use of polymer etch masks in lithography and etching. It is known that Ar⁺ bombardment of various polymers results in the formation of a 1-2 nm deep cross-linked region at the exposed surface, and that virgin polymer sputtering yields can be several orders of magnitude higher than steady state yields after ion bombardment. ^[1]

In this talk, we report results from molecular dynamics (MD) simulations of Ar+ (~ 40 - 150 eV) sputtering of oxygen-containing polymers. The MD data are compared to available experimental results, with special focus on the so-called Ohnishi parameter, which has been shown to correlate with sputtering yields for many O-containing polymers. ^[2] The MD simulations match the published correlations well, and we present a quantitative model of sputtering for these polymers that shows why the Ohnishi parameter (a function of the polymer composition) is proportional to the steady state sputtering yield.

However, we also show that the Ohnishi parameter does not correlate with yields for other polymers, including polyfluoroethylene and polyethylene. The MD simulations show that the validity of this parameterization is dependent on whether or not the sputtering of the polymer transitions between ion-induced scissioning to cross-linking at steady-state.

Finally, we discuss the implications of the dynamics of ion-induced surface cross-linking for synergistic photoresist roughening that occurs in plasmas, especially in the presence of vacuum ultraviolet photon and/or beaming electron exposure.

[1] J. J. Végh, D. Nest, D. B. Graves et al. ” Near-surface modification of polystyrene by Ar+: molecular dynamics simulations and experimental validation” Applied Physics Letter, vol. 91, pp. 233113-1-3, 2007.

[2] H. Gokan, S. Esho, and Y. Ohnishi, “Dry etch resistance of organic materials,” J. EZectrochem. SOC., vol. 130, pp. 143-146, 1983.

Supported by the National Science Foundation under Grant Number DMR-0306582 and the Semiconductor Research Corporation under Contract Number 2008-KJ-1781. The Synchrotron Radiation Center is funded by the National Science Foundation under Grant Number DMR-0537588.