AVS 70 Session PS-ThP: Plasma Science and Technology Poster Session

In this work, the uniformity of photoresist (PR) etch rates was monitored with multi-channel optical emission spectroscopy (OES). Etch rate uniformities were controlled from center-high etch rate to edge-high etch rate by varying gas flow rates, source power, bias power, and CF4/O2/Ar plasma pressure in a 300mm cup-type inductively coupled plasma (ICP) descum chamber. Eight fiber optics in an 8-way jig were mounted on a rectangular viewport of the chamber and the eight optical fibers were connected to a switching module that transfers signals to a single spectrometer. We can exclude variations of signals originating from different spectrometers with this configuration. The spatially resolved endpoints were determined with CO (428-431nm) band in all PR etch processes, while other relatively strong peaks such as O (777nm), F (703nm), and CO (481-483nm) show no difference in spatial variation in PR etch rates. The CO (428-431nm) band corresponds to the de-excitation of CO from the state d³Δ with a vibrational number of 14 or 16 to the metastable state a³π. The results indicate that to monitor etch uniformity effectively, it is necessary to observe emissions from select excited states of etch products, which only transition to a metastable state of those products. The etch profiles across the wafer were determined based on the etch endpoints from 8-channel optical emission spectroscopy (OES) under various conditions, and these results aligned with the trends observed in thickness measurements conducted using a reflectometer.

Plasma nitriding is widely used for mechanical parts, dies and tools as one of the surface modification processes. Plasma nitrided material is improved the surface hardness, wear resistance and fatigue strength etc. while maintaining the core properties. An advantage of plasma nitriding over conventional nitriding is that the former is a clean, nontoxic process that involves a shorter nitriding time than gas nitriding. However, plasma nitrided sample have not be attained low friction coefficient same as samples nitrided by another nitriding method. A mechanical part that contacts two surface requires a reduction in the friction coefficient. Friction accounts for about 23% of all energy consumption worldwide, among which 20% is needed to reduce friction and 3% is required to remanufacture replacements for worn-out parts and equipment. Consequently, studying the characteristics of contact surfaces that reduce friction not only results in energy savings, but also extends the lifetime and dependability of the component. Surface texture technology is crucial for reducing the friction coefficient of contact surface and enhancing friction performance.

There is technology of surface texturing that create a series of regular microstructures on the surface of a sample using process techniques. The textures with suitable size can act as micro-bearing to increase the dynamic pressure between friction pairs, store lubricants, and capture debris produced during the friction process.

In this study, plasma nitriding was performed partially by Electron-Beam Excited Plasma using mask with hole to create surface texturing on the surface of tool steel. The effects of partially plasma nitriding were investigated, and the tribological properties of surface on the formed partially nitrided layer were clarified.

The results of our experiments show that the hardness of the sample was increased only in the open areas that were not masked. The measured hardness of the nitrided layer of all tool steel work pieces were increased by more than two times that of the core material. Regular micro asperities were formed on the nitrided sample surface.

A diverse combination of feed gases is utilized in creating low-temperature plasmas for applications in the microelectronics industry. These plasmas generate a wide combination of reactive and non-reactive species, with a spatial and temporal fluctuation in the density, the temperature, and the energy. Precise understanding of these parameters and their variations is crucial the advancement of microelectronics through validated modeling and designing relevant devices. Laser-induced fluorescence (LIF) offers spatially and temporally resolved information of the plasma-produced radicals, ions, and metastables. However, implementing this diagnostic tool requires the knowledge of the optical transitions, including the excitation and the fluorescence wavelengths. This information is often scattered across extensive literature from widely different fields. This study analyzes and compiles the available transitions for laser-induced fluorescence of more than 160 chemical species pertinent to the microelectronics industry. The plasma generated by the feed gas mixture and its interaction with the chamber walls and materials generates a complex combination of reactive species. Our analysis will show the overlapping LIF transitions that need to be considered when selecting and implementing LIF in plasmas with a complex combination of feed gases, such as those employed in the processing of microelectronics.

Acknowledgement: This work was supported by the U.S. Department of Energy through contract DE-AC02-09CH11466.

As extreme ultraviolet lithography progresses towards achieving finer details, challenges like stochastics and sensitivity emerge for polymeric photoresists. To meet the demands of high-volume manufacturing and overcome the limitations of EUV source power, enhancing resist sensitivity is crucial for achieving high-resolution patterning while ensuring pattern fidelity and uniformity. One promising approach involves incorporating atoms with high photoemission cross-sections, such as iodine, into the resist composition to significantly enhance material absorbance. Exploring the incorporation of iodine into carbon-containing layers presents an intriguing avenue for advancement.

In recent years, Air Liquide has spearheaded the development of a series of low global warming potential chemistries, facilitating the formation of atomically smooth iodine-containing layers with adjustable thickness and iodine concentration through a low-pressure plasma process. These film properties are studied using scanning electron microscopy (SEM), atomic force microscopy (AFM), and X-ray photoelectron spectroscopy (XPS).

The state-of-the-art EDIPIC-2D code [1] now includes self-induced electromagnetic effects using the Darwin Direct Implicit algorithm [2]. The cell size can be greater than the Debye length and the time step may exceed the period of Langmuir oscillations, which are acute constraints to the standard explicit Particle-In-Cell code [3]. Therefore, the code can now model Inductively Coupled Plasma (ICP) Reactors of a realistic size using a modest amount of computational resources. In this work, we showcase the example of an ICP reactor operating in the sub-mTorr range, that could be typically used for etching in the semiconductor industry. We emphasize our analysis of key parameters essential for atomic precision processing such as the ion energy flux and the ion angle distribution function as measured at the wafer.

[1] Sydorenko D., "Particle-in-cell simulations of electron dynamics in low pressure discharges with magnetic fields," Ph.D. dissertation, University of Saskatchewan, Canada, 2006.

[2] Gibbons M. R. and D. W. Hewett, “The Darwin Direct Implicit Particle-in-Cell (DADIPIC) Method for Simulation of Low Frequency Plasma Phenomena”, J. Comput. Phys. 120, 231 (1995).

[3] C.K. Birdsall, “Particle-in-cell charged-particle simulations, plus Monte Carlo collisions with neutral atoms, PIC-MCC,” IEEE Trans. Plasma Sci. 19(2), 65–85 (1991).

Metal-semiconductor-metal (MSM) photodetectors (PDs) have gained significant interest in optoelectronic applications in industrial, environmental, and even biological fields. MSM PDs benefit from simplicity in design, large active area, fast response, and low dark current. PDs fabricated from materials such as GaN, NiO, AlGaN, ZnO, Nb₂O₅, and TiO₂ have attracted attention from researchers owing to their wide energy band gaps. Among these materials, both GaN and NiO semiconductors stand out as promising candidates for ultraviolet/visible photonic devices due to their wide bandgap energies.

Herein, we report on MSM PDs based on GaN and NiO films which were deposited on sapphire and glass substrates respectively, using hollow-cathode plasma-assisted atomic layer deposition (HCP-ALD). GaN was deposited at 200℃ using triethylgallium (TEG) as the metalorganic precursor and N₂/H₂ plasma as the co-reactant. NiO was deposited at 100℃ using nickelocene (NiCp₂) as the metalorganic precursor and O₂ plasma as the co-reactant. The rf-plasma power was maintained at 100W for both films. Aluminum interdigitated electrodes, with widths and spacing of 500 microns, were deposited onto the films using e-beam evaporation.

The deposited films were characterized for their optical and structural properties. Both films deposited showed strong absorption in the UV region (l=190-380 nm) yet demonstrated lower absorption in the visible and near-IR regions. As a result of the analysis using the Tauc relation, the band gaps of GaN and NiO films were found to be 3.32 and 2.95 eV, respectively. XRD analysis revealed a preferred orientation along (002) direction for both films. I-V characteristics of the fabricated MSM PDs were conducted under dark and light conditions. An incandescent light source at a distance of ~10 cm from the sample was applied to generate the photocurrent. Bias-dependent photocurrent signals were observed for both GaN and NiO MSM PD samples. On the other hand, NiO photoresponse showed a significantly slower temporal response indicating a notable persistent photoconductivity.

Nano-Master has developed a hybrid PECVD/PEALD tool capable of depositing ALD and PECVD layers on silicon wafers within the same chamber with no hardware modifications. In this paper, we will propose a novel reactor design capable of performing PEALD and PAALE in the same chamber without breaking the vacuum. Removing native oxides from a Si wafer using other means, such as reactive ion etching or wet etching, results in substrate damage and defects to the wafer. Here, we will discuss the means of removing the native oxide from the Si wafer surface using plasma-assisted atomic layer etching (PAALE) before depositing AlN in the same chamber and without or low substrate damage. The PEALE of native oxide on the silicon wafer will occur in the same PEALD chamber, which prevents re-oxidation between the steps. The objective of the study is to develop a tool that can perform precise atomic layer etching and damage-free atomic layer deposition of high quality AlN in the same chamber.

Introduced in the early 90s to perform deep etching of silicon, plasma cryoetching involves using highly reactive plasmas, such as SF₆- or CF- based plasmas, to etch materials cooled down to temperatures below -100°C with high anisotropy. During cryoetching of silicon in SF6 plasmas, the addition of small amounts of oxygen allows to form a temperature sensitive SiO_xF_y passivation layer on the sidewalls of the trench, which prevents spontaneous lateral etching and desorbs completely when the substrate is returned to room temperature. Cryogenic processes have the advantage of being polymer-free and clean, which prevents process drift and makes them suitable for new applications where smooth sidewalls or reduced plasma-induced damage are required. These features are attractive for etching porous low-K materials in the back-end-of-line of advanced CMOS technology; applications to atomic layer etching of conventional materials (Si, Ge, GaN, InP) or emerging 2D materials (graphene, MoS₂) are also envisaged. Although the understanding and control of plasma cryoetching has improved over the years, the fundamental mechanisms of the formation and desorption of passivation layers are not well understood. And differences between elementary plasma-surface interactions at cryogenic and room temperature remain unclear. In this paper, Molecular Dynamics (MD) simulations are performed to study the interaction between F- and CF- based plasmas with Si and SiO2 materials. The objective is to understand the impact of ion and radical plasma species (nature, dose, energy, etc.) on the structural and chemical modification of exposed materials, both at room and cryogenic temperatures. Quantitative information on surface reaction probabilities (sticking, thermal desorption, surface diffusion, sputtering yields) will be compared and discussed, to better understand the key mechanisms behind cryoetching and cryo-ALE processes.

This work was supported by ANR, under the project name PSICRYO (No. ANR-20-CE24-0014).

In recent semiconductor etching processes, pulse-driven capacitively coupled plasma (CCP) is widely used to achieve high etching selectivity. As technology advances, the equipment structures used in these processes become increasingly complex, requiring various conditions. Consequently, diverse issues are emerging within the CCP equipment, and understanding the various plasma phenomena occurring inside the process equipment is essential for solving these problems. To investigate the nonlinear and transient particle dynamics of CCPs, particle-in-cell (PIC) simulations are required [1,2]. This study uses a parallelized PIC simulation called K-PIC to investigate the onset of arcs in the gap between the wafer edge and the focus ring, which occurs in actual CCP equipment. We report the electron heating concentrated in the gap [3] through spatiotemporal data analysis, examining the effects of driving waveforms and the gap length [4] between the wafer edge and the focus ring.

References

[1] J.S. Kim, M.Y. Hur, C.H. Kim, H.J. Kim, and H.J. Lee,J.Phys. D: Appl. Phys. 51, 104004 (2018).

[2] M.Y. Hur, J.S. Kim, I.C. Song, J.P. Verboncoeur, and H.J.Lee, Plasma Res. Express 1, 015016 (2019).

[3] T. Lafleur, P. Chabert, J. P. Booth, Plasma Sources Sci. Technol, 23, 035010 (2014).

[4] J.S. Kim, M.Y. Hur, H.J. Kim and H.J. Lee, J.Appl.Phys. 126, 233301 (2019).

This study examines the role of carbon monoxide (CO) as an additive gas in the etching of silicon oxide (SiO) and silicon nitride (SiN_x) using Ar, C₄F₆, and O₂ plasmas. SiO and SiN_x are commonly used in the semiconductor industry, as device designs shift from 2D to 3D structures. The etching gases like CF₄, C₄F₈, and C₄F₆, are commonly used for dielectric etch and additive gas like hydrogen (H₂) is used to control the SiO and SiN_x etch rates and selectivity. However, H₂ is typically used without O₂ condition, whereas CO can be used with O2. This research investigates CO’s impact on SiO and SiN_x etch rates and selectivity under different O₂ included process conditions. The SiO and SiN_x surfaces are analyzed using X-ray photoelectron spectroscopy (XPS).

Initial experiments kept the total flow rates of C₄F₆, O₂, and CO constant while increasing the CO ratio. As the CO ratio increased up to 60%, the SiO etch rate rose from 390 Å/min to 1426 Å/min but dropped sharply beyond 80% CO raio. Conversely, the SiN_x etch rate increased steadily from 82 Å/min at 0% CO to 289 Å/min at 100% CO. The selectivity of SiO over SiN_x etching improved from 4.8 to 11.9 with up to 40% CO ratio but declined to 1.4 beyond 60% CO. The XPS analysis showed that the C1s ratio on SiO surfaces decreased until 40% CO, disappearing entirely after 60% CO, correlating with the sharp decline in SiO etch rate. For SiN_x, higher CO ratios reduced the deposition of the FC layer, gradually increasing the etch rate. These findings suggest that increasing CO decreases the partial pressure of C₄F₆, which provides CF_x radicals, thereby reducing FC polymer thickness and affecting etch rates and selectivity.

In another experiment, CO was compared with O₂ as an oxygen source. The SiO etch rate increased from 987 Å/min to 1776 Å/min up to 60% CO, then dropped sharply to 172 Å/min beyond 80%, with deposition occurring at 100% CO. The SiN_x etch rates decreased gradually from 1219 to 0 Å/min as CO increased. The SiO over SiN_x selectivity rose from 0.81 to 3.7 up to 60% CO. XPS analysis indicated that up to 60% CO, SiO was primarily etched, but beyond 80% CO, FC polymer layers formed, inhibiting etching. For SiN_x, the increasing CO ratio reduced Si2p and N1s peaks, suggesting SiN was almost entirely covered by an FC polymer layer, consistent with the gradual decrease in the etch rate. These experiments suggest that the CO serves as more carbon source rather than O radical Source.

Photoresist ashing is a critical and increasingly common step in semiconductor manufacturing. Conventional methods for ashing are either wasteful, cause undesired damage, or lack the high throughput desired by industry. In this study, a large-area, atmospheric-pressure plasma is studied for ashing. The plasma is powered by radio frequency power (13.56 MHz) and interacts with the substrate through the afterglow which limits bombardment by energetic species. Etch rates of 205 nm/min are achieved, with little damage to the underlying substrate. The process is demonstrated up to 4-inch wafers and on patterned structures.

Plasma processes are extensively used in semiconductor device fabrication, but as devices become more highly integrated and critical dimensions shrink to the nanometer and atomic scale, maintaining high accuracy and stability becomes increasingly challenging. In this work, fault detection model was developed for plasma processing using non-invasive optical emission spectroscopy (OES) and recurrent neural network (RNN) with autoencoder. 13.56 MHz RF source power and 12.56 MHz bias power was applied to an inductively coupled plasma (ICP) with CF₄, Ar, and O₂ gases. To train the model, experimental data were gathered by varying process conditions, including RF source power and CF₄, Ar, O₂ flow rates, as well as pressure, from -50% to +50% of the normal process values. OES signals from the ICP reactor were collected with 3648 channels, and collected optical signals were trained by autoencoder consisted with linear unit, batch normalization and hyperbolic tangent function. Then, 16 hidden features were extracted from 3648 intensities in each time step. For the predicting the hidden features in next time step, the 16 hidden features at the current step and 16 values from process variables to train the RNN: RF powers (4), RF matching variables (4), gas flowrates (3), pressures (3), valve position (1), and temperature (1). Faults were determined when the means squared error between hidden features predicted in post time step and the hidden features extracted in present time step. The model was tested under scenarios involving incorrect input of process conditions. An accuracy of 78.2% was demonstrated for RF power errors, 76.3% for bias power errors, and 63.4% for pressure errors.