AVS2009 Session PS-TuP: Plasma Science and Technology Poster Session

As one of the candidates for universal memory, spin transfer torque MRAM (STT-MRAM) based on magnetic tunnel junction (MTJ) shows several important features such as nonvolatility in data storage and fast writing speed (2~4ns). The most critical engineering challenge for the fabrication of STT-MRAM is the development of etching technology. The etching of magnetic films has severe problems due to the nonvolatility of the metallic byproducts at practical processing temperature. The chlorides, fluorides, bromides and iodides of transition metals have much higher boiling temperatures compared to those of the typical materials used in semiconductor industry. In addition to this intrinsic problem, we should make the MTJ etching profile more vertical while minimizing the redeposition in order to avoid electrical shorting between free layer and pined layer in MTJ structure.

The object of this study is to provide characteristics of dry etching process available for sub 100nm STT-MRAM application. In this study, we investigate the effect of noncorrosive gas chemistry on MTJ etch rate, profile angle and hard mask selectivity, as a function of various process parameters such as RF power, working pressure, gas flow rate and ESC temperature in ICP etching system. It is also examined how the post etching treatment including oxygen plasma ashing and wet cleaning process affects the interface of MTJ films. The characteristics of MTJ interface are carefully analyzed by using high resolution transmission electron microscope (HR-TEM), electron energy-loss spectroscopy in the TEM (TEM-EELS) and X-ray photoelectron spectroscopy (XPS). Highly anisotropic profile of nearly 80 degrees is obtained by optimizing the etching condition and hard mask process scheme. Electromagnetic characteristics are also reported as a function of various etching conditions.

Higher porosity and new film chemistries are required to drive down the k value of porous ultra low k dielectrics integrated in advanced BEOL stacks. In the k=2.2 porous SiCOH films used in 32nm and 22nm BEOL stacks, higher carbon content and higher Si-C/Si-O ratio render these films more chemically similar to photoresist and to SiCN barrier layers. Hence, there is greatly reduced selectivity between the mask and low-k dielectric, and between the low-k dielectric and the barrier film. The via etch process has to move to a drastically different plasma regime in order to achieve mask selectivity and barrier selectivity, as well as to control RIE lag. New film chemistry and increased porosity also result in new film/plasma interactions, such as surface roughness phenomena observed both on planar and vertical surfaces. In some cases, plasma modification to the film from one step is only observed several steps beyond the modification point. In this work, film surface roughness phenomena will be examined for a k=2.2 porous SiCOH film utilizing a via first trench last integration scheme. Experiments show that surface roughness can arise from several etch steps if plasma conditions are not carefully controlled. Results will be presented with some of the process regimes explored.

This work was performed by the Research Alliance Teams at various IBM Research and Development Facilities.

For the fabrication of a multilevel resist (MLR) based on amorphous carbon (a-C) layer and Si₃N₄ hard-mask layer (underlayer), etch selectivity of the Si₃N₄/a-C layer becomes increasingly critical with the feature size reduction. In this work, therefore, the highly selective etching process of the Si₃N₄ layer using chemical-vapor-deposited (CVD) a-C etch-mask was investigated by varying the following process parameters in CH₂F₂/H₂/Ar plasmas: etch gas flow ratio, high-frequency source power (P_HF) and low-frequency source power (P_LF) in a dual-frequency superimposed capacitively coupled plasma etcher. It was found that infinitely high etch selectivities of the Si₃N₄ layers to the CVD a-C on both the blanket and patterned wafers could be obtained for certain process conditions. In particular, the etch gas flow ratio was found to play a critical role in determining the process window for infinite Si₃N₄/CVD a-C etch selectivity, due to the change in the degree of polymerization. The etch results of patterned ArF PR/BARC(bottom anti-reflective coating)/SiO_x/CVD a-C/Si₃N₄ MLR structure supported the possibility of using a infinitely high selective etch processes of the Si₃N₄ layer using a very thin CVD a-C etch-mask for reduced overall aspect ratio of MLR structure during patterning.

For the next generation silicon substrates applied to nano-scale semiconductor devices, silicon-on-insulator (SOI) wafer is known to be one of the outstanding candidates because of the advantages such as high speed, high packing density, immunity from latch-up, low power dissipation, high resistance to ionizing radiation, etc.

For the SOI wafer, the surface roughness of SOI wafer is very important because it can change the physical and chemical properties of the top silicon layer of the SOI wafer. Many approaches have been attempted to reduce the surface roughness of the SOI wafer by chemical mechanical polishing, high temperature annealing, wet etching, etc. but these methods are known to have some problems such as long processing time, reliability of exact thickness control, etc.

In this study SOI wafers were etched by a chlorine neutral beam obtained by the low angle forward reflection of an ion beam and the surface roughness of the etched wafers was compared with that of the SOI wafers etched by a chlorine ion beam. The result showed that the surface roughness of the SOI wafer etched by the chlorine neutral beam was significantly improved compared to that etched by the chlorine ion beam. By etching about 150nm silicon of about 300nm-thick top silicon layer of SOI wafer using the chlorine neutral beam, the rms surface roughness lower than 1.5 Å could be obtained with the etch rate of about 750 Å /min while that etched by the chlorine ion beam showed the rms surface roughness higher than 2.5 Å.

The induced defects in the surface area of the SOI wafer by the ion beam and neutral beam were observed by high-resolution-transmission-electron-microscopy(HR-TEM). An atomic force microscopy(AFM) was employed to measure and evaluated the surface roughness of the SOI wafer before and after the etching process, respectively.

ACKNOWLEDGMENT

This work supported by the National Program for Tera-Level Nano devices of the Korea Ministry of Education, Science and Technology (MEST) as a 21st Century Frontier Program.

Low dielectric constant (low-k) materials for interlayer dielectric are important for the improvement of ULSI devices performance. The low-k films tend to be damaged during plasma processes. The damage free plasma processes are strongly required. Although many researchers have been studying on the plasma damage on the low-k films, there has been little in-situ evaluation of plasma damages. The in-situ evaluation is crucial for the clarification of damage generation mechanism because the damaged low-k films are modified after exposing to atmosphere. This work investigated the mechanism of plasma ashing damage on the porous SiOCH films by in-situ evaluation. We examined the effect of ions, radicals, and radiation using PAPE technique. The thickness and refractive index of porous SiOCH films were measured using in-situ spectroscopic ellipsometry. Si-CH₃ bond absorption was measured using in-situ FT-IR.

The ashing plasma was exited in a 100 MHz CCP etcher. We adopted porous SiOCH films (k = 2.3) as low-k films in this study. The ashing process condition was total gas pressure of 2.0 Pa, 100 MHz source power of 450 W, substrate temperature of 20 °C. In the evaluation, a Si plate or a MgF₂ window which transmits the radiation (greater than 115 nm in wavelength) were placed at 1 mm above or just on the low-k film during the ashing. We carried out 4 kinds of experiments : (a) nothing for evaluating of the interaction of ions, radicals, and radiation, (b) Si plate for evaluating of the effect of radicals, (c) MgF₂ window for the interaction of VUV radiation with radicals, (d) MgF₂ window with no space for the effect of VUV radiation.

In the case of H₂ plasma ashing, we confirmed that the interaction of ions, radicals, and radiation or that of radicals and radiation decreased the thickness of the porous SiOCH film. The interaction of radicals and radiation caused the increase of the refractive index. The interaction of ions, radicals, and radiation or that of radicals and radiation caused the decrease of Si-CH₃ bond absorption.

The experimental results showed that H radicals extracted Si-CH₃ bond and that effect was drastically promoted by radiation and ions. The decrease of Si-CH₃ bond caused the decrease of polarizability and density of the film. However, ions made the film contract. Then, the refractive index of the films exposed to ions, radicals and radiation drastically increased. From these results, we proposed a mechanism of the plasma damages on porous SiOCH films.

These days, atmospheric pressure plasmas are being investigated as the application to the flat panel display device processing such as indium tin oxide etching, the deposition and etching of thin film transistor materials (SiO₂, amorphous silicon, and Si₃N₄) in addition to the surface treatment. Especially, among the various atmospheric pressure plasmas sources, much attention has been paid to Dielectric Barrier Discharge (DBD) due to its potential to numerous industrial applications such as plasma ashing, etching, thin film deposition, etc. The DBD, which is consisted of two parallel electrodes covered by dielectric plates, has been studied most widely due to the easier generation of stable glow discharges and the possibility of large-area plasma processing compared with other atmospheric pressure plasma sources.

In this study, using a modified DBD called “pin-to-plate DBD”, SiO₂ was etched and its plasma characteristics were investigated. Especially, the effect of additive gas such as CF₄ and C₄F₈ gas to the gas mixture of N₂ (60 slm)/ NF₃ (600 sccm) on the SiO₂ etch characteristics was investigated. The results showed that the increase of C₄F₈ (200 ~ 800 sccm) to the gas mixture decreased the SiO₂ etch rate continuously, while, the addition and increase of CF₄ (1 ~ 10 slm) to the gas mixture increased the SiO₂ etch rate until 7 slm of CF₄ was added and the further increase of CF₄ decreased the SiO₂ etch rate . The increase of SiO₂ etch rate up to 7 slm CF₄ is from the effective removal of Si in SiO₂ by F atom through the removal of oxygen in SiO₂ by carbon in CF_X in the plasma. However, the decrease of SiO₂ etch rate with further increase of CF₄ was related to the formation of a thick C-F polymer layer formed on the SiO₂ surface. The SiO₂ etch rate of about 243 nm/min could be obtained with the gas mixture of N₂ (60 slm)/ NF₃ (600 sccm)/ CF₄ (7 slm) when input voltage and operating frequency to the source were 10 kV and 30 kHz, respectively.

The role of protein-repelling coatings is to limit the interaction between a device and its physiological environment, by inhibiting the non-specific protein attachment. Plasma-polymerized-PEG (pp-PEG) surfaces are of great interest since they are known to avoid protein adsorption [1]. In this study, pp-PEG films have been deposited on gold and polyvinylfluoride (PVF) surfaces, by means of atmospheric pressure plasma liquid deposition (APPLD) and atmospheric pressure plasma enhanced chemical vapour deposition (APPECVD) processes. A comparison between those two methods has been made by investigating the chemical composition of the films using infrared reflection absorption spectroscopy (IRRAS), X-ray photoelectron spectroscopy (XPS) and secondary ions mass spectroscopy (SIMS). By observing the C1s high resolution XPS spectra of our samples, it appears that for APPECVD samples, the hydrocarbon component (285 eV) is increasing as the power of the plasma is increased, revealing a higher fragmentation of the precursor (tetra(ethylene glycol)dimethylether), while for APPLD samples no changes occur. The same conclusion could be made by observing the typical ToF-SIMS peaks (m/z = 45 (CH₃-O-CH₂⁺ and ⁺CH₂CH₂-OH), 59 (CH₃-O-CH₂-CH₂⁺), 103 (CH₃-(O-CH₂-CH₂)₂⁺)) that are decreasing in the case of high powered APPECVD treatments. The non-fouling properties of our samples have been studied with Bovine Serum Albumin (BSA) adsorption. On that purpose, XPS was used to track the presence of BSA on the surface by using the N1s signal coming out from the protein. For the APPECVD samples, a low plasma power (30 W) leads to an important reduction of BSA adsorption (over 90% reduction). However, higher-powered treatments tend to reduce the non-fouling ability of the surfaces (around 50% of protein adsorption reduction for a 80 W deposition). The same order of magnitude of BSA adsorption reduction (over 90%) is obtained for the APPLD surfaces, whatever is the power of the treatment. Those results show an important difference between APPECVD and APPLD processes in terms of power of the plasma treatment.

The technique of plasma polymerization has attracted much interest for its ability to deposit uniform polymeric coatings whose thickness can be controlled with nanometer precision via the plasma duration. Thus, plasma polymer (PP) films are well suited to application in optical precision instruments and other devices where high quality optical films are required. However, the range of refractive indices (RI) reported for plasma polymers is quite narrow. One objective of our research is to study how higher RI values can be achieved. Another objective is to develop PP coatings whose RI varies gradually, from a value matching the substrate to a higher value. Gradient Refractive Index (GRIN) films have been produced by plasma polymerization, with the RI changing linearly with the film composition [1]. Here we report on the plasma polymerization of bromoethane and other brominated monomers to create a database of RI versus PP composition prior to using such data for producing GRIN PP films. Factors affecting the PP film composition and thus the RI are plasma deposition power, pressure, deposition rate, and the monomers themselves. Characterization of compositionally homogeneous PP films is performed by optical techniques such as ellipsometry and the data obtained can be extrapolated to provide information about graded polymer films. This is also the case when analyzing homogenous polymer films with techniques such as x-ray photoelectron spectroscopy and Time of Flight Secondary Ion Mass Spectrometry (TOF-SIMS). However, several complementary techniques must be employed when analyzing GRIN films. Techniques such as Small Angle X-ray Scattering (SAXS) and X-ray Reflectometry (XRR) have been utilized because the power of x-rays allows the RI in any medium to be assumed to be unity, as opposed to other optical techniques that essentially deal with optical interfaces and assume samples to be optically homogenous. SAXS is particularly useful as it is capable of measuring polymer blends. Similarly, neutron techniques such as neutron reflectometry (NR) and Small Angle Neutron Scattering (SANS) are complementary to x-ray techniques, as shown by previous work where PP films were analyzed using both XRR and NR [2]. Studies are currently underway to examine the composition of homogenous PP films. One interesting method for depth profiling thin polymer films is TOF-SIMS analysis with a C₆₀ gun, which has been used to characterize discrete multilayer structures [3].

[1]Jiang H; et al. Chem. Mater. 2004, 16, 1292

[2]Nelson A; et al. Langmuir 2006, 22, 453

[3]Zheng LL; et al. Anal. Chem. 2008, 80, 7363

Current ULSI technology requires the use of hafnium related high-k dielectrics with ~3 nm thick for MOSFET to lower

the power consumption. HfSiON is the most applicable chemistry for the high-k material with proper energy band

alignment, large area uniformity, and thermal stability. The direct formation of HfSiO film from the Hf overlayer and underlying SiO₂ utilizing the thermal interfacial reaction was previously proposed [1]. The process demonstrates remarkably low inpurity in the film due to the lack of carbon in contrast to the case of MOCVD processes. In our case, 2.5 nm thick Hf metal layer is deposited with e-beam deposition source on SiO₂/Si(100) surface uniformly. The

morphology obtained with the in-situ non-contact AFM measurement revealed the surface consists of the high density array of Hf nano particles with the size of 4 nm in diameter.

The exposure of atomic nitrogen and ions from the non-equiliblium plasma enables the introduction of N into the film and increases the interfacial reaction rate of Hf and SiO. Within the first 1 min., the Hf nano particles are

oxynitrided with the N atoms from the plasma and the O atoms supplied from the lower interface judging from the XPS

analysis. The following plasma exposure (~10min.) enables the diffusion of Si atoms into the higk-k film from the

underlying SiO layer. The Si content in the film increases with the exposure time and becomes comparable to the Hf content with 35 min. exposure. The XPS spectrum shows the Si incorporated is mostly nitrided in the film. The spectrum also indicates the interfacial SiO layer is nitrided and this leads to the minimized EOT of the high-k stack structure.

[1] H. Watanabe, Appl. Phys. Lett. 85, 449 (2004).

We present a study of the near substrate plasma properties in an unbalanced magnetron deposition system used to deposit hydrogenated amorphous germanium thin films. The system is equipped with external Helmholtz coils that allow control over the near substrate plasma density. Four plasma diagnostic methods are used to characterize the plasma; a cylindrical Langmuir probe, a flat probe with a guard ring, a retarding field analyzer, and optical emission spectroscopy. The complementary nature of the diagnostics results in a robust determination of the plasma density, electron temperature, and plasma potential. The plasma density inferred from the cylindrical and flat probe results are corroborated by the relative ion currents to the retarding field analyzer. The latter also allows determination of the plasma potential, which agrees well with that inferred from the cylindrical probe results. The electron temperature inferred from the cylindrical probe is approximately corroborated by the relative intensity of the argon optical emission lines, but there is also some evidence that the electron energy distributions have a non-Maxwellian part. In our system the near substrate plasma density can be varied by about a factor of 25. Higher plasma densities near the substrate result in a lower electron temperature and a slight negative shift in the plasma potential. Hydrogen-argon mixtures results in large increases in both plasma density and electron temperature compared to argon-only plasmas. Possible reasons for this phenomena are discussed.

Formation of dust particles due to plasma-surface interaction has attracted a great deal of attention in many fields because dust particles can cause quality deterioration in semiconductor manufacturing [1, 2] and can contain a large amount of tritium in fusion devices [3], and so on. Therefore, it is important to reveal their formation mechanisms, their transport as well as their accumulation area. Up to now, we have collected carbon dust particles formed due to interaction between graphite target and helicon plasmas using in-situ and ex-situ collection methods [4], and have analyzed them. Here we report experimental results regarding carbon particle formation due to interaction between graphite and helicon plasmas and discuss their formation mechanisms.

Experiments were carried out with a helicon plasma reactor. Hydrogen or deuterium plasmas were generated by applying pulsed rf voltage of 13.56 MHz to a helicon antenna. The ion density and electron temperature obtained in the helicon discharge reactor are 4x10¹⁰-3x10¹² cm^-3 and 4.5-11.8 eV, respectively. Dust particles collected in the helicon plasma reactor can be classified into small spherical particles, agglomerates whose primary particles are around 10 nm in size and large irregular particles. There are many small dust particles of 1 nm-1 μm in size. The typical density ratio among them is 2x10³ : 1: 3. The smaller their size is, the higher their number density is. The size regions of these dust particles are 1-500 nm for small spherical particles, 50-700 nm for agglomerates and 50 nm-6 μm for large irregular particles, respectively. The three kinds of dust particles suggest three formation mechanisms: CVD growth, agglomeration, and peeling from walls. The dust particles of 10 nm in size have the highest probability to be charged positively, whereas those above 30 nm in size are charged negatively [5]. Agglomeration between a negative large agglomerate and a positive small dust particle takes place during the discharging period.

[1] M. Shiratani, M. Kai, K. Koga, and Y. Watanabe, Thin Slid Films, 427, 1 (2003).

[2] N. Hershkowitz, IEEE TRANSACTIONS ON PLASMA SCIENCE, 26, 1610 (1998).

[3] J. Winter, Plasma Physics and Controlled Fusion, 40, 1201 (1998).

[4] K. Koga, S. Iwashita, S. Kiridoshi, M. Shiratani, N. Ashikawa, K. Nishimura, A. Sagara, A. Komori, LHD Experimental Group, Plasma and Fusion Research, in press.

[5] Y. Watanabe, M. Shiratani, H. Kawasaki, S. Singh, T. Fukuzawa, Y. Ueda, and H. Ohkura, Journal of Vacuum Science & Technology, A14, 540 (1996).