Reaction mechanism study on the atomic layer deposition of silicon nitride : 실리콘 질화막의 원자층증착 반응기구 연구

다국어 초록 (Multilingual Abstract)

Over the past 50 years, the semiconductor devices have been densified by two-dimensional (2D) scaling, and their performance have been improved by applying new materials such as Cu interconnect, low-k, SiGe, high-k and metal gates. Recently, new three-dimensional (3D) structures has begun to be adopted. For logic devices, 2D planar transistors are converted to 3D structured FinFETs, and even for NAND flash memory devices, memory cells that have been horizontally integrated are being replaced by memory cells with 3D structures stacked with dozens of vertical layers.
Atomic layer deposition (ALD) is gaining attention as the most promising candidate to replace existing thin film deposition technologies such as physical vapor deposition (PVD) and chemical vapor deposition (CVD) processes. Since ALD is an atomic layer-by-layer growth process based on self-limiting surface reactions, this technique can produce uniform and conformal thin films on complex 3D structures with very high aspect ratios at low temperatures.
Silicon nitride thin film has been widely used because of their chemical stability and resistance to diffusion of impurities as well as their etching selectivity compared to SiO2 and Si. The most important applications of silicon nitride films in next-generation semiconductor devices include the gate spacers in field-effect transistors and the charge trap layers in NAND flash devices. Various precursors and reactants have been studied for the ALD of silicon nitride films for these applications, and there is a growing interest in the reaction mechanism of the ALD of silicon nitride.
Understanding the deposition reaction mechanism of ALD is very important because it affects the composition, structure and the properties of the deposited films. In order to grow silicon nitride films with an excellent step coverage and high quality, this dissertation focuses on predicting the surface chemical reactions by density functional theory (DFT) calculations and on identifying the ALD reaction mechanism by in-situ monitoring techniques
The energies of surface reactions for the 1st half reaction of ALD process, such as adsorption energy, reaction barrier and reaction energy, were calculated by DFT. The bond dissociation energy (BDE), the highest occupied molecular orbit (HOMO), the lowest unoccupied molecular orbit (LUMO) and charge density were also calculated to predict the characteristics of precursors and substrates. The substrate was modeled as two types: NH*/NH2*-terminated β-Si3N4 surface and under-coordinated β-Si3N4 surface. The reaction was monitored by measuring the weight change of the growing film using the quartz crystal microbalance (QCM) and analyzing the chemical species on the pellet surface using Fourier transform infrared (FTIR) spectroscopy.
Thermal ALD SiN was investigated using chloride precursor. The DFT calculation was used for analyzing the BDE and reaction pathways to understand the reaction mechanism. The various reaction paths of Si3Cl8 were calculated and then compared with the in-situ monitoring results and experimental results. Increasing the number of silicon atoms in precursor, the BDE values of Si–Si was weaker from 3.52 eV to 3.33 eV and the BDE values of Si–Cl also weaker from 4.62 eV to 4.35 eV. Among silicon chlorides, Si3Cl8 showed the fastest reaction with the lowest BDE value, which is related in 1st half reaction. The saturation dose was reduced from 1010 L to 107 L and the deposition temperature was decreased to 300℃ with a high growth rate of ~0.2 nm/cycle. However, Si3Cl8 shows different growth kinetics at different deposition temperatures. The adsorption at high temperature is expected to make –SiCl2 species and the formation of 2HSiCl3 by-products after the 1st half reaction. On the other hands, two Si-N bonds generated with the low energy barrier of 1.38 eV, and the deposited film showed Si-rich concentration at low temperature. Because it is difficult to remove –Cl after silicon chloride precursors are adsorbed on the substrate. Therefore, it is necessary to use NH3 and NH3 plasma at a high temperature or to use a stronger nitriding agent.
The PEALD process was studied using aminosilane precursors to grow high quality silicon nitride films at low temperatures. To simulate the PEALD process, under-coordinated silicon nitride surface formed by N2 plasma was modeled. The new precursor was designed to improve film quality, based on TSA. To improve the thermal stability, reactivity and vapor pressure of the precursors, the –SiH3 ligand was substituted with the –SiMe3 ligand and the –(MeSiHNMe2) ligand and then the precursor properties were compared using the DFT calculation. The 1st half reaction was calculated on the under-coordinated silicon nitride surface after selecting bis (dimethylaminomethylsilyl) trimethylsilyl amine (DTDN2-H2) among several TSA derivatives. The reaction energy barrier of DTDN2-H2 was ~1.5 eV. As a low energy barrier, DTDN2-H2 had a wide ALD window from 250℃ to 400℃ and GPC was 0.36 Å/cycle. It showed excellent step coverage of more than 80% in trench structure with AR of 5.
Since the silicon nitride films for next-generation semiconductor devices required a better step coverage, a new precursor was designed that are superior to the TSA derivatives. To increase the thermal stability and reactivity of the precursors, a ring structure of Si-N bonds was designed and a suitable ligand was added to have sufficient vapor pressure for the ALD process. The energy of barrier of the designed 1,3-di-isopropylamino-2,4-dimethylcyclosilazane (CSN-2) was 1.32 eV, which is lower than that of DTDN2-H2 in the 1st half reaction. And CSN-2 showed a wider ALD window from 200℃ to 500℃ with GPC of 0.43 Å/cycle. All of the deposited films in the ALD window region showed excellent step coverage over 90%. However, WER properties of the films deposited on the bottom sidewall at low temperature were not good. This is a limitation caused by plasma recombination loss of the N radical.
In order to solve this problem, a new 3-step PEALD process was designed by adding NH3/N2 plasma to the standard N2 plasma PEALD process. The NH3 plasma can more effectively remove the ligands of the adsorbed precursor, but it is difficult to adsorb the precursor in the next cycle by forming NH*/NH2* surface. The following N2 plasma treatment removes –H on the surface and activates the surface so that the precursor is easily adsorbed in the 1st half reaction. As a result, silicon nitride films having high quality and excellent step coverage were deposited on a trench with high aspect ratio. The bottom sidewall coverage was improved from 81% to 95% and the wet-etch characteristic was improved form 13 Å/min to 3 Å/min.
In this dissertation, the in-situ process monitoring and DFT calculations successfully applied to explain the ALD reaction mechanism for growth of silicon nitride films with excellent film quality. And the precursors are designed by calculation of the precursor properties, and the calculation results well predicts the experimental results.

목차 (Table of Contents)

• Chapter 1 Introduction 1
• Chapter 2 Atomic layer deposition of silicon nitride: Literature survey 6
• 2.1 Atomic layer deposition 6
• 2.1.1 ALD process 6
• 2.1.2 ALD precursors 10
• 2.1.3 ALD of silicon nitride 13
• 2.2 In-situ monitoring for ALD 18
• 2.2.1 Quartz crystal microbalance 20
• 2.2.2 Fourier transform infrared spectroscopy 24
• 2.3 Density functional theory in ALD 26
• 2.3.1 Density functional theory 27
• 2.3.2 Frontier Molecular Orbital Theory 31
• Chapter 3 Experiment setup, method, and data analysis 32
• 3.1 ALD system 32
• 3.2 Data analysis using in-situ monitoring tools 35
• 3.2.1 QCM data analysis 35
• 3.2.2 FTIR data analysis 38
• 3.3 Parameter setting of DFT calculation 40
• 3.4 Analysis of deposited films 44
• Chapter 4 Thermal ALD SiN using chloride precursor 45
• 4.1 Comparison the properties of chloride precursors 45
• 4.2 Density functional theory calculation of 1st half reaction 49
• 4.3 In-situ monitoring of ALD process 57
• 4.4 Experimental results of thermal ALD 63
• 4.5 Summary 73
• Chapter 5 PEALD SiN using silylamine silicon precursor 75
• 5.1 Design of noble silicon precursor for PEALD 75
• 5.2 Synthesis of silylamine compounds 76
• 5.3 Characterization of silylamine compounds 78
• 5.4 Modeling and simulation of PEALD process 80
• 5.5 Growth of PEALD SiN using DTDN2-H2 85
• 5.6 Effect of deposition temperature and RF power 87
• 5.7 Step coverage of PEALD silicon nitride. 93
• 5.8 Summary 95
• Chapter 6 PEALD SiN using cyclosilazane-type silicon precursor 96
• 6.1 Modeling and simulation of PEALD process 97
• 6.2 Growth of PEALD SiN using CSN-2 101
• 6.3 Effect of growth temperature 103
• 6.4 Step coverage of PEALD silicon nitride 105
• 6.5 3-step PEALD SiNx 109
• 6.6 Summary 114
• Chapter 7 Conclusions and discussions 115
• Chapter 8 Reference 118