Atomic Layer Deposition Mechanism Of Silicon Dioxide And Silicon Nitride Thin Films With Promising Precursors Design

SiO2 and Si3N4 thin films are the most important dielectric materials widely used in the microelectronics and semiconductor industries, due to their excellent physical and chemical properties, such as high optical transmittance, dielectric performance, chemical inertness and hardness. With the miniaturization of semiconductor devices and the continue shrinkage of feature sizes, the conventional deposition methods, such as PVD and CVD are no longer suitable for producing high-quality coatings in ultra-high aspect ratio nano-structures at low temperature due to their conformity nature. Atomic layer deposition (ALD) techniques are widely accepted to be the preferred methods for depositing SiO2 and Si3N4 thin films because of their optimal step coverage, excellent conformality and precise thickness control, etc. Nevertheless, the precursors used in the current ALD process are mostly derived from the CVD reaction, and the low reactivity of these precursors results in higher reaction temperature, low deposition rate and impurities in films. Therefore, it’s necessary to develop and screen some precursors with appropriate reactivity for low temperature deposition of SiO2 and Si3N4 thin films via ALD technology, which is of great importance for application of ALD technology in semiconductor industry in the future. Although some experimental researches have been focused on ALD of SiO2 and Si3N4 thin films, the systematical investigations by computational chemistry is still missing. In this thesis, we present a theoretical study based on density functional theory to systematically unveil the thermodynamic behavior and kinetics mechanism of ALD processes, including the dissociative chemisorption of different precursors on the substrate and the removal of residual ligands on the surface. We reveal the effects of structural properties of precursors on the entire ALD reaction mechanism to screen suitable precursors for low temperature deposition of SiO2 and Si3N4 thin films. The main results and conclusions are listed below.1. To identify the ALD reaction mechanism of SiO2 deposition, a full cycle of an ALD process using different aminosilanes with diverse structures as precursors and ozone as the oxidizing agent was investigated using density functional theory. Our calculations indicated that the amounts and species of substituents in aminosilanes had an influence on the ALD reaction mechanism and the deposition process of SiO2 thin films. In the dissociations of conventional chloro-silicane precursors on a-SiO2(001) surface, high activation barriers were observed due to their relatively low reactivity, which was consistent with experimental reports at high temperature. Different aminosilanes with diverse structures were designed, according to the calculated thermochemical and kinetic properties, the aminosilane precursors were found more reactive than the conventional chloro-silicane precursors, and some promising precursors were selected.(a). The ALD growth mechanism of SiO2 thin films with 20 different mono-aminosilanes as precursors was addressed. The mono-aminosilanes precursors firstly underwent dissociations on a-SiO2(001) surface, resulting in the anchoring of-SiH3 on the surface, which can be facilely oxidized into-Si(OH)3 by ozone with highly exothermic energies and modest activation barriers. To grow a denser and purer SiO2 films, two possible pathways for the crosslink reaction between the -Si(OH)3 species and the neighboring surface hydroxyl groups were investigated. The path that led to the growth of the SiO2 layer with crystalline morphology was identified to be both thermodynamically and kinetically preferred. This is important for growing dense and conformal SiO2 thin films. The dehydration crosslink reaction was the rate controlling step of the entire ALD process, which will determine the crystal orientation of SiO2 thin films. We also attempted to explore the influences of substituent species in mono-aminosilanes on the precursor’s reactivity, finding the reactivity been affected by the substituent’s electronic properties and steric effects simultaneously. The results indicated that the precursor’s reactivity can be regulated by tuning the electron distribution of N atoms and the volume of substituents via different substituting policy. The chloride methyl in CMAS and DCMAS molecules were strong electron-withdrawing groups, leading to the very high activation barriers for their dissociations. The double bond or phenyl in VAS, PhAS and MPhAS can conjugate with N atom to form stable structures, lowering the reactivity of these precursors. The dissociations of alkyl substituted mono-aminosilanes were favorable with relatively moderate activation barriers, and these molecules are suitable precursors for ALD growth of SiO2 thin films at low temperature, among of which the SBAS, DPAS, DIPAS and DSBAS molecules are the optimal choices.(b). The ALD growth mechanism of SiO2 thin films with 14 different bis-aminosilanes as precursors was addressed, which was different from the process using mono-aminosilanes. The bis-aminosilanes firstly underwent two sequential dissociations on α-SiO2(001) surface, then the produced-SiH2 can be favorably oxidized into-Si(OH)3 by ozone both thermodynamically and kinetically. The second dissociation steps with relatively higher activation barriers are the rate controlling steps of the entire ALD process, which will determine the crystal orientation of SiO2 thin films. The path that led to the growth of the SiO2 layer with crystalline morphology was thermodynamically more favorable, while the path that led to grow disordered SiO2 thin films was kinetically more facile. The reaction reactivity of bis-aminosilanes was mainly affected by the steric effects. For instance, the BDIPAS, BDSBAS, BDIBAS and BDTBAS molecules with bulky substituents, the activation barriers of their dissociations were very high. When the substituent sizes decrease, such as BMAS, BEAS, BPAS, BIPAS, BSBAS, BIBAS, BTBAS, BDMAS, BDEAS and BDPAS molecules, their dissociations were more facile with relatively lower activation barriers, and these molecules are promising precursors for ALD growth of SiO2 thin films with modest activation barriers. BDEAS is the optimal one among these precursors, while BDPAS requires relatively elevated reaction temperature. Given enough time and suitable temperature to settle, these precursors will fully react with the substrate, leaving no impurities on the surface.(c). The ALD growth mechanism of SiO2 thin films with 7 different tri-aminosilanes as precursors was addressed. The tri-aminosilanes containing three amino groups, need to undergo three sequential dissociations on a-SiO2(001) surface, and the reaction reactivity was also mainly affected by the steric effects. The TEMAS and TDEAS molecules with bulky substituents required much higher activation barriers to dissociate. The TMAS, TEAS, TIPAS, TTBAS and TDMAS molecules with relative small size substituents, the first and second dissociations of these precursors can occur with relatively modest activation barriers, and these molecules can be used as the Si precursors for ALD growth of SiO2 thin films. However, the sequential dissociations of these precursors occur only up to the second step even at high temperature, with the third amino groups anchored on the surface, giving rise to surface impurities. We can utilize H2O to remove these residual amino fragments to reduce the content of impurities in films.2. To identify the ALD reaction mechanism of Si3N4 deposition, a full cycle of an ALD process using different aminosilanes with diverse structures as precursors and NH3 or NH2NH2 as the nitrogen sources was investigated using density functional theory. The amounts and species of substituents in aminosilanes also had an influence on the ALD reaction mechanism and deposition process of Si3N4 thin films. Two Si3N4(001) surface models were considered, and the aminated surface was more active than the hydrogenated surface. The reactivity of conventional chloro-silicane precursors was also found very low on aminated Si3N4(001) surface, which was consistent with experimental reports at high temperature. Similarly, the aminosilane precursors were found more reactive than the conventional chloro-silicane precursors again. The results indicated that the ALD of Si3N4 thin films is an inherent high-temperature process,(a). The ALD growth mechanism of Si3N4 thin films with 8 different mono-aminosilanes as precursors was addressed. The mono-aminosilanes precursors firstly underwent dissociation on aminated Si3N4(001) surface, then NH3 or NH2NH2 was introduced to reduce the-SiH3 to form-Si(NH2)3 species, followed by the crosslink reaction between the-Si(NH2)3 species and surface amino groups. In the reduction reaction, the NH2NH2 was found more reactive than NH3 molecular. The dissociations of eight selected mono-aminosilanes were all slightly difficult with relatively high activation barriers; and the substituent sizes were larger, the activation energy was higher. At relatively high temperature, these mono-aminosilanes can be used as Si precursors for ALD growth of Si3N4 thin films, but are unsuitable for low-temperature deposition. The crosslink reaction was the rate controlling steps of the entire ALD process with very high activation barrier, which could lead to grow disordered and low density Si3N4 films with high H content at high temperature.(b). The ALD growth mechanism of Si3N4 thin films with 6 different bis-aminosilanes as precursors was addressed, which was different from the process using mono-aminosilanes. The bis-aminosilanes firstly underwent two sequential dissociations on aminated Si3N4(001) surface, then the NH3 or NH2NH2 molecular can reduce the-SiH2 to form-Si(NH2)2 species. The first dissociations of six selected bis-aminosilanes on the aminated Si3N4(001) surface were unfavorable with relatively high barriers, and the second dissociations of these precursors were the rate controlling steps of the entire ALD process with higher activation barriers, which will determine the crystal orientation of Si3N4 thin films. The results suggested that, the ALD process for deposition of Si3N4 thin films using bis-aminosilanes as the Si precursors is inherently a high-temperature process. At elevated temperatures, the largely disordered Si3N4 thin films can be obtained, while at a lower temperature, these precursors are unsuitable for deposition of Si3N4 thin films.(c). The ALD growth mechanism of Si3N4 thin films with TMAS or TDMAS as precursors was addressed. The TMAS and TDMAS molecules containing three amino groups, need to undergo three sequential dissociations on the aminated Si3N4(001) surface, then the NH3 molecular can reduce the -SiH to form -Si-NH2 species. All these sequential dissociations required high temperature with very high activation energies, and the second dissociations were the rate controlling steps of the entire ALD process, which will determine the crystal orientation of Si3N4 thin films. Therefore, the ALD process for deposition of Si3N4 thin films using tri-aminosilanes as the precursors was also inherently a high-temperature process, the sequential dissociations of these precursors are expected to be incomplete and some fragments may be trapped in the films as impurities, degrading the device performance.(d). The dissociation activation energies of all aminosilane precursors on the aminated Si3N4(001) surfaces were relatively high. Possibly, the aminated Si3N4(001) surface may not be reactive enough to decompose the precursors, as the H atoms on the surface after the reduction reaction will depress its reactivity. To obtain Si3N4 thin films at low temperature, we should adopt new ALD technology, such as plasma enhanced ALD, using N2 plasma as the nitrogen sources to remove the surface H atoms and enhance the surface reactivity.In summary, computational chemistry has been successfully applied to reveal the ALD reaction mechanism of SiO2 and Si3N4 thin films growth and to screen promising precursors in the semiconductor industry. The results are in good agreement with experimental observations and even capable to provide useful guidance for experimental and industrial applications. Our theoretical methods also have an important reference value for other thin films development via ALD technology.