A Three-Step Atomic Layer Deposition Process for SiNx Using Si2Cl6

Subscriber access provided by Miami University Libraries

Surfaces, Interfaces, and Applications

A Three-Step Atomic Layer Deposition Process for SiN Using SiCl, CHNH, and N Plasma x

2

6

3

2

2

Rafaiel A Ovanesyan, Dennis M. Hausmann, and Sumit Agarwal ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/acsami.8b01392 • Publication Date (Web): 11 May 2018 Downloaded from http://pubs.acs.org on May 14, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

A Three-Step Atomic Layer Deposition Process for SiNx Using Si2Cl6, CH3NH2, and N2 Plasma Rafaiel A. Ovanesyan,1 Dennis M. Hausmann,*,2 and Sumit Agarwal *,1 1

Department of Chemical and Biological Engineering, Colorado School of Mines, Golden, CO 80401

2

Lam Research Corporation, 11155 SW Leveton Drive, Tualatin, OR 97062

Abstract We report a novel three-step SiNx atomic layer deposition (ALD) process using Si2Cl6, CH3NH2, and N2 plasma. In a two-step process, non-hydrogenated chlorosilanes such as Si2Cl6 with N2 plasmas lead to poor quality SiNx films that oxidize rapidly. The intermediate CH3NH2 step was therefore introduced in the ALD cycle to replace the NH3 plasma step with a N2 plasma, while using Si2Cl6 as the Si precursor. This three-step process lowers the atomic H content, and improves film conformality on high aspect ratio nanostructures as Si–N–Si bonds are formed during a thermal CH3NH2 step in addition to the N2 plasma step. During ALD, the reactive surface sites were monitored using in situ surface infrared spectroscopy. Our infrared spectra show that, on the post-N2-plasma treated SiNx surface, Si2Cl6 reacts primarily with surface –NH2 species to form surface –SiClx (x = 1, 2, or 3) bonds, which are the reactive sites during the CH3NH2 cycle. In the N2 plasma step, reactive –NH2 surface species are created due to the surface H available from –CH3 groups. At 400 °C, the SiNx films have a growth per cycle of ~0.9 Å with ~12 atomic percent H. Films grown on high-aspect nanostructures have a conformality of ~90%. *

Corresponding author e-mails: [email protected]; [email protected]

Keywords: silicon nitride, atomic layer deposition, infrared spectroscopy, ellipsometry, dielectrics

1 ACS Paragon Plus Environment

ACS Applied Materials & Interfaces 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 28

1. Introduction As the dimensions of semiconductor devices continue to shrink past the current 14 nm technology node, deposition of ultra-thin, highly-conformal films on high aspect ratio nanostructures becomes necessary.1 Of the many different types of films used in semiconductor manufacturing, highly conformal growth of silicon nitride (SiNx) is critical for applications such as sidewall spacers and etch stop layers.2-3 To further improve the versatility of SiNx films, incorporation of C may be required to tune the dielectric properties for specific applications.4 Due to thermal budget limitations, a substrate temperature ≤400 °C is desired for SiNx growth.5 To meet these stringent requirements, a thin film growth technique known as atomic layer deposition (ALD) has been investigated.6 ALD is characterized by the exposure of a substrate to an alternating series of reactive precursors in an ABABAB… sequence. The heterogeneous gas-solid reactions of the precursors with the substrate are surface reaction limited, leading to exceptional conformality, with a typical film growth of ~1 Å per AB cycle.7-8 The replacement of one of the precursor steps with a plasma can lower the deposition temperature compared to a completely thermally-activated ALD process.9 However, despite the need for SiNx ALD in semiconductor manufacturing, to date, there are few known ALD processes that are capable of depositing highly-conformal, C-containing SiNx films at temperatures ≤400 °C.1 In previous studies, we developed and tested a plasma-assisted SiNx and a plasma-assisted SiCxNy ALD process; one using alternating exposures of Si2Cl6 and NH3 plasma,10 and another using alternating exposures of Si2Cl6 and CH3NH2 plasma,11 respectively. Si2Cl6 was chosen as the Si precursor because it is very reactive at ≤400 °C, and leads to saturated surface reactions at doses of ~1 Torr·s.10,12 In the CH3NH2 plasma ALD process, NH3 was replaced with CH3NH2 to incorporate C into the SiNx films.10-11 While these two ALD


Page 3 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


processes resulted in the growth of highly-conformal SiNx and SiCxNy films at substrate temperatures ≤400 °C, the resulting SiNx films had several shortcomings.10-11 In particular, when the SiNx or SiCxNy films were deposited on high-aspect-ratio three-dimensional (3-D) nanostructures, the material properties of the sidewalls were very different than the planar surface.10-11 The structure of the planar surface and the sidewalls in these 3-D structures was indirectly probed by observing the wet etch rate of the films in a solution of dilute aqueous HF. In both cases, the planar surfaces that are exposed to direct ion bombardment during the plasma step remained largely intact, while the sidewall surfaces were completely etched in the dilute HF solution. A high wet etch rate for SiNx in dilute HF has been attributed to a high H content. For SiNx and SiCxNy films deposited using NH3 and CH3NH2 plasmas, respectively, the H content was typically ~20 atomic percent in planar samples,10-11 but the H content on the sidewalls in 3D samples could not be directly measured. We speculate that the H content is even higher than ~20% in the sidewall surfaces, which leads to the highly nonuniform wet etch rate. Detailed studies of the surface reaction mechanism for the NH3 or CH3NH2 plasma based processes showed that atomic H generated in the plasma was incorporated into the film as –NH and –NH2 groups during the plasma step. In the subsequent Si2Cl6 half cycle, surface –NH groups were unreactive, while –NH2 surface groups reacted with Si2Cl6 and were converted to –NH surface groups, which were then incorporated into the SiNx film. One possible way to decrease the H content in SiNx films deposited by ALD, and to improve film quality, is to replace the NH3 or CH3NH2 plasma step with a N2 plasma step to eliminate atomic H in the plasma. However, in our research, we have found that a N2 plasma combined with chlorosilane precursors with no hydrogen such as Si2Cl6, does not lead to the ALD of SiNx, but instead growth of a poor-quality SiNx film in a non-self-limiting manner that oxidizes rapidly to SiNxOy even under high-vacuum



Page 4 of 28

conditions. To enable the use of a N2 plasma for a SiNx ALD process, a hydrogenated chlorosilane precursor such as SiH2Cl2 can be used, but the issue of low conformality that is typically associated with aminosilane Si precursors in N2-plasma-based ALD processes still remains.13-14 Despite the low deposition conformality, ALD processes based on aminosilanes and N2 plasmas have a positive attribute: the wet etch rate of the sidewalls and the planar surfaces in dilute HF is uniform, and much lower than NH3-plasma-based processes. Therefore, if the conformality of N2-plasma-based SiNx ALD processes can be improved, such films could be incorporated into semiconductor manufacturing. A potential way to enable the ALD of SiNx using Si2Cl6 and N2 plasma is through the introduction of an intermediate thermal step between the Si2Cl6 and the N2 plasma cycles. This intermediate thermal step should deliver both N and H atoms to the surface. Delivering a fraction of N atoms in a thermal step would help overcome the low conformality generally associated with SiNx films deposited using N2-plasma-based ALD processes,13-14 and the presence of H would alleviate the problem of forming a SiNxOy film in completely hydrogen-free SiNx ALD processes. Furthermore, we expect that such a process should improve the wet etch rate of the sidewall surfaces as a lower amount of H would be incorporated due to the absence of atomic H in the plasma.11 Considering N and H containing molecules, at ≤400 °C, NH3 is not reactive with the –SiClx (x = 1, 2, or 3) terminated surface that is left behind following the Si2Cl6 cycle,15 while N2H4 is difficult to handle due to safety concerns. Based on the literature, alkylamines are more reactive with –SiClx (x = 1, 2, or 3) terminated surfaces than NH3. Specifically, Zhu et al. showed that on a Si(100) – (2 × 1):Cl surface created by exposing a

pristine Si(100) – (2 × 1)

reconstructed surface to highly purified Cl2 at 23 °C, n-butylamine, n-octylamine, and aniline react at 177 °C to form Si2N–CxHy terminated surfaces.16-17 Therefore, for the intermediate step,


Page 5 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


we chose CH3NH2, as it is the simplest alkylamine, and is expected to readily react with a –SiClx (x = 1, 2, or 3) terminated surface left behind following the Si2Cl6 partial cycle.15 Since this precursor contains –CH3, it also introduces the possibility of incorporating C atoms into the SiNx film during the intermediate thermal step. Zhu et al. also showed that at temperatures >423 °C, the alkylamine terminated surface starts to decompose to form surface nitrides and carbides, which suggests that at higher substrate temperatures, surface amides created after the CH3NH2 step may lead to a greater degree of C incorporation into the SiNx film.16 Note, the final N2 plasma step in the three-step process is required to activate the –CH3 terminated surface obtained after the CH3NH2 step, which is fairly unreactive at ≤400 °C. In some sense, this three-step SiNx ALD process is similar to ones reported for aminosilane precursors [SiHx(NRR')4-x (x =1, 2 or 3; R, R' = alkyl)] such as bis(di-ethylamino)silane (BDEAS) and an N2 plasma, which also leads to surface amide linkages after the aminosilane precursor step. The key difference is that in the three-step process, after the CH3NH2 step, Si–N–Si bonds are formed on the surface through Si2N–CH3 formation with no –SiHx (x = 1, 2, or 3) bonds.18 Similar to an aminosilane and N2 plasma ALD process, during the subsequent N2 plasma step, –NHx (x = 1, 2) surface species can be created due to the availability of surface H from the surface –CH3 groups, allowing for Si2Cl6 chemisorption in the next step, and thus the continuation of the ALD process.19 In this article, we report on the surface reaction mechanism and the properties of the SiNx films deposited using the novel three-step plasma-assisted SiNx ALD process described above. The film composition, reactive surface sites, and adsorbed surface species were monitored using in situ attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy. Four-wavelength ellipsometry was used to monitor growth per cycle (GPC) and the refractive index of the SiNx films. Additionally, the properties of the SiNx films were characterized using



Page 6 of 28

ex situ diagnostic techniques such as cross-sectional transmission electron microscopy (TEM), Rutherford backscattering spectroscopy (RBS) combined with hydrogen forward scattering (HFS), and X-ray reflectivity (XRR). 2. Experimental Details In Situ ATR-FTIR Spectroscopy and Four-Wavelength Ellipsometry

Figure 1: Cross-sectional diagram of the ALD reactor showing the layout of the in situ ATRFTIR spectroscopy and the in situ four-wavelength ellipsometry setups. The SiNx films were grown in an ALD reactor (see Figure 1) evacuated to a base pressure of 1 × 10-6 Torr using a turbomolecular pump (Pfeiffer TMU-521P) backed by a mechanical pump (Edwards E2M28). The ALD reactor is outfitted with both in situ ATR-FTIR spectroscopy


Page 7 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


and in situ four-wavelength ellipsometry setups. These in situ diagnostics have been described in detail in previous publications.20-26 Briefly, in this setup, SiNx films were simultaneously deposited on a 50 × 20 × 1 mm ZnSe trapezoid with the short faces beveled to 45°, which acts as the internal reflection crystal (IRC) for the ATR-FTIR spectroscopy setup, and a ~25 × 15 mm double-side polished Si wafer, which acts as the substrate for the four-wavelength ellipsometry setup. The infrared beam enters and exits the reactor through KBr windows, while the visible light from the ellipsometer enters and exits through Kodial glass windows that are protected by gate valves during deposition (see Figure 1). The cutoff frequency of the in situ infrared setup is ~700 cm-1 due to the ZnSe IRC and the MCT-A detector. The infrared spectra in this paper were all recorded as difference spectra, where a fresh background was collected prior to each process step. The infrared data was recorded over a spectral range of 700–4,000 cm-1 with a spectral resolution of 4 cm-1. Each spectrum was averaged over 250–500 scans, depending on the requirements of that particular processing step. The ellipsometry measurements were averaged over 4 s and the data was recorded at the end of the purge step. The two substrates are placed on a heater that also acts as the grounded electrode for the capacitively-coupled radio-frequency (rf) (13.56 MHz) plasma. The distance between the grounded substrate heater and the top rf-powered electrode was kept at ~4.5 cm during all processing steps. ALD Process Parameters Figure 2 schematically shows the entire ALD process sequence. The three sequential partial cycles of Si2Cl6, CH3NH2, and N2 plasma in the ALD process were separated by N2 + Ar purge steps. For the entire ALD process, the gases and vapors were pumped through a dry mechanical pump (Kashiyama NV60N-2). To ensure that leftover unreacted precursor or reaction byproduct do not affect the following precursor step, following each N2 + Ar purge step,



Page 8 of 28

the ALD reactor was pumped down for 15 s using the turbomolecular pump. The chamber pressure dropped to ~1 × 10-5 Torr during this pump down step.

Figure 2: Schematic of the gas pulsing sequence during the three-step SiNx ALD process. The shaded regions correspond to the times when the ALD reactor was pumped down using the turbomolecular pump. The gas and vapor flow rates, except Si2Cl6, were metered using mass flow controllers, and the flow direction, whether into the reactor or to bypass, was controlled by pneumatic valves operated via LabVIEW. The complete gas pulsing sequence is shown in Figure 2. Si2Cl6 (SigmaAldrich, 96%) was delivered to the ALD reactor using a fill-release-hold method. The Si2Cl6 ampoule was heated to ~60 °C to increase the Si2Cl6 vapor pressure, while the delivery line to the ALD reactor was heated to ~70 °C to prevent Si2Cl6 condensation. The delivery line fill volume was ~10 cm3. During the Si2Cl6 partial-cycle, the release time during which the valve to the delivery line was open was 2 s. Following release, Si2Cl6 was held in the chamber for 20 s by closing the butterfly valve to the dry mechanical pump. During the Si2Cl6 release step, the chamber pressure was ~200 mTorr for a Si2Cl6 dose of ~4 Torr·s. Following the Si2Cl6 cycle, the ALD reactor was purged for 30 s at a flow of 150 standard cm3/min (sccm) of Ar (99.999%) and


Page 9 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


200 sccm of N2 (99.999%) at a chamber pressure of ~500 mTorr. Following the N2 + Ar purge step, the chamber was again pumped down to 1 × 10-5 Torr for 15 s, after which 150 sccm of Ar and 30 sccm of CH3NH2 (Matheson 99.5%) were introduced into the ALD reactor for 60 s resulting in a total pressure of ~300 mTorr. The partial pressure of CH3NH2 in the reactor during this step was ~50 mTorr, resulting in a CH3NH2 dose of ~3 Torr·s. The CH3NH2 step was followed with a 60 s N2 + Ar purge, and another 15 s pump down. Subsequently, N2 and Ar were reintroduced into the reactor and the pressure was allowed to stabilize for 15 s. Next, a plasma was ignited for a duration of 60 s. The input rf power to the plasma source was 100 W. The last step in the ALD cycle was a final 60 s N2 + Ar purge step at the same flow rates as the previous purge steps, followed by a final 15 s pump down using the turbomolecular pump. Surface Preparation To ensure that the three-step SiNx ALD process was studied on a relevant growth surface, prior to data collection for the SiNx ALD process, the ZnSe IRC and the Si wafer were coated with a ~5-nm-thick amorphous hydrogenated SiNx film deposited by plasma-enhanced chemical vapor deposition (PECVD) using process conditions described previously.4 The pressure in the chamber during SiNx PECVD was ~300 mTorr with 100 W rf power coupled to the plasma. To minimize the memory of the underlying substrate and the SiNx capping layer on the subsequent growth process, prior to collecting any infrared spectra, a thin SiNx film was deposited using ten to fifty Si2Cl6, CH3NH2, and N2 plasma ALD cycles. Ex Situ Film Characterization To determine the conformality of the SiNx ALD process, a ~25-nm-thick SiNx film was deposited at 400 °C over nanostructures with an aspect ratio of ~5. The cross-sections of the SiNx film and the high-aspect-ratio nanostructures were then imaged using a high-resolution TEM



Page 10 of 28

(JEOL 2010F). The elemental composition of a ~25-nm-thick SiNx film deposited at 400 °C on a planar Si wafer was obtained using RBS combined with HFS. The thickness and density of the ~25-nm-thick SiNx film was determined using an XRR (PANalytical X’Pert Pro MRD) with a Cu X-ray tube. 3.

Results and Discussion Figure 3 shows the infrared spectra of a SiNx film deposited using 100 Si2Cl6, CH3NH2,

and N2 plasma ALD cycles at 400 °C, with and without KBr window correction. The reference spectrum for both of the spectra in Figure 3 was recorded prior to the beginning of the first ALD cycle. To obtain the infrared spectrum of the SiNx films such that it is truly representative of their composition, the data collection methodology had to be modified. During the CH3NH2 cycle, this ALD process produces methylammonium chloride (CH3NH3Cl) due to the reaction of CH3NH2 with HCl, the latter being the reaction byproduct of CH3NH2 chemisorption on a –SiClx (x = 1, 2, or 3) terminated surface that is created in the preceding Si2Cl6 cycle. The CH3NH3Cl salt absorbs

Figure 3: Infrared spectra of a SiNx film deposited by 100 ALD cycles of Si2Cl6, CH3NH2, and N2 plasma at 400 ºC with and without the KBr windows in the infrared beam path.


Page 11 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


strongly in the infrared, but is unstable at the growth temperature of 400 °C. Therefore, this salt is not expected to exist on the SiNx film surface, but forms on the chamber walls, which are at room temperature. Thus, the contribution of this salt in the infrared spectrum is from the accumulation of deposits on the KBr windows (see Figure 1) that allow for the entrance and exit of the infrared beam into and out of the ALD reactor. These windows are at room temperature during SiNx deposition. To eliminate this effect, the KBr windows were physically removed after SiNx ALD (see Figure 3), and the entire reactor was purged with CO2-free dry air prior to collection of the infrared spectrum. In the window-corrected infrared spectrum in Figure 3, the features at ~2750–3000 cm-1 disappear, verifying that these features are related to the vibrational modes of methylammonium chloride salts on the KBr windows.27 Since HCl is also the likely surface reaction product produced during the previous Si2Cl6 half-cycle (see discussion later in this section on the surface reaction mechanisms), extended purge times were used for recording the data in Figure 3. Note, this window effect was observed only after ~100 ALD cycles. For partial-cycle spectra, most of the signal is from the IRC: the window contribution is negligible as it gets background corrected after each partial-cycle. In Figure 3, the prominent infrared feature at ~870 cm-1 is assigned to the Si–N–Si asymmetric stretching mode and is characteristic of a SiNx film.28 The absence of a discernible shoulder at ~780 cm-1, which corresponds to the Si–C–Si asymmetric stretching mode,29 indicates that a silicon carbide (SiCx) phase is not present in the deposited film in a significant amount. Thus, C in the SiNx film must be present in a different local bonding configuration, which means that the deposited film is primarily SiNx. Regarding the species responsible for C incorporation, two C-related features are evident in Figure 3: a feature at ~1600 cm-1 that can be unambiguously assigned to >C=N– stretching vibrations,30-31 and a strong feature at ~2180 cm-1



Page 12 of 28

that is attributed to carbodiimide (Si–N=C=N–Si) asymmetric stretching vibrations.32 This assignment is based on previous literature on SiCN ternary blends, where a similar feature was assigned to Si–N=C=N–Si species.32-33 Although this feature has also been assigned to –SiHx (x = 1, 2, or 3) and nitrile (–C≡N) stretching vibrations, the rationale behind assigning this feature to the Si–N=C=N–Si stretching mode is described later in this section. In Figure 3, the shoulder at ~1180 cm-1 is assigned to –NH bending vibrations, while the broad feature at ~3350 cm-1 is attributed –NHx (x = 1, 2) stretching vibrations. The assignment of the infrared feature at ~2180 cm-1 primarily to the Si–N=C=N–Si stretching vibration rather than the –SiHx (x = 1, 2, or 3) stretching vibration, the absence of an observable –NH2 scissor mode at ~1550 cm-1, and the absence of observable –CHx (x = 1, 2, or 3) stretching vibrations at ~2900 cm-1 in Figure 3 suggests that the primary method of H incorporation in the SiNx film deposited via the three-step ALD process is as –NH species. The decrease in the absorbance at ~2050 cm−1 in both of the infrared spectra in Figure 3 is attributed to changes in the underlying PECVD SiNx film, as this feature is present in the infrared spectra of the PECVD SiNx films prior to the start of the ALD process. The partial cycle infrared spectra in Figure 4 shows the absorbance change during the Si2Cl6, CH3NH2, and N2 plasma steps at a substrate temperature of 400 °C. On a post-N2-plasma SiNx growth surface, we can say that Si2Cl6 reacts primarily with –NH2 surface species, as evidenced by the absorbance decreases at ~1550 cm-1 and ~3400 cm-1, corresponding to the


Page 13 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4: Infrared spectra showing the absorbance change during the Si2Cl6, CH3NH2, and N2 plasma partial cycles at a substrate temperature of 400 °C. –NH2 scissor mode34 and the –NH2 stretching mode,28 respectively,. The corresponding increase in the –NH stretching and bending modes at ~3350 and ~1180 cm-1 suggests that the reaction of Si2Cl6 with surface –NH2 species results in the formation of a surface –NH species, and a bound –SixCl2x-1 surface species.15 The presence of –SixCl2x-1 surface species following chlorosilane chemisorption has been verified through presence of –SiClx (x = 1, 2, or 3) stretching modes, which absorb at ~620 cm-1.15,35-36 However, these modes are not directly observable in our infrared setup because of the cutoff frequency of ~700 cm-1. In the subsequent CH3NH2 partial cycle, CH3NH2 reacts with surface –SiClx (x = 1, 2, or 3) species to form Si2N–CH3 species via the formation of Si–N linkages, which manifests as an increase in the –CHx (x = 1, 2, or 3) stretching mode at ~2900 cm-1, and the –CH3 asymmetric deformation at ~1450 cm-1.37-38 The presence of a C–H stretching feature at ~2820 cm-1 is indicative that the –CHx (x = 1, 2, or 3)species are backbonded to N atoms,38-39 confirming the presence of amine ligands on the surface following CH3NH2 chemisorption. The reaction of CH3NH2 with surface –SiClx (x = 1, 2, or 3)produces HCl as the gas-phase reaction product, which results in the formation of 13 ACS Paragon Plus Environment


Page 14 of 28

CH3NH3Cl on the KBr windows of the chamber (see Figure 3). The formation of surface amine groups after the CH3NH2 partial cycle makes the three-step ALD process similar to an aminosilane and N2 plasma ALD process.18 When an aminosilane precursor reacts with the SiNx surface during ALD, one or more of the amide ligands are exchanged with the surface resulting in the formation of –SiHx(NRR')y-x (x ≤ 3, x + y = 3) groups, and NHRR' as the reaction product that is released into the gas phase. The primary difference between the aminosilane and N2 plasma ALD process and the three-step process is that no –SiHx (x = 1, 2, or 3) surface species are created following CH3NH2 chemisorption. Thus, for this three-step process, the aminosilane precursor surface reaction is essentially separated into two steps: the formation of –SixCl2x-1 surface species during the Si2Cl6 step followed by conversion to Si2N–CH3 species during the CH3NH2 step. While this two-step pathway to create amide groups on the surface may appear unnecessary, we will show later that this reaction pathway results in a much higher GPC and deposition conformality compared to aminosilane and N2 plasma processes. In the N2 plasma step, the decrease in absorbance in the ~1450 and ~2900 cm-1 regions indicates that Si2N–CH3 surface species are removed, and the small increase in absorbance in the ~1550 and ~3300 cm-1 regions indicates that surface –NHx (x = 1, 2) species are restored, even though there is no H in the feed gas. The most likely mechanism for this is that H liberated into the N2 plasma during the removal of surface –CH3 groups is recycled back to the surface as atomic H and/or ̶ NHx (x = 1, 2) radicals. The cyclic continuation of the ALD process is enabled by the formation of –NH2 surface species by the plasma, allowing for subsequent Si2Cl6 chemisorption. The rationale behind the assignment of the feature at ~2180 cm-1 in Figure 3 to Si–N=C=N–Si is described here and is based off of an analysis of the 2000-2300 cm-1 region of the partial cycle infrared spectra seen in Figure 4. The feature at ~2180 cm-1 in C-containing


Page 15 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


SiNx films has previously been assigned to –C≡N and –SiHx (x = 1, 2, or 3) stretching vibrations,18,40-42 but the presence of these species is unlikely in SiNx films deposited using the three-step ALD process for the following reasons. Stretching vibrations for –C≡N are generally reported at frequencies >2200 cm-1, which are not observed in the infrared spectra in Figure 3.43 Surface

–SiHx (x = 1, 2, or 3) species tend to absorb at ~2180 cm-1 in amorphous

hydrogenated SiNx films, which is similar to the feature observed in the infrared spectra in Figure 3. However, this peak assignment is unlikely in these spectra due to the following reasons. The lack of silyl groups in either Si2Cl6 or CH3NH2 thermal cycles means that the formation of surface –SiHx (x = 1, 2, or 3)species during these two precursor steps is highly unlikely, even though in Figure 4 infrared absorbance increases are observed in this region. The presence of –SiH bonds in the Si2Cl6 precursor is not expected as an impurity. When Si2Cl6 from the same vendor and of the same purity was used for the deposition of SiNx films using Si2Cl6 and NH3 plasma at 400 °C, the Si2Cl6 half-cycle infrared spectra did not show any changes in the –SiHx (x = 1, 2, or 3) stretching region at ~2100 cm-1.10 In any case, the increase in absorbance in the Si2Cl6 and CH3NH2 partial cycles occur at ~2120 and ~2220 cm-1, respectively, which are unlikely to be related to –SiHx (x = 1, 2, or 3) stretching modes, because they are outside of the acceptable range of the –SiHx (x = 1, 2, or 3) stretching vibrational frequency range in hydrogenated amorphous SiNx films.44 Therefore, –SiHx (x = 1, 2, or 3) species would most likely be created during the N2 plasma step through recycling of surface H from –CH3 groups to form surface –SiHx (x = 1, 2, or 3)groups. However, more detailed analysis of the N2 plasma partial-cycle shows that this possibility can also be mostly eliminated.



Page 16 of 28

Figure 5: Spectral deconvolution of the 1950-2350 cm-1 infrared region for the N2 plasma partial cycle using Gaussian curves. To elucidate what occurs during the N2 plasma step, the region between 1950 and 2350 cm-1 from the partial cycle spectra in Figure 4 was analyzed through spectral deconvolution using Gaussian curves as shown in Figure 5. The deconvolution of the N2 plasma partial cycle in Figure 5 shows three distinct features: a small increase at ~2050 cm-1, a prominent increase at ~2120 cm-1, and an increase at ~2180 cm-1. The frequencies of the peaks at 2050 and 2120 cm-1 are too low to be related to –SiHx (x = 1, 2, or 3) stretching modes in hydrogenated amorphous SiNx films, which typically appear at frequencies >2150 cm-1 and therefore, are assigned here to the ketene imine (>C=C=N–) and carbodiimide (–N=C=NH) stretching modes.44-47 The feature at ~2180 cm-1, is assigned to the Si–N=C=N–Si asymmetric stretching vibration,11,32-33 which we previously reported in C-containing SiNx films deposited using alternating exposures of Si2Cl6 and CH3NH2 plasma.11 As Si–N=C=N–Si asymmetric stretching vibrations absorb very strongly, even relatively small levels of C incorporation could produce a strong infrared peak at ~2180 cm-1.48-50 In a separate set of experiments (see supplemental Figure S1) we show that for a –SiHx (x = 1, 2, or 3) terminated surface, N2 plasmas are more likely to remove –SiHx (x = 1, 2, or 3) 16 ACS Paragon Plus Environment

Page 17 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


surface species than to create them. It is therefore likely that in this ALD process, the N2 plasma step would also preferentially remove –SiHx (x = 1, 2, or 3) surface species, rather than create them on the growing film. We can then assign the feature at ~2180 cm-1 primarily to the Si– N=C=N–Si stretching vibration. That being said, due to the presence of background H in the reactor we are unable to completely eliminate the possibility that the infrared feature at ~2180 cm-1 does not contain contributions from –SiHx (x = 1, 2, or 3) stretching vibrations. The Si– N=C=N–Si and –SiHx (x = 1, 2, or 3) species are incorporated into the film because C and H present on the surface as amine ligands following the CH3NH2 step is recycled from the surface during the N2 plasma discharge and redeposited as carbodiimide or possibly –SiHx (x = 1, 2, or 3) species.18-19,51 Lastly, we have also verified the spectral deconvolution in Figure 5, where the integrated absorbance of the 2180 cm-1 feature matches the integrated absorbance per ALD cycle at 2180 cm-1 in the cumulative cycle ALD spectra of the SiNx film shown in Figure 3.

Figure 6: Schematic of the proposed surface reaction mechanism for the three-step Si2Cl6, CH3NH2, and N2 plasma SiNx ALD process.



Page 18 of 28

The changes observed in the 2000 ̶ 2300 cm-1 infrared region in the Si2Cl6 and CH3NH2 spectra in Figure 4 have previously been attributed to incorporation and/or interconversion of carbodiimide (–N=C=NH) and cyanamide (–NH–C≡N) species,11,47 and appear prominently due to their high absorption cross-sections.48-49 However, as we show later, since the amount of C incorporated into the SiNx films is low, changes in absorbance in this region are not representative of the primary film growth mechanism.47 Change in absorbance in the 2000–2200 cm-1 was also observed by Kessels and coworkers during N2-plasma-assisted ALD of SiNx from bis-tert-butylaminosilane, and assigned to H-C-N complexes.18 Based on our interpretation of the partial cycle infrared spectra recorded during the Si2Cl6, CH3NH2, and N2 plasma, the proposed growth mechanism during SiNx ALD is shown in Figure 6. Growth per Cycle, Film Properties, and Deposition Conformality

Figure 7: Film thickness of a SiNx film deposited using 10 complete ALD cycles measured using in situ ellipsometry. A zoomed-in view of a single ALD cycle is shown in the inset. The thickness, as measured by in situ ellipsometry, of a SiNx film deposited using 10 complete Si2Cl6, CH3NH2, and N2 plasma ALD cycles at 400 ºC is shown in Figure 7. The GPC for this three-step process was 0.9 Å. While we had earlier determined from the partial cycle


Page 19 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


infrared spectra (see Fig. 4) that the reaction pathway for the three-step process is similar to an aminosilane and N2-plasma-based SiNx ALD process, the GPC at 0.9 Å is substantially higher (see Table 1),13-14,18,51 which is a distinct advantage over aminosilane/N2 plasma ALD processes. In fact, the GPC is very similar to that of chlorosilane/NH3 plasma ALD processes. The refractive index (see Table 1) of the SiNx films deposited using the three-step ALD process was ~1.95 at a wavelength of 633 nm, as measured by ex situ spectroscopic ellipsometry, which is a lower than that of SiNx films deposited using aminosilane and N2-plasma-based ALD processes (~2.0), 13,18-19,51 but greater than that of films deposited using chlorosilane and NH3-plasma-based ALD processes (~1.85).10,12 This is consistent with XRR measurements, which show that the SiNx film has a density of 2.91 g/cm3. This density is higher than the typical density of SiNx films deposited using chlorosilanes and NH3 plasmas or CH3NH2 plasmas, but lower than that of films deposited using aminosilanes and N2 plasmas (see Table 1).1 Table 1: GPC and film properties for the three-step SiNx ALD process at 400 °C compared to typical values for the chlorosilane and NH3 plasma and aminosilane and N2 plasma SiNx ALD processes. GPC and Film Properties

Chlorosilanes/Nand H-containing Plasmas >1

Three-Step Process

Aminosilanes/N2 Plasmas

~0.9

A Three-Step Atomic Layer Deposition Process for SiNx Using Si2Cl6

Recommend Documents