Less Energetic Routes for the Production of SiOx Films from Tris(dimethylamino)silane by Plasma Enhanced Atomic Layer Deposition

Abstract

SiO_x films, frequently derived from amino silane precursors, have found several applications with high added value. Although frequently used, the deposition of coatings from Tris(dimethyl amino) silane (TDMAS) has been reported to demand considerable amounts of energy, mainly due to the difficulty of oxidizing such compounds. As is well known, Plasma-enhanced atomic layer deposition (PEALD) is able to improve the oxidation efficiency, even under low processing temperatures. Owing to this, PEALD can be considered a very promising technique for the deposition of SiOx coatings. In this work, the deposition of silicon oxide films using TDMAS at 150 °C has been investigated. The effect of the plasma oxidation time (6 to 18 s) and atmosphere composition (pure O₂ or O₂ + Ar) on the chemical structure, elemental composition, and chemical bonding state of the films has also been evaluated. Increasing the plasma oxidation time in pure O₂ resulted in a larger proportion of retained C (Si-CH₃), whereas N was preserved in the structure (Si-N). On the other hand, the formation of SiO_x films from TDMAS is favored in shorter oxidation times and O₂ + Ar plasmas.

1. Introduction

Silicon oxide is a versatile compound that has been used in microelectronic devices and as a protective film against the corrosion of metallic surfaces [1]. Thus, there is a constant search for economically viable deposition routes resulting in high-quality SiOx films. In this scenario, the atomic layer deposition (ALD) technique is widely used, as it allows for the deposition of dense films with nanometric thicknesses. However, the routes used for the growth of SiO_x films by ALD are thermal routes, requiring high temperatures [2]. Burton et al. [1] studied some precursors using amino ligands, employing the ALD process at high temperatures. They found that TDMAS is a good precursor, but only at temperatures above 450 °C and using H₂O₂ as an oxidant agent. In other conditions, the Si-H bond remained the at the film surface.

Kamiyama et al. [3] used BDMAS (bis(dimethylamino)silane) and TDMAS (tris(dimethylamino)silane) as precursors in the ALD process at 275 °C, applying ozone as the oxidant. The film deposited from BDMAS showed fewer impurities than TDMAS, and good results have been achieved only after annealing the samples at 1000 °C.

Since glow discharge plasmas contain large amounts of oxidative species, Plasma Enhanced-ALD (PEALD) stands out as a very promising alternative, as it allows for the removal of organics at lower temperatures in the oxidation step [4] in comparison to the methods conventionally used in ALD.

Precursors based on aminosilanes (SiN_xH_y), such as BDMAS and TDMAS, are extremely reactive and promote the first stage of the reaction in a self-limiting manner, without the need for high temperatures [5,6]. Despite this advantage, only a few authors have investigated the use of TDMAS as a precursor in PEALD.

In the work of Ma et al. [7], for instance, the deposition of SiO_x films was reported using TDMAS in a PEALD system. The effect of the process temperature (100–250 °C), TDMAS pulse time (2–8 s), and the plasma oxidation time (2–16 s) on the film’s thickness and stoichiometry was investigated. An elevation in the growth per cycle (GPC) was observed with temperature (up to 175 °C), oxidation time (up to 10 s), and TDMAS pulse time (up to 8 s). However, non-stochiometric films were obtained, presenting an abundance of Si and suggesting an SiO_1,6 or SiO_1,8-like structure.

In a subsequent work, Ma et al. [8], proposed the deposition of an SiN_x/SiO₂ superlattice film of TDMAS (275 °C) by PEALD, using O₂ as an oxidant agent and 10 s of plasma oxidation time. Following deposition, some of the films were annealed at high temperatures (up to 1100 °C). The chemical characterizations of the annealed superlattices revealed Si-H and Si-CHN bonds, indicating that the precursor’s residual groups remained in the structure. In other recent works found in the literature, the authors demonstrated the effect of the substrate (metal-oxide) electronegativity [5] and of the type of native oxide interlayer (n-GaN/native oxide or p-Si/SiO₂) [9] on the growth per cycle and on the deposition kinetics of SiO_x films prepared from TDMAS by PEALD.

Even though it is an environmentally friendly, cost-effective, and accessible compound, the low attention devoted to the PEALD of TMDAS is a consequence of the fact that the third dimethyl group in the precursor molecule may hinder the formation of pure SiO_x films, due to the high steric hindrance and consequently, the high energy required to break the bonds involving such carbon-containing groups [1].

Li et al. [10], Jeong et al. [11], and Fang et al. [12] used density functional theory calculations to study the initial reaction and the first principles of reaction of silicon precursors with different numbers of amino ligands. They demonstrated that when the number of ligands grows, the adsorption energy increases, and the precursor’s dissociation stops at the second dimethylaminosilylenyl, energetically forbidding further dissociation. They suggested that precursors with a lower content of dimethylamino groups, such as Bis(dimethylamino)silane (BDMAS) and Tris(dimethylamino)silane (TDMAS), remain the best precursors for this propose.

In this context, this work aims to evaluate the possibility of removing the third methyl group from the structure of films prepared from TDMAS, adjusting the parameters of the PEALD process and thus developing a low-energy route to produce SiO_x films. More specifically, the effect of the plasma oxidation time and composition on the chemical structure of the samples has been evaluated.

2. Materials and Methods

The films were deposited on aluminum and carbon steel substrates of approximately 20 mm × 10 mm in size. The substrates were mirror-polished using 1200 grit sandpaper and subsequently, with a water-based diamond paste. After polishing, the samples were cleaned in an ultrasonic bath with isopropyl alcohol for 480 s and dried with a hot air gun. Despite the fact that 17 different conditions were tested, only the effects of plasma oxidation time (6 to 18 s) and the plasma chemical composition (O₂ or O₂ + Ar), which showed a stronger influence on the properties of the resulting films, are discussed here. The depositions were carried out in an Oxford atomic layer deposition device with inductively coupled plasma—OpAL. For the first set of experiments, in which the effect of the oxidation time on the film properties was evaluated, the deposition step was carried out using a pulse (20 ms) of tris(dimethylamino)silane and 30 sccm of Ar. The pressure in this step was 170–190 mTorr. After the deposition step, the system was purged by flowing 30 sccm of argon for 8 s. The oxidation process was performed in O₂ (100 sccm) plasmas (13.56 MHz, 250 W, 260–270 mTorr), with exposure times ranging from 6 to 18 s. The temperature of the table in the reactor was maintained at 150 °C for the processes. For the second batch of experiments, in which the effect of the oxidative plasma composition on the film properties was evaluated, the deposition process was maintained for 20 ms of tris(dimethylamino)silane with Ar as a carrier (30 sccm) at 170–190 mTorr. The purge process was conducted with Ar flow (30 sccm) for 4 s. The oxidation process was performed in plasmas (13.56 MHz, 250 W, 260–270 mTorr) of O₂ (90 sccm) or O₂ + ArAR (90 sccm + 20 sccm), with the lowest exposure time used in the first set of experiments (6 s). Despite the fact that same oxidation time of 6 s was used in the samples prepared in both the first and second batch of experiments, they cannot be directly compared, since the purge time (8 and 4 s) and O₂ flow (100 and 90 sccm) used differs in each batch. Consistent with the proposal of testing processes with lower temperatures, all the steps of these sets of experiments were conducted at 150 °C. The complete deposition, purging, and oxidation parameters used in the different experiments are summarized in

Table 1. Parameters for the preparation of samples for different conditions.

Condition	TDMAS Deposition	Purge	Oxidation Process (260–270 mTorr, 250 W, 13.6 MHz)
			Composition/Time	Flow
Plasma oxidation time	20 ms, 30 sccm Ar, 170–190 mTorr	8 s, 30 sccm Ar	O2, 6 s	100 sccm of pure O2
	20 ms, 30 sccm Ar, 170–190 mTorr	8 s, 30 sccm Ar	O2, 12 s	100 sccm of pure O2
	20 ms, 30 sccm Ar, 170–190 mTorr	8 s, 30 sccm Ar	O2, 18 s	100 sccm of pure O2
Plasma composition	20 ms, 30 sccm Ar, 170–190 mTorr	4 s, 30 sccm Ar	O2+Ar 6 s	90 sccm of O2 and 20 sccm of Ar
	20 ms, 30 sccm Ar, 170–190 mTorr	4 s, 30 sccm Ar	O2, 6 s	90 sccm of O2

.

The chemical structure was assessed by Fourier transform infrared spectroscopy (FTIR) on a Jasco spectrometer (Jasco Corp., Tokyo, Japan) in the range of 4000 to 400 cm⁻¹ with a resolution of 4 cm⁻¹, with 128 scans, in transmittance mode at room temperature. The bonding state of the samples was investigated by X-ray photoelectron spectroscopy (XPS) using a model K-Alpha device from Thermo Scientific at the National Nanotechnology Laboratory (LNNano) in Campinas, SP. A monochromatic Al Kα radiation source was employed. The analyses were conducted on films deposited on aluminum using a pass energy of 200,00 eV and a scan number of 10, while the high-definition spectra were acquired with a pass energy of 100,00 eV, a scan number of 10, and an energy step of 0.05 eV in three different positions. The spectra were analyzed using Advantage 5.9925 software. Also, to determine whether the presence of C on the surface of the samples resulted from the oxidation process or from exposition to atmosphere, ion bean sputtering was used in XPS analysis in a cluster mode with 8000 eV of ion energy and 30 s of etching time.

The surface morphology of the samples was inspected by Scanning Electron Microscopy (JSM-6010LA, JEOL Ltd., Peabody, MA, USA) equipment, operating at voltages of up to 15 kV. The micrographs were acquired at magnifications of 3000× and 5000× using a voltage of 10 or 12 kV, with a spot size of 50 µm. The semi-quantitative chemical composition of the samples was acquired by X-ray Energy Dispersive Spectroscopy (EDS) (JEOL Ltd., Peabody, MA, USA) with the detector coupled to the SEM equipment, operating at 12 kV.

The chemical composition of the samples was analyzed using Rutherford backscattering spectrometry (RBS). An He⁺ beam from a Pelletron-tandem accelerator produced by National Electrostatic Corporation (NEC), model 5SDH, was directed onto the samples in a vacuum chamber at 7° incidence and 170°, 120°, and 90° scattering angles. The energy of the ions was 2.2 MeV, and the beam charge was 20 µC. Chemical composition and elemental depth profiles were determined by computer simulations using RUMP and SIMNRA software (version 7.03).

3. Results and Discussion

The effect of different plasma oxidation times on the morphology of the surface was evaluated by scanning electron microscopy (SEM). The corresponding micrographs are presented in Figure 1a–c. The micrograph of the uncoated substrate (carbon steel) is also shown (Figure 1d) for reference. It is noted that the substrate exhibits surface defects, including scratches and small pits, originating from the applied finishing process. Even after the exposure to the PEALD process, the scratches are still evident on the treated surfaces due to the mold effect. However, their morphologies are slightly changed by the presence of a flat conformational layer. In this coating, the content of defects and the porosity are reduced, pointing to a dense structure.

The effect of oxidation time, t_ox, on the chemical structure of the films can be determined with the help of Figure 2a, which presents the infrared transmittance spectra of the films exposed to the oxidation plasma for 6, 12, and 18 s. The spectra can be directly compared to each other once the number of cycles of deposition is kept the same; when this was achieved, all the samples presented roughly the same thickness. In general, the infrared spectra of the samples prepared with the highest t_ox (12 and 18 s) presented wider peaks. In the literature, the peak at 3379 cm⁻¹ is assigned to the stretching of N-H bonds [13], which is in good agreement with the appearance of the contribution at 1646 cm⁻¹, also related to N-H groups [11]. However, O-H stretching vibrations may also contribute to this band, since they are reported around 3400 cm⁻¹ [14] in the literature. This attribution is also consistent with the appearance of a band related to O-H stretching vibrations around 3644 cm⁻¹. Free water molecules also produce a contribution around 3700 cm⁻¹ [15], at which low intensity peaks are detected. In the spectrum of the film produced with 18 s of oxidation time, the band at 3379 cm⁻¹ is overlapped with the band ascribed to ν-O-H (3644 cm⁻¹) [13]. Another very prominent band around 1180 cm⁻¹ can be deconvoluted in two peaks, as depicted in Figure 2b, c, for the bands in the spectra of samples oxidized for 12 and 18 s, respectively. Such contributions, centered at 1196 and 1100 cm⁻¹ (t_ox = 12 s) and on 1196 and 1099 cm⁻¹ (t_ox = 18 s), can be attributed to SiO (1230 and 1080 cm⁻¹) [16]; Si-N (1125, 1190 and 1220 cm⁻¹) [16,17]; and Si-CHx (1230 cm⁻¹) [18].

Different features can be observed in the spectrum of the film oxidized for 6 s. The drop in peak intensity at 3379 cm⁻¹, the disappearance of the peak at 1646 cm⁻¹, and the separation of the bands initially centered around 1178 cm⁻¹ indicate that the chemical neighborhood of the Si-O functionals becomes more restricted. The reduction in the concentration of Si-N, C-H, and N-H contaminants makes the peaks sharper, indicating that the formation of SiOx films is favored by the decrease in plasma oxidation time. This hypothesis is supported by the presence of peaks at 430 to 470 cm⁻¹, related to Si-O bonds [19]. Although the most intense peaks of the spectrum indicate the formation of an SiO_x film, the presence of smaller intensity contributions at 450 cm⁻¹ (Si-O-Si) [17], 500 cm⁻¹ (Si-N) [20,21] and 3379 cm⁻¹ (N-H) [13] are consistent with the presence of contaminants. Interestingly, the increase in t_ox does not favor the removal of the third dimethylamino group, but rather the reincorporation of residual groups from the plasma oxidation atmosphere.

The atomic ratios of Si/C, Si/O, and O/C, obtained from EDS results are shown in Figure 2d. According to Renlund et al. [22], the Si/C ratio in silicon suboxides tends to be higher as the C contamination decreases. A decrease in the Si/C ratio is observed with the increase in t_ox to 12 and 18 s, suggesting an increment in the proportion of C retained in the film. A similar behavior is observed for the O/C ratio, whereas the Si/O ratio is roughly independent of t_ox. These results also corroborate the hypothesis that the shortest oxidation time promotes the removal of organic moieties, without favoring the re-incorporation of residual groups.

Figure 3 shows the high-resolution spectra of the Si 2p and C 1s, obtained by XPS, for the samples investigated here. Carbon, oxygen, and silicon, elements present in the precursor molecule, were observed in all the samples, while N was detected only in the samples oxidized for 12 s and 18 s. For fitting the Si 2p peaks of these two last samples, a component related to Si₃N₄ was used instead of Si-CH, since the detection range of this last bond would appear at a higher energy edge. This indicates the absence of Si-CH groups in the films. However, Si-O and Si-O-C groups were included. Consequently, in the adjustments of the C 1s peaks, O-C=O and O-C-O groups were used. The use of the latter corresponds with the contributions of reduced intensities in the infrared spectra of these samples (Figure 2) between 1500–1400 cm⁻¹, attributed to the stretching of C=O and C-O bonds [23]. It is worth mentioning that the ratio between the intensities of the contributions of oxidized groups (O-C=O + O-C-O) in relation to that of non-oxidized groups (C-C) is greater in the XPS spectrum of the film prepared using 6 s of oxidation time, consistent with the greater intensity of the band at 1511 cm⁻¹ (C=O and C-O) in the IRRAS spectrum of the same sample. Thus, these results corroborate the infrared results that suggested the deposition of SiO_x-type films with reduced organic contamination.

The atomic proportions of the elements derived from the XPS spectra are presented in

Table 2. RBS and XPS atomic proportions for films submitted to oxidation plasmas for 6, 12, and 18 s.

Element	6 s		12 s		18 s
	XPS	RBS	XPS	RBS	XPS	RBS
C (%)	23	5	56	23	45	33
Si (%)	26	16	12	24	16	23
O (%)	51	79	27	53	33	44
N (%)	0	-	5	-	6	-

. As it can be noted, the shorter the plasma oxidation time, the smaller the proportions of C (23%) and N (0%) on the film surface. Consistent with the previous results, the proportion of C increases and that of silicon decreases with t_ox larger than 6 s. For the latter, the structure is composed of 26% Si, 51% O, and 23% C, with no detection of a significant amount of N. As C-containing groups were not necessary for the fitting of the Si high resolution peak of this sample, the contribution of C to the elemental composition of this film is assumed to be due to the adsorption of atmospheric C. Indeed, when the sample was plasma-sputtered with Ar ions prior to the acquisition of the XPS spectrum, only 6% of C was detected.

In Table 2, the atomic proportions of C, O, and Si derived from the RBS measurements is also presented. In these cases, N was not detected due the low cross-section of this element to the alpha beam. However, a trend of elevation in the proportion of C with increasing t_ox is promptly detected, thus reinforcing the previous XPS, IRRAS, and EDS findings. The C proportions, derived from RBS inspections are always lower than the detected by XPS, indicating that C is more abundant on the layer surface. This compositional gradient will then influence the atomic proportions obtained by the different methodologies, since they probe different depths.

All these results show that SiO_x-like structures with fewer organic groups are obtained with the shortest oxidation time (6 s) in O₂ plasma, which represents the treatment with the lowest energy cost. It is interesting to mention that the growth in the C content with increasing t_ox is due to recontamination; that is, in all cases, the third dimethylamino group is released. Thus, adding extra time to the oxidation process is not beneficial, but instead, detrimental to the formation of SiO_x. In the resulting structure, IRRAS showed that hydrogen atoms share the oxygen of the siloxane network, forming hydroxyls (O-H), which are normally observed in silicon oxides. It can finally be inferred, by comparing the results obtained from different methodologies, that organic groups (C-H, C-C and C-O) are mainly concentrated on the film surface, suggesting that they are originated by atmospheric contamination.

In the second set of experiments, the effect of the chemical composition of the oxidation plasma on the chemical structure of the films was investigated. To do this, the deposition step was conducted using the same conditions employed in the previous experiments (TDMAS + Ar, 150 °C, 260–270 mTorr), but purging the system for only 4 s, which is the reason why the results of both samples prepared with 6 s of t_ox cannot be compared between the two batches. The oxidation process was conducted in atmospheres containing pure O₂ (100%, 90 sccm) or a mixture of O₂ + Ar (82% O₂, 90 sccm + 18% Ar, 20 sccm), preserving the other parameters (150 °C, 260–270 mTorr, 6 s) used in the previous section. Figure 4a shows the infrared spectra of the samples oxidized in plasmas of O₂, with and without the addition of Ar.

The spectrum of the film produced using only oxygen in the oxidation process reveals the presence of peaks characteristic of the Si-O bond at 1219, 1146, 1038, and 944, and a doublet at 825–755 cm⁻¹. In addition, absorptions ascribed to the Si-CH bond (1300–1500 cm⁻¹) and the N-H and/or O-H bonds (1600 cm⁻¹) [13,14,16,24] are also observed. Furthermore, a broad band observed in the range of 3713 to 2948 cm⁻¹ suggests the presence of peaks assigned to vibrations of N-H (3400 cm⁻¹), O-H and H₂O (3700 cm⁻¹), and C-H (2957 cm⁻¹). It is interesting to note the substantial difference between the infrared spectrum of this sample and that of the sample prepared with the same oxidation time (150 °C, 260–270 mTorr, 6 s), but with 8 s of purging time (Figure 2d). The higher contribution of C-containing groups in the spectrum of the film prepared with 4 s of purging time shows that the deposition and oxidation kinetics are also sensitive to this parameter.

Different features are observed in the spectrum of the film grown with the addition of argon in the oxidation plasma. Whereas the peak at 1390 cm⁻¹, attributed to methylsilyl groups, practically disappears, the one at 1303 cm⁻¹ becomes only a shoulder of a prominent band at 1232 cm⁻¹. These modifications indicate a reduction in the contribution of the methylsilyl groups, along with an enrichment of the density of the SiO_x connections. Furthermore, the broad band in the highest wavenumber region, ascribed to various contributions (N-H, C-H and H₂O), becomes thinner and more prominent, and there is a rise in the intensities of the bands related to Si-O (1232, 985, 862, and 749 cm⁻¹). Taken together, these results suggest that the addition of Ar increases molecular fragmentation by the plasma, favoring the formation of volatile carbon-containing species (CO and CO₂, for instance). In this way, carbon can be pumped out of the discharge by the vacuum system. The addition of Ar to the plasma, within a certain proportion range, tends to increase the average energy and concentration of activated species [25]. Consequently, the oxygen reactivity may also increase, making the etching of deposited carbon groups more effective. This hypothesis is corroborated by the atomic ratio results obtained from the EDS data presented in Figure 4b. The enhancement of the Si/C and O/C ratios with the addition of Ar to the oxidation step is clearly revealed, reinforcing the proposal to reduce the proportion of C proportion.

Figure 5 presents the high resolution XPS spectra of Si 2p and C 1s of films deposited with and without the addition of argon to the oxidation plasma. The Si 2p peaks were fitted with Si-O and Si(OH) components, when the oxidation process was conducted in Ar-containing plasmas, and with Si-O and Si₃N₄, when pure oxygen plasmas was used. This means that, for films oxidized in pure oxygen plasma, radicals incorporated in the SiO_x structure are saturated by N. On the other hand, when oxidation is conducted in Ar-containing plasmas, the radicals are passivated by the OH groups. Considering the high-resolution C peak, it is observed that C is bonded to O and to another C atom. In good agreement with previous results, no N was detected in the film that was oxidized in the Ar-containing process. For the latter, the Si/O ratios were 0.4 (EDS) and 0.52 (XPS), which is near the stoichiometric ratio of the SiO₂ (0.5).

The atomic proportions of the elements derived from the XPS and RBS analyses are presented in

Table 3. RBS and XPS atomic proportions for films submitted to different plasma compositions.

Element	With Ar		Without Ar
	XPS	RBS	XPS	RBS
C (%)	15	1	66	48
Si (%)	29	20	7	14
O (%)	56	79	20	38
N (%)	0	-	7	-

. The C content on the surface of the films (XPS) is always higher than that in the bulk (RBS). Besides affecting the proportion of the elements detected in the different analytical methodologies, this compositional gradient suggests that a significant fraction of the C detected may be derived from atmospheric contamination. Considering the effect of the plasma composition on the chemistry of the films, a substantial reduction in the C content is observed as Ar is incorporated into the process, corroborating the previous results of IRRAS and EDS. Such results show that Ar has a synergistic effect in the oxidation step. It is worth verifying that the C content detected in the sample produced under the best oxidation condition evaluated here was only 15%, but when the sample was plasma sputtered with Ar ions, this value decrease to less than 10%. This phenomenon is the same as that observed in the previous experiments, which were conducted under different conditions. This concurrence of results suggests that, in fact, the carbon detected in both cases results from post-plasma reactions when the sample is exposed to atmosphere and not from the oxidation process. Finally, the third methylamine connection of the precursor was not identified in the chemical structure of the film when this oxidation route was used.

4. Conclusions

This work aimed to evaluate new routes for the complete removal of methylsilyl groups of monolayers deposited from PEALD of TDMAS. The oxidation time in pure oxygen plasmas was decisive for the quality of the resulting SiO_x structure. The lowest oxidation time, 6 s, was enough for removal of the three dimethylamino groups, while preventing the incorporation of residual groups from the oxidative atmosphere. The traces of C observed using the lowest oxidation time were ascribed to atmospheric contamination in the post-deposition reactions. The composition of the gas feed of the oxidative plasma was also relevant for the chemical structure of the film. Argon has a synergistic effect on the O₂ plasma, enhancing the organic ablation and inhibiting N inclusions, favoring the complete removal of the dimethylamino groups. Thus, whereas the reduction of oxidation time inhibited the inclusion of contaminants, the incorporation of Ar accelerated the removal of organics, allowing for the establishment of new routes for the creation of SiO_x networks from PEALD of TDMAS. Nevertheless, the combination of both conditions (a shorter plasma oxidation time and the presence of Ar), seems to be a still better condition for the deposition of this type of film in future studies. Finally, the methodology proposed here may provide thicker SiO_x films by increasing the number of cycles, allowing for their deposition on different types of substrates (polymeric, ceramic, metallic, etc.).