Excellent conformality of atmospheric-pressure plasma-enhanced spatial atomic layer deposition with subsecond plasma exposure times

Atmospheric-pressure spatial atomic layer deposition (s-ALD) is an emerging technology for depositing high-quality thin films for low-cost high-volume applications.¹ Using a plasma as co-reactant has the additional benefit of a more comprehensive materials selection and enlarged set of compatible precursors,^2,3 as well as lower deposition temperatures,⁴ and in some cases improved film properties.^4,5 Although low-pressure temporal ALD is known to be able to deposit highly conformal films, potential challenges arise when using atmospheric-pressure plasma-enhanced s-ALD (PE-s-ALD). First, the partial pressure of reactive plasma species can quickly drop inside high-aspect-ratio structures due to surface recombination while the effect of using high reactor pressures on recombination probabilities is unknown. Second, high pressures can lead to gas-phase collisions and low diffusion coefficients causing reactive species to diffuse less deep than for low-pressure ALD.⁶ 

Continuum models, based on pioneering work by Gordon et al.,⁷ have been used to gain a theoretical understanding of the conformality of thermal ALD.^8–14 By adding a non-saturating reactant loss mechanism to describe recombination of plasma radicals, such models can be applied to PE-ALD,⁹ and can even be used to directly determine the recombination probability $r$⁠.¹⁵ The recombination probability is highly dependent on the surface material. Arts et al. found $r$-values of oxygen radicals during the deposition of oxides to vary from $∼ 10 − 5$ for SiO₂ and TiO₂ to $∼ 10 − 3$– $10 − 2$ for Al₂O₃ and HfO₂.^15,16 So far, only the conformality of low-pressure PE-ALD processes has been studied.^17,18 The step coverage of atmospheric-pressure s-ALD has only been studied for thermal processes.^6,19 At high reactor pressure, the mean free path of oxygen radicals is shorter than at low pressure, so more gas-phase and wall collisions can be expected for atmospheric-pressure PE-s-ALD as compared to low-pressure ALD, potentially leading to lower penetration depths. Additionally, $r$ has been found to be pressure dependent, typically increasing with pressure.^20,21 Yet, Arts et al. have shown reduced $r$ with increasing pressure in the 12–130 mTorr range.¹⁶ Clearly, several questions remain concerning the conformality of PE-ALD.

In this Letter, the conformality of PE-s-ALD deposited SiO₂ and TiO₂ is systematically studied in 3D structures. We demonstrate large penetration depths for subsecond plasma exposure times and we extract $r$ of oxygen radicals during these processes at atmospheric pressure. Moreover, we demonstrate that the plasma exposure times required for achieving saturation in structures with a wide range of aspect ratios are lower than for low-pressure PE-ALD.

The penetration depths of the deposited films in high-aspect-ratio structures were studied by employing the lateral-high-aspect-ratio (LHAR) method enabled by PillarHall LHAR test chips (Chipmetrics Ltd).²² These Si chips of 15 × 15 mm² contain LHAR trenches of various lengths. The type of PillarHall substrate used in this work is the LHAR4 chip, with 500 nm high trenches varying in length from 1 up to 5024 μm, resulting in aspect ratios up to 10 000. These high aspect ratios ensure that the depositions are in the recombination-limited regime. One sidewall of the horizontal trenches consists of Si membranes that can be removed with adhesive tape after depositions to expose the other sidewall of the trenches and allow for thickness and material analysis. The penetration depths of SiO₂ and TiO₂ films were determined for various plasma exposure times and used to determine $r$ of oxygen radicals using the method developed by Arts et al.¹⁵ To use this method at atmospheric pressure, some alterations to the model—related to the diffusion of the oxygen radicals—were required. The diffusion is influenced by the geometry of the structures (Fig. S1 in the supplementary material shows the influence of trench height on the diffusion coefficient at these deposition conditions) and the deposition conditions. In this work, the depositions were performed at 100 °C and atmospheric pressure (⁠ $10 5$ Pa), using a 2% O₂ in N₂ plasma as a co-reactant.

The deposition conditions and trench structures used in this work give a specific scattering length of oxygen $λ 0 , O=2× 10 − 7$ m [Eq. (1)] and result in $Kn=0.4$⁠, corresponding to the transitional diffusion regime.

Here, $a$ denotes a term dependent on the diffusion regime of the system. For Knudsen diffusion, $a$ equals $¾$⁠.¹⁵ In the case of molecular diffusion, $a$ equals $v th h 2 D i j$⁠, where the thermal velocity $v th$ is calculated using $v th= 8 k B T m π$⁠. In the intermediate case, such as in this work, the use of $D eff$ is required, resulting in the substitution of $a$ with $3 D i j + 2 v th h 4 D i j$⁠.

Thin films were deposited using an atmospheric-pressure s-ALD tool with a rotating substrate table (Fig. 1) described in detail by Poodt et al.²⁶ The deposition head was equipped with a precursor inlet and a dielectric-barrier discharge (DBD) plasma source (SparkNano B.V.) positioned opposite of each other.²⁷ Data will be presented for LHAR trench openings that were oriented perpendicular to the precursor and plasma inlets, although it was observed that the trench orientation does not affect the deposition (see Fig. S4 in the supplementary material). SiO₂ and TiO₂ were deposited using precursors bis(diethylamino)silane (BDEAS, SiH₂(NEt₂)₂, bubbler temperature = 50 °C), and tetrakis(dimethylamido)titanium (TDMAT, Ti(NMe₂)₄ bubbler temperature = 40 °C), respectively. A plasma was used with a 10 slm flow of N₂ with 2% O₂ by applying 50 V. For SiO₂ and TiO₂, 200 and 420 cycles were deposited, respectively. Depositions took place at 100 °C and $1× 10 5$ Pa while varying the rotation frequency of the substrate table: 1, 2, and 3 rpm, corresponding to plasma exposure times of 0.73, 0.37, and 0.24 s, respectively. The plasma-source-to-substrate distance was 200 μm. See Fig. S5 in the supplementary material for the influence of the plasma height on the thickness profile of SiO₂.

In addition to the PillarHall test chips, a planar reference wafer was used during each deposition to determine the film composition using x-ray photoelectron spectroscopy (XPS) with a Thermo Fischer Scientific K-alpha x-ray photoelectron spectrometer. The deposited films were nearly stoichiometric, with Si:O ratios of 0.47  $±$ 0.01 for the SiO₂ films and Ti:O ratios of 0.50  $±$ 0.02 for the TiO₂ films. The SiO₂ films were free of carbon, while the TiO₂ films contained trace amounts (<1 at. %). Thickness profiles were measured inside the trenches using a J.A. Woollam Small-Spot size (25 × 40 μm²) RC2 spectroscopic ellipsometer (SE). Measurements were performed over a spectral range of 0.8–5.9 eV. The SiO₂ refractive index was modeled using a Sellmeier-type dispersion, which matched the experimental data from multiple locations (with varying thicknesses). These pre-determined optical constants were fixed to determine the thickness variation of the SiO₂ films. The refractive indices of the SiO₂ films ranged from 1.46 to 1.48 at 633 nm, and the growth per cycle (GPC) was 1.4–1.5 Å/cycle. These values are in line with the earlier work on a similar rotary PE-s-ALD setup using the same precursor and temperature.²⁸ Furthermore, the refractive indices were constant throughout the LHAR structures. For the TiO₂, however, this was not the case, as a lower refractive index was observed inside the trench. Several regions inside the trench were fitted with Kramers–Kronig consistent b-spline functions to create a composition-shifting material file.²⁹ This file was used to determine the thicknesses while accounting for variations in the optical constants throughout the film, within set boundaries (see the supplementary material for more detail). Outside the structures, the refractive indices were 2.35–2.43 at 633 nm and the GPC was 0.60–0.64 Å/cycle. The drop in refractive index has not been reported in the previous work at low pressure, because reflectometry was used to measure the thickness instead of SE, while assuming a constant refractive index.^5,15

Thickness profiles of the films deposited inside LHAR structures are shown in Fig. 2. The thickness of the SiO₂ films remains constant upon entering the trench, up until the so-called deposition fronts, where the thickness quickly decreases to zero. The radical density decreases throughout the trench and at the deposition front the radical density becomes limiting for film growth, before the radicals become depleted. The shape of the SiO₂ thickness curves is similar to what was observed in the earlier work.^15,16 Although the thickness profiles of TiO₂ look mostly similar, they exhibit an increased thickness at the trench entrance. This profile and the changing refractive index throughout the trench will be investigated in more detail in future work. As expected, the penetration depth increases with plasma exposure time for both materials. Even with plasma exposure times as short as 0.73 s, aspect ratios of 73.5 and 218.5 were achieved for SiO₂ and TiO₂, respectively. These aspect ratios are sufficient for most ALD applications. Although depositions with aspect ratios larger than 800 were previously obtained for low-pressure PE-ALD, long plasma exposure times of 120 s were used during those depositions.¹⁵ In contrast, these atmospheric-pressure PE-s-ALD results show remarkably good conformality for subsecond plasma exposure times. For comparison with other research, the aspect ratios obtained for these trenches can be converted to equivalent aspect ratio (EAR), corresponding to cylindrical pores as a reference structure, using $EAR= A R trench/2$⁠.²³ 

In order to further analyze these depositions and to extract the recombination probabilities, $r$⁠, of oxygen radicals during processing, $PD 50 %/h$ was plotted as a function of the logarithm of the plasma exposure time and a linear fit was applied. These results (Fig. 3) show the logarithmic scaling of plasma exposure time with penetration depth expected for recombination-limited depositions. From Eq. (5), it follows that the slope of the linear fit equals $1/ 4.26 r$⁠. The obtained recombination probabilities were $4× 10 − 4$ for SiO₂ and $6× 10 − 5$ for TiO₂. These values are only slightly higher than what was reported by Arts et al. at 50 mTorr (i.e., $9× 10 − 5$ for SiO₂ and $4× 10 − 5$ for TiO₂).¹⁶ Remarkably, the influence of a significant pressure difference on $r$ appears limited, and radical recombination on SiO₂ and especially on TiO₂ is still very small at atmospheric pressure. Note that Arts et al. investigated the pressure dependence over a relatively small pressure range (12–130 mTorr) and found a decreasing trend of $r$ with pressure withing this range.

The exact $t 50 %$ and $t sat$ values vary from reactor to reactor as they depend on many factors (e.g., type of plasma source, plasma power, and plasma-to-substrate distance), yet it is clear that $t sat$ increases more with increasing aspect ratios at atmospheric pressure compared to low pressure. The difference is caused by the difference in diffusion (i.e., shorter mean free path of oxygen radicals leads to more wall collisions, and thus more recombination) as well as the slightly higher recombination probability. Despite this difference, $t sat$ is actually lower at atmospheric pressure for a wide range of aspect ratios, i.e., 0–50 for SiO₂ and 0–300 for TiO₂. This is due to a lower $t 50 %$⁠. Since the atmospheric-pressure plasma generates a higher radical density than the low-pressure plasma, saturation of the plasma exposure step is faster on planar surfaces. This effect is gradually offset by the shorter mean free path and higher $r$ for surfaces with increasing aspect ratios. The short $t sat$ is a clear benefit of using atmospheric pressure PE-s-ALD on 3D substrates.

In conclusion, we have demonstrated that SiO₂ and TiO₂ films deposited using atmospheric-pressure PE-s-ALD with short plasma exposure times can have excellent conformality. Aspect ratios of 74 and 219 for SiO₂ and TiO₂, respectively, were reached using only 0.73 s of plasma exposure, sufficient for coating 3D substrates in many applications. Additionally, we have shown that the established method of determining $r$ of plasma radicals on the deposited material at low pressure can be used at atmospheric pressure by accounting for the difference in the diffusion of the radicals inside the structure. The recombination probability of oxygen radicals was determined to be $4× 10 − 4$ and $6× 10 − 5$ on SiO₂ and TiO₂, respectively. Despite slightly higher $r$ values and lower diffusion coefficients of oxygen radicals compared to low-pressure PE-ALD, atmospheric-pressure PE-s-ALD can yield lower plasma saturation times for aspect ratios up to 50 for SiO₂ and 300 for TiO₂. These findings demonstrate the compatibility of atmospheric pressure PE-s-ALD as high-throughput thin-film deposition method for demanding applications.