ALD2018 Session AS-TuA: Area Selective Deposition II

Area-Selective Atomic Layer Deposition (AS-ALD) is attracting more attention from the semiconductor industry as a possible solution to alignment issues typically faced when scaling down transistor feature sizes. To address this challenge we show a new approach to AS-ALD, where self-limiting thermally-driven atomic layer etching (ALE) is chemically coupled with self-limiting thermal atomic layer deposition (ALD) to yield several nanometers of TiO₂ thin film formation on receptive SiO₂ surfaces at 170°C, with no measurable grown on adjacent areas of hydrogen-terminated silicon (100). When TiO₂ ALD using TiCl₄ and H₂O is done 170°C on non-oxidized hydrogen-terminated silicon, we find an incubation time of ~30 cycles is required before substantial TiO₂ nuclei appear. On SiO₂ surfaces, however, TiO₂ nucleation proceeds rapidly, allowing 10-20 Å of deposition before nucleation on Si-H. Using this inherent nucleation delay combined with a novel self-limiting thermal ALE procedure employing sequential doses of WF₆ and BCl₃, we create a new isothermal “Atomic-Level Process”, where the atomic-scale chemical control of ALD is intimately coupled with that of thermal ALE to build up on a prepared surface, precise nanoscale constructs with pre-selected location and dimension. Using the integrated ALD/ALE sequence, we achieve in excess of 200 TiO₂ ALD cycles, yielding ~ 4 nm of TiO₂ on SiO₂, before visible nuclei form on Si-H, as determined by SEM, ellipsometry and TEM analysis. Process and materials analysis using in-situ QCM and ex-situ AFM and XPS further confirm our findings. To date, extending the ALD/ALE sequence to more than 500 ALD cycles leads to incomplete TiO₂ etch removal from Si-H, ascribed to changes in Si-H during prolonged exposure to deposition and etch species. This demonstrated Atomic Level Process for improved control in selective deposition offers substantial opportunities for integrated area-selective ALD, and provides a viable pathway to explore other Atomic Level Processes for parallel and wafer-scale synthesis of nanoscale and sub-nanoscale constructs.

Ruthenium is a candidate to replace copper in future sub-10 nm interconnects. At these dimensions the resistivity of Ru lines is expected to be lower compared to Cu due to the lower sensitivity to size effects.¹ In addition, it is likely that Ru interconnects won’t require a diffusion barrier, and will show a better electromigration performance.² At feature sizes below 10 nm it will be difficult to align subsequent lithography steps, and the conformality of the deposition method is increasingly important, such that area selective atomic layer deposition (ALD) of ruthenium is of high interest.³

We first report inherent area selective ALD of Ru on H-terminated Si (Si-H) versus SiO₂, using the thermal RuO₄ (ToRuS^TM)/ H₂-gas ALD process.⁴ In situ spectroscopic ellipsometry (SE) on blanket substrates shows that Ru growth initiation occurs from the first cycle on Si-H, while on SiO₂ the growth is delayed, resulting in a substrate selectivity window of ~ 70 cycles (Figure 1, a). Area selective Ru ALD was evaluated using a patterned substrate of 1-10 µm wide Si-H lines separated by 10 µm wide SiO₂ regions, and exposing it to 20 cycles of the RuO₄ / H₂-gas ALD process. Ex situ scanning electron microscopy (SEM) and cross section high resolution transmission electron microscopy (HRTEM) measurements show that a 4.5 nm Ru film could be deposited on the Si-H, with no Ru detected on the SiO₂ (Fig. 2). In vacuo X-ray photoelectron spectroscopy (XPS) experiments showed that exposure of Si-H to a single RuO₄ pulse leads to the oxidation of the Si surface, together with the deposition of RuO₂. On SiO₂ however, the surface is already oxidized, and in vacuo XPS shows that for the same exposure to RuO₄ no Ru is deposited on the surface (Fig. 3). Therefore, we propose that the mechanism behind the inherent substrate selectivity is the oxidation of the Si-H surface by RuO₄. Secondly, we report for a methodology to enhance the nucleation of the RuO₄ / H₂-gas process on oxide substrates. In vacuo XPS and in situ SE experiments show that a single exposure of SiO₂ to trimethylaluminum (TMA) makes the surface reactive towards RuO_4, which allows for Ru growth initiation from the first cycle (Fig. 1, b; Fig. 3). We propose that this is due to the combustion of surface CH₃-groups by RuO₄. As TMA is known to be reactive towards many oxide substrates, this methodology presents a way to achieve Ru metallization of virtually any surface.

¹ S. Dutta et al. IEEE Elec. Dev. Lett. 2017, 38, 949.

² O. V. Pedreira et al. 2017 IEEE IRPS, Monterey, CA, 6B-2.1.

³ P. C. Lemaire et al. J. Chem. Phys.2017, 146, 052811.

⁴ M. M. Minjauw et al. J. Mater. Chem. C. 2015, 3, 132.

Creative surface protecting agents are being used in efforts to explore novel methods for Area Selective Deposition (ASD). These agents include self-assembled monolayers, patterned photoresists, plasma deposited films and others. At the same time, fast nucleation and growth of metal oxide films require creation of fully covered reactive surfaces. Surface treatment ideally will:

• Generate high density or complete surface functionalization
• Eliminate or minimize sub-surface oxidation
• Increase speed and uniformity of nucleation compared to H₂O
• Not react with organic functionality or photoresist on adjacent surfaces

Anhydrous hydrogen peroxide has been largely ignored as a potential novel reactive chemistry. There are several reasons for this. First, there is no precedent in the literature. Second, when delivered H₂O₂ is typically mixed with H₂O, which dominates the reaction. Third, this material has only recently become available in a packaged form that could integrate into selective deposition process equipment.

Hydrogen Peroxide is an attractive chemistry for area selective deposition because of both its oxidation properties and proton transfer properties. The chemistry compares favorably to Ozone (oxidation potential = 2.1V versus 1.8V for H₂O₂). It also has slightly stronger proton transfer than water (water pKa = 7.0 versus 6.5 for H₂O₂). Most critically, H₂O₂ has a very weak O-O bond, with Bond Energy = 36 kcal/mole, suggesting more energetically favorable reactivity at reduced temperatures.

Results from our work show good correlation with selective deposition requirements:

• Nucleation surface density of hydroxyl groups (-OH) is 3-5 times greater using anhydrous hydrogen peroxide versus water on metal surfaces
• No sub-surface oxidation on Si surfaces using anhydrous H₂O₂ for monolayer hydroxyl (-OH) surface functionalization
• Faster nucleation and growth of Al₂O₃ on Si-H surfaces when using anhydrous H₂O₂ versus water
• Little to no photoresist removal at temperatures up to 300°C with anhydrous H₂O₂

Other testing shows that metal oxide film quality grown using anhydrous H₂O₂ are nearly identical to those grown with ozone methods. Metal oxide films include aluminum oxide, hafnium oxide, and zirconium oxide.

The presentation will discuss details of newly discovered reactivity of anhydrous H₂O₂ on several surfaces and will outline potential ASD pathways.

As metal-oxide-semiconductor field effect transistors (MOSFETs) shrink into the <10 nm regime, it becomes a significant challenge to minimize electrical loss with a decreasing pitch especially at the contact regions. To reduce resistance in a compact geometry, selective atomic layer deposition (ALD) of transition metal disilicides is of great interest. In previous studies, selective ALD of tungsten (W) via a fluorosilane elimination process was demonstrated using WF₆ and SiH₄ or Si₂H₆.^1,2 Selectivity was achieved by an inherently favorable reactivity of the precursors on hydrogen-terminated Si versus OH-terminated SiO₂. In this study, sub-stoichiometric MoSi_x (x= 0.7–1.4) was selectively deposited by ALD on H-terminated Si (001) in preference to SiO₂ and SiN using MoF₆ and Si₂H₆ at 120°C. In-situ, X-ray Photoelectron Spectroscopy (XPS) was used to investigate the chemical composition of MoSi_x ateach experimental step. To confirm selective deposition on the nanoscale, MoSi_x was deposited on a Si sample patterned with SiO₂ and Si₃N_4­ and cross-sectional Tunneling Electron Microscopy (TEM) was performed. It was observed that the Si-H surface termination allowed nucleation of MoSi_x on Si in contrast to the inherently chemically passive (non-reactive) SiO_x and SiN surfaces. This substrate-dependent selectivity was retained for MoSi_x growth of up to 10 nm with a proper N₂ purge gas to prevent any CVD components on SiO₂ and SiN. Performing additional Si₂H₆ doses after the ALD cycles allowed the incorporation of more Si into the film and increased the stoichiometry to be closer to MoSi₂. The MoSi_x catalyzes this self-limiting CVD of Si while retaining selectivity over SiO₂ and SiN. In-situ Scanning Tunneling Microscopy (STM) showed that MoSi_x ALD on Si produced an atomically flat surface with a root mean square (RMS) roughness of 2.8 Å. Post-annealing in ultra-high vacuum at 500°C for 3 minutes further decreased the RMS roughness to 1.7 Å. A depth profiling XPS study revealed that the bulk of the MoSi_x film is close to stoichiometric MoSi₂ with <10% oxygen and fluorine. The TEM imaging shows that the selectivity is retained on the nanoscale and that MoSi_x can be selectively deposited on Si without substrate consumption. This is enabled by just taking advantage of the selective ALD of substiochiometric MoSi_x combined with the ability of the substiochiometric MoSi_x films to selectivity induce self-limiting Si deposition from Si₂H₆.

1. Ph. Gouy-Pailler et al., Thin Solid Films, 241, 374 (1994)

2. B. Kalanyan et al., Chem. Mater., 28, 117-126 (2016).

Area-selective atomic layer deposition (AS-ALD) allows nanostructures of arbitrary composition and lateral shape to be built with atomic precision on pre-selected substrate locations. Most current approaches for AS-ALD are based on local inhibition (e.g. with self-assembled monolayers) or activation. However, for some applications of AS-ALD (e.g. in self-aligned fabrication) it is relevant to be able to exploit differences in chemical behavior of a pre-patterned substrate. For this reason, investigating inherent differences in nucleation on diverse substrates is of crucial importance for developing future AS-ALD processes. In this paper we are focussing on substrates of silicon and silicon-based dielectric materials (SiC, SiO₂ and SiN_x) used in electronics.

In order to investigate the possibility of area-selective deposition of Si-based materials using aminosilane precursors, nucleation on four different Si-based surfaces (Si:H, SiC:H, SiO₂:OH and Si₃N₄:NH₂/NH) was analyzed. First, we investigated the difference in precursor adsorption on these surfaces during the exposure of di(isopropylamino)silane (DIPAS), di(sec-butylamino)silane (DSBAS) and bis(t-butylamino)silane (BTBAS) precursors by calculating their adsorption energies using ab-initio modelling. From density functional theory (DFT) calculations, we found that DSBAS is thermodynamically favorable to react with Si₃N₄ and SiO₂, but not with SiC and Si at 0K. To experimentally corroborate these results, SiN_x was deposited using Plasma-Enhanced ALD from DSBAS precursor and N₂ plasma on three different surfaces (H₂ plasma-exposed SiC, HF-last c-Si and c-Si with native SiO₂). In-situ spectroscopic ellipsometry (SE) measurements were performed after every half-cycle to analyze the DSBAS adsorption reaction on these surfaces. During the first DSBAS dosing cycle on the SiO₂ surface a change in the SE signal was observed, suggesting initial DSBAS adsorption, while no changes were observed on the Si and SiC surfaces. The selective adsorption of DSBAS on SiO₂ is in agreement with the DFT predictions. The subsequent N₂ plasma half-cycle modifies the non-growth surface into SiN_x, after which the selectivity is lost. These results illustrate that it is generally difficult to achieve area-selective ALD for nitrides, because of the nitridation of all the exposed substrate surfaces during the plasma step.

We conclude, that during ALD on Si-based substrate materials DSBAS reacts selectively with SiO₂ and SiN_x surfaces but not with Si and SiC surfaces. Our results highlight the role of DFT calculations in predicting possible routes towards AS-ALD process development.

Atomic layer deposition (ALD) is a mainstay technology for the semiconductor industry since it allows deposition of nanometer-thin layers of desired materials onto a substrate in a very controlled and uniform manner. Recently, ALD has been adapted to design and synthesize composite catalysts that allow them to promote multiple chemical reactions. In fabrication of composite catalysts, the selective approaches of ALD are of great importance to exert spatial control of deposition to fabricate three dimensional nanostructures.

In this talk, strategies for selective ALD and enabled nanostructures for catalytic applications will be discussed. Selective ALD allows directional and precise tailoring of the structural size, composition, interfaces, and active sites, that is of great importance for catalysis applications. Two major types of selective ALD approaches are introduced, template selective method via surface modification of self-assembled monolayers (SAMs) and reverse SAMs passivation, as well as inherently selective deposition. With these methods, core shell nanoparticles, oxide overcoating structures ranging from porous coating to ordered structures, and oxide surrounding structures could be fabricated controllably. Theoretical simulations, spectroscopic and microscopic analysis, and catalytic performance are carried out to verify the results. These strategies of selective ALD demonstrate unique advantages to design and fabricate highly stable and active catalysts on the atomic scale, providing unique opportunities to understand the structure–property relationship of catalysis.