ALD/ALE 2025 Session AS1-WeM: Area Selective Deposition II

Area-selective atomic layer deposition (AS-ALD) has become essential for advanced semiconductor manufacturing by enabling precise bottom-up material deposition and reducing process complexity. Various approaches have been developed to achieve selective deposition, including surface activation, growth inhibition using self-assembled monolayers, and growth-etch back processes. These strategies are particularly crucial for fabricating self-aligned features in advanced technology nodes, where traditional lithography and etching face increasing challenges in meeting the demands of device scaling.

Thermal atomic layer etching (ALE) is a method for thin film removal based on sequential surface modification and volatile release reactions. After thermal ALE, atomic layer deposition (ALD) may be needed to deposit another material on the etched surface. Many thermal ALE procedures are defined by sequential fluorination and ligand-exchange reactions. These reactions can leave the surface terminated with fluoride and methyl or chloride species, respectively. These terminated surfaces may inhibit ALD. Additional chemical treatment may be required to reactivate the surface for normal ALD growth.

In this study, zinc sulfide (ZnS) ALD was conducted on Al₂O₃, SiO₂, HfO₂, and native oxide on Si substrates before and after ALE. The ALE was performed at 300 °C using sequential HF and Al(CH₃)₃ (TMA) exposures. ZnS ALD was conducted at 300 °C using diethylzinc (DEZ) and hydrogen sulfide (H₂S) exposures. The ZnS ALD thicknesses were monitored using in situ spectroscopic ellipsometry. When ZnS ALD was performed on as-deposited Al₂O₃ ALD films, the ZnS ALD films grew immediately with no noticeable nucleation delay. In contrast, the ZnS ALD growth was completely inhibited until 50 ALD cycles after the Al₂O₃ film was etched using sequential HF and TMA exposures. Similar results were observed for the other metal oxide substrates.

Various chemical treatments were explored to reactivate the metal oxides after ALE for normal ZnS ALD. The surface species present after ALE using HF and TMA exposures are fluoride or methyl species. These species lead to a nucleation delay for ZnS ALD. These surface species can react with exposures of H₂O₂ or H₂O to add hydroxyl groups to the surface. The ZnS ALD had only a slight nucleation delay using H₂O₂ exposures after ALE. This study reveals that ZnS ALD growth occurred almost immediately after H₂O₂ reactivation of the etched Al₂O₃ substrate. Results will also be shown for different chemical treatments for the various metal oxide substrates.

In recent years, area-selective atomic layer deposition (AS-ALD) has emerged as a superior technique for precise and selective thin-film deposition, surpassing conventional methods. AS-ALD has demonstrated promising potential for 2D and 3D nanoscale patterning by using inhibitor molecules to tailor surface properties. This approach enables thin-film deposition exclusively on desired growth surfaces (GS) while preventing unwanted growth on non-growth surfaces (NGS). Among various inhibitors, small molecule inhibitors (SMIs) have received significant attention for their effective inhibition despite their small size. The choice of surface inhibitor is crucial in determining the selectivity between growth and non-growth surfaces, as well as the degree of surface passivation. In this study, an aldehyde-based inhibitor, trimethylhexanal (TMH), was utilized for AS-ALD to achieve selectivity between oxide and nitride surfaces. TMH displayed selective adsorption on a diluted hydrofluoric (DHF) acid-treated Si₃N₄ surface while leaving the SiO₂ surface unaffected. The DHF pretreatment facilitated favorable TMH adsorption on Si₃N₄ via interactions with NH and NH₂ surface groups, whereas the OH-terminated SiO₂ surface exhibited no TMH adsorption. Density functional theory (DFT) calculations confirmed the favorable adsorption energy of TMH on the nitride surface. To assess the blocking properties of TMH, SiN_x ALD was performed using silicon tetrachloride (SiCl₄) as the precursor and ammonia (NH₃) as the reactant. The results demonstrated that SiN_x could be selectively deposited on the SiO₂ surface with no detectable growth on the TMH-treated Si₃N₄ surface. This process shows great promise for 3D patterning in silicon devices, where AS-ALD is critical to meeting the demands of miniaturization.

Selective deposition on SiO₂/SiN_x patterns, where either oxide-on-oxide or nitride-on-nitride deposition is desired, is currently considered as the “holy grail” in the field on area-selective deposition (ASD). Unlike other silicon-based materials (e.g., Si, SiO₂), SiN_x can be terminated with a large variety of surface groups e.g., (NH₂, NH, N, SiH). As a result, the surface chemistry of a SiN_x surface is very sensitive to the manner in which the layer was deposited, and to how the layer was treated after deposition. In addition, in case of air exposure, this surface chemistry becomes even more complex due to (partial) oxidation of the layer (which adds OH groups). For ASD, the surface chemistry of the SiN_x strongly affects the (selective) adsorption of the inhibitors and precursors. In this contribution, it will be discussed how plasma pretreatments can be used to control the surface chemistry of a SiN_x surface and correct for any unwanted oxidation, as well as how the obtained surface chemistries affect inhibitor adsorption on a SiN_x surface.

Three different plasma chemistries were explored: N₂, H₂, and NH₃ plasma. In-situ reflection adsorption infrared spectroscopy (RAIRS) results show that the N₂ and NH₃ plasma pretreatments were effective in removing the OH groups introduced by air exposure, whereas the H₂ plasma was insufficient and only partially removed the OH groups. The N₂ plasma was observed to result in a largely N-terminated SiN_x surface, while the NH₃ plasma resulted in a largely NH₂-terminated surface. The effect of the plasma pre-treatments on the adsorption of trimethylacetaldehyde (TMAAH) and acetylacetone (Hacac) inhibitor molecules onto the SiN_x was studied using in-situ RAIRS and density functional theory (DFT) calculations. Our results show that the SiN_x surface termination strongly affects whether and how strongly the inhibitors adsorb on the SiN_x surface. Importantly, Hacac was found to require a N-terminated SiN_x surface for adsorption, while TMAAH adsorbs most strongly on a NH₂-terminated surface. Therefore, these results demonstrate the importance of understanding and controlling the surface chemistry for ASD.

Area Selective Deposition is a key technique in semiconductor fabrication processes, enabling the deposition of thin films in a controlled, pattern specific manner. Achieving a high level of selectivity, i.e., no growth on “non-growth surface” and high-density continuous growth on the “growth surface” is essential for ensuring good device yield. Traditional Atomic Layer Deposition processes often face limitations in selectivity due to the inherent challenges in controlling surface chemistry and reaction kinetics.

This talk will explore an approach to enhancing area selectivity through a subsaturated ALD process. By carefully controlling the precursor selection, flux and reaction time to operate in a subsaturated regime, we can significantly reduce unwanted deposition on non-target areas while ensuring high quality film growth on desired regions. This method leverages the intrinsic self-limiting nature of ALD while mitigating issues such as precursor overexposure and undesired nucleation, resulting in improved spatial uniformity and deposition control.

The talk will present experimental results demonstrating the superior selectivity and film quality achieved with sub-saturated aluminum oxide ALD in a dielectric on dielectric (DoD) process compared to a traditional, saturated process. We will also discuss the underlying mechanistic insights, including the role of surface coverage, precursor dynamics and impact of process parameters such as temperature, dilution and pressure. Additionally, a known challenge with ASD – metrology will be discussed. Lam’s successful attempts at image processing to understand large scale selectivity over an entire die will be presented. With the need to shorten the time from R&D to High Volume Manufacturing (HVM), this fast, quantitative metrology is key to rapid validation of new processes.The quick turnaround image processing method allows for testing of marginal processes enabling a larger number of experiments in a short period of time.

This work provides future directions for optimizing selectivity and process uniformity at smaller pitches across a wide range of semiconductor applications.

With the ongoing downscaling of logic and memory devices, one of the main challenges has emerged such as edge placement error issues resulting from top-down patterning. To overcome the limitations of lithography, a recent focus in bottom-up patterning is based on area-selective atomic layer deposition (AS-ALD). Numerous studies have investigated AS-ALD that employed precursor inhibitors such as SAMs or SMIs to prevent precursor adsorption in non-growth areas. However, there is an increasing need for research into inherent AS-ALD strategies, which exploit the intrinsic properties of substrates with the chemical adsorption of the precursor to enable selective adsorption at targeted surface sites. Molybdenum carbide (MoC_x) has attracted as promising materials for metallization, particularly as bottomless diffusion barriers, liners, capping layers, and interconnects, due to their high melting points, low resistivity, excellent thermal stability, and low reactivity with Cu and its area selective deposition methods have been requiring.

In this work, we developed conductive MoC_x films via thermal ALD without the use of halogen-based precursors, at the deposition temperatures of 200−300℃. This process enabled area-selective growth of MoC_x films on metallic substrates (TiN, Ru, Cu) over oxide substrates (SiO₂, Al₂O₃) by utilizing the intrinsic chemical adsorption of the precursor. We investigated the crystallinity, chemical bonding states, impurity, and resistivity of the MoC_x films, and evaluated the selectivity between substratesthrough analysis of Mo areal density and film thickness. Moreover, the selective growth of MoC_x films on metallic substrates was demonstrated on metal/dielectric patterns using auger electron spectroscopy (AES) mapping and energy-dispersive X-ray spectroscopy (EDS) analysis, indicating the feasibility of implementing this process in practical device applications. To elucidate substrate-dependent surface chemistry in MoC_x AS-ALD, density functional theory (DFT) calculations were conducted, revealing the relative adsorption energies of Mo precursor between metal and dielectric substrates. In conclusion, a newly developed inherent AS-ALD of MoC_x films presents a promising alternative to top-down processes, offering a simplified workflow and potential conducting materials for advanced metallization.

Atomic Layer Deposition (ALD) is a process that forms uniform thin films with atomic-level precision. Within this process, Area-Selective Deposition (ASD)—a technique that achieves film growth selectively on specific areas—has garnered significant attention, and recent research has been actively focused on it. With advancements in computational chemistry, the atomic-level mechanisms in ALD processes have been elucidated, enabling the design of ALD processes and ALD precursors using these insights.

In this study, we propose a method for designing ASD-suitable precursor molecules using atomic-level simulations with an universal machine learning interatomic potential (uMLIP). Specifically, we constructed an adsorption energy dataset for various precursor model molecules using adsorption energy calculations for areas where film formation is desired and areas where it is not (modeled as H-terminated Si surfaces and alkyl-protected surfaces, respectively). Through the sparse modeling of this data, we extracted ligand features that enhance adsorption selectivity towards H-terminated Si surfaces and alkyl-protected surfaces. Consequently, we successfully identified ligand candidates that increase adsorption selectivity towards H-terminated Si surfaces.

Furthermore, by combining a uMLIP with the Global Reaction Route Mapping (GRRM) method, we successfully analyzed the ligand dissociation mechanisms of ALD precursor molecules on surfaces. These methods allow for the design of precursor molecules that meet the high adsorption selectivity and ligand dissociation requirements necessary for ASD.