ALD/ALE 2026 Session AS1-WeM: ASD Process I

High selectivity in area selective atomic layer deposition (AS-ALD) requires the effective performance of an inhibitor that must exhibit selectively binding mode on non-growth areas as well as strong thermal, and chemical stability to prevent degradation or decomposition during the semiconductor manufacturing process. N-heterocyclic carbenes (NHCs) have emerged as promising next-generation alternatives to conventional small-molecule inhibitors (SMIs) due to their strong σ-donor character and preferential binding to metal surfaces over a dielectric region. We have prepared and developed triazolylidene NHC (Tz) derivatives as a novel class of NHC inhibitor for selective high-k dielectric growth on SiO₂ over metallic bands reaching a selectivity factor of 93% after 50 dielectric deposition cycles. The selective adsorption behavior and dielectric blocking efficiency were systematically evaluated using time-of-flight secondary ion mass spectrometry (ToF-SIMS), and X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM). The practical applicability of the Tz inhibitor is further demonstrated through bottom-up fabrication of a working field-effect transistor, in which the NHC selectively protects metal electrodes during the deposition of metal oxide dielectric and semiconductor layers. This work paves an innovative pathway for exploring novel class of SMIs toward advanced AS-ALD applications.

Selected references: (1) D.A.R. Nanan, J.T. Lomax, J. Bentley, L. Misener, A.J. Veinot, W-T Shiu, L. Liu, P.J. Ragogna, C.M. Crudden J. Am. Chem. Soc.2025, 147, 5624–5631; (2)J.T. Lomax, E. Goodwin, M.D. Aloisio, A.J. Veinot, I. Singh, W-T Shiu, M. Bakiro, J. Bentley, J.F. DeJesus, P.G. Gordon, L. Liu, S.T. Barry, C.M. Crudden, P.J. Ragogna Chem. Mater.2024, 36, 5500–5507; (3) P.J. Ragogna, C.M. Crudden, D.A.R. Nanan, J.T. Lomax, A.J. Veinot, J. Bentley, “Method of Selective Deposition of Triazolylidenes on Metallic Surfaces”, International Patent Application No. PCT/CA2025/051508, filed: Nov. 12, 2025

Machine learning–guided molecular design is important for accelerating materials discovery, and we have applied it to area-selective atomic layer deposition (AS-ALD). In this work, we integrate machine learning with systematic experimental characterization to guide the development of amino-substituted silane small-molecule inhibitors (SMIs) to optimize volatility, thermal stability, and surface selectivity. By prioritizing data-driven structure–property relationships, this study helps establish a predictive framework for identifying viable SMIs while limiting experimental trial-and-error.

A diverse library of aminosilanes was synthesized using an environmentally benign nucleophilic substitution route involving disubstituted amines and chlorosilanes, with triethylamine as a sacrificial base. This approach avoids pyrophoric reagents and metal-containing intermediates, supporting sustainable scale-up. Thermal properties were evaluated using thermogravimetric analysis, vapour pressure measurements, and differential scanning calorimetry. Surface adsorption and selectivity were quantified using in situ quartz crystal microbalance measurements.

The experimentally derived thermophysical dataset was used to train ML models to map molecular structure and composition to key properties governing AS-ALD performance, including volatility and thermal stability. A chemistry-informed molecular property prediction platform (DeepAutoQSAR, Schrödinger) helped identify optimal model architectures. The resulting ML models were used to predict properties of previously untested silane precursors, enabling targeted selection of candidates for experimental validation. Comparison of predicted and measured properties demonstrates the effectiveness of this ML-guided workflow for accelerating inhibitor design.

Area selective atomic layer deposition (AS-ALD) enables the selective placement of material based on differences in surface chemistry and is thus a promising strategy for device manufacturing by avoiding addition patterning steps and alignment issues.Key developments that are being sought include achieving high selectivity (near 100%) at high film thicknesses on growth surfaces, and expanding the palette of materials (such as new low-k materials) that can be grow in AS-ALD.In this talk I will explore two aspects of our work that work toward these developments: Examination of the role of alkanethiol chain length and examination of the role of alternative metal and co-reactant precursors on selectivity.

The stability and impermeability of monolayer-based blocking layers is critical to preventing film growth in certain regions.Few reports explicitly studied the effect of alkane chain length and temperature on selectivity, so we sought to do so with a series of alkanethiols of various alkane chain length on copper surfaces.We found that longer chain lengths achieved higher selectivity and all chain lengths to be unstable to the highest temperature investigated (180 °C).To isolate the precise breakdown mechanisms, we separately subjected the alkanethiol monolayers on copper to various individual ALD steps including elevated temperature, metal precursor exposure, or water exposure.We found that trimethylaluminum at elevated temperature induced alkanethiol desorption, whereas amido-based Hf precursors did not, thus demonstrating a chemical effect on monolayer stability.

To further investigate the role of film precursor, we studied an array of metal and non-metal precursors including alkyl aluminums, amido aluminum, aluminum alkoxide, and ethylene glycol.Importantly, we found that precursor size, rather than reactivity, was the prime determining factor in realizing high sensitivity.We found that molecular layer deposited films (using ethylene glycol) did not show significantly higher selectivity than traditional ALD growths with the same metal precursor.With certain combinations of large metal precursors and water, we were able to achieve high selectivity (>90%)at alumina film thicknesses greater than 15 nm on the growth surface.Thus, this work builds on existing reports from other groups that the precursor chemistry has a massive role in determining selectivity.

As semiconductor devices continue to scale down, current densities in interconnects increase, leading to serious reliability degradation of Cu interconnects due to electromigration. The introduction of a metallic Co cap layer is expected to enhance adhesion between interconnects and dielectric layers, thereby extending the lifetime of Cu wiring. In Co atomic layer deposition (ALD) using CCTBA (Cobalt Carbonyl Tertiary-Butyl Acetylene) as a precursor, an incubation period exists in which film nucleation on SiO₂ dielectric surfaces is delayed compared with Cu surfaces, enabling selective growth [1]. In this study, the inhibitor effect of NH₃ was evaluated as a method to further enhance selectivity, motivated by previous reports on selective CVD using Co₂(CO)₈, where simultaneous NH₃ supply suppresses growth initiation and improves selectivity [2].

Deposition experiments were performed on Si substrates with a 300 nm thermally grown oxide layer at 120 °C using alternating CCTBA and H₂ pulses for 500 cycles, with in-situ monitoring by spectroscopic ellipsometry. Under standard conditions, no incubation period was observed and a growth-per-cycle (GPC) of 0.016 nm/cycle was obtained. When NH₃ was co-supplied during the CCTBA pulse, an incubation period of approximately 75 cycles appeared and the GPC decreased to 0.0099 nm/cycle, demonstrating that NH₃ functions effectively as an inhibitor.

Furthermore, surface reaction calculations using a machine-learning potential (PFP: Preferred Potential by Matlantis Corp.) suggest that NH₃ selectively adsorbs on OH groups on oxide surfaces, which serve as potential adsorption sites for CCTBA molecules, thereby blocking these reactive sites. These results indicate that NH₃ can act as a small-molecule inhibitor in Co-ALD using CCTBA, providing a promising approach for enhancing selective growth in Co capping processes for advanced interconnect applications.

The authors gratefully acknowledge Daikin Industries, Ltd. For their support and valuable discussions.

References

[1] J. Yamaguchi et al., AVS 24th International Conference on Atomic Layer Deposition, AA1-TuM-7 (2024).

[2] Z. V. Zhang et al., J. Vac. Sci. Technol. A 38, 033401 (2020).

To address this, we reconsidered a previously studied “waterless” TiO₂ ALD process using TiCl₄ and titanium isopropoxide (TTiP) at 210℃.³ Using TiCl₄/TTiP ALD on blanket SiO₂ and SiN_x substrates, we found (Figure 1a and b) that after pre-treating the SiO₂ and SiN_x with only a brief dip in dilute HF_(aq), the TiCl₄/TTiP process showed delayed deposition on both surfaces. However, the extent of delay was much more substantial on the SiO₂ vs SiN_x, allowing several nm of ASD. Subsequently, we tested the same SiO₂/SiN_x surfaces after dilute HF_(aq) followed by exposure to MoF₆ for 1 second at 210℃. As shown in Figure 1c and d, initial results indicate substantially improved selectivity for the fluorine-passivated water-free TiO₂ ALD, enabling ~ 25 nm of TiO₂ on SiN_x vs. SiO₂ with exceptionally high selectivity. The lines in the figures correspond to fits obtained from an analytical nucleation model.⁴ Confirmation of these findings will require testing the process on nanoscale patterned substrates. Overall, these results demonstrate how the combination of pretreatment and precursor selection can help achieve chemical contrast for ASD, even on starting surfaces with similar chemical structure and composition.

The effectiveness of the silylation of both silicon oxide and copper surfaces using N-(trimethylsilyl)dimethylamine (TMSDMA) as a passivation agent in atomic layer depositions (ALDs) was evaluated and contrasted by using x-ray photoelectron spectroscopy (XPS). It was determined that on SiO₂ such silylation does indeed block the nucleation centers where the ALD precursors are activated and therefore inhibit film growth, but only temporarily; after a few ALD cycles, deposition becomes evident. By testing this chemistry on two types of SiO₂ surfaces, prepared by plasma-enhanced chemical vapor deposition (PE-CVD) and by chemical (RCA) treatment of Si(100) wafers, it was concluded that the nature of the initial substrate does not play a crucial role in the silylation or ALD blocking processes. The material being deposited, on the other hand, does make a difference: TiO₂ film growth can be blocked for almost 10 ALD cycles, whereas HfO₂ starts building up on the surface after less than 5 ALD cycles. Moreover, the steady-state deposition rate for TiO₂ was determined to be lower than for HfO₂. One of the key findings of this work is that the silylation can be carried out using either gas- or liquid-phase treatments. It was found that the extent of silylation and the inhibition of the subsequent ALD were comparable in both cases, but the gas-phase method was determined to be cleaner and to deposit less carbon contaminants.