ALD/ALE 2025 Session AF1-MoA: ALD on 3D Structures

Powder-based processes are critical to a wide range of industries, from battery materials to pharmaceuticals. Achieving the desired performance often depends on precise surface modifications.Over the past decade, Powall has been developing a scalable nanocoating technology that combines the atomic-level precision of Atomic Layer Deposition (ALD) with the speed and throughput of Chemical Vapor Deposition (CVD), enabling continuous and cost-effective production.

In this presentation, Dr. Moitzheim (CTO) will share Powall’s key learnings on transitioning gas-phase coating methods from lab to pilot-scale and beyond. Specific attention will be paid to how the technology can be adapted to accommodate powders of varying sizes, shapes, and porosities, while still ensuring uniform and high-quality coatings. The talk will also provide an update on the upcoming commercial pilot-scale equipment and outline the path toward large-scale manufacturing. By illustrating the core challenges and advancements in scaling powder nanocoatings, this talk will offer a balanced perspective for researchandindustry.

As of late there were two key research questions regarding the conformality of plasma-enhanced spatial ALD (PE-s-ALD): can the technique be used to deposit highly conformal films, and how do the material properties – like crystallinity and composition – of such films change throughout the coated 3D structures? We have recently answered the first question by showing that exceptionally conformal films can be grown by PE-s-ALD with subsecond plasma exposures, thanks to the high radical density in the atmospheric plasma [1]. However, various crystallization and growth effects can influence the film profile and crystallinity, and understanding these effects and their interplay is key.

In this work, we demonstrate the complex growth mechanism of TiO₂ using PE-s-ALD, and study conformality not just in terms of films thickness, but also in terms of film properties. TiO₂ films are deposited both on planar substrates and inside lateral high-aspect-ratio (LHAR) test chips (PillarHall^TM by Chipmetrics Ltd). These LHAR structures uniquely allow for spatial material property mapping in trenches using spectroscopic ellipsometry, Raman spectroscopy and X-ray photoelectron spectroscopy. Conditions that result in the anatase phase on a planar surface only partially form this phase inside LHAR structures, with the deepest part of the film being amorphous. This partial crystallization is ascribed to the film thickness inside the LHAR structure gradually dropping below the critical thickness for crystallization. In turn, the partial crystallization is shown to have a significant effect on the resulting thickness profile, due to an enhanced growth per cycle on crystalline surfaces. A framework of the interplay between effects is proposed, offering insights that enable better control of the crystallinity and thickness throughout the entirety of coated surfaces of 3D structures by PE-s-ALD.

Additionally, the recombination probability of O-radicals during this atmospheric-pressure PE-s-ALD process at 200 °C is determined to be 3×10^-5, which is similar to low-pressure PE-ALD [2]. This result indicates that differences in conformality between the two types of ALD are not the result of differences in recombination probability, but rather of differences in initial radical density and diffusion behavior.

[1] van de Poll, M. L., Jain, H., Hilfiker, J. N., Utriainen, M., Poodt, P., Kessels, W. M. M., & Macco, B. (2023). Applied Physics Letters, 123(18), 182902.

[2] Arts, K., Deijkers, S., Puurunen, R. L., Kessels, W. M. M., & Knoops, H. C. M. (2021). Journal of Physical Chemistry C, 125(15), 8244–8252.

The increasing demand for high-bandwidth memory necessitates the development of devices with 3D structures, such as DRAM. These devices rely on the deposition of conformal, particle-free films with complete coverage in high-aspect-ratio (HAR) spaces. Transmission electron microscopy (TEM) is the standard technique for verifying these parameters, but it is costly, time-intensive, and only inspects a very small area of the surface of interest. This study demonstrates a rapid and non-destructive alternative involving thermally bonded chips that provide HAR spaces for deposition. The chips can be debonded and analyzed using scanning electron microscopy (SEM) and atomic force microscopy (AFM).ALD titanium nitride (TiN), typically utilized as a 2 nm diffusion barrier between tungsten (or copper) and SiO₂ (or SiCOH) in HAR spaces, was employed using TiCl₄ and N₂H₄ precursors, to deposit a 20 nm layer in a thermally bonded chip with a 2000:1 aspect ratio (fig. 1). This is 10x the normal thickness of diffusion barriers to increase sensitivity to particle formation.

A control test was conducted on a clean, thermally bonded sample with no TiN deposition (fig 2). The debonded dies were analyzed as planar samples using AFM, SEM, and energy-dispersive spectroscopy (EDS), confirming no detectable titanium (0% atomic percent) in any region (edge, middle, center). The AFM RMS was measured at 2.06 nm. For TiN deposition, increasing precursor doses 1×, 2×, and 3× normal exposure (1x = 500 ms TiCl₄, 2750 ms N₂H₄, 30 sec purge, substrate temperature 475°C) were employed to evaluate penetration depth and particle formation. The RMS roughness values for the top die remained low at 1.92 nm, 1.77 nm, and 1.64 nm, respectively. This indicates no particle formation. Atomic Ti percentages from EDS decreased as the effective aspect ratio increased from 475:1 to 2000:1 aligning with predictions from the Gordon Model [1]. Additionally, the 3x smaller %Ti decreases at the center between the 3× and 2× dose sample further support the model’s estimation of precursor penetration. Temperature variations across the sample may have prevented complete surface saturation. The low RMS roughness and absence of large features suggest that CVD particle formation did not occur in the HAR space, consistent with the TiCl₄ + N₂H₄ chemistry avoiding NH₄Cl(s) formation [2].

AFM was used to analyze an area of over 111 µm², revealing no defects or CVD particles. In comparison, a typical TEM survey covers only 0.004 µm². This means SEM/AFM in debonded chiplets examined a region ~30,000x larger than TEM allowing large area determination of particles formation in HAR.

As semiconductor technology advances, improvements in performance, power efficiency, and area optimization continue to drive innovation. However, increasing device complexity and higher integration levels have exacerbated step coverage challenges during the deposition of high-k thin films in DRAM capacitors, metal gate insulators, and block oxides in 3D NAND. In particular, achieving uniform thin-film deposition on high-aspect-ratio patterns has become increasingly difficult. Additionally, the application of high-temperature processes to enhance film quality can lead to precursor decomposition, further deteriorating step coverage. To address this issue, this study proposes the use of inhibitor technology as an effective approach to improving step coverage in high-temperature atomic layer deposition (ALD) processes. Our experimental results demonstrate that the inhibitor remains stably adsorbed on the substrate surface under various ALD conditions, effectively mitigating step coverage issues. Notably, this inhibition effect persists even at temperatures above 600°C during the growth of Al₂O₃ thin films. Furthermore, a newly developed hydrocarbon-based inhibitor has been identified as optimal, as it fully decomposes into water and carbon dioxide upon reacting with ALD precursors, leaving no residual impurities in the film.