ALD/ALE 2026 Session ALE1-TuA: Plasma and Energy-enhanced ALE II

Iridium (Ir) is a promising replacement for copper (Cu) in next-generation interconnects, yet its plasma etching is limited by low-volatility reaction products. This work investigates Ir plasma atomic layer etching (ALE) using Cl/O₂ and CF₄/O₂ chemistries. The ALE sequence employs oxygen-assisted halogenation for surface modification, followed by low-energy ion activation to enable directional desorption of Ir-containing byproducts.

Etching mechanisms are examined using X-ray photoelectron spectroscopy (XPS), time-of-flight secondary ion mass spectrometry (TOF-SIMS), and transmission electron microscopy (TEM). XPS identifies Ir–Cl, Ir–F, and Ir–O bonding states, TOF-SIMS quantifies modified layer thickness and molecular distribution, and TEM probes small critical-dimension feature profiles and sidewall chemical composition. Correlating these surface and structural analyses with etch-per-cycle, profile anisotropy, line-edge roughness, and hardmask selectivity enables a better understanding of Ir ALE characteristics and guides the optimization of Ir interconnect pattern fidelity and dimensional control.

The critical dimensions of semiconductor devices continue to shrink, reaching nanometer and even angstrom scales in both 2D and 3D structures. Consequently, the demand for atomic-scale precision in etching processes is rapidly increasing. This talk will present various examples of plasma-based atomic layer etching (ALE) with emphasis on etch selectivity, surface roughness, and residue control for both metal and dielectric materials. Materials discussed include silicon nitride, titanium nitride, zirconium oxide, silicon oxide, aluminum oxide, hafnium oxide, molybdenum, ruthenium, and tantalum nitride. Typical ALE processes consist of two sequential steps: surface modification and material removal. Surface modification is achieved through various reaction pathways, including fluorocarbon film deposition, surface fluorination, chlorination, and oxidation using plasma-generated radicals. In the subsequent removal step, the modified layers are eliminated through mechanisms such as ion bombardment, thermal desorption, ligand exchange, ligand volatilization, or halogenation. This talk will also address key characteristics of plasma-based ALE processes, including etching selectivity, surface roughness, and residual surface contamination.

Atomic layer etching (ALE) of AlGaN is well suited to the critical gate recess step of AlGaN/GaN normally-off metal insulator semiconductor-high electron mobility transistors (MIS-HEMT) [1]. A two-dimensional electron gas (2DEG) forms at the AlGaN–GaN interface (Fig. 1) enabling the mobility and carrier density required for high–performance power devices. As shown in Fig. 2, 2DEG mobility directly governs device efficiency [2]. Its formation depends strongly on AlGaN film properties and thickness. Precise control of the AlGaN gate–recess thickness is essential for engineering threshold voltage (V_TH) and blocking capability, ideally ~5 nm thickness [3]. Low–damage processing is necessary to minimise interface traps, leakage, and electric–field crowding [4].

Using a reflectance-based endpoint technique, sub-5 nm AlGaN films have been etched with Cl₂/BCl₃ ALE. Fig. 3 plots dose time vs etch–per–cycle (EPC) for 30 - 120 ms doses, with a clear EPC jump at 40 ms and plateauing thereafter, indicating saturation. Process repeatability is demonstrated in Fig. 4: across 11 runs, EPC ranged from 0.79–0.88 nm/cycle. The normal distribution of this data set is plotted in Fig. 5, indicating run-to-run EPC uniformity of 5.4 %. However, reflectance end pointing only measures the targeted point and gives no cross–wafer EPC uniformity. Such uniformity is difficult to measure on <10 nm AlGaN due to surface roughness, ellipsometry limits, and FIB/SEM damage. We aim to infer etched film uniformity by correlating reflectance–measured AlGaN thickness with sheet resistance obtained via Hg–probe analysis. Fig. 6 shows in–situ thickness measurements for targeted 5 nm and 0 nm endpoints, illustrating the strength of the technique: the ALE sequence stops automatically when the desired thickness is reached [5]. Fig. 7 displays raw reflectance data; the initial smooth region occurs before the “turning point,” after which true thickness is extracted, shown by the stepped profile. Accurate end thickness does not require knowledge of the initial AlGaN thickness.

In this submission we will quantify EPC uniformity across Ø150 mm AlGaN–GaN substrates by combining in–situ reflectance thickness measurement with local Hg–probe C–V/I–V sheet–resistance mapping. We will also electrically characterise surface roughness of sub–5 nm AlGaN films produced by ALE and ICP etching. Finally, we will expand the current dataset to show how etch–process parameters influence run–to–run uniformity across multiple AlGaN–GaN wafers.

As the GaN light emitting diode (LED) device technology is developed from conventional LED to micro LED, the lateral dimension of LED devices is decreased, and the ratio of sidewall area relative to overall device area is increased. Especially, as the GaN LED device size decreases below 5 microns, the performance and reliability of GaN devices are significantly degraded. One of the reasons for the degradation of the micro LED device is the damage from the sidewall due to the ion bombardment during the reactive ion etching (RIE).Ion bombardment during the RIE can lead to various issues such as surface composition changes, surface defects, surface contamination, and increased leakage current.

In this study, we focused on the etching of GaN LED devices with a multi quantum well (MQW) layer, which is composed of multiple InGaN and GaN layers. A mesa structure of GaN LED device consisted of p-GaN/MQW/n-GaN/undoped-GaN layer on sapphire wafers (or silicon wafers) was etched using BCl₃/Cl₂/Ar RIE, and the sidewall damage remaining after the RIE was removed using wet etching and/or ALE methods using conventional ICP etch system and an ion beam-based etch system. The effect of wet etching of damaged GaN LED sidewall using a KOH-based solution was compared with the sidewall damage removal using ALE methods. The effect of ion beam-based ALE on the removal of sidewall damage for vertical GaN LED devices will be also discussed. The results showed that, the wet etching improved the sidewall stress or defects due to the damage by RIE, however, the optimized ALE processes removed the sidewall stress and defects on MQW layers of patterned GaN structure almost completely. This indicates that, as GaN-based devices decrease in size, the effectiveness of ALE increases, making ALE more effective than wet etching for removing sidewall etch damage for next-generation device fabrication.