ALD/ALE 2026 Session ALE-MoP: Atomic Layer Etching Poster Session

Atomic layer etching (ALE) with high selectivity, low damage, and minimal thermal budget is increasingly important for advanced microelectronics and quantum devices. In this work, we present a plasma-based ALE approach for SiO₂ and Al₂O₃ that enables controlled, isotropic material removal using sequential SF₆, trimethylaluminum (TMA), and low-power Ar plasma exposures. The processes are designed to operate at room temperature or modest heating, expanding the accessible parameter space for temperature-sensitive substrates.

For SiO₂, we implement a near-room-temperature ALE sequence using alternating SF₆ and Ar plasma steps. Ellipsometry measurements indicate self-limiting behavior in both half-reactions, with no measurable etching during isolated SF₆ or plasma exposures and an expected etch-per-cycle (EPC) of ~1 Å. A defined temperature and plasma-power window supports stable etching and strong selectivity to underlying Si, making the process suitable for applications where substrate preservation is critical.

We further extend this strategy to Al₂O₃ by introducing a TMA pulse following SF₆ surface fluorination. This fluorination–ligand-exchange mechanism enables isotropic removal of the modified layer, yielding an EPC of ~0.9 Å/cycle at 150 °C. The process provides reliable etching on both horizontal and vertical surfaces, demonstrating applicability to 3D structures. XPS, AFM, and SEM confirm reduced surface roughness and uniform material removal, consistent with self-limited etching.

Together, these results establish a versatile ALE platform capable of high-precision, low-temperature oxide etching. The approach offers a safer and more controllable alternative to HF-based methods and supports the fabrication of complex geometries in emerging semiconductor, superconducting, and quantum technologies.

This work is carried out through a collaboration between Lawrence Berkeley National Laboratory, the Molecular Foundry, and the University of Copenhagen as part of the NQCP program, enabling coordinated development of low-temperature ALE for next-generation nano- and quantum device fabrication.

Fluorine-based quasi-atomic layer etching (quasi-ALE) of SiO₂ is widely used as an ALE process. However, in this process, fluorocarbon (CF) polymers deposited on the reactor chamber re-evaporate and redeposit on the wafer surface, causing the etch rate per cycle to gradually increase with increasing cycle number [1,2]. This behavior originates from the fact that, unlike ideal ALE, the modification step in quasi-ALE does not proceed in a self-limiting manner, and the etch rate depends on the thickness of the deposited film. As a result, even when cycles are repeated under identical process conditions, the etch amount per cycle cannot be maintained at a constant value. In this study, we focus on the instability of the etch amount per cycle and propose a method to stabilize it. A SiO₂ quasi-ALE process using C₄F₈/Ar gas was employed, in which modification and removal steps were alternately repeated in a cyclic manner. To remove CF polymers deposited on the reaction chamber and stabilize the etch amount per cycle, a short O₂ plasma step (several seconds) was inserted after the removal step of each cycle. The etch amount per cycle was evaluated and compared with and without the O₂ plasma step.
Without the O₂ plasma step, the etch amount per cycle increased after approximately 50 cycles, leading to a loss of linearity in the etch rate. In contrast, when the O₂ plasma step was introduced in every cycle, the etch amount per cycle remained nearly constant throughout all cycles, resulting in stable linear etching characteristics. Furthermore, the self-limiting behavior observed during the removal step exhibited similar saturation characteristics regardless of the presence or absence of the O₂ plasma step. These results demonstrate that the proposed method effectively stabilizes the etch amount per cycle in fluorine-based SiO₂ quasi-ALE processes by suppressing cycle-to-cycle variations without altering the underlying reaction mechanism.

[1] D. Metzler et al., J. Vac. Sci. Technol. A 32, 020603 (2014).
[2] A. Ronco et al., J. Vac. Sci. Technol. A 42, 062601 (2024).

GaN is a wide-bandgap semiconductor that exhibits excellent electrical and optical properties, making it suitable for LED device. In particular, it is well-suited for light-emitting diodes due to its direct bandgap property. [1] Light-emitting devices adopt a mesa structure to expose multi-quantum well (MQW) layer composed of InGaN and GaN emitting layer. A direct plasma-based etching process is commonly used to form these structures. [2] Unfortunately, reactive ions accelerated by the combined potential of the self-bias and sheath inevitably induce structural and physical damage, which particularly lead to electrical and optical degradation of the MQW. To mitigate plasma-induced damage, wet etching techniques using KOH solutions are employed to remove the damaged GaN surface layer. [3] Subsequently, atomic layer etching (ALE) has been investigated to eliminate the remaining damage layer. However, ALE conducted in direct plasma cannot efficiently target vertical GaN structures including MQWs, because accelerated ions are incident perpendicularly to the surface. To address this issue, ALE utilizing an ion-beam source has been investigated, enabling selective targeting of MQW layers through the use of tilted ion beams.

In this study, atomic layer etching (ALE) utilizing an ion beam is introduced after RIE process to remove the damage caused by RIE in GaN-based structures, including multi-quantum well (MQW) layers composed of InGaN and GaN. X-ray photoelectron spectroscopy (XPS) analysis reveals that RIE-induced damage alters the N/Ga and Ga/In atomic ratios relative to reference data. However, after ion-beam ALE treatment following RIE, the atomic ratios return to values similar to the reference data, whereas wet etching only partially restores the atomic ratios. Consequently, ion-beam ALE is a promising approach for enhancing device performance by providing efficient surface damage removal, particularly in nanoscale vertical GaN-based devices.

References:

[1]I. Akasaki, Nobel Lecture: Background story of the invention of efficient blue InGaN light emitting diodes, Rev. Mod. Phys. 87, 1119 (2015).

[2] K. Eriguchi, Jpn. J. Appl. Phys. 56, 06HA01 (2017).

[3] C. H. Kim et al., Appl. Surf. Sci. 714, 164429 (2025).

As demand for high-performance chips increases, not only the critical dimension (CD) of device gates but also the CD of interconnect metals is decreasing. Cu is currently used as an interconnect material with a barrier material in a damascene structure because of low resistivity. However, there is a limit in scaling down of Cu width due to the existence of barrier metal, the need for new materials to replace Cu is increasing to overcome this limitation. Molybdenum (Mo) is one of the candidates for next-generation interconnect materials, as it has lower apparent resistivity (electron mean free path × bulk resistivity; λ × ρ₀) than Cu when the critical dimension (CD) is below 10 nm.

In this study, anisotropic atomic layer etching (ALE) of Mo was performed using O₂ and Cl₂ plasma radicals as adsorption gases to precisely control the etch depth of Mo. The etch depth per cycle (EPC) of Mo and hard mask materials such as TiN, Si₃N₄ were compared in the Mo ALE process under Cl₂ and Cl₂:O₂ conditions. The results showed that, by adding O₂ to Cl₂, even though the Mo EPC is slightly decreased, highly selective Mo etching relative to TiN could be achieved. X-ray photoelectron spectroscopy (XPS) showed the formation of volatile MoO_xCl_y on the Mo surface during ALE which can be easily removed during the desorption step.

Gallium Nitride (GaN)-based devices have emerged as key components in optoelectronic and power applications because their wide direct bandgap and high energy conversion efficiency enable superior performance. Light-emitting diodes (LEDs) and power semiconductors, in particular, strongly benefit from these material properties [1]. As the critical dimensions of LED structures continue to scale down, reactive ion etching (RIE) has been widely adopted to form anisotropic profiles during device fabrication. Nevertheless, energetic ion bombardment during the RIE process introduces surface and lattice damage, which degrades both the electrical and optical characteristics of GaN devices. Conventionally, wet etching has been employed as a post-treatment to mitigate this damage; however, its effectiveness is often limited by insufficient damage removal and residual surface defects [2–3].

To address this issue, atomic layer etching (ALE) was employed as a post-RIE treatment to selectively eliminate the damaged surface region in GaN-based structures, including InGaN/GaN multi-quantum well (MQW) layers. Transmission electron microscopy (TEM) images revealed that the MQW layers became indistinct after RIE, indicating structural deterioration. Although subsequent wet etching partially restored the layer contrast, TEM images showed that ALE produced much clearer interfaces, demonstrating more effective damage removal.Photoluminescence (PL) measurements indicated that RIE-induced damage increased as device dimensions decreased, but after the ALE, PL intensity was improved after ALE and became more significantly improved for smaller devices.Overall, as GaN-based LEDs continue to shrink in size, an efficient damage-recovery process becomes increasingly important. Compared with conventional wet etching, ALE offers superior surface restoration and is therefore a more suitable approach for enhancing device performance.

References:

[1]Mohanta, Antaryami, et al. "Observation of weak carrier localization in green emitting InGaN/GaN multi-quantum well structure." Journal of Applied Physics 117.14 (2015).

[2] Ding, Kai, et al. "Micro-LEDs, a manufacturability perspective." Applied Sciences 9.6 (2019): 1206.

[3] Ladroue, Julien, et al. "Deep GaN etching by inductively coupled plasma and induced surface defects." Journal of Vacuum Science & Technology A 28.5 (2010): 1226-1233.

Gallium oxide (Ga₂O₃), featuring an ultra-wide bandgap and a high breakdown electric field, is regarded as a promising material for high-power semiconductor devices; however, its etching process remains challenging in terms of achieving both high precision and low plasma-induced damage. Atomic layer etching (ALE), which employs cyclic self-limiting surface reactions, provides a potential solution for realizing highly controlled and low-damage etching.

In this study, a plasma-based ALE process for Ga₂O₃ using tetrafluoromethane (CF₄) plasma and tin(II) acetylacetonate (Sn(acac)₂) as sequential reactants is developed. The CF₄ plasma half-reaction is first investigated to verify the surface fluorination behavior of Ga₂O₃. The CF₄ plasma is operated at a power of 30 W and a working pressure of 0.5 Torr, with a reaction time of 5 s per cycle. A total of 5 cycles (25 s) and 25 cycles (120 s) are performed to examine the evolution of surface chemical states. X-ray photoelectron spectroscopy (XPS) analysis reveals that the fluorine concentration on the surface increases with the number of plasma cycles (Figure 1). Meanwhile, the intensities of Ga and O signals decrease, indicating a signal attenuation effect caused by the formation of a fluorinated surface layer, which confirms effective surface fluorination of Ga₂O₃ by the CF₄ plasma. After confirming the surface fluorination behavior, Sn(acac)₂ is introduced as the second reactant to complete the full ALE process. During the full reaction cycles, the CF₄ plasma is applied at a power of 50 W with a reaction time of 10 s per cycle, while the process temperature is maintained at 450 °C. The Sn(acac)₂ exposure consists of a 0.1 s pulse followed by a 1 s soak, enabling the conversion of surface fluorides into volatile reaction products and their subsequent removal, thereby completing a single ALE cycle. Following the full ALE process, spectroscopic ellipsometry measurements show a continuous decrease in film thickness with increasing ALE cycle numbers, confirming effective atomic layer etching achieved by the combined CF₄ plasma and Sn(acac)₂ reactions (Figure 2). The average etching rate is approximately 0.105 nm per cycle.

Overall, this work demonstrates a repeatable and precisely controllable plasma-enhanced ALE process for Ga₂O₃, highlighting its potential for applications in ultra-wide bandgap semiconductor devices and advanced manufacturing technologies.

Y-doped HfO₂ (Y:HfO₂) is known to stabilize the orthorhombic/tetragonal (o/t) ferroelectric phase of HfO₂ and widen the ferroelectric thickness window. However, Y-doping in HfO₂ could also affect the thermal ALE of Y:HfO₂. In this study, Y:HfO₂ was etched using sequential exposures of hydrogen fluoride (HF) and boron trichloride (BCl₃). The initial Y:HfO₂ films with thicknesses of 10 nm were deposited on 10 nm of TiN with an underlying W layer of 30 nm. The Y:HfO₂ film thicknesses were then monitored during thermal ALE using in situ spectroscopic ellipsometry (SE) measurements.

The in situ SE measurements showed that sequential HF and BCl₃ exposures at 270 °C resulted in the linear decrease of the Y:HfO₂ thickness. The etch rate was 0.22 Å/cycle during the first 80 ALE cycles. Subsequently, there was a gradual reduction of the etch rate to ~0.15 Å/cycle over the next 60 ALE cycles (Figure 1). X-ray photoelectron spectroscopy (XPS) analysis before and after ALE showed an increase in the Y concentration from ~2 to 4 at.%. This increased Y concentration indicated preferential Hf removal during the ALE cycles. The reduced etch rate was attributed to the gradual Y accumulation on the surface. The Y buildup was consistent with thermochemical calculations that showed low favorability for BCl₃ ligand exchange with YF₃. YF₃ + 2BCl₃(g) → BF₃(g) + YCl₃(g) has an unfavorable ΔG(270 °C) = +41.1 kcal. In contrast, HfF₄ + 4/3BCl₃(g) → 4/3BF₃(g) + HfCl₄(g) has a favorable ΔG(270 °C) = −10.9 kcal.

Atomic force microscopy (AFM) measurements revealed only a minor increase in root-mean-square (RMS) roughness. The initial RMS roughness was 7.0 Å. The RMS roughness increased to only 7.7 Å after removal of 3 nm by etching. Grazing-incidence X-ray diffraction (GIXRD) analysis of the initial 10 nm Y:HfO₂ film showed only metastable o/t phases. After etching removal of 3 nm, GIXRD revealed no phase transformation and reduced peak intensity and peak broadening consistent with decreased crystallite size. In contrast, HZO ALE of a 10 nm HZO film thickness using HF and dimethylaluminum chloride (DMAC) showed transformation from predominantly o/t-phase to monoclinic phase after etching at 250 °C.

Silicon materials such as Si, Si₃N₄, SiO₂, and SiC play a critical role in the semiconductor industry. Selective etching of silicon materials while leaving nearby dielectric materials such as HfO₂ or ZrO₂ intact is a challenge in semiconductor device fabrication. In this study, the thermal etching of silicon materials by chlorination using SOCl₂, SO₂Cl₂, and HCl was explored using quadrupole mass spectrometry (QMS) studies on powders. The measure of etching was the appearance of SiCl4+ ion intensity during the QMS studies. Temperature ramp studies were performed to determine the onset temperature of each etching reaction.

The thermochemistry of etching silicon materials by SOCl₂, SO₂Cl₂, and HCl to produce SiCl₄ is favorable. However, reaction kinetics may dictate the feasible etching temperatures. Si chemical vapor etching (CVE) was demonstrated using SOCl₂, SO₂Cl₂, and HCl exposures above 200°C (Figures 1 & 2). SiO₂ CVE was achieved using SOCl2 exposure above 400°C. Si₃N₄ CVE was observed for SOCl₂, SO₂Cl₂, and HCl exposures above 200°C (Figure 3). SiC CVE was demonstrated for HCl exposures above 200°C (Figure 4). The SiCl4+ ion signal increased progressively at higher temperatures above the threshold temperature for all silicon materials. In contrast, there is selectivity for etching these silicon materials compared with HfO2 and ZrO2. Thermochemical calculations suggest that SO₂Cl₂ and HCl will not chlorinate either HfO₂ or ZrO₂. Etching HfO₂ and ZrO₂ using SOCl₂ requires temperatures above 350°C.

Etching of silicon materials by SOCl₂, SO₂Cl₂, and HCl required the removal of the native oxide on the initial silicon-containing powders prior to exposure to the chlorination precursors. The native oxide was removed by a 5 minute exposure to HF at 400°C. Time-resolved QMS data showed the evolution of H₂O+ and SiF₄+ ion signals during SiO₂ removal. The H₂O+ ion signal decreased during the HF exposure consistent with the native oxide removal. Si, SiC, and Si₃N₄ powders that were not treated with this initial HF exposure did not produce any SiCl₄+ ion signals during the SOCl₂, SO₂Cl2, or HCl exposure.

Atomic layer etching (ALE) of metal oxide thin films is of increasing importance for advanced device fabrication requiring precise thickness control and minimized surface damage. Most reported metal oxide ALE processes rely on halogen-based chemistries and, in many cases, metal-containing etchants to enhance surface reactivity [1-4]. In this work, we report a halogen- and metal-free ALE process for metal oxides enabled by halogen-free liquid phase organic etchants.

The ALE cycle consists of an etching step employing fully organic etchants followed by a reactive removal step that regenerates the surface for subsequent cycles (Fig. 1). The organic etching chemistry exhibits a clear saturation behavior with respect to etchant feed time, resulting in a self-limited etch-per-cycle (EPC) governed by etch-stop formation through surface reaction saturation. During sequential Etchant/Remover ALE, the etched thickness per cycle remains constant once the etchant feed time exceeds a critical window, demonstrating feed-time-limited material removal by the organic etchant (Fig. 2). This saturation behavior was consistently observed for both TiO₂ and Nb₂O₅ thin films.

In addition, a linear reduction in film thickness with increasing ALE cycle number was observed for TiO₂ thin films, demonstrating cycle-by-cycle controllability of material removal (Fig. 3). Similar linear etch behavior is being evaluated for Nb₂O₅ thin films under the same ALE scheme. The reactive removal step oxidatively eliminates or transforms etch-inhibiting surface species formed during the etch step; within the tested process window, the film thickness change is insensitive to the removal-step feed time, indicating that material removal is defined by the etch step rather than the regeneration step.

Unlike many reported metal oxide ALE processes that rely on halogen chemistry and/or metal-containing etchants to enhance reactivity, the present approach achieves self-limited etching through organic reaction saturation and reactive surface regeneration alone. These results demonstrate a halogen- and metal-free ALE approach for metal oxides based on organic reaction saturation, providing a controllable and damage-mitigated etching method for functional oxide materials.

References

[1] S. M. George, J. Vac. Sci. Technol. A 39, 030801 (2021).
[2] S. M. George, Acc. Chem. Res. 53, 1151–1160 (2020).
[3] K. J. Kanarik et al., J. Vac. Sci. Technol. A 35(5), 05C302 (2017).

[4] P. C. Lemaire and G. N. Parsons, Chem. Mater. 29, 6653–6665 (2017).

Atomic layer etching (ALE) is increasingly used in the manufacturing of semiconductor tools as it enables more control over the resulting etching profiles than traditional etching techniques. While different approaches to ALE exist, many of them employ plasmas in one or more steps of the ALE process, be it to use the neutral radicals produced in the plasma to alter the surface or the ions to remove the altered top layer [1].

For such applications, independent control of ion flux and ion energy is highly desirable. Single frequency capacitively coupled discharges (CCPs) do not offer this, as the input power affects both flux and energy. Dual frequency discharges allow this to some extent, but it is limited due to, for example, increased ionization by secondary electrons at high powers of the low frequency. Furthermore, traditional CCPs usually produce bimodal ion energy distribution functions which can cover several 10s to 100s of eV with sharp peaks at either end, so the ion energy cannot be easily limited to a small interval of energies, which is desirable especially for ALE applications, so that the ions remove the top layer of the surface but do not damage the underlying bulk [1].

An alternative approach to achieve this desired control are tailored waveforms. These can range from so-called asymmetric waveforms combining a fundamental frequency with even multiples [2] to non-sinusoidal waveforms typically consisting of sharp voltage peaks [1] followed by a relatively long interval of an almost constant voltage. While it has been demonstrated that these types of CCPs offer independent control of ion flux and energy and/or are able to limit the ion energy to narrow energy intervals, they have not yet been well studied in industrial applications.

This presentation will show our continued efforts to simulate industrial applications of ICPs combined with tailored waveforms to allow precise control of ion energies. The simulations are carried out with the Hybrid Plasma Equipment Model [3]. The simulated results are compared to experimental results with a focus on how the precise shape of the tailored waveform affects the ion energy distribution measured at the electrode for different pressures and ICP powers.

[1] T. Faraz et al. J. Appl. Phys. 128213301 (2020)

[2] U Czarnetzki et al Plasma Sources Sci. Technol.20 024010 (2011)

[3] M. Kushner J. Phys. D42 194013 (2009)

As middle–of–line (MEOL) scaling advances, robust source–drain (S/D) contact formation—especially in self–aligned contact (SAC) schemes—has become critical to device performance, yield, reliability, and power efficiency. At aggressively scaled nodes, small deviations in contact geometry can significantly increase resistance and leakage. A major challenge in current plasma–based etch processes is metal–gate corner loss, which degrades gate integrity and is expected to worsen with further CD shrink.

To address this, we developed an innovative thermal etch process capable of minimizing corner loss while achieving high metal gate selectivity. The method utilizes liquid precursor and HF gas delivered under vacuum in cyclical or co–flow operation. Liquid is vaporized through a heated delivery system, while HF is introduced through the gas manifold; these species react in situ to form a neutral–dominant, chemical etch pathway that selectively removes oxide without physically damaging metal gates. This enables highly selective oxide etching while preserving the sensitive gate corners that are typically eroded in ion–driven plasma processes.

Because thermal etching relies predominantly on neutral–chemistry reactions rather than directional ions, controlling the lateral component of the etch and maintaining profile fidelity are key challenges. To overcome this, we introduced a conformal liner deposition step, which shapes the sidewalls prior to thermal etch, followed by a breakthrough process that removes polymer at the etch front to enable clean thermal–oxide removal. This hybrid approach stabilizes the feature profile, suppresses lateral encroachment, and significantly reduces corner retreat.

To optimize process behavior and quantify the impact of neutral–driven reactions, we modeled the thermal etch using the SEMulator3D (SEM3D) profile simulator. SEM3D was calibrated to wafer data and used to capture lateral–etch tendencies, reaction penetration depth, and sensitivity to liner thickness and thermal–etch conditions. Modeling results guided the development of a combined liner + thermal etch sequence that minimizes lateral loss, improves uniformity, and preserves gate–corner geometry. The SEM3D–based DOE accelerated tuning of precursor flow, sticking behavior, cycle time, and breakthrough parameters, enabling precise control of the etch front.

Across customer wafers and device test structures, the integrated process demonstrated excellent metal–gate selectivity, near–elimination of corner loss, and strong compatibility with existing MEOL integration schemes. By coupling a novel thermal etch chemistry with physics–based profile modeling and a liner–assisted control strategy, this work provides a scalable and manufacturable solution to one of the most critical challenges in future SAC–based S/D contact formation.

Atomic layer deposition (ALD) is the backbone of the modern semiconductor industry, allowing for the deposition of films with atomic scale control for the creation of sub 5 nm nodes. Atomic layer etching (ALE), a relatively more recent process, has proven itself to be just as essential as its ALD counterpart, allowing for the removal of material with atomic scale precision. Despite this, ALE lacks the diversity of precursors afforded to ALD processes due in part to the challenge of modifying a surface while also ensuring the resulting reaction products are volatile and thermally stable enough to be removed cleanly. Herein we demonstrate a method to model and isolate the products of an ALE process, encompassing ZnO and a series of acetylacetonate precursors, utilizing thermogravimetric analysis (TGA) and small scale solvothermal reactions (synthesis).

Acetylacetone precursors with more electron withdrawing functionalities exhibit increased reactivity with ZnO, while acetylacetone precursors with more electron donating functionalities exhibit reduced to no reactivity with ZnO. Isolation and subsequent TGA analysis of the resulting zinc acetylacetonate complexes reveal that more fluorinated complexes are more volatile at temperatures relevant for an ALE process. This approach demonstrates a fast and efficient process to screen precursors that could effectively etch a surface to yield volatile thermally stable byproducts. This bypasses the need for an ALD/ALE reactor in the preliminary studies, and as a result expedites the discovery phase while lowering the associated costs and efforts around ALE process optimization.

Controlled ALE of metallic copper without surface modification has proven challenging, because copper surfaces are relatively chemically inert. But modifying the surface by oxidation (i.e., forming a CuOx surface layer) facilitates the etch step, which consists of removal of the modified layer as a volatile metal complex. Therefore, Cu oxidation is especially important for enabling ALE. We have developed a novel ALE process for Cu metal that is halogen-free and operates at temperatures less than 200°C. This poster describes our in situ characterization of the ALE mechanism and ALE results on Cu thin films.