Atomic Layer Deposition & Etching

• Atomic Layer Deposition and Etching (ALD/E) are self-limiting, cyclic techniques that enable atomic-scale deposition and removal with precise thickness control.
• They employ alternating exposure to gaseous precursors to produce conformal, uniform films essential for applications in electronics, optics, and quantum devices.
• Integration of ALD and ALE enhances surface engineering by enabling defect removal, passivation, and improved performance in high-precision device fabrication.

Atomic layer deposition and etching (ALD/E) are self-limiting, cyclic, surface-chemical techniques for precise, atomic-scale thin film fabrication and removal. By leveraging alternating exposure to carefully chosen gaseous precursors, these methods enable the growth or etching of conformal, uniform layers with sub-nanometer control over thickness, chemical composition, and interface properties. ALD/E plays a foundational role in the manufacturing of advanced electronic, optical, and quantum devices, where exacting material specifications, defect and interface engineering, and large-area uniformity are critical.

1. Fundamental Principles of ALD and ALE

Atomic layer deposition is characterized by alternating pulses of a metal–organic precursor and a co-reactant (commonly H₂O, ozone, or a plasma species). Each half-cycle proceeds via self-limited chemisorption, saturating available reactive surface sites and resulting in a digitally controlled addition—typically a fraction of a monolayer—per cycle. The general ALD sequence can be represented by:

1. Adsorption: Metal precursor chemisorbs on available surface functional groups (e.g., hydroxyls).
2. Purge: Unreacted precursor and byproducts are removed.
3. Reaction: A second reactant transforms the chemisorbed species into the target material and regenerates functional surface sites.
4. Purge.

This cyclic process enables high conformality, pinhole-free coverage, and atomic-scale thickness control even over high-aspect-ratio geometries, as confirmed by ellipsometric mapping and thickness uniformity on 100–300 mm wafers (Bulkin et al., 2019, Kennedy et al., 2021).

Atomic layer etching (ALE), the inverse process, alternates a surface modification step (e.g., fluorination) with a ligand-exchange or volatilization step to remove material in a similarly self-limited fashion. Each cycle etches a fixed amount (typically 0.5–1.2 Å/cycle for Al₂O₃) (Hennessy et al., 2017, Mahuli et al., 20 Jun 2025). Process chemistry, such as TMA/HF for Al₂O₃, is chosen to enable layer-by-layer removal with strict monolayer control, surface selectivity, and minimal roughness.

2. Process Chemistry and Surface Reaction Mechanisms

ALD and ALE reactions exploit thermodynamically favorable, surface-limited chemisorption and ligand exchange. For Al₂O₃, a canonical ALD cycle utilizes trimethylaluminum (TMA) and H₂O as follows:

• C–half-cycle:

$$\mathrm{AlOH^* + Al(CH_3)_3(g) \rightarrow AlOAl(CH_3)_2^* + CH_4(g)\uparrow}$$

• D–half-cycle:

$$\mathrm{AlOAl(CH_3)_2^* + H_2O(g) \rightarrow AlOH^* + CH_4(g)\uparrow}$$

Each ALD cycle deposits ∼1 Å (monolayer limit: 3 Å for AlO₁.₅) of conformal Al₂O₃ (Henning et al., 2021, Bulkin et al., 2019). Self-limitation ensures saturation with respect to pulse duration and pressure within operational windows.

ALE for Al₂O₃ typically combines:

• A–half-cycle (HF fluorination):

$$\mathrm{Al_2O_3^* + 6\,HF(g) \rightarrow 2\,AlF_3^* + 3\,H_2O(g)\uparrow}$$

• B–half-cycle (TMA methylation/volatile removal):

$$\mathrm{2\,AlF_3^* + 4\,Al(CH_3)_3(g) \rightarrow 6\,AlF(CH_3)_2(g)\uparrow}$$

Process rates are typically 0.5–1.2 Å/cycle, strongly temperature- and exposure-dependent (Hennessy et al., 2017, Mahuli et al., 20 Jun 2025). Surface conditioning agents such as LiF facilitate conformal, uniform etching over large areas (Hennessy et al., 2017).

3. Reactor Design, Kinetics, and Uniformity Control

Reactor-scale phenomena impose additional constraints on ALD/E. Uniform precursor distribution, pulsed dosing times, purge efficiency, and kinetic parameters—especially sticking probability ($\beta$)—directly influence spatial film uniformity, cycle throughput, and overall process reproducibility (Yanguas-Gil et al., 2021). Plug-flow and CFD reaction–diffusion models, incorporating Langmuir kinetics and detailed transport, predict:

• High $\beta$ ($10^{-2}$): sharp growth fronts, <1% thickness variability over 300 mm wafers at modest dose times (≥0.2 s).
• Low $\beta$ ($10^{-3}$ or below): broadened, less-uniform coverage requiring longer exposures or higher pressures.

Non-idealities such as soft-saturating pathways and competitive byproduct adsorption result in incomplete site coverage and increased thickness gradients, even at long overall dose times (Yanguas-Gil et al., 2021).

Table: Comparison of Ideal vs. Non-Ideal Kinetic Effects on Uniformity

Kinetic Regime	Mechanism	3σ Thickness Variability (300 mm)
Ideal ($\beta=10^{-2}$)	Fast, saturation	<1%
Slow/soft-saturating ($f=0.1$)	Mixed kinetics	>2%
Byproduct site-blocking	Incomplete sites	~25%

QCM and QMS in situ diagnostics, interpreted via such models, distinguish process-limiting surface vs. transport steps (Yanguas-Gil et al., 2021).

4. Advanced ALD/ALE Process Engineering and Monolayer Limits

Precise interface and monolayer control is central for applications in nanoelectronics, optics, and catalysis. For GaN, plasma-assisted ALD achieves true pinhole-free, 3 Å AlOx monolayers via an oxidant-free TMA/H₂-plasma process, exploiting the self-limiting conversion of native Ga₂O₃ (Henning et al., 2021). The three-step kinetic regime—rapid exponential site conversion, zero growth on closure, then slow defect-driven accumulation—enables unparalleled interface abruptness.

AlOx monolayer functionalization modifies GaN work function by $-0.38$ eV and enables densest possible self-assembled monolayers, critical for charge-transport and sensor applications (Henning et al., 2021). A plausible implication is that such strategies generalize to other III–V semiconductors with reactive native oxides.

Process windows for both ALD and ALE must balance temperature (225–300 °C for TMA/HF cycles), precursor dosing, and surface conditioning (LiF, plasma) to prevent over-etch, surface roughening, or gas-phase parasitic reactions (Hennessy et al., 2017, Henning et al., 2021).

5. Integration of ALD and ALE for Functional Device Surfaces

The combined application of ALE and ALD as in situ surface engineering steps yields devices with improved functional metrics. In aluminum-based superconducting qubits, an ALE–ALD sequence (50 cycles ALE, 10 cycles ALD at 300 °C) doubles transmon $Q$ and $T_1$, verified by median values of $Q_{\text{eff}}=3.69\times10^6$ and $T_1=196\ \mu$s maintained long-term (Mahuli et al., 20 Jun 2025).

Mechanistically, ALE removes not only the native oxide but also polymeric and oxyhydroxide defect species, while ALD provides a dense, conformal, stoichiometric Al₂O₃ passivation. This reduces two-level system (TLS) defect density at metal/air and substrate/air interfaces, halving single-photon TLS loss in microwave resonators. Such encapsulation impedes regrowth of high-loss native oxides and preserves device performance for many months (Mahuli et al., 20 Jun 2025). Similar conjunctions of ALE (controlled removal) and ALD (precise encapsulation) are critical in UV optics, yielding reflectance improvements in Al mirrors across the 120–400 nm band through sequential oxide removal and AlF₃ capping (Hennessy et al., 2017).

6. Applications and Performance Criteria

ALD/E-based surface treatments are foundational in multiple domains:

• Optics: ALD-protected Ag mirrors (>15 nm Al₂O₃ by thermal ALD, $GPC\approx0.10$ nm/cycle at 150 °C) maintain >95% reflectivity in 300–2500 nm even after oxygen plasma exposure. Minimum protective thickness ≥15 nm required for reliable barrier performance (Bulkin et al., 2019). UV mirrors combining ALE and ALD (AlF₃) achieve reflectance up to 94% at 300 nm and improvements of 4–6% in far-UV (Hennessy et al., 2017).
• Quantum Devices: ALD-applied Al₂O₃ (50 nm films) used as strain-relieving buffers in silicon/superconductor heterostructures narrows strain distribution, reduces spin-resonance linewidths, and leaves spin coherence time $T_2\approx23$ ms unchanged, with no detected electric/magnetic noise from the dielectric (Kennedy et al., 2021). ALE–ALD treated transmons reach $T_1>400~\mu\text{s}$ and $Q_{\text{eff}}>9\times10^6$ in best cases (Mahuli et al., 20 Jun 2025).
• Nanoelectronics: Monolayer ALD (3 Å AlOx on GaN) enables maximal self-assembled monolayer densities and tunable work functions (Henning et al., 2021). Conformal ALE removes native/interfacial oxides for low-resistance ohmic contacts and patterned nanostructures (Hennessy et al., 2017).

A plausible implication is that monolayer-precision ALD/ALE strategies underpin continued device scaling and new interface functionalities across semiconductor, photonic, and quantum technologies.

7. Challenges, Non-Idealities, and Future Directions

ALD/E scalability and fidelity depend critically on understanding and mitigating process non-idealities. Soft-saturating sites, competitive byproducts, and kinetic limitations degrade uniformity and reproducibility, as evidenced in 3σ variability metrics for large wafers (Yanguas-Gil et al., 2021). Reactor engineering—including flow path design, injector configuration, and thermal management—must be tightly coupled to multiscale kinetic modeling for robust process translation from laboratory to high-volume manufacturing.

Further work is required to elucidate failure mechanisms in sub-15 nm barrier ALD films for Ag mirrors (Bulkin et al., 2019), optimize ALE/ALD temperature/purge windows to balance throughput and minimal roughness (Hennessy et al., 2017), and generalize monolayer ALD approaches to broader material classes at industrial scale (Henning et al., 2021). Open-source frameworks and in situ diagnostics are rapidly evolving as essential tools for rapid process characterization and scale-up (Yanguas-Gil et al., 2021).

A final consideration is the sustained integration of ALE and ALD sequences as composite surface engineering protocols—simultaneously removing interfacial contaminants and reconstructing ideal dielectric or passivation layers—for all next-generation nanoscale devices (Mahuli et al., 20 Jun 2025).