Depositing uniform ALD gate oxides for 2D materials is challenging because surfaces such as graphene and molybdenum disulfide (MoS2) have few dangling bonds. Conventional atomic layer deposition (ALD) may therefore nucleate preferentially at defects and step edges rather than across the full surface. The resulting nonuniform dielectric can compromise insulation and gate control. A low-temperature process called nanofog ALD offers a different approach: it creates oxide nucleation sites during deposition, without a separate surface treatment or seeding layer.
Why oxide nucleation is difficult on 2D materials
ALD builds films through alternating precursor exposures and purges. On reactive surfaces, these cycles can support controlled, conformal growth. But the relatively inert basal planes of 2D materials provide fewer sites for the initial precursor reactions. As a result, oxide may form at edges and defects while leaving broader terraces less covered during early growth.
Surface functionalization, reactive metal layers and polymer-based seeds have been used to address this limitation. However, chemical treatments can affect the underlying material, while some seeding approaches add process complexity or dielectric thickness. The nanofog method was developed to initiate Al2O3 growth directly on 2D surfaces without those additional steps.
How nanofog ALD creates nucleation sites
The process deposits Al2O3 using trimethylaluminum (TMA) and water at low temperature. By controlling the time between precursor pulses, the researchers intentionally introduce a chemical vapor deposition (CVD) component alongside ALD. The proposed mechanism is that precursor reaction in the gas phase generates very small AlOx particles. These particles can adsorb across the substrate, including on surfaces that otherwise offer few nucleation sites, and provide centers for subsequent oxide growth.
This combination is termed nanofog ALD. It differs from conventional ALD because gas-phase particle formation is part of the growth strategy. Pulse and purge conditions influence particle formation and, in turn, the resulting surface morphology. The study reports that longer purge times produced smoother surfaces, consistent with smaller particles.
AFM observations showed continuous Al2O3 coverage on highly oriented pyrolytic graphite (HOPG) and MoS2 at 50 °C. At higher temperatures, deposition on HOPG became concentrated at step edges and defects; MoS2 remained continuously covered at 80 °C but not at 100 °C. The authors attributed the difference between the materials to their surface polarizability and the resulting interaction with the oxide nuclei. These results underline that deposition conditions may need to be tuned to the substrate.
Conformality and electrical performance
To assess coverage on a three-dimensional structure, the team deposited a thin Al2O3 layer on patterned Si3N4/SiO2/Si fins with an aspect ratio of 5:1. A film about 2 nm thick had reported step coverage of 91%. This conformality supports the potential of the process for coating high-aspect-ratio features, although the test structure was not itself a 2D material device.
Electrical testing used metal-oxide-semiconductor capacitors (MOSCAPs) fabricated on MoS2, HOPG and Si0.7Ge0.3. For 50 cycles of Al2O3 deposited at 50 °C, the reported maximum capacitance was approximately 1.1 µF/cm2 across the substrates. Similar capacitance values were interpreted as evidence of comparable oxide growth without a substantial nucleation delay. The measured leakage currents on MoS2 and HOPG were also consistent with uniform, insulating oxide coverage in the tested capacitors.
The interface-state results are especially relevant to device integration. For the Al2O3-only MOSCAPs, the reported density of interface states, or Dit, was 9.84 × 1011 eV−1cm−2 on MoS2, compared with 6.89 × 1012 eV−1cm−2 on Si0.7Ge0.3. In the bilayer-stack measurements, Dit on MoS2 was also lower than on Si0.7Ge0.3. The authors associated this difference with van der Waals interactions at the oxide–MoS2 interface, rather than covalent bonding that can create dangling bonds.
Building higher-capacitance bilayer stacks
To increase capacitance, the researchers used low-temperature Al2O3 as a nucleation layer before depositing HfO2 at 300 °C. The resulting Al2O3/HfO2 bilayer reached a reported maximum capacitance of about 2.2 µF/cm2 with nickel gates, roughly twice the value for the Al2O3-only capacitors. The study also tested Pd/Ti/TiN gates. Because titanium-containing gates can scavenge oxygen from oxide stacks, this configuration was evaluated for its effect on the measured capacitance. With these gates, the reported maximum capacitance was approximately 2.7 µF/cm2 on MoS2 and HOPG, without a reported loss in leakage performance.
Practical implications for process integration
For engineers evaluating gate dielectrics on 2D semiconductors, the central process insight is that initial nucleation need not rely on chemically modifying the substrate. Controlled gas-phase formation of AlOx nuclei can provide a route to continuous Al2O3 coverage, while that layer can enable subsequent HfO2 deposition.
- Process tuning matters: Pulse and purge timing influence particle formation and surface roughness.
- Substrate behavior differs: HOPG and MoS2 showed different temperature limits for continuous coverage.
- Stack design affects capacitance: Adding HfO2 increased measured capacitance, while the gate-metal choice also influenced the result.
- Device topography remains important: The study linked higher leakage in MoS2 bilayer capacitors to tall step edges, where oxide coverage may be less conformal.
Nanofog ALD is therefore a promising deposition strategy, not a guarantee that every 2D material, geometry or process recipe will behave identically. Its practical value lies in combining direct deposition on inert surfaces with a demonstrated path to thin, conformal oxide layers and lower measured interface trap density on MoS2. Process developers can use these findings to guide evaluation of gate-stack nucleation, substrate-specific conditions and device topography.