Past exam of the MPhil in micro- and nanotechnology enterprise course of the University of Cambridge / 2022 / Paper 2 / Section A(i) / 14

In atomic layer deposition, expose the surface to a molybdenum precursor, purge unreacted precursor and volatile products, expose it to a sulfur reactant, and purge again. Each half-reaction should saturate the available reactive sites rather than continue indefinitely with dose. Repetition builds a molybdenum disulfide film, with nucleation, incubation and layer closure checked independently.

Self-limited surface chemistry is not the same as depositing exactly one complete crystal monolayer per cycle. A cycle may add only a fraction of a layer, and initially isolated islands can thicken before all bare substrate is covered. Reactions on hydroxylated silicon dioxide can also differ from those on the comparatively inert MoS${}_2$ basal surface. A precursor pair must actually show saturation under the chosen conditions; simply alternating gases is insufficient evidence of ALD.

In metalorganic chemical vapor deposition, metalorganic and sulfur-containing precursors are typically supplied simultaneously or with overlapping residence times, and film growth depends on decomposition, transport and surface reaction rates. Both methods involve chemistry, but ALD separates saturating half-reactions and purge steps. An ALD process need not use a metalorganic precursor, and pulsed CVD is not automatically ALD.

Potential advantages are reproducible thickness increments, large-area uniformity, conformal access to features and compatibility with moderate-temperature integration. Saturated exposure can make coverage less sensitive to small precursor-flux variations than an unsaturated deposition rate. The cycle count provides a useful process variable for matching an optical or electrical thickness target.

For a monolayer device, control of nucleation and coalescence is especially valuable, because excessive second-layer growth and uncovered patches both change the optical response. The advantages are conditional: a reliable ALD window and suitable precursor chemistry are required, and conformality does not imply a single-crystal film. Long exposure and purge steps reduce throughput, residual ligands or impurities can remain, and post-growth crystallization may require temperatures that the integrated device cannot tolerate.

Many independently oriented nuclei form on an amorphous support and then coalesce before they can become one large crystal. The SiO${}_2$ substrate provides no unique epitaxial orientation. At relatively low deposition temperature, short surface diffusion lengths and slow bond rearrangement restrict the distance over which deposited species can find a well-ordered configuration. Island collisions therefore leave grain boundaries; precursor residues and incomplete sulfurization can add disorder.

Raising temperature or annealing can improve crystallinity, but may conflict with the ALD reaction window, substrate compatibility or sulfur retention. Small grains are consequently a frequent process outcome, not a fundamental theorem that ALD can never form a well-ordered layer. A direct ALD experiment on SiO${}_2$-covered wafers found nanocrystalline MoS${}_2$ by diffraction and electron microscopy, while demonstrating self-limited precursor exposure behaviour.^[ref]

The pressure is $10^{-8}\mathrm{mbar}=10^{-6}\mathrm{Pa}$. The molecular arrival flux from an ideal gas isBecause the residual-gas composition, temperature and sticking probability are not specified, the answer is an order-of-magnitude monolayer-equivalent exposure time, not a unique coverage prediction. Take $T=300\mathrm{K}$, a nitrogen-like molecular mass $m=28u=4.65\times 10^{-26}\mathrm{kg}$, and unit sticking probability. ThenUse the given $0.5\mathrm{nm}$ molecular-layer scale as an approximate intermolecular spacing, giving $N_s\sim (0.5\mathrm{nm}{)}^{-2}=4\times 10^{18}{\mathrm{m}}^{-2}$. ThusFor $s=1$, this is a few minutes; assuming instead $10^{19}{\mathrm{m}}^{-2}$ sites gives about $350\mathrm{s}$, the same rough scale. A clean-looking vacuum pressure does not imply a contamination-free surface. Desorption and a small sticking probability can greatly lengthen net accumulation, while residual water or hydrocarbons can dominate the chemistry. Full saturation in a site-blocking adsorption model is not exactly the same as one monolayer-equivalent incident fluence.

Use a thin, continuous cap of hexagonal boron nitride, preferably assembled without exposing the cleaned MoS${}_2$ surface to ambient contamination. It is chemically inert and relatively transparent in the visible range, and can isolate the film from adsorbing water, oxygen and hydrocarbons during optical measurements. A transferred sheet avoids depositing a strongly reactive precursor directly on the monolayer.

The cap must cover the illuminated region and overlap its edges; pinholes, wrinkles and unsealed edges can allow ingress. Trapped residues or bubbles can themselves change the optical response. Record the capped background and account for optical interference and dielectric-environment shifts rather than assuming that encapsulation leaves every spectral peak unchanged. A thin oxide or polymer coating may also provide protection, but deposition damage, optical background and permeability require checking. An atomically thin sheet with a perfect basal plane is not proof that the whole assembled enclosure is sealed.