An idealized ALD process. Precursor 1 (red circles) is introduced (a), forming a self-limiting monolayer (b). The second precursor (white circles) is introduced (c), forming another self-limiting monolayer and forming the first layer of target material (d). This sequence is repeated until the desired thickness is reached (e).
An idealized ALD process. Precursor 1 (red circles) is introduced (a), forming a self-limiting monolayer (b). The second precursor (white circles) is introduced (c), forming another self-limiting monolayer and forming the first layer of target material (d). This sequence is repeated until the desired thickness is reached (e).

Nanotechnology has become ubiquitous, integrated into everything from phones to clothing to sunscreen. As Moore's Law marches on, manufacturing of electronic devices demands increasingly precise nanoscale methods. Manufacturing is approaching the atomic scale in many applications and physical methods are no longer an option for creating devices; we must now turn to methods that leverage the manner in which individual molecules and atoms interact. An excellent example of this approach is atomic layer deposition (ALD), which creates thin films from gas phase precursors. ALD provides the greatest possible control over the thickness of two-dimensional nanoscale structures. The core idea behind this control is to use the intrinsic properties of gaseous chemical compounds to enforce the desired physical architecture: a continuous coating that adheres to all available surfaces and can be grown with sub-nanometer precision. As you might expect, the chemicals used for ALD are necessarily very picky about where and how they react. The aim of this article is to describe in general terms the ALD process, and how precursors are tailored to possess the properties that are necessary.

These precursor properties are best illustrated with the workhorse of the ALD world, the trimethyl aluminum (TMA) and water process for the deposition of aluminum oxide. A commonly invoked model has surface hydroxyl groups on a substrate (say, silicon) react with individual TMA molecules to form a dimethyl aluminum surface species and a gaseous methane byproduct. Once every available surface site has reacted in this manner, the first monolayer is complete. Furthermore, since there are no longer any reactive surface species, no additional growth is possible. This self-limiting growth is the defining characteristic of the ALD process. This of course must also be true for your next reagent (water, in the current example). Water can be introduced without fear of gas phase reactions precipitating alumina powder once any excess TMA (and methane) has been removed from the reaction chamber (typically by vacuum). A water pulse then reacts with every available surface species creating another self-limiting monolayer in the form of a single atomically-thick, hydroxyl-covered aluminum oxide film. This step re-creates the reactive sites for the next pulse of TMA so that the cycle may continue. This system illustrates the essential characteristic of any ALD process: stepwise, self-limiting reactivity.

This black-and-white view of the reactions that occur during an ALD process do not of course always match the reality, but it does provide a framework for thinking about how you would go about inventing a new process. Your precursors (two or more) ideally have very particular characteristics: high volatility, reactivity at the surface and not in the gas phase, and surface reactivity that produces a self-limiting monolayer. These demands easily generate your precursor design checklist: low mass, weak intermolecular interactions, stability in the gas phase, reactivity at the surface. Some of these criteria are easier to satisfy than others. Ligands with branching organic groups and without aromatic rings that might participate in pi-stacking can help to reduce your intermolecular interactions, for example. Saturating the coordination sphere of your metal centre can minimize (but not necessarily eliminate) the formation of heavy dimers which might reduce volatility. There are many excellent examples in the literature of how a simple ligand system, such as amindinates or guanidinates, can be modified to tune the stability and volatility of a precursor.

Verification of a true ALD process (as opposed to continuous growth as seen with chemical vapour deposition or physical methods) is typically established with a “saturation curve”. A series of trials are performed keeping the amount of one precursor constant and varying the other. When the growth rate (thickness divided by number of cycles) plateaus as the amount of precursor used increases, self-limited growth is demonstrated. The experiment is then repeated with the roles of the precursors reversed, to find the minimum amounts for each precursor in the process. Watching the process from the inside is much more interesting however: there are several clever ways to observe the chemistry of ALD in-situ. Ellipsometry can observe film growth and characteristics such as density in real time. A quartz crystal microbalance can produce a timeline of mass gains and losses with respect to each precursor, providing excellent insight into surface reactions. Just beyond the reaction chamber, exhaust gases can be analyzed by mass spectroscopy to observe byproducts to help discern the chemistry happening at the surface. The self-limiting nature of the process can be interrogated by these techniques, providing excellent data for modelling and understanding the chemistry of the reactions.

A wide range of materials can be deposited by ALD, for a wide range of applications. Many oxides, ternary materials, and metals (including noble metals) are known. Challenges remain, of course; new inorganic and organic material processes await discovery. Metals such as copper tend to grow polycrystalline films with poor conductivity, whereas epitaxial films might be preferred.  Faster growth rates for known materials are always welcome as well. There is also much work to be done discovering the mechanism by which these processes operate; understanding these mechanisms will enable better precursor design. Control over reactivity at the surface can be further complicated by techniques used to impose patterns; organic masks and the like. So there is much to explore and discover, and as a fairly young process which has demonstrated unparalleled uniformity and thickness control for the deposition of the thinnest possible films, ALD will be a mainstay in nanofabrication for decades to come.