Atomic Level Processing

Many of today’s high-tech devices are manufactured by processing materials at the nanoscale. Prominent examples include computing, data storage and communications devices, sensors, solar cells and batteries. Making devices smaller, more powerful and more energy-efficient means developing new patterning, deposition and etch techniques at ever finer resolution, in some cases down to just a few atoms thick. Atomic layer deposition (ALD) is a processing technique that can achieve this level of control through self-limiting surface chemistry, delivering the required conformality, uniformity and purity. The search is now on to find ALD processes for new materials. Related chemistries for atomic layer etch (ALE), multiple patterning and substrate-selective deposition are also being developed at present, opening up the prospect of atomic-level control of all steps of device fabrication.

Finding and optimizing ALD chemistries in the lab is challenging and time-consuming, and researchers are now turning to computer simulations to accelerate the discovery process and give a deeper understanding.¹ The Schrödinger suite of software for atomic-scale simulation is particularly suited to tackling this problem, as the following case studies illustrate.

Automated Precursor Screening
Precursor chemicals must be carefully designed so as react on surfaces with the atom-by-atom control of ALD and ALE, while at the same time being volatile and stable enough for delivery. Organometallic complexes can be used as precursors for metals or metal oxides, as long as the ligands react in a clean and self-limiting fashion. A vast range of possible ligands can be proposed for each metal. However, synthesis of the complexes in the lab can be time-consuming, complicated by the fact that many complexes react violently in air. We therefore turn to computational chemistry to screen larger numbers of chemicals than can ever be practically synthesized, narrowing down the chemical space and discovering the most promising candidate molecules for synthesis. Schrödinger provides easy-to-use software for automated screening of hundreds or thousands of precursor gases.² A ligand library can be custom-built and systematically extended into novel chemistries. The library is then used to enumerate all possible precursor complexes, including the heteroleptic precursors with mixtures of ligands that allow fine-tuning of saturation and selectivity. For example, with tetravalent titanium, a library of just 8 ligands can form 330 distinct complexes (Figure 1). Efficient quantum chemical calculations are then carried out with Schrödinger’s Jaguar code, producing high quality energetic and structural data for assessing the viability of each precursor. 

Figure 1: Automated enumeration of heteroleptic precursors: a sample of the 330 distinct structures enumerated when eight ligands are combined around a tetravalent metal center (purple=Ti, grey=C, white=H, red=O, blue=N, green=F). Feeding the enumerated structures into quantum mechanical calculations allows them to be screened for properties of interest.

Linking Surface Reactions to Process Chemistry
Self-limiting reactions between precursor gases and a substrate are key to the atomic-level control of ALD. It is important to understand which reactions can occur at useful temperatures and pressures and whether there are competing reactions that lead to impurities or non-uniformity. Quantum mechanical slab calculations are a powerful way to investigate these surface reactions, quantifying how the atomic structure and energetics change from one reaction step to the next. Figure 2 shows a simulation of postulated reactions for the ALD of gallium oxide from trimethylgallium. Experimentally, this process is much more problematic than the well-known aluminum oxide analogue³, and DFT simulations of the reaction pathway can reveal the reasons why. Slab models, where surfaces are repeated periodically in 3D, are readily generated with the Schrödinger Materials Science suite and computed with the Quantum Espresso periodic DFT code. Schrödinger workflows provide a straightforward way to find out how the reaction pathway is affected by chemical modifications, such as alternative ligands. Properties like film growth per cycle and FTIR vibrational frequencies are predicted and compared with experiment.⁴

Figure 2: Surface structures during postulated reactions for the ALD of gallium oxide (yellow=Ga, red=O, white=H, gray=C). (a) The hydroxylated gallium oxide surface provides reactive sites for (b) the chemisorption of trimethylgallium, from which (c) the by-product CH₄ can be eliminated. Successive chemisorption and elimination reactions lead in the ideal case to (d) a surface saturated with methyl groups, so that the reaction self-limits.