Use of a custom quartz crystal microbalance vacuum chamber to investigate area-selective atomic layer deposition

Abstract

Atomic layer deposition (ALD) is a technique in which a substrate is exposed to precursor gases, which undergo self-limiting and irreversible chemical reactions, allowing for creation of conformal thin films with particular thickness and stoichiometric composition. ALD has shown tremendous potential for microelectronics fabrication and has been incorporated into various essential applications. In some cases, ALD is the only technique capable of controlling chemical reactions and thin film thickness with atomic precision. In area-selective ALD, one seeks to prevent irreversible growth in areas of the substrate where deposition is not desired. A novel idea to promote selective deposition is to introduce a co-adsorbate with the precursor, which should compete for binding sites and prevent precursor adsorption on one substrate, while allowing deposition on the other substrate to proceed unrestrained. The co-adsorbate molecule is chosen based on its expected binding interactions with dielectric vs. metallic surfaces. We designed and constructed a vacuum chamber incorporating a quartz crystal microbalance (QCM) to explore methods to achieve area-selective ALD. Quartz wafers are manufactured with a specific orientation relative to optical axes within the crystal, and these wafers have an intrinsic vibrational frequency, which can be exploited to measure deposition, adsorption behavior, and desorption behavior on the crystal. Because we could observe the crystal response and edit the process in real time, we were able to study the behavior of individual chemicals on various substrates and rapidly assess how experimental conditions such as chemical dose length, partial pressure, carrier gas flow rate, substrate temperature, and chamber pressure affect ALD experiments. We utilized three different reactant delivery feedthrough configurations to investigate the influence of mixing the chemicals before delivery to the substrate. In the first, the metal precursor, the co-reactant, and the co-adsorbate were pre-mixed. In the second, the chemicals were all delivered independently. In the third, the metal precursor and the co-adsorbate were mixed, while the co-reactant was delivered separately. By comparing similar experiments performed using each of the feedthrough configurations, we discovered that pre-mixing the metal precursor and the co-adsorbate had important consequences on the ALD results.