Simulation approach of atomic layer deposition in large 3D structures

Abstract

We present a new simulation method predicting thicknesses of thin films obtained by atomic layer deposition in high aspect ratio 3D geometries as they appear in MEMS manufacturing. The method features a Monte-Carlo computation of film deposition in free molecular flow, as well as in the Knudsen and diffusive gas regime, applicable for large structures. We compare our approach to analytic and simulation results from the literature. The capability of the method is demonstrated by a comparison to experimental film thicknesses in a large 3D structure. Finally, the feasability to extract process parameters, i.e. sticking coefficients is shown.

Sticking probabilities of H2O and Al(CH3)3 during atomic layer deposition of Al2O3 extracted from their impact on film conformality

“…As recently reviewed by Cremers et al, 11 several high AR structures have been used in the literature to assess the conformality of ALD processes, such as vertical trenches, [16][17][18][19] pillars, [20][21][22][23] and porous materials. 24,25 Alternatively, lateral structures can be employed 6,7,26,27 which allow for top-view diagnostics to easily and accurately quantify the conformality and properties of the deposited film. The microscopic lateral-high-aspect-ratio (LHAR) trenches developed by Puurunen and co-workers, 7,14,28,29 named as PillarHall ® technology, have been adopted in this work.…”

“…Comparing experimentally obtained thickness profiles with simulated profiles can provide the initial sticking probability s 0 of the used ALD reactant, as shown for instance by Rose and Bartha 12 and Ylilammi et al 14 Here, s 0 indicates the sticking probability toward the initial surface on which all reaction sites are still available. In several models, e.g., ballistic, 8,31-35 continuum, 14,[36][37][38] and Monte Carlo, 9,10,12,27,30,[39][40][41][42] it is observed that s 0 governs the slope of the leading front of the thickness profile. In this work, the relation between the slope of the profile front and the initial sticking probability has been quantified using the continuum model reported by Yanguas-Gil and Elam 36 to directly determine sticking probabilities from ALD thickness profiles without further modeling.…”

Monte Carlo and Kinetic Monte Carlo Models for Deposition Processes: A Review of Recent Works

“…Schwille et al [54,55] proposed a new MC method to study the deposition inside micro-structures. Their direct simulation MC-derived method models only the walls of the geometrical structure, rather than partitioning the entire domain into grid cells.…”

“…Their direct simulation MC-derived method models only the walls of the geometrical structure, rather than partitioning the entire domain into grid cells. This assumes that deposited particles can only collide with the molecules of the carrier gas and can only interact with the walls of the structure [55]. The simplification drastically decreased the computational demands.…”

Multiscale computational fluid dynamics modeling of thermal atomic layer deposition with application to chamber design

“…For the gas-phase model, traditionally, analytical solutions of the mass, momentum and energy conservation equations are obtained to characterize the macroscopic gas-phase spatio-temporal evolution in simplified geometries and with very low computational demand. For example, 2 [15] assumed a 2D vertical flow gas-phase profile, and [16] constructed the reactor chamber geometry with a simple array of equally sized square-shaped cells. However, the first-principles models with simplified assumptions fail to provide meaningful results that are applicable to the industrially-relevant ALD systems, which often involve complicated and detailed reactor geometry designs.…”