AVS 70 Session TF1+AP-ThM: Thin Films: Fundamentals of ALD

The deposition of thin solid films is central to many industrial applications, and, thanks to the isotropic nature of the adsorption of chemical species (which affords even coverages on surfaces with rough topographies), chemical vapor deposition (CVD) methods are particularly useful for this task.Moreover, by splitting the overall film-depositing reactions into two or more complementary and self-limiting steps, as it is done in atomic layer depositions (ALD), film thicknesses can be controlled down to the sub-monolayer level.Thanks to the availability of a vast array of inorganic and metalorganic precursors, CVD and ALD are quite versatile and can be engineered to deposit virtually any type of solid material.However, the surface chemistry that takes place in these processes is often complex and can follow undesirable side reactions leading to the incorporation of impurities in the growing films.Appropriate precursors and deposition conditions need to be chosen to minimize these problems, and that requires a proper understanding of the underlying surface chemistry.

In this work we have examined explicitly how the choice of ALD chemistry (precursor and co-reactant) can affect the process temperature window, composition of the deposited thin film, and possible modification of the underlying substrate.Growth of Al₂O₃ from Al(CH₃)₃ (TMA) and H₂O may be the most studied and robust ALD process.Here we consider two perturbations from this benchmark process by considering both a different precursor (a heteroleptic Al complex, void of halogens and Al-C bonds, L′₂AlL″), and a different organic oxidizing compound (RO_x).We have employed a quartz-crystal microbalance (QCM) to monitor ALD in situ and in real-time, and the deposited thin films have been characterized using X-ray photoelectron spectroscopy, spectroscopic ellipsometry, and X-ray reflectivity.Growth of thin films of Al₂O₃ using TMA and either RO_x or H₂O as the co-reactant at T = 285 °C produces thin films of similar physical properties (density, stoichiometry, minimal carbon incorporation), and the growth rate per cycle is similar for the two co-reactants at this temperature.At a lower temperature of T = 120 °C, the behavior is starkly different, where growth occurs with H₂O but not with RO_x.At either process temperature, we find no evidence for significant coverages of a long-lived intermediate from the reaction of RO_x. Growth from the heteroleptic Al precursor is similar to that observed with TMA—essentially stoichiometric Al₂O₃ is formed, and growth occurs readily with H₂O at T = 120 and 285 °C, while there are differences in the species formed at the end of the precursor half cycle.Growth with the heteroleptic precursor and the organic oxidizing species exhibits a similar dependence on temperature as observed with TMA—deposition at T = 285 °C, but not at T = 120 °C.Deposition of thin films of Al₂O₃ on metal surfaces of Cu and Co has been examined for evidence of interfacial oxidation.^[1] While growth with either co-reactant (H₂O or RO_x) does not lead to the oxidation of the underlying Cu substrate, use of H₂O leads to the oxidation of Co, while use of RO_xas the co-reactant does not. We postulate that the size of the co-reactant in this case plays a role, where the diffusion of species through the deposited Al₂O₃ thin film could result in oxidation at the Al₂O₃|metal interface, a process that is strongly hindered in the case of RO_x due to its size.

[1]ACS Appl. Mater. Interfaces 16, 16983−16995 (2024).

Close-proximity atmospheric-pressure spatial atomic layer deposition (AP-SALD) is a powerful technique to control nanoscale surfaces and interfaces at scales relevant to large-scale manufacturing. To investigate the manufacturing tradeoffs in an AP-SALD system (film quality-throughput-cost-sustainability), many studies utilize computational modeling to gain insight into the difficult to observe process region that contains the isolated precursor zones. However, there is limited investigation into the effects of process parameters on the spatial separation and distribution of the precursor zones and their resulting impact on the deposited film. Additionally, few previous works have paired a computational model with an experimental system that has comparable, closed-loop control over the many system parameters, including gap size, relative alignment, gas flow rates, and substrate velocity.

In this work, we introduce a transport-reaction model to study how variations in process parameters affect the resulting deposited film. A finite element method model uses laminar flow fluid mechanics and transport of dilute species to give insight into the pressure, velocity, and partial pressure gradients within the process region. These outputs are then input into a surface reaction model that simulates the chemical reactions, yielding predicted film thickness maps. The model is validated and parameterized by comparing the simulated film geometries with TiO₂ films deposited using a mechatronic AP-SALD system. By coupling an experimental system with the transport-reaction model, process parameters can be tuned both physically and digitally to provide new insights into AP-SALD process control and optimization for thin film deposition at large scales.

Many of the properties that are prized in films deposited with atomic layer deposition (ALD) such as precise control of composition and thickness, as well as conformal coverage on high aspect ratio structures, rely on the self-limiting nature of ALD surface reactions.A deeper understanding of the fundamental thermodynamics, kinetics, and mechanisms of ALD precursor reactions is essential to enable emerging applications such as site-selective ALD and area-selective ALD, which rely on chemical selectivity between precursor molecules, inhibitors, and the substrate surface.Differences in the surface composition and availability of reactive sites, as well as the precursor ligands can have a significant influence on the properties of the deposited film and the formation of interfaces, even within a given material system.

ALD pyroelectric calorimetry measures the in situ, time-resolved heat generated by ALD surface reactions under standard ALD conditions. Previously, we used this technique to measure the heat generation rates of trimethylaluminum (TMA), tetrakis(dimethylamino)zirconium (TDMAZr) and water to form Al₂O₃ and ZrO₂.Here we extend this technique to study the reaction between bis(ethylcyclopentadienyl)magnesium (Mg(CpEt)₂) and water to form MgO.To fully extract the heat generated by these ALD surface reactions over longer time scales, the data was fit to a calibrated model to account for cooling and electronic decay.These results were then compared to computational first principles models of various surfaces and reaction steps as well as calculations of the standard reaction enthalpies.Pyroelectric calorimetry provides a unique opportunity to quantitatively benchmark the values provided by computational models.This can help evaluate the accuracy of proposed processes and mechanisms.Heat generation rates provide new insight into ALD processes, helping us to untangle the thermodynamics, kinetics, and mechanisms of these heterogenous surface reactions.

In situ film and surface stresses were measured during tungsten (W) atomic layer deposition (ALD) using wafer curvature techniques. Tungsten was deposited at temperatures between 130 °C and 300 °C using tungsten hexafluoride (WF₆) and disilane (Si₂H₆) as the reactants on aluminum oxide (Al₂O₃) ALD surfaces. The stress was initially compressive during nucleation over the first 8-10 W ALD cycles as expected from a Volmer-Weber growth model. The stress was then tensile as the film became continuous.

Large tensile stresses were measured for the W ALD films. The film stress was ~2.0 GPa at a deposition temperature of 200 °C (Figure 1). For comparison, Al₂O₃ ALD films deposited at low temperatures have much smaller tensile film stresses of ~0.45 GPa. The large tensile stress in W ALD films may be generated at the grain boundaries. Atoms are too far apart at grain boundaries and experience an attractive force that wants to pull them together.

Surface stresses were also measured during the individual WF₆ and Si₂H₆ exposures (Figure 2). The WF₆ exposures produce a tensile stress of ~+2 N/m. This tensile stress is correlated with WF₆ reacting with SiF₂H* sacrificial surface species to deposit WF₅* surface species (SiF₂H* + WF₆(g) -> WF₅* + SiF₃H(g)). The Si₂H₆ exposures then partially release the tensile stress with a stress change of ~-1 N/m. The Si₂H₆ exposures redeposit SiF₂H* sacrificial surface species (WF₅* + Si₂H₆(g) -> WSiF₂H* + SiF₃H(g) + 2H₂(g)).

The large tensile stress in the W ALD films can lead to film cracking and delamination. This phenomenon is clearly evidenced by the in situ stress measurements (Figure 3). A large reduction in tensile stress is observed during film failure. The film failure was also obvious by visual inspection of the W ALD films when the W films began to peel off the underlying wafer.