AVS1997 Session TF1-ThM: Modeling of Thin Film Deposition
Thursday, October 23, 1997 8:20 AM in Room B1/2
Thursday Morning
Time Period ThM Sessions | Abstract Timeline | Topic TF Sessions | Time Periods | Topics | AVS1997 Schedule
Start | Invited? | Item |
---|---|---|
8:20 AM | Invited |
TF1-ThM-1 Computer Modeling as a Tool to Predict Deposition Rate and Film Composition in the Reactive Sputtering Process
S. Berg, T. Nyberg, C. Nender (Uppsala University, Sweden) Reactive sputtering is a widely used technique to deposit oxides and nitrides etc. A serious drawback of this technique, however, is the drastic decrease in deposition rate that almost always occurs when depositing compound films as compared to depositing the pure metal film. In many cases it is possible to overcome this problem by operating the process in the narrow critical processing region between metallic and compound mode sputtering. In this processing region, however, the composition of the deposited compound thin film is extremely sensitive to the processing conditions. It is therefore very interesting to investigate the sensitivity of different processing parameters on the overall processing behaviour. Computer modeling can be used as a tool to increase the understanding of this complex process. We will show that such results can be used to illustrate the individual effects of the major involved processing parameters. It also serves to assist in the work of process optimization. The relationship between target poisoning, argon current density, deposition rate, reactive gas supply, partial pressure, total pressure, and film composition have been carefully examined. In particular we will point out a serious limitation that may arise during high rate multielement reactive sputter deposition. The effect of different reactivity of the involved elements may cause that one of the elements during deposition will not be fully oxidized. |
9:00 AM |
TF1-ThM-3 Ion Beam Sputter Deposition on Large Area Substrates - Comparison of Modeling and Experiment
A.V. Hayes, C.C. Fang, H. Hegde, B. Druz, V. Kanarov, K. Williams (Veeco Instruments, Inc.); R.J. Gambino (State University of New York, Stony Brook) Practical applications of ion beam sputter deposition (IBSD) in the manufacture of Thin Film Magnetic Heads and optical coatings, such as Extreme Ultraviolet Reticle Masks, have driven the development of extremely uniform coating systems. The primary elements of these systems are an ion source aimed at a target material and a substrate situated in the path of the sputtered flux. Design of these systems has been guided by mathematical modeling at various levels of sophistication. Of particular importance is the thickness uniformity of the deposited film as a function of the position of the ion source, target, and the substrate and the deposition parameters. In order to determine this by modeling, the distribution of the sputtered flux as a function of the ion mass and energy, the beam distribution on the target, and relative angle of the ion source to the target must be known. Frequently, the sputtered material is assumed to leave the target with a cosine distribution. This is an oversimplification, but is often used due to the scarcity of experimental sputter distribution profiles for the ions, materials, and energy range of interest for IBSD. In this work, the distribution of sputtered flux of various target materials including Mo, Au and Ni, for Argon ion sputtering in the range of energy of 300 to 1000 eV is calculated using the TRIMTM code simulation and this information is then incorporated into a program which calculates the deposited film thickness on the substrate. Simulations were run for various experimental geometries and process conditions. This program also calculates the sputtered atom energy distribution and the distribution of angle of incidence on the substrate. The model results are compared with experimental results and with predictions based on the simple cosine distribution. |
|
9:20 AM |
TF1-ThM-4 High Rate Deposition of Stoichiometric Al2O3 using Non-arcing dc Magnetron Sputtering.
K. Macak, U. Helmersson (Linkoping University, Sweden); M. Kharrazi, T. Nyberg, S. Berg (Uppsala University, Sweden) Dc sputter deposition of stoichiometric Al2O3 is usually difficult due to formation of an oxidized target surface which reduces the deposition rate drastically and causes charge buildup and arcing at the target. To avoid this situation the arrival rate ratio O2/Al must be high enough at the substrate position that a stoichiometric film can form but low enough at the target such that a conducting target surface is maintained. We have utilized Monte-Carlo simulations to estimate flux distribution of sputtered particles for the different geometries. These results supplemented by Bergs’ standard steady state model for the reactive sputtering process made it possible for us to predict the composition at different surfaces in the processing chamber. Experimental studies were carried out for several different target-to-substrate distances, a range of sputtering gas pressures, and different target diameters. Our results clearly shows that process can be tailored in order to achieve stoichiometric Al2O3 at the substrates while keeping the target in the metallic state. This is achieved providing a high enough inert gas pressure or small enough target diameter or long enough target-to-substrate distance. Thick stoichiometric Al2O3 films were successfully deposited at these conditions with low or no arcing during deposition. A considerable back-deposition of Al on to the non-etched part of the target keeping the surfaces conducting is also an important reason for the reduction of arcing. |
|
9:40 AM |
TF1-ThM-5 Formation of Ohmic Contacts to III-V Compound Semiconductors
T.J. Kim, P.H. Holloway (University of Florida, Gainesville) AuGeNi metallization has been provided practical low resistance ohmic contacts to n-type GaAs even though the mechanism(s) is not fully understood. Solid-phase regrowth has been reported to be important to formation of ohmic contacts, even though several critical issues still have not been clarified. We propose a refined solid phase regrowth model which includes factors controlling the Ge concentration in regrown GaAs. This model clarified the evolution of interfacial phases and their effects on the path(s) for dopant site selection and concentrations during regrowth of GaAs. The general applicability of this model to other GaAs as well as InP metallizations will be reviewed. Implications for contacts to GaN will also be discussed. This work supported by ONR Grant N00014-92-J-1985. |
|
10:00 AM |
TF1-ThM-6 Simulation of Metal Multilayer Deposition
X.W. Zhou, R.A. Johnson, H.N.G. Wadley (University of Virginia) Some metal multilayers composed of thin magnetic layers separated by thin conductive layers exhibit giant magnetoresistance (GMR). Examples include NiFe/Cu/NiFe and Co/Cu/Co sandwich structures. These structures are produced by physical vapor deposition, and it has been shown that the magnitude of the GMR effect is a sensitive function of the microstructure and defect content (especially those defects located near the boundaries between the magnetic / con- ductive layers). For instance, the interfacial roughness, interfacial chemical mixing, vacancies, stacking faults, twins, dislocations etc. are all likely to contribute to the magnitude of the GMR effect. To perfect the physical vapor deposition synthesis of GMR materials, the relationships between these multilayered material defects and the methods / conditions of processing need to be understood. Atomistic modeling techniques have been used to study the deposition of copper-nickel multilayers under various incident energies. The results indicate that although increasing incident energy may reduce the interfacial roughness and vacancy concentration (and therefore improves the magnitude of the GMR effect), it also causes interfacial mixing and can result in a deteriora- tion of the GMR effect. In experimental studies, some researchers have reported that increasing the incident energy improves the GMR ratio, while others have indicated that increasing incident energy has a negative effect on the GMR. The results of the present calculation clarify some of the confusion surrounding the effect of incident energy, and predict the existence of an optimum inci- dent energy determined by a trade-off of the interfacial roughness and interfacial mixing. The simulations also indicate that copper atoms preferentially segregate on the film sur- face because they have a larger radius than nickel (which can release the surface stress) and are associated with a lower surface energy. The net result of this segregation is that during deposition, copper atoms preferentially migrate into the nickel layer deposited on top of them, which leads to a reduction of the GMR effect. In fact, segregation of copper on the surface has been suspected to be the major cause of reductions in the GMR effect. The problem is often solved in experiments by introducing oxygen into the system which, when adsorbed on the surface, can prevent copper from segregating onto the surface. The present calculation provides a theoretical base for various experimental approaches aimed at controlling this type of segregation. |
|
10:20 AM | Invited |
TF1-ThM-7 Atomistic Modelling for Simulating Physical Vapor Deposition
H.N.G. Wadley (University of Virginia) Physical vapor deposition is widely used for depositing microelectronic interconnects, giant magneto resistive devices, and other metal films. Experiments have shown that the growth conditions (i.e., deposition rate, substrate temperature, bias voltage, background pressure, plasma power, source-substrate distance, etc.) have large effects upon the surface morphology, microstructure, defect populations and residual stress of sputter deposited films. Predictive models that relate these "outcomes of processing" to the methods and conditions of deposition are needed for reactor design, process optimization and eventually, for "intelligent" feedback control. The emergence of computationally efficient embedded atom potentials (that can be "calibrated" against large scale ab-initio calculations and/or measured physical constants) has led to a renewed interest in atomistic simulations of physical vapor deposition. Molecular dynamics simulations based on EAM potentials can be used to analyze short duration events (~1ps) encountered during vapor deposition, e.g., inert gas ion sputtering, metal atom resputtering, atomic reflection, bias diffusion, and bombardment induced (athermal) diffusion/damage. The MD method cannot be directly used to simulate deposition at realistic growth rates because too few atoms can be deposited in the 1-5ns of real time computation achievable with a high end workstation. New two and three dimensional Monte Carlo techniques are able to overcome this computational limitation and are beginning to be used to establish relations between morphology, structure and defects and growth conditions. Novel MD-continuum (hybrid) models are also being developed. These essentially relax the kinetically trapped structures predicted by high deposition rate MD simulations and enable surface morphology, point defects concentrations and twin structures to be simulated for any process condition. These modelling methods are described, and their contribution to the simulation of thin film sputter deposition reviewed. |
11:00 AM |
TF1-ThM-9 Vapor Transport during Directed Vapor Deposition
J.F. Groves, H.N.G. Wadley (University of Virginia) Directed Vapor Deposition (DVD) is being developed as a flexible vapor phase material synthesis tool to produce engineered films for such diverse applications as turbine blade thermal barrier coatings (TBCs) and semiconductor via filling. The primary DVD system configuration utilizes electron beam evaporation in combination with a focussed, supersonic gas jet to transport vapor to a substrate for film creation. A model of vapor transport during DVD has been developed using Direct Simulation Monte Carlo (DSMC) to reproduce the gas jet flowfield and Monte Carlo techniques based upon bimolecular collision theory to follow individual vapor atoms through the flowfield from source to substrate. The model has been verified against DVD experimental results and against experimental data of vapor transport during sputtering. The model produces individual vapor atom trajectories during transport and records the energy, angle, location, and efficiency of vapor deposition, information vital to an understanding of film growth microstructure. Initial model results facilitate understanding of the microstructure of silicon, copper, and zirconia films produced via DVD. The importance of the model for optimal process design and synthesis of engineered films and the direction of further model development are discussed. |
|
11:20 AM |
TF1-ThM-10 Modeling of Hyperthermal Molecular Beam Film Deposition
G. Chen, I.D. Boyd (Cornell University) The use of supersonic molecular beams as sources represents a novel approach to film deposition. The seeding technique employed allows considerable control over the incident kinetic energy of the reactant molecules striking the substrate. Hyperthermal kinetic energy greater than 1 eV can be readily achieved and the surface reaction probability is greatly enhanced. The current study considers expansion of a mixture of 1% Si2H6 and 99% H2 from a nozzle orifice, through a conical skimmer, and into the growth chamber. Silicon thin film is deposited over the substrate surface. The direct simulation Monte Carlo method (DSMC) is employed to simulate this flow in the transition regime. This particle based method provides detailed flow field information in the deposition chamber and is expected to aid in optimization of the system design. Results are presented for disilane incident energy, impact angle and flux distributions across the substrate to study the uniformity of the thin film. The beam intensity prediction from DSMC is in good agreement with Quadrupole Mass Spectrometer (QMS) measurements. The film growth rate calculated also has excellent agreement with experimental data. The effect of the skimmer interference is discussed and an optimal nozzle-skimmer distance is suggested. Finally, some studies are performed on the scale-up of the film deposition system, such as using slit nozzles or multiple nozzle sources. This work is supported by the National Science Foundation. |
|
11:40 AM |
TF1-ThM-11 Simulations of Thin Film Deposition Via Hyperthermal Cluster Impacts
L. Qi, S.B. Sinnott (University of Kentucky) Molecular dynamics simulations of collisions between organic molecular clusters and non-rigid surfaces have been performed to study the growth and properties of the resulting thin film. Previous studies have shown that at hyperthemal velocities, polymerization reactions occur when the incident cluster velocity corresponds to an external kinetic energy that is within 3 eV/molecule of the bond energy of an individual cluster molecule. Some of the chemical products chemisorbed to the surface in the initial stages of thin film growth but the exact nature of the thin film was not clear. The goal of the present study is to clarify the mechanisms by which the thin film grows and examine its properties. A second-generation version of the reactive empirical bond order potential for hydrocarbons developed by Brenner, that has been modified to include long-range van der Waals interactions, is used in the simulations. The velocities considered are in the hyperthermal region and are comparable to those that result in shock-induced chemistry in energetic materials and that occur between particles and solid surfaces in interstellar space. |