Elsevier

Acta Materialia

Volume 49, Issue 4, 23 February 2001, Pages 583-588
Acta Materialia

Kinetic model of low pressure film deposition from single precursor vapor in a well-mixed, cold-wall reactor

https://doi.org/10.1016/S1359-6454(00)00356-6Get rights and content

Abstract

A phenomenological model was derived to describe the deposition kinetics of oxide film growth from thermally activated decomposition of vapor precursor on a heated surface at low pressure. A Langmuir derivation of mass balance on the growing film surface with surface saturation condition was used to model the deposition. Growth rate of solid oxide film was modeled as a function of molecular arrival rate, adsorption rate, and surface reaction rate. The model was applied to a novel metalorganic chemical vapor deposition process (Pulsed-MOCVD). The process features precisely controlled pulsed injection of precursor solution with ultrasonic atomization to deliver the precursor vapor to the reactor with no carrier gas. Growth behavior was predicted for an experimental investigation in which a dilute solution of Ti(OPr)4 (isopropoxide) in toluene was used as the liquid precursor for TiO2 rutile film on nickel substrate.

Introduction

Most CVD systems which have been modeled are either high pressure, diffusion controlled processes [1], [2], [3] or low pressure kinetic controlled processes [4], [5], [6], [7]. The unique precursor delivery system employed in the Pulsed-MOCVD process results in reactor conditions which fluctuate between molecular flow and viscous flow within a 10 s time period. We used the Langmuir analysis of surface adsorption coupled with reaction kinetics to derive a phenomenological model of film growth encompassing both control regimes. The model was developed for single component, heterogeneous, thermally activated reactions in a low pressure reactor.

The motivation for this work is the patented Pulsed-MOCVD process [8] which operates in a manner fundamentally different from typical CVD [9]. Figure 1 schematically illustrates the experimental Pulsed-MOCVD system. The typical CVD carrier gas and bubbler delivery system is replaced with pulsed liquid delivery and ultrasonic atomization. Each pulse cycle delivers an exact amount of liquid precursor solution to the reactor through the ultrasonic nozzle. Precise control of injection rate is achieved without a liquid pump or mass flow controllers by using a computer to alternately actuate the solenoid valves B and C. The volume of precursor solution in tube length L is delivered to the ultrasonic nozzle at the desired pulse interval [10], [11]. The injection rate of precursor to the reactor is further controlled by adjusting the concentration of metal alkoxide in the solution. Solutions of a high volatility solvent, i.e. toluene, and up to 20% volume of precursor were used.

The Knudsen number ranges between 0.001 and 0.15 during each pulse, and the time constant for steady flow is typically less than 2 s at the end of each 10 s pulse cycle. These conditions make modeling by ordinary viscous flow analysis quite problematic. Our approach is to model the flow through the vertical reactor as well-mixed plug flow. At the beginning of each pulse, the precursor evaporates instantaneously, filling the reactor with a homogeneous mixture of vapor. As the reactor is pumped back down during the rest of the pulse, it is assumed that the vapor in the reactor remains well-mixed, that is, that no precursor concentration gradients develop. Since most of the film deposition occurs in the first few seconds of each pulse when the precursor partial pressure is highest, the assumption of a well-mixed vapor should be reasonably accurate. The significant processing parameters are substrate surface temperature, Ts, reactor total pressure which is a function of time through each pulse, P(t), and precursor solution molar concentration, cmo. The reaction activation energy, EA, and the adsorption energy, Q, are determined by the precursor chemistry.

Section snippets

Derivation of the kinetic deposition model

The phenomenological model of the growth kinetics is derived from the conservation of mass on the surface and the physics of molecular adsorption. Thus, the model describes the deposition of a single phase solid oxide from the thermal decomposition surface reaction of a single precursor. Figure 2 shows the schematic representation of the deposition process. We assume that only physical adsorption occurs and that the low reactor pressure is below the precursor condensation pressure. If there

Results: comparison of the model to Pulsed-MOCVD experiments

We were able to compare experimental growth rate data to both forms of the kinetic model. Using the color shift method [11], the TiO2 film growth rate was measured as a function of time during the pulse, as in equation (9). The growth rate for the whole process was determined from measurement of the final film thickness and compared to equation (17). Figure 4 shows the results for both instantaneous and integrated growth rate. The fact that the growth rate data exhibit reasonably good agreement

Conclusion

The phenomenological model describing the reaction kinetics of the growing film was derived from a Langmuir model of the mass balance on the growing film surface. The growth rate was normalized by the molecular arrival rate, making the normalized growth rate model a very general result, applicable to many types of single precursor deposition systems. Both forms of the model, for growth rate at a given temperature and precursor concentration with time, and for normalized growth rate as a

Acknowledgements

This research was supported by the National Science Foundation under Grant No. DMI-9796114. Dr. Krumdieck also received support from an ARCS Foundation Scholarship and an SAE Doctoral Scholars Award. The pulsed ultrasonic nozzle method was developed in collaboration with, and with contribution from, Dr. Harvey Berger, Sono-Tek Corporation, Milton, NY [8]. The Pulsed-MOCVD kinetic deposition model was developed with a major contribution from Professor Rishi Raj, University of Colorado at Boulder.

References (12)

  • Y. Akiyama et al.

    J. Crystal Growth

    (1995)
  • K.F. Jensen
  • A.S. Kelkar et al.

    Trans. ASME

    (1996)
  • K.L. Siefering et al.

    J. Electrochem. Soc.

    (1990)
  • C.P. Fictorie et al.

    J. Vac. Sci. Technol. A

    (1994)
  • M. Masi et al.

    J. Electrochem. Soc.

    (1999)
There are more references available in the full text version of this article.

Cited by (14)

  • Al<inf>2</inf>O<inf>3</inf> coatings on stainless steel using pulsed-pressure MOCVD

    2013, Surface and Coatings Technology
    Citation Excerpt :

    Thus, we have selected a low-cost precursor and aimed to produce a continuous, crack-free, adherent coating with controllable thickness at reasonable growth rates. The cold-wall PP-MOCVD reactor has been described previously [15,16]. The process does not use a carrier gas.

  • Scale-up design for industrial development of a PP-MOCVD coating system

    2013, Surface and Coatings Technology
    Citation Excerpt :

    These process design factors have been examined individually [17], as specific subsystems of an MOCVD system and as a whole [18,19]. The PP-MOCVD process utilizes direct liquid precursor injection of a metered liquid volume at timed intervals through an ultrasonic atomizer into a continuously evacuated reactor [20]. PP-MOCVD is distinguished from other MOCVD approaches in that it does not use a steady flow of carrier gas to transport low volatility precursor to the substrate [21].

View all citing articles on Scopus
View full text