For now, we will not consider the pressure of the reaction nor the kind of reactor configuration, since these can vary widely and are often process-specific. The main common characteristics among different types of CVD processes is that, in all cases, the reactions or critical aspects of the reactions are thermally driven, and this requires that energy be added to the process. This energy can be in the form of heat to the substrate, plasma energy to the reactants or a combination of both.
The precursors used for CVD reactions must be of extremely high purity since any impurity will end up incorporated into the deposited film. Such impurities introduce uncontrolled changes in the material properties of the films that are detrimental to device performance. As a corollary to the statement on purity, CVD precursors must also be stable under storage conditions since any decomposition produces impurities that will be fed to the process. Ideally, CVD precursors only react under the temperature and pressure conditions that exist within the CVD process.
Typical precursors for CVD processes include:
The first step in the actual CVD process is the controlled introduction of the precursors and any required diluent gases into the reaction chamber; this introduction is depicted on the left side of the chamber in Figure 1. To do this, the precursors must have sufficient vapor pressure to produce a stable, controllable flow to the process chamber. High pressure gases such as silane, hydrogen, ammonia, etc. meet this criterion and are easily delivered to the process using MFCs. Many liquids (e.g. halides such as dichloro- and trichlorosilane, titanium tetrachloride, tungsten hexafluoride, tantalum pentachloride; organometallic compounds such as TEOS, trimethylphosphate (TMP), aluminum alkyls, tetrakis(diethylamido) hafnium, tetrakis(dimethylamido)titanium, etc.) also have sufficient vapor pressure to produce stable, controllable flows using MFCs. Potential precursors having lower volatility, including volatile solids, present a problem for delivery to CVD reactors since most methods employed to date suffer from repeatability problems. In practice, a substrate is loaded into the process chamber and heated to the required process temperature under inert gas flow. Once the substrate is at temperature, the precursor/diluent gas mixture is introduced to the process chamber.
While it may not be the controlling factor for growth rate in all forms of CVD, the adsorption and reaction of the precursor(s) on the substrate surface are obvious pre-requisites for thin film growth and it is worthwhile to take a moment to understand the processes that occur on the surface during thin film deposition. The factors that control the surface reaction are:
Once the precursor reaches the surface, it must adsorb and remain there long enough to react either by decomposition or with a co-reactant similarly adsorbed on the surface. The concentration of adsorbed precursors on a substrate surface is governed by the rate of arrival of precursor molecules at the surface; the proportion of incident molecules that stay on the surface; and the density of sites available for the adsorption of precursor molecules. Precursor arrival rates are also called the precursor flux and the probability of an incident molecule actually sticking to the surface once it arrives there is known as the "sticking coefficient". The rate of adsorption of precursor molecules is directly proportional to both factors and to the density of free adsorption sites on the surface. If reaction by-products remained adsorbed, they would block fresh precursor from adsorbing and thus slow the rate. Similarly, if impurities are present they may occupy surface sites limiting the adsorption of fresh precursor and slowing film growth rates.
The elemental adatoms (atoms that are adsorbed on a surface) that are created in the surface reactions are not stationary; in the case of a single crystal substrate such as a silicon wafer, they migrate across the surface until they encounter a high-energy surface site. High energy sites on a single crystal surface are sites such as atomic vacancies in a crystal face, lattice edges where a crystal plane ends and sites where a lattice edge experiences a "kink" (Figure 3).
Thin Film Deposition