Steam delivery method improves oxide growth rate and uniformity
Using new steamer design, silicon oxidation process shows up to 20 percent improvement in oxide uniformity
By Jeffrey Spiegelman, RASIRC
Oxidation of silicon is a common and frequent step in the manufacture of integrated circuits (ICs). The success or failure of the IC manufacturer often depends on the ability to grow a uniform oxide film quickly and repeatedly. Water vapor is commonly used to grow oxides; a new steamer design replaces water bubblers, direct water injection, and pyrolytic torches for the delivery of water vapor into oxidation processes.
Starting with deionized (DI) water, the steamer uses a non-porous hydrophilic membrane that selectively allows water vapor to pass through it. Membrane selectivity is significant with water molecules passing through it 1 million times faster than nitrogen molecules. In the vapor or steam phase all other molecules are greatly restricted, so contaminants in water such as dissolved gases, ions, total organic carbons, particles, viruses, bacteria, pyrogens, and metals can be removed. Total metals have been verified to less than 10 ppt.1
Test results were collected from installations at three different fabs. Each steamer installation replaced a different form of water vapor delivery. Expected benefits were lower cost, improved safety, and reduced film contamination. These expectations were met, and unexpected improvements in uniformity and growth rate were also reported during wet thermal oxidation of silicon wafers.
This article will review the thermal oxidation process for silicon, present the field results for the steamer installations, and then discuss the results.
Oxidation is a process used in wafer fabrication. The goal of oxidation is to grow a high quality oxide layer on a silicon substrate. During oxidation, a chemical reaction between the oxidants and the silicon atoms produces a layer of oxide on the silicon surface of the wafer. It is often the first step in wafer fabrication and will be repeated multiple times throughout the fabrication process.
Oxidation takes place in an oxidation tube. During the reaction, silicon reacts with oxidants to form silicon oxide layers. Typical operating temperature is between 900° and 1,200°C. The oxide growth rate increases with temperature. Generally, this technique is used to grow oxides between 60 and 100,000 angstroms thick.
Thermal oxidation of silicon is divided into two classes-dry and wet.
- Dry oxidation
Si (solid) + O2 (gas) → SiO2 (solid)
- Wet oxidation
Si (solid) + H2O (gas) → SiO2 (solid) + 2H2
Dry oxidation. During dry oxidation, dry oxygen is introduced into the process tube where it reacts with silicon. Dry oxidation is a slow process that grows films at a rate between 140 and 250 angstroms/hour. It is only used in industry to grow thin oxides (less than 1,000 angstroms).
Wet oxidation. During wet oxidation, water vapor is introduced into the heated oxidation tube. Because water molecules are smaller in size than oxygen molecules, they diffuse faster in silicon dioxide and the oxide growth rate increases. The wet oxidation growth rate is 1,000 to 1,200 angstroms/hour, so wet oxidation is the preferred method to grow thick oxides.
As a general principle, the amount of silicon consumed in the oxidation reaction is 45 percent of the final oxide thickness. For example, growing 10,000 angstroms of oxide consumes 4,400 angstroms of silicon.
Linear parabolic model
The kinetics of SiO2 growth occurs in three steps. First, the oxidant (H2O or O2) reacts with silicon atoms, then silicon atoms are consumed by the reaction, and finally a layer of oxide forms on the silicon surface.
The linear parabolic model developed by Deal and Grove3 demonstrates how silicon dioxide is grown on a silicon substrate during oxidation under both wet and dry conditions. The model identifies and defines two different stages in the oxidation of silicon: linear and parabolic.
Linear (first) stage. As the first phase in oxide growth, the linear stage refers to the chemical reaction resulting from the direct contact between the silicon and the oxidants at the surface of the wafer. The reaction is limited by the number of silicon atoms available to react with the oxidants. For approximately the first 500 angstroms, the oxide grows linearly with time. From that point on, the reaction rate begins to slow down as a direct result of the silicon dioxide layer covering the silicon atoms. As the silicon dioxide layer grows, it eventually prevents the oxidants from coming in direct contact with the silicon atoms and the parabolic stage of oxidation begins. The reaction of the oxygen at the silicon/silicon dioxide interface limits the oxide growth in this stage.
Parabolic (second) stage. The parabolic stage of oxidation begins when approximately 1,000 angstroms of silicon dioxide have been grown on the silicon substrate. At this point, the silicon atoms are no longer exposed to the oxidants and the oxidants begin to diffuse through the silicon dioxide in order to reach the silicon. The oxidation of silicon during this stage occurs at the silicon/silicon dioxide interface. As oxidation continues, the silicon dioxide layer thickens, and the distance the oxidants must travel to reach the silicon increases. The oxide growth rate is limited by the diffusion of the oxidants through silicon dioxide.
The details of the Deal-Grove Model are left to the reader. To summarize the portion relevant to our discussion, the growth rate depends on the concentration of oxidant at the oxidized surface to drive the oxidant through the oxide layer to get to the unoxidized silicon surface.
Henry’s Law states that the concentration of an oxidant in the solid is proportional to the partial pressure of the oxidant in the surrounding gas. Increasing the water vapor pressure or oxygen pressure in the process gas environment will increase the oxidation growth rate.
An increase in the water vapor pressure will directly increase the oxidation rate. Many researchers have reported increased growth rate with increased water pressure. Choe et al.4 reported a fourfold increase in aluminum arsenide (AlAs) oxide growth rate with increased water vapor pressure and no carrier gas at 440°C.
Figure 3 shows how an increase in temperature or pressure increases the growth rate. The oxide thickness increases with both pressure and temperature.
In practice, there are limitations to both temperature and operating pressures. Operating temperatures are kept below 1,200°C due to limitations of the equipment and thermal effects on materials. Although diffusion furnaces running above atmospheric pressure have been built, they are far more dangerous and become significantly more expensive to purchase and operate. Oxidation furnaces are generally run at atmospheric pressure.
Steamer test results
Test steamers were installed at three separate locations. All three installations were horizontal furnaces that were dedicated to wet thermal oxidation. The first steamer installation was on a 6-inch furnace tube that replaced a water bubbler that used oxygen as a carrier gas. The second steamer was installed on an 8-inch furnace tube and replaced a direct water injection system that used an oxygen carrier gas to purge the vaporizer. The third installation was on a 300 mm furnace and replaced a pyrolytic torch that generated water vapor from combustion of oxygen and hydrogen. The elimination of hydrogen from the process allowed the user to expand production and still remain within local fire ordinances. All customers expected the same or better film quality but did not expect a change in oxidation growth rate or process uniformity.
Process recipe temperatures and run times were kept constant. The only changes made were the elimination of oxygen and/or hydrogen gas and installation of the steamer. Total amount of steam supplied was initially the same as the previous technology and then adjusted to maximize performance.
Installation 1. The furnace was operated at 1,000°C. Initial results from replacing the bubbler with the steamer generated a consistent improvement of greater than 16 percent.
The recipe was then changed to eliminate the oxygen flow of 1 slm. Wafers were loaded at three locations within the tube. Run time was 104 minutes. Eliminating the oxygen purge improved the growth rate by 7 percent on 50,000 angstrom film.
Installation 2. At the second installation, the steamer was attached to an 8-inch horizontal furnace. The steamer replaced an existing DI water direct injection system running at 8 grams/minute. The oxygen purge through the injector was also eliminated. Multiple lots of 130 wafers were run through the furnace for different periods of time. Every fifth wafer was measured to get an average growth rate along the furnace tube and from run to run. The results were compared to film grown with water injection vs. the steamer.
Thin films reported better than 20 percent increase in growth rate. This rate slowed to more than 5 percent at thicknesses greater than 3,000 angstroms.
Installation 3. The 300 mm furnace was operated at 900°C. The pyrolytic torch was replaced with a steamer. Flow rate was 30 slm of steam. Time to grow the 1,000 angstrom film was reduced from 32 minutes to 28 minutes, representing a 7 percent increase in growth rate.
Installation 1. A 6-inch horizontal furnace was operated at 1,000°C. Wafers were loaded at three locations within the tube. Run time was 104 minutes. By eliminating the oxygen and increasing the water vapor flow rate, the non-uniformity across the tube decreased from ±3% to ±0.2 percent.
Installation 2. Three separate runs were made of 8,000 angstrom films. Previous customer requirements were ±5 percent. Before steamer installation, oxide thickness uniformity often exceeded the acceptance criteria with film thickness failing to meet 95 percent of target. This required measurement of each wafer and reworking a partial load. The tighter uniformity eliminated rework for definitive oxide thickness.
Installation 3. Initial data was collected on the 300 mm tool. The steamer was moved from the 300 mm tool to the 8-inch furnace for additional tests. Uniformity was average 995.410, min 992.63, max 999.43, range 6.809, std. dev. 3.325, and wafer uniformity 0.342000.
Increased growth rate with increased partial water vapor pressure
Users of the steamer demonstrated increased growth rate while keeping operating pressure fixed at atmosphere and increasing the partial pressure of water vapor. Similar increases in growth rate with increasing water vapor pressure had been reported by Geib et al.5
While increasing water vapor flow rate seems obvious to improve growth rates, technical difficulties interfere with increasing the actually quantity of delivered water vapor. For bubblers, the water cannot be heated near boiling or uncontrolled flow will result. The carrier gas flows needs to be increased to increase the delivery rate, which can then slow the diffusion of water vapor to the surface. With direct liquid injection systems, increasing flow rate leads to incomplete vaporization. This increases micro-droplet formation, which increases non-uniformity and ionic contamination on the wafer. Torches become larger and the thermal load from pyrolytic combustion of hydrogen and oxygen becomes more difficult to handle. There is also an increase in operating cost and safety issues.
Three different facilities using three different methods-a bubbler, vaporizer, and pyrolytic torch-all confirmed that a reduction in background oxygen gas and an increase in water vapor pressure resulted in an increased oxide growth rate. This increase in water vapor pressure agrees with the Deal and Grove Model for oxidation.
When the water vapor pressure is increased, the oxide growth rate is increased. According to the model of Deal and Grove, the growth rate of the oxide layer is directly related to the effective diffusion coefficient of the water molecules into the oxide layer and the equilibrium concentration in the immediate area. When a carrier gas is used to deliver water vapor, the carrier gas molecules generate a partial pressure. This partial pressure lowers the partial pressure of water vapor and slows the diffusion of water into the oxide film. The result is lower driving force and slower growth rate. For a given temperature and process pressure, oxide growth rates are fixed if the gas ratio is also constant. However, for a given operating temperature, this growth rate is not maximized until the water vapor pressure is equal to 100 percent of the operating pressure.
Localized effect and uniformity
At the gas entrance to the furnace tube, there is a ratio of partial pressure of oxygen to water vapor. As the water molecules travel toward the exit of the tube, they are absorbed into the oxide film at a much faster rate than the oxygen molecules. The oxide growth results in the gradual reduction in water molecules relative to the oxygen molecules. This localized reduction in water molecule concentration slows the available oxidation reaction and the growth rate as the exit of the furnace is reached, leading to a typical slower growth at the exit of the furnace and front-to-back non-uniformity as can be seen in the uniformity results from Installation 1.
As long as oxygen is part of the process recipe, the partial pressure within the furnace tube will not be uniform. By eliminating the oxygen gas, the water vapor pressure stays relatively constant and film uniformity improves across the chamber.
Replacement of previous water vapor delivery systems with the new steamer design resulted in significant increase in oxide growth rate and improvement in uniformity throughout the furnace.
For a given temperature and process pressure, the oxide growth rate is not maximized until the water vapor pressure is equal to 100 percent of the operating pressure.
If oxygen makes up part of the process recipe during wet oxidation, it will slow the overall growth rate by reducing the water vapor partial pressure. The oxygen gas will also lead to non-uniform growth throughout the furnace tube, since as the water is consumed from front to back of the furnace, the partial pressure of oxygen increases, and the partial pressure of water decreases. This difference in water partial pressure leads to variability in the oxide growth rate and non-uniformity throughout the furnace tube.
Jeffrey Spiegelman is president and CEO of RASIRC, a company specializing in liquid purification and delivery products.
- J.J. Spiegelman, R.J. Holmes, “Alternative Method and Device to Purify and Deliver Water Vapor,” SPWCC, February 2006.
- J. Salzman, “Microelectronic Processing Oxidation,” Microelectronics Processing Course Presentation,
- B.E. Deal, A.S. Grove, J. Appl. Phys. 36, p. 3770, 1965.
- J.-S. Choe, S.-H. Park, B.-Y. Choe, H. Jeon, “Letter to the Editor: Lateral Oxidation
of AlAs Layers at Elevated Water Vapor Pressure Using a Closed Chamber System,” Semiconductor Sci. Tech. 15, pp. L35???L38, 2000.
- K.M. Geib, K.D. Choquette, H.Q. Hou, B.E. Hammons, “Fabrication Issues of Oxideconfined VCSELS,” Center for Compound Semiconductor Technologies, Sandia National Laboratories.