Maintaining Gas Purity–A Systems Approach to Contamination Control
Controlling particles is not enough to guarantee contamination-free manufacturing. Gas purity must be maintained from production to wafer processing.
By Ralph J. Richardson
Contamination Free Manufacturing (CFM) of Integrated Circuits (ICs) has historically involved particle control and removal. Although yield-destroying defects are often traced to process non-uniformities, a majority of the effort to control contamination in IC manufacturing centers has focused on particle control. Cleanrooms are defined by the particle concentrations in the air and designers of processing equipment strive to minimize the number of particles-per-wafer processed through the equipment.
However, when gases and chemicals are included in the equation, control of particles isn`t enough to ensure CFM. Molecular impurities and degradation of construction materials are equally, if not more important.
Although, correlating molecular contamination and system degradation with device performance or yield is much more difficult than determining the impact of particle contamination on device production, manufacturers must determine the appropriate level of purity for process gases used in IC manufacturing and more importantly know how to maintain that purity from gas production to the wafer processing environment.
Characteristic Contaminant Levels for Various Wafer Environments
Historically, the purity of process gases used in semiconductor manufacturing has been driven more by improvements in analytical technology than the needs of semiconductor process yields or device performance.
Figure 1 shows the improvements that have been made in nitrogen, over the last decade. Nitrogen, the most common gas in semiconductor fabrication, is used widely to purge process gas lines and process equipment. Whether the nitrogen contains 1 ppb or 10 ppb O2, it`s unlikely to affect the performance or yield of the devices produced in the equipment. In fact, nitrogen is probably the cleanest environment that the wafer experiences during the entire fabrication process. Figure 2 lists the characteristic contaminant levels for some typical wafer environments in a semiconductor fabrication facility. As can be seen, although state-of-the-art cleanrooms are generally superior to Class 1 specifications, cleanroom air contains very high levels of particles and molecular impurities compared to nitrogen. Presently, there`s a high level of concern about molecular impurities, especially hydrocarbons, in cleanroom air.
With the advent of multi-chambered cluster tools, wafers can now undergo sequential processing steps without exposure to cleanroom air. However, it takes an ultrahigh vacuum chamber to approximate the cleanliness level of ultrahigh purity nitrogen. Most of the wafer processing steps performed in a vacuum are accomplished at conditions which are only marginally better than the Class 1 cleanroom environment.
According to Texas Instruments, the migration of contamination sources from people to equipment to process will be almost complete by the year 2000. (See Figure 3.)
ULSI Process Gases
The challenge for gas and chemical suppliers is, of course, not the bulk gases, which are cleaner than any other environment experienced by the wafer, but the process gases and chemicals used to add or subtract patterned thin films on the wafer. Not only are there a large number of process gases and chemicals, but the market for any one item in particular is relatively small. Figure 4 contains a partial list of gases and chemicals commonly used in semiconductor manufacturing. Although there are over 50 chemicals on the list, the total worldwide market for high purity versions of these gases and chemicals probably doesn`t exceed a few hundred million dollars. Consequently, the R&D dollars available to improve any specific process chemical are relatively small.
A hidden but overwhelming factor associated with improving the purity of process gases and chemicals is the cost of the analytical instrumentation required to measure and verify their purity levels.
Figure 5 lists a few common instruments currently in use to measure impurities in gases and chemicals in semiconductor manufacturing. Traditional instruments, such as gas chromatography (GC), have been supplemented by new technologies, Atmospheric Pressure Ionization Mass Spectrometry (APIMS) and Inductively-Coupled Plasma-Mass Spectrometry (ICP-MS), which cost hundreds of thousands of dollars instead of the tens of thousands of dollars of more traditional instruments. An additional, and often overlooked factor is the cost of properly calibrating and maintaining this state-of-the-art instrumentation that must reliably measure impurities in the part per billion (ppb) range.
In the past, it was generally accepted that improvements in gas and chemical purities would be required before the industry could move to the next-generation design rule. Specifications have traditionally moved to the limits of detection of the available analytical instrumentation. There has been very little correlation of process gas and chemical purity with device performance. (A major exception has been metals in liquid chemicals.) The cost of moving to the next level of purity may no longer be justified without some means of verifying the need.
From Specifications to Process Control
Semiconductor manufacturers have been slow to adopt the process control techniques used in the chemical process industries. The complex chemistry of producing a working semiconductor device is spread among many process steps involving equipment from many vendors. The constant change inherent in moving to each smaller generation of devices makes it difficult to optimize the process chemistry or unravel the complex interactive chemistry among the many process steps. Process models attempt to capture the effect of multiple processing steps (e.g., diffusion of dopants subjected to further thermal processing) for idealized structures but rarely include the consequences of possible impurity ranges in process gases and chemicals. As the industry becomes more interested in process control, the need for process chemical consistency becomes more and more important.
Although chemical consistency has always been recognized as an important requirement for a stable semiconductor fabrication process, the tendency to over specify chemical purity has limited the value of true chemical process control. The cost of ever more stringent process-chemical purity specifications and the interest in distributed process control for manufacturing semiconductor devices are combining to force a reconsideration of traditional pass/fail specifications.
A more useful approach to process chemical specifications for semiconductor manufacturing process control would entail establishing a range of acceptable process purity levels that permit the semiconductor manufacturing process to remain in control.
Purity at the Wafer
Control of the wafer process environment implies an understanding of all of the potential contributors of impurities:
1. Process gas as produced.
2. Cylinder contamination.
3. Gas distribution system.
4. Process equipment gas system.
5. Process chamber.
6. Process generated
contamination.
This discussion is concerned with a gas phase process, but a similar analysis could be performed for liquid chemicals.
Just as the concept of a thermal budget is useful when discussing a series of processing steps, the concept of a contamination budget may be useful in evaluating trade-offs in costs and performance. To maintain process control, important contaminants must remain within an acceptable range, no matter what their origin. If gas impurities are being specified to detection limits that are orders of magnitude greater than the levels of contaminants present in a process chamber from outgassing of the chamber walls or from the wafers themselves, processing costs will be unnecessarily high.
Moreover, process consistency may be unacceptable because of unknown variations in chamber surface conditions after scheduled maintenance. It`s important to understand all possible sources of contamination during processing to ensure reproducible process conditions. In some cases, the process itself may not determine process gas specifications, but stringent specifications may be required to prevent degradation of the gas distribution system (e.g., H2O in HBr) that leads to component failure (e.g., MFCs) or transport of contaminants to the reaction chamber such as corrosion by-products (e.g., H2 from HCl decomposition). Purity at the wafer requires a systems approach to contamination control.
Purity at the Cylinder
Process gas purity specifications will continue to improve, but maintaining process gas purity for CFM requires evaluating all aspects of the system. Once the gas is produced under contamination-free conditions, the first opportunity for degradation is the cylinder itself. With proper treatment (i.e., surface modification) and preparation (i.e., surface cleaning) the cylinder can maintain product purity for long time periods. For example, Figure 6 illustrates that, with proper treatment and preparation, a cylinder can be made virtually particle free. Each data point is the average of four measurements taken after a purge/ evacuation sequence.1
Having filled a clean cylinder with pure gas, however, doesn`t guarantee that clean gas will be extracted from the cylinder, unless proper gas management procedures are followed. Figure 7 is a plot of measured H2O concentration in HCl extracted from two clean cylinders filled under identical conditions. Corrosion in the valve of one cylinder released H2O previously adsorbed on the porous scale in the valve that resulted from corrosive attack of the valve by a condensed phase of H2O/HCl.
Dry HCl will not corrode 316L stainless steel. How dry does the HCl have to be to prevent corrosion? Insight into the corrosion of metal systems by HCl and HBr can be obtained by inspecting Figure 8.
The isotherms on the graph depict at what relative humidity a given concentration of gas in air will condense into a corrosive acid-water mixture. For example, at 20°C and a concentration of 1,000 ppm of HCl in air, if the air is at a relative humidity of 50 percent or higher, condensation and corrosion will occur. The valve with corrosion in Figure 7 was exposed to air before all of the HCl was removed from the manifold.
Corrosion can also occur if all of the air isn`t removed from the manifold before the cylinder is opened and the manifold is exposed to product gas. The situation is worse for HBr which will form a condensable acid-water mixture at even lower concentrations of HBr or H2O. Even though seemingly adequate purge/ evacuation procedures have been followed, cooling from gas expansion through a regulator, or compression from the impulse of a fast opening valve, could result in the system being on the wrong side of the vapor-liquid equilibrium curve and to condensation of a corrosive mixture.
Point-of-use purifiers have become widely used to ensure process gas purity. These systems can provide some protection in the event of an “upset” or a mistake in gas management, but they`re not a substitute for high-purity process gas, because system component degradation in front of the purifier is still possible, and their limited capacity and high-cost-of-ownership makes them questionable for low-purity gas. Most importantly, under certain conditions, some point-of-use “purifiers” have been known to evolve impurities and actually degrade the quality of gas.
Purity at the Distribution System
The next link in the chain between process gas production and wafer processing is the gas distribution system. This system is usually made-up of at least three separate systems: gas cabinet, facility gas system, and process tool gas panel. Poor design (e.g., “dead legs”), underperforming components (e.g., particle shedding from valves) and inadequate manufacturing methods (e.g., loose fittings) can all lead to contamination entering the gas and ultimately the wafer environment from these systems. More subtle problems can arise from materials compatibility issues (e.g., CO and Ni). Although gas management of WF6 used for tungsten plugs has centered on reaction with H2O desorbing from gas lines, tungsten has also been found to displace chromium in unpassivated 316L stainless steel exposed to WF6. Consequently, there`s a potential for chromium to be transported downstream to the process reactor.
Once the gas reaches the process chamber, the potential for contamination from the process chamber or even the wafers themselves must be considered. Figure 9 shows how the level of contamination in a process chamber can routinely reach the high ppm or even percent level from outgassing or cross contamination from other chambers in cluster tools.2 In this case, the high background levels (~ 0.3 percent H2O, 50 ppm O2) doubled when the gate valve between the loadlock (L/L) and the wafer transfer chamber (WTC) was opened. The level of O2 in the WTC was observed to jump to almost 200 ppm when the ballast was turned off on the vacuum pump evacuating the WTC.
Finally, the process itself can contribute contaminants to the wafer environment, either as by-products of the reaction (e.g., HF from reduction of WF6 to form tungsten films) or from outgassing of moisture or other gas phase species from the wafer itself. Figure 10 illustrates the observed high concentration of H2O formed from H2 outgassing from a wafer and reacting with O2 during a thermal oxidation process step.3
Future Directions
Although the largest gains in CFM can probably be achieved by reducing process equipment contamination, progress will also continue on improving gas purity, but with a shift toward product consistency rather than specification at the limits of detection. Figure 11 is a generic roadmap projecting process gas purity based on extrapolation of previous product purity trends.
Before these purity levels are established, however, industry must establish validation structures at 0.25 and 0.18 design rules to permit a Parato analysis of process gas impurities to more rationally guide molecular impurity product improvement efforts. The roadmap emphasizes a range of acceptable product purities.
Advances in analytical technology such as, low cost sensors for molecular impurities (e.g., H2O, O2, CH4) are critical to chemical process control. These sensors could become part of a distributed analytical network to monitor systems from cylinder to wafer and determine process limits as a function of chemical contaminants. This data could then be used to maintain process control within the acceptable limits.
Summary
Bulk gases (i.e., N2, O2, Ar, He, H2) used in semiconductor manufacturing provide the most contaminant-free environment the wafer is exposed to during wafer processing. Although the process gases aren`t as clean as the bulk gases, neither are they the major source of wafer contamination. This is dominated by the process equipment and the process itself.
The large number of gases and chemicals and the relatively small markets for these process chemicals doesn`t provide sufficient resources to meet the challenge to relentlessly remove all impurities from all process chemicals. The cost of the analytical instrumentation required to perform analysis at lower and lower concentrations of impurities is escalating at a rate that will make it prohibitive to continue to lower levels of impurities without evidence of real value to semiconductor device performance or manufacturing yield. Analytical instrumentation also remains a bottleneck for distributed process control. There`s a real need for distributed analytical sensor networks.
The concept of a contamination budget may be useful when evaluating the role of all the potential contributors of impurities, especially if process control parameters are to be developed for semiconductor processing. Development of process control parameters for ULSI processing will require cooperation among semiconductor manufacturers, process equipment manufacturers and materials suppliers.
References
1. J. Hart and A. Paterson, “Evaluating the Performance and Outgassing Performance of High-Purity, Electronic-Grade Specialty Gas Cylinders”. Microcontamination p.63, 12 (7), July 1994.
2. B. Huling, S. Y. Lynn, M. Su, “Characterization and Improvement of Gaseous Contamination Levels in a Multichamber Etch Tool”. IES Proceedings, 39th Technology Meeting, p.124, May 1993.
3. M. A. George, D. A. Bohling, J. J. Wortman, J. A. Melzak and G. A. Hames, “Hydrogen Content of Silicon and Thermal Oxidation Induced Moisture Generation in an Integrated Rapid Thermal Processing Reactor”. J. Vac. Sci. Tech. B11 (1) 86 (1993).
Dr. Ralph J. Richardson is responsible for coordination of all technology programs for the Electronics Division of the Air Products Gases and Equipment Group. He directs current research programs in gas purification, analytical technology, semiconductor process applications (e.g. plasma etching, chemical vapor disposition), reactor analysis, contamination control, and synthesis of new process chemicals and gases. Dr. Richardson`s previous assignments at Air Products include manager of the performance ceramics business with the Advanced Materials Division and manager of Electronics Diversification for the corporate Technology Diversification Departments. Dr. Richardson held previous positions as program manager of chemical laser development at McDonnell Douglas Research Laboratories and director of technology and innovation management at SRI International. Dr. Richardson has authored 22 technical papers and holds five patents. n
 |
Figure 1.
Minimum line width and product purity trends. As shown, nitrogen (N2), commonly used to purge gas lines, is unlikely to affect the performance or yield of semi- conductor devices.
 |
 |
Figure 3. Sources of wafer contamination shown over time. By the year 2000, contamination in wafer fabs will come mostly from the process.
 |
 |
 |
Figure 6.
With proper treatment and preparation, a cylinder can be made virtually particle-free.
 |
Figure 7. The effect of valve corrosion on product purity in identically filled HCl cylinders.
 |
Figure 8.
The isotherms on the graph depict at what relative humidity a given concentration of gas in air will condense into a corrosive acid-water mixture.
 |
Figure 9.
Contamination in the wafer transfer chamber increases when the loadlock gate opens during wafer transfers.
 |
Figure 10. Moisture levels attributed to H2 outgassing from the wafer and reacting with ambient O2 for two subsequent thermal oxidations in UHP O2.