Subscribe to receive more articles like this: Print/digital

The requirement for ultrapure steam in many applications can be met by a new hydrophilic membrane for purifying water vapor that is nonporous and extremely selective.

Abstract
The Semiconductor industry faces technical challenges at each node. At the 65 nm and 45 nm nodes, processes in use include wafer cleaning, lithography, ALD, RTP, and diffusion. All of these processes are sensitive to the quality of the water or water vapor used. This paper describes a new method for generating ultrapure water vapor or “clean steam” for use as a process gas or a feed that could be condensed to generate ultrapure water. Data will be presented that positively supports the validity of the new method. This new method was able to reduce high ppb levels of contaminants in deionized water to ppt levels. Under class 100 sampling conditions, metals in the purified steam were found to be below detection limits of a single ppt except for sodium found to be 9 ppt. Significant reduction in TOC (total organic compounds) was also found. The purifier device was found to have a linear relationship between pressure differential across the device and the mass flow rate.

Introduction
At the 65 nm node, the high aspect ratio of silicon structures and small feature sizes has exposed the marginal performance of older cleaning technologies. Those processes that still function can exhibit significant wafer damage when they are most effective at removing particles.

Wafer Cleaning using supercritical carbon dioxide (SCCO2) was once heralded as the next great cleaning process. The advantages to using a gas over a liquid were significant. They included: the ability to rapidly move into and out of high aspect ratio nanometer structures; ease of chemical modification of the cleaning solution; solvation of contaminants; and ease of drying. However, serious issues became apparent when put into practice. These issues included: designing a reliable tool to work at high temperatures and pressures; particle formation from the fluid system; and the cost of the raw material. The failure of SCCO2 has extended the use of megasonic cleaning, which becomes increasingly less effective and more destructive as feature sizes shrink.

Immersion lithography has been generally accepted as the lithography technology for the next several silicon generations. Impurities in the water such as dissolved gases and ionic salts can change the index of refraction, which directly affects image quality projected onto the wafer. Initial immersion systems used house de-ionized water and may have included secondary chemical filters and degassers to provide point-of-use purity. This leads the immersion process to be susceptible to the variable capacity of resin beds to remove impurities and the limited effectivity of vacuum degassers.

Atomic Layer Deposition (ALD) depends on water vapor for High-K flm formation. The technique requires the proper molecule be available and not replaced by competitive species that will disrupt the lattice structure.Today, water is commonly delivered via bubblers. The water quality becomes a living history of contact with all piping materials as well as gases used for bubbling. Temperature control of both the gas and water, as well as level control, affect delivery rate. The bubbling process is a single stage distillation step that concentrates contaminants in the remaining water in the vessel. These contaminants can be traced to the supply water and the carrier gas as well as the continuous leaching of contaminants from the vessel itself, which is commonly heated. The bubbling process is not smooth but violent, which can lead to the entrainment of microdroplets of the contaminated water that carry nonvolatile metal, boron and silica into the process chamber with unknown effects.

Rapid Thermal Processing (RTP) and diffusion need water vapor for oxide growth. The move to larger wafers and higher throughputs has only increased these flow requirements. Since the direct delivery of water vapor from water has not been considered pure enough for these processes, pure oxygen and hydrogen are burned. The combustion process occurs either in a stainless steel catalytic oven or on a silicon torch at the entrance to the furnace. This process is inherently dangerous, and the higher flow rates only raise the level of risk. In addition to safety issues, contamination from particle formation of the torch, thermal management of the heat profile of the furnace, and incomplete combustion affect furnace performance. As for RTP, metal impurities from the combustion chamber and higher flow rates are a challenge as are the higher acquisition and operating cost of these systems.

While the above processes are very different, they all use water or water vapor. With a reliable source of ultrapure steam, applications such as wafer cleaning, immersion lithography, ALD, RTP, and diffusion could benefit significantly.

Advantages of Ultrapure Steam
Steam generated from de-ionized water is an excellent cleaning agent. It is extremely aggressive at absorbing ionic and hydrocarbon contaminants. This key benefit has historically been limited, since it has not been practical to generate either ultrapure steam and or keep it from entraining contaminants before it has reached the wafer to be cleaned.

Steam cleaning of wafers would be ideal, since vapor phase delivery allows rapid penetration into high aspect ratio structures. It could easily have its chemistry adjusted to follow existing wafer cleaning recipes. Steam is extremely aggressive at removing molecular contaminants from the wafer surface. In regard to particle filtration, liquid water can only be filtered to 0.1 micron while gas can be filtered to 0.001 micron. The contaminant carrying capacity of water is very large, being able to either dissolve or physically remove particulates by the condensed gas stream. From a raw materials standpoint, the use of ultrapure steam, as a replacement for dip and dunk water process for cleaning, could reduce water needs from gallons per wafer to grams per wafer. Existing wafer drying processes could easily be adapted to this cleaning technique and finally, the by-products of the process could be easily handled by existing fab wide waste management systems.

The development of a metallic-free water vapor generator and purifier would meet the needs of emerging RTP, diffusion, and immersion technologies. With regard to immersion, condensed ultrapure steam should be consistent over long periods of time without regard to house DI system variability. In addition, dissolved gases and other absorbed airborne molecular contaminants should also be easily removed through the steam purification process.

For ALD, water vapor would be available without entrainment of nonvolatile impurities or dissolved gases. For thick and thin oxide growth, a safer and less expensive way to meet increased water vapor requirements would also be available.

New Purification Material
Up to now no steam purification technology has been commercially available. Water vapor has been transferred to carrier gas through porous hydrophobic membranes that have been chemically treated to allow water vapor, but not water to pass through them. They have little selectivity, have a tendency to wet out, and have an upper temperature limit of 50°C which prevents their use in above-atmospheric applications. We have developed a new hydrophilic membrane that is nonporous and selective for water vapor. Figure 1 shows selectivity of up to a million-toone water molecules over nitrogen. In addition, the glass transition temperature is above 180°C, well above the boiling point of water.

Testing the New Material
To determine the viability of this new purification material, an experiment was performed to determine if water vapor that permeates across the hydrophilic membrane will be free of contamination. A steam generator system was developed to provide the controlled delivery of pure steam to the purifier. The membrane was tested for metallic, total organic carbons (TOC), and mass flow rate.

Manifold Setup
Figure 2 is a schematic of the manifold used for this experiment. The water was fed to the system through Valve 1 (V-1), which is controlled via Program Logic Controller (PLC).

Information about the water level was fed to the PLC through lasers that worked as level indicators. The indicators from top to bottom are full, low-level, and empty. Whenever the water level reached the low-level indicator, V-1 would open until the water level reached the full level indicator. The PLC also controlled the pressure within the boiler by raising its temperature. A pressure transducer relayed the pressure within the boiler to the PLC. The boiler’s pressure was set to 850 Torr. The steam that was generated rose into the Steam Purifier Assembly (SPA). The SPA consisted of a hydrophilic membrane within a 0.5 inch outer diameter PFA tubing. The water vapor that did not permeate through the membrane was condensed and sent to drain. V-2 and V- 4 were left open during this experiment. The water vapor that did permeate through the membrane was condensed and sent to the ballast. The ballast was kept at pressures between 45 and 195 Torr with a diaphragm pump. V-6 was left open when the system was being operated. The water flowing through the shell side of the condensers was kept at 22.5°C with a chiller. Sample collection was done within a glove bag under a purified N2 purge or under a laminar Flow bench. All samples were sent to a third party for analysis.

Steam Sample Collection
V-5 was the condensate’s sample port. Initially, this sample port was rinsed with the water generated by the system. After purging the sample port, one of the sample bottles was rinsed with the water from the condensate’s sample port. After the bottle rinse process, the bottle was left to collect water until it was approximately two-thirds full.

Permeate Sample Collection
V-7 was opened and the collection tube filled with water. Once the collection tube was filled, V-7 was closed and V-8 was opened to pressurize the collection tube. The water was then drained into the beaker. V-8 and V-9 were closed. This collection tube rinsing process was repeated two more times. After the third rinse of the collection tube, the collection tube was allowed to fill. Once the collection tube was filled with water, V-7 was closed and V-8 was opened to pressurize the collection tube. The cap was removed from one of the sample bottles and the bottle was placed under V-9. This valve was then opened and the water was allowed to drain into the bottle. The sample bottle was capped and the water was swirled around and then discarded. This bottle rinsing process was repeated two more times. After the bottle rinse process, the collection tube was allowed to fill for sample collection. Once the collection tube was filled with water, V-7 was closed and V-8 was opened to pressurize the collection tube. The cap was removed from the sample bottle and the bottle was placed under V-9. The valve was then opened and the water was allowed to drain into the bottle. The bottle was then capped. The sample collection steps were repeated until the bottle was two-thirds full.

Mass Transport Test
The test system was modified to allow permeate to flow through a condensing column that was maintained at 10°C. The condensate and permeate that condensed was collected in glass flasks open to atmosphere. The flasks were seated on digital scales that had a resolution of 0.1 gram. The data was collected over 15 minute periods and an average value for the collection period was used to calculate the gram per minute values. A total surface area of 48.7 square inches was used in the test.

Results and Discussion High Challenge Metals Test
The initial testing was not conducted in a controlled environment. The results indicated the membrane had the ability to reduce metallic impurities from source water from sub ppm levels to sub ppb. Lower level numbers were questionable due to the lack of control in the environment. Table 1 provides the details of measured metals in the source DI water, condensate, and permeate samples.

TOC Test
The TOC test was conducted within the same conditions as the high challenge metals test. The TOC measured in the steam sample was 380 ppb, while the TOC measured in the permeate sample was 22 ppb. This was a 94% decrease of TOC. Unfortunately, the results do not specify which contaminants permeated through the membrane. For example, the 22 ppb of TOC in the permeate may be light alcohols that can permeate through the hydrophilic membrane. While these results were favorable, they may have been better because the five-day time period between sample collection and sample analysis could have added contamination to the samples. Furthermore, the permeate sample could have been contaminated due to the sample gathering problems discussed in the high challenge metals test results.

Trace Metals Test
The following test was performed in a Class 1000 cleanroom. Samples were then collected under a laminar flow hood that measured particles to below Class 100 specifications. The use of the glove bag was discontinued. The move to a clean test environment allowed the detection limits to be extended to single digit ppt for most contaminants. The total number of trace metals tested for was 67. Table 2 shows the trace metals that were actually measured within the two samples.

Out of the 67 metals tested for only sodium was found to be present at 9 ppt in the purified steam (permeate). The source steam or condensate was found to contain 18 ppt aluminum, 49 ppt boron, 55 ppt potassium and 25 ppt sodium. This indicates that the membrane effectively prevented transport of these 4 metals and did not leach other metals from the polymer.

Mass Transport Test
As shown in Figure 3, the condensed water vapor output per minute was found to be linear with pressure differential. The slope was found to be Torr. The linear regression curve was a surprising 0.99999. The minimum pressure to drive water vapor across the membrane was found to be 40 Torr. The mass transfer rate based on the exposed surface area was calculated to a value of 8x10-⁴ grams/( inch). If flow rates of 50 to 100 slm of clean steam were needed to clean a wafer, the entire purifier assembly would require a foot print of under 16 square inches and 16 inches high, about the size of a standard water filter housing. The linearity with pressure and simplicity of design implies scalability to meet a wide range of process flow rates.

Summary
The test results indicate that a steam purifier can be used to reduce high levels of contaminants that may concentrate in de-ionized water. When 18 meg DI water is used as the source, water vapor with virtually no measurable metals can be delivered. A significant reduction in TOC was also observed. The mass transfer rate across the membrane was found to be directly proportional to the upstream pressure. Furthermore, this mass transfer rate was high enough to be commercially viable to meet future semiconductor requirements.

JEFFREY SPIEGELMAN IS PRESIDENT OF RASIRC, 11760 SORRENTO VALLEY ROAD, SAN DIEGO, CA, 9212. HE HOLDS A BS IN BIOENGINEERING AND MS IN APPLIED MECHANICS FROM UNIVERSITY OF CALIFORNIA AT SAN DIEGO. HE HAS OVER 50 INTERNATIONAL PATENTS AND PUBLICATIONS. PREVIOUSLY, HE WAS FOUNDER AND PRESIDENT OF AERONEX UNTIL IT WAS PURCHASED BY ENTEGRIS IN 2003. IN 2005, HE FOUNDED RASIRC TO ADDRESS PROCESS PURITY AND DELIVERY ISSUES AROUND NEXT GENERATION CHEMISTRIES, WITH AN INITIAL FOCUS ON WATER VAPOR. HE CAN BE REACHED AT 858-259-1220 OR JS@RASIRC.COM

RUSSELL J. HOLMES IS ENGINEERING TEST LAB MANAGER AT RASIRC. HE HOLDS A BS IN CHEMICAL ENGINEERING FROM THE UNIVERSITY OF CALIFORNIA AT SAN DIEGO. PREVIOUSLY, HE WAS EMPLOYED AS AN APPLICATIONS ENGINEER AT A E R O N E X / M Y K R O L I S / ENTEGRIS FOR MORE THAN 5 YEARS. HE HOLDS SEVERAL PATENTS AND IS THE AUTHOR OF PUBLICATIONS C O N C E R N I N G PURIFICATION FOR THE SEMICONDUCTOR INDUSTRY. HE CAN BE REACHED AT 858-259-1220 OR RHOLMES@RASIRC.COM.

To read the entire article, sign up for a free subscription and then click on
Click here for the digital edition of the current issue