Resources
Gas Filtration: Where it's Headed for the Next Generation of Semiconductor UHP Gas Processes
Submitted by Brian Sullivan || Valin Corporation
Within any of the semiconductor manufacturing processes that utilize gases, filtration is one of the most critical elements. Whether discussing inert or specialty gases, this holds true from the bulk source all the way through to the final point of use.
The reason for this is simple: if the gases are not properly filtered, contaminants will cause defects on the surface of the wafers being processed. Specifically, the improperly filtered gases will lower the yield of die per wafer, decrease the wafer processing system’s productivity, increase the number of wafers that need to be processed to achieve the fab’s production targets, and increase the overall costs of manufacturing.
What evidence of this exists? All one needs to do is follow a gas line from point of origin to point of use and count the number of filters installed along each gas line. In a typical fab, as a minimum, there will be one at the bulk source, one at the gas distribution panel, one at the gas drop to the process tool, one at the process tool’s gas box, and one more at the process chamber. Some fabs use additional filters as do some OEMs between their gas box and the process chamber.
The requirements of filtration continue to evolve. As the line geometries in production continue to decrease and the need for improved filtration results come into play, one form of filtration that has been commonly used for decades will most likely need to change.
Before considering the future of filtration, it is worthwhile to consider how the semiconductor industry arrived at this point. Most of the pioneers who worked in the original fabs have retired and the history of their work has been either mostly forgotten or is completely unknown by those now at work in today’s fabs. I started working with engineers in the fabs of southern California in 1990 and had many conversations with them about their filtration methods. At that point, filters made specifically for their newly sub-micron processes were still being developed and their need vs. their costs strongly debated. One of the more experienced equipment engineers told me the way the original gas delivery lines were filtered was to shove a piece of felt into the line. When it finally made its way to the other end, they’d “hog the line” and then replace it. I thought he was joking, but he was completely serious and the few people I know from those days have confirmed it was the way filtration was accomplished in those early years.
Filters, like fittings going from threaded NPT to compression to face seals and micro-welds, have evolved to take the place of felt plugs. Interestingly, the primary filter elements used in the industry have not really changed much in nearly 30 years while the other primary components of gas delivery and distribution systems have made significant changes.
Figure: 1 Evolution of Instrumentation from 1990-2019
Fittings: NPT to compression to Face Seal
Weld Fittings: Socket Weld to Butt Weld to Mini/Micro Weld
Valves: Needle/Ball/Plug to Bellows to Diaphragm
Regulators: Non-Tied diaphragm to Tied Diaphragm, to Mini-Diaphragm
Pressure Measurement: Pressure Gauges to UHP Gauges to Transducers and Monometers
Flow Control: Measuring pressure drop across a fixed orifice to Rotameters to MFCs
As depicted in Figure 1, the evolution in nearly all other forms of instrumentation used in gas delivery systems have evolved significantly but the types and technology of gas filters have remained fairly constant. The most commonly used elements inside of the sealed filter housings installed today are the very same types as they were in 1990: Teflon Membranes, Sintered Metal elements, and Sintered Ceramic elements.
There have been more sintered metals added to the mix, going from Stainless Steel to Nickel to Hastelloy (and other exotic metals in rare use). Additionally, the material forming the elements through the sintering process has been broadened from powdered metals to also include sintered metal fibers for some manufacturers, but the end result is still a monolithic metal element.
It is especially interesting how the semiconductor industry’s processes have gone from the sub-micron geometries produced as “cutting edge technology” in the 90s to the 10nm processes currently in production and 7nm processes in development, and how these filters have been able to endure, with very little change, until today. Some of the credit goes to the fact that many of the filters were made to a standard that was well ahead of its time and thus provided an amazing level of particle reduction. The Semi Standard for gas filtration is a 9-log reduction of particles down to .003µm. That means if one billion particles of .003 µm or larger were somehow introduced into a gas line, only one of them could make it through the filter.
Monolithic filters (sintered metal and ceramic element filters) have been able to provide this level of filtration for decades, if used within their designed constraints for specific flow, pressure, and temperature. Teflon and other polymeric membrane type filters have claimed this level of efficiency, but with many of them being only one thin layer of the membrane material (essentially equal to a piece of Teflon tape wrapped around a plastic support structure), their actual performance may be far less, especially after a few cycle purge events. In many cases the Teflon tape itself becomes shredded or develops large holes created by the high differential pressure shifts cycle purging causes.
The low ΔP provided by Teflon and other similar elements (when contrasted with the sintered elements) is a big factor in why the tearing or shredding of the element is not recognized by the end users. Additionally, the fact that these elements are held inside a welded shut stainless steel housing that prevents any means of inspection of the element without cutting it open in a machine shop keeps this flaw “out of sight and out of mind” for those who are relying on them.
Awareness on this issue in the industry is lacking. There are a disproportionately small number of engineers I have encountered who had the wherewithal to uncover this issue. They would take used filters from their inert gas lines that were changed out after a fixed service period (e.g., an annual filter change-out event) or a PM cycle and take them to their fab’s machine shop. Next, they would have the housings cut open, exposing the elements, or the lack thereof. This helped them learn for themselves just how fragile these filters truly can be.
What drives the use of these filters?
For the filters used in the OEMs systems, where many different gases are commonly mixed at the entry point of or very near to the process chamber, this compatibility issue is crucial. The OEM’s customers commonly demand a “final filter” right before the gases enter the process chamber to prevent any last particle from contaminating their wafers. Unfortunately, this leaves the OEMs caught in a dilemma: What the end user believes they need vs What the experienced engineers know the end user needs.
However, a few of the more senior and experienced engineers understand that after a Teflon tape-type filter has been run through a series of cycle purges, the element itself will likely be damaged and not even be functioning as a filter any longer. Unfortunately, the number of these engineers are dwindling, and there are far more engineers who do not take the time or effort to research the issue, but instead, blindly accept the product literature and “believe” the filters are truly functioning. The fab’s facilities and equipment engineers and the OEM product engineers rarely complete rigorous testing of their components, including ordering post-test cutaways of the components to inspect what happens.
A common practice is for the OEMs to put Teflon filters in place to meet the fab’s demands but put their trust in the sintered metal or ceramic elements they’ve installed upstream of the Teflon filters to do the actual filtration. They have no other choice that makes practical sense. Additionally, many of the current engineers are not aware of the flaw in these Teflon filters and truly believe the filters are functioning. The issue is not one which many engineers have been exposed to or one the Teflon filter manufacturers are keen to discuss.
There is also a significant difference between the way a sintered, monolithic filter element functions and the way a Teflon tape-type filter does.
Sintered filters act as true depth filters, providing particle filtration through a number of different methods combined into one. sieving, impaction, diffusion, surface energy, attraction, and entrapment. Particles that slip past the initial sieve of the element’s surface are forced to then go through a tortuous, thick series of channels and pathways where each of the other filtration methods are at work to prevent the particles from escaping the matrix they were introduced into. Also, the sintered membrane has so much surface area and depth to it, the particles only have the slightest chance of making their way through. That is how a 9-log particle reduction is realized.
A Teflon tape-type filter membrane is a very thin layer of sieve-type media that lacks nearly all the benefits found in sintered elements. The Teflon membrane is formed and either stretched in a controlled manner to form holes or perforated by lasers. Particles encountering the membrane are strained out because they are too large to get through the holes. That is until the membrane ruptures or has larger holes torn into it by exceeding its ability to withstand high differential pressure. When this occurs, the larger hole(s) created, or the area of the filter where the Teflon element is ripped away, become the path of least resistance for the gas and any entrained particles. Those particles will then bypass the damaged element and pass through the filter unhindered.
What about the need for improved filtration efficiency as the line widths continue to shrink with the latest EUV technologies?
As the old proverb suggests: “Necessity is the mother of invention.”
Particles that did not cause failures in the line widths of the past will cause critical failures now. In 1990, one could line up 166 particles with a size of 0.003µm (3nm) between the walls when the line width of the die was a 0.5µm (sub-micron) process. Today, that same 3nm particle in a 7nm line is a huge boulder that will fail a die in test.
The need for robust filters that remove much smaller particles from more harsh process gases has certainly arrived.
There are improvements in filter performance that have come to the forefront in just the last couple years. One manufacturer of sintered metal element filters has released third-party testing showing an efficiency of at least a 9-log reduction down to 0.0015µm (1.5nm).
The use of sintered metal fiber media has improved the ΔP performance of sintered metal elements and decreased the cost when compared to sintered metal powder elements.
The use of Hastelloy in sintered elements, though more expensive, has addressed certain needs with a few very problematic process gases and where the robustness of a monolithic filter element is needed because a higher process gas temperature is in use.
Today, no sintered metal or ceramic filter currently available is the cure-all for the industry. No Teflon tape-type filter can meet the requirements for robust performance or function for the long term in process gas lines that go through purging cycles or in those processes that utilize gases at increasingly higher temperatures.
Additionally, the OEMs are being driven by the fabs to lower the price for their wafer processing systems. Replacing their existing, low cost, Teflon filters with a series of monolithic filters to meet both the higher performance requirements and the compatibility requirements of the many gases used and blended in those systems is not currently seen as a viable option.
However, neither was the option of a fab buying an EUV photolithography system costing more than $100 million.
When the need for a specific level of performance outweighs the concern for cost, the solution will be found. The answer to the filtration requirements of the near future will not be a piece of Teflon tape far thinner than a single piece of paper, wrapped around and welded to a polymeric support structure and that element then welded into a metal housing.
About the Author
Brian Sullivan has worked with and for fabs and semi OEMs since 1990. He is the Director of Sales - Technology for Valin Corporation.
The reason for this is simple: if the gases are not properly filtered, contaminants will cause defects on the surface of the wafers being processed. Specifically, the improperly filtered gases will lower the yield of die per wafer, decrease the wafer processing system’s productivity, increase the number of wafers that need to be processed to achieve the fab’s production targets, and increase the overall costs of manufacturing.
What evidence of this exists? All one needs to do is follow a gas line from point of origin to point of use and count the number of filters installed along each gas line. In a typical fab, as a minimum, there will be one at the bulk source, one at the gas distribution panel, one at the gas drop to the process tool, one at the process tool’s gas box, and one more at the process chamber. Some fabs use additional filters as do some OEMs between their gas box and the process chamber.
The requirements of filtration continue to evolve. As the line geometries in production continue to decrease and the need for improved filtration results come into play, one form of filtration that has been commonly used for decades will most likely need to change.
Before considering the future of filtration, it is worthwhile to consider how the semiconductor industry arrived at this point. Most of the pioneers who worked in the original fabs have retired and the history of their work has been either mostly forgotten or is completely unknown by those now at work in today’s fabs. I started working with engineers in the fabs of southern California in 1990 and had many conversations with them about their filtration methods. At that point, filters made specifically for their newly sub-micron processes were still being developed and their need vs. their costs strongly debated. One of the more experienced equipment engineers told me the way the original gas delivery lines were filtered was to shove a piece of felt into the line. When it finally made its way to the other end, they’d “hog the line” and then replace it. I thought he was joking, but he was completely serious and the few people I know from those days have confirmed it was the way filtration was accomplished in those early years.
Filters, like fittings going from threaded NPT to compression to face seals and micro-welds, have evolved to take the place of felt plugs. Interestingly, the primary filter elements used in the industry have not really changed much in nearly 30 years while the other primary components of gas delivery and distribution systems have made significant changes.
Figure: 1 Evolution of Instrumentation from 1990-2019
Containment: Threaded pipe to tubing to welded tubing to electropolished tubing
Fittings: NPT to compression to Face Seal
Weld Fittings: Socket Weld to Butt Weld to Mini/Micro Weld
Valves: Needle/Ball/Plug to Bellows to Diaphragm
Regulators: Non-Tied diaphragm to Tied Diaphragm, to Mini-Diaphragm
Pressure Measurement: Pressure Gauges to UHP Gauges to Transducers and Monometers
Flow Control: Measuring pressure drop across a fixed orifice to Rotameters to MFCs
As depicted in Figure 1, the evolution in nearly all other forms of instrumentation used in gas delivery systems have evolved significantly but the types and technology of gas filters have remained fairly constant. The most commonly used elements inside of the sealed filter housings installed today are the very same types as they were in 1990: Teflon Membranes, Sintered Metal elements, and Sintered Ceramic elements.
There have been more sintered metals added to the mix, going from Stainless Steel to Nickel to Hastelloy (and other exotic metals in rare use). Additionally, the material forming the elements through the sintering process has been broadened from powdered metals to also include sintered metal fibers for some manufacturers, but the end result is still a monolithic metal element.
It is especially interesting how the semiconductor industry’s processes have gone from the sub-micron geometries produced as “cutting edge technology” in the 90s to the 10nm processes currently in production and 7nm processes in development, and how these filters have been able to endure, with very little change, until today. Some of the credit goes to the fact that many of the filters were made to a standard that was well ahead of its time and thus provided an amazing level of particle reduction. The Semi Standard for gas filtration is a 9-log reduction of particles down to .003µm. That means if one billion particles of .003 µm or larger were somehow introduced into a gas line, only one of them could make it through the filter.
Monolithic filters (sintered metal and ceramic element filters) have been able to provide this level of filtration for decades, if used within their designed constraints for specific flow, pressure, and temperature. Teflon and other polymeric membrane type filters have claimed this level of efficiency, but with many of them being only one thin layer of the membrane material (essentially equal to a piece of Teflon tape wrapped around a plastic support structure), their actual performance may be far less, especially after a few cycle purge events. In many cases the Teflon tape itself becomes shredded or develops large holes created by the high differential pressure shifts cycle purging causes.
The low ΔP provided by Teflon and other similar elements (when contrasted with the sintered elements) is a big factor in why the tearing or shredding of the element is not recognized by the end users. Additionally, the fact that these elements are held inside a welded shut stainless steel housing that prevents any means of inspection of the element without cutting it open in a machine shop keeps this flaw “out of sight and out of mind” for those who are relying on them.
Awareness on this issue in the industry is lacking. There are a disproportionately small number of engineers I have encountered who had the wherewithal to uncover this issue. They would take used filters from their inert gas lines that were changed out after a fixed service period (e.g., an annual filter change-out event) or a PM cycle and take them to their fab’s machine shop. Next, they would have the housings cut open, exposing the elements, or the lack thereof. This helped them learn for themselves just how fragile these filters truly can be.
What drives the use of these filters?
- First, Teflon tape-type filters are generally much less expensive than the sintered element filters.
- Second, the porous Teflon membrane itself and its more rigid Teflon support structure (typically a cylindrical mesh of molded Teflon the membrane is wrapped around and ultrasonically welded to) are compatible with nearly every gas used in the semiconductor industry. This is unlike the metal elements which must be thoughtfully selected to match their compatibility with the specific gases used in the process.
For the filters used in the OEMs systems, where many different gases are commonly mixed at the entry point of or very near to the process chamber, this compatibility issue is crucial. The OEM’s customers commonly demand a “final filter” right before the gases enter the process chamber to prevent any last particle from contaminating their wafers. Unfortunately, this leaves the OEMs caught in a dilemma: What the end user believes they need vs What the experienced engineers know the end user needs.
How is this manifested?
The OEM must install some kind of filter to satisfy the end user’s demand. Furthermore, the OEM knows the only filter element currently available that can meet all the compatibility requirements for the broad spectrum of gases and combinations of gases their customers can use is a Teflon element. Finally, the Teflon filter is the least expensive and makes the burden as light as possible.However, a few of the more senior and experienced engineers understand that after a Teflon tape-type filter has been run through a series of cycle purges, the element itself will likely be damaged and not even be functioning as a filter any longer. Unfortunately, the number of these engineers are dwindling, and there are far more engineers who do not take the time or effort to research the issue, but instead, blindly accept the product literature and “believe” the filters are truly functioning. The fab’s facilities and equipment engineers and the OEM product engineers rarely complete rigorous testing of their components, including ordering post-test cutaways of the components to inspect what happens.
A common practice is for the OEMs to put Teflon filters in place to meet the fab’s demands but put their trust in the sintered metal or ceramic elements they’ve installed upstream of the Teflon filters to do the actual filtration. They have no other choice that makes practical sense. Additionally, many of the current engineers are not aware of the flaw in these Teflon filters and truly believe the filters are functioning. The issue is not one which many engineers have been exposed to or one the Teflon filter manufacturers are keen to discuss.
There is also a significant difference between the way a sintered, monolithic filter element functions and the way a Teflon tape-type filter does.
Sintered filters act as true depth filters, providing particle filtration through a number of different methods combined into one. sieving, impaction, diffusion, surface energy, attraction, and entrapment. Particles that slip past the initial sieve of the element’s surface are forced to then go through a tortuous, thick series of channels and pathways where each of the other filtration methods are at work to prevent the particles from escaping the matrix they were introduced into. Also, the sintered membrane has so much surface area and depth to it, the particles only have the slightest chance of making their way through. That is how a 9-log particle reduction is realized.
A Teflon tape-type filter membrane is a very thin layer of sieve-type media that lacks nearly all the benefits found in sintered elements. The Teflon membrane is formed and either stretched in a controlled manner to form holes or perforated by lasers. Particles encountering the membrane are strained out because they are too large to get through the holes. That is until the membrane ruptures or has larger holes torn into it by exceeding its ability to withstand high differential pressure. When this occurs, the larger hole(s) created, or the area of the filter where the Teflon element is ripped away, become the path of least resistance for the gas and any entrained particles. Those particles will then bypass the damaged element and pass through the filter unhindered.
What about the need for improved filtration efficiency as the line widths continue to shrink with the latest EUV technologies?
As the old proverb suggests: “Necessity is the mother of invention.”
Particles that did not cause failures in the line widths of the past will cause critical failures now. In 1990, one could line up 166 particles with a size of 0.003µm (3nm) between the walls when the line width of the die was a 0.5µm (sub-micron) process. Today, that same 3nm particle in a 7nm line is a huge boulder that will fail a die in test.
The need for robust filters that remove much smaller particles from more harsh process gases has certainly arrived.
There are improvements in filter performance that have come to the forefront in just the last couple years. One manufacturer of sintered metal element filters has released third-party testing showing an efficiency of at least a 9-log reduction down to 0.0015µm (1.5nm).
The use of sintered metal fiber media has improved the ΔP performance of sintered metal elements and decreased the cost when compared to sintered metal powder elements.
The use of Hastelloy in sintered elements, though more expensive, has addressed certain needs with a few very problematic process gases and where the robustness of a monolithic filter element is needed because a higher process gas temperature is in use.
Today, no sintered metal or ceramic filter currently available is the cure-all for the industry. No Teflon tape-type filter can meet the requirements for robust performance or function for the long term in process gas lines that go through purging cycles or in those processes that utilize gases at increasingly higher temperatures.
Additionally, the OEMs are being driven by the fabs to lower the price for their wafer processing systems. Replacing their existing, low cost, Teflon filters with a series of monolithic filters to meet both the higher performance requirements and the compatibility requirements of the many gases used and blended in those systems is not currently seen as a viable option.
However, neither was the option of a fab buying an EUV photolithography system costing more than $100 million.
When the need for a specific level of performance outweighs the concern for cost, the solution will be found. The answer to the filtration requirements of the near future will not be a piece of Teflon tape far thinner than a single piece of paper, wrapped around and welded to a polymeric support structure and that element then welded into a metal housing.
About the Author
Brian Sullivan has worked with and for fabs and semi OEMs since 1990. He is the Director of Sales - Technology for Valin Corporation.
Printer-friendly versionA lesson for me is that I need to involve you earlier in the program.
You were tireless in your support and it will not be forgotten!