*AV*-- The following 5/24 Morning Commentary out of RadarView was published on the CFMT thread at the request of a reader. Instead of linking to that response, I am printing it here also. Sorry for the duplication but I got in trouble the last time I posted a link<GGG>.
OPENING COMMENTARY MAY 24, 1999
For today, we would like to use some simplistic terms to discuss CFMT, a company that is in accumulation mode at the present price levels. We present this now due to upcoming 0.18u implementation and because one of our readers sent a private email asking me what the big deal was for CFMT. Therefore, as a change of pace, we present, hopefully, a layman's explanation of CFMT and why we believe it will regain its historical price levels or eventually be a take over target. As long as we mentioned takeovers, we are just 7 weeks away from SEMICON West in California. This year looks to be shaping up to be a blowout show with some heightened attendance. As we come out of this downturn, there is a great deal of catching up to do. To this end, it is quit possible some of the weaker companies may be on shaky ground, making them vulnerable to the larger players in this sector. SEMICON is a great venue to make these types of announcements. Before we move onto the CFMT discussion, we would like it known we do not consider ourselves gifted in the art of analogies.
CFMT = Contamination Free Manufacturing Technology.
Background Analogy - Imagine 100 golf balls randomly placed on a football field and 2 competing teams are engaged in a game. The golf balls pose no immediate threat to the majority outcome of the game since the weight of the player stepping on the golf ball might not cause an accident. The golf balls are more of a nuisance since there are hundreds of plays that can be run that do not come in contact with those balls. Therefore, for all the plays where the golf balls are not stepped on, there is no net affect on the outcome of the game since they did not factor into the plays. For those plays that did involve stepping on the golf balls, some of the do not affect the play while others cause the player to lose his footing and slip. Along with these golf balls, there are tens of thousands of lifesaver candies scattered across the field. These flat lifesavers should pose no threat to any of the players and are insignificant, when running plays, since they would be stomped into the turf. The outcome might be different if there were 10,000 golf balls on the playing surface or if there were 100 huge "medicine balls" on the field to act as obstacles.
Now, here is the kicker. The "environment" is such that spectators at the football game can sneak in a certain number of golf balls (like sneaking in beer) and an unlimited amount of lifesavers to throw on the field during each game. Security can catch a majority of the golf balls but not 100% of them. IT is highly unlikely they would go to the levels required to make sure no lifesavers are smuggled in.
Now suppose you put a proportional amount of golf balls and lifesavers on the stage of your local performance theatre for a ballet or gymnastics exercise event. If a group of little girls were doing ballet moves with leaps and jumps, and cartwheels with tumbling exercises. More than likely you would have many broken/sprained legs/ankles, sprained backs, girls flipping all over the place, a regular "Carnage de Ballet or Gymnastics".
The major job as a groundskeeper or janitor would be to clean up the "work surfaces" for both situations. For the football field, we might attach a rake to the back of a tractor with forks that are 1" smaller in diameter than the golf balls and rake up all the golf balls off the field so every play can be run without issue or danger. The lifesavers would slip through the rakes, remain on the field since there is no reason to get down to that level of cleaning detail. The rake on the back of the tractor is real inconvenient for the gymnastic/ballet stage area. You might use a smaller hand-rake designed to clean off all the golf balls in the smaller area but you have not addressed the lifesavers that pose a great danger to the gymnasts and dancers.
At the gymnasium, security is probably tighter. You go through a body search where they inspect everyone for spherical bulges. It is extremely unlikely that any golf balls can be smuggled into the area based on the level of detail used for inspection. However, you might be able to hide a roll of lifesavers in a clenched fist or in your sneakers. A few golf balls might elude detection but you really haven't stemmed the tide of the marbles unless you do a full body strip search. Therefore, you need an improved process or piece of equipment to remove the lifesavers, which were not required for the football field.
BTW - Consider that each of the plays run by the football team or each move by the dancer or gymnast as the equivalent of an active device (IC die) on a wafer. The situations are world's apart and it is apparent that different processes and procedures are needed for both.
Analogy is over.
The environment of a wafer fab is such that it is not contamination free. The pumps, the humans, the chemical recombinant fumes, the filtration systems themselves, humidity or lack there of, the equipment, and many other things, introduce contaminants into the wafer processing area.
On a side note, the air filtration systems used inside these facilities, filter out viruses and bacteria from the fab airborne environment. Defects the size of bacteria and viruses are not only killer defects, but the chemical and organic make up of these contaminants are enough to contaminate entire processes and equipment for a few manufacturing runs. As a matter of fact, If you touch just one wafer with one of your fingers, the fingerprint you made on the wafer has enough contaminants in it to destroy a majority of the other 100-249 wafers that were processed with it in the high temperature diffusion tube process. The fingerprint would then contaminate the next few manufacturing runs through that process tube. The wafers out of that tube that are contaminated would contaminate other processes as well, if there were no cleaning operations in between.
Back on topic. When you are making 2.0u devices with 3-D structure that are 500 - 10,000 angstroms in thickness (500 being the critical gate oxide), there are defects that exist which don't affect the performance of the electrical device that is created with the overlaying polysilicon film. However, as you shrink the feature sizes, the width and thickness of the features, also shrink.
The golf ball - Lifesaver example could represent airborne contamination and their affect on the different device structures, at varies device critical dimensions. There are ways to reduce the airborne contaminants, with the ultimate choice being a SMIF enclosure (Different topic altogether but the reason we are bullish on ASYT). In addition to airborne particles, there are process, equipment, and human defects that are introduced into the process. As you go from one process to another, there are cleaning operations that are done to remove contaminants that come in contact with the wafers through all sources. The cleaning processes clean off many of these surface defects that are induced by the process or the equipment, and acts as a redundant safety procedure for the prevention of cross contamination from one manufacturing operation or from one piece of equipment to another.
These cleaning processes were never a major problem before, at larger geometry sizes or were at such low levels that they were overlooked and yield or reliability issues. If we use the gymnast analogy again, before you start investigating "body grease or sweat" as a cause for slipping on the floor, you probably have addressed the golf ball, Lifesaver and floor wax issues first.
As you get down to 0.25u or 0.18u (you could argue whether it is 0.25 or 0.18), the residual films that are left on the wafer as a result of the cleaning process itself (the wafers are not perfectly clean but extremely close), start affecting the performance of the devices. This can be a source of contamination, causing manufacturing issues further into the process (like affecting growth rates and integrity of films).
Here is a homework assignment for you that require a dictionary and maybe the encyclopedia.
Look up the definition of Meter - millimeter - micron - angstrom - nanometer, and find out what 1 nanometer or angstrom is equal to in each of the larger units of measure. Then look up the dimensions of a virus or bacteria. You might have to use the Internet to look for filtration tables or information on bacteria and viruses. Then find out what Class 10 and Class 1 means in terms of cleanliness and particulate size and control. This should provide some perspective about the sizes we are talking about in the manufacturing of ICs.
At 1.0u, we might have a critical gate thickness of 500A. The tolerance for thickness control is +/- 50A, which is real good since most gate furnaces can control to +/- 5A and could actual hit a target thickness within 20A. At 0.25u or 0.18u the gate thickness is under 200A and the control needed is much tighter. When you grow oxides from the base silicon, it is a precise process of gas flows, time, and temperature since you are heating the silicon to a certain temperature, then introducing the gases in precise volumes for a specific timeframe. Any surface contamination on the bare silicon affects the growth rate and introduces defects (not necessarily particulate) into the gate oxide. The defects are on a molecular level and could be ionic contamination contained within the cleaning solutions. Whatever they are, it is molecular in nature and can alter the electrical characteristics of the gate oxide. They could also affect thickness control, which can affect performance.
At 1.0u, the effects of certain surface contamination resulting from the existing cleaning processes, or the inability to remove certain levels of contamination, are not sufficient enough to affect the quality, quantity, reliability and performance of critical film layers within the IC manufacturing process. The variability they do create is within the tolerances of the process. When you get down to 0.25 or below feature sizes, the present cleaning processes are incapable of meeting the stringent specifications required. It all boils down to the traditional cleaning processes not being sufficient to meet the needs of the advanced manufacturing processes and become sources of contamination, affecting device yield.
CFMT has the next generation of cleaning processes that addresses this deficiency. There are 1 or 2 other competitive processes but they either infringe upon CFMT's patent or they have not been commercially implemented successfully. FSII (I speak about them infrequently), has an IBM derived cryogenic cleaning process that freezes off the contaminants. But that is another story.
If I we haven't totally confused the reader, we will now give our reason why CFMT should perform over the course of the next few years. CFMT's problem has been that not enough 0.25u or below production was ramped up prior to this last downturn. In addition, there was controversy as to whether it was 0.25u or 0.18u manufacturing when this technology was needed. Most of the time the certain yield failures were attributed to process equipment not being capable to run those processes. As devices started shrinking, we found that the cleaning processes were not sufficient enough to prevent the process variability in the equipment, caused by inadequate cleaning technology..
BTW - the same type of explanation could be done for CMP and we may take that on as our next attempt to explain some of the enabling technologies that will be introduced as the more advanced 0.18u technology ramps up. CMP deals with 3D topography and the need to "planarize" the wafer surface as you add grow or deposit more levels of metalization and inter level dielectrics. . If you do not planarize and you increase the levels of interconnect, you run the risk of not being able to create the structures you need..
A good deal of this really boils down to doing brain surgery with the Thanksgiving carving knife. At sometime in the process, you have to change the tools you use to perform the process. Such is the case with cleaning technology. Traditional methods are running out of steam and must be replaced with more sophisticated processes.
Andrew Vance
RadarView Financial Newsletter
avance@radarview.com (Site is not up but the email portion works<g>) |
