Photostrip Faces 300 mm, Copper and Low-k Convergence
Alexander E. Braun, Senior Editor -- Semiconductor International, 9/1/2000
 | At a Glance | | |
| | ||
| |||
"Strip technology faces two challenges: technology and economics," said Kevin Fairbairn, vice president and general manager of the conductor etch organization at Applied Materials (Santa Clara, Calif.). "With low-k materials and extreme low-k porous oxides, it's clear that existing strip technology is inadequate. We're using experience gained from our background in strip and conductor etch to work with our module development groups in the damascene area. We've made progress on the stripping of these materials, without degrading the k value. Today one needs process sequence integration knowledge, because you can't just look at the strip process; you have to look at all the processes."
Brian O'Donnelly, business manager for polymer ancillary products at Arch Chemicals (Norwalk, Conn.), sees growth in post-etch residue cleaners. "Ours was a strip business — chemicals to remove photoresist after processing. That market is relatively flat and declining, as facilities replace wet stripping with plasma ashing, O2 ashing."
Jeff Butterbaugh, applications development manager at FSI International (Chaska, Minn.), agreed, pointing out that even stable chemistries must evolve. "The basic sulfuric/peroxide mixture used at the FEOL, where you don't have to worry about attacking metal and other materials, continues as a mainstay of our stripping business; but we're constantly fine-tuning to improve process efficiency."
According to Bernie Wood, director of corporate marketing for Mattson Technology (Fremont, Calif.), strip has matured from a simple fab application — relatively low-tech, low-cost — to a key process. "Considerable progress has been made, not just to remove resist more effectively, but residue as well. Almost all tools sold today are single-wafer processing tools with advanced residue-removal capability."
"Above everything else new materials pose a challenge to the industry," said Graham Cable, director of product marketing at Axcelis Technologies (Beverly, Mass.). "Copper's relatively easier to handle than low-k materials, which is why everybody's rushing to it; whereas the adoption of low-k materials is fragmented. Organic materials, etch and strip issues remain unresolved."
Dana Scranton, director of strategic marketing for Semitool (Kalispell, Mont.), views plasma as strip's dominant mechanism. "Wet stripping is being done to a far lesser extent and is obviously becoming the minority choice," he said.
The integrated approach
Applied's Fairbairn views the integrated approach as the solution to the photostrip problem. "Companies have asked us to develop processes with new chemistries, and modifications, to change the post-strip process — which is a wet process — from being solvent-based to one less solvent-aggressive, and therefore less expensive, while simultaneously reducing environmental considerations (Fig. 2)."
|
|
Chris Lane, Applied's program director of strip technology, is concerned about strip processing's sensitivity to upstream processes. "While everybody's developing upstream processes and bringing in others, the expectation is that strip will scale almost linearly from 200 to 300 mm and the chamber will pick up all this in one step," he said. "Much is changing simultaneously, and it's not going to be just a linear scale-up of process technology."
Although device manufacturers want equipment for processes not fully proven, Fairbairn sees Applied's module approach as a solution to the problem. "Historically, we'd have had to wait for customer feedback before discovering a need for a new process and then begin developing it," he said. "Now the work is done in parallel with our other groups." Even so, Fairbairn admitted there has never been a time when so many changes were occurring simultaneously.
Jim Papanu, Applied's senior technology manager, expects manufacturers' expectations for residue removal to grow and increasingly burden the process. "We still must maintain material integrity while tackling residual removal," he said. "There's going to be a shift in the process chemistries being used. Historically, fluorine is used to address residue removal; but usually it's done in an oxidizing environment. With low-k materials, there'll be a big change, possibly to reducing chemistries."
Post-etch cleaning
Arch's O'Donnelly indicated that when people talk about strippers, they mean wet stripping as well as post-etch residue cleaning. "The growth is in the post-etch residue cleaning area," he said. "That's where chemicals are used to remove sidewall polymers-ash residues on metal lines or vias."
The shift toward post-etch residue cleaners is altering chemistries. Wet resist-strip chemicals are organic solvents. Post-etch residue cleaners initially were also organic-based; but, as the need to remove organic residues diminishes, chemical companies are introducing aqueous-based residue cleaners. "For aluminum/silicon-based IC manufacturing there's a trend toward the ultimate replacement of organic with aqueous-based chemistries," said O'Donnelly.
The transition to copper is changing everything. With the introduction of organic low-k dielectric materials, the choice of a post-etch residue cleaner is critical. Suppliers are working on chemistries for residue cleaning in copper low-k applications. This is difficult, because the various low-ks being considered differ. "Each has its own characteristics," said O'Donnelly, "such that if you develop a post-etch residue cleaner, you must ensure its chemistries are compatible with the specific low-k and the metallurgy on the wafer. Additionally, everything is complicated by thinness and porosity, and the fact you're dealing with copper, not aluminum, and products with low etch rates on aluminum don't necessarily have low etch rates on copper."
Copper/low-k integration is the biggest headache facing the chemical industry. "The fact that there's no chosen integration scheme, and that it'll change over the next five years as fabs get comfortable with copper and select various low-k materials, is intimidating," said O'Donnelly.
This is an understatement. Each low-k choice requires different plasma etching processes, or chemistries, to etch vias into the low-k materials; therefore, the nature of the residues can be different. No one product can meet all copper/low-k applications, and not too many existing chemistries are tweakable — or even desirable — for the new processes.
The expertise question
Axcelis' Cable observed that the industry expects suppliers to carry a greater applications development burden, for both new materials and conventional approaches. "We're no longer just selling equipment," he said. "We're being asked to provide global support and coverage for equipment and applications — softer services. Customers want our expertise at their venues." This leads to more mergers and acquisitions. "This definitely isn't the era of the little guy," he added.
The new materials' quandary lies more in upstream and downstream process integration. "Nothing happens in isolation anymore," said Cable. "For example, high-k and low-k are two different deposition approaches for different applications; they fit with some technologies and not others." This poses an interesting dilemma for equipment suppliers: their expertise must be greater than the customer's. "Suppliers must have expertise about what happens upstream and downstream from the process step they provide equipment for," said Cable. "We must do it to raise sales and afford more R&D resources to be able to provide process integration solutions."
There is some concern in the industry, however, that companies may place too much reliance on suppliers' expertise. The problem is many users are too resource-limited to acquire the needed critical talents in an industry where the company with the best skills set wins. The impact on resources caused by the steepness of the industry ramp's upward curve is becoming an issue for suppliers and users. "The industry must address 300 mm-associated product development needs while simultaneously delivering current application solutions — without overshooting, because we know there's a downturn coming sometime," said Cable. Suppliers and users must work more closely. "Customers expect delivery, but are unwilling to help," said Cable. "Things you can do in the demo lab can turn out differently in production because of upstream and downstream interactions. The industry wants production-worthy solutions, but equipment manufacturers don't have fabs. Doing this without access to production facilities and information is difficult."
"People are doing little more than ashing photoresist," said Bob Small, R&D director at EKC Technology (Hayward, Calif.). "Also, DUV resist doesn't always respond well to conventional photoresist-removing chemistry, so we're reformulating these old industry workhorses (Fig. 3).
Process kinetics concern Small. "If you're using 200 mm wafers, doing stripping — even with organic film — you're probably using a wet bench with process times of 5 to 30 minutes and running 25, maybe even 50, wafers with the newer pitches on the tools," he said. "With 300 mm, process windows are narrower, with most manufacturers using single-wafer equipment. This means you can't spend over 4 minutes per wafer in a single-wafer processing tool. This changes kinetics; now you must strip photoresist that's gone through a plasma process and hardened to a certain extent. You must hit hard on the films very rapidly with these chemistries, aggressively remove the organic films and leave the damascene structure undamaged." This will not be easy. New technologies are coming along, such as supercritical CO2 (see sidebar) — but although these cannot be discounted, they are not completely adaptable to back-end processes.
Since surface tension around a particle must be altered, traditional organic solvents have a particle-removal limit, because there is a limit to how much an organic solvent's surface tension can be reduced. "Do you do it all with water in the rinse step?" asked Small. "Or modify the initial organic material to do more of the work during the first step? We'll have to do more two-step processes and ensure the second step is robust. DI water works well, but we'll need more than that now; perhaps DI water with surfactants — but you can't leave any of those behind."
FEOL, BEOL and implant levels
As FSI's Butterbaugh put it: "Resist materials are getting difficult to work with; as they change, the process must be fine-tuned. We're dealing with these issues as we introduce ozonated water processes, eliminating sulfuric." Butterbaugh added that while basic, i-line and non-implanted photoresist strip can be done fairly easily, antireflective coatings, BARC coatings and DUV resists are another story.
Implant levels are a front-end-of-line (FEOL) issue. Piranha and ozone strip chemistries leave off at about the same point: when the process gets to ~ 1014/cm2 implant levels. It is here where ashing becomes necessary. FSI and others are working with SEMATECH on tantalum oxide and BST, considering them high-k materials.
Nothing is getting easier. As Butterbaugh said, "In three years, when we get to new materials in the resists — which have a very thin imaging layer at the top and use dry transfer techniques, and with 193 nm going deeper into the resist — what will the stripping rates be? Can we remove them after patterning, after ion implantation and plasma processing? As we've progressed from i-line to DUV resist, stripping has become increasingly difficult."
Back-end-of-line (BEOL) poses different requirements. There, device manufacturers generally ash everything off after both the metal and via etch, making it impossible to use acids as with FEOL. The situation is complicated by the new semi-aqueous chemistries to strip off etch residues.
Some stripping issues lie in the selectivity between resist and low-k material, although some integration techniques being considered do not leave much photoresist since it erodes at the same rate as the low-k etching, requiring residue removal. "We've worked with controlled levels of dissolved oxygen in dilute HF mixtures, as a cleaning chemistry for copper," said Butterbaugh. "We'll be expanding this to copper low-k, focused more on residue removal than resist strip." Carbon-doped low-k materials such as Coral and Black Diamond are ozone-resistant. Potentially it might be possible to use a DI ozone strip process for those low-k materials, depending on how the patterning is integrated.
"With bulk photoresist removal, interest lies in throughput and CoO areas, with the latter as the key driving issue," said Graham Hills, vice president and CTO of GaSonics (San Jose, Calif.). "For some devices there are charge sensitivity concerns, and there's a preference for microwave downstream technology for some of the more damage-sensitive flash-type products." However, there are more sophisticated applications, where an ash is done followed by a polymer clean. Also, where it is necessary to remove something such as a carbonaceous layer in a high-dose implant or, in the case of an etch, where polymeric material deposited in the etch must be removed. Multi-step processes are more successful when it is possible to vary temperature and rf bias and the microwave downstream excitation to alter the ion content and radical generation.
"Processes can be used that span the space from reactive ion etch (100%), downstream (100%), or some variation of the two." Hills added that many of the residue cleans are lower-temperature processes — not running at 250°C, but at 100°C and below. "Those are the more sophisticated cleaning applications, where resist removal is one step in many. We've incorporated water vapor and hydrogen into many of our cleans. Research is needed on these processes' mechanisms. We're used to using, somewhat empirically, oxygen, nitrogen with CF4 and small hydrogen additives, but lack fundamental knowledge about how they work, and certainly have very little knowledge of what water vapor actually does in a clean."
Cleans with low-ks, such as some of the organosilicate glasses (OSGs), are etcher-dependent. "Much depends on the etched scheme, whether or not it is the via-first, which tends to be the most dominant one," Hills said. Depending on the etcher and chemistry, there seems to be a dependency on the type of polymers left behind, requiring different approaches to the ash and clean steps.
The many choices in low-k films offer challenges to removal. There are many OSG films (either spin-on or CVD) versus organic films under consideration. This will influence how these films are dealt with because, compared to fluorinated silicate glass (FSG), an OSG film must be treated respectfully — however, it is still silica glass. Removing organic films will require completely different approaches.
In three years, manufacturers will be working at k values of 2.0. How these films will evolve is anybody's guess. Some might withstand today's ashers. Whether they will be integratable and what the ashing issues will be is unknown. It does not appear wet clean limitations will be reached for several years, and some sort of wet processing after ashing steps probably will be needed until below 0.13 µm. From a wet cleaning perspective, fundamental barriers will not be reached until well below 0.10 µm. Then there will be concerns over liquids going into and out of small, high-aspect features. By then, a dry approach may be required.
Plasma and dry stripping
"Three years ago there was interest in electrical charge damage caused by resist strip equipment," said Mattson's Wood. "This is no longer true due to progress made in the design of plasma systems, providing a more uniform plasma, remote plasmas and a low-ion-current wafer surface. Much advancement has been made to ensure no gate surge damage."
Several low-k materials are being evaluated and implemented. Some are organic-based, spin-on dopant (SOD) materials such as SiLK, the Dow material IBM will use; Honeywell's FLARE; and others. Then there are Novellus' and Applied's Coral and Black Diamond, respectively. Even these last two are carbon-doped oxides, SiOC-based films with a higher porosity than normal dielectric films, and a lower k in a carbon content. These materials, particularly SODs, are problematical. The manufacturer ends up with an organic-based material with an organic-based resist on top, which must be selectively removed without touching or oxidizing the low-k material. If it is oxidized, performance is degraded, and it turns hydrophilic, causing additional problems.
Mattson views its inductively coupled plasma source system as useful in low-k material evaluation because it operates at lower temperatures. "We get more directionality in the plasma flux," said Wood. "For those evaluating low-k materials, lower temperatures with a more directional plasma flux and alternative chemistries are of major interest for cleaning." Wood added they have fielded systems that are running copper and soon will be implementing low-k. "We have the only tools for resist strip that can process pure hydrogen, making them more like etchers," he said (Fig. 4).
The hydrous implant strip area traditionally has been a difficult dry stripping application. "You have heavily implanted resist, and the surface becomes very carbonized," said Wood. "Then solvents remain underneath this carbonized layer, and issues arise when you try to remove this material: It etches slowly on the surface, and solvents that expand cause popping and blistering and particle problems." Typically, hydrous implant is defined at a 1014 dose level or higher. There may be two to four layers, depending on the device, that get this type of process in the implants.
Uniformity is key for 300 mm. "Many dry stripping systems in the industry use microwave sources," said Wood. "The issue here is that when the wafer comes into the chamber area, energy is delivered basically through a small tube, with baffles and diffusers used to attempt to produce a more uniform flow over the wafer."
Semitool's Scranton sees a continuation of the dry stripping trend, but conceded the industry is trying to do without plasma. "Particularly with some of the low-k and, for that matter, even high-k films, plasma use can be damaging," he said.
The leading technologies in that area fall into two categories. One is advances in wet strip to discard sulfuric acid and, in some cases, to eliminate plasma stripping in medium and low-level implant situations. "We're trying to eliminate sulfuric acid and peroxide, as it pertains to what wet stripping does and that chemistry's dominant application," said Scranton, "which is post-ash clean."
Though plasma is the dominant method of resist removal, there still is enough variation in plasma strips and the resists themselves to require a cleaning step to ensure inconsistencies that may have occurred in the dry strip process are accounted for. Wet cleans are needed to remove the etch polymer that remains after ashing. Long-term, wet cleaning will remain an important part of stripping. High-level implant doses — 1015 and up — crosslink the resist, carbonizing it to a point at which wet cleaning in its current state cannot replace plasma. But, there are other technologies that use dry gas in combination with other forms of energy, which may replace plasma.
"Some of these chemistries, although able to remove resist without plasma, still need a wet clean," said Scranton. "Even though they can dissociate some of the implant and resist's carbon breakdown, in and of themselves they can't remove contaminants that would cause problems for follow-up deposition steps."
A problem for most strip providers is the lack of consistency in user needs at the 300 mm node. There is a difference in expectations about what they want in the 300 mm realm, and since there is little experience and few proven equipment sets, conflicting messages are sent out about what is needed in software, and whose FOUP will be used. There are even conflicting messages about processing formats — batch or single-wafer? All this makes a difference, particularly when manufacturers want to bring up 300 mm lines with as much proven 200 mm technology as possible.
Until a few 300 mm fabs come into high-volume production, the equipment has been shaken out, and the capability of doing 300 mm processing at 0.18, 0.15, and 0.13 µm is validated, there will be an uneasiness similar to the one during the conversion from 150 to 200 mm. Device manufacturers could not get everything they needed for 200 mm because the 150 mm supplier base had not matured sufficiently.
So far, semiconductor development has been welded to Moore's Law. Some wonder, however, whether it is so critical to meet — and often exceed — this 18-month device-capability-doubling-cycle. For specific applications such as wireless communications, where packages must be smaller and work at lower voltages, this is clearly needed; but is prodigious processing speed essential for the rest? The Internet is viewed as our industry's next prime mover; but, considering that it is bandwidth, not processing speed, that limits the efficiency of computing devices — whether PCs or hand-helds — for Internet uses, it is unclear why the entire industry must turn head-over-heels in ever-shortening cycles to attain that elusive extra 100 MHz.
As one goes lower in the supply food chain the cycle's impact is amplified. Once R&D time might have been as much six months, with a six- to eight-month window to work with customers, eliminate bugs and gain a clear understanding of where (or if) improvements were needed. Now if we can stick a wafer in and get a new product out in two months — preferably one month — we consider ourselves fortunate. Smaller suppliers lack resources to cope with such a cycle, which means they are always running behind. This leads to diminishing market choices. Then, a year and a half later, the whole rigmarole restarts, and the call for something new goes out, even though most of the "old" chemistries, processes or equipment can still do the job.
Moore made an observation that was arbitrarily transmuted into law. It will be interesting to watch what happens in three years or so, when we hit fundamental walls erected by those genuine laws of physics which, as Scotty so often remarked to Captain Kirk, we "canna bend." •
| Supercritical Fluids as Wafer Cleaners This heating can be continued until the liquid's density has been so reduced — and that of the vapor phase has been so increased — that both densities become equal. This point is known as the critical temperature. Since temperature and density inside the sealed container are everywhere equal, thermodynamics dictates the critical pressure inside the container be everywhere equal. A fluid brought to conditions above its critical temperature and pressure is known as a supercritical fluid. Supercritical fluids are used as solvents in commercial applications, including the extraction of caffeine from coffee, fats from foods, and essential oils and spices from plants. CO2-based supercritical fluids are particularly attractive because CO2 is non-toxic, non-flammable and inexpensive. Its critical conditions are easily achievable with existing process equipment. This unique combination of physical, chemical and economic properties of CO2-based supercritical fluids has prompted their evaluation as a replacement for process solvents currently used in manufacturing, such as precision cleaning and photoresist stripping.
The attractiveness of supercritical fluids as process solvents stems from their combination of gas- and liquid-like properties. The Table compares some physico-chemical properties of a typical organic fluid in liquid, supercritical and gaseous states, showing that the supercritical fluid state possesses gas-like values of viscosity and diffusivity as well as a high, liquid-like density. Figure 1 shows the density-pressure-temperature surface for pure CO2. The critical point for pure CO2 (31°C and 1072 psi) is depicted by the large, solid circle. To a first approximation, a fluid's solvent power is proportional to its density. Thus, high, liquid-like densities achievable in supercritical fluids allow for liquid-equivalent solubilities. Figure 1 also shows that relatively small changes in temperature or pressure in the critical point's vicinity result in large density changes. This means the density — and solvent power — of a supercritical fluid is "tunable."
Figure 2 is a schematic pressure-temperature phase diagram for CO2, incorporating the process flow for a closed-loop supercritical fluid treatment system. No uncontrolled waste streams can exit such a closed-loop system. All the materials solubilized in the supercritical fluid are retained in the separation vessel for subsequent analysis, recycle, treatment and/or disposal. Further, these materials are concentrated in the separator, greatly reducing waste volume.
GT Equipment Technologies (Nashua, N.H.) is working on a photoresist stripping tool based on a closed-loop supercritical fluid system. It will incorporate manifolds for the introduction of modifying chemicals to assist in the swelling, breakdown and solubilization of photoresist polymers. There also are a variety of surfactants and metal chelating compounds that are known to be soluble in supercritical CO2, which can be used, if needed, for particulate removal and trace-metal scavenging. Finally, by incorporating a final rinse step in the liquid phase of the CO2 (Fig. 2), there is potential for eliminating the need for purified water. | ||||||||||||||||||||||
