Spin-on or CVD. This is the very basic choice that semiconductor producers must make in the coming months about crucial insulating materials known as dielectrics.

	WELL SUITED Honeywell's Star Center in Sunnyvale, Calif., where researchers develop new semiconductor materials. DUPONT PHOTO

It won't be an easy decision to make. Chip makers are spoiled because silicon dioxide, their standard dielectric for many years, is well known and easy to work with. But SiO₂ isn't a good enough insulator to prevent "cross talk" between the closely spaced copper wires of the latest generation of semiconductors.

Most companies have made their dielectric choices for the computer chips with 0.13-µm circuit lines that are just starting to come out now. However, they face another round of decision-making for the 0.10-µm chips that will be state of the art a few years from now.

Some 10 electronic chemical and materials companies are jostling for the industry's dielectric business. Right now, there's not much of an opportunity--the consulting firm Kline & Co. figures that only about $6 million will be spent around the world on new dielectric materials this year.

But Joseph Stockunas, worldwide marketing manager for electronics at Air Products & Chemicals, says these new low-dielectric-constant, or low-k, materials are likely to be the fastest growing electronic chemicals category for the next several years. Kline predicts they will be a $300 million market by 2005.

Chemical and materials suppliers are promoting an astonishing variety of low-k solutions. In one corner are several carbon-doped SiO₂ materials that are applied to silicon wafers by chemical vapor deposition (CVD) in the same way that SiO₂ itself has been deposited for years. These organosilanes tend to be marketed by CVD equipment makers, of which Applied Materials and Novellus are the most prominent.

Applied Material's Black Diamond CVD film, for example, is based on trimethylsilane, a precursor manufactured by Dow Corning, among others, and distributed by Air Products. Novellus' Coral CVD film is based on tetramethyl silane and tetramethylcyclotetrasiloxane (TOMCATS), which Air Products supplies.

IN BOTH CASES, the organosilane serves to reduce the dielectric constant, or k value, of the CVD film to 3.0 or lower, compared with 4.2 for conventional SiO₂ generated from tetraethyl orthosilicate. Yet the films handle much like SiO₂ and can be applied with CVD equipment already installed at the semiconductor fabrication plant.

In the other corner are materials that are applied by spin-on techniques. The spin-on process is fairly new to the dielectric realm, although it is already used in semiconductor plants to deposit photoresist onto the wafer. Spin-on dielectrics are being advanced by companies including Dow Chemical, Honeywell, Rohm and Haas, Air Products, Sumitomo Chemical, and JSR Corp.

Some spin-ons are organic polymers, some are inorganic, and some are blends of the two. Their k values range all over the map, depending on the material, but are generally below 3.0.

The semiconductor industry first broached the subject of new dielectrics in 1998 with the 0.18-µm circuitry that is now standard in high-end chips like the Pentium 4. At this production node, manufacturers found a relatively easy answer in fluorinated silicate glass (FSG), a CVD material made by doping traditional SiO₂ with silicon tetrafluoride.

Air Products benefited from this switch, Stockunas notes, because it's the world's largest producer of SiF₄.

At the 0.13-µm node, spin-on materials jumped out as a front-runner in April of last year when IBM announced that it had picked Dow Chemical's SiLK polyphenylene polymer as its low-k dielectric. But early this year, three other big semiconductor makers--TSMC, Advanced Micro Devices, and Motorola--announced that they would go with Applied Material's Black Diamond CVD dielectric film. Texas Instruments has said it will use Novellus' Coral.

THE DUST is only now beginning to settle at 0.13 µm--chips with this circuitry won't be in consumer products until next year--and it's still not entirely clear which technology is ahead.

Mark McClear, business director for Dow's semiconductor materials group, says SiLK will capture about one-third of the low-k market for 0.13-µm chips. He figures carbon-containing SiO₂s like Black Diamond and Coral will take another third and FSG will take the rest.

However, Meggy Gotuaco, Applied's Black Diamond product manager, sees the market sticking largely with FSG. Low-k CVD materials and SiLK will split what's left, she says, with CVDs dominating.

It's well known that Intel, one of the world's largest chip makers, is sticking with FSG at the 0.13-µm node. Plus, Gotuaco expects that some Applied Materials customers will use Black Diamond only in their high-performance products and continue to use FSG for their workhorse chips.

Gotuaco says this picture is "a change from 12 months ago," when the semiconductor industry seemed to be leaning toward SiLK and other spin-on materials in the wake of IBM's announcement. Since then, though, IBM has pushed back its 0.13-µm launch by some six months until early next year. Plus, several of the other chip makers that have adopted spin-on dielectrics for 0.13-µm devices--firms such as Altis and Infineon--have technology ties to IBM.

McClear counters that companies outside the IBM sphere--Sony and Fujitsu, for example--are also adopting SiLK. "We're very satisfied with our progress, especially in the face of entrenched competition," he says. "If SiLK wasn't everything we said it was, we would have been eliminated a long time ago." 

	Split decision Semiconductor firms adopting low-k at 0.13 mm

FRESH BATTLE LINES are now being drawn at the 0.10-µm technology node. Michael A. Corbett, an analyst at Kline, says these chips will be a more important market for low-k dielectrics than are the 0.13-µm devices because fabricators won't be able to use FSG: Its k value of 3.7 is well above the 2.5 or less limit that 0.10-µm chips will likely require.

Although chips with 0.10-µm circuitry won't hit the market until at least 2003--and won't be at high-volume production until 2005--McClear says semiconductor fabricators are already narrowing down their dielectric options. In fact, he expects some to make their decisions in the first quarter of next year.

McClear is counting on two to four more SiLK adopters beyond the six who have said they will use it at 0.13 µm. "People who picked SiLK at 0.13 will reap the benefits and be seamless at 0.10," he adds.

The SiLK used at 0.10 µm will probably be a "porous" version with a k value of 2.0 that the company began sampling earlier this year. Instilling SiLK with air-filled pores may be necessary because the k value of the current product, 2.6, is a little too high for 0.10-µm devices.

McClear concedes that other spin-on dielectrics are also likely to be adopted for 0.10-µm chips as part of what he expects will be a general semiconductor industry migration to spin-on technology. "CVD will still be around, but competing spin-ons will come in at its expense," he says.

Dow Corning and Air Products--which are already selling CVD starting materials such as trimethylsilane, tetramethylsilane, and TOMCATS for 0.13-µm devices--are developing spin-on dielectrics of their own as well.

Air Products has a porous silica material called MesoELK that Stockunas says is in the prototype stage today and could be in the marketplace by next year. It was developed through a joint effort of materials scientists at Air Products' main R&D center in Allentown, Pa., and silicon specialists at the company's Schumacher subsidiary.

Dow Corning's candidate is called XLK. According to Chris Konitzer, the firm's marketing manager for semiconductor fabrication materials, XLK is a porous version of an existing Dow Corning dielectric called FOx, which is based on hydrogen silsesquioxane.

MesoELK and XLK are inorganics and as such are on the other side of a spin-on materials divide from polymeric materials like SiLK. This divide isn't as wide as the chasm separating spin-on and CVD, but marketers on both sides try to capitalize on it nonetheless. 

ONE COMPANY extolling the virtues of inorganics is Rohm and Haas's Shipley subsidiary, which in May debuted a new dielectric called Zirkon LK. Zirkon is a matrix of methyl silsesquioxane that has been made porous through a novel approach: incorporation of an acrylic copolymer that volatilizes upon heating to leave tiny air-filled spaces.

Eric Alling, director of Shipley's dielectrics program, says Zirkon's pore-forming technique--the product of years of Rohm and Haas experience in coatings and other acrylic polymer fields--yields small and uniformly dispersed holes that other firms can't match. "This has proven to be a challenge for quite a few companies," he claims.

Alling maintains that inorganic materials like methyl silsesquioxane are better dielectrics than organics such as SiLK because, like SiO₂, inorganics are less prone to thermal expansion. This means less chance of heat-induced swelling, which can lead to chip failure, particularly in high-temperature applications such as automobile controls.

Dow's McClear claims that thermal expansion hasn't shown up as a problem in testing by IBM and others. Furthermore, inorganics have their own problems, he says, including an inherently open pore structure that can lead rinse chemicals and other unwanted semiconductor materials to become trapped, potentially poisoning the photoresist.

Another spin-on dielectric contender, Honeywell, is playing both sides of this materials divide by offering organic and inorganic solutions. The company's GX-3 and GX-3P aromatic polymer-based dielectrics are organic, while its Nanoglass porous silicas are inorganic. Honeywell's HOSP, meanwhile, is a hybrid organosilicate/polymer.

However, Jeff Kinnicutt, Honeywell's global dielectrics director, downplays the difference between organic and inorganic spin-on materials. "It's a matter of flavor," he says. "They're both ice cream."

Kinnicutt considers the differences between spin-on and CVD to be much more significant, and he sees spin-on winning in the end, in part because spin-on materials can be extended more easily from one chip generation to another through the introduction of porosity.

FOR EXAMPLE, Honeywell's GX-3 polymer has a k value of 2.6 whereas GX-3P--P is for porous--has a k value as low as 1.9. "The only thing a customer will need to do to move from one to the other is put a new bottle on the spin-on track," Kinnicutt says. "With CVD it's more difficult."

So far, Kinnicutt adds, CVD dielectric suppliers haven't come forward with k values in the sub-2.5 range needed for 0.10-µm semiconductors.

According to Gotuaco, however, scientists at Applied Materials are hard at work on this very task by playing with new organosilanes and fine-tuning the deposition technique. She notes that last year Belgium's IMEC microelectronics research center demonstrated the extendability of Black Diamond technology to k values below 2.4. Plus, low k isn't everything, Gotuaco points out. "We believe the difference in composition between CVD and spin-ons means different usability. CVD is just more compatible with a typical interconnect," she says.

Indeed, for all the breakthroughs and competing claims, outsiders such as Kline consultant John C. Davis say the race to 0.10 µm between spin-on and CVD materials is still too close to call. "When the year started, the momentum was with spin-on," Davis says. "However, as the year comes to a close, CVD seems to have caught up."

With numerous suppliers of electronic chemicals and materials competing in the low-k race, there will no doubt be plenty of losers. Given the amount of time and money suppliers are investing in the effort, the only sure winners will be customers in the semiconductor industry.

[Previous Story] [Next Story]

Chemical & Engineering News
Copyright © 2001 American Chemical Society