A whole new state of purity

By Hank Hogan

Process gases build up integrated circuits (ICs), touching and molding them layer by layer at virtually every manufacturing step. In turn, gas-borne imperfections can destroy these precious circuits at any step, making gas “control” perhaps the most critical contamination control issue that exists, especially now as the industry marches eagerly toward the 90-nanometer process node.

It's generally understood across the semiconductor community that current gas control requirements are met, in part, by some clever engineering and painstaking attention to detail in today's gas delivery systems. “However, when we move into the next node, the purity of some of the specialty gases will probably have to be different,” predicts Jack Martinez, a senior scientist in the Office of Microelectronics Programs at the National Institute of Standards and Technology (NIST; Gaithersburg, MD). “There will also have to be some new materials used.”

Difficulties are also expected to arise in the delivery of lower-impurity-level bulk gases such as nitrogen and oxygen. These lower levels will have to be achieved more economically than in the past. Fresh contamination problems will also appear due to new techniques and materials such as liquids vaporizing near the point of use, carbon dioxide pressurized to higher levels and the appearance of relatively unused gases such as xenon.

While the 2001 International Technology Roadmap for Semiconductors (ITRS) reveals plenty of gas-related contamination control challenges, here are some possible solutions, both practical and predicted.

What price purity?

For gases, contamination issues break down into two categories: impurities and particles—the latter scales with the process node because a particle that is large compared to typical circuit features is a problem.

Purity is less clear cut. Traditionally, purity has changed by adding “nines.” Gas purity generations moved from four (99.99), to five (99.999), to six (99.9999) nines.

That trend, according to NIST's Martinez and others aligned in the field, tracks more with advances in analytical capabilities than with actual, demonstrated impact on circuitry. Once metallic contaminants are removed, other factors besides purity may be more important.

“It isn't so much about purity in terms of most of traditional gases for the fab, but it's more about consistency,” comments Christopher Case, chief technology officer for semiconductor process materials, equipment and services specialist BOC Edwards (Crawley, UK).

Today's bulk gases—nitrogen, oxygen, etc.—are in the low parts-per-billion or high parts-per-trillion impurity and contamination levels. These gases, in particular nitrogen, are the gold standard when it comes to freedom from contamination. The purity levels are so high that nitrogen is often used to flush out a line, with liters of gas flowing through tubing to whisk away any problems that bubble out of the stainless steel or other materials.

According to Mike Fitzpatrick, program director of the advanced technologies group at the engineering design and construction company Lockwood Greene (Spartanburg, SC), the purity levels at the tools are boosted through the extensive implementation of point-of-use filters and purifiers. These devices remove particles and trap such impurities as water.

This approach has resulted in contamination levels that are in the high parts-per-trillion, a figure in line with the ITRS' call for the value to be less than 1,000 parts-per-trillion. This is supposed to fall to less than 100 parts per trillion for leading-edge manufacturing technology by 2006, while the cost of overall manufacturing is slashed by 25 percent or more annually.

That's a tall order in many ways, not the least of which is the need to build costly systems capable of such performance levels and then to verify the system using expensive instrumentation. For that reason, industry experts are predicting that gas delivery systems will follow a path similar to that taken by minienvironments in cleanrooms. The idea is to use filters and purifiers at the point of use, with the rest of the tubing and delivery system designed, built and operated at a less stringent level.

“Those of us who design and build these things have been trying to get owners to go with dirtier lines, if you will, and then have point-of-use filtration and purification,” says Fitzpatrick. “Evidently, the prices of these systems haven't hit the threshold of pain yet that they considered it necessary to do so,” says Fitzpatrick. “Now that manufacturers are seriously talking about this tells me they are hitting that threshold of pain.”

Under pressure: Low

This quest toward lower cost, higher performance and better contamination control could be achieved by some imminent technical innovations. But no matter what the future may hold, the issue will almost always boil down to one of “pressure.”

In a traditional gas delivery system, regulators sit at the source end and at various points along the way while a mass flow controller (MFC) sits at the consuming end. That arrangement allows the regulator to maintain pressure in the line while the flow controller limits what enters the processing chamber. Flow controllers measure the incoming reactant stream and employ various software algorithms to adjust the flow via a control valve.

According to Kaveh Zarkar, vice president and general manager for the material delivery group of MKS Instruments Inc. (Andover, MA), an MFC manufacturer, MFCs do not produce particles because they are not shutoff valves and so never grind two surfaces together.

However, an impending MFC innovation has contamination control implications. Regulators limit the pressure fluctuations that cause flow controllers to overshoot or undershoot a set point. These fluctuations arise because of variable gas demand. The semiconductor industry is demanding MFC changes that will eliminate regulators.

“They don't want the MFC to react to these pressure perturbations,” explains Zarkar. “They want it to be totally insensitive to pressure perturbations to enhance the performance of the MFC and to simplify the gas delivery system.”

However, there are a number of possible ways to achieve this pressure insensitivity, and such an MFC is expected to be used in the next generation systems. When that happens, gas delivery will be simpler, more reliable, and cleaner.

The “pressure” issue also rears its head in various ways when new materials appear in semiconductor facilities. These involve the use of both very low and very high pressures. Each has its own contamination control issues and each is prominent in the ITRS.

The first category, low pressure, includes the various liquid precursors that are expected to provide semiconductor manufacturing with materials of different capacitance. These low-k materials often have liquid origins—fluids that essentially evaporate off and produce a gas.

This is delivered to the tool and then combined with other materials to create a film with the desired electrical characteristics. From a contamination control point of view, these are gases that flow with very low pressure and may or may not have liquid droplets in them. The droplets, if present, can cause the output of downstream MFCs to oscillate. These droplets are another form of contamination, if the desired reactant is a gas. The low pressure also means that there isn't as much pushing of the resulting gas through various filters and purifiers. Manufacturers have to account for this.

“As you use more of these low vapor pressure gases, you want to minimize your pressure drop and yet maintain your filtration efficiency,” notes Gregory Leggett, applications development manager for Mykrolis Corp. (Billerica, MA), a maker of semiconductor gas delivery subsystems, including a line of filters and purifiers.

According to Leggett, these filter improvements will be achieved by altering the design and materials. Purification, on the other hand, will probably be done while the material is still a liquid.

Under pressure: High

High pressure contamination control challenge arises due to the imminent use of supercritical carbon dioxide. At the right pressure and temperature, the properties of liquid carbon dioxide become the same as those of the gas. In this supercritical state, carbon dioxide has some of the physical characteristics of both. This allows the solubility to be tailored for specific uses such as selective removal of a top-layer polymer without disturbing the other polymer layers underneath.

It's these capabilities that place supercritical CO2 on the semiconductor roadmap, but from a process and contamination point of view there are hurdles to overcome.

For one thing, the pressure needed is quite high, at a figure estimated to be thousands of pounds per square inch. This is not the highest currently used in a semiconductor facility. Nor is supercritical CO2 a brand new substance. It is currently used for, among other things, decaffeinating coffee. But it is a new bulk gas for semiconductors, one that has not gone through the optimization process needed to bring it to the purity and contamination standards demanded by the industry.

“It would be used under very high pressure, under supercritical conditions,” says BOC Edwards' Case. “And there isn't much infrastructure to support that in the fab, nor is there necessarily a lot known about how you purify, or perhaps reclaim and recycle, CO2 for semiconductor applications.”

A similar new material situation may also apply to xenon, a noble gas that today is little used in semiconductor manufacturing. Tomorrow, however, xenon could play a prominent role in advanced lithography.

Such lithography tools probably may require very high-purity nitrogen to prevent films from forming on the vacuum-based optics. At present, it's unclear if it will be economical to prevent contamination through the use of nitrogen or if it will simply be better to clean things up as needed.

What you don't see

The final problem has little to do with actual contamination but everything to do with knowing if things are clean. The ITRS calls for cleanliness levels that are next to impossible—or perhaps even impossible—to measure.

“Many of the projected limits are below the measurement capabilities of current analytical methods,” says Ralph Kirk, director of programs for the San Jose-based Semiconductor Equipment and Materials International (SEMI) North America trade association.

The problems reside in both contamination and impurity requirements. The situation is not totally hopeless, as advances in technology promise both better and less expensive detection. Currently, the best available method is often atmospheric pressure ionization mass spectrometry, or APIMS. Such devices are expensive, costing upwards of $250,000 per system. They also tend to be somewhat temperamental and require skill in setup and operation.

Alan Bandy, a professor of chemistry at Drexel University and president of a consulting company, MicroAnalytics Inc., is an acknowledged expert on APIMS and offers his services to companies interested in information such as the level of moisture in ammonia after purification.

Ammonia is a notoriously wet gas and low-level moisture detection is difficult. But Bandy says he can measure water levels down to parts-per-billion—a capability that is popular with various gas-delivery subsystem vendors.

There are other techniques on the horizon that might challenge APIMS by providing nearly the same or a better level of detection while requiring less skill to set up and operate. These include tunable diode, ion mobility and cavity ring-down spectroscopy. These techniques potentially lower the threshold detection limit, for instance, of water in such corrosive gases as HCl and HBr.

There are claims of significant improvements to the low-parts-per-billion measurement level. These advances may be needed in order to do on-site analysis and qualification, an increasing trend within the semiconductor industry.

Suhas Ketker, manager of analytical technology in the electronics division of Air Products and Chemicals Inc. (Lehigh Valley, PA), sums up the future of measurement this way: “Analytically, it may be possible to do things in the near future that we weren't able to do just a year or two ago.”

Hank Hogan is a special correspondent to CleanRooms magazine. He lives in Austin, TX.


Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.