Addressing the challenges of LP-epi exhaust management
05/01/2001
VACUUM TECHNOLOGY AND APPLICATIONS
Katsunori Takahashi, BOC Edwards, Tokyo, Japan
Andrew Seeley, Peter Mawle, BOC Edwards, Clevedon, United Kingdom
overview
Alternating high flows of acid vapors and hydrogen with additional toxic and pyrophoric gases presents a new challenge for exhaust treatment. These processes can be sensitive to exhaust pressure surges, limiting the use of switching between dual exhaust systems. Burner lifetime can be severely reduced by high input flows of acid gas; input flows of hydrogen varying from zero to 200 slpm and back to zero affects both efficiency and cleanliness of combustion. An interesting new development based on inward fired flameless combustion developed for PFC abatement has provided solutions. Results from this system mean that a single effective exhaust treatment for these processes is now available to the industry.
 |
Wafer processing tools for depositing epitaxial or polycrystalline silicon, silicon germanium, or silicon germanium carbide layers are becoming increasingly common in semiconductor manufacturing. These tools operate at atmospheric to subatmospheric pressures. They use process gases, including silane, germane, chlorosilanes, methylsilanes, arsine and other reactants, in high-volume carrier gas flows of hydrogen. Because of the high operating temperatures involved in the tool deposition chambers, hydrogen typically flows continually. Often a cleaning step involving 10 to 30 liter/min of hydrogen chloride (HCl), sometimes mixed with hydrogen, is performed at set intervals to remove material deposited on the internal structures of the processing chamber.
Abatement demands resulting from these various processes and gases are varied, ranging from maintaining emissions of flammable materials below their lower flammable limit (LFL) to removal of toxic materials to an appropriate recognized level, such as "immediately dangerous to life and health" (IDLH) or "eight-hour threshold limit" (TLV) (see table).
Exhaust management problems
The deposition gases mentioned above and hydrogen are only effectively removed from an exhaust stream by combustion, but regular flows of 30 slpm of HCl during a chamber clean step, for example, into a hot combustor (i.e., an exhaust treatment system combustion chamber) will cause severe corrosion of any metallic parts.
 Figure 1. Flammable limits of hydrogen (left). Figure 2. Hydrogen emissions from a standard TPU at various hydrogen flows. |
Some have proposed using exhaust systems with three-way valves, allowing switching between a combustor for the deposition step and a water scrubber for the clean step. Reliance on three-way valve switching presents both reliability and potential safety issues, however. Significantly, in these applications, where tools also have an atmospheric vent line and a vacuum line from the reaction chamber, any exhaust switching creates transient exhaust pressure changes that perturb the vent line and can cause a tool to shut down.
Others have proposed an alternative that places two wet scrubbers in series with a combustor, one before and one after. But adding components in series increases exhaust loading on the facility, and costs and maintenance. The most effective solution for handling exhaust from low-pressure epitaxial deposition, as outlined above, would use a combustor capable of handling high HCl flows.
Also, as we have noted, epitaxial and polycrystalline deposition of silicon, silicon germanium, and similar electronically significant layers are typically achieved using group IV hydrides, silane, germane, etc. These materials are usually toxic and often pyrophoric. Their abatement results in the formation of particles as the main byproducts of combustion; silicon dioxide and germanium dioxide are byproducts from oxidation of silicon and germanium hydrides. Incomplete reaction of some commonly used hydrides can cause the formation of partially oxidized byproducts that may subsequently continue to react, often in an unpredictable way. Abatement systems that cannot guarantee the complete reaction of such materials should be avoided.
Combustion of hydrogen
For combustion of common flammable materials, such as methane or hydrogen, the oxidant is ordinarily air, and sources of ignition a spark or a naked flame. (For pyrophoric materials, such as silane at a suitably high concentration, the material is spontaneously flammable and thus its own chemical nature is the source of ignition.) Materials are not flammable under all conditions; they are bound by LFL and upper limits of flammability (UFL), which are usually quoted for mixtures in air at room temperature. Hydrogen's LFL is 4% and UFL 75%, but a more accurate picture is given by a flammable limits diagram (Fig. 1).
 Figure 3. Inlet head design of a TPU modified for high hydrogen flows. |
Because of the nature of combustion processes, simply passing a process exhaust stream through an abatement system cannot guarantee the removal of hydrogen to a safe level. Sufficient oxygen concentration and temperature are required; otherwise hydrogen will pass unabated. With this in mind, we measured the emissions of hydrogen from our standard thermal processing unit (TPU) and thermal conditioning system (TCS), as seen in Fig. 2.
We found that the robust design of the radiant combustor maintains hydrogen emissions below the LFL even at flow rates of hydrogen significantly above the unit's maximum recommended input flow (i.e., 50 slpm). In other abatement technologies, for example an electrically heated oxidation chamber, such high hydrogen excursions can be a significant problem. With relatively low input reactant air these high hydrogen flows will result in flammable mixtures in the exhaust duct.
 Figure 4. Exhaust hydrogen concentration vs. input hydrogen flows, showing the effect of excess airflows. |
Since the ability to burn hydrogen in a combustion system is directly related to oxygen concentration within the combustor, we performed a series of experiments to determine hydrogen emissions from a radiant combustor into which an additional flow of excess air was introduced. For these tests, we modified the inlet head assembly of a standard TPU to allow excess air to flow via passages around the process gas inlet nozzles such that the process exhaust (containing hydrogen) was allowed to mix with the excess air only within the confines of the combustion chamber (Fig. 3). This modification allows up to 250 liters/min of excess air to be added (with no hydrogen flowing) without any impact on burner emissions.
We found that an additional airflow of just 200 liters/min was sufficient to allow 200 liters/min of hydrogen to be passed into the abatement system and still maintain an exhaust concentration of hydrogen such that a flammable mixture could never form downstream. Additional excess air is required to reduce the hydrogen level further (Fig. 4), but excess airflow without additional hydrogen input as a fuel lowers the combustor temperature and raises the outlet level of carbon monoxide (CO). Without hydrogen flowing, carbon monoxide output ranged from 63 ppm for 50-150 liters/min of excess air flow up to ~170 ppm for 300 liters/min of excess air flow.
Specifically, we found that the addition of 600 liters/min of excess air is sufficient for the complete abatement of process flows up to 200 liters/min of hydrogen. The robust and resilient nature of the combustor design results in a surprisingly wide window of operation; the airflow required for the combustion of 200 liters/min of hydrogen still gives clean combustion (no CO emissions) when the hydrogen flow is reduced to just 100 liters/min (Fig. 5).
Below 100 liters/min of hydrogen input with 600 liters/min excess air, however, we see that carbon monoxide emissions rise rapidly. At this point excess air input must be reduced; if it is reduced to 200 liters/min, then hydrogen is still completely removed (see Fig. 4) and clean combustion is maintained.
 Figure 5. Combustor performance with 600 liters/min excess air vs. various hydrogen flows in 200 liters/min of nitrogen. |
Our conclusion here was that we needed to set the base level of the variable air input at a value that ensured that all hydrogen input flows would be kept below the LFL in the exhaust. In addition, to maintain the exhaust as a clean mixture, it is necessary to interface the unit to the actual hydrogen flow being used on the low-pressure epitaxy process tool that it serves. When >50 slpm of hydrogen is flowing, additional air is added to the input to ensure clean combustion and maintain low carbon monoxide emissions.
Abatement of acid gases
The modified TCU (named Helios) performed best for low-pressure epitaxy combined with a three-stage wet scrubber as used in the standard product. Proven corrosion resistance and particle handling properties are the result of careful design and choice of construction materials. With this set-up, HF emissions from abatement of even the most demanding flows of PFC gases (e.g., CF4, C2F6, and NF3) are routinely below TLV (3ppm). The system can hande high flows of HCl; for example, with an input HCl flow of 30 slpm, the HCl output did not exceed 1ppm, well below the TLV of 5ppm.
Conclusion
Low-pressure epitaxial processing using high flows of hydrogen is rapidly increasing in use. The untreated exhausts from these process tools cannot be safely discharged into a common exhaust, so effective point-of-use treatment is essential. This cannot be achieved by simple exhaust abatement units because of the varied and alternating flows of process gases used. However, a refinement of inward fired combustion technology, which was originally developed for PFC gas treatment for CVD processing, provides a system that can reliably handle corrosives, pyrophorics, powders, and hydrogen in one system. This system could prove valuable to wafer fabs adding low pressure epitaxial processing to their lines.
Katsunori Takahashi received his chemical engineering degree from Niigata University. He leads the exhaust gas management group at BOC Edwards in Japan.
Andrew Seeley received his BSc in chemistry and his PhD in inorganic chemistry from the University of Bristol. He has spearheaded the development of several new exhaust management technologies at BOC Edwards.
Peter Mawle received his BSc in biochemistry from the University of Bristol and an MBA from London Business School. Mawle is global product manager for exhaust management systems at BOC Edwards, Kenn Rd., Clevedon, BS21 6TH, UK; ph 44/1275-337100, fax 44/1275-337200, e-mail [email protected].