Home
About Future-Fab
Download New Issue
Contact Us
Volume Archives
Editorial Panel

FUTURE FAB ARCHIVES


Reducing PFC Emissions: A Technology Update
(1/7/2000) Future Fab Intl. Issue 9
By Walter Worth, International SEMATECH
Print Article Print this paper
Send As Email Send as email



1.0 Introduction

As was stated above, the US semiconductor industry’s effort was spear-headed by the microchip manufacturers through International SEMATECH with the collaboration of tool suppliers, chemical and gas suppliers and universities (see Figure 2). It initially addressed the use of PFCs in the cleaning of tool chambers after plasma enhanced chemical vapor deposition (PECVD). Back in 1993, about 60-75% of the emissions were believed to originate with the chamber cleaning process, the rest coming form etch processes. Furthermore, early tool exhaust analyses indicated that at best only 30-35% of the PFCs are consumed in the process and that the rest is emitted to the atmosphere. Additional, more rigorous analyses of the tool emissions by quadrapole mass spectrometry showed that 5-15% of the PFCs fed to the tool are converted to CF4 as a by-product of the chamber cleaning process. This was a very important finding as CF4 is probably the most stable, long-lived PFC of them all and appears to be an unavoidable by-product of any plasma process using PFCs.

Figure 1. Global Warming Potentials and Lifetimes of PFCs.

Figure 2. Industry Collaboration.

2.0 A Chronology

A) Abatement

The first approach was to treat these emissions with point-of-use abatement at the tool using “burn-boxes” (thermal oxidation), a technology the industry was already intimately familiar with in dealing with pyrophorics. However, evaluations of these devices quickly showed that efficient destruction of PFCs, especially the very stable CF4, required higher temperatures, in the 900-1000ºC range. Such elevated temperatures may lead to the production of other undesirable by-products, such as nitrogen oxides, one of the precursors of ground level air pollution. In addition, these devices require significant quantities of water for cooling and scrubbing out the large amounts of HF released by the destruction of the PFCs[2]. Similarly, treatment of the PFC emissions by the existing, packed (dry granules) reactor columns used for abatement of toxics such as arsine and phosphine was considered, but it was found that these reactor columns can only partially treat PFC emissions. In addition, the spent reactor contents always require disposal, and the final resting place of these canisters, potentially in landfills, implies an undesirable risk of future liability (eg, the Superfund sites in the USA).

 

B) Alternative Chemistries

This led the industry on the search for PFC replacement chemicals with significantly lower or, better yet, no global warming effect. SEMATECH engaged the Massachusetts Institute of Technology (MIT) to screen whole family of chemicals such as hydrofluorocarbons (HFCs), unsaturated fluorocarbons (UFCs) and iodofluocarbons (IFCs). In the meantime, C3F8 was tested and shown to be a potential “drop-in” replacement for C2F6 in some tools[3]. Although C3F8 is still a PFC, it is dissociated to a higher degree in the plasma during the chamber clean, resulting in significantly higher utilization of the chemical, ie, ~65-70% for C3F8 vs ~25-35% for the standard process using C2F6. While this does not completely eliminate the PFC emissions, it does take a significant step in the right direction without the need for a change in equipment or a negative effect on the process; hence it is referred to as “drop-in replacement”.

More recently IBM developed a dilute NF3 (with helium), in-situ chamber cleaning process for both Novellus and Applied Materials CVD tools that appears to be very promising. According to IBM, in going from C2F6 to NF3, the gas utilization increases from 31-37% to 85-91% and the emissions are reduced by 97-99%. Novellus is actively beta testing the process for oxide and tungsten deposition and hopes to release this process soon for its Concept Two Sequel 200mm tools.

 

C) Recovery and Recycle

Given that PFC utilization is low in both the chamber cleaning and the etch process, and considering that there are a large number of PECVD and etch tools in a typical 200 mm production fab, the idea of combining and collecting the exhausts from the various tools in a common duct for recovery of the unreacted PFCs has considerable appeal. Several technical approaches from pressure swing absorption to membrane separation to cryogenic capture were tested. Figures 3 and 4 show a simplified process flow diagram and a photograph of the pilot plant used to evaluate the cryogenic process at Texas Instruments. Of these techniques at least the latter two showed real promise. However, pre-conditioning of the exhaust gas mixture to remove hydrogen fluoride (HF), particles, moisture, etc. prior to the separation of the PFCs from the nitrogen-rich stream proved quite challenging and costly. Even the disposition of the recovered PFCs became a question mark, as the recycled material would always be suspect by the process engineers.

Figure 3. Simplified Process Flow Diagram for Cryogenic PFC Recovery.

Figure 4. Photograph of Pilot Plant for Cryogenic PFC Recovery.

D) Process Optimization

Work by the tool suppliers showed that there was an opportunity to optimize the existing chamber clean process by small hardware modifications such as the addition of Applied Materials’ plenum pumping plate and moving from a two-step to a one step cleaning process for post-TEOS deposition chamber cleaning. The results were dramatic and resulted in a 30% reduction in clean time, a 50% reduction in gas usage and a 30% lower cost of ownership, according to Applied Materials’ announcement[4]. Around the same time, one of the chip manufacturers showed that through monitoring of the chamber cleaning endpoint by a technique such as plasma diagnostics, the PFC emissions could be significantly reduced for the same Applied Materials CVD tool. The chamber clean process in use was using a time-based recipe, which included a significant safety factor. It was found that clean gas (C2F6) flow and clean time could be reduced by 35% and 25%, respectively, without any adverse effects on the process[4].

E) Plasma Techniques

Most recently two plasma techniques have appeared on the scene, which promise to become cost-effective technologies for reducing PFC emissions. One is the remote plasma technology being marketed by Applied Materials for chamber cleans. It uses NF3 as cleaning gas and dissociates it almost completely in a microwave plasma upstream of the PECVD chamber eliminating PFC emissions almost totally. The other new technology involves the use of an Rf plasma in the tool foreline, upstream of the dry pump, to abate the emissions from etch tools. Again, this technique achieves >95% destruction efficiencies for the PFCs commonly used in dielectric and nitride etch processes.

F) Lessons Learned

As can be seen, the US search for a practical, cost-effective solution for reducing PFC emissions from semiconductor manufacturing has come full circle in just seven years. The industry started with abatement, then looked for other solutions such as process optimization, alternative chemistries and recovery with, ideally, recycle of the recovered, purified chemicals back to the tools. On this road of discovery it became apparent that no one solution fits all and the “greener” technologies are frequently more difficult to achieve and more costly to implement.

For example, what makes sense for a small fab may not be cost-effective for a large fab. Similarly, abatement may make sense for a few tools, while collection of tool exhausts for either PFC recovery and recycle or destruction of the recovered PFC mixture may make sense for a large fab. Furthermore, a solution such as the installation of more efficient, “cleaner”, next-generation tools, may be available for a new fab, but may be out of the question for older production fabs.

It also became apparent that, over the last six or seven years, the tool suppliers had made great strides in reducing the emissions from their new tools by better gas utilization through higher density plasmas.

3.0 Solutions Currently Favored for CVD Chamber Cleans

Based on all the research and development over the last seven years, several technologies have risen to the surface and are currently favored by the industry and are actually being implemented in preparation for the year 2010 deadline. The following technology solutions are now commercially available for CVD chamber cleans:

Switching to PFCs that are more fully dissociated in the plasma

Installing reactive fluorine generators upstream of the chamber

Installing point-of-use (POU) thermal oxidizers

 

These options will now be discussed in more detail with reference to recent technology evaluations sponsored by International SEMATECH and performed at its member companies.

A) Switching from C2F6 to C3F8 in a Novellus Tool

An extended evaluation of C3F8 as a possible replacement for C2F6 has shown that indeed C3F8 is a “drop-in” replacement for C2F6 for some tools. The evaluation was made in 1997 at Advanced Micro Devices (AMD) on a Novellus Concept Two dielectric process and included a full Design of Experiments (DOE) followed by a 3000 wafer marathon[3]. The project included a complete set of film characterizations including film thickness, film uniformity, film stress, refractive index and particles. All these parameters were found to be well within specifications and without any sign of process drift over time. In addition to the excellent process performance, it was found that the utilization of the C3F8 was considerably greater than that of the standard C2F6 process, 63-71% versus the 24-35% experienced with C2F6 (see Figure 5). This resulted in an approximately 60% reduction in emissions over the standard C2F6 chamber clean process, measured in terms of carbon equivalents and accounting for the global warming potentials (GWP) of the unreacted C3F8 as well as the by-products (C2F6 and CF4). Since then, several major chip manufacturers such as AMD, Motorola and Texas Instruments have implemented the conversion to C3F8 in many of their fabs with excellent process results and chemical cost savings. However, in the meantime, Novellus as well as Applied Materials have gone back and optimized their tools and process recipes making significant progress in both the gas utilization and the emissions reduction. For, example, between 1995 and 1999, Applied Materials was able to reduce emissions from its TEOS oxide film CVD tools by 87% in going form the two-step, in-situ C2F6 clean process in the older, lamp-heated tool to the in-situ C2F6 process in the newer generation DxZ tool (see Figure 6[6]). Thus, today the advantage of the C3F8 process over the newer tool C2F6 chamber clean process may no longer be significant.

Figure 5. C2F8 verses C3F8 Gas Utilization.

Figure 6. Example of Emissions Reductions with New Tools, Applied Materials, 1995-1999 Time Period.

B) Installing Applied Materials’ Remote Clean™ on DxZ Chambers

Installation of Applied Materials’ Remote Clean™ technology on the DxZ chamber has opened up another avenue for significantly reducing emissions (up to 99%). The latest evaluation[7] completed in 1999 at Motorola showed that this technology works very well. It uses NF3 and dissociates the gas in a high density, microwave plasma, upstream of the CVD chamber. The device mounts on the chamber lid and the reactive fluorine species travel from there through the existing “shower head” into the chamber (see Figure 7). During the cleaning step, there is no plasma in the chamber, which eliminates the ion bombardment and results in a “soft” clean and increased life of chamber consumables. The process appeared to be very robust and required no “tuning” (ie, adjustment) during the whole evaluation, which included a DOE as well as two separate 1000-wafer marathons. Again, film properties including film thickness, film uniformity, film stress, refractive index and particles were well within the specifications of the baseline in situ C2F6 process. There are only three features that currently detract from this technology. They are the relative high cost of NF3 compared to C2F6 on a per pound basis, the additional cost of the NF3 gas distribution inside the fab to the tool (if not already in place), and the capital cost of the microwave device of approximately $60K per chamber.

Figure 7. Schematic Diagram of Applied Materials Remote Clean Technology.

C) Installing a Thermal Oxidizers

Less desirable, or possibly less “green”, is the use of abatement to reduce PFC emissions through a thermal oxidizer. Not only is the capital cost significant, but the cost of operation or cost-of-ownership (CoO) can be sizable with cooling water, fuel and waste treatment/disposal being the major contributors to that cost. However, there are cases when a thermal oxidizer such as Edwards Thermal Processing Unit (TPU) makes sense. For example, a POU thermal oxidizer may already be required at the tool to treat silane, arsine, phosphine or other pyrophoric and toxic materials. In addition, the use of reclaimed water and cooling coils can lead to a reduction in the fresh water requirement and cost of operation of such devices.

International Sematech has supported the development and evaluation of the first commercial unit, Edwards TPU 4200. While this unit had some difficulty in destroying CF4, subsequent modifications have led to the TPU 4214, which can effectively destroy all PFCs at >99% efficiency (see Figure 8)[8]. The unit uses natural gas as fuel and conserves water by recycling and heat exchange. In addition, it has a novel inwardly fired ceramic matrix burner that results in a very even oxidation of the emissions, which tends to minimize the generation of the nitrogen oxides (NOx) normally associated with open-flame combustion. As precursors in the formation of ground-level smog, the nitrogen oxides are highly regulated in the US as hazardous air pollutants (HAPs).

Figure 8. Example of a Combustion Device, Edwards TPU 4214 (Photograph courtesy of BOC Edwards).

4. Solutions Currently Favoured for Etch

Although 70-90% of the PFC emissions of a typical semiconductor manufacturing facility are attributed to CVD chamber cleans, several technologies including process optimization, alternative chemistries, plasma abatement and catalytic conversion are being pursued for reducing the PFC emissions from etch tools.

A) Process Optimization

Some of the chip manufacturers have shown that it is possible to cut emission for existing tools and process recipes. For example, STMicroelectronics in collaboration with Lam Research has been able to reduce emissions by up to 50% without any impact on process performance by optimizing Rf power, pressure, gap and PFC flow[9]. As mentioned earlier, the tool suppliers, in turn, have successfully increased source gas utilization, with a concomitant reduction in emissions, in their newer generation tools by going to higher density plasmas.

 

B) Alternative Etch Chemicals

The search for alternative etch chemicals has been pursued for several years, but has so far not produced any replacements as ideal as the PFCs currently in use from a process performance point of view. It goes without saying that any alternative chemical has to meet current process performance, as a minimum, to make it acceptable to the process engineer. Many of the chemicals tested to date do not have the right etch rate, resist selectivity or sidewall polymer formation that is required for the anisotropic etching at high aspect ratios.

PFC emissions in etch originate from two sources: 1) incomplete utilization of the etch gas and 2) the formation of CF4, the seemingly unavoidable by-product of every plasma process that uses PFCs. To circumvent this problem, MIT is exploring a novel approach[10]. MIT wants to deliver the fluorine that does the etching in the form of an inorganic fluorine source with the addition of just enough hydrocarbon for the sidewall passivation required for anisotropic etching. The assumption is that in a plasma the origin of the reactive species is irrelevant in as long as the atomic species are present in the correct proportion. Preliminary results from proof-of-concept experiments using NF3 and acetylene (C2H2) or ethylene (C2H4) in a medium density etcher were very positive with an 80% reduction of emissions over the C3F8 baseline etch process and very similar process performance (see Figure 9). This work is currently being continued at MIT.

Figure 9. Comparison of NF3 Etch with Conventional C3F8 Etch.

However, in contrast to the chamber cleaning, for etch there is little hope of finding “drop-in” replacement chemicals. Any new etch chemistries would require extensive tool/process re-qualification. Therefore, the new alternative chemicals being developed will most likely have to be introduced through future generation etch tools.

 

C) Abatement

This leaves abatement as probably the best current path for reducing emissions from etch tools. There are two promising technologies that International SEMATECH has recently evaluated. The one that appears to be most cost effective is an Rf plasma device such as the Litmas “Blue” which is mounted in the foreline, between the turbo and the roughing pumps (see Figures 10, 11 and 12) [11]. An eight-month long evaluation showed that the device was easily integrated with the tool, was very reliable and required no service over the test period. For the etch recipes tested it reduced emissions by 97–99% and only produced HF and other water soluble by-products, when water vapor was added to the plasma to avoid CF4 re-formation. There was concern that the increase in foreline pressure from approximately 200 to 375 mTorr due to the device and, potentially, back-streaming of water vapor into the etch chamber would affect the etch process. However, no impact on the process due to the change in foreline pressure was observed and an optical emission spectroscopic analysis did not detect the signature peaks of OH and H that would indicate the presence of water vapor in the chamber. In addition, the fear that the formation of HF in the foreline in the presence of the moisture would corrode the pump internals or any o-rings downstream of the pump was unfounded. Visual inspection of the pump internals and pump grease after the eight-month evaluation revealed no signs of corrosion, particles or other deposits. At $20-25K per device for the capital cost and an annual operating cost of ~$1,000, this technology appears to be quite affordable. These costs assume that the acid ductwork can handle the HF produced without the need for POU a water scrubber.

Figure 10. Specifications for the Litmas “Blue” Abatement Device (Photograph courtesy of Motorola).

Figure 11. Schematic of the Litmas “Blue” Test Installation.

Figure 12. Litmas “Blue” Abatement Results.

The other etch abatement option still under evaluation is the Hitachi catalytic converter. This technology uses a catalyst bed to convert the PFCs, in the presence of air and moisture, to HF, SO3 and CO2 at a temperature of 500-750oC. The catalyst life is estimated to be greater than 6 months. In the current application, the device abates the emissions from a SCALPEL mask (silicon) making operation, which uses alternating flows of SF6 and C4F8 (300 and 160 sccm of SF6 and C4F8, respectively). The ongoing evaluation shows that even at these relatively high PFC flows the unit successfully reduces PFC emissions by >99%. The system includes a pre-scrubber, to remove particles and add moisture to the gas to be treated, and a post reaction scrubber to remove the HF and SO3. A post-scrubber blower pulls the gases through the system. This system is more complicated and uses about 3KW of power for the heater and 4.5 lpm of water. With a capital cost of >$100K for a 60 lpm unit and a sizable foot print of roughly 37” x 22”, this device appears not to present as elegant a solution as the Litmas “Blue” device. However, in this case the device is downstream of the roughing pump and eliminates process risks almost completely. This additional security is very important to some process engineers.

5.0 Conclusions

The semiconductor industry became aware of the PFC global warming issue only about 7 years ago, but since then, awareness of this dilemma has spread around the globe and significant strides have been made in developing the technologies to address this challenge. While it was too optimistic to assume that one solution would fit all needs – etch as well as chamber clean processes, small as well as large fabs, new tools as well as older generation tools, a number of options are now commercially available. For PECVD chamber cleans switching to PFCs that are more fully dissociated in the plasma or installing reactive fluorine generators upstream of the chamber currently appear to be the favored technologies. For etch tools, there is the abatement of the PFCs by a plasma device in the foreline or the post-pump catalytic conversion process. More long-term the industry is holding out hope that eventually alternative chemistries will be found which are non-global warming and, at the same time, can match the ideal process performance of the PFCs.

Figure 13. Process Flow Diagram for the Hitachi Catalytic Converter.

6.0 References

[1] US Climate Change Report, Chapter 3, 1997.

[2] Laurie Beu et al, “Results of DELATECH Controlled Decomposition/Oxidation (CDO) Unit Testing for PFC Emission Abatement Applications”, SEMATECH Tech Transfer (TT) document #94092543A-ENG, Oct 31, 1994.

[3] Sey-Ping Sun, “Evaluation of C3F8 as a Cleaning Gas in a Novellus Concept Two Sequel Tool, SEMATECH TT document #97053282A-TR, May 31, 1997.

[4] Applied Materials Technical Update for Precision 5000, Mach II D-CVD, “TEOS Single-Step Clean with Plenum Pumping Plate”, Nov 1993.

[5] Jim Pinto, IBM Burlington, VT, “AMAT 5000 CVD Fourth StateTechnology RF Metrology Project TEOS/PSG”, presentation to SEMATECH EPIT Group, February 27, 1995.

[6] Hichem M’Saad, “CVD low k Solutions for Sub-0.18µm Technology”, presentation at NSF/SRC Engineering Research Center for Environmentally Benign Semiconductor Manufacturing, Thrust D, TeleSeminar, March 2, 2000.

[7] Laura Mendicino et al, “Evaluation of Applied Materials DxZ Remote CleanTM Technology”, TT document # 99113842A-ENG, Nov. 30, 1999.

[8] Tom Walling et al, “Evaluation of an Edwards TPU 4214 and an EcoSys Phoenix IV for CF4 Abatement”, TT document #97073319A-TR, Sept 30, 1997.

[9] F Illuzzi, M Molgg, L Colombo, L Atzei, “Etching Process Optimization: A Way to Reduce Emissions”, 4th International Conference, Milan, Italy, June 22-24, 1997.

[10] Laura Pruette et al, “NF3-Based Oxide Etch Processes for PFC Emissions Reduction Feasibility Study”, TT document # 99113851A-TR, Nov 30, 1999.

[11] Victor Vartanian et al, “Long-Term Evaluation of the Litmas ‘Blue’ Plasma Device for POU PFC and HFC Abatement”, TT document # 99123865A-ENG, Jan 7, 2000.

Biography

Walter Worth is a Program Manager in International SEMATECH’s Environment, Safety and Health Technology Development Department. He earned a BASc at the University of Toronto and a MS and PhD at the Massachusetts Institute of Technology, all in Chemical Engineering. He is a Professional Engineer registered in Texas and an active member of the Semiconductor Safety Association, the American Institute of Chemical Engineers and the Electrochemical Society. He has 25 years of environmental experience in the petrochemical and the semiconductor industry.

 
 
Search



Published By:
?¡Àcheap #167
38 Miller Ave.
Mill Valley, CA 94941
www.mazikmedia.com
converse@mazikmedia.com
Disclaimer | Privacy Policy