More Than Just Carbon Dioxide

As discussed in Part Two of this series, lifecycle analyses of greenhouse gas emissions consider both direct and indirect sources. Indirect CO₂ emissions, attributed to electricity and other forms of energy purchased by the fab, are the semiconductor industry’s single largest environmental impact. Of those emissions, a large fraction are attributable to plasma-based etch and deposition steps, the most energy-intensive processes in the fab.

Those same deposition and etch processes are also major contributors to direct, Scope 1, fab emissions. Up until 1999, perfluorocarbons (PFCs) such as CF₄ and C₂F₆ were routinely used in dry etch and post-deposition chamber cleaning steps. With good stability and relatively low toxicity, they gave fab managers a versatile library of potential etch and cleaning chemistries.

Trace amounts, big impact

Unfortunately, PFCs also have substantially greater global warming potential than CO₂. As the name suggests, greenhouse gases in the upper atmosphere absorb infrared radiation, trapping it near the Earth’s surface rather than allowing it to escape into space. The global warming potential of a gas — measured in 100-year CO₂ equivalents — reflects both its IR absorption and the rate at which it will break down through atmospheric processes or be reabsorbed at the Earth’s surface. PFCs are problematic on both counts. They absorb strongly in the IR range and, being stable molecules that are almost entirely generated by human activity, there are few processes that will break them down or extract them from the atmosphere. (In contrast, CO₂ is naturally absorbed from the atmosphere by plants.)

Source: IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories,

Because of their stability, PFCs often have relatively low utilization rates in fab processes. Researchers at Chia Nan University in Taiwan estimated that as much as 60% to 70% of the CF₄ and C₂F₆ used in CVD reactor cleaning failed to break down and passed through to the chamber exhaust. For the same reason, thermal abatement is extremely challenging, requiring temperatures in excess of 1000 ºC.

In 1999, the World Semiconductor Council set voluntary targets for PFC emissions, setting its goal at 90% of 1995 emissions by 2010, and an additional 30% reduction relative to 2010 by 2020. The other large global PFC source, the aluminum smelting industry, likewise set voluntary reduction targets under the auspices of the Aluminum Industry Association. Both industries appear to have worked diligently to meet those goals.

In the aluminum industry, fluorine is used to help reduce oxygen from bauxite ore, with CF₄ as a by-product. There, according to Mike Czerniak,  environmental solutions business development manager at Edwards and an expert witness on CF₄ for the relevant IPCC working group, engineers are examining every step of the electrolysis process, identifying fault conditions that can increase emissions. For its part, the semiconductor industry has examined alternative etch and clean chemistries that can still meet process requirements. The 2010 WSC target was met primarily by substituting NF₃ for PFCs in CVD chamber cleaning.

Replace or abate?

However, as Czerniak pointed out, CF₄ is more difficult to replace in etch processes, where carbon “polymer” helps control the etch profile. Nor, as the table shows, is NF₃ a panacea. Though not a PFC, it too has high IR absorption and high atmospheric persistence, giving a 100-year CO₂ equivalent global warming impact of 8,000. On the positive side, NF₃ does break down more easily under process conditions, increasing utilization and improving abatement efficiency.

With the “easy” greenhouse gas reductions accomplished, efforts to meet the WSC’s 2020 target are focusing on abatement. According to Czerniak, Edwards abatement systems both supply energy to break the strong C-F binds, and add oxygen and hydrogen to give the carbon and flourine ions, respectively, alternative reactants. Vijay Venugopal, senior director of engineering management at Applied Materials, observed that some legacy fabs have little or no PFC abatement. Installing a pre-pump scrubber using heat and water vapor can substantially reduce PFC emissions from those fabs.   

Paul Stockman, head of market development at Linde Electronics, said that some companies have been using F₂ as an alternative to NF₃. Both break down to molecular fluorine, so both have similar process characteristics. F₂ is not a greenhouse gas, though, and can be generated on-site from anhydrous HF. (A future article will discuss on-site F₂ generation in more detail.)

Adding up the global inventory

A larger issue with NF₃, and with PFC emissions generally, is that, globally, substantial emissions are unaccounted for. Top-down measurements of these gases in the atmosphere find higher concentrations than can be explained by the sum of bottom-up emissions reported by the aluminum and semiconductor industries. It’s not clear where these “missing” emissions are coming from, making it difficult for regulators to address them. For example, these compounds are also used for etching and chamber cleaning in the display industry, where the large process chambers required by large sheets of glass increase the sheer volume of gas needed. The World LCD Industry Cooperation Committee, a counterpart to the WSC, has also set voluntary reduction targets for its members.

A 2014 paper from Jooil Kim and colleagues made a valuable contribution to the debate about missing PFCs. They measured the emissions plumes in Australia, where there are several aluminum smelters and no semiconductor manufacturers, and also in South Korea. This second location can detect emissions from South Korea, Japan, and Taiwan, which together account for large fractions of both global semiconductor production and global display production, but have few or no aluminum smelting businesses. From these results, they were able to estimate the fractions of C₂F₆ and CF₄ attributable to the two industries, and thus their respective contributions to missing PFC emissions.

The uncertainties in this kind of study are necessarily quite high. In the aluminum industry, most emissions calculations are based on “anode events,” where the level of bauxite ore in the electrolysis chamber is low and the carbon-based anode is exposed, but other phases of the smelting process may also contribute. Different smelters have different emissions characteristics, which may depend on both the smelting process and the abatement systems being used. In the semiconductor industry, the exact process gas mixes used by particular fabs are trade secrets, as are gas utilization, abatement efficiency, and other pertinent parameters. Emissions at gas manufacturer facilities, for instance during bottling, are essentially unknown. It is also unclear whether the authors of this paper recognized that display industry emissions are not included in World Semiconductor Council reporting. Still, this study suggests that the aluminum industry’s contribution to missing PFC emissions began to rise after 2001, which coincides with the growth of aluminum smelting in China. The semiconductor industry’s contribution to excess PFC emissions appears to have peaked in 2002, suggesting that manufacturing improvements, including increased use of NF3, have had a significant impact.

Conclusions
An ongoing theme in this series of articles has been the need to consider sustainability from the perspective of the full process. That is especially true with gases, like PFCs, that are difficult to either remove from the process or abate from the fab emissions stream. Once the easy solutions have been implemented, continuing to reduce emissions requires the same level of innovation in waste reduction and waste management that engineers have been applying to process improvement for decades.

Related Stories
Making Manufacturing Sustainable For Chips
Part 1: What is sustainability and why does it matter?
Saving Energy In The Fab
Part 2: Improving fab sustainability and better equipment and practices.
Sustainability Saves Water, Too
Part 3: How semiconductor manufacturing impacts water supplies and quality.
Think Globally, Act Globally (blog)
Fabs consume huge amount of electricity – does the industry have a responsibility to act sustainably?