Sustainable Chamber Cleaning Solutions: The Back End of the Front End

In most manufacturing processes, tool cleaning is often an afterthought in the process design and execution. Its impact is generally thought of as little more than housekeeping, akin to sweeping the shop floor at the end of the day. However, in microdevice fabrication, cleaning plays a critical role in the manufacturing process. The choice of technique affects not only device yield, but also fab productivity and environmental impact.

Because cleanliness is such a concern for device manufacturers, production equipment is housed in ultraclean environments, while the substrates themselves are transported in sealed carriers and transferred robotically into vacuum chambers and coater tracks. And no wonder. Microscopic detritus can ruin high-end processors retailing at $1000 a piece, extinguish pixels on the flashiest of flat screens, and darken conversion efficiencies of solar panels to gray skies-only returns.

Despite all of this best practice, deposition processes, such as chemical vapor deposition (CVD), coat parts of the chambers as well as the substrates with their gas-to-solid chemistries. While wafers and glasses move on to further device-building steps, chamber walls and mechanicals accumulate ever thicker and non-homogenous layers of unwanted waste. As a result, timely and efficient cleaning of production chambers to remove device-threatening particles and maintain process stability is required to protect hard-won yields.

Most cleaning processes employ reactive gas-surface chemistries to volatilize the dross. The typical active cleaning agent is free fluorine. However, the traditional gases used to deliver this fluorine tend to be hard-to-abate, perfluorocarbon (PFC) gases, which contribute substantially to global warming.¹

Design trends toward more complex devices — smaller features and higher aspect ratios on silicon wafers, and thicker and more numerous layers on ever-larger glasses — dictate that more cleaning must take place. Despite the added complexity and cost of these new and additional processes, the chamber cleaning step is expected to hold the line for aggregate process cost with little or no impact on overall tool throughput.

Without cleaning process changes, more cleaning will result in more global warming emissions. This poses big challenges, because the industry is committed to a 10% reduction in CO₂ emissions from its 1995 levels. The need to develop sustainable ways to create more, while minimizing environmental impact by using less, is a major factor behind the development of new chamber cleaning solutions.

Two of these solutions, sulfur hexafluoride (SF₆) recovery and on-site generated fluorine (F₂), can help to overcome these challenges because, in different ways, each provides both increased productivity and reduced environmental impact. For example, SF₆ recovery allows existing users of highly economical SF₆ to significantly reduce their per-clean usage and eliminate the extremely high environmental impact of this otherwise benign gas. Meanwhile, on-site generated fluorine cleans both small and large chambers faster at lower cost with a lower environmental impact than the commonly used competing cleaning gases, such as nitrogen trifluoride (NF₃).

On-site generated fluorine

For semiconductor thermal deposition processes, thermolytic cleaning with fluorine is the preferred method for most end users over offline, off-site wet cleaning, which is process-disruptive and necessitates duplicate chamber equipment kits to maintain acceptable productivity. Because of its much lower bond strength, fluorine can be thermolytically activated at process temperatures that would be ineffective with the much less-reactive NF₃ or PFCs (Table 1). This has increasing importance as more front-end processes are carried out at lower and lower temperatures.

On-site generated fluorine (see sidebar) is the safest and most cost-effective way of supplying that fluorine for both medium and large fabs. High-pressure cylinder fluorine already has significant distribution and handling risks with various restrictions on cylinder size and fluorine concentration. In Asia, there are plans to increase the fluorine cylinder pressure for greater productivity, further increasing the risks during filling, transportation and handling — risks that can be eliminated with an on-site generated supply and at a lower cost for a higher-purity product.

Greater productivity

For larger cleaning supply schemes, like those supplying thin-film transistor (TFT)-LCD and thin-film solar production fabs, on-site generated fluorine provides value in both scaled costs and process improvement relative to nitrogen trifluoride, which is often considered the default choice for chamber cleaning. However, NF₃, which is produced by electrolysis — either directly or by using fluorine as a starting material — is more expensive to produce and has a larger carbon footprint than elemental fluorine, and that disparity scales with larger-use applications. It also requires 3.5× the input energy to create the same amount of atomic fluorine from elemental fluorine caused by the large disparity in bond strengths (Fig. 1).

1. F2 vs. NF3 throughput for typical remote plasma source.

This is not simply unnecessary and wasteful consumption; it also limits the useful throughput of the remote plasma source (RPS) used to energize the cleaning gas to form reactive atomic radicals. Because chamber cleaning rates are directly related to the throughput of atomic fluorine, on-site fluorine can clean faster than NF₃ in a set equipment configuration.

Although higher flows of fluorine are required to provide the same throughput of atomic fluorine as NF₃, less mass is used because of the differences in molecular weights, resulting in a lower gas cost per clean. At the same time, faster etch rates lead to a reduction in clean time and, thus, an improvement in tool throughput.

The atomic nitrogen produced from NF₃ plasmas not only contributes to the higher activation energy, but its recombination produces unwanted heat, potentially requiring the cost and complexity of additional cooling systems.

The cleaning gas requirements for Gen 8 size panels already exceed the actual capacity of standard RPS units with NF₃, requiring either multiple units be used or new high-power sources developed. Either option adds unnecessary and significant cost, and the problems will continue to grow as panel sizes continue to increase. On-site fluorine can meet projected clean flow requirements using standard equipment for all announced panel sizes.

Environmental advantages

Most cleaning gases have a strong global warming potential (GWP) and, as the square meter size of both glasses and annual fab throughput increases, so do concerns about the environmental damage associated with chamber cleaning. The GWP associated with NF₃ is particularly high.³ Manufacturers estimate that RPSs are only 95% efficient in dissociating NF₃. Added to surface recombination and typical process cleaning recipe over-etch, a significant amount of unreacted NF₃ passes through the process chamber. End users vary to the extent and the methods with which they abate NF₃. However, researchers have shown that the most common form of NF₃ abatement may just be trading one high GWP gas for another, as thermal combustors produce nearly like amounts of CF4, as well as NO_x, from NF₃.⁴ Table 2 illustrates the GWP values of chamber cleaning gases.

In contrast, on-site generated fluorine provides a sustainable environmental solution. It has a GWP of zero, and its overall environmental impact is low because unused fluorine is readily abated using liquid scrubbers and, ultimately, sequestered as solid waste.

Production proven

On-site fluorine generation (Fig. 2) has proved very reliable in high-volume production (more than 25 units have been placed in fabs worldwide) and safe, because its use greatly reduces the need for operators to interact with the product. In addition, its inherently low temperature, pressure and inventory meet both industry standards and end users' safety and licensing needs.

2. Typical on-site fluorine generator installation.

In TFT production, replacement of NF₃ with fluorine has reduced cleaning times by as much as 40%, enabling a tool throughput increase at the same time as eliminating the release of significant quantities of global warming gases. By analogy, the same benefits should be realized for thin-film solar panel production, helping drive down cost/watt while minimizing the carbon footprint.

SF6 recovery: lower cost, GWP

In the production of large solar and LCD panels, SF₆ is often a logical choice for chamber cleaning because it is non-toxic and non-corrosive, making it a material that is easy to handle while also enjoying the advantage of low cost. However, it is a known greenhouse gas with a GWP greater than 22,000; thus, any untreated emissions of SF₆ carry a severe environmental impact. In addition, only a fraction of the SF₆ is consumed in the CVD reactor clean cycle. The unconsumed gas is exhausted to the house abatement system, resulting in low process efficiency, which offsets the cost advantage of SF₆.

However, SF₆ recycling systems designed to eliminate these disadvantages are commercially available and, once installed and integrated into the fabrication process, can serve as the point of supply for SF₆ used in the CVD reactor clean step of the manufacturing process. An example of an SF₆ recycling system is shown in Figure 3.

3. SF6 recovery process flow diagram.

In these systems, the tool's process exhaust is diverted to the fully automated and self-contained recovery system where it is processed through a series of unit operations to purify the residual SF₆ of the etchant byproducts from the chamber. These typically include fluorinated compounds, acid fluorides and residual fluorine. The purified process exhaust is then compressed and dried prior to a final recovery and purification step to produce recovered SF₆, which is >99.9% pure. The recycled SF₆ is then supplemented with virgin gas to make up the quantity of SF₆ consumed in the reactor clean. Finally, it is returned to the tool to feed the clean cycle on an on-demand basis.

Because all unconsumed SF₆ from the chamber clean is recovered by the system, the effective use of SF₆ in the chamber clean step is 100%, resulting in good cost efficiency and the total elimination of a known greenhouse gas from the exhaust.

Cost efficiencies are further enhanced as systems typically manage exhausts from multiple tools. Remote operation capability is built into the system to enable easy troubleshooting in the event of any operating problems.

Summary

Commercialized and production-proven chamber cleaning solutions now allow device manufacturers to deliver increased productivity and reduce their environmental impact. On-site generated fluorine achieves the fastest clean at inherently competitive prices and with zero GWP contributions. Alternatively, SF₆ complete recovery systems permit equipment manufacturers and end users to take advantage of a highly economical cleaning reagent, while eliminating the GWP emissions and further reducing the material consumption. Either way a fabricator chooses, these chamber cleaning solutions are part of a new portfolio of sustainable processes engineered to maintain both the bottom line and the environment.