Technical Insight
An increased portfolio for waste gas abatement
The planned merger of CS Clean Systems and Centrotherm Clean Solutions creates a single-source vendor offering a full suite of exhaust gas abatement products
WORK in the compound semiconductor industry and you will face far greater challenges than colleagues working with silicon. It's not just that you have to strive far harder to minimize defects, and are unable to access design tools and measurements systems that are as sophisticated as those in the silicon sector "“ safety is more difficult to realise, due to the use of a wider range of gases that require careful handling. And on top of this, the abatement of the waste process gas requires more thought, partly because it can contain a greater variety of by-products.
CS Clean Solutions will offer a very broad portfolio of approaches to help with the safe treatment of exhaust gases resulting from the production of compound semiconductor chips. The new company will combine the expertise of CS Clean Systems, which has a pedigree in dry bed removal of hazardous gases that dates back to the 1980s, with the capabilities of Centrotherm Clean Solutions, a leading manufacturer of burner-wet and related thermal scrubbers.
All these companies, whether high-volume chipmakers or small start-ups, can soon draw on our full suite of trusted abatement products and our enlarged team of waste gas management experts. We are best prepared to support our customers and meet the challenges of the future.
At the heart of the manufacture of compound semiconductor devices is the growth of thin films, either by MBE or MOCVD. The former finds its greatest use in the deposition of films for high-frequency power amplifiers, HEMTs and HBTs. As MBE tends to involve solid-source precursors, the need to handle waste gases downstream is often avoided, with the focus on the safe entrapment and disposal of solid by-products, which often include pyrophoric white phosphorous and arsenic.
The situation is very different when films are grown by MOCVD, the most common method for making LEDs. This growth technology involves group III metal-organics and group V hydrides that are either pyrophoric, highly toxic, or fit into both of these categories. The safety challenges associated with these materials surpasses that in the silicon industry by some margin, due to litre-per-minute flows of toxic hydrides, matched with simultaneous usage of pyrophoric metal-organics, and the predominance of hydrogen carrier gas in the exhaust.
Adding to the challenge is the length of the growth process. It can last for several hours, which is significantly longer than that associated with plasma etching or PECVD. And complicating matters even further, throughout the process cycle the concentrations of the individual exhaust species fluctuate strongly as growth steps stop and start.
Heading the list of concerns are two highly toxic, widely used gases: arsine and phosphine. In silicon foundries, these gases provide a source for n-type doping, with typical flow rates of 5 sccm to 20 sccm. That is about two orders of magnitude less than that used for the growth of III-V films, where arsine flow rates typically range from 300 sccm to 3000 sccm, and that for phosphine can hit 5000 sccm. Consumption rates for metal-organic precursors are rarely disclosed by chipmakers, but can be of the order of a gram or so per minute.
Dry bed scrubbing
Dry bed scrubbing is the most common technology for handling waste products generated during growth of GaAs and related material systems. This approach is favoured, because it requires very little maintenance, so operators rarely have to intervene to replace or service contaminated components. Unlike wet-scrubbing methods, the surplus III-V reactants and their MOCVD by-products end up as solids within the absorber column, so there is no liquid or sludge waste that must be handled and disposed by on-site service personnel. For our customers, there is also the benefit of a network of local service centres for emptying and refilling absorber columns.
There are two different classes of dry bed scrubbing. One option is based on physisorption, which is short for physical adsorption. What we advocate, however, is to use chemisorption "“ in other words, chemical absorption.
The physical adsorption approach for treating MOCVD exhaust products involves bonding of arsine and phosphine gases to a high-surface-area medium, such as charcoal. To enhance the retention performance of these materials, commercially available charcoal or "˜activated carbon' media are often impregnated with metal oxides. However, despite this modification to the material, most of the adsorbed arsine and phosphine adheres to the carbon surface in its original "“ highly reactive "“ gaseous state.
One of the steps taken with physical adsorption is to periodically treat the bound hydride gases with a dilute mixture of air. This converts them into solid arsenic and phosphorous oxides. During this step, when the expended cartridge is air-treated, incoming exhaust gas is routed through a second cartridge. This cycle is reversed and repeated, typically every few days.
Rewind the clock and there was a time when impregnated charcoal was widely used for the abatement of MOCVD exhaust gases. However, recently there has been a sharp decline in the use of this approach, due to the transition of MOCVD from research and development to mass production, and a heightened awareness of worker safety.
Our preference for dry scrubbing is to use chemisorption, which converts by-products to solid at an ambient temperature. Chemisorption can work with a range of hydrides, including arsine, phosphine, germane, stibine and disilane; and also commonly used metal-organics. Our product line, known as Cleansorb, involves the reaction of gases with a chemisorber granulate in an absorber column. The reaction is immediate, with no reliance on other media or heating.
When we install our Cleansorb systems at customer sites, we specify the retention efficiency of our dry bed columns in terms of the absolute concentration at the column outlet, rather than as a percentage of inlet concentration. The gas concentration at the column's outlet lies below the detection level of commonly used gas detectors − right up until the end of absorber capacity lifetime.
Selectivity is essential in the dry bed chemisorption process used during epiwafer growth by MOCVD. It is critical that the chosen absorber granulate must be reactive enough to enable effective, safe removal of waste gases and precursors, but it must not undergo any reaction with the bulk carrier gas, hydrogen. Of all the common hydride species, phosphine is the most reactive, followed by arsine, silane, germane and then hydrogen.
The challenge of removing group IV and V gases to below detection level, while allowing hydrogen to pass unabated, is even tougher at higher hydride flowrates. That's because the exothermic scrubbing reaction generates heat, which promotes the reaction between hydrogen and the absorber bed. If non-dedicated absorber media are deployed, reactions with hydrogen can take place at arsine and phosphine flow rates as low as around 1 slm. If this occurs, the temperature can rapidly ramp up, due to "˜thermal runaway'. Chipmakers then lose time and money, because the scrubber must be shut down due to overheating.
Preventing scrubber overheating is not the only challenge. It is also a necessity to tailor chemisorber chemistry to ensure controlled conversion of process by-products, while avoiding vigorous chemical reactions "“ especially with highly pyrophoric species such as hydrazine and metal alkyls.
To address all of these concerns, we have developed a range of Cleansorb granulates that are specifically tailored to the needs of individual MOCVD applications. Examples include selective removal of the highly poisonous, but relatively unreactive germane from hydrogen carrier gas; safe scrubbing of arsine and phosphine at combined flows of 5 slm and well beyond, while maintaining a safe energy barrier to reaction with hydrogen; selective removal of Group III metals from GaN and other MOCVD exhausts; and non-charcoal based removal of chlorine gas, which is increasingly used for in-situ cleaning of MOCVD chambers.
Handling hydrogen
Hydrogen is employed as the carrier gas in most MOCVD recipes for depositing arsenides, phosphides and nitrides. Though sometimes matched in flow rate by the nitrogen dilution added to the pump, hydrogen is often the major component of the exhaust gas. It is not relevant in terms of its toxicity, and it doesn't contribute to global warming "“ but it can form explosive mixtures with air over a wide concentration range, so its safe handling is obviously the primary concern.
One step MOCVD engineers must take to prevent any explosions associated with hydrogen gas is to ensure that exhaust lines are free of leaks and other sources of air incursion. On top of this, measures must be taken to deal with hydrogen that are tailored to the particular materials being deposited. These measures may differ from those employed in silicon semiconductor facilities, which have traditionally used thermal and burner-wet scrubbers to incinerate hydrogen at the point of exhaust. If toxic metals are a concern for regulatory compliance, chipmakers may use dry-bed/ burner-wet scrubber combinations to prevent the contamination of scrubber waste water. This approach is particularly common when arsenic is used.
For the production of optoelectronic devices, LEDs and photovoltaics, and for the growth of wafers in III-V foundries, hydrogen tends to be diluted with air. With this approach, known as safe dilution, engineers aim for a maximum hydrogen-in-air concentration of 2 percent or even 1 percent. This corresponds to about 50 percent to 25 percent of the lower explosion limit. In some facilities, there are purpose-built dilution stations that feature a side-channel blower, which extracts the hydrogen stream into an auxiliary high flow of air via a series of non-return flaps. Once safely diluted, hydrogen can be released to atmosphere without any negative impact on the environment.
Gases for GaN
Ammonia is nearly always used for the growth of GaN. This precursor is used regardless of whether the substrate is sapphire, SiC or silicon, and whether the device will be an LED, laser or power transistor. In the growth chamber, ammonia reacts with a metal-organic precursor to produce GaN and a hydrocarbon gas.
At first glance, it appears that one of the advantages of the nitride MOCVD process is that it is less toxic than its phosphide and arsenide counterparts. For example, according to the American Conference of Governmental Industrial Hygienists, for ammonia the eight-hour threshold-limit (time-weighted average) is 25 ppm, compared with just 0.005 ppm for arsine and 0.3 ppm for phosphine.
However, this far higher ceiling is offset by the incredibly high consumption of ammonia gas for the growth of some devices. For LEDs, ammonia flow rates can hit 150 slm or more, exceeding that for arsine and phosphine by more than an order of magnitude. Due to these very high flow rates, dry bed scrubbing does not offer a viable solution for the abatement of GaN processes. Even when GaN devices are produced in research and development facilities, ammonia flow rates can be 30 slm. A weekly change-out of the absorber filling is required at this flow rate, rendering dry scrubbing commercially unfeasible.
Treatment of GaN exhaust gas is an important issue, due to the incredibly high volumes of GaN chips that are produced every year. The approaches that are taken appear to vary by region, and in Asia, large, sometimes self-built, wet scrubbers are prevalent. However, even in aqueous solution, ammonia can exert a high vapour pressure, so effective removal requires an acidic scrubbing media, such as sulphuric acid. With this approach, the scrubbing by-product is an ammonium sulphate sludge. Freed of metallic and other contaminants, this has the potential to be reclaimed and recycled to fertilizer.
In other parts of the world, a popular approach for abatement is to use heated scrubbers. This is energy-intensive, rather than maintenance-intensive, and can involve catalytic bed designs, which decompose ammonia into its constituent nitrogen and hydrogen elements. Here again, MOCVD facilities with a policy of at-source hydrogen removal tend to use burner/wet scrubbers. With this approach, a key challenge is to balance the combustion of hydrogen with the formation of secondary nitrogen oxide pollutants.
Figure 1: Typical flow ranges used in the MOCVD production of arsenide and phosphide based III-V devices. Right: Typical flow ranges used in the MOCVD production of GaN-based LEDs and power amplifiers.
III-Vs on silicon
Clear divisions between the compound semiconductor and silicon industries are blurring, with III-Vs seen as a potential saviour to maintain the march of Moore's law. Within the silicon industry, the design of equipment and facilities to support next-generation 450 mm wafers is already well underway. One of the most important objectives for the Facilities 450 mm Consortium is to ensure that energy savings, conservation of resources and emissions reduction are factored into facility design, right from the early planning stages.
InGaAs is tipped to enter the manufacturing roadmap for silicon devices at around the 7 nm node. The introduction of this III-V will bring with it environmental, health and safety challenges that have not escaped the attention of the larger silicon device makers and foundries. They are currently evaluating the implications of introducing these high mobility materials, and considering how to accommodate the growth of these films into their facilities.
Due to the delicate circuitry and the associated limited thermal budget of the silicon IC, the group V precursor that is most likely to be used in production is either tertiarybutylphosphine or tertiarybutylarsine. The good news, from a safety perspective, is that both of these precursors are far less toxic than their phosphine and arsine analogues. What's more, thanks to the nanometre dimensions of the transistor gates, flowrates will probably be far lower than those used for the growth of other devices by MOCVD.
Despite these comforts, MOCVD processing is still uncharted terrain for most silicon wafer fabs. Some working in that sector have anxieties related to the prevention of emissions to air and waterways. In particular, there are concerns about the safe containment of arsenic, both from a deposition processes perspective and as a by-product of III-V etching.
It is clear that the waste gas abatement of III-V processes covers an ever-widening spectrum of applications, each with its own challenges requiring individual, optimized solutions. Companies entering this industry should not be alarmed, though, but take reassurance from the well established III-V community: it already spans several industries, and has a remarkably high track record of safety and environmental compliance.
All these companies, whether high-volume chipmakers or small start-ups, can now draw on our full suite of trusted abatement products and our enlarged team of waste gas management experts. We are best prepared to support our customers and meet the challenges of the future.
