By Jeff Dorsch
The cutting edge in semiconductor manufacturing has meant not only big changes in IC design and process technology, but also in semiconductor materials. What follows are responses from Linde Electronics; Kate Wilson of Edwards Vacuum; David Thompson, Technology Director, Process Chemistries, Silicon Systems Group, Applied Materials; and Ed Shober, General Manager, Advanced Materials, Air Products and Chemicals.
1. What changes are being made in materials in fabrication of FinFETs, gate-all-around transistors, vertical NAND and fully-depleted silicon-on-insulator processes? Are there other new developments and trends in semiconductor materials (in interconnects, for example)?
Kate Wilson: We are seeing an increase in MOCVD precursors, low-temperature precursors and switching of process gases for ALD and multilayer films.
Linde Electronics: These devices will be implemented at aggressive nodes, e.g., 14/16 nm and below. Due to the unavailability of EUV lithography in HVM for at least the next 2—3 years, chip makers are forced to use multiple patterning techniques, which have led to several additional deposition and etch steps being incorporated relative to previous generations.
Of the devices mentioned, FinFETs and vertical NAND are becoming or will soon become mainstream in leading-edge logic and NAND fabs. Given that key parts of the device structures are approaching dimensions that are tens of atoms across, the tolerance for variability in the manufacturing process is significantly reduced, which in turn imposes special demands on materials suppliers to control variation in the quality of the electronic materials (EM) products supplied to a fab. This requires additional metrology and quality control techniques to be used across the EM supply chain, from incoming raw material to in-process material to finished product.
In addition, these advanced devices are more sensitive to any unspecified species in the EM products, and it is crucial to measure trace levels of unspecified impurities and understand their potential interactions with the thin films and interfaces involved in these devices. Collaboration across all the supply chain participants is thus key. Looking ahead, novel channel materials such as germanium and III-V compound semiconductors will be required, which bring their own set of challenges in deposition and etch.
David Thompson: For all these architectures the big trends are the introduction of new materials to enable scaling and the increasing criticality of interfacial materials. Consistently meeting new material innovation and interface engineering requirements is what we call precision materials engineering. Numerous new materials are being used today to enable low-power, high-performance foundry/logic devices. Selective epitaxy and metal gate films deliver >2 nodes of performance scaling with no litho-scaling. The introduction of CVD Cobalt liner and selective Cobalt cap layers in interconnect improves device reliability by 80x by completely encapsulating interconnect with Cobalt. More potential new materials are being evaluated at 10nm and beyond in transistors and interconnect. In 3D NAND, SiN film is used to store electrons using charge trap storage technology, compared to planar NAND where polycrystalline Si film is used using floating-gate MOSFET technology.
Interface engineering is also becoming increasingly critical. Whether it’s the transistor metal films, the interconnect cladding, or 3D NAND, there are essential enabling films that require sub-angstrom uniformity control across the wafer that have thicknesses between 10 and 40 angstroms – that’s less than the diameter of an atom. Additionally, while the bulk properties of materials are a useful roadmap, developing an understanding of how barrier, electrical, and other properties are impacted when the material is so thin that it doesn’t exhibit bulk properties is the challenge of the day. In many cases, a very particular pretreatment is required to enable the specific material interface to tune these properties. We’re finding that increasingly these steps need to be carried out sequentially within a vacuum environment with no air breaks.
Ed Shober: For logic devices and the adoption of 3D transistors, such as FinFETs, there are many needs for new chemical precursors to deposit — for example, silicon nitride and silicon oxide at temperatures far lower than previously required. In earlier nodes the thermal budget was in the 450 to 600C range, now the budget has been reduced to the 250 to 400C range, and the expectation is that it will go lower in the future. The shallow dopant profiles around the fin structure is one of the reasons for this drive to lower temperatures. Atomic layer deposition (ALD) is playing a greater role in depositing films. Thus, the chemistries employed must adsorb and react on the surface quickly and allow for deposition of highly conformal films over high aspect ratio features. More metal precursors are being employed in the FEOL for logic in order to tune the work function of the transistors. In the BEOL the biggest metal change has been the adoption of cobalt as a copper capping film and as a barrier liner film.
Needs for CMP are also increasing with 3D transistor structures. There are at least three new CMP steps that need to be performed to fabricate the fin. Finally, cleaning continues to play a major role in preparing the structures for the next deposition step.
Vertical NAND is moving memory off the lithography road map and onto a track that is driven by number of SiO/SiN films in a stack. The material needs here are again ALD-based precursors for lining and filling the channels etches into these film stacks. A major driver in cleans is products that limit particles left on the wafer and the scale of these particles must be in the nm size range.
For all devices, but especially DRAM and logic, the delays in adopting EUV are driving the need for self-aligned double, triple and quadruple patterning. These patterning strategies require new materials for forming the structures needed to reduce the pattern dimensions.
2. What changes are necessary in pumps and abatement?
Wilson: Increased variety of precursors requires flexible product operating range and tailored set-ups by process. Collaboration with semiconductor tool manufacturers, collaborative research organizations and end-customer development facilities is becoming more critical to ensure best known methods are applied.
Shober: Compatibility with the chemical precursors which can be highly reactive is one necessary requirement. Particle generation by the components is also a major concern and must be addressed by the suppliers to the same scales as discussed in the above answer around semi materials.
3. What are the risks involved with certain materials? Can they be disposed of safely? What about EPA regulation of these materials?
Wilson: Metal byproducts can be very toxic and containment of them can be challenging. Special care needs to be applied to their capture and disposal. Many of the new precursors are flammable and pyrophoric as well as being highly toxic so abatement is essential. Safe handling of new flammable or pyrophoric deposition precursors puts additional challenges on the process equipment and its maintenance – leaks of material out of, and air leaks into, process equipment can have serious consequences and therefore have to be diligently avoided to prevent accidents. It’s essential that sub-fab equipment designed to support advanced CVD processes should be designed from the outset with safe operation and servicing in mind. Better yet, an integrated sub-fab system design and a single point of ownership for the whole sub-fab system provide some assurance that the system can be operated with the minimum risk of accidents due to inadequate maintenance. Advanced integrated dry-pump/exhaust system/abatement/thermal management systems are available from at least one reputable equipment supplier to support such advanced processes, and have been widely adopted by several top-tier device manufacturers to provide exactly such assurance of maximized risk reduction in their advanced CVD processes.
Shober: With the drive to lower thermal budget there is a trend to using chemical precursors that are more reactive and stable. Thus, how they are produced, packaged, shipped and used by the customers are risks that we must address as new products are introduced. Shipping is especially a concern because there are more limitations on what materials can be flown from one point to another. This is driving for more localization of production/purification to shorten supply chains. BCP is of course another concern. Customers are looking for multiple and secure supply of materials.
4. How can semi materials be made “greener”?
Wilson: Once upon a time there was a concerted effort by a number of companies to find greener alternatives to persistent PFC gases used in etching processes to address the “green problem” at source – in general they weren’t very successful and attention switched to abating the PFCs effectively instead. In the current environment, where increasingly exotic CVD precursors are being introduced into advanced device node manufacturing, it’s also likely that “greenness” will be more a result of diligent treatment of the waste precursors, their decomposition products and the solid residues left behind in the process equipment than efforts to make the materials themselves “green.” That puts the onus on abatement and waste treatment system manufacturers to develop suitable products to meet the emerging challenges, and the end-user community to accept responsibility for installing suitable waste treatment facilities.
In some instances, careful consideration has to be given to the balance between risk of gas release and cost to the environment of treating it – abatement of nitrous oxide (N2O) being a case in point. N2O is a greenhouse gas widely used in oxide CVD processes, and device manufacturers would prefer to abate it to reduce their GHG emissions. However, combustion of N2O consumes natural gas, generates carbon dioxide (CO2) and under adverse conditions can generate significant quantities of nitrogen oxides (NOx); so the question arises – which is the least bad situation?
Linde Electronics: Materials suppliers can contribute to greening of semi materials by:
- Limiting emissions and waste over product life cycle
This includes material production, delivery and return/reclamation/disposal.
Sometimes, despite the material selection constraints, direct, process-compatible substitutions can be made such as F2 (fluorine). See reference to this in the article “Material Support: Helping Displays Deliver Higher Performance” in the September, 2012 issue of Solid State Technology.
- Packaging and processing for efficient use
Often headspaces are exhausted and heels are unused to prevent light and heavy contaminants. Better purification, quality control, packaging, and material property knowledge can reduce the amount of material lost to safeguarding quality.
- Recovering material from waste streams
Examples of this are He (Helium), Ar (Argon), Xe (Xenon), H2SO4 (Sulphuric Acid). See reference to this in the SemiMD.com blog post “Sustainability through Materials Recovery” at http://semimd.com/materials-matters/2014/07/21/sustainability-through-materials-recover.
Thompson: In many respects, semi materials are the greenest materials known and need to be taken in the context of not just as the materials in the chip but in what they end up consuming, for instance, power. A good example is the UNIVAC that ran at 1,905 floating point operations-per-second on 125 kW, while processors like today’s Tegra K1 run at 326 GFLOPS on 10 W – which uses a number of different materials that in effect reduce the power required per floating point operation by a factor of 1 trillion. Net reduced environmental impact will almost always favor choices that continue the power scaling trajectory. That being said, we need to be vigilant in managing the dangers and impact of some of the more hazardous materials that are being used which provide this net green benefit.
Shober: The industry made a major step to become greener when NF3 was adopted rather than fluorocarbons for cleaning CVD reactors. Solvent recovery and re-purposing for other use in either the fab or other industries is now being adopted. More and more cleaning processes are employing water-based formulations which reduces solvent usage at the fab. In the future IDMs will be looking for more ways to recapture and purify, recycle and/or re-purpose wastes into other uses at the site or outside into other industries.
5. How are advanced processes, such as atomic-level deposition, affecting materials use?
Wilson: Diverting precursor and low utilization rates in the process cause higher unreacted material and waste.
Linde Electronics: Both atomic layer deposition (ALD) and atomic layer etch (ALEt) are key new processes required in leading-edge device manufacture because of the new elements being incorporated and the aggressive geometries being adopted to keep Moore’s Law on track. Several new EM products are now used in ALD, e.g. organometallic molecules. For ALEt, quite a few traditionally used EM products (e.g. chlorinated gases) are currently being evaluated. The selection of process materials becomes very challenging when multiple films are in close proximity, e.g. requiring high selectivity for etching one thin film without affecting two or more nearby materials.
Thompson: What we find is that the deposition technique or specific chemistries employed strongly impact material properties. There are almost no situations where when we migrate from one technology to another – say PVD to ALD – where we don’t see significant change in materials properties associated with the technique.
Additionally, in many cases the materials that the industry is accustomed to using are no longer available for a new technique. Usually, there’s a ripple effect on retuning other materials or processes to enable the new material. It’s an exciting time – there’s a renaissance of metallurgy in both the front end and back end.
Shober: ALD does have a tendency to reduce material consumption, but not to the degree one may expect as a result of the process itself. The biggest impact ALD is having is on chemical costs. Materials capable of being deposited by ALD are oftentimes novel and there only use is within the semi industry. Thus, the cost on a gram basis can be much higher than what the industry has come to expect from use of materials like TEOS.