3D-NAND flash memory devices (hereafter, called 3D-NANDs) are at the centre of still competition in creating more layers, according to The Gas Review.
This is having a large influence on the specialty gas market, including the application of new gases, such as tungsten hexachloride (WCI6) and sulphur dioxide (SO2).
Increased competition in increasing the number of layers including Samsung at 64, SK Hynix at 72 and Toshiba at 96 layers
Samsung Electronics, at their Pyeongtaek semiconductor plant (line 18), started full-scale production of 64-layer 3D NANDs on 4th July. This plant was constructed with a mammoth capital investment of YEN 1.6 trillon. Plans are in place to expand production there from 250,000 wafer sheets per month at the end of 2016 to 660,000 per month by the end of 2017.
SK Hynix announced the development of the world’s first 72-layer 3D-NAND on 1st May. Mass-production is slated to start at the Icheon Plant, M14, in the last half of 2017. SK Hynix is also in the process of building a new state-of-the-art plant in Chungju targeted at completion in June 2019.
Toshiba announced on 28th June that they have made a prototype 3D-NAND with a 96-layer process and have confirmed basic operation. Scheduling places sample shipping in the last half of 2017 and the start of mass-production shipping at the Yokkaichi Y6 building in 2018. Although 48-layer versions are currently at the centre of 3D-NANDs, it appears that 64-, 72- and 96-layer versions will soon take over centre stage.
It is said that for each additional layer, four or five more processes are required, and it is estimated that 64-layer versions require around 300 processes, 72-layer versions around 350, and 96-layer versions around 450. While the number of layers is being increased, design changes are simultaneously in process to form thinner films and downsize the devices. The amounts of gases used increase as the number of layers increases, but thinner films means that the increase is not linear.
Increase in hydrocarbon gases
Manufacturing processes appear to differ among companies but the type of gases that they used are basically the same. A new process that came with 3D devices is hard mask formation with PECVD in lithography process. Propylene or acetylene is used to protect the wiring film etched on the substrate by using propylene or acetylene to fabricate amorphous carbon. A feature of this process is that multi-pattern etching and PECVD are used together as a set. It appears that diborane, which has been used for 40 years, is also used to increase the resistance property of the amorphous carbon.
As the hard mask carbon material, Lam Research uses high-purity acetylene while Applied Materials and Tokyo Electron used propylene in gradually becoming the main material. It seems it is selected more for safety and ease of use than for quality.
This process has been introduced not only to 3D-NANDs, but also to FinFETs and DRAMs. It is used by Micron Technology and Taiwan Semiconductor Manufacturing (TSMC) as well, among other companies, and the estimated global demand is several hundred tonnes per year.
Chlorine over fluorine?
A new specialty gas that is receiving attention is WCI6. Lam Research, a US semiconductor production equipment manufacturing giant, announced the world’s first low-fluorine tungsten ALD process ALTUS last August.
Chief Operating Officer Tim Archer of Lam Research said, “As the number of layers in 3D-NAND memory cells was increased, two latent problems in embedding tungsten in word-lines surfaced. One of them is that fluorine spreads to the insulating film from the tungsten film causes physical defects. The second is for devices that exceed 48 layers, layer stress causes excessive wafer bowing, degrading yield and device electrical characteristics, leading to reduced reliability. To solve these problems, we developed the low-fluorine tungsten ALD technology.”
This announcement does not specify any of the gases that were used, but after the announcement, the gas supplier was actively asked, “Is WCI6 available?” Although being on circumstantial evidence, we assume that in part of the tungsten process, a switch was made from using the previous tungsten hexafluoride (WF6) to using WCI6.
Also, in the “Tungsten film forming method,” the patent application of which was published on 4th June 2015, Tokyo Electron said “as miniaturisation of a design rule advances, the devices may be adversely affected by fluorine in the case of using a source material containing fluorine”and that tungsten chloride gas and a reducing gas (eg hydrogen) were used to form a tungsten film.
Specific details are not given, but the “adverse” influence of the fluorine component on the device seems to have been a previously known problem. However, although there is the impression that the device would also not be very compatible with chlorine, it might be that the adverse influence of chlorine on 3D devices is less than that of fluorine.
It is however inconvenient that WCI6 is a solid source. According to the MSDS, WCI6 is a dark purple solid with a vapor pressure of 43.11 at 215°C. For application in ALD, WCI6 must ultimately be vaporised, and advanced technology is required to stably supply minute amounts of a solid material. If this problem can be solved, the application in mass-production processing would become possible. That said, it is undoubtable that WCI6 would be used only in some processes and that WF6 would remain the main material.
Toward mass-production of 3D transitors
On the other hand, shifting to 3D transistors is making progress. Although the planer type is traditional, the FinFET (or Tri-gate) technology that was used for microprocessors and CMOS’s is being introduced to achieve miniaturization to 16nm or less. It is being introduced by TSMC, Intel, Qualcom, Samsung, Micron and other manufacturers.
This technology suppresses leak current by exploring structures in which gates surround the channels. It is thought that three or four additional processes are required over previous production, but multi-etching is being introduced and hydrocarbon materials such as propylene are used to cover the substrate while using PECVD to form amorphous carbon.
Otherwise, products that use organic hafnium, a high-k material, have been around for 10 years, and they are now coming into their own and are attracting great interest.
For etching, there are many assist gases that could be used. To produce 3D devices, deeper etching and more selectivity are required. Although previously carbon monoxide (CO), carbonyl sulfide (COS) and other gases were used, specialty gases that have not been used much in the past, such as Sulphur dioxide (SO2), are being used. To etch an intricate multi-layer structure, the current trend is to study a wide range of specialty gases.
However, as much as possible, the assumption is that currently existing gases will be used. Synthesizing or chemically combining new materials would require greater expense and both safety and environmental characteristics would have to be checked, resulting in too much time to product release. If materials that are already stipulated in the High Pressure Gas Safety Act or Industrial Safety and Health Act are used, experiments can be conducted much more easily as long as material purity is achieved. There are also greater chances of achieving mass-production.
It is thought that achieving ultra-multi-layered 3D-NANDs would lead the way to future 3D processing of DRAMs and logic, and that would definitely lead to greater expansion in the global market for specialty gases.
The Gas Review, issue no. 442