Part One: The Issue with 450mm Ingots
In the last article, I mentioned that there appears to be a route toward a cost-effective (upgradable), reduced cost (faster), and safer means of pulling 450mm silicon crystals (also referred to as ingots or boules). This technology may be a potential game changer for the industry.
Background
First some background on the key issues and roadblocks toward developing a steady supply of 450mm prime, test wafers. The overall industry standard batch 300mm silicon boule pulling process leads us to look at 450mm and why things haven’t been as reasonably scalable as previous wafer size changes. Experts, bear with my pedantic discourse, you’ll see the clear logic in this if you make it through to the end. For a full description, please refer to a textbook such as Handbook of Crystal Growth[1].
Assuming a supply of electronics-grade polysilicon is already available, in both large and small chunks, as well as granular form[2], the first major step in the process is to fill a large crucible with polysilicon for later melting in a furnace. The manual loading process is more an art than science, as Figure 1 demonstrates. A number of patents[3][4] have been issued on the loading process, each describing a more efficient means of packing.
Proper loading ensures only minimal voids and helps safeguard that as the polysilicon melts, pieces don’t get stuck on the side of the crucible. Pieces that “stick” up high and later fall into the molten silicon can splash, which is quite dangerous. It is also important to position the silicon pieces in such a way that they do not entrap hydrogen, which is outgassed as the melting proceeds. Side-scrape of the crucible as the chunks melt and fall must also be prevented. Since melting of silicon occurs at over 1410 degrees Celsius (with the crucible, necessarily hotter than this), any scraping at this temperature not only shortens the life of the quartz crucible, but sloughs off impurities into the molten silicon, contaminating the final product to some level.
With even with the very best polysilicon loading techniques, there’s only an 80-90% packing efficiency, and voids (air pockets) exist, so the molten silicon level is less than the initial polysilicon load level. An important characteristic to remember is that with a batch process, you can only use what is loaded into the crucible prior to melting. This initial load is also known as the “charge” of silicon. Additionally, the crystal pulling process cannot begin until the entire crucible of silicon has been melted. This can take 10 to 12 hours or more for a 300mm charge of 450kg of silicon.
History
Back in the late 1990’s, a consortium in Japan was exploring a transition from 200mm to 400mm wafers (rather than to 300mm wafers). This group, the Super Silicon Crystal Research Institute[5] or SSi, performed ground-laying work that has helped set the stage for later exploration in 450mm. They were successful in pulling 400mm crystals of up to 1.1 meter in length. One of their limitations was the size of crucible. Their experiments were completed with 36 inch (~90cm) and 40 inch (102cm) crucibles. Any larger crucible was deemed too great a safety risk. Fortunately, their results were well published. A number of challenges were overcome while some remain, proving to be significant for 450mm crystals.
450mm Crystals using Batch CZ
On initial consideration, it might make sense to try to extend 300mm CZ crystal pulling size and scale to 450mm. For comparison, using the historical 3x rule for crucible size, a 450mm wafer crucible would be 135cm (over 5 foot) in diameter. This is a really large crucible. In fact, one this size has not actually been produced for silicon wafer test or production. This size comparison is shown in Figure 3.
Without going into details, the fact is that the industry is not manufacturing crucibles this big that can survive at 1450+ degrees Celsius for the number of hours, well, literally, days required to melt a 980kg charge and then pull a 2m ingot. The crucible would possibly fail, which would lead to a disastrous safety incident. Imagine about 2200 pounds of the equivalent of molten lava spilling out of a container in your workroom. It is no wonder SSi was limited in the size of the crucible they were willing and able to test.
Publication by MEMC[6] of some findings in making 450mm ingots and wafers show that there appear to be no plans for crucibles larger than 42”, and that the crystal diameter to crucible diameter ratio will be significantly less than historical trends, as shown in Figure 4.
In fact, their chart is somewhat misleading in that it’s ordinate axis only goes to 44”, when, actually the historical 3x ratio would require a 54” inch crucible for 18” (450mm) wafers. For this publication, 36”, 40”, and 42” crucibles were counted as contenders for 450mm batch systems. Commercially, I was only able to find 40” crucibles, from Shin-Etsu[7]. I believe this is the crucible shown in Figure 5, which is an image from their catalog, though only 36” is actually listed for purchase.
Status
The key message is that with a smaller crucible, in a batch system, there will be a smaller (shorter) ingot for 450mm than historical transitions. It’s not possible to load enough polysilicon in this crucible to do otherwise. MEMC’s chart is consistent with the SSi’s results making it safe to say that we can expect ingots of only half the length of the historical 2m long ingots, given the initial charge allowed. On the bright side, a 36” crucible will fit into an existing PVA TePla EKG-3000 crystal puller. Its 450kg maximum load will at least allow ~1m length crystals to be pulled. However, if only Shin-Etsu is making 36” production crucibles and only PVA TePla is manufacturing 450mm capable pullers, it seems that an industry-defacto standard may be evolving. Only one supplier for each critical piece in the supply chain is less than ideal, but the risk vs. reward seem to be leading in this direction.
Next. Part Two: CZ Batch, Pros and Cons
References
[1] Handbook of Crystal Growth.
[2] Blending FBR Granular Polysilicon and Siemens Chunk.
[3] Method of loading a charge of polysilicon into a crucible, US 20140060422 A1
[4] Polysilicon system US 20120260845 A1
[5] Large diameter silicon technology and epitaxy H. Yamagishi, et. al. Microelectonic Engineering 45 (1999).
[6] Growth of 450mm diameter semiconductor grade silicon crystals, Zheng Lu and Steven Kimbel, Journal of Crystal Growth 318 (2011), pps 193-195.
[7] Semitransparent Quartz Glass Crucible for Silicon Crystal Pulling Applications
Lead story image:
https://upload.wikimedia.org/wikipedia/ru/7/72/Chohralsky_Silicon_Crystal_Growth.jpg
