Sizes of silicon wafers for chip manufacturing

Some readers probably know that the transition to a new standard promises a significant reduction in the cost of chips. With this advantage of using larger plates, the reasons for delaying the transition to the new standard do not seem obvious. It may even seem that all this is nothing more than a global conspiracy of corporations, striving to stop progress for super-profits, while all that is needed is to buy new machines. “It’s so simple!” The lay man will think.

I propose to remove the cap from the foil and understand the problems that arise when scaling, and the technical details of the production of silicon wafers. This material lists only a few of them.

Investment and return on investment

To begin with, let's figure out why all this gimp with an increase in the size of the plates? The fact is that with an increase in the plate, the cost of each chip decreases. Calculations show that the transition from the current 300 mm to 450 mm will ultimately save about 30%. About the same as what happened during the transition from 200 to 300 mm plates.

A 30% reduction in the chip production price is really significant. But with a note that this will not happen right away, since initially the cost of new plates will be much higher. Sq cost cm 450 mm of the substrate is equal to the current price per square meter. cm 300 mm of the substrate not earlier than 2025. And at the beginning of use it will be 4-5 times more expensive.



General transition trends are as follows:

• 30-100% increase in cost
  production facilities, tech. maintenance and equipment for
  automation.
• 20-50% increase in value
  lithographic equipment.
• 10-30% reduction in throughput
  abilities of projecting equipment. For the rest of the equipment
  - without changes.
• 15% reduction in labor costs.
• 70% increase in reagent costs.

Investment costs are estimated to reach $ 60 billion, and therefore, the largest producers will have to join forces to move to a new standard. The return on such a large investment is about 8 years. At least the R&D of the transition from 200 to 300 mm plates paid off in approximately this period.



Well, where is the savings? The answer is not so simple.

The area of ​​the silicon substrate increases by a factor of 2.25. In addition, you can place even more chips, since rectangles fit into a circle with a large radius with greater efficiency. In simple terms, if you draw a circle with squares, the circle will be “rounder" the larger it is or the smaller the squares.

Take, for example, the Tahiti XT core with an area of ​​352 mm2, which is the basis for HD 7970 video cards. Provided that the chips are square, about 160 chips can be placed on a 300 mm substrate, and 386 on a 450 mm substrate (see the figure below). Which is 2.41 times more. Of course, the increase will vary depending on the size of the chip.



It follows that a factory with a production capacity of 40-45 thousand wafers is capable of producing as many chips as a 300 mm factory with a capacity of 100 thousand wafers. And just this moment allows you to save! A 450 mm factory for 40 thousand plates will cost 25% cheaper than a 300 mm factory for 100 thousand plates. The savings are largely due to lithographic equipment, the cost of which will not increase commensurate with capacity, as noted above.

But there is a nuance. With the other factors listed increasing the cost of production, the current cost of lithographic equipment would not compensate for the costs if it remained unchanged. Fortunately, the cost of equipment for each more advanced tech. the process is growing, and with it the benefits of switching to a larger plate diameter.

Technical difficulties

Crystal weight

The weight of the grown crystal increases 3 times, up to one ton. There is a problem with the vertical suspension of the crystal by a narrow neck, a thickness of several mm, during the growing process. The seed crystal alone cannot support the weight of 450 mm of the crystal. An additional way to maintain the crystal is needed.

Growing time

Any external vibration (such as an earthquake) can cause a change in the spatial lattice of the crystal. If this happens, the crystal has to be completely remelted and growing starts from the very beginning. This takes a considerable amount of time (up to a month) and thereby increases the cost of production.

So the consequences of the earthquake in Japan in 2011 reduced the global production of silicon single crystals by 24.5%. In fact, the grown crystals turned into garbage. It took more than a month to restart production.

Quartz crucibles

Quartz crucibles for melting polycrystalline silicon should be larger and should withstand twice the time required for production than used to grow 300 mm crystals.
If crucibles of 81.3 cm in size are used to grow 300 mm crystals, then crucibles up to 111 cm in diameter are required for 450 mm crystals. To withstand such a long growing time, the quality of the crucibles must also be improved.

Cooling

Cooling time can increase up to 4 times.

The thermal past, during the production of the crystal, affects the number, size and distribution of defects in the form of impurities. Unlike offsets, such defects are usually present. And although they do not necessarily lead to rejection of the crystal, they can affect the performance of the final chips. Therefore, it is necessary to precisely control the process in order to minimize their appearance. Due to the enormous size of 450 mm crystal, it will be cooled more slowly, which exposes it to a greater temperature difference and affects the appearance of defects. This requires an innovative cooling process.

Conclusion

Summing up, the transition to 450 mm plates is the need to compensate for the ever-increasing cost of lithographic equipment. Boarding this train is very expensive, but not boarding at all means putting yourself at a disadvantage in relation to competitors in the long run. In addition, it is important to know the best moment for landing.

So manufacturers can be easily understood, especially considering the possible risks that may arise over such a long period of time. It can be everything, starting from global economic problems, ending with the fact that more promising technologies may appear even before the investment pays off.

By the way, if you haven’t understood yet, for us consumers, the transition will not be reflected in the reduction in the cost of processors, GPUs and memory. At least not in this decade.

Sources:
www.sumcosi.com/english/products/next_generation/problem.html
A Simulation Study of 450mm Wafer Fabrication Costs
Economic Analysis of 450mm Wafer Migration
Reaping the Benefits of the 450mm Transition
Indistinguishable From Magic: Manufacturing Modern Computer Chips
www.silicon-edge.co.uk/j/index.php?option=com_content&view=article&id=68