The Next-Generation Wafer

Manufacturing single crystal silicon ingots via the Czochralski (Cz) method first involves forming crystals into a neck several millimeters in diameter. This neck is created by dipping a seed crystal in a crucible containing molten silicon. The neck is then suspended, rotated and grown in order to create a dislocation-free single crystal ingot (or simply “crystal” or “ingot” for the purposes of this article). Manufactured ingots consist of conical top and tail sections and a cylindrical body section in between. Dislocations are undesirable linear crystal defects that limit the portion of ingot usable for wafer production.

The top, tail and peripheral body sections and residual raw materials inside the crucible are effectively material losses as they are not processed into wafers. SUMCO has examined the technical issues relating to manufacturing 450mm single crystal ingots based on the assumption that material loss rates would be comparable to those for 300mm ingots (i.e. proportional).

Crystal weight: The total weight is expected to be somewhere in the region of one ton, approximately three times the weight of a 300mm crystal.; Whereas 300mm crystals are suspended and grown using only a seed crystal, strength limitations will not allow this for production of 450mm crystals. It will therefore be necessary to develop a new means of suspending the neck.

Growth times: 450mm crystals are expected to take roughly twice as long to grow as 300mm crystals.; This extended growth time increases the risk of dislocations in the crystal (resulting from external forces such as an earthquake).
If dislocations occur, the growing process has to be terminated and the crystal gradually melted down and re-grown from scratch. This takes a considerable length of time and also increases manufacturing costs.

Quartz crucibles: Crucibles need to be larger than those used to grow 300mm crystals and have to be able to withstand manufacturing times that are roughly twice as long.; Whereas 32 inch (81.3cm) diameter crucibles are standard for 300mm crystals, it is thought that 450mm crystals would require crucibles with a diameter of 40-44 inches (101.6-111.8cm). In order to withstand such long growth times, the quality of crucibles will also need to be improved.

Thermal history: Cooling times for 450mm crystals are expected to be two to four times longer than for 300mm crystals.; Thermal history during the crystal growing process determines the distribution, size and density of so-called grown-in crystal defects, including crystal-originated particles (COP), oxygen-induced stacking faults (OSF) and oxygen precipitates. Unlike dislocations, ingots typically possess a certain amount of these defects. While their presence does not necessarily preclude use for wafer production, functionality of the product can be affected. Therefore, the formation of these defects must be controlled. Due to the massive size of ingots, 450mm crystals will cool at a slower rate, increasing the length of time during which crystals are susceptible to temperature ranges that can cause defects to form. As a result, there are concerns that crystal defects could increase in size and density or unexpected defects could form. The 450mm process will therefore require innovative cooling technology.

Essentially, manufacturing 450mm crystals will require innovation in terms of both manufacturing equipment and manufacturing processes.

300mm wafers came into commercial use in 2001. Double-sided polishing (DSP) technology was introduced and applied in order to precisely control the flatness of both sides of the wafer. Having undergone various modifications since then, the DSP process has helped to substantially reduce device line widths and is now capable of handling 45nm generation devices. It is thought that merely scaling up the existing DSP process in order to shape 450mm wafers however would have a serious impact on flatness. It is estimated that the level of flatness required for wafers 1.5 times larger in diameter would necessitate a return to device line width processes from two generations ago. Moreover, as device line widths are expected to continue to be reduced in the future (from 32nm to 22nm to 16nm), some form of groundbreaking innovation in wafer shaping technology will be needed in order to handle generations from 22nm onwards at 450mm.

If you were to compare a 300mm wafer to the site of the Tokyo Dome, the level of flatness is so precise that the height difference for the entire site would be a mere 0.1mm.

As the area would increase by 2.25 times at 450mm, a more precise level of flatness will undoubtedly be required.

As it stands, there is no established groundbreaking technology capable of handling 450mm wafers. Further research and development will therefore be essential.