450mm Back from the Dead? The Ultimate Wafer may Help
Solutions from an Unlikely Place
In the last article, the evolution of the crystal pulling methods was analyzed, and there presented was the promise of a new method that would continue the improvement over the continuous CZ, fed with dry pelletized silicon, with a shallow crucible (called “Generation Three”). While Generation Three provides improvements over batch and continuous CZ there are still issues one has to face with the method. Namely, there is a natural “fight” between the high temperature required to melt the silicon in the outer double-walled crucible, and the lower temperature required at the same surface level to cause crystallization. As such, the temperature gradient in not optimal, resulting in a compromise in the maximum pull rate one can achieve and still maintain good defect density in the ingot.
Fortunately, during the cold war between the USSR and the United States, while commercial Western allies focused on advancing the batch CZ method for ever-larger silicon wafers, the Soviet Union was quite interested in developing methods to grow large, single-crystal, scintillating material crystals. Effectively, with the iron curtain between the societies, parallel, yet different approaches were developed in improving the CZ method. A methodology, developed in the Ukraine prior to the fall of the wall, had been used for growing very large, ultra-pure crystals. The intent of these crystals was to detect nuclear submarines. Scintillating materials, such as Sodium Iodide, doped with Thallium (an alkali metal halide), emit photons when struck with gamma radiation. With a slight possibility that gamma radiation might escape from U.S. nuclear submarines, large, ultra-pure, transparent, scintillating crystals were developed in order maximize their detection.
ROST Method
This method developed and perfected, also known as the ROST method, is an improvement to the shallow-crucible, continuous CZ method. By vertically separating the melt and recrystallization regions of the dual-walled crucible, separate heating elements for each region allowed better optimization of the melt and recrystallization regions of the crucible. This method[1] has been used to grow crystals grown greater than 550mm in diameter, weighing over 500kg (see Figure 1).
Though the expertise was developed at the Ukrainian Institute for Single Crystals, after the cold war, through a series for fortuitous events, the process was licensed and commercialized for medical use by Siemens[2]. The quality of the crystals produced was so superior that much of the Ukrainian team was moved to the U.S. to work with Siemens.For purposes of these articles, this method is referred to as “Generation Four,” and is shown in Figure 2. The advantages of the ROST method over batch is multifold: faster loading and setup times due to minimal initial charge of material required, greatly reduced or eliminated scoring and damage to the crucible wall vs. batch CZ, tight doping control since mixed as the process proceeds, no need for a magnetic field to minimize turbulence, and optimization of temperature gradients resulting in a defect-free maximum crystal pulling rate.
Generation Five
In order to overcome these last challenges, the scientists have come up with a further development, which is to provide an external silicon (and dopant) melter that provides a liquid feed through the wall of the furnace, into the crucible (Figure 3). In this article, this is being called “Generation Five.” A double-walled crucible, similar to the ROST, Generation Four, method is used, however, with liquid being fed rather than granules, the outer region can be made even thinner. With the reduction in crucible size afforded, a 450mm boule can be pulled from a retrofitted puller previously sized for 300mm boules, and possibly inside a puller sized for 200mm boules, though the boule weight would be an issue.
With only a few inches of molten silicon required in the crucible before starting the pull, one can save over 24 hours in time versus the batch CZ method’s of loading and then melting a large crucible full of silicon, at much less power consumption. Staff at 450mm.com has calculated that the overall liquid feed ROST process time saved (over batch CZ) to be about 20%. This improvement would similarly apply to 300mm boules. One of the largest, upfront cost savings comes from the ability to retrofit an existing 300mm puller to be used for 450mm boules. The differences between the generations of CZ pulling capability are shown in Figure 4.
Certainly, the transition to 450mm has stalled; with economic factors being one of the primary obstacles. The cost and availability of 450mm test and production wafers, along with the associated development of batch CZ 450mm furnaces and purchase of production lines is a significant driver in the cost equation. Liquid-feed ROST can help balance this equation. When one factors in all of the benefits of moving to this Generation Five method of wafer production, including: reduced cost, increased safety, reduced energy consumption, and quality (defects and doping consistency) of the final product, it would make sense to seriously considering pursuing this revolutionary approach to 450mm, and for that matter, 300mm wafer production.
References
1. Eidelman L.G., Goriletsky V.I., Protsenko V.G. et.al. Automated Pulling from the Melt – an effective method for Growing Large Alkali Halide Single Crystal for Optical and Scintillation Applications. J.Crystal Growth. -1993. -Vol. 128. -P. 1059-1061.
2. Van, Jon. Crystal Technology: Medical Gem. Once A Nuclear Detector, Research Now Detects Illness. Chicago Tribune. March 21, 1995.
3. Holder, John, et. al. MEMC. US Patent Number 6284040 B1. Process of stacking and melting polycrystalline silicon for high quality single crystal production. Granted September 4, 2011.