Building Specialized Silicon at Industrial Volume

Danny Kohanek, SVP of Manufacturing, SUMCO

Over my forty-one-year career in making silicon wafers, I have certainly seen a lot of changes, even in front-end operations that most people do not even know exist. When I started in 1983, two-inch, three-inch and four-inch silicon wafers were already in full-blown production. At that time, several different dopants were already being used to change the electrical characteristics of wafers. Types of dopants used included boron, the most heavily used, along with some red phosphorus and antimony.

It was not until the early 1990s that six-inch (150 mm) wafers became the industry standard. Many different companies made large investments to expand front-end operations in order to increase silicon supply during the home personal computer boom.

In the 2000s, wafer diameter technology advanced to eight-inch (200 mm) wafers. This introduced an entirely new level of quality expectations for silicon parameters that had not previously been considered critical. Thickness, dopant type and resistivity, now collectively referred to as substrate characteristics, began to determine the operating conditions required for building integrated circuits on the silicon wafer.

During this period, the industry also saw several advances in the use of integrated circuits across new categories, including CMOS, MEMS, MOSFETs, diodes and rectifiers. This progressed into large-scale integration (LSI), very large-scale integration (VLSI) and ultra-large-scale integration (ULSI). Once again, wafer diameter changed in the 2010s to 300 mm, which became the latest platform for high-volume products such as CPUs and memory chips. Even today, 200 mm and some 150 mm wafers are still being produced and sold to customers.

U.S. operations specialize in extremely low-resistivity products, including arsenic-doped wafers with resistivity as low as 0.00159 ohms and red phosphorus-doped wafers down to 0.0009 ohm-cm. 

Reaching this point over the years required many distinct groups with different disciplines and functions, beginning with a manufacturing operation capable of producing consistent products in high volume.

The U.S. operation is the primary wafering plant in the country and produces about 8,000 silicon wafers per day. It supplies a dozen different customers with more than 250 specialized product types, which continue to change as customer products evolve. To manage this complexity, one of the key team meetings is Technology, Production and Sales (TPS). This team includes application engineers, manufacturing managers and the sales team. Together, they review customer trends in applications and wafer requirements, analyze potential sales and determine what is required to manufacture these products.

"The U.S. operation is the primary wafering plant in the country and produces about 8,000 silicon wafers per day."

This evaluation covers every step, from growing mono-silicon ingots using the Czochralski (CZ) method to wafering operations, optional chemical vapor deposition (CVD) processes and optional epitaxial layers. The epitaxial operation adds additional layers with different dopant types, thicknesses and resistivity values. As a result, the 250 distinct wafer types can be further customized into another 200 to 300 variations to meet very specific product demands.

Another critical group is the process engineering team, which is responsible for setting process conditions and controls to ensure products are consistently produced within 1.5 to 3.0 sigma of customer specifications.This is achieved by establishing control parameters and testing methodologies across all process steps, supported by statistical process control (SPC).

The production control group also plays a major role. They receive customer sales orders up to three months in advance, based on product specifications and anticipated changes in demand. These specifications include dopant type, resistivity ranges, oxygen ranges and CVD requirements for optional layers. Based on product mix and volume, production control schedules mono-silicon ingots to be grown.

Each ingot grown is approximately 1,500 to 2,000 mm long, with each millimeter yielding one silicon wafer. The ingot’s intrinsic silicon characteristics change along its length, with resistivity and oxygen profiles varying from top to bottom. To maximize CZ yield, production control works closely with CZ engineers to optimize yield based on anticipated sales volume and product mix for the next three months.

As ingots grow, wafers may be produced that have no immediate sales opportunity. Unfortunately, resistivity and oxygen profiles naturally change during growth, depending on the dopant type and target resistivity ranges. All manufacturing plants operate 24/7/365, which requires a comprehensive preventive maintenance program to keep tools operational at least 80 percent of the time. Quality engineering teams oversee testing methodologies to ensure all metrology tools remain calibrated and controlled, enabling wafers to meet or exceed customer requirements.

Equally vital is employee training. New employees are trained to ensure their work performance meets manufacturing demands. This includes learning to follow product-specific requirements across all tool groups in wafering and epitaxial operations, as well as implementing required changes to tool recipes to meet specific product requirements.

Throughout manufacturing, product quality checks are performed and SPC is used to control tools and process conditions, ensuring wafers are produced with minimal scrap loss. From start to finish, the operation sells, on average, about 70 percent of the polysilicon purchased as customer-specific silicon wafers.