Pyrometers Improve Quality and Yield in Silicon Growth

Radiation pyrometers are particularly well-suited to overcome these challenges for several reasons. First, they provide non-contact measurement, which is important because many processes (such as deposition processes) do not allow contact instrumentation, as it can contaminate and heat-sink the product during process. They also eliminate junction deterioration problems typically seen with thermocouples, and they can achieve better accuracy in higher-temperature ranges. Following are a few examples of how the use of pyrometers can allow equipment designers to improve equipment performance and process engineers to improve quality and yield.

Polysilicon Growth

The Siemens process is the traditional way of converting purified trichlorosilane to polysilicon. The cold-wall reactor is usually water-cooled in this process. Seed rods inside the reactor are heated to process temperature, and chemical vapor deposition (CVD) takes place on the surfaces of the rods. The rods grow as the deposition accumulates, and this growth continues throughout the process until the rods reach the desired diameter. The surface temperature of the growing rods is of the most interest in this process as the deposition occurs on the surface and temperature control is important in CVD processes. Figure 1 shows CVD as part of the Siemens Process [1].

Figure 1. Time-temperature cycle for multi-crystalline silicon growth.

Conventionally, two-color (ratio) radiation pyrometers are used for this application. The optical access for the pyrometer to the rods is through a viewport (window). During the process, the window can get contaminated, which could result in errors in temperature measurement. By taking the ratio of the measurement at two wavelengths, the changing transmission of the window can be canceled out. The measurement therefore can be made reasonably immune to window deposition. This is based on the assumption that the spectral transmission of the coated window is identical at the instrumentation wavelengths or at least with a constant ratio between them at all temperatures. In reality, neither can be completely true. In practice, by carefully choosing the pyrometer wavelengths, the assumption can be satisfied to an acceptable degree. With proper emissivity slope setup, the pyrometer would serve the purpose.

Some windows can have a non-flat spectral transmission curve. In those cases, the user should check with the pyrometer manufacturer to ensure the factory calibration of the two-color pyrometer includes the window spectral transmission.

While two-color pyrometer helps correct for window coating, it will not work with too-low window transmission. As the window contamination continues, eventually, the transmitted infrared radiation can become too low. The time it takes for the transmission to reach the threshold depends on the reactor design. When the threshold is reached, the pyrometer will stop working or won’t function properly, which would then require window replacement/cleaning. Therefore, it is important to check the window transmission before each run to avoid unnecessary interruption during the growth.

It is worth noting that a two-color pyrometer is not necessarily the only solution. There are other ways to reduce growth-induced coating on the window. If a window is kept clean, single wavelength pyrometers can be used as well. A single wavelength pyrometer normally can measure lower temperatures than a ratio pyrometer. Emissivity slope adjustment normally required in ratio pyrometers is not needed with single wavelength pyrometers. With a clean or near clean window, a single color pyrometer may achieve a better accuracy/repeatability.

After the growth, the rods are allowed to cool down. Below roughly 700°C, silicon becomes transparent in a fairly wide spectral range in infrared. That makes it hard for most low temperature radiation pyrometers to measure the rod temperature correctly. If the silicon rod temperature in the lower temperature range is also of interest, the instrumentation wavelength has to be carefully selected.

Figure 2. Chemical vapor deposition (CVD) as part of the Siemens process.

Aside from the Siemens process, new methods have been proposed for polysilicon production. One potential method is to use a fluidized bed. In this process, seed silicon particles instead of rods are used, and the particles are fluidized. The particles grow into pallets with the CVD process occurring on the surface. A typical reactor design uses a quartz envelop. In that type of design, heaters are usually installed around the quartz envelop, which puts more constraints on the use of pyrometers. When planning for temperature instrumentation, the stray radiation from the heater must be considered, as it can affect the temperature measurement. Because there is no efficient way of cooling the quartz envelop, parasitic deposition on the quartz is hard to avoid. A quartz nipple on the envelope, therefore, is recommended. This allows the use of a replaceable window, which can be kept at a much lower temperature to help minimize window coating.

Multi-crystalline Silicon Growth

Multi-crystalline silicon usually is grown with directional solidification or a similar method. Silicon feedstock is loaded in a crucible and melted. By controlling the cooling from the bottom side, the crystal grows in a vertical direction. Figure 2 is a temperature profile for multi-crystaline silicon growth [2].

There are three temperature instrumentation applications in this process: the heater temperature, the silicon melt surface temperature, and the liquid-solid interface temperature.

Heater Temperature. The growth temperature is typically controlled with a thermocouple or radiation pyrometer on the heater. If used properly, radiation pyrometers can be much more reliable. The selection of the pyrometer largely depends on the furnace’s design details. For example, if the end user wants to retrofit a furnace that previously used a thermocouple, the thermocouple typically leaves a narrow optical path, thus dictating the need for a pyrometer with a narrow measurement beam.

Silicon Melt Surface Temperature. When melted, silicon behaves like metal. Its emissivity drops and its reflectivity increases dramatically at the on-set of melting. Temperature measurement on the silicon melt is commonly monitored by a two-color radiation pyrometer. This, in general, helps handle the changing emissivity. However, there normally is a heater above the silicon melt. The stray radiation can affect the temperature measurement and must be taken into consideration in the instrumentation design.

Liquid-solid Interface Temperature. In directional solidification, it is important to keep the moving liquid-solid interface as flat as possible. Quite commonly, a quartz rod is inserted into molten silicon to determine the liquid-solid interface height. That provides some information, but temperature measurement at different locations would be more useful. New solutions are being developed that would allow manufacturers to measure different locations and depth.

Upgraded Metallurgical-grade Silicon

Upgraded metallurgical silicon is a low-cost way of producing solar grade silicon, and many different manufacturing approaches have been proposed. One method uses a plasma torch on top of the molten silicon to get rid of the unwanted impurities. Depending on the gases uses, the plasma could have many different emission lines and can strongly interfere with the pyrometer measurement. One possible solution is to have the plasma switched off for a short time, and use a pyrometer with a short response time to take the measurement.

Conclusion

Certain environmental conditions are required to grow silicon – ironically, these conditions cause many challenges to creating a high-quality, high-yield product. An important key in overcoming these challenges is precision temperature measurement. By gleaning more information about the overall production environment, manufacturers gain greater control of the final product.

Ji-Dih Hu received his PhD in mechanical engineering from North Carolina State U. and is Sr. Applications Engineer at LumaSense, 3301 Leonard Court, Santa Clara, CA 95954 ; ph.: 408-235-3820; email: j.hu@lumasenseinc.com

References

1. From quartz rock to electronic circuits. Redrawn from http://www.hscpoly.com/content/hsc_prod/manufacturing_overview.aspx

2. V. Chandrasekaran, C. Wang and G. J. Hildeman, “Minicaster” - A Research-Scale Directional Solidification Furnace.” Redrawn from http://www.algor.com/news_pub/tech_white_papers/minicaster/default.asp