"Minicaster" - A Research-scale Directional Solidification Furnace

Introduction

A program aimed at expanding the supply of silicon feedstock for producing ingots by the directional solidification technique was funded by the Pennsylvania Energy Development Authority and executed at Solar Power Industries, Belle Vernon, Pennsylvania. Solar Power Industries is a vertically integrated manufacturer of multi-crystalline silicon solar cells. It currently has 13 commercial directional solidification furnaces that cast 265 kg silicon ingots. A research-scale crystallization furnace was constructed to screen candidate silicon feedstock materials and to explore experimental processes to enable their use.

Production scale ingots have melt volumes of 0.629 m³ and a mass of approximately 265 kg, whereas the research furnace charge size was scaled down by a factor of 15 or more to a melt volume of 0.042 m³ at ingot mass near 10 kg. Production of high quality solar cells begins with controlling processing variables during ingot casting. Since casting of commercial size 265 kg ingots is expensive and requires times of approximately 48 hours, there was a need to develop a capability to efficiently cast small ingots for research. The smaller size of the minicaster will allow for the evaluation of candidate feedstock sources and growth techniques on less material having faster turn around times. This paper describes the development and capabilities of the minicaster which is used to conduct research on variables such as silicon feedstock, doping and solidification cycles.

Description of the Minicaster

In the minicaster, the silicon charge resides in a 195 mm x 195 mm x 108 mm vitreous quartz crucible. Graphite plates surround the crucible and serve as mechanical support for the crucible once it begins to soften above 1200°C. A graphite block rests atop pedestal supports and serves as a heat sink for the crucible during growth. Surrounding the crucible is a bank of resistive heaters that uniformly heats the charge. A movable insulation cage serves as the primary means by which the desired cooling rate and directional solidification growth is achieved. Temperature monitoring is conducted using three thermocouples and an optical pyrometer. These sensors are positioned in specific positions throughout the hot zone to aid the thermal profile and modeling. Encapsulating the hot zone is a water-cooled chamber where pressure and inert atmosphere can be regulated.

Operation of the Minicaster

Figure 1 is a plot of a time-temperature cycle for research solidification which is about 12 hours. The primary features of the casting cycle are illustrated in the figure. The crucible that is loaded with silicon feedstock is heated slightly above the silicon melting point of 1410°C. Time at full melt is restricted to minimize the incorporation of impurities from the crucible and heater power is reduced to achieve melt temperatures just below the melting point.

Figure 1: Temperature control scheme for the minicaster.

Nucleation of silicon crystals at the crucible bottom begins, followed immediately by the onset of grain growth. The insulation cage is slowly raised to cool the melt approximately 50°C over an extended growth time. Once solidification is completed, the temperature is lowered to an annealing temperature to relieve internal stresses in the ingot and the system is cooled to room temperature for ingot removal.

Finite Element Modeling of the Solidification Process

In order to assess the design of the "minicaster" hot zone prior to fabricating the components, thermal modeling was carried out using the ALGOR Professional Heat Transfer Analysis finite element package. Questions to be answered included the power and time required to melt the silicon charge, temperature profiles within the melt and system components at the beginning of solidification (no gap in the insulation), the required insulation gap to achieve controlled directional solidification, and the shape of the melt solid interface during solidification. Temperature dependent material properties (density, thermal conductivity, specific heat, and emissivity) were used for silicon, the quartz crucible, the graphite crucible holder and heat sink, the graphite heaters, and the graphite insulation board.

The latent heat of fusion for silicon was incorporated into the model via a spike in the silicon specific heat over a five degree range starting at the melting point (1410°C). Heat transfer via radiation and conduction were incorporated into the model.

Case 1: Transient solidification process with 30 mm insulation lift

Solidification process begins from the bottom of crucible. The ALGOR model was used to calculate thermal profiles in the minicaster ingots. Figure 2 shows the temperature distribution of minicaster during growth period. It is clearly seen that the surface solidification is initiated before the main silicon crystal grows from the bottom to the melt surface. Solidification at the top of the molten silicon is undesirable and thought to be due to the shorter graphite heater and low heater position. The heater could not provide enough heat to the top part of silicon melt to maintain a temperature higher than melting point.

Figure 2: Case 1: Temperature distribution of minicaster furnace
with 30 mm insulation lift, showing solidification on the top surface.

First "Mini"-Ingot

Weight of the initial ingot was 5.5 kg, with ingot dimensions of 195 mm x 195 mm x 62 mm which is suitable for 156 mm square solar cell substrates. As shown in Figure 3, most of the ingot surface is flat and smooth, but there are some regions at the top of the ingot where the solidification proceeded erratically. This is thought to be associated with an undesired solidification at the top of the melt which initiated while solidification was occurring from the bottom upward. Such a solidification was predicted by the finite element thermal model of the growth. Since silicon expands upon solidification, the last of the melt to freeze may have been partly trapped between the solidified lower portion and upper portion of the ingot and have been squeezed out along the edges of the ingot to give the shape observed.

To eliminate the surface solidification, the following were considered: 1) Increase the heater power level; 2) Change the graphite heater position; 3) Change the insulation lift distance.

Figure 3: Photograph of first silicon ingot produced in August 2006,
before being removed from the minicaster (5.5 kg, 195 mm x 195 mm x 62 mm).

Case 2: 30 mm insulation lift with higher heater power level

To eliminate surface solidification, the graphite heater power level was increased by 25%, the temperature distribution of which is shown in Figure 4. With more heat introduced into crucible, the surface solidification has been eliminated. However, since more heat has been added from side walls of crucible, the solidification interface is much more convex to the melt which will make it take much longer time for corner growth after central growth finished. Such a long period of time may increase the chance of contamination from crucible or graphite parts of the furnace into silicon melt.

Figure 4: Case 2: 30 mm insulation lift and 25% increase of
graphite heater power level showing a highly convex solidification profile.

Case 3: 30 mm insulation lift with higher heater position and power level

In addition to increasing the heater power level by 25%, the heater position was also raised. Figure 5 shows the temperature distribution of the minicaster during the solidification process. No surface solidification was predicted during the growth process. The solidification interface is flat and a little convex to the silicon melt which is beneficial for high quality silicon crystal growth.

Figure 5: Case 3: 30 mm insulation lift with higher
graphite heater position and 25% increased heater power level.

Summary of Modeling Results

In the current minicaster design, the graphite heater position is relatively lower than the production DSS furnace. Such low position might give surface solidification during growth process. Several possible solutions have been provided. Increasing the heater power may cause longer growth process time which will enhance the chance of the secondary contamination from other thermal components inside furnace.

By adjusting the graphite heater to a higher position, the temperature of silicon melt surface can be maintained higher than melting until the solidification interface reaches the surface.

Another effective way to adjust the thermal gradient is by adjusting the insulation lift distance. More research will be done in the future. Moreover, the thermal model will be calibrated using experimental data from minicaster furnace operation.

Acknowledgements

The authors gratefully acknowledge Robert Stoehr, Barry Munshower and John Belcher for their contributions to the development of the minicaster as well as Dan Meier, formerly at Solar Power Industries. The Pennsylvania Energy Development Authority provided a grant to fund this program.

To read a customer application story about SPI's work, see "ALGOR FEA Helps Engineers Energize Solar Cells".