Advancements in the commercial production of polysilicon

Advancements in CVD reactor design, improvements in TCS production technology, and increased plant scaling have driven reductions of polysilicon manufacturing costs …

Bruce Hazeltine, Chad Fero, WENJUN QIN, GT Solar Inc., Missoula, MT USA

Advancements in CVD reactor design, improvements in TCS production technology, and increased plant scaling have driven reductions of polysilicon manufacturing costs from hundreds of dollars per kilogram (adjusted for inflation) to less than twenty five to thirty dollars per kilogram.

The feedstock for nearly all crystalline silicon PV cells is polycrystalline silicon (polysilicon). Metallurgical grade silicon (MGSi) is purified to make polysilicon. MGSi is first converted to a volatile chlorosilane and then distilled before turning it back into silicon in a chemical vapor deposition (CVD) process. This fundamental technology has changed little over its history. However, the cost of manufacturing has decreased significantly by advancements in CVD technology, improvements in intermediate gas production technology and plant economies of scale.

Chemical vapor deposition technology

The majority of polysilicon sold worldwide is produced through the trichlorosilane (TCS) Siemens process. In this process, purified TCS and hydrogen are mixed and introduced into a reactor which has a base plate, bell jar and multiple silicon rods that are electrically heated to about 1100?C. This is a batch process where the polysilicon is removed from the reactor when the rods reach a predetermined diameter. The chemical reactions involving hydrogen, chlorine and silicon at deposition temperature and concentrations are quite complicated. However, they can be simplified as follows:

SiHCl3 + H2 Si + 3HCl
SiHCl3 + HCl SiCl4 + H2
2SiHCl3 SiH2Cl2 + SiCl4
SiH2Cl2 Si + 2HCl

Although the process is often referred to as the TCS Siemens process, as shown in the above reactions, dichlorosilane (DCS) participates in the process and contributes to the making of polysilicon. For this reason, most polysilicon producers choose not to separate DCS from TCS. As a result, the TCS stream usually contains 6% to 9% DCS.

Click to Enlarge
Figure 1. Influence of pressure and number of rods on CVD reactor productivity.

To prevent deposition on the inner surfaces of the reactor and to maintain the structural integrity of the reactor, the base plate and the bell jar are cooled by running coolant through them. Because of the huge temperature difference between the polysilicon rods and the reactor shell, a significant amount of heat is lost to the cooled surfaces.

Click to Enlarge
Figure 2. Influence of pressure and number of rods on CVD reactor electricity usage.

Production of polysilicon started in the 1950s. Aided by knowledge and innovation, polysilicon Siemens CVD reactors have evolved and improved tremendously over the decades. These reactor improvements are one of three key elements of success in the modern polysilicon business.

In the history of Siemens CVD reactor technology development, the number of rods has increased from two rods (one hairpin) to over a hundred in the largest low pressure reactors. The operating pressure has increased from atmospheric to ~6barA. The effects of reactor size (number of silicon rods) and reactor pressure on productivity and electricity usage are shown in Figs. 1 and 2.

As the size of the reactor increases, the productivity of a single reactor increases as expected. Reactors operating at atmospheric pressure have about a third of the productivity of reactors running at 6barA. Considering energy consumption, it takes more than twice the energy for the atmospheric reactors to make the same amount of silicon as reactors at 6barA. As the size of the reactors increase, the electricity usage decreases as heat loss per mass of silicon is reduced. However, this benefit levels off at a certain size and further increases result in little or no meaningful savings.

It is important to note that the optimal size of a polysilicon CVD reactor is not based on maximizing the number of rods. Instead, it is about finding the balance between maximizing throughput, reducing energy consumption, and minimizing capital cost for the entire plant.

Figure 3 shows the capital cost to build a polysilicon plant ranging from 1000 metric tons annually (MTA) to 8000MTA capacity using different CVD reactor technologies. This data illustrates the importance of CVD reactor technology to the capital cost of polysilicon plants. Polysilicon made from many small reactors is much more costly than that made from fewer large reactors.

Click to Enlarge
Figure 3. Total plant capital cost by CVD.

TCS production technology

TCS and silicon tetrachloride (STC) have been made from direct chlorination (DC) reactors since the 1940s or earlier. Initially, the dominant demands for TCS and STC were for derivative silanes and fumed silica, and these are still part of the demands today. This process is normally carried out in a fluidized bed reactor (FBR) where finely ground MGSi is reacted with HCl gas at about 300°C and 2-4 barG. TCS is purified primarily by distillation steps.

At the onset of polysilicon production, most producers had a captive TCS supply built for other purposes. TCS FBR capacity of existing facilities had typically been greater than the TCS demand for polysilicon, and byproduct STC could be used to make fumed silica or other silane derivatives. As the demand for polysilicon grew, TCS capacity was increased specifically to support the polysilicon demand.

As the size of TCS plants increased, the lowest cost producers began scaling up their direct chlorination FBRs and generally operated two FBRs per plant site, regardless of size. Even so, there are examples to this day where many more than two FBRs are installed due to scale-up concerns, which significantly increases costs.

Once the byproduct STC produced from polysilicon production exceeded the demand for STC, processes to make TCS from STC were developed. Two dominant technologies were developed to deal with the STC oversupply; STC converters and hydrochlorination (HC). Both technologies entered initial commercial production at roughly the same time, but STC converters represented more production capacity than hydrochlorination in the period from 1985 to 2005. Both technologies are being built into new facilities, but HC has become much more prevalent since about 2004 when patent restrictions lapsed.

At the point TCS capacity was built specifically for polysilicon production, and again when the supply of byproduct STC exceeded demand, the TCS cost for polysilicon plants actually went up slightly, despite other economies of scale.

STC Converters (hydrogenation). The basic technology of STC converters is the hydrogenation of STC at ~1000°C by the following reaction:

H2 + SiCl4 HSiCl3 + HCl

STC converter technology has evolved over time around the following improvements: insulating the pressure vessel from the hot section, improved thermal efficiency by interchanging feed gas and exhaust gas, higher capacity converters, increasing conversion, reducing exposure of heater elements and insulation materials to corrosive gas, and developing components less susceptible to thermal stresses and failures.

There are substantial economies of scale for larger plants with respect to the feed system, compressors, off gas recovery (OGR), and tanks associated with converter operations. However, the cost of converters themselves per unit capacity is nearly linear because multiple units are required.

Cost savings have been contemplated using a shared converter OGR with the CVD reactor OGR. There are obvious economies of scale capital cost savings afforded by installing one large OGR vs. two smaller units, but there are also hidden capital and operating costs. Sharing OGRs generally results in high carbon concentration in polysilicon that must be resolved. At a minimum, sharing OGRs will force all of the un-reacted TCS recycled in the CVD loops to run through a fairly large and energy intensive distillation step to remove methyldichlorosilane (MDCS) and measures must be implemented to remove methane from the recycled hydrogen.
Hydrochlorination. Hydrochlorination (HC) is a process where STC and hydrogen are contacted with MGSi in a FBR to form TCS at reaction temperatures typically >500°C and pressures >20 bar:

3SiCl4 + Si + 2H2 4HSiCl3

As shown in the reaction above, the process hydrogenates the STC and chlorinates the silicon in the same step. Any remaining HCl from the CVD process can be fed back to this FBR and reacts similar to the
DC process.

The hydrochlorination process is continuous and can operate for a year or more between shutdowns. In the HC process, impurities in the MGSi become volatile and leave the FBR with the gaseous reactor effluent and do not accumulate in the FBR. This is not the case for the DC process, which requires bed dumps every 6-10 weeks.

DC-Converter plants, hydrochlorination

In aggregate, the raw material costs are about the same for both processes. The removal of solids downstream of the FBR is more complicated for a hydrochlorination FBR than for a direct chlorination FBR, but these complications have been successfully addressed in modern HC plants.

The OGR in the HC process is analogous to that in the DC process with the addition of a recycle hydrogen compressor. However, it is much less complex than a converter OGR where HCl must be removed.

Because of the higher temperatures and pressures for hydrochlorination, more expensive materials of construction are required for the HC FBR reactor, quench system and pre-heat train than are required for a DC plant.

One of the greatest advantages for hydrochlorination plants versus converters is economy of scale. Those skilled in the art have designed single fluid bed systems having equivalent polysilicon capacities >5000MTA. This same advantage exists for DC FBRs, but to date, no one has commercially offered an STC converter with a capacity over 500MTA polysilicon equivalent.

Because of the number of converters required and corresponding OGR System, the cost of a HC process train is less than the cost of a converter process train. Even if one has an existing DC FBR plant with excess capacity, it will cost more to build a converter train from scratch than a HC process of equivalent capacity. Electrical energy costs for a HC plant are also substantially less than a converter based plant given the lower operating temperature of HC compared with converters.

Capital cost estimates for producing TCS were made using a consistent estimating system for direct chlorination/converter plants and hydrochlorination plants based on USA Gulf Coast construction costs in 2009 dollars. Estimates were made for plants of several different sizes for each technology. The scope of the plant considered is essentially everything required to produce and purify the TCS for the CVD reactors. It does not include the cost to separate un-reacted TCS from the CVD loop back to the CVD loop, which is the same for both processes. For hydrochlorination, the cost estimate includes the hydrochlorination FBR, purification of all TCS coming from the FBR, and any intermediate tanks. For direct chlorination, the cost of the DC FBR, the converters, purification and tanks associated with those operations are captured. In both cases, the cost of purified TCS tanks is assigned to CVD costs. STC storage tanks are assigned to TCS operations.

Click to Enlarge
Figure 4. TCS plant capital cost by TSC technology.

Those familiar with chemical plant capital costs will quickly recognize the economies of scale of traditional chemical plant operations such as distillation and continuous reactors scaled to meet plant capacity. These economies of scale apply to essentially all equipment in a hydrochlorination plant. However, the economies of scale benefit for converters are less than the HC process given the capacity constraints of individual converters.

Click to Enlarge

One can see from Fig. 4 that larger capacity converters reduce the capital cost assignable to TCS, but even a 750MTA (polysilicon equivalent capacity) converter-based plant is still significantly more costly than the same sized plant making TCS by hydrochlorination. The cost of HC is even lower when only one FBR is employed.

Polysilicon manufacturing costs over four decades

Capital and operating cost estimates were made for the prevalent technologies used from 1975 to present. Table 1 shows the summation of this effort. Capex is based on 10 year straight line depreciation and 6% financing costs. All values are based on 2009 dollars. This compilation of data illustrates the following points: capital cost is the largest differentiator in polysilicon cost for modern polysilicon plants; plant size and/or process train capacity is the largest determinant in the cost of TCS; CVD reactor technology (individual reactor capacity and electricity consumption) is the largest differentiator in CVD costs; and the rate of decline in the cost to produce polysilicon has declined more quickly in the past five years than in any previous period.

Manufacturing cost comparison

TCS Siemens polysilicon technology is the most widely used technology for producing the feedstock for crystalline silicon solar cells. The largest polysilicon producers in the industry – Hemlock, Wacker and OCI – utilize this manufacturing method. The fourth largest producer, Renewable Energy Corporation (REC), also maintains production of polysilicon via the Siemens method. However, REC uses silane as the feedstock to the CVD reactor instead of TCS. More recently, REC has chosen to expand with a silane fluid bed reactor (FBR) instead of its silane Siemens technology. MEMC is the other producer using an alternative technology to TCS Siemens. Like REC, MEMC also produces polysilicon by silane FBR. In addition, MEMC has TCS Siemens polysilicon technology and has recently expanded with this technology.

Figure 5 shows the estimated cost of manufacturing for the three major polysilicon technologies. Also included are two subsets of the TCS Siemens polysilicon technology: 1) TCS production via hydrochlorination, and 2) direct chlorination.

The most significant factor for the overall cost is the depreciation and interest charges for the capital equipment. For this study, a straight-line depreciation over 10 years and interest charges of 6% of capital were used. The capital cost of 6500MT TCS Siemens polysilicon plant was estimated at $565 million ($87 per installed kg of capacity) [1]. The capital cost of a 6500MT silane FBR polysilicon plant is estimated at $850 million ($131 per installed kg of capacity) [2]. The capital cost for a silane FBR polysilicon plant would be higher to meet the product quality of a Siemens based polysilicon plant. The capital cost of a 6500MT silane Siemens plant is estimated based on the announced cost of REC’s Butte silicon plant [3] (including scaling).

Click to Enlarge

Figure 5. Manufacturing cost by technology.

It should be noted that the silane FBR technology has lower variable costs. The cost of raw materials is lower because starting filaments and carbon chucks are not required. The cost of energy is lower as a result of the lower electricity usage of an FBR reactor. However, the total energy usage for a silane FBR polysilicon plant is not necessarily lower than a TCS Siemens polysilicon plant. This is because the heat from the CVD reactor can be recovered for distillation in a TCS Siemens polysilicon plant, whereas it cannot be recovered in a silane FBR polysilicon plant. The labor cost is also slightly lower for a silane FBR plant given the easier processing of the granular material. The maintenance costs for a silane FBR polysilicon plant are higher than for a TCS Siemens polysilicon plant, however, thereby reducing the gains of lower cost raw materials and electricity cost. The higher maintenance costs are a result of overall higher capital costs and the requirement for the CVD reactor to run with a hot wall.

The total cost advantage of a TCS Siemens polysilicon plant over a silane FBR polysilicon plant is approximately $5/kg. Where Siemens polysilicon must be crushed to form a flowable product of the same size as granular polysilicon, some or all of this difference is negated from processing costs and fines losses.


1. D. Keck, “Polysilicon Production Costs,” Photon’s 7th Solar Silicon Conf., March 3rd, 2009.
2. Renewable Energy Corporation ASA, Fourth Quarter Results 2009, February 10th, 2010.
3. /Aug2001.htm

Bruce Hazeltine is director of business development at GT Solar Inc., 101 E. Front Street, Missoula, MT 59802 USA


More Photovoltaics World Issue Articles
Previous articleConcentrator photovoltaics: a mature technology for solar power plants
Next articleDrawing on the strength of hybrids to enable greater coverage when screen printing

No posts to display