Polysilicon

2006
Polysilicon
Supply, Demand, & Implications
for the PV Industry
Produced by the Prometheus Institute for Sustainable Development
Hilary Flynn
Travis Bradford
About the Prometheus Institute
Many people provided information

and guidance for this report. We
would like to acknowledge these
individuals and express our sincere
thanks for their assistance.
Romy Acosta, Christof Bachmair
Johannes Bernreuter, Peter Bostock
Nolan Browne, Rob Bushman, Goran Bye
Shreya Dave, Iain Dorrity, Joerg Duske
Giso Hahn, Tor Hartmann, Gary Homan,
M. Iizuka, Christian Keck, Wieland Koornstra
Sanjeev Lahoti, Mikhail Lekhtman, Zijun Li Ian
MacLellan, Kiyoko Matsuyama
Paul Maycock, Vince Mayeda
Kelly McGrotha, Hiroyuki Oda, Leila Panzner
Jesse Pichel, Michael Rogol, Jörg Rottstegge Colin
Rusch, Fred Schmid, Howard Somberg Our Vision
Keiji Takebuchi, Takeshi Yamagami, Deren Yang
Our goal is to accelerate the deployment of socially-beneficial sustainable
technologies, including those of energy, water, and food, by educating
industry participants, advocates, and policymakers about their advantages.
We achieve our mission by collecting and disseminating reliable data,
quantitative analysis, and practical information about these industries.
Today, the Institute strives to be the world’s leading source of publicly

available primary data on the photovoltaic (PV) supply chain and end-
markets through a series of data collection and outreach projects and the
publication of the industry’s oldest newsletter, PVNews™. Currently, the
Institute focuses solely on its initiatives in the solar energy industry.
Contributing Staff
Hilary Flynn
Author
Travis Bradford
Author
Robert Schlussler
Design and Layout
Kate Cell
Editing
Cover photo courtesy of Renewable Energy Corporation.
Polysilicon: Supply, Demand, & Implications for the PV Industry

Table of Contents
Foreword v
1 Polysilicon—An Overview 1
2 Polysilicon Processing 5
2.1 Solar Grade vs. Electronic Grade 5
2.2 From Silica to Metallurgical Silicon 5
2.3 Metallurgical Silicon Processing 6
2.4 Silicon Ingots: Monocrystalline vs. Multicrystalline 8
3 Polysilicon Production Capacity—Historical Production and Major Players 13

3.1 Prometheus Institute Research Methodology 13
3.2 History of Polysilicon Production and Prices 13
3.3 2005 Polysilicon Production Capacity 14
3.4 Polysilicon Producer Profiles 14
3.5 Polysilicon Recyclers—IC Castoffs 20
4 Emerging Polysilicon Producers and Supplemental Silicon 23

4.1 New Producers Using Conventional Silicon Processing 23
4.2 Alternative Silicon Production 26
5 Geography of Global Polysilicon Capacity 31

6 Projections and Forecast Model 33
6.1 Polysilicon Production Capacity Projections to 2010 33
6.2 Changing Market Dynamics for Polysilicon Producers 34
6.3 Demand Forecast Model Assumptions 35
6.4 Model Results 36
6.5 Sensitivity Tests 37
7 Polysilicon Constraint Implications 39

7.1 Polysilicon Prices 39
7.2 Changing Industry Structure 39
7.3 Efficiency Gains 40
7.4 Emergence of Alternatives: New Silicon Processing and Gains in Thin Film Production 41
7.5 Concluding Thoughts 42

Table of Figures
Figures
Figure 1: Polysilicon 1
Figure 2: 1996 to 2005 cell production (MW) and % growth rates 1
Figure 3: Silicon metal 5
Figure 4: Silicon production 5
Figure 5: Production of metallurgical silicon (metal silicon, MG) 6
Figure 6: Percent of polysilicon produced by technology, 2005 and 2010 6
Figure 7: Chunk polysilicon 7
Figure 8: Granular polysilicon 7
Figure 9: Fluidized bed reactor 7
Figure 10: Vapor-to-liquid deposition process 7
Figure 11: Sample route for refining MG 8
Figure 12: CZ drawing 9
Figure 13: Float zone diagram 9
Figure 14: Cast ingot 9
Figure 15: String ribbon machines 10
Figure 16: Sheet (or EFG) 10
Figure 17: Wire saw machine 10
Figure 18: Historical production capacity 13
Figure 19: Polysilicon production 2005 14
Figure 20: Chinese producers 2005 - 2010 25
Figure 21: Map of production 31
Figure 22: 2005 - 2010 producers 33
Figure 23: 2005 polysilicon producer market share 34
Figure 24: Base case model results 36
Figure 25: Sensitivity test #1: 6 g Si/ W by 2010 36
Figure 26: Sensitivity test #2: 20% thin film market share by 2010 37
Figure 27: Sensitivity test #3: 20% thin film market share and 6 g Si/W by 2010 37
Figure 28: Average silicon prices 2003 to 2005 with projections to 2007 39
Tables
Contents
Table 1: Silicon production announcements from April–September 2006 3
Table 2: Silicon producer capacity in 2005, projected and potential in 2010, and technology employed 15
Table 3: New silicon producer capacity in 2005, projected and potential in 2010, and technology employed 23
Table 4: Geographic distribution of polysilicon production as a percent of total production 31
Table 5: Piper Jaffray solar industry production estimates 35
iii
Foreword October 2006
This polysilicon supply report is the first of a series of comprehensive Travis Bradford
reports from the Prometheus Institute that delves deeply into the various President,
links in the photovoltaic (PV) supply chain. These reports meet our
mandate to provide comprehensive industry information and analyses
Prometheus Institute
in order to accelerate the deployment of sustainable technologies such for Sustainable Development
as solar electricity. We believe that delivering good data and information
about the industry’s stages of production and its prospects for growth
will help industry stakeholders, users, and policy makers make the best
decisions about how and where to increase the use of PV.
Given the amount of industry speculation this year about silicon supply,
a comprehensive review of this vital feedstock is timely. With raw
material prices rising and rumors of idle cell production capacity, silicon
has become the bottleneck for the growth of the PV industry. In 2005,
nearly 95% of the cells produced used silicon-based technologies, and
long lead times to deploy new production plants mean bottlenecks will
persist, slowing the industry’s growth rate, historically in excess of 30%
per annum. Thin films, while promising, will not increase their market
share fast enough to keep the PV industry from relying almost entirely on
the polysilicon supply through 2010. Understanding the supply of silicon,
therefore, is essential to understanding the prospects for the PV industry
as a whole.
In this report, we look at all of the factors that affect the availability
of this feedstock for the PV industry. We survey seven major polysilicon
producers and nearly 20 emerging producers (a third of them in China),
and we project production capacity through 2010 under a variety of
scenarios. In so doing, we determine that the current silicon supply
shortage will begin to ease by 2008, and that there will be enough silicon
to allow the PV industry to grow to eight GW per year by 2010.
We also look at various technological approaches to refining silicon,

different methods of increasing efficiency of silicon use, and dynamics
shaping the industry. What emerges is a view of a stable base of producers
that is currently enjoying increased pricing power and profitability.
Meanwhile, however, many new players have been attracted to silicon
manufacturing. These newcomers threaten to erode the market share of
established producers and perhaps even to flood the market by the end
of this decade. While much new capacity is planned and outcomes are
uncertain, what is certain is that polysilicon supply is in for a dynamic
transformation in technology, industry structure, pricing, and margins
over the next four years.
I want to thank all the staff, advisors, and company representatives

who helped put this report together. Much thought and work went
into creating this report, and many people provided data, insight, and
suggestions for improvement. However, this report, like all the research
of our Institute, remains a work in process. It is vital to us and to the
mission we serve that we continue to improve the information and
analysis we provide to our readers. Please help us to make this report
better by sending your comments, corrections, and suggestions. They
are always welcome.

1 Polysilicon—An Overview
In 1954, Bell Labs created the first silicon solar cell. Prior to that, solar
cells were made of selenium, and not particularly productive, with a
0.5 percent sunlight-to-electricity conversion efficiency (sometimes
referred to simply as “cell efficiency”). Silicon-based cells on the other
hand achieved six percent conversion efficiency in the early years.
Today, some companies report cell efficiencies of 20 percent or more.
While solar cell semiconductors can be made from various compounds, Figure 1: Polysilicon
including silicon, cadmium telluride, copper indium diselenide, etc.,
the easily transferable developments in silicon technology from the
electronics industry and the long history of silicon-based semiconductors
make silicon by far the most popular material for solar cells. Currently,
over 90 percent of the global PV production is silicon wafer-based.
Until recently, solar cell manufacturers could obtain sufficient silicon
feedstock supply in the form of off-spec and waste material from the
electronics industry. This lower quality material was sold at half the price
of high quality material. With the strong growth of the PV industry in
2003, 2004, and 2005, the demand for silicon for PV cells has exceeded
the amount available through traditional supply channels and has led
to industry-wide material shortages and rising silicon feedstock prices.
The PV industry has grown 40 percent per annum, on average, over Courtesy of Sumitomo
the last decade. The semiconductor industry, on the other hand, has a
cyclical demand pattern. In the semiconductor bust of 2001, polysilicon
producers found themselves with significant overcapacity, making them
until recently reluctant to increase capacity to meet aggressive growth
predictions from the PV industry. In the past few months, most of the
seven major polysilicon suppliers worldwide have announced new
capacity expansion plans in response to the growing demand from the
PV industry. In addition, numerous companies outside of these veteran
producers have announced plans to produce silicon using existing and
new methods of silicon production.
Figure 2: Cell production (MW) and % growth rates 1996 to 2005
Polysilicon production is a capital-intensive
and highly technical business that requires
2,000
large amounts of electricity for melting and
purification. It is not surprising then, that
more than 50 percent of the polysilicon
production in the world occurs in the United 1,500
47%
Production (MW)
MW Produced
States which has a highly-trained labor force

and relatively low energy costs. The rest of % Growth
the worldwide production also occurs in 1,000
%
61
technology-oriented countries, including

%
Japan (24 percent) and Germany (18 32
percent). In the next few years, the industry is 500 44%
projected to see capacity added in countries 36%
43%
like Norway, China, Spain, and Korea. Even as 23% 30%
42%
polysilicon capacity becomes more distributed 0
throughout the world, however, the United
96
97
98
99
00
01
02
03
04
05
20
20
19
19
19
19
20
20
20
20
States is expected to continue to be the top

producing country through 2010. Year

SECTION 1 Polysilicon—An Overview
The existing polysilicon producers traditionally catered to the electronics

industry, with the majority of their product being sold for integrated
circuits. We expect 2006 to mark a transition in polysilicon consumption,
with the majority of the world’s polysilicon henceforth consumed by the
“While many new PV industry. Most of the existing polysilicon producers are now responding
to the growing PV demand by adding capacity, and a number of new
companies are companies will cater to the solar industry. Some of these companies are
even developing technology to produce silicon specifically for the PV
entering the silicon industry with appropriate purity levels and lower costs.
business, those with These new polysilicon producers fall into three groups. The first group
established upstream is companies that already produce the silicon feedstock used to make
polysilicon (e.g. Elkem, DC Chemical, and Degussa). The next group consists
experience will of companies that are in the solar industry and looking to become more
integrated along the supply chain (e.g. GiraSolar, SolarWorld, and ARISE).
be most likely to The third group of companies is completely new to both silicon processing
and PV manufacturing (e.g. Hoku). We believe companies that already
succeed.” have experience in silicon processing upstream of polysilicon production
will be the most successful entrants to the market.
Table 1 illustrates the spate of new capacity expansion announcements

since April, 2006. We sorted these companies into three categories:
existing producers; new producers using existing technology; and
new companies using new technology. Existing technology, as we will
explain in more detail in Section 2, includes the Siemens process and
the fluidized bed reactor (FBR) process. Siemens is the industry standard,
but a few companies are involved with FBR-based production. The newer
technologies being developed (which have been researched for decades
“Between April and but not yet brought to commercial scale production) are more focused
September, plans for on a solar grade, or lower quality and thus cheaper, silicon product. This
is achieved by modifying and simplifying the refining process.
70,000 additional The advantages of building capacity based on the traditional Siemens reactors
tons of capacity to be include the transparency and low risk of using a well-established and well-
understood process. The Siemens process was used to produce over 90 percent
built by 2010 were of polysilicon in 2005. For producers using this established technology, there
is little risk of patent infringement and it is comparatively easy to build the
announced — this production facility. The newer technology of FBR has higher intellectual
property concerns and technology risks, but is being deployed by a few
would triple total credible producers today. Both of these technologies share the advantage that
they produce silicon pure enough to be used in both PV and IC applications.
polysilicon supply and However, the ability to dramatically cost-engineer these technologies and
quadruple the amount reduce the polysilicon prices for cell manufacturers is limited. A cheaper
alternative would be to build capacity strictly for the PV industry, the costs of
available to PV.” which are targeted to be half of what Siemens or FBR costs are.
The majority of new production recently announced will be undertaken by

new companies using existing technology. Approximately 36,000 MT of new
capacity falls into this category. The most interesting information in Table
1 is the amount of capacity under development using new technology. At
the end of the day, this is where PV could ultimately reduce cost per watt,
which is crucial for the industry to stimulate long-term demand.

Polysilicon—An Overview SECTION 1
Table 1: Silicon production announcements from April–September 2006

New Producers &
Month Existing Producers New Producers
Technology
April Sumitomo; 400 MT ARISE; 2,000 MT
Mitsubishi Polysilicon
America; 300 MT
Hemlock, 4,500 MT
May REC; 6,500 MT CSG Holdings investing $150m, capacity GiraSolar files for Si patent
to reach 4,000 to 5,000 MT
Hoku; 1,500 MT
Shanghai City to invest in
4,000 MT
June Oregon/Japanese Si producer; Over 20,000 MT/yr possible in China2
5,000 MT1
Wacker; 4,500 MT Solar EnerTech, PAIS, Jin Yu Silicon &
Wuhai Gov’t
July SunPower/DC Chemical; 3,000 MT
Isofoton, Endesa, gov’t agency; 2,000 MT
August Econcern, Photon Power Technologies SolarValue; 10,000 MT
(PPT) and NorSun ; 2,000-3,000 MT
M.Setek; 6,000 MT (not announced)
JFE Steel; 100 MT
September Dow Corning SoG blend; Global PV Specialists using
1,000 MT rice hulls; 2,000 MT
Total ~22,000 ~36,0003 ~14,000
The combination of new production by existing suppliers, new entrants,

and cell producers engineering their products to use existing silicon
feedstock more efficiently, will ensure that the PV industry can continue
its historic trends of market growth and price reductions. This polysilicon
report will explore each of these trends in detail and then look at how
they will collectively affect the market through 2010.
In the next section, we explore the different silicon production

technologies as well as the process of making a solar cell through ingot
and wafer manufacturing.
1. Read, Richard. “Enticing Japan Back to Oregon,” June 6th 2006, The Oregonian.
2. “Analysis: China’s Polycrystalline Silicon Production,” June 14th 2006, AsiaPulse.
3. We have not accounted for China’s entire 20,000 MT here because some of the companies
in the article were already known to be in the construction or planning phases of
polysilicon production.

2 Polysilicon Processing1
2.1 Solar Grade vs. Electronic Grade

Silicon is the second most abundant element in the earth’s crust. In
nature, it is found as an oxide (in the form of sand and quartz) and as a
silicate (in the form of granite, clay, and mica).2 Silicon is sold in a range of
purity levels, or grades. For silicon to be a useful semiconductor material, Figure 3: Silicon metal
it must be highly purified. For solar cells, the silicon must be 99.9999
percent pure (often referred to as “six nines” or 6N pure).3 The silicon
grade used in electronics is even more pure, typically 9N to 11N. The
type of impurities also makes a difference. While carbon and oxygen are
less significant, boron and phosphorus concentrations must be managed
since they are important in the electrical functioning of a cell.
For silicon to reach semiconductor grade, whether for solar cells (i.e.
solar grade, SoG) or integrated circuits (electronic grade, EG) it must
undergo a significant amount of processing. Silicon that is purified for Courtesy of Resource Committee, US
the semiconductor industry is referred to as polycrystalline silicon, or House of Representatives
polysilicon (poly-Si, or poly). Silicon used in solar cells has historically
come from off-spec and waste silicon, produced either during the
polysilicon purification process or during ingot and wafer production.
But this is no longer the only source of polysilicon for the PV industry. As
demand increases, solar companies are increasingly buying higher quality
silicon. An industry rule of thumb is that 10 percent of the polysilicon sold
to the semiconductor industry will eventually become available to the
solar industry. However, polysilicon producers are now producing silicon
specifically for PV companies. In addition,
new technologies are being developed to Figure 4: Silicon production
produce silicon that caters to the needs of
the PV industry.
2.2 From Silica to

Metallurgical
Silicon
Silicon used in the semiconductor and
PV industry must go through several
steps before it is a practicable feedstock.
Figure 4 illustrates one route from silica
mining to the solar module or integrated
circuit using the Siemens purification
process. This section will describe the
process from mining to metallurgical
silicon (referred to as metallic silica in
the figure). The next section will discuss
the routes for further purification.
Courtesy of Tokuyama

SECTION 2 Polysilicon Processing
Figure 5: Production of metallurgical silicon (metal The first step in polysilicon production is the extraction
silicon, MG) of quartz from a silica mine. The quartz is then put into a
furnace with a carbon source—a mixture of coal with coke,
woodchips, or charcoal. The mixture is then heated and the
silicon reduced in a process called Carbothermic Reduction.
This produces liquid silicon, carbon dioxide, and silica fumes.
The silica fumes are used for other industrial processes, while
the liquid silicon is poured out of the furnace. The liquid is
further refined, then allowed to solidify. The resulting silicon
material is referred to as metallurgical silicon, or metal
silicon (MG-Si). Producers then crush the MG-Si before it is
sold. The purity level at this stage is 96-99 percent, with an
average purity of 98.5 percent. The cost of MG-Si is relatively
low, estimated to be around $0.77/lb or $1.70/kg.4
US MG-Si production in 2005 was 148,000 MT.5 Globally,

more than one million MT was produced. Approximately
half the MG-Si produced in a year is used by the aluminum
industry; the other half is used in various chemical processes.
MG-Si is not pure enough for use in integrated circuits or
PV cells; further refining is necessary.
There are several ways in which MG-Si can be refined. We

Courtesy of Elkem will describe these processing options below.
2.3 Metallurgical Silicon Processing

The majority of polysilicon used by the semiconductor and PV industry
is produced via a process of chemical deposition, whereby a chlorosilane
gas is deposited onto a heated rod. The first step is
Figure 6: Percent of polysilicon produced by technology, 2005 to react HCl with MG-Si. This forms a liquid that
and 2010 is distilled and then vaporized. The resulting gas
is then deposited onto heated silicon rods (1100°).
11% 1% The earliest process, named after the company that
developed the process (Siemens), uses trichlorosilane
9% gas (TCS) as the deposition material. TCS (HSiCl3) has
13% many advantages, including a high deposition rate
and high volatility (which makes it easier to remove
2005 2010 two compounds that are problematic in solar cells,
boron and phosphorous). One of the disadvantages
of using TCS is the high electricity requirement to
maintain process temperatures.
91%
75%
Another process further refines TCS to produce
monosilane (SiH4). This gaseous monosilane is
then deposited on heated silicon rods. Monosilane
is a higher purity starting material which leads
Siemens MG-SoG
to more pure polysilicon. This higher purity also
makes it more expensive to produce. This process
FBR Other was developed by Union Carbide in the 1970’s
through research funding from the US government

Polysilicon Processing SECTION 2
to produce solar grade silicon. The funding was then rescinded and
Figure 7: Chunk
the company’s focus shifted from PV to the semiconductor industry.
polysilicon
The company has changed hands several times, but is now owned by
Renewable Energy Corporation which has converted it back to a strictly
solar grade silicon producer.
The final product of the above two processes is a rod of polysilicon that
is broken up into smaller pieces; at this point the product is called “chunk
polysilicon”.
A third process for polysilicon production uses a fluidized bed reactor

with a final product of granular silicon.
Figure 9: Fluidized bed reactor

Courtesy of Mitsubishi Polysilicon
Figure 8: Granular
polysilicon
Courtesy of ScanWafer
Ethyl Corporation developed the FBR process for polysilicon production.

The company used silicon fluoride instead of MG-Si, which is converted
to monosilane, and then the silicon seed is dropped into the reactor while
the silane and hydrogen gases pass through. The theoretical advantages
of this process are lower capital and electricity costs than Siemens
reactors. To date, however, only a couple of producers have established Courtesy of MEMC
FBR capabilities, so the production economics are still unclear. Ethyl
Corporation’s FBR business unit was eventually sold to Albemarle
Corporation, which was in turn bought by MEMC. Renewable
Energy Corporation has also announced new FBR capacity under
construction. Depending on a company’s wafering process, granular Figure 10: Vapor-to-liquid deposition process
polysilicon may be preferred over chunk polysilicon.
A fourth and still experimental polysilicon technology is Tokuyama’s

“Vapor-to-Liquid Deposition” (VLD) process. As figure 10 shows, this
process uses TCS as does Siemens, but the reactor is a tube in which
liquid silicon forms. The silicon product could thereby be produced
faster than through the Siemens process.7 The company’s focus is
on making a low cost product that would be suitable for solar use.
Based on results from the semi-commercial plant (200 MT), Tokuyama
believes it can provide a cheaper feedstock material to the industry
and is in the design phase of a large scale operation.
The polysilicon industry, regardless of the technology, requires a

significant amount of capital, large amounts of electricity and a Courtesy of Tokuyama

highly trained labor force. As a general rule, it costs $100/kg to build a

polysilicon facility. For comparison purposes, a 5,000 MT plant would cost
$500 million to build and could produce enough silicon for 450 MW of
annual PV cell production at today’s average efficiency rates. Assumed
capital recovery plus operating cost for this means that the cost of silicon
is at least $0.60 to $0.75 per watt, or 15 to 18 percent of the entire cost
of today’s solar modules (average of $3.50 per watt). If the industry wants
to bring average costs for modules down to $2.00 per watt, a level widely
thought to be economically transformative in global markets, additional
cost reductions in silicon processing will be required.
In an attempt to lower silicon production costs, as well as to ensure

feedstock is available to the PV industry, methods dedicated specifically
to solar grade silicon have been under development. This research includes
variations on existing purification processes, as well as
Figure 11: Sample route for refining MG completely novel processes. To date, minimal amounts of
product have resulted from this research, though many
companies have promised product over the next few years.
Upgraded Metallurgical Silicon

Upgrading the metallurgical silicon process could be a cost
effective way to produce silicon for the PV industry.
Companies such as Dow Corning, Elkem, and others are

pursuing this route of SoG silicon manufacturing. The process
involves a series of refining steps and the employment of
directional solidification (described in more detail in section
2.4). While this route offers the promise of lower costs than
the Siemens process, product quality remains an issue. To
date, only Dow Corning is commercially producing SoG
from MG-Si, though the quality is not high enough for the
product to be used on its own; it must be blended with purer
Courtesy of Crystal Systems, Inc. silicon. Elkem has not brought its technology to commercial
scale, though it has ambitious plans to do so by 2008.
2.4 Silicon Ingots: Monocrystalline

vs. Multicrystalline
Wafer and cell producers receive the silicon in chunk or granular form
but then need to shape it into a form that can be sliced. The shape can be
an ingot, block, ribbon, or sheet. The product can also be monocrystalline
(single crystal) or multicrystalline (polycrystalline). The Czochralski
(CZ) and float zone methods produce monocrystalline ingots, while
directional solidification/casting, ribbon, and sheet techniques produce
multicrystalline structures.

CZ Method Figure 12: CZ drawing

Czochralski crystal growing starts with melting the silicon in a crucible.
Then a rod with a silicon seed is dipped into the molten silicon and as it
is drawn up, a monocrystalline silicon crystal is grows on the seed crystal.
Figure 12 illustrates the concept of CZ pulling. In 2005, monocrystalline
wafers made from CZ wafers accounted for 35 percent of global
production. CZ pulling takes a significant amount of time and is more
expensive than the other methods, but the resulting cells have some of
the highest conversion efficiencies in the industry.
Float Zone
Float zone ingot formation is used for producing even more pure wafers. Courtesy of EERE
A float zone ingot has fewer impurities than a CZ ingot; it is particularly
lower in oxygen which can decrease the efficiency of a cell. As with CZ
crystal pulling, a seed crystal is exposed to molten silicon. But instead of Figure 13: Float zone
being dipped into a crucible with a silicon melt, a heating coil passes along diagram
an ingot, effectively separating the newly crystallized monocrystalline
ingot from the input silicon. Crystallization occurs between the solid and
molten regions referred to as the “float zone” (figure 13).
Directional Solidification/Casting
Multicrystalline blocks are formed via casting or directional solidification.
While this process takes less time than monocrystalline production,
efficiencies are lower due to the variable silicon crystal sizes, dislocations,
and impurities. Casting can take place in the crucible in which the silicon is
melted or the silicon can be poured into a second crucible.
With directional solidification, the silicon remains in the crucible after

heating. Once the silicon is melted the entire crucible is moved down, away
from the heating element and the silicon solidifies as it cools. Figure 14
shows the process of casting. Courtesy of University of
Delaware
Ribbon/Sheet
Two methods have been developed to reduce the amount of slicing, Figure 14: Cast ingot
and thus waste, associated with the above methods. The first, String
Ribbon technology, is a proprietary technology employed exclusively by
Evergreen Solar. A crucible melts the silicon, and then two ribbons of
silicon are pulled up out of the crucible. Once the ribbons are slightly less
than two meters long, they are removed and sliced into wafers.
A second method is Edge-defined Film Fed Growth (EFG), developed and

exclusively employed by SCHOTT Solar. An octagonal hollow tube is pulled
up from a silicon melt. Once it reaches six meters, the tube is removed
from the machine and sliced into wafers with a laser.
Crystal Growing Systems (CGS) and SCHOTT recently succeeded in

improving the Edge Defined Film Fed Growth (EFG) process for wafer
production. Previously, SCHOTT employed octagonal EFG process, which Courtesy of EERE

Figure 15: String ribbon machines produces eight-sided tubes of silicon that are then separated with a laser.
The two companies developed a process for 12-sided (dodecagonal) tubes.
While the increase in silicon pulled obviously has productivity benefits, the
thickness of the 12-sided tubes is also reportedly more homogeneous. Both
factors could offer significant cost advantages over current methods.
2.5 Wafering
Ingots, blocks, ribbons and sheets of silicon are sliced into wafers prior to cell
manufacturing. Wafer-based PV is the predominant cell type produced: 94
percent of PV production in 2005 was based on silicon wafers. The wafering
process wastes a significant amount of silicon. Wire saws are used to slice the
wafers from ingots and blocks. Wafers are typically in the range of 200 – 300
Courtesy of Evergreen Solar µm thick. The wires destroy 220 - 230 µm of silicon as they slice though the
block. Wire saw performance is improving, and new techniques are under
investigation to reduce waste. Lasers are an option, though there the heat
Figure 16: Sheet (or from the laser slicing through the ingot causes the outer silicon to degrade.
EFG)
In 2005, it took an average of 12 grams of silicon to produce one watt of
PV, including silicon lost in the wafering process. This requirement should
decrease over the next few years to 9 g Si/W or less. (We perform sensitivity
tests in our projections to account for greater advances in silicon use per
watt. See section 6.) In order to achieve targeted improvements in grams
per Watt, three methods are being pursued:
• First, sawing thinner wafers may help to reduce silicon use, though
sawing losses will persist. A silicon requirement of 6 g Si/w or less
is theoretically achievable, but cells this thin also increase breakage
concerns for wafers in sawing and material handling.
• Second, the ribbon and sheet methods of wafering significantly
reduce waste. Evergreen Solar’s wafers reportedly use 6 g Si/w
today, with no material lost from wire slicing.
• Finally, efficiency improvements in cell manufacturing can reduce
grams per Watt through more power from the same chip. This
Courtesy of SCHOTT Solar
method has been effectively pursued by SunPower, whose 20 percent
to 21.5 percent efficient cells reportedly only use 8 g/W.
Figure 17: Wire saw machine
Each of these will help to improve the efficiency of silicon

use in PV. But to meet the dynamic growth projections of
the industry and the current global demand, significant
new production capacity will be required. The next section
looks at the current polysilicon producers and their plans to
expand capacity through 2010.
Courtesy of LDK Solar
10
1 This section relied on several sources: PV handbook, Elkem’s website, http://www.wafernet.

com/PresWK/h-ptl-as3_wsc_siltronic_com_pages_training_pages_Silicon_TCS-1.htm;
http://www.udel.edu/igert/pvcdrom/; and personal communication with Dr. Giso Hahn of
the University of Konstanz
2 http://periodic.lanl.gov/elements/14.html/
3 EERE Solar Energies Technologies Program www1.eere.energy.gov/
4 http://minerals.usgs.gov/minerals/pubs/commodity/silicon/silicmcs06.pdf/
5 http://minerals.usgs.gov/minerals/pubs/commodity/silicon/simis1205.pdf
6 http://www.tokuyama.co.jp/eng/news/release/pdf/2004/20040908_silicon_e.pdf#search
=%22tokuyama%20vapor%20to%20liquid%20deposition%22
7 Oda, Hiroyuki, Tokuyama VLD at a Glance, 3rd Solar Silicon Conference, April 2006,
Munich, Germany
11
3 Polysilicon Production Capacity
—Historical Production and Major Players
For the purposes of this report, we consider production capacity to be

equal to the amount of polysilicon a company will produce in any given
year. This number is a base estimate, however, as the amount can be
increased five to ten percent depending on the desired quality of the
product and improvements over time of the manufacturing process (such
as de-bottlenecking). For example, a lower quality silicon destined for a
solar company could be produced faster than higher quality material,
which would allow for more to be produced in any given time frame. It
“Due to a glut of
is possible that some companies have reported theoretical capacities that silicon after the
are not physically achievable, but we have attempted to standardize with
actual production capacities throughout this report. tech crash, little
new capacity
3.1 PI Research Methodology
has been added
This section provides polysilicon production capacity in 2005, with
projections to 2010. The Prometheus Institute for Sustainable since 2000.”
Development, in preparing this report, obtained information from each
polysilicon company concerning its current and projected production
capacity at each of its facilities. We also asked several general questions
that are presented in aggregated form. In almost all cases company
representatives provided production data. For the few companies whose
representatives were unable to comment, we used company press releases
and websites to gather the needed data. The research also included
surveying cell producers for their assessment of the silicon supply situation,
as well as reviewing previously published reports on global silicon supply.
In April of 2006, Solar Verlag GmbH hosted the 3rd Solar Silicon Conference
in Munich, Germany. Over 700 people, including silicon producers, ingot
and wafer producers, cell manufacturers, and solar researchers, attended
this one-day event. Industry analysts, polysilicon companies and research
institutes gave presentations and took questions from the audience. We
attended this event and obtained information not previously
available. We also attended the Semiconductor Equipment Figure 18: Historical production capacity
and Materials International’s SEMI Europa event in Munich
that week, where solar equipment and cell companies
35,000
discussed the issues their companies were facing.
30,000
3.2 History of Polysilicon
Capacity (MT)
25,000
Production and Prices 20,000
Before we discuss current and future polysilicon capacity, 15,000
we will briefly mention the capacity and price environment
10,000
in the years leading up to the current shortage.
5,000
In 2000, polysilicon production was an estimated 24,000
MT. Around this time, demand for polysilicon from the 0
01
02
00
03
04
05
semiconductor industry dropped, leaving the polysilicon

20
20
20
20
20
20
producers with excess capacity. As figure 18 shows, very

little capacity was added from 2000 to 2005. Year
13
SECTION 3 Polysilicon Production Capacity
3.3 2005 Polysilicon

Production Capacity
In 2005, the global capacity for polysilicon
Figure 19: Polysilicon production 2005 production was 31,280 MT. There are seven
companies that dominate the polysilicon supply:
8,000
7,700 Hemlock, Wacker, Renewable Energy Corporation,
7,000 Tokuyama, MEMC, Mitsubishi, and Sumitomo
Titanium Corporation. Hemlock, located in the
6,000
Capacity (MT)
US, is the largest of the polysilicon companies

5,500
5,000 5,300 5,200 with a production capacity in 2005 of 7,700
metric tons (MT). It was followed by Wacker
4,000 (5,500 MT), Renewable Energy Corporation (5,300
3,800
3,000 MT), Tokuyama (5,200 MT), MEMC (3,800 MT),
2,850 Mitsubishi (2,850 MT), and Sumitomo (800 MT).
2,000 There was a small amount (130 MT) supplied by
1,000 two companies in China. Each of these companies
800
130 is discussed in greater detail in section 3.4, except
0 for the Chinese companies, discussed in greater
detail in Section 4.1.
ck
na
ish
ke
am
m
RE
EM
lo
hi
ac
ito
ub
m
uy
lC
M
W
m
He
its
Polysilicon capacity has not increased significantly

k
ta
To
Su
M
To
over the last five years due to production

overcapacity that resulted from a downturn in the
semiconductor demand versus the growth forecast. In 2000, polysilicon
manufacturers report that there were approximately 24,000 MT of
capacity worldwide.2 From 2000 and 2004, capacity expansions across all
producers were very modest, on the order of 1,000 MT per year or less.
Between 2004 and 2005 there was a jump in capacity of 3,000 to 4,000
MT to bring capacity above 30,000 MT.
3.4 Polysilicon Producer Profiles

Table 2 provides a list of the silicon producers in 2005, with 2005 actual
capacity and 2010 projected capacity. The technology each company uses
for polysilicon production is also listed. In the 2010 production column,
we give two estimates, projected and potential capacity. In an effort
to give a realistic idea of future capacity we evaluated each capacity
Future Capacity expansion plan based on four criteria. Projects that met these criteria were
considered likely to occur and thus included in the projected capacity.
Projected = Probable Potential capacity includes projects that do not meet these criteria and
thus are unlikely to occur. The criteria were used were:
Potential = Possible
1. Is the company reputable in the solar or silicon space?
2. Has a technology been chosen and provided?
3. Is a funding source identified?
4. Has a timeline for project construction been provided?
14
Polysilicon Production Capacity
SECTION 3
In a few rare cases a project did not meet all of these criteria, but it was
our opinion that the project would be completed and thus was included
in the projected capacity. As table 2 shows, we believe existing producers
will achieve the production capacities they announced.
The annual estimates produced throughout include not only partial year
results, but ramp up periods as appropriate. We believe that additional
positive news (such as marginal productivity improvements) and negative
news (construction delays and reductions of new plant sizes) will likely
occur, but these projections reflect our best estimates of the future
capacity at this time.
Table 2: Silicon producer capacity in 2005, projected and potential in 2010, and
technology employed (* unconfirmed, assumed technology)
2010 Production capacity Technology
2005 Production Capacity
Company (MT) (Siemens, FBR,
(MT)
(projected/potential) VLD, MG to SoG)
Hemlock 7,700 19,000/-- Siemens
Wacker 5,500 14,500/-- Siemens/FBR (trial)
REC 5,300 13,500/-- Siemens/FBR
Tokuyama 5,200 5,400/-- Siemens/VLD (trial)
MEMC 3,800 8,000/-- FBR/Siemens
Mitsubishi 2,850 3,150/-- Siemens
Sumitomo 800 1,300/-- Siemens
Total China 130 7,300/-- Siemens*
Total 31,280 72,150/--
Hemlock
After producing polysilicon for several years, Dow Corning formed Hemlock Hemlock
Semiconductor Corporation, located in Hemlock, Michigan, as a wholly-
owned subsidiary. In 1984 Hemlock became a joint venture between three Michigan, US (Siemens)
companies, Dow Corning (63 percent ownership), Shin-Etsu Handotai Co.,
Ltd. (25 percent) and Mitsubishi Materials Corporation (12 percent). Hemlock 2005: 7,700 MT
rapidly increased production capacity, and by 1994 it became the largest 2010: 19,000 MT
polysilicon supplier in the world. Hemlock is still the largest polysilicon
producer, and will likely continue to be through 2010 based on stated and
likely capacity additions.
Hemlock uses Siemens reactors and trichlorosilane gas to produce chunk

polysilicon at two plants. The company has expressed an interest in developing
granular polysilicon depending on the market. Hemlock’s raw materials come
from one of Hemlock’s parent companies, Dow Corning, which has silicon
mining and processing operations in South America and the United States.
15
In 2005, the company reported a production of 7,700 MT. It will increase

production to 10,000 MT in 2006, with an investment of $400-$500
million.3 By 2008 the company will have 14,500 MT of capacity and by
2009, 19,000 MT. Hemlock is expanding capacity at its Michigan facility,
but is contemplating other locations for some of the additional capacity.
One news report quoted a company representative as discounting
Taiwan as a potential location due to political uncertainty. A location in
Asia might be attractive to the company, considering the huge market
potential there for solar energy and the fact that of Hemlock’s current
production is exclusively in the United States. Approximately 40 percent
of Hemlock’s customers in 2005 were from the PV industry.
Hemlock 2005 2006 2007 2008 2009 2010

Capacity(MT) 7,700 10,000 10,000 14,500 19,000 19,000
Wacker
Wacker At the turn of the 20th century, Dr. Alexander Wacker founded “Dr.
Alexander Wacker Gesellschaft für elektrochemische Industrie KG” in
Traunstein, Germany. This consortium moved to Munich two years later
Burghausen, Germany (Siemens)
and started Wacker-Chemie GmbH.
2005: 5,500 MT In 1916, the Burghausen facility, now the company’s largest production
2010: 14,500 MT plant, came online. With several hundred employees, the operation
produced acetaldehyde, acetone, and acetic acid. Wacker now has five
business segments: semiconductors (Siltronic), silicones, polymers, fine
chemicals, and polysilicon. The company has 20 production plants, 14,400
employees, and over 100 Wacker sales offices worldwide.
The polysilicon segment of Wacker experienced a 12 percent increase in

sales in 2005 over 2004, slightly higher than the overall company sales
growth. In 2005 polysilicon sales reached €288.1 million. The company
credits the booming PV market as the main reason for its increase in
polysilicon sales. Wacker also disclosed that its facility ran at full capacity
in 2005. Though the company did not disclose how much polysilicon it
had in inventory from the previous year, it did indicate that its stock
helped the company meet some of the 2005 demand.
Wacker’s polysilicon production capacity was 5,500 MT in 2005. It will

expand capacity to 6,500 MT in 2007. Wacker originally announced
expanding capacity to 9,000 MT by 2009, but has since revised that estimate
to 10,000 MT. A company representative attributes this extra capacity to
progress the company expects to make in de-bottlenecking the production
process. Wacker will add another 4,500 MT by 2010. Like Hemlock, 40
percent of Wacker’s polysilicon was sold to the solar industry.4
Wacker 2005 2006 2007 2008 2009 2010

Capacity(MT) 5,500 5,500 6,500 10,000 10,000 14,500
16
SECTION 3
REC
Renewable Energy Corporation, with headquarters in Norway, was first REC
incorporated as a private company in 1996 with a mission to invest in
renewables.5 The company is now one of the few fully-integrated PV Montana, US (Siemens)
companies in the world, with full or part ownership of subsidiaries
throughout the PV supply chain. The company has a polysilicon 2005: 2,800 MT
division, REC Silicon; an ingot division, REC SiTech; a wafering division, 2010: 4,000 MT (est)
ScanWafer; a cell division, ScanCell; a module division, ScanModule;
and an integrator/distributor division (for installations in developing Washington, US (Siemens)
countries), Solar Vision. REC also has a 23 percent stake in CSG Solar and
a 33.3 percent stake in EverQ. In May of 2006, REC listed on the Oslo 2005: 2,300 MT
Stock Exchange. 2010: 3,000 MT (est)
Union Carbide Corp started a polysilicon business in Moses Lake, WA Washington, US (FBR)
in 1984.6 Initial production was 1,000 MT/yr, expanded to 1,400 MT of
capacity in 1987. This was followed by another expansion to 2,100 MT 2005: pilot
in 1996. Komatsu bought the facility in 1990, changing the name to 2010: 6,500 MT
Advanced Silicon Materials Inc (ASiMi). In 2002, REC and ASiMi entered
into a joint venture, converting the company to the first silicon producer
dedicated to the solar industry and changing its name to Solar Grade
Silicon LLC. REC assumed full ownership of SGS in 2005.7 SGS uses the
Siemens process and TCS to produce polysilicon, which the company
reports to be 99.999999 percent (8N) pure.8
In 2005, REC purchased a polysilicon plant in Butte, MT, formerly owned

by Advanced Silicon Materials Inc. This facility uses monosilane gas to
produce polysilicon in a Siemens reactor.
REC has been running a fluidized bed reactor which it is now ready to
bring to commercial scale. On May 23, 2006, the company approved a
plan to invest $600 million for the construction of a 6,500 MT FBR plant
on the Moses Lake site. REC signed a contract with the Fluor Corporation
to build the new plant in June and celebrated groundbreaking of the
project in mid-August.9
REC produced 5,300 MT in 2005, 2,500 MT of which was solar grade. The
6,500 MT FBR plant will be online in 2008, and fully operational in 2009.
With this additional capacity and de-bottlenecking at existing facilities,
REC will have over 13,000 MT of capacity by 2010. The company believes
that it supplies 20-25 percent of polysilicon used by the PV industry.10
REC 2005 2006 2007 2008 2009 2010

Capacity(MT) 5,300 5,800 5,800 10,250 13,500 13,500
REC supplies Evergreen and EverQ with polysilicon. REC’s vision is “[t]o
become the most cost-efficient solar energy company in the world, with
a presence throughout the value chain.”11 The company has a target to
reduce costs (per watt of module produced) by 50 percent by 2010.12
17
Tokuyama
Tokuyama Tokuyama was started in 1918 under the name Nihon Soda Co. Ltd., which
was changed to Tokuyama Soda Co., Ltd in 1939. The focus of the company
Yamaguchi, Japan (Siemens) was soda ash until 1938, when it moved into the cement business.
2005: 5,200 MT Over the years, the company has expanded operations into three business
segments (chemicals, building materials, and specialty products) that serve
2010: 5,200 MT
various industries. Tokuyama has offices and factories throughout Japan,
Yamaguchi, Japan (VLD) as well as offices in the United States (California), Germany (Düsseldorf),
Singapore, etc. In its 2005 annual report, Tokuyama identified the silicon
2005: 200 MT component of their business as having the most potential for growth and
2010: N/A profitability in the future.
In September 2004, Tokuyama announced its plans to expand production

of polycrystalline silicon. Citing the growth in solar energy, in addition
to the rebounding semiconductor industry, as the reasons for dwindling
inventory, the company increased capacity from 4,800 MT to 5,200 MT
at its Higashi plant. They also have plans for a 200 MT verification plant
for vapor-to-liquid deposition (VLD). VLD technology allows for faster
production and produces a product more appropriate for PV applications
(quality is inferior for semiconductor applications).
Construction of the new VLD plant was initiated in February 2005. The
Tokuyama Factory is expected to start shipping polycrystalline silicon
in 2006. The New Energy and Industrial Technology Development
Organization (NEDO) in Japan offered funding for the verification plant.
If successful, a commercial plant should be operational by 2008. The
company believes VLD could put it in position to become the world’s
second largest supplier of polycrystalline silicon. However, when
contacted, the company declined to provide estimates for production
capacity aside from existing and pilot-scale capacity. Evidently the
company is still trying to perfect the VLD process, and will need to make
significant progress before it can bring VLD to commercial-scale.
Tokuyama 2005 2006 2007 2008 2009 2010

Capacity(MT) 5,200 5,400 5,400 5,400 5,400 5,400
MEMC
MEMC
MEMC (Monsanto Electronic Materials Company) was founded in 1959
Texas, US (FBR) in St. Peters, Missouri as a division of Monsanto Chemical Company. The
focus of the company was to produce silicon wafers (19mm diameter) for
2005: 2,700 MT the transistor and rectifier industries. Today, the target markets for the
2010: 6,000 MT (est) company’s silicon wafers are the electronics and PV industries.
Merano, Italy (Siemens) The German company Hüls, AG (a VEBA AG subsidiary) assumed ownership
of MEMC in 1989, and then consolidated the company with an Italian
2005: 1,100 MT company, Dynamit Nobel Silicon Holdings. The consolidated company
2010: 2,000 MT (est) was renamed MEMC Electronics Materials, Incorporated.
18
SECTION 3
MEMC produces polysilicon feedstock usable in PV cells at two factories,

one in Pasadena, Texas and the other in Merano, Italy. MEMC is the only
company producing granular polysilicon at industrial scale, though
both Renewable Energy Corporation (REC) and Wacker have pilot plants
producing this form of purified silicon. Some cell companies prefer
granular polysilicon, including Evergreen Solar for its String Ribbon
technology. This puts MEMC at a competitive advantage for this product,
for now. MEMC also produces chunk polysilicon.
Annual polysilicon production capacity at MEMC was 3,800 MT in 2005,

or 12 percent of the global total. The Merano plant’s capacity was 1,100
MT in 2005, while the Pasadena plant’s capacity was 2,700 MT. This is
expected to remain flat for 2006. In 2007, however, the capacity at the
Merano Italy plant will grow to 1,600 MT for a company-wide total of
4,300 MT annual production capacity. By 2008, MEMC plans to almost
double its annual production capacity to 8,000 MT between the two
existing plants. The company may also build a third plant; the proposed
location has not yet been announced.
MEMC 2005 2006 2007 2008 2009 2010

Capacity(MT) 3,800 3,800 4,300 8,000 8,000 8,000
Mitsubishi
Mitsubishi
Alabama, US (Siemens)
Mitsubishi Materials Corporation holds 100 percent ownership of two
polysilicon companies, Mitsubishi Polycrystalline Silicon America Corporation 2005: 1,250 MT
and Mitsubishi Materials Polycrystalline Silicon Co. Mitsubishi Polycrystalline 2010: 1,550 MT
Silicon America Corporation is located in Mobile, Alabama, with a capacity
of 1,250 MT. Construction is underway to build an additional 300 MT at this Yokkaichi, Japan (Siemens)
site, the only expansion the company has publicly announced. Mitsubishi
Materials Polycrystalline Silicon Co., located in Yokkaichi, Japan, has a 2005: 1,600 MT
capacity of 1,600 MT, with no publicly-known plans for expansion. 2010: 1,600 MT
Mitsubishi 2005 2006 2007 2008 2009 2010

Capacity(MT) 2,850 2,850 3,150 3,150 3,150 3,150
Sumitomo Sumitomo
Sumitomo Titanium Corporation, located in Japan, is an affiliated company
of Sumitomo Metal Industries and Kobe Steel. It caters to the electronics Japan (Siemens)
industry, producing only EG polysilicon. In 2005, its production capacity
2005: 800 MT
was 800 MT. It will have 900 MT of capacity in 2006. In early 2006, it
announced that would expand capacity by another 400 MT to respond to 2010: 1,300 MT
demand from the electronics industry.
Sumitomo 2005 2006 2007 2008 2009 2010

Capacity(MT) 800 900 1,100 1,300 1,300 1,300
19
3.5 Polysilicon Recyclers − IC Castoffs

Several thousand metric tons of reclaimed silicon were recycled in 2005.
While some cell companies have in-house recycling divisions, there are
a few companies that specialize in this type of silicon processing. A rule
of thumb is that 10 percent of the silicon sent to the semiconductor
industry eventually makes its way to the PV industry. This number will
likely increase in the next few years as the gap between silicon supply
“Recycling of silicon and demand increases, and as companies improve their recycling abilities.
Here we provide a sample of silicon recyclers.
will be more important
Silicon Recycling Services
in the next few years
Silicon Recycling Services (SRS) recycles off-spec and waste silicon. 13 SRS
as a way of lowering has facilities in Camarillo, CA and Beijing, China. In 2005, SRS provided
700 MT of silicon on the spot market, 99 percent of which was sold to the
material costs, and PV industry. In addition, the company was contracted to recycle several
increasing usable hundred metric tons worth of silicon for companies that used the silicon
in-house. In early 2006, ErSol AG, a German cell producer, purchased SRS
silicon supply.” for an undisclosed amount.
ReneSola
ReneSola started recycling polysilicon in July 2005 in Zhejiang province,
but has already made quite a mark on the solar industry. ReneSola recycles
scrap silicon and broken wafers primarily from the semiconductor industry,
but also from the PV industry. It has 54 monocrystalline furnaces, each
of which produces 0.5 MT of silicon ingots per month. ReneSola’s wafer
capacity is 48 MW (as of August 2006), which it hopes to expand to 80
MW by the end of 2006 and 125 MW in 2007. The company uses the
labor of 600 employees to hand-sort the scrap material for recycling.
The company estimates that over 2,000 metric tons sold to PV companies
was recycled material. The company believes there is even more available
to recycle annually (on the order of 5,000 MT), in addition to scraps
from past production. ReneSola has signed wafer supply contracts with
Jiangsu Linyang Solarfun Co. and Motech. It has also signed feedstock
supply contracts with Holy Technology Corporation and Komex Inc, for
the delivery of scrap silicon.14
Poseidon Chemical
Poseidon Chemicals is a wafer recovery and processing company in Chennai,
India. It processes wafer discarded by the semiconductor industry. Germany-
based Solar-Fabrik signed a letter of intent in May 2006 to hold 80 percent
of the shares of Poseidon Chemicals at a price of €3 million. The pending
acquisition will turn Solar-Fabrik, a module manufacturer into an integrated
solar company. This wafering capacity will be a welcome addition to Solar-
Fabrik’s module production, which did not operate at full capacity in the
beginning of 2006 due to insufficient cell supply. Excess wafers not used by
the company will be sold to the market. The recycling capacity of Poseidon
Chemicals was not available at the time this report was printed.
20
SECTION 3
SolarWorld
the cell companies that have in-house recycling capability, SolarWorld
Of
is worth mentioning due to the large scale of its operations. On June 13th
SolarWorld’s subsidiary Deutsche Solar celebrated the company’s nearly
completed second materials recycling facility in Freiberg, Germany. By
the fourth quarter of 2006 the two plants will be capable of processing
a combined total of 1,200 MT of silicon, or 40 percent of SolarWorld’s
feedstock needs. Half of the material recycled comes from internal
operations of the company’s wafering activities, while the other half is “Recycling is important
obtained via outside sources, including the semiconductor industry. The
company also plans to provide recycling services to other companies in
to maintaining PV’s
the fourth quarter of 2006.15 environmentally-
oriented image.”
Recycling of silicon will be more important in the next few years as a way
of lowering material costs, and increasing usable silicon supply. It is also
important in maintaining PV’s environmentally-oriented image. In this
section we presented a historical context for polysilicon production, with
a detailed description of the top seven current polysilicon producers.
Next, we will discuss the companies that have stated plans to enter the
market in the next few years.
1 Price estimates provided by Mike Rogol

2 Hemlock and Mitsubishi presentations at Solar Silicon Conference
3 www.dowcorning.com November 15, 2005 press release. “Demand in Solar Energy
Industry Drives $400 million Hemlock Semiconductor Expansion.”
4 www.wacker.com “Preparing Ourselves for the Photovoltaic Boom: An Interview with
Ewald Schindlbeck, President of WACKER POLYSILICON.”
5 http://www.scanwafer.com/default.asp?V_ITEM_ID=488
6 Grant Country EDC, December 5, 2005.
8 REC 2004 annual report.
9 http://hugin.info/136555/R/1067341/180538.pdf; Company Press release, August 16, 2006.
10 Company Press release, August 16, 2006.
12 http://hugin.info/136555/R/1067341/180538.pdf
13 Prometheus Institute, PVNews, April 2006
14 http://www.renesola.com/download/FirstDayDealings.pdf
15 http://www.solarworld.de/
21
4 Emerging Polysilicon Producers
and Supplemental Silicon
Table 3 shows the new silicon producers with their production capacity
goals and technology. The most striking piece of information in Table 3
when compared with Table 2 is the greater number of companies planning
to employ direct MG to SoG technologies as way to reduce the cost of
silicon for solar applications. We used the same criteria outlined in section
3.4 for projected and potential capacity for the companies listed in Table 3.
Table 3: New silicon producer capacity in 2005, projected and potential in 2010, and
technology employed (* unconfirmed, but assumed technology)
2010 Production capacity Technology

2005 Production
Company (MT) (Siemens, FBR, VLD,
Capacity (MT)
(projected/potential) MG to SoG)
DC Chemical 3,000/-- Siemens

Hoku 1,500/-- Siemens
Isofoton et al. 2,500/-- Siemens*
French Consortium 2,000/3,000 Siemens*
Crystal 1,200/-- Siemens
Russia/FSU --/14,500 Siemens*
M.Setek 6,000/-- Siemens*
Elkem 5,000/10,000 MG to SoG
JSSI 850/-- Silane to SoG
ARISE --/2,000 MG to SoG*
JFE Steel 100/-- MG to SoG
SolarValue 5,300/10,000 MG to SoG
Global PV Specialists --/2,000 MG to SoG
Total China 130 7,300/22,100 Siemens*
Total 130 34,750/63,600
4.1 New Producers using

Conventional Silicon Processing
This section describes companies that have not yet entered the polysilicon
production business, but have stated plans to do so using production
technologies that are well-established.
DC Chemical DC Chemical —
High-efficiency cell-maker SunPower signed an agreement to provide chemical-handling
advance payments to Korea’s DC Chemical (DCC) for the construction of
a 3,000 MT polysilicon facility. This project marks DCC’s first venture into expert finds strong
the polysilicon business. The DCC will employ Siemens technology and
use TCS as the feedstock gas. In all, SunPower will pay $250 million in partner with SunPower
the multi-year agreement, though the total amount of silicon DCC will
deliver has not been disclosed.
23
SECTION 4 Emerging Polysilicon Producers and
Supplemental Silicon
Hoku Scientific
Hoku Scientific, a fuel cell company located in Hawaii, announced in
Hoku — fuel cell May 2006 that it will enter the solar business in an effort to diversify.1
The company has plans to build a 1,500 MT Siemens-based polysilicon
company enters PV production facility as well as a 30 MW module plant. Excess polysilicon
(approximately 1,200 MT per year) will be sold to the electronics and solar
industry as vertically- industries. Hoku will invest $250 million in the project, some of which may be
in the form of up-front payments from other solar companies for contracts
integrated business to secure feedstock material. Hoku first indicated that the location would
likely be in Singapore for several reasons, including government support of
new energy technologies, its proximity to Hoku’s potential customers, and
political stability. However, Hoku later announced a decision to locate the
polysilicon and module facilities in the State of Idaho. The exact location
in Idaho has yet to be determined; the company is still considering three
locations along the Snake River. CH2M Hill Lockwood Greene will design
Isofoton et al. — Spanish and construct the polysilicon facility.
group has no previous Isofoton, Spanish government and Endesa

poly experience, but A joint project between Spanish cell producer Isofoton, an Andalusian
government agency (the Department of Innovation, Science and
otherwise is the right Business), and Endesa, a Spanish utility, will result in the first polysilicon
plant in Spain.2 The 2,500 MT facility will be built in Los Barrios. This
combination of players has exciting implications for the PV industry, as the plant will supply
much needed solar grade silicon and is located in a region poised to be
in a promising market the next big market for solar installations. The utility’s involvement is
particularly interesting considering the huge costs associated with the
energy required to produce polysilicon.
French Consortium (Econcern, NorSun and

Photon Power Technologies)
French Consortium — In an August press release, Econcern, a sustainable energy holding
REC founder teams up company, announced that it, along with two partners, is planning to build a
polysilicon plant dedicated to the solar industry. The two other companies
to start a poly plant in are NorSun, a Norwegian wafer company started by Renewable Energy
Corporation founder Dr. Alf Bjorseth, and Photon Power Technologies of
southern France France. The consortium is considering Saint Auban, in the southeast of
France, for the location of the 2,000 to 3,000 MT facility, which is slated
to be operational in 2008.3
Crystal/Russia and Former Soviet Union Companies

According to Dr. Lebedev of Swiss Wafers, Russia and countries of the Former
RU/FSU — region still Soviet Union have the potential to produce upwards of 14,500 MT provided
poised to take off, but the necessary investment funds are secured. He believes that it is likely that
3,000 MT of capacity will be online in this region by 2009.4 We found one
progress is slow company with firm plans to produce polysilicon over the next few years,
Crystal. Located in Kyrgyzstan, Crystal did not produce polysilicon in 2005,
but anticipates 60 MT in 2006. The company’s goal is 1,200 MT by 2008.
24
Emerging Polysilicon Producers and
SECTION 4
RU/FSU 2005 2006 2007 2008 2009 2010

Capacity (MT) 60 660 1,200 1,200 1,200
Potential Capacity (MT) 14,500
M.Setek M.Setek — Japanese

M.Setek, a Japanese silicon-wafer company, is adding polysilicon
production to its business operations. M.Setek has not provided many
silicon wafer expert
details on this new area for the company, other than the below capacity
projections. We have not confirmed that M.Setek will be using the
builds capacity
Siemens process, but we assume it will. upstream
M.Setek 2005 2006 2007 2008 2009 2010
Capacity (MT) 1,100 3,000 3,000 6,000
Chinese Companies
With many of the top Chinese and Taiwanese cell manufacturers having
aggressive expansion plans, polysilicon production within the region is an
important consideration. There are no existing polysilicon manufacturers
in Taiwan. However, Taiwan’s Bureau of Energy is seeking approval from
the legislature to allocate $6.2 - $9.2 million in 2007 to the advancement
of domestic silicon production technologies.
In 2005, two Chinese polysilicon companies, Emei Semiconductor Material,

and Luoyang Monocrystal Silicon, produced approximately 130 MT.5 At the
end of 2005, Luoyang Zhonggui, a joint venture company in which Luoyang
Monocrystal Silicon is involved, came online with a production capacity of
300 MT. Once this company was in production, Luoyang Monocrystal Silicon
closed.6 Emei and Luoyang Zhonggui will have a combined capacity of 400
MT in 2006. In addition, three companies (Aostar, Luoyang, and Shangxin
Silicon) are in the construction phase and will
come online between 2007 and 2008. Three more Figure 20: Chinese producers 2005 - 20108
companies (Linghai, Shizuishan, and Jinyi Silicon) are
8,000
in the planning stage. CSG Holdings 7,300
7,000 Jinyi Silicon
On May 23rd, CSG (China Southern Glass) Holding of Shizuishan
China announced that it will invest $150 million to build 6,000
Production (MT)
Linghai
a polysilicon production facility in Hubei Province. The Luoyang
company plans to eventually build capacity to 4,000- 5,000
Shangxin Silicon
5,000 MT annual production. Although a construction 4,000 Xinguang
time of 18 months was provided, the start date for the Luoyang Zhonggui
project was not disclosed. It is possible that this plant 3,000 2,800 2,800
Luoyang Monocrystal
could come online by mid-2008. The process CSG Emei
2,000 1,800
will be using to produce the polysilicon is reportedly
based on technology developed at a Russian research 1,000
institute. CSG Holding’s core business since its 400
130
inception in 1984 has been glass manufacturing. In 0
2005 2006 2007 2008 2009 2010
1992, it became one of the first Chinese companies to
list on the Shenzhen Stock Exchange.7 Year
25
By 2010, these nine companies in China could produce over 7,000 MT

Chinese companies — of polysilicon, far short of the anticipated demand for silicon in China
and Taiwan over that time period. Seven thousand metric tons may be
the wild card for future a significant underestimate of the potential supply from China. A June
14th news article claimed that within the next three to five years, annual
supply anywhere from production in China could be over 20,000 MT. The estimate was based
on “public data.” The large projects mentioned in the article include a
7,000 to 22,000 MT letter of intent signed between Jiangsu Sunshine, the Ningxia Dongfang
Nonferrous Metal Group, and Ningxia Yinlite Electronics Group to build
a 4,000 MT plant, as well as an agreement signed by Nanyang Bulk and
Yinchang City (Huibei Province) for a 4,500 MT polysilicon plant. As many
of these projects are still in the planning stage, whether they will actually
come online is still in question. Furthermore, the quality of feedstock is
also a concern.
China 2005 2006 2007 2008 2009 2010

Capacity (MT) 130 400 1,800 2,800 2,800 7,300
Potential Capacity (MT) 130 400 1,800 7,800 8,800 22,100
Other rumored projects

There have been several news articles that suggest still more developments
in polysilicon capacity. In a recent article in the Oregonian, state officials
disclosed that the Oregon has been in talks with a Japanese company to
build a polysilicon factory in the state. The $500 million project could
start in 2007, with production commencing two years later. The same
company may also build a module manufacturing plant, and two other
solar panel companies have expressed interest in locating facilities in
Oregon as well. The companies have visited manufacturing facilities that
Komatsu and Sumco have left vacant.
EEtimes online reported that MEMC and Hong Kong Specialty Gases are
rumored to be considering a partnership to build silane gas and polysilicon
capabilities in China.
All of the companies in this section are planning to use well-understood

technologies (primarily Siemens) for polysilicon production. In the next
section, we turn to companies using new or not-yet-commercial silicon
production methods.
4.2 Alternative Silicon Production

The majority of companies that have decided to pursue alternative silicon
production routes are already involved in silicon processing at some level.
Therefore, it is a logical step for them to build on this experience by further
refining their product instead of using traditional processing techniques.
Elkem, Joint Solar Silicon, ARISE, SolarValue, Global PV Specialists, JFE
Steel, GiraSolar, and Dow Corning have all been investigating alternative
silicon processing methods.
26
SECTION 4
Elkem
The Norwegian Elkem Solar, a division of Elkem AS, is one of the most Elkem — MG
promising of the new polysilicon companies. Elkem AS produces the
precursor feedstock for polysilicon production, silicon metal. The company powerhouse could offer
claims to have 50 percent of the silicon metal market. Elkem Solar has been
developing a new process for solar grade polysilicon production. Currently in cheap Si if successful
pilot scale production, the company expects to be producing on a commercial
scale by 2007. Initial capacity will be 2,500 MT, ramped up to 5,000 MT in in commercial-scale
2008. A further 5,000 MT of capacity could be added in 2010. Like much of
the new capacity coming online, initial production is already slated to go to
operations
companies that have supported the development of this project.
Elkem’s contribution to the future silicon supply is important for two

reasons. First, it will constitute a large percentage (~9.8 percent) of the total
production capacity available by 2010. The scale of expansion is enormous,
and could make Elkem the fourth largest polysilicon producer in the world,
after Hemlock, Wacker, and REC. Second, the silicon will be solar grade,
and thus presumably lower in cost, though questions remain about the
product’s usefulness as a stand-alone silicon replacement. Some blending
with traditional silicon may be required. Elkem meets all of our requirements
for projected capacity, but given that it will use a new process, we have
projected that it can meet its initial goal of 5,000 MT. If the company
reaches this goal we will include the next 5,000 capacity expansion it plans
for 2010. We took this approach with Elkem and SolarValue (see below) to
avoid skewing the 2010 capacity total, in the event that the commercial
scale production at either company is delayed. With any company that is
employing MG refining to produce SoG, the time between confirmation of
industrial production and subsequent capacity expansions is expected to
be very short – especially in comparison to building capacity with Siemens
reactors which takes approximately two years.
Elkem 2005 2006 2007 2008 2009 2010

Capacity (MT) 2,500 5,000 5,000 5,000
Potential Capacity (MT) 2,500 5,000 5,000 10,000
Joint Solar Silicon GMBH & Co. KG (JSSI) JSSI — solid

In July 2006, SolarWorld announced that its joint venture with chemical
giant Degussa for a polysilicon production process would proceed partnership between
to commercial scale production. The JV is currently operating a pilot
plant, the product of which goes directly to SolarWorld’s wafer division,
chemical company
Deutsche Solar. The new plant, which will begin producing polysilicon in Degussa and vertically
2008, will have a production capacity of 850 MT for 10 years. SolarWorld
currently procures silicon from Wacker and Hemlock. integrated PV company
JSSI 2005 2006 2007 2008 2009 2010

Capacity (MT) 850 850 850
27
ARISE
Until relatively recently, Canada-based ARISE Technologies concentrated
its efforts on the downstream end of the supply chain. This year, ARISE
announced plans to integrate upstream by adding polysilicon and cell
ARISE — little manufacturing capabilities. The company will receive $6.5 million (CAD)
from Sustainable Development Technology Canada (SDTC) for use towards
experience, but strong development of a proprietary solar grade polysilicon production process.
SDTC was created by the Government of Canada to provide support to
collaborators clean technology projects. The money committed to ARISE by SDTC will be
added to the $13.3 million of funding in the form of cash and donations
from a consortium of partners. The SDTC funding will be spread out over
three years, pending final agreement and the achievement of project
benchmarks. ARISE is collaborating on this project with the University
of Toronto, University of Waterloo, Topsil Semiconductor Materials A/S,
and others.
A 200 MT pilot-scale facility will begin producing polysilicon in 2007. If

everything goes according to plan, the company expects to have a 2000
MT plant operational by 2010.
ARISE 2005 2006 2007 2008 2009 2010

Potential Capacity (MT) 50 200 200 2,000
JFE Steel
JFE Steel, established in 2003 and headquartered in Japan, is a company
JFE Steel — Japanese within the JFE Group. In July 2006, JFE Steel announced that it started
construction of a 100 MT solar grade silicon plant. With help from NEDO
wafer company with (New Energy and Industrial Technology Development Organization), JFE
Steel launched an R&D project in 2001 to develop SoG from metallurgical
novel process grade silicon (figure 25). The company revealed that it is in the process of
designing a larger facility.
SolarValue
The solar start-up company SolarValue is in the final stages of purchasing
TDR Metalurgija, a metallurgical silicon company in Ruše, Slovenia. The
purchase will secure feedstock for SolarValue for the production of
solar grade silicon. SolarValue was started in 2005 in Germany and was
SolarValue — new to listed on the Frankfurt Stock Exchange in September 2006. The company
formed a subsidiary in Slovenia, SolarValue Productions d.d., that will
silicon processing, but use a four-step refining process to convert metallurgical silicon to solar
process has been verified grade silicon. The company hopes to produce 2,000 MT in 2007 and 5,300
MT in 2008. After fine-tuning its operations, SolarValue hopes to produce
10,000 MT per year by 2010. As with Elkem, we have not included the
additional 5,000 MT of additional capacity in 2010 in an effort to remain
conservative in our estimates. If the company is able to meet capacity
goals on time, we will revise the current estimates.
28
SECTION 4
SolarValue 2005 2006 2007 2008 2009 2010

Capacity (MT) 2,200 5,300 5,300 5,300
Global PV Specialists
Global PV Specialists is a consulting firm based in California that provides
Global PV Specialists
turnkey factories for cell manufacturing as well as other services — unique process using
pertaining to the operation of the facilities. In an effort to support
companies building new cell factories, Global PV Specialists will begin rice hulls
producing silicon specifically for its own customers. The technology
it will use includes a unique approach with rice hulls as the feedstock
material. The final location for the production facility has not been
decided, though it will likely be built in the US.
Global PV Specialists 2005 2006 2007 2008 2009 2010

GiraSolar
GiraSolar, located in the Netherlands, is an umbrella company for solar
energy companies. It has subsidiaries, partners, and affiliates performing
research and development, as well as doing business at several levels of the
supply chain. GiraSolar serves as an investment and managing company for GiraSolar —
the group, allowing the companies to remain independent yet benefit from
their association with the group. Scientists at GiraSolar have announced they in research phase
have applied for a patent for a proprietary solar grade silicon production
process. The company anticipates production costs for the material to be
$10-$12/kg, less than half of current production costs of conventional
silicon. The company CEO, Wieland Koornstra, indicated that the project is
still in the research phase, but once the patent goes through they expect
to build a pilot-scale facility. This project is still a long way from producing
usable silicon; hence there are no capacity estimates at this time.
Dow Corning
With the PV industry in the midst of a silicon shortage, Dow Corning
believes its new solar-grade silicon (SoG) offers a unique solution to
the problem. The product, called PV 1101, derives from the purification
of metallurgical silicon, and is used in a blend with purer silicon to
manufacture flat plate silicon solar cells. Dow Corning is supplying a few
PV companies with the product to test the new material in their own
production processes. The undisclosed companies have blended PV 1101
into their silicon feedstock at ratios of 10 percent or more. Due to non-
disclosure agreements, Dow Corning refused to provide details on specific
blend ratios and resulting cell efficiencies, but it did disclose that no
reduction in performance was reported by its customers.
29
Current production capacity of PV 1101 is around 1,000 MT, but the

Dow Corning — Si company could ramp up to large scale production quickly in response
to demand. Dow Corning is producing the product exclusively at its
giant with stake in new facility in Santos Dumont, Brazil. In the short term, only current
Dow Corning customers can purchase the limited quantities of PV 1101
polysilicon producer available. While the availability of public information on this new product
is limited, it appears that this product could serve to add feedstock to the
Hemlock produces first currently tight global silicon supply.
commercially available
SoG but it must be As this overview of current and future polysilicon producers shows,
blended with pure poly production is becoming more distributed throughout the world. We
examine the current and projected regional breakdown of polysilicon
supply in the next section.
1 http://www.shareholder.hokuscientific.com/releasedetail.cfm?ReleaseID=198969/
2 www.solarbuzz.com/
3 http://www.econcern.com/cms/download/PressRelease_Econcern_SildeProv_300806_
EN.pdf/
4 Lebedev, Alexander, Revival of Polysilicon Production in Countries of the Former Soviet
Union? 3rd Solar Silicon Conference, April 2006, Munich, Germany
5 Yang, Deren, Overview of Silicon from Chinese Manufacturers, 3rd Solar Silicon Conference,
April 2006, Munich, Germany
6 Deren Yang, personal communication
7 http://www.csgholding.com/en/about/instruction.asp/
8 All capacity projections except for CSG were provided by Deren Yang (see endnote #5)
30
5 Geography
of Global Polysilicon Capacity
In 2005, the US dominated silicon production, providing over 50 percent

of the product available on the market. Japan and Europe (mostly
Germany) split the remaining production amount. In Figure 26 we show
the geographic distribution of polysilicon production in 2005 and 2010.
For 2010, we once again use the criteria outlined in section 3.4.
Figure 21: Map of production

Production Capacity (MT)
50,000
40,000
30,000
20,000
10,000
0 * * *
US Japan Europe China Russia/FSU RoW
2005 2010 projected 2010 potential
The US share of production is expected to fall over the next few years as
production increases in Europe (particularly in Norway) and China (Table
4). The Japanese companies are not keeping pace with their US and Europe
counterparts; however, this could change. The Japanese companies have
been slow to respond to growing demand thus far, but may announce
new expansions in the future.
The countries that have the potential to play a much larger Table 4: Geographic distribution of polysilicon
role in polysilicon supply than they currently do are China production as a percent of total production
and the FSU countries. If all of the potential capacity we 2010 2010
report actually comes online, we expect the US and Japan to Country/Region 2005
(projected) (potential)
lose a significant amount of the polysilicon market. (In the
third column of table 4, we assume that all potential capacity US 54.2% 40.7% 27.6%
comes online, which affects the percentage of production Japan 24.3% 14.8% 9.6%
for all regions/countries). Europe 21.1% 32.6% 30.2%
While the geographic distribution of polysilicon is interesting China 0.4% 7.5% 18.8%
given the relatively low cost of shipping silicon, it does Russia/FSU N/A 1.2% 10.4%
not necessarily dictate where future markets for PV will RoW N/A 3.1% 3.3%
develop. In the next section we examine overall polysilicon
production in the next five years and offer a forecast of PV
production based on several variables. We also discuss the assumptions of
our projections, and perform sensitivity analyses to investigate the range
of possibilities for future industry growth.
31
6 Projections and Forecast Model
The PV industry has grown an average of 40 percent per year over the last
decade. In 2005, cell production grew 46 percent over 2004. We estimate
that in 2005, the PV industry required roughly 19,000 MT of polysilicon,
only 3,000 MT less than the semiconductor industry. In 2006, polysilicon
demand from PV producers will likely be greater than demand from the
electronics industry. But with no significant new capacity available this
year, PV growth will be limited. In this section we present a forecast
model of PV production growth to 2010.
6.1 Polysilicon Production Capacity

Projections to 2010
All companies that currently produce polysilicon are expanding capacity
to varying degrees. Mitsubishi has the most modest capacity expansion
plans (300 MT), whereas Hemlock anticipates nearly 12,000 MT of new
capacity over the next five years.
Figure 22: 2005 - 2010 producers
Hemlock
Wacker
REC Year
Tokuyama 2005
2006
MEMC 2007
Mitsubishi 2008
2009
Sumitomo 2010
China
Other
0 5,000 10,000 15,000 20,000 25,000
Capacity (MT)
By 2010, the total silicon supply is expected to grow to over 97,000 MT. This
is our projected estimate, according to the criteria outlined in Section 3.4.
Year 2005 2006 2007 2008 2009 2010

Capacity (MT) 31,280 34,710 44,260 73,900 83,000 97,100
The top three companies will maintain their positions through 2010. Tokuyama,
fourth in production in 2005, is also expected to increase capacity if and when
it perfects its vapor-to-liquid deposition (VLD) process. Engineering setbacks
appear to be keeping the company from moving forward with the few
thousand metric tons of capacity it would like to build using this technology.
33
SECTION 6 Projections and Forecast Model
In making our projections, we have distinguished between companies

that have solid reputations in the industry and clear production goals,
and those firms that are less well-known with unclear or questionable
plans. We are aware that there may be new announcements from existing
producers as well as new entrants. But we also expect that not all of
the announced expansion plans will be brought to completion on time
or indeed at all. For example, there has been very little activity from
the Japanese producers. One could speculate that expansion plans at
Tokuyama and Mitsubishi are under consideration but not made public.
However, a counterargument is that perhaps they are not as committed
to the PV industry for future sales as, for instance, Hemlock or Wacker.
Furthermore, some companies may worry that with all of the expansions
by existing companies and the spate of new companies, there will be
an oversupply. Therefore, our estimates represent a probable outcome of
production over the next few years.
6.2 Changing Market Dynamics for

Polysilicon Producers
In 2005, Hemlock was the largest polysilicon producer and accounted for
approximately 25 percent of the global supply.
While some companies had pilot plants

Figure 23: 2005 polysilicon producer market share in operation during 2005 (i.e. Elkem, JSSI,
2010 etc), the amount produced is not provided
97,100 MT here. Chinese companies produced a total
of 130 MT, or approximately 0.4 percent
20% of the global market for polysilicon.
2050 26%
31,280 MT Under the projections presented here
for capacity expansion plans, by 2010
the seven largest companies producing
3% <1%
9% silicon will lose market share to new
25% entrants. Chinese companies are expected
12% 15% to grow collectively from less than one
8% percent to 7.2 percent market share. The
most dramatic change portrayed in this
1%
figure is the increase in market share
17% 18% 3%
for the “other” category. This category
8% 14% includes all non-Chinese producers that
17% 6% are one of the top seven producers.
These companies - which in 2005 did
Hemlock Tokuyama Sumitomo
not produce any polysilicon (aside from
Wacker MEMC China some pilot scale batches) – will capture a
REC Mitsubishi Other quarter of the market in 2010.
34
Projections and Forecast Model SECTION 6
6.3 Demand Forecast Model

Assumptions
Silicon-based cells accounted for 94 percent of production in 2005 (91 percent
were flat plate and three percent were ribbon/sheet). In our base case scenario MODEL ASSUMPTIONS (2005)
we assume thin film production reaches 500 MW by 2010 (corresponding to
a 36 percent growth rate). With so many new thin film capacity expansions Silicon Production31,280 MT
announced in the spring and summer of 2006, this is likely a conservative
estimate. However, we will give results of sensitivity analysis where we explore Polysilicon Inventory3,000 MT
the possibility of greater market penetration by thin film.
Other7,600 MT
Following Woditsche and Koch (2002),1 we assume a reduction in silicon (IC castoffs & recycling, pilot batches,
usage per watt of five percent per year. In 2005, wafer-makers required 12 & excess production)
g Si/W. By 2010, with thinner wafers and higher conversion efficiencies,
we expect silicon use to be around 9 g Si/W. This is also a conservative IC Demand22,000 MT
estimate. While high breakage rates are a concern for thinner cells, a great
deal of research and development effort is currently devoted to producing PV Demand19,800 MT
thinner cells. A target of 6 g Si/W has been promoted, and may be achieved
in the best applications, but is unlikely to be an average over the entire Thin Film Production107 MW
industry within the next few years.
For 2004 to 2008, the Semiconductor Industry Association predicts

a seven percent increase in shipments of wafers. In 2005, demand for
polysilicon from integrated circuit (IC) producers was approximately
22,000 MT, approximately 10 percent of which was assumed to have
eventually made its way to the PV industry.
A list of our model assumptions is on the upper right-hand side of this

page. For comparison we have included supply and demand estimates
from Piper Jaffray (table 5).
Table 5: Piper Jaffray solar industry production estimates2

Other
Poly Poly Residual Poly Surplus Total Poly Wafer c-Si Solar
Solar
Year Production Demand Poly for from Past Available Production Production
Production
(MT) IC / Semi Solar Productions for Solar (MW) (MW)
(MW)
2003 26,700 17,000 9,700 11,000 20,700 671 80 750
2004 28,800 19,350 9,450 11,700 21,150 1,142 114 1,256
2005 31,080 20,627 10,453 7,118 17,571 1,523 170 1,693
2006E 35,800 22,277 13,523 - 13,523 1,610 258 1,868
2007E 40,210 24,282 15,928 - 15,928 1,828 449 2,277
2008E 57,220 26,710 30,510 - 30,510 3,162 720 3,882
2009E 71,370 27,245 44,125 - 44,125 4,641 918 5,559
2010E 77,370 28,062 49,308 - 49,308 6,035 1,098 7,133
35
6.4 Model Results

Figure 24 shows our projection model through 2010. For
Figure 24: Base case model results reference, it also shows three different tracking lines for
establishing various compound growth rates for the PV
16,000 industry at 10 percent, 30 percent, and 50 percent per
PV Production (MW)
50%
Growth Trend Lines
14,000 Thin Film

annum. These tracking lines help to conceptualize the
30% market dynamics and the repercussions caused by silicon
12,000 capacity constraints through 2010.
PV Production (MW)
Silicon
10%
10,000 In 2005, we estimate there were approximately 3,000 MT
8,000 in inventory draw-downs, and another 2,000 MT recycled
from previous years’ production. This leaves over 3,000 MT
6,000 more demand from the PV industry than was thought to
be available. There are two possible explanations for this
4,000
discrepancy. First, polysilicon producers are able to produce
2,000 more than they state with production capacity estimates.
Lower purity silicon takes less time to manufacture, allowing
0 for more to be produced in a given time period. With
9,8 5
9,8 6
2,7 8
9,9 9
5,7 7
2,1 0
)
)
(2 200
)
(1 200
(1 200
(5 200
(5 200
(7 201
strong demand from the PV industry, it is very likely that

80
55
59
33
88
05
polysilicon producers modified the production process for

Year
(Silicon Supply for PV (MT)) PV applications, i.e., produced lower quality material, and
thus were able to make more material over the course of the
year. Another explanation for the discrepancy is that pilot
batches made their way to the market, but were not necessarily reported
by companies. Lastly, we also believe it likely that cell manufacturers
overstated their production, to some degree, in 2005. All of these factors
helped the PV industry report a 46 percent growth in production in 2005.
In 2006 we anticipate that there will be no excess inventory draw-downs

from 2005, but there will likely be more recycled material from previous
years as companies scour the globe for wafers discarded
Figure 25: Sensitivity test #1: 6 g Si/ W by 2010 from past production. Our estimate for recycled material in
2006 is 3,000 MT. In 2007 this will drop, perhaps 1,000 MT.
16,000 We doubt that after 2007 any remaining silicon material will
PV Production (MW)
50%
Growth Trend Lines
14,000 Thin Film

be worth the processing to make it usable. We also assume
30% that pilot production and understated silicon production
12,000 will continue at a level of more than 3,000 MT per year.
PV Production (MW)
Silicon
10%
10,000 With all considerations embedded in the model, we expect
8,000 2006 to see very modest production growth of around
five percent. This will pick up again to over 35 percent in
6,000 2007 as more production comes online and more recycling
occurs. A spike in production in 2008 due to several plant
4,000
expansions coming online will theoretically allow the
2,000 industry to grow over 100 percent that year. While many
PV companies are already expanding capacity in anticipation
0 of greater feedstock supply, a jump from 35 percent to 100
9,8 5
9,8 6
2,7 8
9,9 9
5,7 7
2,1 0
percent in one year is unlikely. What is more likely is that the

)
)
(2 200
)
(1 200
(1 200
(5 200
(5 200
(7 201
80
55
59
33
88
05
2008 production will be incorporated into 2009’s available

Year
(Silicon Supply for PV (MT))
supply, allowing average growth rates between 2008 and
36
Projections and Forecast Model SECTION 6
2010 to hover around 50 percent. By 2010, based on announced expansion

plans and assumed silicon use and thin film market share, we forecast PV
production to be just over eight GW.
With any model, certain assumptions must be made. Further Figure 26: Sensitivity test #2: 20% thin film market
analysis using sensitivity testing can show how things may share by 2010
change with a change in assumptions. In our model, the 16,000
two assumptions that may be the most controversial are
PV Production (MW)
50%
Growth Trend Lines

14,000 Thin Film
the potential growth of thin film production (we assume
30%
that thin films will account for only eight percent of 2010 12,000
PV Production (MW)
Silicon
PV production in 2010) and silicon requirements per watt 10%
(we assume that less than 9 g/W by 2010 is unlikely). 10,000
For this reason we have performed sensitivity tests with 8,000
modifications in these two variables. The results of these
tests are discussed in the next section. 6,000
4,000
6.5 Sensitivity tests 2,000
The first conservative estimate we included in our model
0
was the amount of silicon that cell producers need to
9,8 5
9,8 6
2,7 8
9,9 9
5,7 7
2,1 0
)
)
(2 200
)
(1 200
(1 200
(5 200
(5 200
(7 201
80
55
59
33
88
05
produce a watt of electricity.
Year
In our model, silicon use decreases from 12 g Si/W in 2005 (Silicon Supply for PV (MT))
to just over 9 g Si/W in 2010. In our second sensitivity test
we project average silicon use decreasing more dramatically
to 6 g Si/W. As figure 25 (page 36) shows, this allows for much greater
growth in the industry than the base case. Even in 2006, if cell producer
could decrease their silicon use to less than 11 g/W on average, the
industry could grow nearly 15 percent. By 2010, the industry would reach
just over 12 GW of production.
In our next sensitivity test we maintain our original

Figure 27: Sensitivity test #3: 20% thin film market
assumption for silicon use, but modify the amount of
share and 6 g Si/W by 2010
growth in thin film production (Figure 26). For this we
assume silicon-based production is based on the available 16,000
silicon, and thin film reaches a level equivalent to 20
PV Production (MW)
50%
Growth Trend Lines
14,000 Thin Film

percent of the overall PV market. This results in more
30%
modest improvements in industry growth, bringing total 12,000
PV Production (MW)
Silicon
production in 2010 to just less than 10 GW. 10%
10,000
The third sensitivity test we performed combined the first
8,000
two scenarios, the results of which are presented in Figure
27. If the industry were able to reduce silicon use to 6 g Si/ 6,000
W and if thin films were able to achieve 20 percent market
share by 2010, the industry could see more than 50 percent 4,000
growth after 2007 and over 15 GW of production by 2010. 2,000
Our base case model shows that a conservative estimate 0
of PV production based on projected capacities is around
9,8 5
9,8 6
2,7 8
9,9 9
5,7 7
2,1 0
)
)
(2 200
)
(1 200
(1 200
(5 200
(5 200
(7 201
80
55
59
33
88
05
eight GW. Production could be as high as 15 GW according

to our sensitivity tests. These estimates are much higher Year
than some other estimates; EPIA for example believes the (Silicon Supply for PV (MT))
37
market for PV in 2010 will be only 5.5 GW. Michael Rogol believes the
industry will be over 10 GW in 2010. An important consideration to keep
in mind when projecting potential production is that it needs to be met
with adequate demand. The strong markets of Japan and Germany are
not expected to maintain the growth they have seen in years past, while
new markets, such as Spain, Italy, and China, may take some time before
they rival Germany at hundreds of MW installed per year. The US also
has a huge potential; it will interesting to see the degree to which that
potential is realized. So while we project significant production levels by
2010, we stress that is important to remember the role public policy has
played in PV demand and will likely play in the next five years.
While it is likely that 2006 and 2007 will be difficult for the PV industry
due to primary feedstock constraints, the silicon shortage is having a
more profound impact than simply two years of higher prices and under-
utilized cell and module capacity for some producers. In the next section
we explore the long-term structural changes that the silicon shortage
will create throughout the PV Industry.
1 Woditsche and Koch, Solar Grade Silicon Feedstock Supply for PV Industry, Solar Energy
Materials and Solar Cells, 72: 11 -26, 2002.
2 Pichel, J. and Yang, M. PJC Solar Industry Note (Volume II, Issue 5) - Is The Solar Bloom Off
The Rose? July 25, 2006 p.3. Reprinted with permission.
38
7 Polysilicon Constraint Implications
The PV industry’s dependence on silicon has made it extremely vulnerable

to supply shortages. Prior to 2000, the industry could survive on the
small amount of feedstock not consumed by the semiconductor industry.
At present, however, PV silicon demand is on the verge of surpassing
IC silicon demand. In this section we provide our assessment of what
the feedstock constraint means for the PV industry in terms of material
prices, as well as the industry’s fundamental structure.
Silicon constraint will
7.1 Polysilicon Prices have short-term and
PV companies have been paying increasingly higher prices over the past
three years as demand has exceeded the supply available through traditional
long-term consequences
channels of the IC industry’s off-spec material. Prices for EG polysilicon for PV industry.
just a few years ago were in the $50/kg range, while PV companies were
paying nearly $25/kg.2 In 2004 and 2005, contract prices began to rapidly
increase as the demand growth became imminent and inventories were
depleted. Now, long-term contract prices for PV companies mirror those
semiconductor companies pay for feedstock. Today‘s contract prices are
above $50/kg; spot market prices are $200 - $300/kg.
Of the polysilicon suppliers surveyed for this report, the majority indicated
their sales were primarily via contracts with cell manufacturers; however
this practice was not universal. Furthermore, some companies that primarily
sold via contracts also sold on the spot market. Many of the long-term
silicon contracts have fixed prices for the duration of the agreement. For Figure 28: Average silicon prices
this reason, as well as numerous other uncertain variables, it is difficult to 2003 to 2005 with
predict the future price of silicon. Michael Rogol, Managing Director of projections to 20073
Photon Consulting, believes prices will not go down until after 2008.4
Profit margins for silicon producers are no doubt rising. With silicon 60 $60
selling at two to three times its 2003 price, there is greater incentive $55
for more investment in the industry. As we have shown, there are now 50
more newcomers to the business than existing players. However, even $45
with these high prices, many of the industry veterans, (e.g. Mitsubishi, 40
Price in $/kg
Sumitomo and Tokuyama) remain cautious—they remember all too $32

painfully the burst of the tech bubble and the implosion of the world’s 30
demand for silicon. $24
20
Actual
With thousands of metric tons of new capacity announced in the summer Projected
of 2006, it is possible that the industry will have enough silicon to supply 10
the projected demand, causing a stabilization or drop in prices. On the
other hand, if the semiconductor industry sees an upturn in sales, or 0
2003 2004 2005 2006 2007
if public policies are put in place that cause the PV market to expand Year
even more quickly, silicon demand may once again outpace supply past
2008. Furthermore, if announced capacities from new market entrants
fall through, demand could exceed supply.
7.2 Changing Industry Structure

Several trends have become apparent as the industry reacts to the
polysilicon shortage. First, the industry is solidifying by increasingly
39
SECTION 7 Polysilicon Constraint Implications
entering into joint ventures and long-term contracts, and by building

in-house silicon processing capabilities.
Long-term contracts and joint ventures

Joint ventures and long-term supply contracts allow for cell producers to
guarantee their feedstock supply without committing to the expensive
construction of their own polysilicon production facility. REC, Q-cells,
and Evergreen have entered into this type of joint venture agreement,
whereby REC will provide silicon to a JV entity previously established
between Evergreen and Q-cells (dubbed EverQ) and in return gain 33.3
percent ownership of EverQ.
MEMC and Motech were in discussions to sign a long-term supply

PV supply chain agreement, but they eventually parted; MEMC signed a deal with Suntech
instead, and Motech signed a supply contract with ReneSola. The MEMC/
is solidifying Suntech deal requires Suntech to provide MEMC with advance funding
for MEMC’s wafer capacity expansion. MEMC will deliver wafers to
Suntech for ten years. The Motech/ReneSola agreement will last only
two, January 2007 – December 2009. ReneSola is contracted to deliver
80 MW of wafers to Motech. Incidentally, ReneSola hopes to bring its
wafer production capacity up to 80 MW by the end of 2006, and Motech
is expanding its cell capacity to 200 MW by the end of 2006.
Other examples of joint ventures and long-term contracts include Joint

Solar Silicon, a joint venture discussed in section 4.2, and Elkem Solar,
which has long-term agreements with cell manufacturers who have been
involved in Elkem’s MG to SoG silicon research and development.
Vertical integration
It has never been more difficult for PV cell manufacturers to secure
feedstock supply. The shortage has caused many to re-evaluate their
business models. With the increasing scale of cell manufacturing, the
The industry is shortage is causing many cell producers to explore the possibility of
acquiring sufficient in-house silicon production, “vertically integrating”
making strides in to ensure long-term access to vital feedstock. The costs of vertical
integration will, however, remain prohibitive for most companies.
reducing silicon REC is a leader in vertical integration, with a presence at each level of
waste the PV supply chain. Others are moving toward vertical integration by
building polysilicon capacity for their cell production. ARISE, Hoku, and
GiraSolar have plans to build polysilicon production plants specifically
for their own use.
7.3 Efficiency Gains

The industry is also reacting to the silicon supply constraint by making
strides in reducing silicon waste through increasing recycling efforts,
improving the conversion efficiencies of cells, slicing thinner wafers, and
devising innovative processing methods.
40
Polysilicon Constraint Implications SECTION 7
Recycling
Silicon recycling may be a cheaper option for smaller cell producers looking
to secure silicon feedstock, although this still depends on available raw
material. Furthermore, when silicon supplies are constrained and prices
are high, recyclable material is also constrained and therefore more
expensive. Two German companies have incorporated recycled material
into their operations. SolarWorld doubled the capacity of its Solar Material
“Recycling will likely
Division this year and can now recycle 1,200 MT of polysilicon annually. continue to play an
In 2006, ErSol AG acquired Silicon Recycling Services (SRS). ErSol has not
yet decided if it will offer SRS’s recycled material or recycling services important role in
to outside companies. Recycling is already becoming a bigger part of PV
manufacturing and will likely continue to play an important role in the years to come.”
years to come.
Increased efficiency of silicon use and energy

conversion
Cell companies, such as ErSol Solar, Evergreen Solar, Q-Cells, and
SunPower, are improving their silicon requirements per watt. One way
they accomplish this is by producing thinner cells; ErSol is making great
strides towards this end and hopes to release a new line of cells under
200 μm in 2006.5 Increasing the efficiency of the cells is another method
of lowering silicon use. SunPower has achieved over 20 percent cell
conversion efficiencies. Ingot processing offers another opportunity to
reduce silicon waste. Solaicx, located in California, has a continuous CZ
process that the company claims is faster and less wasteful than the
traditional method for making monocrystalline ingots.11
7.4 Emergence of Alternatives: New

Silicon Processing and Gains in
Thin Film Production
Silicon-based cell manufacturing using chunk or granular polysilicon
represents the vast majority of PV production, but alternatives are beginning Thin film is gaining
to break through. Thin film’s benefit of not requiring silicon has made it
more attractive—evident in the numerous thin film expansion plans that ground
have been announced recently. More companies, such as SolarValue and
JFE Steel, to name a few, are also investing in MG to SoG efforts.
Potential competition from thin films

Technologies that do not use silicon are becoming more cost-
effective despite their lower conversion efficiencies. In recent months
announcements of increased thin-film cell production have outnumbered
those touting increased silicon cell production. Several silicon-based
companies are looking to diversify into thin films. This year, Shell Solar sold
all of its silicon-based operations to SolarWorld to focus on thin film.12 Q-
Cells announced that they will invest $9 million in the Swiss thin film PV
41
SECTION 7 Polysilicon Constraint Implications
developer VHF-Technologies SA. The company produces a product called

“Flexcell,” an amorphous silicon technology for commercial applications
that uses a roll-to-roll manufacturing process. While Q-Cells’s ownership
in VHF is currently a mere 15.5 percent, the agreement includes an option
“MG silicon could to increase its stake to 51 percent. The investment will enable VHF to
ramp up to an initial production volume of two MW per year.6
offer the industry
United Solar Ovonic, First Solar and other thin film companies are rapidly
a low-cost silicon expanding capacity. CGS in Germany produces crystalline silicon on glass
option, leading to cells. The feedstock for these cells is gaseous silicon, which presently has
no supply constraints.
declining cell and
Further expansion of metallurgical to solar grade
module costs.” material
Polysilicon now available on the market is almost exclusively made using
FBR or Siemens technology. The single exception is the PV1101 product
from Dow, which must be blended with purer polysilicon. However,
various producers are attempting to bring the MG-Si to SoG process to
commercial production. It remains to be seen if companies like Elkem
and SolarValue can achieve their announced, ambitious production goals
while maintaining adequate purity of the feedstock and keeping the
price low enough to justify the lower cell efficiencies resulting from MG-
Si to SoG product. MG silicon could offer the industry a low-cost silicon
option, leading to declining cell and module costs. This in turn would
make PV more competitive in more markets.
“The PV industry 7.5 Concluding Thoughts

will have an Based on our assessment of the planned new silicon production capacity,
we predict that the worst of the silicon shortage will be over by the
average production end of 2008, and that the PV industry will have an average production
growth rate between growth rate between 2008 and 2010 of approximately 50 percent per
year. By 2010, we forecast production to be just over eight GW.
2008 and 2010 of While circumstances may cause production in 2010 to be higher or
approximately 50% lower than eight GW, it is certain that the fundamental structure of
the PV industry is changing. The polysilicon shortage has already caused
per year.” companies to shift business strategies; the past three years of high prices
and tight supply has led to consolidation all along the supply chain. By
2010 we expect the industry to be dominated by a smaller number of
large, vertically-integrated companies.
1 Parts of this section were published by ASES for Solar 2006.

2 Price estimates provided by Mike Rogol.
3 Price estimates provided by Mike Rogol.
4 Mike Rogol, Personal Communication.
5 http://www.ersol.de/en/pm.php?type=pm&id=77&year=2006
6 http://www.eet.com/news/latest/showArticle.jhtml?articleID=190200061/
42
Contributing Staff
Hilary Flynn is a senior researcher at the Prometheus Institute where she Hilary Flynn
reports on the solar energy industry. She has researched wind and biomass Senior Researcher
technologies, as well as market-based policies designed to promote clean
energy, such as renewable portfolio standards with REC trading and cap-
and-trade programs. Hilary also has training in computer-based analysis
using system dynamics modeling and geographic information systems.
She earned a BS from Rutgers University, and an MS from Washington

State University under the mentorship of Dr. Andrew Ford. She has co-
authored articles in Energy Policy, System Dynamics Review, Renewable
Energy World and Solar Today, and has presented her research at
conferences in the US and abroad.
Travis founded the Prometheus Institute in 2003 as a means to connect the Travis Bradford
vast reach and power of industrial and capital markets with the technologies Founder, President, and Director
necessary to sustain and develop long-term economic well-being for people
around the world. Having spent time in nearly 40 countries, he has seen first-
hand the state of economic development and the need to develop markets to
make existing sustainable technologies more available and cost-effective.
Prior to founding the Prometheus Institute, Travis was a partner at Steel

Partners II, L.P., a hedge fund based in New York City focused on the
acquisition, growth, and sale of small publicly traded and privately owned
businesses. In this capacity, Travis served as both a board member and
active management participant in these types of businesses in industries
ranging from industrial filters to fertilizer distributors. It is this practical
business experience which he hopes the Prometheus Foundation can
bring to the sustainable technology industry.
Travis has worked for the Federal Reserve Bank, has lectured at top Universities
including Columbia University, Duke University, and New York University on
finance and entrepreneurship, and is co-author of a paper in the Journal of
Applied Corporate Finance entitled Private Equity: Sources and Uses. He is
also a partner at Atlas Capital, a hedge fund based in Cambridge, MA.
Robert serves as the Institute’s Office Administrator and Graphic Designer. Robert Schlussler
He attended the Defense Language Institute in Monterey, California Office Manager
where he studied Korean. Robert has several years of experience
working in non-profit. Prior to joining the institute, he was the System
Coordinator for Lesley University, where he collaborated to develop and
maintain software and policies with regard to the University’s housing,
registration, financial aid, and billing systems.
Kate Cell comes to the Institute with six years of experience in nonprofit Kate Cell
media and policy outreach and fundraising, specializing in the area of Director of Development
helping the public and their political representatives appreciate the
and Communications
implications of economic research and analysis.
Kate studied writing at the University of Iowa’s Writer’s Workshop and

Cognitive Pscyhology and English at Macalester College. She is a fellow
of Economists for Peace and Security, the editor of the EPS Quarterly,
and serves on the editorial board of the Economics of Peace and Security
Journal, a new online peer-reviewed publication.

PROMETHEUS INSTITUTE FOR SUSTAINABLE DEVELOPMENT
1280 Massachusetts Avenue, Suite 203
Cambridge, MA 02138 USA
Phone: 1.617.868.0337
Fax: 1.617.868.0076
Email: [email protected]
www.prometheus.org
© Copyright 2006 The Prometheus Institute for Sustainable Development. All Rights Reserved.
Printed on recycled paper

Polysilicon

Uploaded by

Polysilicon

Uploaded by

2006

Produced by the Prometheus Institute for Sustainable Development

Many people provided information

Today, the Institute strives to be the world’s leading source of publicly

 Cover photo courtesy of Renewable Energy Corporation.

Polysilicon: Supply, Demand, & Implications for the PV Industry

3 Polysilicon Production Capacity—Historical Production and Major Players 13

4 Emerging Polysilicon Producers and Supplemental Silicon 23

5 Geography of Global Polysilicon Capacity 31

7 Polysilicon Constraint Implications 39

We also look at various technological approaches to refining silicon,

I want to thank all the staff, advisors, and company representatives

States which has a highly-trained labor force

technology-oriented countries, including

States is expected to continue to be the top

The existing polysilicon producers traditionally catered to the electronics

Table 1 illustrates the spate of new capacity expansion announcements

The majority of new production recently announced will be undertaken by

Table 1: Silicon production announcements from April–September 2006

The combination of new production by existing suppliers, new entrants,

In the next section, we explore the different silicon production

2.1 Solar Grade vs. Electronic Grade

2.2 From Silica to

US MG-Si production in 2005 was 148,000 MT.5 Globally,

There are several ways in which MG-Si can be refined. We

2.3 Metallurgical Silicon Processing

A third process for polysilicon production uses a fluidized bed reactor

Figure 9: Fluidized bed reactor

Ethyl Corporation developed the FBR process for polysilicon production.

A fourth and still experimental polysilicon technology is Tokuyama’s

The polysilicon industry, regardless of the technology, requires a

highly trained labor force. As a general rule, it costs $100/kg to build a

In an attempt to lower silicon production costs, as well as to ensure

Upgraded Metallurgical Silicon

Companies such as Dow Corning, Elkem, and others are

2.4 Silicon Ingots: Monocrystalline

CZ Method Figure 12: CZ drawing

With directional solidification, the silicon remains in the crucible after

A second method is Edge-defined Film Fed Growth (EFG), developed and

Crystal Growing Systems (CGS) and SCHOTT recently succeeded in

Figure 17: Wire saw machine

Each of these will help to improve the efficiency of silicon

Courtesy of LDK Solar

1 This section relied on several sources: PV handbook, Elkem’s website, http://www.wafernet.

For the purposes of this report, we consider production capacity to be

semiconductor industry dropped, leaving the polysilicon

producers with excess capacity. As figure 18 shows, very

3.3 2005 Polysilicon

US, is the largest of the polysilicon companies

Polysilicon capacity has not increased significantly

over the last five years due to production

3.4 Polysilicon Producer Profiles

Hemlock uses Siemens reactors and trichlorosilane gas to produce chunk

In 2005, the company reported a production of 7,700 MT. It will increase

Hemlock 2005 2006 2007 2008 2009 2010

The polysilicon segment of Wacker experienced a 12 percent increase in

Wacker’s polysilicon production capacity was 5,500 MT in 2005. It will

Wacker 2005 2006 2007 2008 2009 2010

In 2005, REC purchased a polysilicon plant in Butte, MT, formerly owned

REC 2005 2006 2007 2008 2009 2010

In September 2004, Tokuyama announced its plans to expand production

Tokuyama 2005 2006 2007 2008 2009 2010

MEMC produces polysilicon feedstock usable in PV cells at two factories,

Annual polysilicon production capacity at MEMC was 3,800 MT in 2005,

MEMC 2005 2006 2007 2008 2009 2010

Mitsubishi 2005 2006 2007 2008 2009 2010

Sumitomo 2005 2006 2007 2008 2009 2010

3.5 Polysilicon Recyclers − IC Castoffs

1 Price estimates provided by Mike Rogol

2010 Production capacity Technology

DC Chemical 3,000/-- Siemens

4.1 New Producers using

group has no previous Isofoton, Spanish government and Endesa

French Consortium (Econcern, NorSun and

Crystal/Russia and Former Soviet Union Companies

RU/FSU 2005 2006 2007 2008 2009 2010

Cover photo courtesy of Renewable Energy Corporation.

3 Polysilicon Production Capacity—Historical Production and Major Players 13

4 Emerging Polysilicon Producers and Supplemental Silicon 23

5 Geography of Global Polysilicon Capacity 31

7 Polysilicon Constraint Implications 39

Polysilicon: Supply, Demand, & Implications for the PV Industry