POLYSILICON PRODUCTION PROCESSES
Why the Siemens process is still the prevailing technology to manufacture polysilicon
Despite more than a dozen attempts to develop less expensive alternatives, the Siemens process has remained the dominant technology to produce highly pure polysilicon. Low-cost plants in China have driven the production costs of the process down to unprecedented levels.
Two massive waves to develop alternatives – but almost no success
The sixty-four-thousand-dollar question: When was the following statement made? “Other processes that depart from conventional technology are also under consideration for the production of polysilicon.” If you think the statement is from the period of polysilicon shortage between 2005 and 2008, you are wrong.
The riddle’s solution: The quote originates from a paper presented by Carl Yaws of the Lamar University in Beaumont, Texas at the U.S. “Flat-Plate Solar Array Project Workshop on Low-Cost Polysilicon for Terrestrial Photovoltaic Solar-Cell Applications” in Las Vegas, Nevada in October 1985. Yaws listed no less than 17 manufacturing alternatives to the standard Siemens process.
Triggered by the energy crisis in the early 1970s, the Flat-Plate Solar Array Project ran from 1975 to 1986. When it was finished, oil prices had fallen again and left photovoltaics (PV) in a niche market too small for a company to dedicate a polysilicon plant to it. Instead, the PV branch lived on scrap silicon from the semiconductor sector (polysilicon uses).
With the explosion of the German PV market in 2004, however, the scenario for a severe polysilicon shortage was perfect (polysilicon market analysis), and many of the approaches presented by Carl Yaws in 1985 were resuscitated. In view of the high polysilicon prices caused by the shortage, the goal was again to develop less expensive methods than the prevailing Siemens process to produce solar-grade polysilicon for the solar industry.
Around a dozen alternative routes were investigated this time. In contrast to the dynamically evolving market, however, nearly all attempts have remained unsuccessful.
Top dog: the Siemens process (rod reactor)
What practically everyone underestimated: The Siemens process did not stand still. Developed by the German companies Siemens and Wacker in the 1950s to produce hyper-pure polysilicon for the semiconductor industry, the process has made enormous technical progress. It has proven to be a moving target that could hardly be caught by alternative approaches. Therefore, it is no wonder that the market share of the Siemens process in polysilicon production has dipped below the threshold of 90% only once since 2004, namely in 2008 when the polysilicon shortage reached its peak.
One major reason why the process has turned out to be so resilient against new approaches is the rise of the polysilicon industry in China. Initially, China-based polysilicon plants obtained the key equipment, notably hydrochlorination and chemical vapor deposition reactors, from providers in the United States, Germany and Italy. In recent years, however, they have procured low-cost equipment from domestic suppliers.
On top of that, Chinese manufacturers are benefitting from subsidized capital expenditures (capex), low raw material and labor costs as well as cheap electricity rates in northwestern China; companies using the Siemens process at these locations have driven the production costs of solar-grade polysilicon down to levels that were inconceivable a decade ago.
The Siemens process – How polysilicon rods and chunks are made
To remove the 0.5% to 1.5% of impurities contained in metallurgical-grade (MG) silicon, the Siemens process creates trichlorosilane (SiHCl3, or briefly TCS), a highly volatile liquid, as intermediate product.
For that purpose, MG silicon is ground up into small particles which react with hydrogen chloride (HCl). The resulting TCS has a low boiling point of 31.8 degrees centigrade (°C) so that it can be purified in tall distillation columns relatively easily.
Silicon is then deposited from the TCS on highly pure, slim silicon filaments that are electrically heated to up to 1,150 °C in a steel bell-jar reactor (see image on the top of this page) until they have grown to polysilicon rods with a diameter of 15 to 20 cm. This energy-intensive step is called chemical vapor deposition (CVD). The long rods are broken into small chunks.
The by-product silicon tetrachloride (SiCl4, or briefly STC) is recycled to TCS mostly through hydrochlorination: STC is fed along with hydrogen (H2) and MG silicon particles into the reactor for TCS production.
Depending on how thoroughly TCS is distilled and whether impurities on the surface of the polysilicon chunks are etched off, different levels of polysilicon purity can be achieved:
- solar grade for multicrystalline cells (multi grade): 99.99999% (7N) to 99.999999% (8N);
- solar grade for monocrystalline cells (mono grade): 9N to 10N;
- electronic grade for semiconductors: 10N to 11N.
The Siemens process has made enormous progress in cutting manufacturing costs since its invention; during the last few years in particular, leading polysilicon manufacturers have driven the cash production costs (excluding depreciation) down to below US$10 per kilogram.
Eternal contender: fluidized bed reactor technology
Long before the polysilicon shortage loomed, fluidized bed reactor (FBR) technology was already applied by MEMC Electronic Materials to produce granular polysilicon in Pasadena in the U.S. state of Texas.
But so far, REC Silicon has been the only company able to establish new large FBR plants: at first, one in Moses Lake in the U.S. state of Washington in 2009, then a second in Yulin in the Chinese province of Shaanxi in 2017.
To date, four obstacles stand in the way of the FBR approach reaching a higher market penetration:
- The technology is protected by many patents.
- The complex fluid dynamics require a great deal of time, experience and capital to scale an FBR up from lab to pilot to industrial scale.
- A liner has to be used to prevent the reactor wall from contaminating the polysilicon granules produced, which drives up costs.
- The advantage of low electricity consumption can largely be eaten up by a high share of unusable silicon dust in the output.
Therefore, the technology has yet not made greater strides on the polysilicon market.
The fluidized bed reactor process – How polysilicon granules are made
A fluidized bed reactor (FBR) has the shape of a tube. Silicon-containing gas is injected together with hydrogen (H2) through nozzles at the bottom to form a fluidized bed that carries tiny silicon seed particles fed from above.
REC Silicon uses monosilane (SiH4) as feed gas whereas the small FBR facility of Wacker Chemie works with trichlorosilane (SiHCl3, or briefly TCS). While SiH4 decomposes in the reaction zone at temperatures of 650 to 700 degrees centigrade (°C), TCS only does so at 1,000 °C.
When the respective temperature is reached, silicon deposits on the seed particles until they have grown to larger granules that drop to the bottom of the reactor. From there, they can be withdrawn continuously – in contrast to the Siemens process, which is interrupted when the polysilicon rods are harvested.
A monosilane-fed FBR consumes only one tenth of the electricity needed for heating a conventional rod reactor in the Siemens process. Mixing polysilicon chunks with granules from an FBR in a fifty-fifty ratio can shorten the time to fill a crucible by 40% and increase the charge weight by 30%.
Solitary existence: upgraded metallurgical-grade silicon
It was only a short heyday that upgraded metallurgical-grade (UMG) silicon enjoyed on the polysilicon market, with ten manufacturers operating at the zenith in 2008. Owing to its low quality, UMG silicon was only a work-around, since it could solely be used for blending with conventional polysilicon at the time.
Later on the quality improved significantly, and multicrystalline solar cells made of UMG silicon reached efficiencies that were on par with those of cells produced from conventional polysilicon. Nonetheless, UMG silicon has never achieved a real breakthrough on the market; the current upswing of monocrystalline solar cells, which require purer feedstock than multicrystalline cells do, makes it all the more difficult for UMG to flourish. The last commercial manufacturer – REC Solar Norway (formerly Elkem Solar) in Kristiansand – stopped production in the spring of 2020.
Upgraded metallurgical-grade silicon – How it is purified physically
Unlike in the standard production process for polysilicon, manufacturers of upgraded metallurgical-grade (UMG) silicon do not pursue a chemical route to purify the raw material of metallurgical-grade silicon (also called silicon metal).
Instead, they use physical methods, such as vacuum melting of the silicon metal, blowing of reactive gases through the melt, treating it with slags, leaching of solidified and crushed silicon with acids or directional solidification of molten silicon.
All these methods serve to extract impurities directly from the silicon metal, and consume much less energy than the standard Siemens process.
Initially, UMG silicon only reached a purity of 99.999% (five nines or 5N for short). In the meantime, it has improved to 6N.
However, the cost advantage of UMG silicon over the Siemens process has become negligible in recent years.
Conclusion: The Siemens process will solidify its leading position
Only fluidized bed reactor technology has a chance to grab some percentage points of market share from the prevailing Siemens process. Upgraded metallurgical-grade silicon is fighting an uphill battle in a PV market where monocrystalline solar panels are dominating more and more. In view of emerging Chinese polysilicon giants with annual production capacities of more than 100,000 metric tons, the Siemens process will solidify its leading position.
Published on June 29, 2020. Last update: September 11, 2020 © Bernreuter Research
About the author
Johannes Bernreuter is head of the polysilicon market research specialist Bernreuter Research. Before founding the company in 2008, Bernreuter became one of the most reputable photovoltaic journalists in Germany because of his diligent research, clear style and unbiased approach. He has earned several awards, among others the prestigious RWTH Prize for Scientific Journalism from the RWTH Aachen University, one of the eleven elite universities in Germany.
Originally an associate editor at the monthly photovoltaic magazine Photon, Bernreuter authored his first analysis of the upcoming polysilicon bottleneck and alternative production processes as early as 2001 (Publication List). After preparing two global polysilicon market surveys for Sun & Wind Energy magazine in 2005 and 2006, he founded Bernreuter Research to publish in-depth polysilicon industry reports.
Get Your In-depth Technology Assessment: The Polysilicon Market Outlook 2024
- Learn all about the latest developments of polysilicon manufacturing technologies (Siemens process, fluidized bed reactor, upgraded metallurgical-grade silicon)
- Obtain comprehensive data on production volumes and capacities of 42 solar-grade and electronic-grade polysilicon plants from 2016 through 2024
- Gain insight into decisive market trends, based on four sophisticated scenarios of supply and demand through 2024
- Get valuable guidance with cash cost data on 25 solar-grade polysilicon plants and spot price forecasts through 2024
- Benefit from 76 pages full of rich data, in-depth analyses and detailed forecasts on the polysilicon, solar and semiconductor industries