2019 Volume 36 Pages 264-270
The Swirl reactor is an innovative concept for performing reactions in gas phases or solid and gas phase mixtures. It was developed during the initial boom of the solar cell industry, where the need for more energy- and cost-efficient means of producing high-purity solid silicon became important. The Swirl reactor was designed to fulfil the following requirements: continuous production of high-purity silicon; efficient energy transfer from the reactor to the gas; controlled transport of the Si-fines and the gas formed; the use of silane as feed for silicon. By employing a stainless steel tubular reactor honed on the inside and heated from the outside, and using argon as the carrier gas, the requirements were all shown to be fulfilled. The Swirl reactor throughput can be increased by having more than one injector introducing new swirls. Also, two swirls—each containing reacting agents—may form product where they mix, i.e. along the path where the swirls overlap. Thus, reactions can be controlled but still be run continuously. The possible uses of the Swirl reactor are numerous.
In the first decade of this century, the photovoltaic industry had a growth of 10–30 % annually. Thus there was a strong incentive to produce high-purity silicon at a lower operating cost, e.g. at a lower energy consumption. For a comprehensive review of photovoltaic production at the time of development of the Swirl reactor, we refer to the Handbook of Photovoltaic Science and Engineering (Luque and Hegedus, 2003).
The standard method for the production of high-purity silicon was the use of the so-called Siemens reactor where trichlorosilane (HSiCl3, called TCS) is thermally decomposed at silicon rods heated by electric current. This process is very energy-intensive and alternatives have been designed and tested. The Siemens reactors used to produce solid silicon from TCS are batch-operated and consume a huge amount of thermal and electric energy in addition to producing a very corrosive gas, HCl, at high temperatures: HSiCl3 + H2 → Si(s) + 3HCl. The use of silane, SiH4, instead is a much more attractive idea since SiH4 → Si(s) + 2H2. Union Carbide Corp. had started the development of a fluidized bed reactor (FBR) based on silane, but were not successful. They sold their silicon facility to Komatsu Ltd. who formed the company Advanced Silicon Materials Inc. (ASiMI), which continued the development of the FBR. However, it was not until REC purchased the plant that this alternative finally became the process of choice for growing Si-crystals from the thermal decomposition of silane (Filtvedt et al., 2013; Du et al., 2014). The company REC Silicon is today the leading producer using this technology1). First, PV-quality silicon is produced, then the standard method of producing polycrystalline silicon wafers for PV-cells is applied, i.e. melting the silicon granules in a crucible, then performing a slow crystallization from the bottom creating what is called a bowl, and finally using a wire-saw to cut the bowl into ca 200-μm-thick sheets. This also turns almost half of the silicon into sawdust, called kerf. Since the saw uses glycol for cooling, the kerf is not easily regenerated. Today the Norwegian company ReSiTec is focusing on this part of PV production.
However, during the competitive days, free-space reactors were tested for the decomposition of silane, but with minor success. The silicon formed was amorphous and of a very small particle size and high surface area. Such fines tend to stick to all kinds of surfaces and grow aggregates. This phenomenon is studied in the frames of homogeneous and heterogeneous nucleation of silane pyrolysis (Slootman and Parent, 1994; Onischuk et al., 2000). Attempts to develop free-space reactors stranded on the handling of the silicon fines. This was also the major obstacle to overcome in development of the FBR.
To meet the challenges of both energy transfer from a hot surface to the gas and to control the movement of the formed fines and agglomerates, the concept of a Swirl reactor was developed. Since the concept was also continuous, the succeeding production of wafers could also be included, but this part of the project was not initiated at the time when the Swirl reactor was developed. Later, a float-casting program was initiated in cooperation with a group at Carnegie-Mellon University in Pittsburgh, PA, USA (Ydstie et al., 2009).
The production of silicon was a case study, but the development of the reactor was in fact based on other kinds of powder and gases, and the technology is therefore not limited to the decomposition of silane.
The Siemens reactor is a device where the gases to be decomposed are separated from air (to avoid the formation of SiO2) and is operated at high temperature at the Si-rods (1100 °C), where the TCS is thermally reacted with hydrogen and decomposed to form polycrystalline silicon. To avoid reaction of the gas at other sites in the reactor than at the hot rod, the surface of the reactor must be cooled. Thus there is a huge heat loss in the process. In addition, the Siemens reactors must be operated in batch mode and involve manual interactions. To represent an attractive alternative, the Swirl reactor was designed to fulfil the following requirements:
The concept can be expressed as: silane gas is swept along the inner surface of a tubular reactor being heated from the outside. The silane is diluted in an inert gas, e.g. argon or hydrogen, which acts as a carrier gas. The carrier gas will sweep the surface and bring the solid silicon formed to the end of the reactor for further processing. Of course, the Swirl reactor is not limited to silane, argon and hydrogen. In principle, it will be possible to use any kinds of fines and gases. Also, there is no reason why liquids cannot be used.
The first test of the reactor was performed using simple tubular transparent reactors made of plastic or glass, using air as the gas and silica, SiO2, or smoke as the solid particles. The particles’ trajectories could then be measured by Laser Doppler Anemometry (LDA). (A primer of LDA can be found in Mathiesen et al. (1999)). In this way, we were able to easily make conical reactors to see how the swirl would be affected by a decreasing or increasing diameter. All reactors had smooth inner surfaces.
With LDA, the velocities’ magnitude in the axial and tangential directions of the particles can be measured. It was shown that the path of the particles was a helix and that this was maintained even when the fluid mixture was heated, causing an expanded volume of fluid gas.
In the second part, the objective was to decompose silane. Then a reactor operating at temperatures up to at least 600 °C was needed. This is above the stability of glass. We therefore attempted to make a reactor based on a cylinder of quartz. However, we had to abandon the visibility due to technical problems with leaks in the connecting parts of the quartz reactor. This part therefore used a tubular stainless steel reactor. The challenge was then to prove that the swirl was maintained during the process.
The LDA equipment was made by Dantec Dynamics who also supplied the lasers and control and analysis software. The gases used were argon and 5 % silane in argon, both supplied by AGA. The reactors were all produced exclusively for this project at the Institute for Energy Technology (IFE), Kjeller, Norway. IFE also constructed and manufactured the in- and outlet parts of the reactor according to our requirements. The injector part had a central inlet for measuring temperature and four peripheral inlets for gas and solid reactants or for other measuring devices. The outlet was too hot for standard filters and a water-cooled filter holder was therefore needed and constructed.
For the initial LDA tests, cylindrical and conical reactors were made from transparent plastic and glass materials, and pressurized air was injected into the reactor with a lance injector through an inlet disc. Smoke was mixed with the air during low-temperature tests in order to make the flow patterns detectable by the LDA equipment, while SiO2 dust particles were used during high-temperature tests.
In the high-temperature tests, the gas temperature inside the reactor was increased linearly along the reactor from room temperature (approx. 300 K) to 600 K, simulating an expanding gas flow as would be the case for a reaction where SiH4-gas is decomposed completely to Si(s) and H2(g). The increase in temperature was achieved by wrapping heating tape around the reactor on the outside and with decreasing space between neighboring tape windings along the reactor.
A cylindrical stainless steel (SS316) reactor was constructed, 500 mm long, with an inner diameter of 56.3 mm, and a wall thickness of 2.0 mm. The tube was honed inside to ensure a smooth reactor wall. At the ends, flanges were welded on, resulting in a neck of 40 mm on each side of the reactor. A separate injector piece was made and attached to the front flange. This was convenient when the injection angle or nozzle size were to be changed. At the back flange, a filtering piece was attached. As the gas exiting the reactor is hot and the dust particles to be collected are very small, we made a water-cooled filter housing. A steel mesh to support the glass-fiber filter also provided cooling to the filter.
Fig. 1 is a simplified Process Flow Diagram (PFD) of the set-up for a useful process including recycling of the carrier gas.
Simplified Process Flow Diagram (PFD) of a Swirl reactor with pumps/compressors, valves and pressure flasks. The solid formed is separated and collected and the carrier gas is returned for reuse. This diagram is based on the thermal reduction of silane forming Si and H2, which also is the carrier gas.
We have only tested the reactor at laboratory scale, producing a few grams per hour. One of the objectives in the project was to obtain design parameters for scaling-up the reactor. The reactants’ length of trajectory along the inner surface of the reactor depends on the flow rate, the diameter, and the injection angle, whereas the heat transferred to the reaction is a function of the temperature of the reactor and the entering carrier gas, the nozzle size, the flow rate and the reactor length. A software package simulating this was planned as a part of the project, but this part had to be abandoned when the project was stopped.
The silicon produced was analysed for structure by X-ray diffraction (XRD) using an INEL XRG 3000 diffractometer, with a multichannel curved position sensitive detector, recording all 2θ positions simultaneously. Particle size distributions were measured by a Malvern Mastersizer 2000, using laser scattering.
The particles’ axial and tangential velocities were measured by LDA. The trajectories can then be extracted in various ways. Figs. 2 and 3 show two possible representations. Both are real measurements of SiO2-particles blowing in a conical reactor which is heated by a heating tape wound around the outside of the reactor wall. The laser was employed between the windings.
Iso-surface plots for non-expanding (left panels) and expanding (right panels) swirling gas flows of SiO2-particles. The expansion is due to heating.
Measured velocities in an isothermal, non-expanding conical swirl flow reactor. The upper panel shows the axial velocity, and the lower panel shows the tangential velocity.
Fig. 4 shows an XRD spectrum of silicon produced by thermal decomposition in a small quartz tube reactor not used for the swirl measurements. The XRD measurements confirm that the bulk of the produced silicon is amorphous. The lack of crystalline structure gives the wide peak in the spectrum. The peak is possibly also widened by the small particle size and the small amount of silicon. The freshly made Si-particles tended to form aggregates. To enable particle size distribution measurement, the produced silicon was absorbed in glycol and the particle distribution measured a short time after (in the order of minutes) by laser scattering. This method was not qualified, but the best measurements gave a distribution ranging from 0.04 to 1 μm. If the particles were allowed to aggregate, the distribution would range from 200 to > 2000 μm. This gives an illustration of the property of Si-fines to form aggregates. The average particle sizes can be used for estimating the surface area, and simple calculations using the atomic distance of Si-Si bonds as the atomic diameter (235.2 pm; Winter, 2018)—assuming spherical Si-particles—gives estimates of > 100 000 m2/g. Considering that the Si-atoms at the surface have non-saturated bonds, it is not surprising that agglomerates are formed.
XRD spectrum of silicon produced by the thermal decomposition of silane, SiH4. The spectrum shows the intensity of reflections as a function of channel numbers, which is proportional to the scattering angle. The narrow peaks are reflections of elements in the backing of the silicon and have nothing to do with the Si.
Fig. 5 shows the structure of the silicon absorbed on two filters. It is shown that applying the same conditions, i.e. mainly nozzle injection angle and flow, the flow conditions inside the reactor are then maintained and there can be no other cause than a swirl for creating such a complex pattern. The difference in color is believed to be dependent on the time the filter was exposed to air. The silicon particles have a high surface area, as explained, and can easily form oxides which will show up as white or gray.
Silicon produced in the Swirl reactor at one set of injection angle and flow. The difference in color is due to the different time of exposure in air before the picture was taken.
Moreover, Fig. 6 represents a different condition, but again it was shown that the flow pattern was maintained as four equivalent tests gave the same pattern.
Silicon produced in the Swirl reactor at a different set of injection angle and flow compared to Fig. 4. The same operational conditions were tested four times.
With a fixed steel reactor length, the three parameters to vary were temperature, injection angle and flow. Nozzle size was not usually changed. Figs. 7 and 8 show the relative variations in yield of silicon as a function of these three parameters. The yield in this case was unfortunately not a well-defined amount in this experimental set-up. It was determined as the amount of Si accumulated on the filter relative to the amount injected. But no Si outside the filter or unreacted silane was detected or included in the calculation of the yield reported. The fact that the yield approaches 100 % at the highest flow rate tested indicates that the highest flow rate brings out more Si from the reactor and therefore gives a more efficient swirl. The silicon not measured at the outlet for the lower flows is believed to stick to the interior of the reactor. The temperature reported is the one at the outside of the reactor. The temperature in the center of the reactor was monitored, but only to check that the gradient between the outside and inside was small.
Relative yields as a function of temperature and injection angle. Gas flow 2.0 L/min. The line is drawn to guide the eye.
Relative yields at 600 °C as a function of the carrier gas flow rate.
The results presented in the foregoing paragraph clearly show that the swirl flow is maintained in the reactor. Moreover, finely grained, amorphous material with a high affinity for making agglomerates and a tendency to stick to any surface could be handled well.
When our work on the Swirl reactor was ongoing, the focus was primarily on the production of photovoltaic (PV) silicon. From the above results the optimum seems to be 45° injection angle, 3.0 L/min gas flow at a temperature of 650 °C (923 K). Both injection angles of 30° and 60° seem to give lower yields than 45°. Overly small angles may give too little force along the reactor and then lose solids to the central part of the reactor, whereas overly large angles will not wipe the whole surface effectively. These statements are hypotheses and need to be investigated further.
The Swirl reactor has similarities with the downer reactor, but that is based on cyclone technology. A downer reactor is vertical and uses gravitation for avoiding back-mixing. Cheng Y. et al. (2008) have published a review of this reactor technology. Downer reactors utilize plug flow whereas the Swirl reactor—where the fluid is heated at the inner surface of the reactor—develops a radial kinetic energy gradient that causes turbulence. Estimates of the Reynolds number indicate figures well above the limit for a turbulent regime. This implies good mixing between solids and gas; important for reactions which comprise these.
Downer reactors were primarily developed for the same purposes as fluidized bed reactors are used for: reforming catalysis of natural gas, cracking of hydrocarbons, anaerobic combustion, etc. The solids employed in the downers, e.g. catalyst and absorbents, are usually recycled. The production of silicon is not a common theme, but Levin H. and Ford L.B. (1982) have patented a vertical cyclone reactor for transforming silicon gas directly into a melt of silicon. We believe the Swirl reactor may be operational in the same segments as the downer reactors.
The Swirl reactor throughput can be increased by having more than one injection point. As shown in Fig. 2, there is space without solids between the swirls. This can be utilized by another swirl. Also, two swirls containing reacting agents may then form product wherever they touch, i.e. along the path where the swirls overlap. Thus reactions can be controlled but still be run continuously.
By combining Swirl reactors in a “honeycomb” pattern, heat loss can be reduced and throughput increased. Increasing the diameter of the reactor increases the volume even more, but the central part of the Swirl reactor is of less use, so there is a possibility of turning the reactor into an annular reactor and thus saving space to be heated.
The concept of the Swirl reactor was patented (Eriksen D.O. and Gorset O., 2008) by the Institute for Energy Technology (IFE) where the authors were all once employed, but was later stopped and there is presently no patent hindering the use or development of this reactor concept. In the patent, means for separating the solids formed from the carrier gas, flow distributions inside different kinds of reactors and verification tests are described.
We believe the Swirl reactor may be useful also for liquids and for the treatment of ores such as carbo-chlorination reactions. Here, the oxidic mineral or solid feed can first be mixed with carbon and then in the Swirl reactor the solid mixture meets chlorine gas. The carbon binds the oxygen and the metal will form gaseous chlorides. Metals forming gaseous chlorides are iron, titanium, silicon, vanadium, zirconium, etc. This can also be used for removing such metals from other non-volatile metal chlorides like the rare earths.
A new segment for use may be the coating of solid particles. For example, combining finely ground graphite and silane may form Si-coated graphite which is of great interest as an electrode material for use in Li-ion batteries.
This scientific work was financed as a Strategic Institute Project at IFE by a grant from the Research Council of Norway. The authors are grateful to IFE for releasing the patent so that interested parties can develop the technology further.
Abbreviations used in the text:ASiMI
Advanced Silicon Materials Inc.FBR
Fluidized Bed ReactorIFE
Institute for Energy Technology, Kjeller, NorwayLDA
Laser Doppler AnemometryPV
Process Flow DiagramREC
Renewable Energy CorporationTCS
Dag Øistein Eriksen
Dag Øistein Eriksen has a PhD degree in nuclear chemistry from the University of Oslo, 1976. He is currently the owner and CEO of Primus.inter.pares AS—a private consultancy company excelling in separation science and the use of radioactivity as tracers. Eriksen was eight years in the rare earth industry and 15 years as a senior scientist at the Institute for Energy Technology (IFE). He was appointed as a member of the Expert Group of Prometia and has for the last four years served as senior advisor to the Department of Chemistry, University of Oslo.
Oddvar Gorset has his MSc. degree in chemical engineering from the Physical Chemistry Department at NTNU (the Norwegian University of Science and Technology) in 1995. He has held positions as process engineer, researcher, project manager and senior process engineer in the engineering companies ABB Environmental, Alstom and Aker Clean Carbon, and the research institute IFE. Oddvar now works as Specialist Engineer and Project Manager in Aker Solutions, Fornebu, Norway, within Aker Solutions’ CO2 capture technology.
Håvar Gausemel has a Dr. scient. in nuclear chemistry from the University of Oslo, 2004. He is currently the Radionuclidic Sourcing Manager and Radiation Safety Expert at Bayer AS, a pharmaceuticals company. Gausemel held a post-doc. position at IFE while working on the swirl reactor and worked as a senior scientist at Renewable Energy Corporation AS for four years, focusing on the thermal decomposition of silane.