Marisa Di Sabatino ^1* Bruno Ceccaroli ²

• ¹ Norwegian University of Science and Technology, Trondheim, Norway
• ² Consultant, Oslo, Norway

SiO2 + 2C → Si+ 2CO in electric arc furnace with submerged electrodes. The process is more than a century old, well proven and optimized with respect to many performances. The product tapped from the furnace is commonly called silicon metal (98-99% purity) because of its appearance and in spite the fact that silicon by chemical classification is not a metal but a metalloid. Its first commercial and industrial application was as alloying element to aluminium, a metal discovered simultaneously with silicon in the second part of the 19 th century. Massive usage of aluminium alloys started around the second world war with the demand being stimulated by the aircraft industry. Referring to application in metallurgy, the product was designated as Metallurgical Grade Silicon MGS or MG-Si, the main impurities (1-2% in total) being Fe, Ca and Al. The latter impurity mentioned has no significance when silicon dissolves in molten aluminium and, therefore, is never counted in the specifications by this industry.Approximately at the same time around the second world war emerged the silicones or polysiloxanes, a new class of inorganic polymers forming oils, elastomers and resins exhibiting remarkable and unique properties. The backbone of the polysiloxanes consists of silicon atoms alternating with oxygen atoms, each silicon atom carrying two alkyl groups, most commonly methyl groups. The manufacture of the silicones starts with silicon metal as main raw material. The purity required to silicon is 98-99%, i.e. the same level as for aluminium alloys but in case of siloxanes Al, Ca and the ratio between the concentration of both elements (Al/Ca) must be carefully controlled to achieve high productivity and reactivity in the direct process of making the monomer precursor (dialkyldichlorosilane). The exact composition of impurities is achieved by oxidative refining of the tapped molten silicon in the ladle before casting/solidifying and crushing. The silicon grade dedicated to the silicones is sometimes designated as Chemical Grade Silicon by vendors and buyers to this industry. However, in the semiconductor and photovoltaic industry making use of silicon of far higher purity, this distinction is not made and all silicon grades below 99% are by extension designated as MGS, regardless the application is metallurgy or chemistry.With the invention of the transistor and the development of solid-state electronics in the 1950s and onwards, ultra-high pure silicon was required. Impurities had to be at the part per trillion (ppt) level. Early in the 1960s a process was widely adopted to achieve this challenging target. In this process, silicon (MGS grade) is reacted with chlorine (or hydrogen chloride) to form chlorosilanes (preferably trichlorosilane), easily volatile compounds which can be purified by distillation and thus enable the required purity for semiconductor silicon (i.e. between 99.99999% and 99.999999%). The chlorosilane is then reduced to elementary silicon of ultrahigh purity by Chemical Vapor Deposition (CVD) at high temperature (approximately 1150 °C) on a heated filament (preferably of silicon). This process is commonly known as the Siemens process according to the name of one of the companies developing it around 1960. The resulting product is shaped as long rods, which can be used as such in Float Zone (FZ) single crystal pullers or as broken chunks and chips in Czochralski (CZ) single crystal pullers. The product is commonly called Polysilicon by reference to its polycrystalline structure, and by opposition to the monocrystalline structure obtained through the FZ and CZ pulling. It is also designated as Electronic Grade Silicon EG-Si or EGS by comparison with UMG-Si and other silicon grades. The single crystals successfully grown and qualified are further processed to wafers. From these wafers, both solar cells and integrated circuits can be obtained.The oil crisis in the first part of the 1970s pushed the development of alternative energy sources to petroleum. Solar energy was identified as a major challenger to oil. The first solar cell was manufactured in the Bell`s laboratory in U.S. in 1954 with an efficiency of around 6% [1] and started the development of photovoltaics (PV). Silicon solar cells made with the silicon technology known from the semiconductor industry were initially produced for aerospace satellites with not much attention to costs. With the oil crisis striking in the 1970s, ambitious programs to develop commercial terrestrial PV were launched in the U.S. (by Department of Energy/DoE), in Japan (Ministry of Economy, Trade and Industry/METI) and in selected European countries (European pioneers were Italy and France) to make solar electricity competitive with oil. In the mid 70s a symbolic cost target assigned by the DoE to the developers was 1 US dollar per watt-peak (Wp) on the module level, an objective which took 30 years to reach. These ambitious programs and objectives trigged the development of low-cost silicon of purity suitable for solar cells, i.e. not as high purity as for semiconductors but much purer than MGS. The concept of Solar Grade Silicon SGS or SoG-Si arose without an exact definition of the purity. Fifty years later the question remains quite open; 1 ppm (weight) was indicated as a possible target (i.e. 99.9999% Si) but most important was the content of donors (mainly phosphorous) and acceptors (mainly boron). One avenue to achieve this goal (moderate purity at low cost) was to make use of large scale, well proven metallurgical techniques, supposed to be cost efficient and scalable to the level needed for the future deployment of PV. The demand was calculated in millions of tonnes at the time EG silicon was produced in less than 10 000 tonnes a year. It was argued that the only way to achieve the large volumes was by the metallurgical route. SGS produced by the metallurgical route as opposed to Polysilicon by the chemical/CVD/Siemens process was designated as Upgraded Metallurgical Grade Silicon or UMGS or UMG. The chemical route through volatile silicon precursors was also challenged to develop alternative variants compatible in terms of purity and cost with the demand of terrestrial PV deployment. Fluidized Bed Reactor (FBR) as alternative to hot filament CVD in bell-jar Siemens reactors and new silicon precursors (e.g. monosilane SiH4) were proposed and became the topics of advanced R&D programs conducted in parallel with UMG for the benefit of the PV industry.Remarkable results have been achieved over the past 50 years. For more information on the history of silicon science, industry and market the interested reader is invited to consult our publications and the multiple references herein [2][3][4].Regarding the upgraded metallurgical route (or UMG) several strategies were pursued. Combining several steps appeared necessary to reach the optimal goal of purity and cost. Most consideration has been devoted to the following:• Selection of the raw materials (various types of silica ranging from ultra-pure quartz to silica rich rice husk) and reductants (including synthetic carbon black). This allowed to have low content of donors (P) and acceptors (B) in the silicon.• Alternative furnace to adjust to high purity raw materials, including two steps furnaces going via silicon carbide as intermediate.• Advanced refining techniques in ladle to remove by oxidation all elements less noble than silicon (e.g. Ca, Al, C). This includes gas, slag, plasma, electron beam treatment and more.• Aluminothermic reduction replacing carbothermic reduction.• Alloying silicon with other metals, e.g. Al, Ca, Ba, Mg and forming intermetallic phases or eutectics with silicon able to absorb and concentrate the impurities, dissolving these intermetallic phases and impurities by acid leaching. This technique was proven efficient to remove Fe and transition elements which all are detrimental to the solar cell performance. It was also proven to partially remove P.• Directional Solidification (DS) playing on the distribution of elements with low segregation coefficients. Obeying to similar mechanisms as for the formation of intermetallic phases in alloys, DS is shown particularly efficient for removing metallic (transition) elements as well as P and C.The metallurgical route never achieved the same level of purity as the chemical route to polysilicon, especially regarding P and B, respectively charge donor and acceptor to silicon. However, playing on "compensation" (neutralizing excess of one charge carrier by its contrary until the nett concentration of the suitable charge carrier is reached) and mitigating imperfection by "defect engineering" the industrial developers, hand in hand with academic researchers developed materials (wafers) and cell architectures performing at par with solar cells made of higher purity polysilicon. The best results were achieved when combining UMG with multicrystalline cast silicon, a far less costly and energy consuming technique than monocrystalline CZ silicon. Regarding UMG performance and limitation, in addition to the chapters in books referred in [2][3][4] more specific information is gathered in some others of our publications [5][6][7][8][9].This special issue of FRONTIERS delivers two review articles showing great details on this strategy.• Production of UMG silicon for low-cost high efficiency and reliable PV technology by José Manuel Miguez Novoa et al. [10]. In this paper, the authors present an innovative technology for purifying silicon, specifically designed for solar applications. This technology spans the entire value chain, from the raw material feedstock to the final photovoltaic (PV) system. At its core, the process involves several metallurgical steps, i.e. slagging, evaporation, and solidification that upgrade metallurgical silicon into upgraded metallurgical silicon (UMG-Si). This purified material contains low levels of impurities, including boron, phosphorus, and metals, all of which are compatible with solar cell production. The obtained multicrystalline Si ingots show uniform resistivity along their height due to the addition of gallium for compensated resistivity. Solar cells using this material have achieved efficiency levels of 18.4% for Al-back surface field (BSF) and 20.1% for passivated emitter and rear cell (PERC) technologies. These cells also show promises for future advancements, with potential efficiencies of up to 22% in next-generation solar cell designs. Beyond performance, UMG technology stands out for its environmental benefits. Compared to conventional polysilicon, it delivers a significant reduction in climate change emissions by over 20% and cuts the energy payback time by 50% [10]. These advantages make UMG-Si a cost-effective, low-CAPEX, and environmentally friendly alternative for producing PV silicon.• A review of defect mitigation strategies for UMG-Si wafers by Rabin Basnet and Daniel Macdonald [11]. In this work, the authors illustrate different defect mitigation strategies like tabula rasa, gettering, and hydrogenation for improving the performance of UMG-Si wafers and UMG-Si based solar cells. While UMG-Si still lags behind conventional silicon, some of these strategies can be applied together with further advancements in purification and ingot growth, which are expected to boost UMG-Si based solar cells performance. Continued research and detailed techno-economic studies will be key to fully realizing UMG-Si's potential as a cost-effective, competitive alternative in the solar industry.A third article proposes the use of magnesium as impurity dissolving agent completing the knowledge on more common calcium and aluminium as alloying metals.• Effect of Mg-alloying and cooling rate in the microstructure of silicon by Mengyi Zhu et al. [12]. They report on the effects of magnesium alloying (5.5 wt% and 9.0 wt%) and cooling rates (3, 10, 25, 40, and 80°C/min) on the microstructure of MG-Si. Faster cooling and higher magnesium concentrations reduced the primary silicon grain size, with faster cooling leading to more equiaxed grains and slower cooling producing elongated platelet shapes. The microstructure investigation shows an increased crystallographic orientation heterogeneity with higher cooling rates. This study emphasizes optimizing cooling rates and alloying content for efficient silicon purification, balancing cost and process effectiveness.Production of wafer from silicon ingots (both multi-and monocrystalline) generates lot of silicon losses, as small particles commonly called kerf. It is assumed that 35-40% of the incoming silicon feedstock ends up as kerf. The last development in sawing, i.e. abrasive free sawing using diamond coated steel wire and an organic lubricant, generates kerf of high chemical purity relatively easy to clean by acid washing as well as more uniform wafer thickness. The widely used former technique based on uncoated steel wire entraining SiC abrasives in a liquid suspension (slurry) generated a heavy contaminated silicon mud of low commercial value. The abrasive free kerf can be used into high value applications for silicon, e.g. to Li-ion battery anodes.The fourth article in this special issue illustrates this recent opportunity and shows how to exploit major knowledge acquired through UMG development.• Silicon kerf-loss as potential anode material for Li-ion batteries by Anne-Karin Søiland et al. [13]. This study shows that industrial silicon kerf particles, with a size of ~700 nm and a 1-2 nm oxide layer, can serve as an effective alternative to nano-sized silicon (40-100 nm) in lithium-ion battery (LIB) anodes. The kerf particles demonstrated comparable initial capacities, efficiencies, and cycling stability, maintaining performance beyond 120 cycles. The findings suggest that silicon kerf could offer a low-carbon-footprint alternative to nano-silicon, with potential for further improvements in cycling stability through methods like silicon-carbon composites or particle coatings.Initiating this special edition, we invited several scientists from academia and industry who we knew had been active and productive in the UMG field. However, the number of enthusiastic responses was rather disappointing. We regret that the present edition does not reflect the tremendous efforts devoted to the topic during the five past decades. Although regrettable, it is nonetheless well reflecting its current commercial status. In 2014 we published the list of the most advanced UMG projects. The companies we referred to were: JFE (Japan), Dow Corning (Brazil), Timminco (Canada), Elkem Solar (Norway), Jaco (China PR), Silicor Materials (Canada, USA), Photosil (France) Ferroatlantica/Ferrosolar/Ferroglobe (Spain), Evonik/Solsic (Norway). Several of these were already either terminated or on hold [8]. Ferroatlantica/Ferrosolar/Ferroglobe (Spain) and especially Elkem Solar/REC Solar (Norway) were after 2014 among those pursuing the most continuous efforts in R&D and industrialization. We supported a publication of Ferrosolar as late as in 2019 [9]. Ferrosolar was planning a plant in Spain which was unfortunately never completed. Elkem Solar commissioned in 2010-12 a commercial UMG plant in Norway with a capacity of 5000 tonnes which was several times temporarily shut down and reshaped to different process concepts. Elkem Solar sold the plant to REC Silicon who further sold it to Reliance of India. A last full attempt was done by REC Solar in Norway around 2020 making use of its UMG technology and plant equipment, turning some of the huge amount of silicon kerf generated by wafer manufacturing into single crystal feedstock.Announcements by the company concluded on the full validity and the overall benefit of the method, energy consumption and carbon footprint being advantageous with positive environmental and economic impact. The project demonstrated that recycling a significant part of the silicon waste into a semi-closed loop was possible and suitable. Assuming conservatively 30% of the feedstock ending up as kerf (worldwide roughly half a million of tonnes per year as of today) it is a golden opportunity for the industry to show social responsibility and reduce this way its carbon footprint by a significant factor. Nevertheless, it did not stop the owner of the company to stop all these efforts. The new owner shut down definitively the plant in 2023-24 putting an end on long lasting developments of UMG in Norway. To our best knowledge no one in our list of 2014 is currently producing solar grade silicon via the UMG route and no newcomer has since 2014 made any attempts to enter this market segment.It is symptomatic that the only article in the present edition dealing with kerf aims at serving a totally different business application, i.e. anodes to lithium-ion batteries. The article makes a point of achieving low carbon footprint by using wasted silicon from solar industry. It must be noted that the silicon market for battery anodes is still at the exploratory stage and until it really takes off, millions of tonnes of valuable solar grade silicon continue to be wasted by the solar industry.The development on UMG have taken place historically between 1975 and 2020 in Europe, USA and Japan, three regions which have fostered the PV pioneers. China showed interest for PV after 2000 but in a very short time (5 to 10 years) built up a word class industry surpassing by momentum and effectiveness all its western competitors. To move fast into the market the Chinese enterprises adopted mature and straight forward technology which they could purchase from the West. Regarding the silicon feedstock, the new entrepreneurs made the choice of polysilicon by the classical Siemens process which then gave the best guarantees in terms of critical purity to make the cells. India is now making similar attempts to penetrate the PV marked adopting a similar strategy as the Chinese 20 years ago.Recycling of materials from the silicon solar cell manufacturing is, however, important and will be increasingly necessary in the coming years. This includes recycling of silicon, crucibles, cells/wafers, metallic contacts, PV module glass and PV module structural materials. Recently, a review about the status of silicon solar cells and PV market development was published [14], which also gives some perspectives about the use of machine learning in the PV process and solar cell manufacturing as well as recycling.In general, it is recommended that the industry consider more use of UMG silicon where its economic and environmental benefits are clearly demonstrated. We fully recognize the merits and the need of other technologies, e.g. polysilicon via the Siemens and the fluidized bed reactors as well as single crystal ingots in making diversified and markets tuned solar systems. However, we think that banning solutions like UMG, multicast and kerf recycling to solar feedstock is counterproductive when these proposed solutions have proven outstanding performances on environmental impacts.In conclusion we call for a rethinking of solar silicon feedstock strategy considering a holistic approach and not only circumstantial interests.