Upgraded Metallurgical-Grade (UMG) Silicon: What It Is, Why It Faded and What It Means for India

Two routes from metallurgical silicon to solar silicon

Every solar wafer starts as quartz reduced with carbon in an arc furnace to make metallurgical-grade silicon (MG-Si), at roughly 98–99% purity. That is far too impure for a solar cell, so it has to be refined. The dominant route is chemical: MG-Si is turned into trichlorosilane, distilled, and deposited back as ultra-pure silicon in a Siemens reactor or as granules in a fluidised bed reactor. PV-Manufacturing.org describes the Siemens process as the most common purification route, and also as energy-intensive and hazardous.

Upgraded metallurgical-grade (UMG) silicon takes the other road. It never leaves the solid and liquid state. Instead it removes impurities from molten silicon by metallurgical steps. A 2024 review in Frontiers in Photonics by the developers of one UMG process lists slagging, vacuum evaporation and directional solidification as the core tools, and says the route has the potential to cut energy use and cost compared with the Siemens process.

• Chemical route: MG-Si → trichlorosilane gas → Siemens rods or FBR granules
• Metallurgical route: MG-Si → slag, vacuum and solidification refining → UMG silicon
• Both start from the same quartz and the same arc-furnace silicon

How UMG purification works

Most metals prefer to stay in liquid silicon rather than freeze into the solid. Directional solidification uses this: molten silicon is frozen slowly from the bottom up, metals are pushed toward the last part to freeze, and that top layer is cut away. PV-Manufacturing.org describes one patented sequence of directional solidification, remelting in an arc furnace under argon, and then plasma treatment, in which an oxidising stage strips boron and carbon from the melt surface and an argon or argon-hydrogen stage removes oxygen as SiO gas.

Two elements resist this approach. Phosphorus has a segregation coefficient of about 0.35, according to PV-Manufacturing.org, so a large share of it freezes into the crystal rather than staying in the melt, which makes it very hard to remove by solidification. UMG processes therefore add steps such as vacuum evaporation. Boron is likewise removed by oxidising slag or plasma treatment rather than by solidification. In practice, these two elements decide whether a UMG process works.

The compensation problem

Boron is a p-type dopant and phosphorus is n-type. When both are left behind in small amounts, they partly cancel each other, and the silicon is called compensated. The Frontiers review gives the Ferrosolar UMG specification as boron below 0.2 ppmw, phosphorus at 0.1–0.3 ppmw and total metals below 0.5 ppmw.

Compensation brings two problems. First, the two dopants enter the growing crystal at different rates, so the net doping and therefore the resistivity can drift strongly from the top to the bottom of an ingot. The Ferrosolar team adds gallium, at 5–10 ppmw, during crystallisation to hold resistivity near 1 Ω·cm along the ingot height. Second, a wafer carrying both dopants behaves differently from one with a single dopant, so cell makers have to tune processes such as gettering to it. The same review reports that phosphorus diffusion gettering raised bulk lifetime from about 10 µs to a best of 722 µs.

• Boron (p-type) and phosphorus (n-type) partly cancel: compensated silicon
• Different segregation means resistivity can vary along the ingot
• Gallium co-doping and gettering are the usual counter-measures

What the published results show

UMG can make working, efficient cells. The Frontiers review reports Al-BSF cells averaging about 18.4%, PERC cells at a representative 20.1% with a record of 20.8%, and outdoor modules holding a performance ratio above 0.97 after two years. It also projects that more than 22% is possible on a TOPCon-based design; that figure is modelled, not measured. The authors include employees of the companies that developed the process, so these are best read as the developers' own results.

On carbon, the claims are attractive but come from producers. REC Solar Norway, formerly Elkem Solar, stated in 2021 a footprint of 11.2 kg CO2-eq per kg of silicon, verified by EPD Norway, and said its metallurgical route used 75% less energy than the Siemens process, PV Tech reported. The Frontiers review found about 20% lower climate-change emissions for its UMG Al-BSF modules than for polysilicon modules, using Spanish grid electricity.

Why UMG stayed a niche

UMG’s case rests on beating chemical-route polysilicon on energy and cost. As chemical-route polysilicon has scaled up and its price has fallen, that gap has narrowed, while UMG's doping variability sat badly with the move to high-efficiency cells. Most results above are on Al-BSF and PERC, and the industry has moved on to n-type TOPCon, heterojunction and back-contact cells, which demand long carrier lifetimes and tight resistivity windows.

The European producers show the pattern. The Frontiers review says Ferrosolar's planned industrial plant in Puertollano, Spain, was stopped because of unfavourable market conditions. REC closed its Norwegian polysilicon operations at Kristiansand and Porsgrunn in November 2023, citing electricity prices, pv magazine reported, and sold REC Solar Norway to Elkem in May 2024, pv magazine India reported. In the public record, low-carbon claims have not so far turned into large-scale UMG supply.

What it means for India

India already makes metallurgical silicon and has quartz resources, and it is planning domestic polysilicon. That makes UMG a fair question in a feedstock strategy discussion. The honest answer from the public record is that UMG is a specialist route: it can suit cost- or carbon-focused niches and multicrystalline or p-type products, but the mainstream n-type wafers India's ALMM List-III chain will supply are built on chemical-route polysilicon.

For ingot makers, the practical lesson is about feedstock specifications. Any non-standard or blended silicon should come with stated boron, phosphorus and metals limits, lot traceability and a resistivity test on pilot ingots before it enters production. This guide is general awareness; feedstock choices should be validated with cell customers, suppliers and test labs.