Kerf Loss: India's Biggest Silicon Loss Is Happening Now, Not in 2050

The loss is structural, not a yield problem

Every wire that cuts silicon destroys silicon. PV-Manufacturing.org notes that the older slurry process used hard steel wire around 180 µm thick and produced a kerf loss of approximately 200 µm per cut, and that the industry has now moved fully to diamond wire, which cuts a narrower kerf, works faster and leaves roughly half the saw damage. Diamond wire improved the ratio. It did not remove the loss.

The scale is easy to underestimate because the material leaves as a slurry rather than as scrap you can see. A 2024 review of hydrometallurgical recovery from diamond wire saw silicon powder puts the loss at approximately 35–45% of the silicon entering the cutting process, and estimates global generation at 679,000–873,000 tonnes in 2023 against about 1.94 million tonnes of crystalline silicon consumed — a quantity the review values at over USD 20 billion at the pricing it assumed. Earlier work on carbon in kerf, from 2017, put the figure at around 40% of ultrapure silicon lost during slicing.

For a wafer plant, this is not a yield problem to be tuned out. It is a structural property of the process. The question worth asking at design stage is not how to eliminate it, but what happens to it.

• Treat kerf as a designed-in output stream, not an operational defect
• Diamond wire narrowed the kerf; it did not remove the loss
• The material leaves as slurry, which is why it is easy to under-account
• Decide the kerf handling route before the plant layout is frozen

Why the powder is hard to put back

Kerf is not simply fine silicon. At an average particle size around 0.5 µm it carries enormous surface area, which makes it reactive, oxidation-prone and difficult to handle. The 2024 review identifies iron, nickel and aluminium as the primary metallic contaminants, and traces them: the diamond wire is stainless steel coated with nickel and iron, and aluminium comes from organic additives used in cutting and from the backplane material that holds the ingot during the cut.

Carbon is the harder problem. The 2017 study attributes it to the sawing lubricant — polyethylene glycol with a mass-average molar mass of around 400–500, sometimes combined with diethylene glycol — and measured total carbon at 1.3% of the sample, with total organics at 4%. Its central finding is that carbon will not simply be baked out. Under inert nitrogen, 12% of the original total carbon content survived heating to 900 °C as a tarry residue, because bond cleavage competes with radical recombination that builds heavier residue. In air the carbon can be removed completely, but at the cost of oxidising the silicon, which defeats the purpose.

That contradiction — inert atmospheres leave carbon, oxidising atmospheres consume silicon — is why kerf recovery is a chemistry problem before it is an equipment problem.

• Sub-micron particle size drives both the reactivity and the handling difficulty
• Fe and Ni trace to the wire itself; Al to additives and the mounting backplane
• Carbon from PEG-based lubricant resists pyrolysis and is the binding constraint
• Cutting-fluid and wire choices upstream decide the contamination you must remove

What published recovery routes have achieved

The literature is further along than the industry. The 2024 review compares routes and reports that direct acid leaching in a combined hydrochloric, sulphuric and hydrofluoric acid system — at 343 K, nine hours, with a liquid-to-solid ratio of 1/10 — achieved metal removal efficiency of 94% or better and a final purity of around 5N. A second route, hydrofluoric acid pretreatment at 40 °C for two hours followed by directional solidification, reached 5.5N across approximately 82.5% of the resulting ingots at a recovery yield of 86.89%.

The review's own conclusion is the important part: acid leaching followed by pyrometallurgy is currently the most effective process, but reaching 6N solar-grade silicon remains challenging. That is the honest state of the art. Recovered kerf is closer to a metallurgical-grade or battery-anode feedstock today than to a drop-in replacement for polysilicon in a Czochralski puller.

Anyone reading a recovery proposal should hold it against those numbers. Laboratory yields on prepared samples are not plant yields on live slurry, and a purity claim means nothing without the element list and the measurement method behind it.

• Ask for the element list and method behind any purity claim, not just an N number
• Distinguish laboratory yield on prepared powder from plant yield on live slurry
• 6N solar-grade from kerf is not yet a settled industrial outcome
• Consider adjacent offtake routes before assuming a return to the crystal puller

India's framework was written for modules, not for kerf

India does have a solar-waste framework, and it is tightening. Solar PV waste is categorised as e-waste under Chapter V of the E-Waste (Management) Rules, 2022, though it is exempt from the EPR recycling targets that apply to other electronic waste. CPCB issued draft guidelines on 4 June 2025 for the safe storage, handling and transportation of discarded solar PV modules, panels and cells, with mandatory registration, storage permissions to 2034-35, annual return filings and compliance with CPCB standard operating procedures.

Read the scope closely and the gap is visible. That framework is built around discarded modules, panels and cells — the end-of-life stream. The projections driving it are end-of-life projections: a 2023 CSTEP study cited alongside the guidelines expects 4.5 million tonnes of Indian PV panel waste by 2050, and IRENA work cited in 2026 puts global PV waste at 78 million tonnes by 2050 with India among the top five creators.

Kerf is a different stream with a different clock. It is produced on day one of wafer production, it is a process residue rather than a discarded product, and as India builds domestic ingot and wafer capacity ahead of ALMM List-III in 2028, the volume arrives well before the module waste does. A plant being designed now should not assume that the module-waste rules answer its kerf question, and should confirm its obligations with the relevant pollution control board directly rather than by analogy.

• Do not assume module end-of-life rules settle a manufacturing residue stream
• Kerf volume arrives with first production, not with first decommissioning
• Confirm handling and disposal obligations with the state pollution control board
• Track the framework — draft guidelines and rules are still evolving

Silicon from end-of-life modules: the second stream

Discarded modules are the other source of recoverable silicon, and they pose a different problem from kerf. The silicon in a finished cell is already crystalline and fairly pure, but it is laminated behind glass and polymer encapsulant and coated with aluminium, silver, lead-bearing solder and an anti-reflective layer. Recovery is mostly a question of stripping those layers off cleanly.

Indian research has worked on that route. pv magazine reported in 2021 that researchers at the KPR Institute of Engineering and Technology described a chemical sequence that avoids hydrofluoric acid: sodium hydroxide to remove the aluminium layer, nitric acid for the silver electrodes and lead, and phosphoric acid for the silicon nitride coating, after the outer module layers were removed by heating. The researchers claimed recovered silicon purity of up to 99.9984%. The work was published in the International Journal of Photoenergy; the coverage did not describe industrial-scale trials.

Even at the researchers' best case, that is roughly four nines. Polysilicon for Czochralski growth is specified to far higher purity, so silicon recovered from old modules is, like kerf, better thought of as a feedstock that needs further refining than as a drop-in substitute for virgin polysilicon. Whether it can be economically upgraded for ingot growth is still an open question in the published literature.

• End-of-life silicon is cleaner than kerf but buried under glass, encapsulant and metals
• Reported Indian lab route: HF-free chemical stripping, purity claimed up to 99.9984% (researchers' claim)
• Not a substitute for polysilicon in a Czochralski puller without further refining
• Module waste volumes arrive on a decommissioning clock; kerf arrives with first production

Crop ends, tails and squaring slabs: the loss before the saw

Kerf is not the only silicon that never becomes a wafer. Before a Czochralski ingot reaches the saw, its seed end and shoulder (the cone where the crystal widens to full diameter) and its tail cone are cropped off, and in solar practice the round ingot is squared into a brick, which removes curved side slabs. A pull that loses its single-crystal structure part-way, for example after a dislocation, can also leave a length of material that is not usable as monocrystalline wafer stock.

Unlike kerf powder, these pieces are solid, large and come straight from a crystal grown out of high-purity polysilicon, so in principle they can be cleaned and returned to the crucible as part of a later charge. Whether that is acceptable depends on three things: the dopant they already carry (a boron-doped p-type piece is not a neutral input to an n-type pull), surface contamination picked up during cropping and handling, and how impurities segregated along the crystal, since the last-to-freeze tail end tends to hold more of them than the body.

That is why off-cut recovery is a material-control decision rather than a free gain. Plants that remelt their own crop typically keep it segregated by dopant type and resistivity, etch or clean it before reuse, and track how much recycled material goes into each charge, because a contaminated input can cost a whole pull. The same purity chain that runs from high-purity quartz sand through the crucible to the hot zone sets how much of that scrap remains worth melting again.

• Crop: seed end, shoulder and tail cone cut from every Czochralski ingot.
• Squaring slabs: curved sides removed when a round ingot is cut into a solar brick.
• Failed or structure-lost sections from pulls that did not stay single-crystal.
• Kerf: fine powder lost in the saw itself, much harder to reuse (see above).

Decisions worth taking before the plant is built

Most of what determines whether kerf is recoverable is decided upstream of any recovery equipment. Cutting fluid chemistry sets the carbon burden. Wire specification sets the iron and nickel load. Mounting and backplane materials set the aluminium. Segregation discipline decides whether the slurry stays a single clean stream or becomes a mixed one that no process can economically sort.

The practical sequence for a new Indian wafer project is to design the capture and dewatering route into the layout rather than retrofitting it, to keep kerf segregated by cutting campaign so that material is traceable to a wire and fluid specification, to establish the contamination baseline by actual assay rather than by assumption, and to treat any recovery partner's claims as a qualification exercise with its own test plan. None of this requires committing to a recovery process on day one. It requires not foreclosing one.

• Design slurry capture, dewatering and storage into the layout from the start
• Specify wire, fluid and backplane with the downstream contamination in mind
• Segregate and trace kerf by cutting campaign so assays mean something
• Qualify any recovery partner against your own assay, not their brochure
• Keep the option open — avoid choices that make recovery impossible later