Oxygen in Cz Silicon: The Defect the Ingot Makes and the Cell Line Reveals

Where the oxygen comes from, and what it becomes

Oxygen enters the silicon melt mainly by dissolution from the quartz crucible and is built into the crystal lattice as it solidifies. PV-Manufacturing.org puts typical Cz concentrations at roughly 5–10 × 10¹⁷ cm⁻³, with carbon around 5–10 × 10¹⁵ cm⁻³, and notes that oxygen solubility falls from about 10¹⁸ cm⁻³ at the melting point to several orders of magnitude lower at room temperature. The crystal therefore cools through a long window in which its oxygen is supersaturated and looking for somewhere to go.

Where it goes defines the defect. Some stays interstitial, where in boron-doped p-type material it takes part in the boron-oxygen complex behind light-induced degradation, whose recombination strength scales with the square of the oxygen concentration. Some forms small clusters that act as thermal donors, which the same source says can affect the resistivity of the material. Some precipitates. The Fraunhofer ISE authors add that interstitial- and vacancy-rich regions form depending on temperature gradients and pulling speed, and that these regions affect precipitation differently — so the growth recipe, not only the crucible, shapes the outcome.

• Treat crucible, hot-zone design and pull recipe as oxygen levers
• Remember oxygen solubility falls sharply as the crystal cools
• Separate interstitial oxygen, thermal donors and precipitates in any discussion
• Expect oxygen behaviour to vary along the ingot and across the radius

Thermal donors: misleading, usually curable

Thermal donors are the most common oxygen signature a wafer maker will meet. In the Fraunhofer ISE study, published in Photovoltaics International in 2012, in-line measurement of wafers specified at 1–6 Ω·cm returned base resistance values from 1.19 to 82.7 Ω·cm, and the authors attributed increased and inhomogeneous resistance in most cases to thermal donors. After an in-line thermal treatment at a peak of 800 °C for 10 seconds, all thermal donors had dissolved and resistivity returned to the expected range.

That makes thermal donors a measurement problem first. They can push as-cut wafers outside a resistivity specification that the silicon would actually meet after the first high-temperature cell step. The practical consequence is that as-cut resistivity readings on Cz material need to be interpreted with its thermal history in mind, and a resistivity dispute between wafer maker and cell maker should establish whether both sides are measuring the same thing before arguing about the silicon.

The caveat is the cell process. Thermal donors are cured by heat, and not every cell line supplies it: PV-Manufacturing.org describes silicon heterojunction manufacturing as running entirely below 300 °C. It follows that thermal donors present in the wafer may not be removed on such a line, so the donor state matters more for low-temperature cell customers — a point to raise with the customer rather than assume.

• Read as-cut resistivity on Cz wafers with thermal donors in mind
• Agree whether resistivity is specified as-cut or after a donor anneal
• Ask whether the customer's cell process has a high-temperature step
• Flag donor-sensitive supply for low-temperature cell lines explicitly

Precipitates: invisible as-cut, decisive after diffusion

Oxygen precipitates are the more dangerous form because they can be activated later. The Fraunhofer ISE team found that commercially available Cz wafers labelled high quality produced cells ranging from 14.4% to 18.4% efficiency on the same process, with the losers showing dark concentric rings in photoluminescence images. The ring defect strength correlated with efficiency at R² = 0.94. FTIR showed interstitial oxygen in a ringed wafer falling from 1 × 10¹⁸ to 1 × 10¹⁷ cm⁻³ through emitter diffusion, and preferential etching revealed dense precipitates in the low-lifetime centre. The same defect appeared in gallium-doped material, indicating it is independent of dopant.

The hard lesson is about inspection. As-cut, harmless thermal-donor rings and dangerous precipitate rings looked alike in PL images and in QSSPC lifetime. The authors concluded that where thermal donors cannot be excluded, critical wafers can only be identified reliably after emitter diffusion. In their two-year incoming inspection statistics, covering roughly 40,000 wafers, around a third showed rings and 4% formed critical precipitate defects. The authors say the figures do not claim to be representative; the data are also more than a decade old. The mechanism, however, is exactly what a new Indian wafer plant needs to design against.

• Do not rely on as-cut PL or lifetime alone to clear Cz wafers
• Use post-diffusion or post-anneal test samples for oxygen-sensitive lots
• Track ring patterns by ingot position to find the growth-side cause
• Treat 'high quality' labels as claims to be verified by lot data

Gettering: where oxygen helps

Precipitated oxygen is not only a liability. PV-Manufacturing.org notes that precipitated oxygen can facilitate internal gettering, in which metallic impurities form less recombination-active species; nickel and copper are described as effectively gettered by oxygen precipitates, while cobalt and lead hinder the effect. External gettering removes impurities from the bulk altogether: phosphorus diffusion collects highly diffusive metals such as chromium, copper, iron and titanium in the diffused region, and aluminium-silicon alloy regions act as a sink. The same source states that phosphorus gettering is more effective for iron than gettering by oxygen precipitates.

In integrated-circuit manufacturing, the Fraunhofer authors point out, precipitation is deliberately engineered to sit in the bulk under a precipitate-free denuded zone near the surface. Photovoltaics has no such luxury, because the whole thickness of the wafer is the active device. That is why the gettering benefit of oxygen is not a reason to tolerate it: for a solar wafer, the target is controlled oxygen and clean material, with gettering in the cell process treated as a safety margin rather than the plan.

• Know which gettering steps the customer's cell process includes
• Do not count on oxygen precipitates to fix metal contamination
• Keep metal contamination control upstream, in the hot zone and handling
• Remember the whole wafer thickness is active in a solar cell

What an Indian wafer plant should measure and agree

Oxygen belongs in the specification, with a stated method. SEMI MF1188 determines interstitial oxygen by FTIR from the 1107 cm⁻¹ absorption band using a short baseline between 1040 and 1160 cm⁻¹, over a range from 1 × 10¹⁶ atoms/cm³ to the solubility limit, on specimens from 0.4 mm to 4 mm thick. That thickness range sits well above a production solar wafer, which in practice points to measuring on dedicated samples cut from known ingot positions and correlating them with wafer-level lifetime and PL data.

The rest is traceability and agreement. With ALMM List-III bringing wafers into the approved-list framework from June 2028, Indian wafer plants will be supplying cell makers who judge them on cell-line yield. Agreeing up front how oxygen, resistivity and lifetime are specified and measured, at which process state, and on what sampling plan turns a potential yield dispute into a controlled parameter. Current ALMM requirements should always be confirmed directly with MNRE.

• Specify interstitial oxygen with a named method and sample position
• Correlate FTIR sample data with wafer lifetime and PL by ingot position
• Agree the process state at which resistivity and lifetime are judged
• Keep wafer lots traceable to crucible, hot zone and growth run