Product Introduction

A process engineer once described this scene to me.

One day, a PL image from a boron-diffusion sampling check suddenly showed a few wafers with obvious concentric ring striations. His first instinct was to pull the incoming inspection data for that batch: minority carrier lifetime above 1500 μs, oxygen precipitate absorbance passing, micro-defect density within spec. On paper, every light was green.

He called the lab for a routine EBIC recheck. Nothing showed up. Switched to preferential etching plus optical microscopy. Still clean.

But those rings on the PL map were still sitting right there. They didn't disappear.

Incoming inspection passes, recheck finds nothing, and PL still shows a dark circle. This three-way mismatch is one of the most common silent losses an N-type process engineer runs into.

The opponent behind it is what this article takes apart: concentric ring defects (CRD) in N-type photovoltaic Czochralski single-crystal silicon. It's one of the most underrated yield killers in N-type cells, and in the worst case it can eat 4% absolute cell efficiency.

From P-Type to N-Type, Engineers Switched Opponents

Let's clear up one thing first.

In the P-type era, the biggest old opponent on the wafer side was the boron-oxygen pair (BO defect): a B-Cz PERC cell under 12 hours of illumination could lose 3-5% absolute (the number reviewed in Vicari Stefani's 2022 PhD thesis). P-type multicrystalline silicon also had LeTID, which at its worst could drop 16%. The whole industry spent more than a decade fighting these light-induced losses, from PERC process tweaks to UV-filtering encapsulants on the module side.

In the N-type transition, the industry once thought this fight was over. N-type wafers are phosphorus-doped, so there's no mandatory B×O pairing and the BO defect simply can't form.

But people soon found out: BO was gone, and oxygen precipitates (OP) stepped up on their own. They just wore a sneakier disguise this time: concentric ring defects.

Li Guixiu from Zhejiang University (in Professor Yuan Shuai's group) presented on this at the 21st CSPV conference in 2025, and published related work in Applied Physics Letters in 2024. Together they lay it out clearly: the essence of the concentric ring defect is an oxygen precipitate that's a bit too small. Its three traits are all "invisible" by nature:

• Low electrical and chemical activity — not the kind of oxygen precipitate you spot at a glance

• Shallow defect level (0.42-0.46 eV, and even shallower after PDG)

• Invisible in native state — the as-grown wafer shows nothing; you have to finish high-temperature steps like diffusion and annealing before it appears

That last point is where engineers get burned: it's a "delayed developer." By the time you see it on the cell PL, the wafer step's accounts are already closed.

This Enemy Picks Its Weapon — Standard Gear Can't Touch It

Concentric ring defects overturn the traditional consensus that "if you can measure it, it's the enemy."

Point different weapons at the same wafer with concentric striations:

Translated into production line language, it's one sentence: this enemy picks its weapon. Standard gear can't touch it. On the line, the only tool that catches it daily is PL; to truly quantify it in the lab you need low-temperature EBIC or copper decoration.

That's also why so many engineers feel "the data all passed but the cell still slaps me in the face." The data isn't faked. The weapon in hand is wrong.

Technical Parameters

12 ppma: The Life-or-Death Line for N-Type Wafer Oxygen

Since the concentric ring defect is an oxygen precipitate, the source is the oxygen concentration [Oᵢ] inside the wafer.

Li Guixiu's report draws a very clear line: [Oᵢ] > 12 ppma enters the high-recombination-activity oxygen precipitate zone (the "black-core wafers" old engineers know well); [Oᵢ] < 12 ppma enters the small-size OP zone, which is the concentric ring we're talking about today.

12 ppma is the life-or-death line for N-type wafer oxygen (per the SEMI M6 standard for silicon materials, roughly 6×10¹⁷ cm⁻³). Industry data shows current mainstream single-crystal furnace technology can only reach about 12.5 ppma; push lower and yield drops sharply. The oxygen floor a wafer plant can hit lands right on the trigger line of the concentric ring defect. That's exactly why concentric ring defects are so common in the N-type era.

Li Guixiu's report gives a clear conclusion: in the worst case, wafers crossing 12 ppma [Oᵢ] can lose up to 4% absolute cell efficiency. "Worst case" here means the extreme situation of oxygen crossing 12 ppma + pull-rate fluctuation causing uneven vacancy distribution + head-and-tail ingot defects stacking. It's not an average; a real line more often sees losses on the order of 0.4-1%.

Worth noting: Li Guixiu's 2024 Applied Physics Letters study points out that even in wafers with oxygen below 7×10¹⁷ cm⁻³ (~14 ppma), concentric striations can still cause up to 0.86% absolute efficiency loss. That means the defect risk stays present even under 12 ppma. Holding 12 ppma is the bottom line, not the finish line.

What does 4% absolute mean on a production line? By 2026, N-type cell mass-production binned average efficiencies have spread into tiers: TOPCon at 25.6-26.2%, HJT at 26.0-26.5%, BC at 26.5-26.8%. A normally running line keeps shift-average fluctuation within ±0.05% absolute; once a batch average drops more than 0.1%, the line stops to investigate and calls a quality review. A worst-case 4% drop from concentric ring defects is equivalent to kicking a whole batch from the "mainstream tier" down to the "downgrade tier" or even "scrap tier" — a whole technology route's efficiency ladder gets punched through.

But for wafer and cell plants, the real pain in this ledger isn't power generation. It's that low-efficiency wafers can't be sold:

• Below the customer's minimum efficiency bin means instant dead stock: mainstream customers generally set N-type cell minimum bins at above 25.4% (some top customers set them higher). If a batch's average drops below 25%, the customer won't take it and it can only be consumed internally or scrapped

• Downgraded sales eat margin directly through bin price gaps: each bin down cuts price by a few cents to a dime per watt; over a batch of hundreds of MW, the gap can mean millions to tens of millions in gross profit evaporating

• Concentric striations found in sampling means full-batch traceback plus return risk: once customer-side EL/PL rechecks catch it, the accountability chain traces all the way back to the wafer plant

That's the ledger an engineer really watches — not "how much less power the plant generates," but "will the customer take this batch."

Why Did This Problem Suddenly Get Worse in the N-Type Era

The same thing existed in the P-type era, but it wasn't this much trouble. Three reasons amplify it in the N-type era.

Reason one: the thermal budget changed.

N-type cell thermal windows are a completely different system from P-type. P-type PERC phosphorus diffusion peaks at 800-850°C — not high, but combined with long high-temperature annealing it could partly repair small defects. In the N-TOPCon route, boron diffusion peaks pull up to 1000-1050°C — higher temperature, but with completely different dwell times and atmospheres, which instead more easily "activates" latent oxygen-related defects. HJT is more extreme: the whole flow is low temperature (around 200°C), losing any "high-temp anneal to dissolve defects" post-processing window. Once the wafer side has a hidden flaw, the cell side is nearly powerless to save it.

Reason two: bigger crucibles, worse oxygen introduction.

300mm large-diameter Cz + bigger crucibles + longer pulling cycles cause the total oxygen dissolving out of the quartz crucible to rise exponentially. In the ITRPV roadmap, the N-type wafer [Oᵢ] target line tightens year after year.

Reason three: low contamination makes the "old weapons" fail.

Oxygen precipitate problems used to rage largely because metal contamination amplified recombination activity. Wu Ruokai et al.'s 2025 paper in Solar Energy Materials and Solar Cells (DOI: 10.1016/j.solmat.2025.113739) quantified this with EBIC:

• Native oxygen precipitate (no contamination) → EBIC contrast ≈2% (nearly "invisible")

• Oxygen precipitate after iron contamination → EBIC contrast ≈12% (recombination activity up 6×)

In recent years metal contamination levels dropped sharply, which ironically made oxygen precipitates more "invisible." The black-core wafers old engineers could spot on PL by experience are gone, replaced by concentric rings that need specialized weapons to identify. This is the mismatch between the "metal contamination ledger" and the "oxygen ledger."

Note: saying "lower contamination makes oxygen precipitates more invisible" absolutely does not mean "more contamination is better." Once iron gets in, oxygen precipitate recombination activity explodes 6×, doing more overall harm. Reducing contamination is the right direction; it just makes "pure oxygen precipitate" risks harder to catch with old methods. So treating contamination and controlling oxygen are both required and can't replace each other.

Technical Advantages

Mechanism Translation: One Twitch in Pull Rate, One Ring of Striations

The most elegant part of Li Guixiu's report explains the concentric ring mechanism clearly.

In production line language: the concentric ring isn't caused by too much oxygen, but by uneven radial distribution of vacancies [V].

Li Guixiu's report uses CGSim simulation data to show that at a fixed pull rate, the radial vacancy concentration in a silicon ingot is naturally "high in the center, low at the edge," differing by more than an order of magnitude. FTIR measurements also confirm the [Oᵢ] radial distribution itself is quite uniform (center 6.0×10¹⁷ cm⁻³ vs edge 5.1×10¹⁷ cm⁻³). So the "ring" is drawn by vacancies, not by oxygen.

Oxygen precipitate nucleation needs "moderate [V]": too low and it can't nucleate, too high and it forms voids directly. When the pull rate fluctuates during pulling, the radial [V] distribution fluctuates with it, and the OP nucleation position drifts along the radius — that's how the ring of striations gets "drawn."

One line: steady pull rate, defects cluster; twitchy pull rate, defects ring.

Many line engineers mistakenly think the concentric ring means "more oxygen at the edge" and go tweak the hot zone oxygen path — wrong direction. The "ring" is drawn by vacancy fluctuation, not by uneven oxygen concentration.

Product Application

Three Lines of Defense: How the Production Line Fights This Battle

With the mechanism unpacked, here's the part engineers care about most: how to fight this? Ordered by investment from large to small, from far to near the line, concentric ring defects have three lines of defense.

Line one: source oxygen reduction (the harshest cut at crystal growth)

Core action: push [Oᵢ] below 12 ppma.

Li Guixiu's strongest evidence is MCz (magnetic Czochralski) measured data — with [Oᵢ] controlled at 4 ppma (~2×10¹⁷ cm⁻³), both the as-grown wafer and one after 750°C/16h + 1000°C/8-16h annealing show completely uniform radial [Oᵢ], and the concentric ring defect disappears.

The cost is blunt too: MCz needs a magnetic field system, raising ingot manufacturing cost. This defense suits top wafer makers on high-end N-type products; not every line can absorb it.

Line two: process stabilization (the daily homework at crystal growth)

Even without MCz, there's plenty to do:

• Pull-rate fluctuation control — the key is "steady," not "fast." Better to sacrifice a bit of pulling efficiency than let [V] fluctuate

• Nitrogen-doped pulling — measured data from Jinko's Wang Pengfei 2026 report: minority carrier lifetime up 7%, cell efficiency up 0.01%. Nitrogen molecules bind excess vacancies, suppressing void and oxygen precipitate formation, and later high-temp steps release the nitrogen again

• Shorten dwell in the 850-650°C window — during ingot cooldown, oxygen aggregates faster with vacancy assistance; this temperature window is a "defect incubator," so pass through it as fast as possible

Line three: incoming wafer screening (the cell plant's last gate)

How to screen incoming wafers? Wang Pengfei gives two hard metrics:

• Micro-defect density < 40 per mm²

• Oxygen precipitate absorbance < 0.5 (FTIR absorption peak at 1230 cm⁻¹)

For HJT processes, add two more:

• PL imaging to screen for "swirl-shaped dark zones" — the only visible evidence of the concentric ring defect on the wafer side

• Prefer two-step phosphorus pre-gettering (2nd PDG) over single-step — Wu Ruokai's paper verifies that even after PDG, defective-wafer PCE is still 0.4% absolute lower than standard wafers (defective 24.68% vs standard 25.08%, lab data). Though this is small-area lab cell data, the magnitude serves as a reference: 0.4% absolute on a mass line means a whole batch drops two bins, disrupting the product bin distribution and creating order-delivery problems — a loss far more painful than the "how much power" ledger

If the cell process allows, introducing a "defect-dissolving" anneal before boron diffusion (1100°C fast ramp, hold 10-30 minutes, fast cool) gives about 1000 PL brightness gain per Wang Pengfei's report, with an estimated 0.02-0.03% cell gain. This is the smallest change you can slot into an existing line.

Three Things the Report and Papers Don't Tell You

To close the technical breakdown, the boundaries of the papers must be made clear too.

First, "eating 4% efficiency" is the worst case after crossing the line. 12 ppma is a warning line, not "cross it and you definitely lose 4%." After oxygen crosses this line, if vacancy fluctuation stacks on, the loss floats between 0 and 4% absolute; 4% is the ceiling, and Wu Ruokai's paper shows the actual residual of defective vs standard wafers is 0.4% absolute. The three data layers relate like this: 4% is the extreme ceiling of line-crossing + vacancy fluctuation + head-tail stacking; 0.86% is the lab measurement when oxygen is slightly above 12 ppma (Li Guixiu APL 2024); 0.4% is the residual after PDG (Wu Ruokai 2025). The longer you're over the line and the more that stacks, the closer you get to that 4% ceiling. 12 ppma holds the bottom line of "don't enter the high-recombination-activity zone."

Second, the MCz cost ledger isn't detailed. Academic reports solve "can it be done"; engineers still have to calculate "is it worth it." At what scale of line does MCz break even? That depends on N-type cell premium room — currently HJT high-end product lines may support it, standard N-TOPCon still struggles.

Third, the coupling of nitrogen doping and HJT is under-covered in literature. Will nitrogen interact with hydrogen in the HJT process? Existing literature mostly validates on the N-TOPCon route; HJT-route data is still insufficient.

One-Line Summary

The P-type era was about "shaking off the BO pair"; the N-type era is about "locking down oxygen precipitates." The opponent changed disguise, so the engineer's weapons have to change too — PL imaging watches the site, low-temperature EBIC quantifies, [Oᵢ] < 12 ppma holds the death line, pull rate stays steady, two-step PDG backs it up.

The invisible killer isn't scary. What's scary is bringing standard weapons to fight it.

Ooitech's View

What strikes me here is how much of an N-type line's fate gets decided upstream, at crystal growth, long before any cell equipment sees the wafer. A concentric ring seeded by a twitchy pull rate can't be fully undone downstream, so the cell line is really inheriting a problem it didn't make. On our module production lines we see the flip side of this — good wafers wasted by process drift, or marginal ones saved by tight screening — which is why PL imaging discipline matters just as much on the module side as it does at incoming inspection. If you want to see how this plays out on a real automated line, our YouTube channel at www.youtube.com/ooitech has plenty of factory footage worth a look. Bottom line: hold 12 ppma, keep the pull steady, and trust PL over the paperwork.

References

Li Guixiu (Zhejiang University). Concentric Ring Defects in N-type Photovoltaic Czochralski Single-Crystal Silicon. 21st CSPV, 2025-11-27

Li G, Yuan S, Zhou S, et al. Separated striations in n-type Czochralski silicon solar cells. Applied Physics Letters, 2024, 125(25)

Wang Pengfei (Jinko Solar). PV Single-Crystal Silicon Quality Characterization and Defect Suppression. 2026

R. Wu, et al. Effect of phosphorus diffusion pre-gettering on electrical properties of oxygen-related defects in n-type crystalline silicon heterojunction cells. Solar Energy Materials and Solar Cells 290 (2025) 113739. DOI: 10.1016/j.solmat.2025.113739

B. Vicari Stefani. Investigation of Bulk Defects in p-type Silicon Wafers and Solar Cells (PhD Thesis), 2022