Perovskite Full-Stack Metrology: Why Thickness Alone Is Not Enough
A perovskite module is a stack of optically active layers, and each one fails differently. A metrology channel that reports one number can tell you that something moved. It cannot tell you which layer moved, and that is the difference between a process adjustment and a stop-the-line investigation. This guide covers the diagnostic patterns that only emerge when multiple optical outputs are read together.
Which signal sees which failure
The SE600X inline metrology platform publishes five output groups in a single acquisition: thickness (5–1200 nm), absorption, reflectance, transmittance, and PL as bandgap, peak intensity, peak position and FWHM — at 10 Hz sampling across 31 measurement points, with no sample preparation. Those outputs are useful mainly because they disagree with each other in informative ways.
| Stack element | What goes wrong | Signal that sees it | Signal that goes blind |
|---|---|---|---|
| ETL / HTL | thickness drift, roughness, incomplete coverage | thickness, reflectance | PL (buried layer contributes little) |
| Perovskite absorber | wet-film → dried thickness mismatch | thickness | — |
| Perovskite absorber | composition / halide ratio drift | PL bandgap, peak position | thickness alone |
| Perovskite absorber | crystallinity, grain-size spread | PL FWHM | thickness, T/R |
| Interfaces, defects | non-radiative recombination | PL peak intensity (used as a proxy for radiative efficiency) | thickness |
| Encapsulation, back contact | optical loss at the top interface | transmittance, reflectance | PL |
No single column of this table is measurable with a thickness-only tool. That is the entire argument for multi-output inline metrology.
Read the pattern, not the number
Diagnostic value comes from combinations. Treat these as hypotheses to validate on your line, not as substitutes for J-V or efficiency testing:
| Observed pattern | Most likely reading | First thing to check |
|---|---|---|
| Thickness ↓, bandgap stable, PL intensity stable | pure deposition drift | pump speed, gap, web speed |
| Thickness ↓, bandgap stable, PL intensity ↓ | right amount of the wrong film — coverage, pinholes | ink wetting, substrate energy, dryer vacuum profile |
| Thickness stable, PL peak position shifting (bandgap ↑) | composition drift | precursor ratio, solvent ratio, anneal atmosphere |
| Thickness stable, FWHM ↑ | crystallinity spread | anneal ramp, drying rate |
| PL intensity ↓ only at specific points across the web | localised, mechanical or wetting origin | die edge, slot-die clean lines, head temperature |
| T/R change with thickness unchanged | top-interface or encapsulant optical loss | later layers, not the coating |
A single-metric channel collapses all six rows into "out of spec."
Disclosure: The diagnostic table above is an engineering inference based on published optical principles, not a measured drift record from a specific production line. Validate each pattern against your own process data before using it for root-cause decisions.
31 versus 11 is a spatial-resolution statement
Evenly spaced points across width W have spacing W ÷ (N−1). So 31 points sample every W/30; 11 points sample every W/10 — three times finer, and 2.8× more measurements per substrate.
Why that matters in perovskite specifically: the patterns in the right-hand column above are mostly *spatial* — a coated edge, a die line, a local wetting failure. An 11-point grid can be blind to a feature between its nodes, and no amount of averaging recovers it. Point count is not a luxury spec; it sets the smallest spatial defect you can see at all.
Time works the same way: at 10 Hz, the interval between readings along the travel direction is `line speed ÷ 10`. If your line runs 60 mm/s, that is 6 mm.
No sample prep is a data-quality argument, not a convenience one
The published advantage is that sampling runs at 10 Hz with no sample preparation, and TDM also lists improved handling safety. The less obvious benefit is causal: an offline coupon is a different piece of glass than the module that failed. Every correlation dataset built by pulling samples has a hidden variable — the sample is not the product. Inline measurement removes that variable from the record you use to make process decisions.
What the SE600X does not do
Worth stating plainly, because a spec sheet that claims everything claims nothing:
- Its published outputs do not include n/k dispersion fitting. If you need optical constants on coupons, that is an offline station — SE100 or SE200 (which publishes n/k).
- It is an inline platform, not a vacuum-chamber in-situ channel (VCD process control module covers drying-stage measurement).
- It does not publish EL. Defect imaging at production speed is SE200x inline PL mapping and AOI.
What the Literature and Other Methods Report
Perovskite thickness measurement is not an unsolved question, and the published record is worth reading before choosing a channel:
- Ellipsometry is described in the peer-reviewed literature as the most appropriate technique for fast, accurate thickness measurement for large-scale production, and is used on complete MAPI multilayer stacks (AIP Advances, 2019; Texas State University thesis repository). Published absorber thickness: 125–350 nm; HTL: 15–65 nm.
- Electron transport layers are commonly reported at 15, 25, 30 and 40 nm in the 2025 MDPI review of film characterisation in perovskite solar cells — layers thin enough that the optical contrast, not the instrument resolution, is the limiting factor.
- In-line optical monitoring is established practice: Laytec's application note for an in-line perovskite thickness system reports accuracy "typically 1–2 %", with the sensor integrated at an optical viewport reading the moving substrate.
- EDXRF (Shimadzu) measures ITO and perovskite layer thickness non-destructively to the nanometre — an elemental route rather than an optical one, and therefore unable to report bandgap or FWHM.
Read together, these set the bar any inline optical claim should be judged against: a published accuracy figure, a layer-by-layer thickness budget, and a stated sample-prep cost. Where TDM does not publish a figure, it is stated as not published rather than filled in.
Frequently Asked Questions
How to determine film thickness?
For a perovskite stack, determine it layer by layer rather than as one number: the published outputs that let you do that on-line are thickness (5–1200 nm), absorption, reflectance, transmittance and PL (bandgap, peak intensity, peak position, FWHM), acquired together at 10 Hz across 31 measurement points. A single derived thickness tells you the stack moved; the combination tells you which layer moved.
How to calculate the thickness of thin film?
During coating, wet thickness is calculated from flow rate, coating width and web speed, then converted to dried thickness through solids fraction. On a finished stack you do not calculate thickness, you fit it: the measured spectrum is matched against a layer model, so the result is only as valid as the layer order and the n/k values given to the fit. Published layer thicknesses in the literature give the scale of the problem — SnO2 electron transport layers are commonly reported around 15–40 nm, with MAPbI3 absorbers in the 125–350 nm band.
Can ellipsometry be used to measure thickness?
Yes — and in the perovskite literature it is treated as the appropriate technique for fast, accurate thickness measurement at production scale. That is why the TDM platform that publishes n/k dispersion fitting is an offline bench (SE100 combines ellipsometry, reflectometry and transmission; SE200 publishes n/k). The SE600X inline platform does not publish n/k fitting — see What the SE600X does not do.
What are the downsides of using perovskite?
From a metrology point of view, the downside that drives instrument choice is that perovskite is a stack of optically active, process-sensitive layers: the absorber drifts in thickness, composition and crystallinity at the same time, and those three failures look identical to a thickness-only tool. The published industry accuracy for in-line perovskite film-thickness monitoring is on the order of 1–2%, which is a useful yardstick for any vendor claim, including ours.
Related guides and products
- SE600X Inline Perovskite Metrology
- SE200x Inline PL Mapping and AOI
- EL vs PL Inspection Guide
- Solar Manufacturing Real-Time Process Control
- PV Module Testing Solutions
