A drop in the efficiency of a perovskite device tells an engineer only that the result has worsened. The problem may be the test light source, a particular spectral region, a local position on the substrate, or voltage loss in the material and interfaces. Each possibility requires different information before the team can choose its next experiment.

At the forum, 廖清霖 of Enlitech moved from solar simulators to external quantum efficiency (EQE), light-beam-induced current mapping (LBIC) and photoluminescence (PL). Viewed together, each measurement narrows a different part of the problem.

QuestionMeasurementWhat it narrows down
Were two efficiency measurements made against the same illumination baseline?Solar simulatorSpectrum, uniformity, temporal drift and calibration differences
Which wavelength range converts abnormally?External quantum efficiency (EQE)Absorption, charge collection or current matching between tandem subcells
Where on the substrate is the problem concentrated?Light-beam-induced current mapping (LBIC)Coating edges, scribe lines, contacts or local defects
Where is voltage potential being lost?Photoluminescence (PL) and voltage-potential analysisMaterial quality, non-radiative loss and the range of interfaces still to investigate

A solar simulator first establishes the comparison baseline. Ground operation, space, long-duration aging and tandem devices require different spectra and stabilization times. External quantum efficiency then separates the total current by wavelength, allowing a researcher to see where the response has fallen. Tandem devices are especially sensitive because the upper and lower subcells absorb different spectral regions; change the test spectrum and the current match moves as well.

LBIC then returns the problem to physical space. After a focused beam scans point by point, two devices with similar average efficiency may show entirely different distributions: one uniform, the other already darkening near a coating edge or scribe line. For a process engineer, the map first reveals where the abnormality is concentrated.

廖清霖 said LBIC equipment had once been more like a technology reserve inside the company, but had suddenly “become active” over the past year or two. He named India, Bangladesh, Vietnam and Thailand as places where more solar manufacturers had begun making inquiries. The talk described an increase in inquiries, not shipments or market share. It nevertheless suggested that spatial measurement was moving beyond research demonstrations and into discussions of process inspection.

The forum presentation stated that an LBIC scan of a 10 × 10 cm area could be completed in 40–50 seconds, while also mentioning roughly 1–2 million pixels. Enlitech's current LSD4 product page lists the same area as 100 × 100 mm, but at 50 μm resolution and in under four minutes. Change the pixel count, resolution, step size, dwell time or data processing and the same area will produce a different scan time.

Even 40–50 seconds is too slow for some production lines. 廖清霖 described another option as a more “direct and forceful” PL screen: first view a full-area emission image without waiting for a focused beam to finish scanning point by point, then decide which regions deserve closer inspection. The emission signal can also estimate the voltage potential the material can sustain under illumination and how much energy is lost through non-radiative pathways. Absorption, excitation, thickness and light extraction all affect brightness, so an abnormal region on the image does not by itself tell an engineer which layer is responsible.

The talk also demonstrated how to record emission over time and then connect luminescence, current and voltage measurements to trace where efficiency loss accumulates. For those results to be compared, the same sample still has to remain under consistent test conditions.

In 廖清霖's comparison, PL and LBIC are not two mandatory stations every sample must pass in sequence. When fast cycle time matters, PL can first identify non-uniform regions. If the problem then requires a local current distribution, LBIC can spend the time to scan them. Whether to investigate further depends on whether the previous image can answer the process question at hand.

Sources and further reading

Talk and organizer material

Company product information

Standard and primary paper