Chiral 2D metal halide perovskites are a class of thin film materials designed for LEDs and data links that use the handedness of light. But nominally identical samples have produced absorption values spanning more than two orders of magnitude.
The reproducibility problem in chiral 2D metal halide perovskites has been hiding in plain sight. Labs around the world were making what looked like the same thin-film material and measuring absorption values for circularly polarized light that differed by more than two orders of magnitude. Some samples responded cleanly to left-handed light, others to right-handed, and many were nearly inert, with no consensus on why.
That uncertainty has been the central blocker for a class of materials that researchers want to use in spin-based LEDs, photodetectors, and optical data links that encode information in the handedness of light rather than its intensity. Without a way to predict which synthesis choices actually move the needle, every group has been running its own version of trial-and-error.
A team at Lawrence Berkeley National Laboratory (Berkeley Lab) has reframed the problem as a tractable engineering question. In a data-driven study published as a 2025 ChemRxiv preprint and noted in a Berkeley Lab press release as having a Matter journal version, the group built a workflow that maps the synthesis parameters of chiral 2D metal halide perovskites against their actual chiroptical performance.
The metric the team tracks is the absorption dissymmetry factor, or g_abs, which is the standard way to express how selectively a material absorbs left- versus right-handed circularly polarized light. The workflow pairs high-throughput thin-film fabrication with statistical correlation analysis and machine-learning methods run through CAMERA, Berkeley Lab's Center for Advanced Mathematics for Energy Research Applications. The goal is not to crown a single best recipe but to identify which knobs, including precursor chemistry, solvent and antisolvent choices, annealing temperature, and film thickness, actually control circular-polarization selectivity, and which ones do not.
The work was led by Carolin Sutter-Fella's lab at Berkeley Lab's Molecular Foundry, with co-first authors Raphael Moral and Maher Alghalayini. Crystallization was characterized with X-ray techniques at the Advanced Light Source, another Berkeley Lab user facility. The team describes the workflow as a practical guide that other labs can follow to tune their synthesis knobs, which is what makes the work portable: any group working on these materials can now aim at the same variables.
That framing matters because the underlying problem is not a defect in any individual experiment but a coordination problem across the field. Two groups following superficially similar recipes were not, in fact, running comparable experiments. The small choices that distinguish one lab's protocol from another's were quietly driving most of the spread.
This is a roadmap, not a finished product. The >2-orders-of-magnitude reproducibility gap that originally blocked the field has not been closed by any single paper, including this one. What has changed is the search space. Other groups now have a shared, transferable method for turning synthesis parameters into predictable chiroptical performance, so progress can compound rather than reset with each new lab.
The stakes extend past the lab. Circularly polarized light can encode and transmit data with built-in handedness, which is why a thin, cheap, tunable film that responds reliably to it is attractive for next-generation displays, sensors, and optical interconnects. None of those applications can ship on materials whose performance is unrepeatable.
Two things to watch. First, the underlying paper is currently a ChemRxiv preprint with a Matter journal version referenced in the Berkeley Lab release. The peer-review status will tighten once that DOI is confirmed; the Sutter-Fella Lab publications page will show whether it has been posted. Second, the >2-orders-of-magnitude spread cited here is the authors' own characterization of the field, not an independent audit, and there is no third-party replication yet. If outside labs follow the mapped synthesis knobs and still see that wide a spread, the model will need to be revised.
For a category of materials that has been stuck on reproducibility for years, that test is exactly the point.