Atsushi Wakamiya opened by connecting materials research at Kyoto University with the work of bringing that research into practice at EneCoat. According to the company's public profile, EneCoat was founded in 2018 on research originating in the Wakamiya laboratory, and Wakamiya is one of its co-founders.

That background explains the way he framed the entire talk. High efficiency was only the starting point. His next question was why an academic result could begin to drift in production when the formula stayed the same but the raw-material lot changed or the coated area grew.

He traced the answer upstream until it reached something that appeared almost trivial: trace water in the precursor.

A trace variable can leave its mark across the entire process

Wakamiya used lead halide quality to show that chemicals carrying the same purity label do not necessarily enter a process in the same state. Trace water changes coordination in the solution and the crystallization that follows. He also described how dehydration or purification can improve lot-to-lot reproducibility.

Film formation then amplifies the difference. The precursor first coordinates with the solvent. An antisolvent induces nucleation and creates a solvent-containing intermediate, which only later converts under heat into the dark perovskite crystal phase. To the eye, the film merely turns black. Inside the material, several processes have already competed with one another: where nuclei appear, how quickly crystals grow, and when the solvent leaves.

“The same formulation” therefore has at least two meanings. Composition, concentration and ratio may be identical while water content, storage and preparation history differ. In a laboratory, that difference may produce a few poor devices. In production, it becomes a distribution across an entire lot, along with rework and scrap.

This was one of the most distinctive points in Wakamiya's talk: manufacturing control does not begin at the coating machine. If incoming-material specifications miss the state that actually governs crystallization, engineers may repeatedly adjust spin speed, temperature and annealing time while compensating for an upstream variable they have not identified.

Scaling means preserving the same crystallization pathway over a larger area

With small-area spin coating, a researcher can rely on experience to time the antisolvent and then select a usable region of the substrate. Once the process moves to printing, roll-to-roll production or gas quenching, that hand-controlled moment has to become a repeatable process window.

As area increases, temperature, airflow and solvent evaporation develop gradients across the substrate. A shift in time becomes a difference in space on the film. The central scaling question is therefore whether material at different positions can still pass through the same intermediate and crystallization pathway.

Later in the talk, Wakamiya extended this chain of control to p–i–n devices. The light-facing hole-selective contact must be thin and transparent while still transporting charge effectively. He compared conventional polymers with ultrathin self-assembled molecular contacts and discussed using multiple anchoring groups to improve interfacial interaction. In a laser-scribed module, local fluctuations in the film and interface can be amplified by the series-connected structure.

A 2023 JACS paper from Wakamiya's group provides a direct example. Molecules with multiple phosphonic-acid anchors adopt an orientation closer to face-on against the electrode, while molecules with a single anchor are more tilted. That difference in orientation affects how readily holes leave the perovskite layer. “Multiple anchors” are therefore more than a structural modification to the molecule: they turn interfacial orientation into a design parameter.

Precursors, crystallization, contact layers and modules are not four separate subjects. Together they answer one question: after the material passes through each transformation, has the reproducibility established earlier in the process survived?

In the final minutes, Wakamiya pulled the discussion away from crystals and interfaces toward application demonstrations, factory activity and team growth. This update shifted the scale of evaluation to manufacturing: all the preceding materials control ultimately has to support actual production and delivery.

One of those applications can be traced to a primary paper. In 2023, Wakamiya's group used a bilayer ITO structure with mutually compensating internal stresses to fabricate perovskite devices on a 4 μm PEN film. After 1,000 bending cycles, the devices retained 95% of their initial efficiency. The point is not another efficiency record. It is that once an application demands an ultrathin, flexible device, electrode stress and substrate deformation join the same process-control chain.

From understanding crystallization to protecting lot-to-lot yield

The Wakamiya group's 2014 crystallographic study already connected lead-halide complexes, water content and perovskite formation directly. Outside the talk, two studies published in 2025 pushed similar control problems into large-area processing. A Chinese–French team used crystallization control to make a slot-die-coated mini-module with an active area of 56.5 cm² and a certified efficiency of 20.3%. Another team at Germany's Karlsruhe Institute of Technology used optical signals collected during film formation and deep learning to predict efficiency.

The problem has moved from whether crystallization can be observed to whether the same indicator still works after the formulation, equipment or raw-material lot changes. A peak value from one large-area device can demonstrate a successful run. The distribution across lots—and the result after changing materials—shows whether the crystallization pathway has entered manufacturing. Wakamiya began with a trace amount of water in a precursor and ended at the same requirement: materials control must work more than once.

Sources and further reading

Talk and organizer material

Speaker and company context

Primary papers