Chemical-mechanical co-design for scalable flexible perovskite manufacturing

INTRODUCTION

Flexible perovskite photovoltaics, enabled by low-temperature solution processing and compatibility with polymer substrates, have emerged as a promising platform for lightweight, conformable, and large-area energy harvesting. Roll-to-roll (R2R) processing represents a viable pathway for industrial deployment due to its continuous operation^[1,2], high material utilization^[3], and efficient multilayer integration^[4], in contrast to batch-based laboratory deposition, which lacks scalability, and large-area uniformity. Nevertheless, R2R fabrication introduces complex mechanical constraints that are rarely incorporated into the state-of-the-art perovskite chemistry design. During continuous coating and drying, the precursor layer is subjected to hydrodynamic shear^[5], extensional flow^[6,7], substrate tension^[8], pressure gradients^[9] within the coating head, and spatiotemporally varying evaporation fronts, which collectively reshape solute transport, alter solvation equilibria, and influence intermediate-phase formation and stress development throughout crystallization^[10]. As a result, ink formulations developed under static or near-static conditions often fail to accommodate the dynamic environments intrinsic to large-area coating^[11].

Against this backdrop, chemical-mechanical co-design can be understood as a manufacturing-oriented framework in which precursor chemistry, solvation equilibria, and crystallization pathways are intentionally designed in synchrony with the mechanical fields inherent to roll-to-roll processing, rather than optimized under static or mechanically decoupled conditions. From this perspective, the key bottleneck for scalable flexible perovskites lies not only in material efficiency, but in the lack of a framework that explicitly couples chemical design with the mechanical fields governing R2R processing. During continuous coating, these chemical and mechanical processes operate on comparable and often competing timescales, encompassing shear deformation, solvent evaporation, intermediate-phase conversion, and stress relaxation. Ink rheology determines shear distribution; shear bias nucleation and adduct conversion; evaporation gradients set supersaturation trajectories; and substrate tension governs residual stress accumulation, collectively determining the polycrystalline film morphology and device performance. Film formation in R2R processing should therefore be viewed as a cross-scale, co-evolutionary process, underscoring the need for chemical-mechanical co-design to enable uniform, defect-resistant, and industrially scalable flexible perovskite photovoltaics.

RETHINKING HOW FLEXIBLE PEROVSKITES ARE BUILT

As summarized in Figure 1A, numerous deposition methods (spin-coating^[12,13], blade-coating^[14,15], press-induced crystallization^[16,17], inkjet printing^[18,19], slot-die coating^[4,20], spray coating^[21], and thermal evaporation^[22]) have enabled high-performance laboratory devices, yet they operate as discrete, batch-style processes with limited control over large-area film uniformity or thickness variation. Transfer printing has also been reported as an alternative fabrication route in which perovskite crystallization is completed prior to mechanical handling, providing a useful conceptual contrast for distinguishing mechanically decoupled and flow-coupled manufacturing strategies^[23]. In contrast, Figure 1B illustrates the integrated R2R architecture, where each functional layer must be deposited with controlled thickness and stable interfaces under continuous substrate motion, i.e., requirements that become more stringent when using flexible transparent conductive oxide (TCO) substrates with low-temperature and mechanical-compliance constraints^[24-26]. Furthermore, the energy-level alignment in Figure 1C emphasizes that electronically optimal interfaces depend on precise composition and crystallinity (or molecular ordering levels), properties highly sensitive to coating hydrodynamics and drying behavior during continuous processing^[27-29]. Scalable flexible perovskite photovoltaics require co-engineered flow, drying, and crystallization.

Figure 1. (A) Schematic overview of representative deposition techniques used for perovskite film fabrication, including spin-coating, pressure-assisted crystallization, inkjet printing, blade coating, slot-die coating, spray coating, and thermal evaporation; (B) Illustration of a fully roll-to-roll (R2R) printing process for flexible perovskite solar cells and the corresponding device structure; (C) Summary of the energy-level alignment of representative ETLs, perovskite absorbers, and HTLs commonly used in perovskite solar cells. TCO: Transparent conductive oxide; IML: interfacial modification layer; HTL: hole-transport layer; ETL: electron-transport layer; FTO: fluorine-doped tin oxide; ITO: indium tin oxide; PCDTBT: poly[N-9’-heptadecanyl-2,7-carbazole-alt-5,5-(4’,7’-di-2-thienyl-2’,1’,3’-benzothiadiazole)]; PTAA: poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine]; P₃HT: poly(3-hexylthiophene); Spiro-OMeTAD: 2,2',7,7'-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9'-spirobifluorene; PVB-DAAF: polystyrene-functionalized 9,9-diarylfluorene-based triaryldiamine; PEDOT:PSS: poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate).

MECHANICALLY ASSISTED CRYSTALLIZATION OF PEROVSKITE

In continuous R2R processing, the chemistry of the perovskite precursor ink is inseparable from the mechanical fields imposed during coating. Conventional non-volatile Lewis-base coordinating solvents [Figure 2A], such as N,N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), γ-butyrolactone (GBL), and N-methyl-2-pyrrolidone (NMP) were originally optimized for spin-coating^[30,31], where slow adduct dissociation, antisolvent/high-vacuum quenching, and high-temperature annealing govern crystallization. Under R2R conditions, however, their strong Pb²⁺ coordination slows adduct conversion and stretches crystallization over minutes to hours^[32], whereas shear, extensional flow, and meniscus deformation occur on millisecond timescales. Recent process-resolved and in situ studies quantitatively confirm that crystallization kinetics in conventional high-boiling-point solvent systems lag coating-induced flow and shear by orders of magnitude under scalable printing conditions^[33-35]. As a result, crystallization rarely overlaps with mechanically active windows, motivating the use of volatile or reactive solvent systems to restore chemistry-mechanics synchronization. This temporal decoupling introduces three fundamental drawbacks: (i) Because nucleation begins only after the mechanical fields have dissipated, shear cannot influence early-stage ordering, resulting in random nucleation and spatially disordered grain formation^[36]; (ii) Without shear-crystallization overlap, the wet film loses hydrodynamic leveling during the onset of supersaturation, causing evaporation rates and solute redistribution to diverge across the coating width; this produces unbalanced drying, compositional heterogeneity, thickness variations, and ribbing-like defects over centimeter scales^[37]; (iii) Slow densification under non-uniform drying allows tensile stress to accumulate within the film, generating microcracks and limiting bending durability^[38], especially on flexible substrates that cannot tolerate the high-temperature annealing required to fully dissociate strong solvent-Pb adducts. Together, these limitations show that slow-crystallizing solvents inherently suppress mechanochemical feedback, limiting scalability in flexible perovskite manufacturing.

Figure 2. (A) Comparison of solvent systems used in perovskite precursor inks. Left: non-volatile Lewis-base solvents illustrated with a schematic of the ionization-solvation process leading to stable Pb-solvent adducts. Right: volatile/reactive solvent systems illustrated with the formation of transient, gel-like intermediates composed of dynamic colloidal dispersions; (B) Lateral distribution of mechanical shear stress during blade shearing and contrasting responses of volatile versus non-volatile precursor systems. From left to right: dimensional characteristics (DLS spectra), dominant chemical interactions (Raman spectra), rheological behavior (dynamic viscosity vs. shear-rate curves), and charge-transport characteristics (current-time transients)^[35]. Reprinted with permission. Copyright 2024, Springer Nature; (C) Multi-scale ordering enabled by chemical-mechanical coupling compared with chemically dominated (decoupled) crystallization. Top left: HRTEM images showing nanoscale lattice coherence. Top right: 2D GIWAXS patterns with schematic illustrations of orientational differences. Bottom left: cross-sectional SEM images revealing mesoscale grain alignment and film densification. Bottom right: PCE-area scalability of state-of-the-art perovskite modules and long-term operational stability of devices fabricated from volatile versus non-volatile precursor systems^[35]. Reprinted with permission. Copyright 2024, Springer Nature. DLS: Dynamic light scattering; HRTEM: high-resolution transmission electron microscopy; 2D: two-dimensional; GIWAXS: grazing-incidence wide-angle X-ray scattering; SEM: scanning electron microscopy; PCE: power conversion efficiency.

Volatile or reactive solvent systems such as acetonitrile-methylamine^[39,40] or methylamine-ethanol^[41] mixtures [Figure 2A] offer a fundamentally different pathway by generating transient, weakly coordinated precursor clusters that crystallize within the same hydrodynamic window as shear. Their rapid solvent release, occurring on the order of tens of milliseconds^[42], aligns with the shear rates of 10²-10⁴ s^-1 experienced during slot-die coating^[43,44], enabling the mechanical field to directly control the crystallization process. This synchrony arises from the distinct chemical nature of volatile solvent systems, in which the precursor evolves into a gel-like intermediate composed of a methylamine-stabilized colloidal network^[45]. These intermediates exhibit larger colloidal dimensions, hydrogen-bond dominated interactions, viscous rheological behavior, and a largely nonionic coordination environment, making them highly responsive to shear and more capable of reorganizing under hydrodynamic forces [Figure 2B]. When this synchronization is fully exploited, as exemplified by the use of volatile inks to form FA_0.75MA_0.25PbI_3-xCl_x films [Figure 2C], the resulting microstructure can exhibit an orientation factor of -0.31 (an 11× improvement), exhibit 1.5× higher mobility, and deliver 25.90% cell and 21.78% module (70 cm²) efficiencies^[35]. Overall, matched chemical and mechanical timescales allow shear to direct crystallization across atomic-to-mesoscale levels, enabling dense, aligned, and flexible films for R2R manufacturing. Beyond enhancing microstructural ordering and efficiency, mechanically assisted crystallization also imprints residual stress distributions, grain connectivity, and interfacial cohesion that persist into device operation. These features could influence the device reliability, including resistance to cyclic bending, crack initiation, and interfacial degradation^[46]. Mechanical fields during crystallization thus may be viewed not only as film-formation variables but as determinants of long-term device stability for industrial considerations. As a concrete reference for the proposed chemical-mechanical co-design framework, Table 1 compares representative solvent systems, deposition methods, and device performances in both rigid and flexible perovskite photovoltaics under scalable manufacturing conditions.

CONVERGENT ROADMAP TOWARD INTELLIGENT R2R

Volatile, shear-responsive precursors enable mechanically programmable crystallization, where shear, tension, bending, and evaporation are tuned to direct nucleation and grain evolution during R2R deposition. Because volatile ink systems synchronize crystallization with hydrodynamic timescales, mechanical fields can now act as ‘active inputs’ rather than constraints. Adjustment of parameters including tension, die pressure, substrate curvature, and shear gradients can therefore modulate nucleation density, grain orientation, intermediate-phase collapse, and stress relaxation simultaneously. In this framework, mechanical fields function analogously to “external stimuli” in colloidal systems: programmable, reversible, and capable of imprinting order from the atomic scale to the mesoscale, providing a powerful route to achieving dense, aligned, defect-minimized films that remain mechanically compliant under bending, key requirements for flexible R2R perovskite modules^[52].

Building on this mechanochemical foundation, real-time feedback and data-driven control offer complementary tools to stabilize and optimize the crystallization pathway across meter-scale manufacturing. In practice, such feedback can be derived not only from ink-side diagnostics, but also from direct, in situ or quasi-in situ monitoring of the photoactive layer during or immediately after film formation, using optical reflectance or scattering imaging^[53], photoluminescence (PL) mapping^[54], ultraviolet-visible (UV-Vis) absorption spectroscopy^[55], or grazing-incidence X-ray scattering techniques^[56], which have been shown to track crystallization fronts, phase conversion, and film uniformity in real time during scalable coating processes. Inline viscometry, interferometry, meniscus imaging, and hyperspectral monitoring further enable detection of viscosity drift, evaporative flux, and intermediate-phase signatures at sub-second temporal resolution.

From a process-design perspective, the objective is not to define universal operating values, but to ensure that the characteristic timescales of shear-induced flow, solvent evaporation, and phase conversion remain comparable during coating. These ranges are intended to be representative, literature-informed guidelines that emphasize relative timescale matching rather than prescriptive process parameters. Under conventional representative roll-to-roll or slot-die coating conditions (coating speeds of ~0.1-1 m·s^-1 and wet film thicknesses of ~10-100 μm)^[57], hydrodynamic residence times fall within the ~1-10 ms regime, whereas volatile or reactive solvent systems have been shown to compress evaporation and crystallization into sub-second to second-scale windows (typically ~0.1-1 s)^[45], based on scalable coating studies, thereby enabling crystallization to occur while mechanical fields are still active. Deep learning or adaptive control algorithms can then map these signals to actionable adjustments, such as altering flow rate, substrate temperature, solvent ratio, or tension, to preserve optimal shear and drying conditions. In parallel, recent progress in interlayer materials for large-area flexible perovskite devices has converged toward scalable polymeric and hybrid interlayers that simultaneously accommodate mechanical deformation by providing compliant stress buffering and stabilize buried interfaces, thereby aligning interfacial engineering with roll-to-roll manufacturing requirements^[58,59]. Finally, low-temperature curing or in-line molecular encapsulation immediately after crystallization can lock in mechanically imprinted microstructures, suppressing relaxation, oxygen ingress, and downstream stress accumulation. These sensing and control layers do not replace chemical-mechanical co-design; rather, they operationalize it by stabilizing shear-responsive crystallization windows in R2R. Together, mechanically programmable crystallization and real-time closed-loop control define a convergent pathway toward intelligent platforms delivering industrially scalable flexible photovoltaics.