Universal crystallization kinetics of solution-processed semiconductors Authors: Zhitian Ling, Mateusz Brzezinski, Shuanglong Wang, Konstantinos Bidinakis, Lei Gao, Chongyao Li, Xin Zhou, Hai I. Wang, Stefan A. L. Weber, Paul W. M. Blom, Wojciech Pisula, Tomasz Marszalek Published: 2026 ABSTRACT Heterogeneous crystallization kinetics critically govern the morphology and performance of solution-processed semiconductors, yet morphology control is often achieved through trial-and-error optimization. Quantitative understanding of crystallization is essential for precise morphology control but is limited by the simultaneous occurrence of heterogeneous nucleation and crystal growth. Here, a time-defined induced nucleation strategy has been developed to initiate crystallization at specific stages of solution processing. Systematic variation of nucleation timing reveals three distinct morphological regimes, whose emergence is explained by the concentration-dependent critical cluster radius in bulk, interfacial, and cusp environments. On this basis, a quantitative, dimensionless model is developed for heterogeneous crystallization kinetics across diverse processing conditions. The model demonstrates universality across perovskite and organic semiconductors. In situ studies reveal identical growth kinetics for induced and spontaneous nuclei, validating its generality. The resulting morphology control enables tuning of crystallinity and charge transport in solution-processed transistors. Introduction Solution-processed semiconductors have been widely exploited for scalable, low-cost, and flexible electronic devices, including solar cells, light-emitting diodes, and field-effect transistors (FETs)1–3. The morphology of the semiconductor film, particularly the size, orientation, and distribution of crystalline domains, plays a decisive role in determining charge transport and overall device performance4,5. While empirical strategies have enabled partial control over film morphology, the underlying crystallization kinetics remain not fully understood. A predictive and quantitative description of the collective impact of nucleation and growth on the microstructure is therefore critical for both advancing material functionality and deepening fundamental insight into thin-film formation. Among the various physical processes that contribute to thin-film growth, heterogeneous nucleation plays an outsized role that governs the spatial and temporal onset of the crystallization, directly influencing size, density, and orientation of domains6,7. Unlike homogeneous nucleation, which occurs uniformly, heterogeneous nucleation is triggered by local perturbations or interfaces, turning it highly sensitive to processing conditions and consequently challenging to predict8–10. While many studies have attempted to control crystallization for improving the film quality and device performance through solvent, additive or processing engineering, the individual contributions of nucleation and growth are not taken into account11–13. Common models offer valuable insights into the overall crystallization, but fall short in resolving crystallization of individual domains to describe domain-specific growth kinetics under heterogeneous conditions. For instance, the Lamer model explains the homogeneous nucleation, but lacks the spatial and temporal resolution needed to accurately describe heterogeneous nucleation events14–16. The Avrami model captures the overall crystallization kinetics, but averages out domain-to-domain variability17–19. Models related to the late stage of the crystallization process like Ostwald ripening focus on grain coarsening and are not applicable to early-stage thin-film formation20–22. The main challenge in describing the crystallization of individual domains is the occurrence of parallel existing subprocesses: nucleation and growth of different domains occur in overlapping and unpredictable timeframes23,24. Consequently, systematically decoupling the nucleation time from the growth stage via experimental strategies is crucial for establishing a comprehensive model capable of quantitatively describing the heterogeneous crystallization, which is essential for both mechanistic insight and control over the film morphology25–28. To address these challenges, we have established a time-defined induced nucleation (TDIN) strategy that enables deterministic control over the heterogeneous nucleation and allows a separate analysis of the nucleation and growth mechanisms. In this approach, artificial nucleation sites are mechanically induced into the film after defined spin-coating times, followed by thermal annealing to continue the subsequent crystal growth. The systematic variation of the coating time exposes three well-defined morphological regimes: large domains (LDs), unidirectional domains (UDs), and small domains (SDs), each emerging within a distinct spin-coating time. The transitions between these regimes occur as short as a few seconds, which is mechanistically understood based on the classical nucleation theory (CNT). To quantitatively describe the growth, we construct a piecewise model based on the mass transfer theory, capturing concentration changes during the coating and annealing stages. The model reliably describes the domain evolution for diverse materials (e.g., 2D/3D perovskites, organics, and salts), independent of specific processing conditions, and accurately predicts crystallization trajectories for both induced and spontaneous nucleation events. Based on the understanding of the crystallization mechanism, film growth controlled by induced nucleation exhibits nearly one order of magnitude higher charge carrier mobilities in field-effect transistors than the spontaneously crystallized ones. This result emphasizes the critical correlation between precise crystallization control and device functionality. While classical crystallization theories describe an ensemble-averaged behavior, the present work establishes a model to resolve and control the evolution of individual domains under heterogeneous conditions. The developed model quantitatively describing the crystallization process of solution processed semiconductors into different film morphologies provides a fundamental basis to control and optimize the film processing for specific electronic applications. Results and discussion In most solution-processed materials, nucleation remains a stochastic event, governed by complex and poorly defined interfacial and concentration conditions. To address this challenge, we have developed a TDIN strategy, which is based on induced heterogeneous nucleation at defined timepoints during solution processing, effectively decoupling nucleation from subsequent crystal growth (Fig. 1a). In this approach, a mechanical scratch is applied to the spin-coated film after a specific spin-coating time to artificially induce nuclei, followed by thermal annealing to initiate crystal growth. The exact moment that the nucleation is induced, termed the induced nucleation time (ts0), determines the crystallization pathway and final film morphology. Early-stage nucleation yields large domains (LD), intermediate nucleation generates unidirectional domains (UD), and late-stage nucleation results in small domains (SD). These regimes reflect distinct crystallization kinetics during spin-coating.Fig. 1Time-defined mechanically induced nucleation enables deterministic control over crystallization and thin-film morphology.a Schematic illustration of the time-defined induced nucleation TDIN process during spin-coating, producing large domains LDs, unidirectional domains UDs, and small domains SDs. Nucleation is mechanically induced at different coating times to define nucleation onset, followed by thermal annealing to continue crystal growth. b Three types of heterogeneous nucleation sites in a wet film: bulk (within the liquid), interface (liquid/air or liquid/substrate boundary), and cusp (triple-phase solid-liquid-gas region at the contact line). c Change in Gibbs free energy (ΔG) for nucleation at different sites, schematically illustrating how nucleation barriers vary between bulk, interface, and cusp locations. Changes in surface energy (ΔGs) and volume energy (ΔGv) are also indicated. The interface curve represents heterogeneous interfacial nucleation in a general sense; the nucleation barriers at the liquid/air and liquid/substrate interfaces may differ depending on the material system and interfacial properties. d Red curves represent theoretical dependence of critical nucleus radius and cluster size on concentration for bulk (solid), interface (dotted), and cusp (dashed) regions. The blue curve describes the change in cluster radius with the concentration. Cs is the solubility; Ci denotes the critical concentration for mechanically induced nucleation; Cn represents the concentration for spontaneous nucleation at the interface, which is located between Ci and Css due to local variations in interfacial states; Css is the critical concentration for spontaneous bulk nucleation. e Schematic illustration of the relation between concentration and spin-coating time. Blue, red and yellow boxes correspond to LD, UD and SD morphologies, respectively. f Optical microscopy images of (PEA)2SnI4 films nucleated at coating times from 34–52 s, demonstrating the transition between LD (blue box), UD (red box), and SD (yellow box) morphologies. The scale bar is 300 µm. g Demonstration of the universality of the TDIN strategy across various material systems, yielding distinct morphologies (left: LD, middle: UD, right: SD) of 2D/3D perovskite (MeOPEA)1.8Cs0.1SnI3.9, 3D perovskite MAPbI3, organic molecule TIPS-pentacene and inorganic salt NaH2PO4. The scale bars are 400 µm for (MeOPEA)1.8Cs0.1SnI3.9, TIPS-pentacene and NaH2PO4, and 100 µm for MAPbI3. Time-defined mechanically induced nucleation enables deterministic control over crystallization and thin-film morphology. a Schematic illustration of the time-defined induced nucleation TDIN process during spin-coating, producing large domains LDs, unidirectional domains UDs, and small domains SDs. Nucleation is mechanically induced at different coating times to define nucleation onset, followed by thermal annealing to continue crystal growth. b Three types of heterogeneous nucleation sites in a wet film: bulk (within the liquid), interface (liquid/air or liquid/substrate boundary), and cusp (triple-phase solid-liquid-gas region at the contact line). c Change in Gibbs free energy (ΔG) for nucleation at different sites, schematically illustrating how nucleation barriers vary between bulk, interface, and cusp locations. Changes in surface energy (ΔGs) and volume energy (ΔGv) are also indicated. The interface curve represents heterogeneous interfacial nucleation in a general sense; the nucleation barriers at the liquid/air and liquid/substrate interfaces may differ depending on the material system and interfacial properties. d Red curves represent theoretical dependence of critical nucleus radius and cluster size on concentration for bulk (solid), interface (dotted), and cusp (dashed) regions. The blue curve describes the change in cluster radius with the concentration. Cs is the solubility; Ci denotes the critical concentration for mechanically induced nucleation; Cn represents the concentration for spontaneous nucleation at the interface, which is located between Ci and Css due to local variations in interfacial states; Css is the critical concentration for spontaneous bulk nucleation. e Schematic illustration of the relation between concentration and spin-coating time. Blue, red and yellow boxes correspond to LD, UD and SD morphologies, respectively. f Optical microscopy images of (PEA)2SnI4 films nucleated at coating times from 34–52 s, demonstrating the transition between LD (blue box), UD (red box), and SD (yellow box) morphologies. The scale bar is 300 µm. g Demonstration of the universality of the TDIN strategy across various material systems, yielding distinct morphologies (left: LD, middle: UD, right: SD) of 2D/3D perovskite (MeOPEA)1.8Cs0.1SnI3.9, 3D perovskite MAPbI3, organic molecule TIPS-pentacene and inorganic salt NaH2PO4. The scale bars are 400 µm for (MeOPEA)1.8Cs0.1SnI3.9, TIPS-pentacene and NaH2PO4, and 100 µm for MAPbI3. To understand the mechanistic basis of the TDIN strategy, we first discuss the formation of clusters that act as potential nucleation sites in a wet film during this event. The nucleation sites can be located in the bulk (homogeneous nucleation), the liquid/air or liquid/substrate interface (heterogeneous nucleation), where nucleation may preferentially occur depending on the material system, and the cusp-like regions (induced nucleation), illustrated in Fig. 1b. In classical nucleation theory (CNT), the change in Gibbs free energy (ΔG) for the cluster formation consists of two competing terms (Equation S51) graphically presented in Fig. 1c: a positive surface term (ΔGS) scaling with the square of the radius (r), and a negative volume term (ΔGV) scaling with the cube of r. A spherical approximation of the cluster is used for conceptual and analytical simplicity29,30. Although the actual crystal growth may deviate from an ideal spherical geometry, the key conclusions of the model rely on the relative evolution of the nucleation barrier and critical radius rather than the exact nucleus geometry. At very small cluster sizes, the ΔGS dominates, resulting in an increase in the overall ΔG. As the cluster grows, the ΔGV governs, lowering the total ΔG. The transition point, the critical radius (r*), marks the smallest cluster size at which further growth becomes energetically favorable, such that clusters can grow further to a site enabling nucleation. Compared to the bulk (solid line), clusters at the interface (dotted line) bear a reduced ΔGS, lowering ΔG and finally reducing the barrier to form larger clusters with subsequent nucleation. Clusters at the cusp (dashed line) show the lowest barrier due to the largest interfacial contact, turning nucleation most favorable in this region. To understand the nucleation kinetics, it is important to realize that both the cluster size (r) and critical radius r* are related to the concentration (C) of the solute in the solution in a reciprocal manner, as shown in Fig. 1d. The critical radius decreases with concentration, as derived from Supplementary Fig. 1a and explained by Equation S52. The analysis in Fig. 1d provides three curves: bulk (red solid), interface (red dotted), and cusp (red dashed), each representing a different surface energy (ΔGS). In contrast, r in solution (blue curve) increases with C and intersects the three curves at key concentrations: critical concentration for mechanically induced nucleation (Ci), concentration for the spontaneous interfacial nucleation (Cn), and critical concentration for spontaneous bulk nucleation (Css). According to CNT, nucleation begins only when r exceeds r*. These thresholds directly map to the morphological regimes LD, UD, and SD in Fig. 1a, as discussed below. The relation between C and ts0 in Fig. 1e follows a LaMer-like trend (detailed explanation is given by the description of Supplementary Fig. 1b). The concentration first increases during solvent evaporation, followed by a decrease as nucleation and growth consume solute, eventually returning to the solubility (Cs) limit31,32. Mapping Ci and Css from Fig. 1d into this curve divides the TDIN process into three time regimes, each with a distinct film morphology upon induced nucleation. (1) Blue regime (C < Ci, short ts0) At short times, not much solvent has evaporated yet, and the concentration of solute in the wet film is still low. In this regime, r (blue curve in Fig. 1d) is smaller than r* at the cusp (red dashed curve), and therefore nucleation cannot be mechanically induced. The formed small clusters of solutes at this low concentration will just disassemble back into the solution. During subsequent annealing, C gradually increases, and spontaneous interfacial nucleation is initiated only when r exceeds r* at the interface (red dotted curve at Cn). By this time, the scratch tip has left the wet film, and the continuous cusp line is no longer active for crystal growth. Consequently, the LD morphology is formed via delayed spontaneous nucleation. (2) Red regime (Ci ≤ C