Regulating the morphology of two-dimensional perovskite single-crystals via high-throughput experimentations Authors: Jingyan Qi, Wei Meng, Kang An, Xujie Hui, Liqun Liu, Yuguang Ma, Ning Li Published: 2025 ABSTRACT AbstractControllable solution-based synthesis of semiconducting micro-structured materials with tailored morphologies and specific properties is crucial for the development of cost-effective microelectronic devices. Here, we present a high-throughput solvent evaporation crystallization experimental approach that integrates in-situ absorption spectroscopy and molecular dynamics simulations to systematically explore ligand-mediated crystallization dynamics in 2D perovskite single crystals. Based on automated experimental platform, 98 synthesis conditions were systematically investigated, including 9 alkyl ammonium ligands, 5 aromatic ammonium ligands, and varying precursor solution dilutions. We establish a clear correlation between organic molecular structure and perovskite morphology, showing that shorter ligands promote 1D nanowire formation, while longer ligands favor 2D nanosheet growth. In-situ spectroscopy and molecular simulations reveal that larger ligands induce conformational changes within the perovskite lattice, shifting crystallization from direct nucleation to lamellar exfoliation. Transmission electron microscopy (TEM) and density functional theory (DFT) calculations confirm that such transition is driven by enhanced solute–solvent binding energy, which modulates the crystallization pathway of lead halide intermediates. Our findings provide valuable insights into solution-phase crystallization kinetics and offer a rational strategy for designing perovskite materials with tailored optoelectronic properties, facilitating their scalable integration into advanced semiconductor applications.Subject terms: Organic-inorganic nanostructures, Synthesis and processing Introduction The relentless advancement of information technology has propelled the demand for micro and nanoscale semiconductor electronic devices. Conventional miniaturization techniques face high costs, environmental concerns, and scalability constraints limitations, such as vacuum deposition, epitaxy and etching1. Solution-processed mesostructured semiconductors provide a promising alternative, offering cost-effective fabrication with precise control over composition and structural design, making them highly attractive for next-generation optoelectronic and photonic applications. Among solution-processed semiconductors, two-dimensional (2D) layered metal halide perovskites (L2An-1MnX3n+1) have garnered significant attention owing to their promising characteristics, such as high charge carrier diffusion length and high defect tolerance. Here, L represents a bulky organic cation, A a small organic cation (e.g., methylammonium, MA), M a metal cation (e.g., Pb or Sn), X a halide anion, and n an integer dictating the inorganic layer thickness, effectively functioning as a quantum well2,3. The tunability of structural and electronic properties makes them ideal candidates for mesoscopic optoelectronic applications4. The dimensionality of perovskites fundamentally influences charge transport and light–matter interactions, thereby determining the optoelectronic performance5. For instance, perovskite nanowires, with 1D electron confinement, exhibit ultrahigh carrier mobility and enhanced sensitivity, rendering them suitable for photodetectors and lasers. Conversely, 2D perovskite nanosheets, characterized by large surface areas and strong quantum confinement, enhance light absorption and charge separation, benefiting solar cells and photocatalysis6. Therefore achieving precise control over crystallization is critical for tailoring perovskite morphologies to optimize the semiconductor functionality7,8. Various synthetic methodologies have been developed to fabricate perovskite materials with tailored dimensionalities, including vapor-phase growth, lithographically templated solution-phase growth, ligand-assisted solution-phase growth, and alcoholic solvent-assisted growth9–16. In solution-phase growth, the reduction in saturation due to temperature drop and the increase in solution concentration to saturation due to solvent evaporation under heating conditions are the two major methods driving crystal formation. However, methodologies developed so far largely rely on empirical synthetic experiments, necessitating iterative functional group modifications to organic ligands to regulate crystal morphology. These methods suffer from high processing complexity, elevated costs, and limited scalability, restricting their applicability mainly in specific materials design. A fundamental challenge remains: the lack of generalizable principles governing ligand-induced crystallization in 2D perovskite systems. The complex interplay of multidimensional parameters further complicates precise morphology control, underscoring the need for a deeper understanding of perovskite crystallization kinetics17. Here, we present a high-throughput solvent evaporation crystallization experimental approach integrating in-situ absorption spectroscopy and molecular dynamics simulations to elucidate ligand-mediated crystallization in 2D perovskite single crystals. Our automated system systematically explores 98 synthesis conditions with programmable robotic arms, enabling precise control over crystal morphology. By screening nine alkyl ammonium ligand, five aromatic ammonium ligand, and varying precursor solution dilutions, we establish a direct correlation between ligand chain length and crystallization behavior: shorter ligands favor 1D nanowire formation, whereas longer ligands promote 2D nanosheet growth. In-situ absorption spectroscopy and molecular dynamics simulations reveal that increasing ligand size modulates the ligand conformation within the 2D perovskite lattice, shifting crystallization from direct nucleation to lamellar exfoliation. Transmission electron microscopy (TEM) and density functional theory (DFT) calculations further confirm that morphological transition is driven by enhanced solute-solvent binding energy, governing the crystallization pathway of lead halide intermediates. Our study provides insights into solution-phase dynamics governing perovskite lattice formation and offers a rational strategy to design perovskite materials with tailored optoelectronic properties, advancing their integration into next-generation semiconductor technologies. Results High-throughput synthesis of 2D perovskite nanorods and nanoplates We selected a total of 14 ligands throughout the high-throughput experiments (Supplementary Fig. 1), including alkylammonium ligand ranging from tert-butylammonium to dodecyl ammonium, i.e., alkylammoniums with main chains ranging from 2 Cs to 12 Cs as alkyl ligands and aromatic ammonium ligand ranging from phenyl ammonium to phenyl butylammonium, i.e., aromatic ammoniums with phenyl-ring side-chain Cs ranging from 0 to 4. The stock solution was diluted with a mixture of three antisolvents, which were chlorobenzene (CB), acetonitrile (ACN), and o-dichlorobenzene (DCB) to reduce the crystal density of the single crystals18. 2D perovskite single crystals were prepared by the drop injection method (Fig. 1a). We have recently set up a fast and versatile high-throughput experimental platform for the development of optoelectronic materials and devices (Supplementary Fig. 2). The platform employs an 8-channel pipetting robot (FCA, Flexible Channel Arm), in which four pipetting fingers are able to precisely prepare different perovskite precursor solutions in 24-well plates at arbitrary ratios. In addition, a robotic gripper arm (RGA) is responsible for transferring the substrates between the various functional modules of the platform in order to complete the different experimental steps. Since four pipetting tips can simultaneously aspirate and dispense liquids in different volumes, high-throughput automated workstations can increase experimental efficiency by at least fourfold. It can perform 240 drop-coating operations and repeat the entire high-throughput experiment 2 times per hour with prominent reliability and accuracy. For alkyl ligands, tert-butyl ammonium19 and isopropyl ammonium20 were chosen, since the molecular size of ethyl ammonium is in the range of 3D perovskites, and it is not able to be used as an intermediate layer to form 2D perovskites21. The absorption spectra of the 2D perovskite films prepared with 14 ligands showed that the absorption peaks of the films are located at 300-400 nm and 510 nm (Supplementary Fig. 3a), the peaks of the PL spectra of the films are in the range of 485-555 nm (Supplementary Fig. 3b), and the fitting lifetimes of the TrPL spectra of the films are in the nanosecond order of magnitude (Supplementary Fig. 3c and Supplementary Table 3), indicating that all of the selected organic ligands could form 2D perovskite structures22. In the experiment, the dilution ratio of nine alkyl ammonium ligand perovskite precursor solutions was set at seven ratios ranging from 60 to 960 times, and the dilution ratio of five aromatic hydrocarbon ligand perovskite precursor solutions was also set at seven ratios ranging from 30 to 480 times (Supplementary Fig. 4). We set different dilution concentrations to screen out the most favorable crystallization concentration for investigation, and eliminated the influence of the aggregation property of the solute in the solution on the morphology of single crystals. We set four temperatures for the stock solution under the optimal dilution conditions of each two-dimensional perovskite and screened out the optimal crystallization conditions (Supplementary Fig. 5). The variable parameter space is composed of three variables: ligand molecule type, solution dilution ratio, and crystallization temperature, which can yield 154 sets of high-throughput experimental results (Fig. 1b). In 154 sets of crystallization experiments, 14 characteristic crystals were screened out based on crystal morphology and single crystal integrity, corresponding to 14 different types and molecular sizes of ligands, as shown in Fig. 1c. High-throughput experimental procedures and results. a Flowchart of high-throughput experiments. An automated platform robotic arm was used to draw up the solution and drop-cast it onto a silicon substrate for evaporative crystallization. b 98 high-pass experiments were designed with 14 ligand types and 7 precursor solution dilution times as variables. c The morphology of two-dimensional perovskite crystals gradually changed from rod-like to lamellar with the increase of the size of ligands in the characteristic crystals of 14 kinds of ligands. d The length of the ligand molecules. e Rod and flake crystals were defined by calculating the ratio of length to diameter of the crystals. Each perovskite included 12 data on the aspect ratio of single crystals. The statistical histogram of the ratio of length to diameter of a single perovskite crystal was prepared by using the coordination system with the limit of 1.5. The white shaded regions correspond to the formation of rod-shaped single crystals by the ligands, and the gray shaded regions correspond to the formation of sheet-shaped single crystals by the ligands in d, e. Scale bars, 10 μm in c. Among 14 2D perovskite crystals, with the continuous growth of ligand carbon chain and the gradual increase of ligand volume, the morphology of 2D perovskite single crystals gradually changes from rod-like to sheet, as shown in Fig. 1d. When the ratio of length to diameter of perovskite crystals is above 1.5, the crystal form is considered to be rod-like, and when the ratio of length to diameter is below 1.5, the crystal form is considered to be sheet23,24. Among the alkyl ammonium ligands, the 2D perovskite single crystal morphology formed by tert-butylammonium (t-BA), isopropyl ammonium (i-PA), n-propylammonium (PA) and n-butylammonium (BA) ligands with smaller volume have rod-like morphology. Five large ligands, pentylammonium iodide (PentA), hexylammonium (HA), octylammonium (OA), nonylammonium (NonylA) and dodecylammonium (DDA), were formed in sheet morphology. Among the aromatic ammonium ligands, the 2D perovskites formed by phenylammonium (PhA) and phenylmethylammonium (PMA) are rod-like crystals, while the 2D perovskites formed by phenylethylammonium (PEA), phenylpropylammonium (PPA) and phenylbutylammonium (PBA) showed sheet-like morphology (Fig. 1e). Among the 14 ligands, both alkyl ammonium ligands and aromatic ligands showed a rod to lamellar transition with a gradual increase in ligand length. The perovskite single crystals formed by the crystallization of different ligand precursor solutions, whether rod-like or sheet-like, exhibit stable green light emission under the fluorescence field (Fig. 2a to d). The length of rod-like 2D perovskite single crystal reach 15 μm, and the size of sheet-like single crystal reach 10 × 10 μm. Under the electron microscope, the single crystals showed clear rod-like (Fig. 2e and g) and sheet-like (Fig. 2f and h) morphology. Meanwhile, the photoluminescence peaks of each 2D perovskite single crystal measured by single crystal PL are 527 nm (Fig. 2i), 508 nm (Fig. 2j), 530 nm (Fig. 2k), 538 nm (Fig. 2l) respectively, indicating the pure 2D (n = 1) perovskite crystal obtained25. The microscopic imaging, PL mapping and corresponding PL spectra of the other 10 types of 2D perovskite single crystals are shown in Supplementary Fig. 6. Optical characterization of two-dimensional perovskite single crystals. a–d Fluorescence microscopy images of a (BA)2PbI4, b (DDA)2PbI4, c (PMA)2PbI4, d (PEA)2PbI4 single crystals. e–h, SEM images of e (BA)2PbI4, f (DDA)2PbI4, g (PMA)2PbI4, h (PEA)2PbI4 single crystals. i–l PL spectra of i (BA)2PbI4, j (DDA)2PbI4, k (PMA)2PbI4, l (PEA)2PbI4 single crystals. Scale bars, 10 μm in a to d, 2 μm in e, 10 μm in f, 5 μm in g, 1 μm in h. In-situ absorption spectroscopy We used in-situ absorption spectroscopy to monitor the evaporation crystallization process of precursor solutions baked at 70 °C (Supplementary Fig. 7), to explore the crystallization process of 2D perovskites formed by different ligands. In the process of evaporation and crystallization, the droplet on the Si sheet moves due to the volume contraction of the solution, resulting in a phase of signal absorption disturbance in the test. The duration of this phase varies from 3 to 10 s depending on the type of solution used. Among the 14 ligands, the in-situ spectra of the representative short-chain ligand BA, and the long-chain ligand DDA showed different features. The two stages before and after signal disturbance were divided into Stage I and Stage II, representing solvent evaporation stage and stable crystallization stage respectively, as shown in Fig. 3a, d. The characteristic absorption peak of perovskite is at 520 nm, and PbI2 has the strongest absorption of light at 425 nm (Supplementary Fig. 8). For BA, the absorbed signal of PbI2 at 425 nm and the absorbed signal of (BA)2PbI4 at 510 nm both appear at Stage II, as shown in Fig. 3b. For DDA, the absorption signal of PbI2 at 425 nm has appeared in Stage I, while the absorption signal of (DDA)2PbI4 at 510 nm does not appear until the time reaches Stage II. As shown in Fig. 3e, the absorbed signal at 425 nm does not increase, while the absorbed signal at 510 nm increases. In-situ absorption analysis. a In-situ absorption spectra of (BA)2PbI4 crystallization show the simultaneous emergence of PbI2 characteristic peak around 445 nm and (BA)2PbI4 peak around 500 nm during Stage II. b Spectra at 180 s and 200 s reveal concurrent signal enhancement for both PbI2 and (BA)2PbI4, supporting the direct nucleation mechanism for perovskite formation. c Schematic diagram of the direct nucleation mechanism and equation. d (DDA)2PbI4 crystallization exhibits only the PbI2 peak at 445 nm in Stage I, with the perovskite characteristic peak at 525 nm appearing later in Stage II. e At 225 s, PbI2 absorption is prominent, while at 255 s, (DDA)2PbI4 absorption increases significantly, indicating a PbI2 nanocrystal stripping mechanism. f Schematic diagram and equation of the stripping mechanism. g Binding energies between 14 ligand ions and CB molecules in precursor solution increase with ligand size. h–i The different binding energies of BA and DDA with CB lead to varied CB dissociation rates, regulating the instantaneous solubility of PbI2 and thereby inducing distinct perovskite crystallization pathways. Based on the different absorption signal characteristics, we conclude that short-chain ligands and long-chain ligands have different crystallization processes. For short-chain ligands like BA, the PbI2 absorption signal at 425 nm and the (BA)2PbI4 absorption signal at 510 nm both appear at Stage II, indicating that neglectable PbI2 crystallization occurs during the solvent evaporation process. In the stable crystallization stage, PbI2 and (BA)2PbI4 crystals begin to appear simultaneously, which is dominated by (BA)2PbI4 direct nucleation mechanism. The growth of (BA)2PbI4 crystal is directly induced by (BA)2PbI4 crystal nuclei in solution26, as shown in Fig. 3c, which is represented by Eq. (1). For long-chain ligands like DDA, the absorption signal of PbI2 at 425 nm appears first in Stage I, while the absorption signal of (DDA)2PbI4 at 510 nm does not appear until Stage II, indicating that more PbI2 crystals appear first in the solution during solvent evaporation. (DDA)2PbI4 crystals appear in large quantities during stable crystallization. It is assumed that the process is dominated by the stripping mechanism of PbI2 crystals. In solution, the first PbI2 nanocrystals are stratified27, and the detached PbI2 sheets induce the insertion of DDA+ into the [PbI4]2- octahedral structure28, and further formation of (DDA)2PbI4 crystals, as shown in Fig. 3F. The process is represented by Eqs. (2–4). where Eq. (2) represents the generation of PbI2 nanocrystals, Eq. (3) represents the process of layer exfoliation of PbI2 nanocrystals, and Eq. (4) represents the process of formation of 2D perovskites by DDA+ insertion. The in-situ absorption spectra of (BA)2PbI4, (DDA)2PbI4, and 12 other ligands during crystallization are shown in Supplementary Fig. 9a to n. For the ligands with smaller sizes, such as t-BA, i-PA, PA, BA, PhA, PMA, low PbI2 signals appeared in the absorption spectra at the evaporation crystallization stage, and no obvious perovskite absorption appeared. In stable crystallization stage, the absorption intensity of PbI2 increases, and the perovskite absorption signal appears. However, larger ligands, such as PentA, HA, OA, NonylA, DDA, PEA, PPA, PBA, PbI2 absorption signals appear obviously in the evaporative crystallization stage. In stable crystallization stage, perovskite absorption signal appears based on PbI2 absorption signal. The variation characteristics of the two absorbed signals correspond to the direct nucleation mechanism and the PbI2 crystal stripping mechanism respectively29. Crystallization mechanism We used molecular dynamics simulations to explore the relationship between different types of ligands and solvent binding energies. The kinetic state of the precursor diluent is simulated at a drop casting temperature of 343 K (Supplementary Table 2). Among the four solvents, DMF with strong solvation energy can dissolve perovskite precursors effectively, but its crystal growth kinetics are slow. Other three antisolvents (CB, ACN, and DCB) have weak solvation energies and limited solubility for precursors, but have the potential for rapid crystallization30. By comparing the crystallization of the ternary anti-solvent and pure CB (Supplementary Fig. 5b), we demonstrated that a small amount of ACN and DCB did not affect crystalline morphology. Therefore, in the solution, the component that plays a crucial role is CB, which constitutes the largest volume proportion. CB, as a polar anti-solvent, the chlorine atom end exhibits low potential, which attracts the amino end of the ligand ions in the solution, as shown in Supplementary Fig. 9. We calculated the intermolecular binding energies of the 14 ligand ions in solution with the CB molecules in solution to determine the change of the solution system during the heating process, which is shown in Eq. (5): where A = t-BA+, i-PA+, PA+, BA+, PentA+, HA+, OA+, NonylA+, DDA+, PhA+, PMA+, PEA+, PPA+, PBA+, and B = CB, with EAB as the overall energy of AB, EA as the energy of A, EB as the energy of B, and Eint as the interaction energy between A and B. We obtained the relationship between the binding energy between 14 alkane and aromatic ligands and CB with ligand size (Fig. 3g), where Eint(BA+−CB) is −16.92×103 kcal/mol, and Eint(DDA+−CB) is −17.37×103 kcal/mol (Supplementary Table 2). The negative value of short-chain ligands’ binding energy with CB is small, so only a small amount of heat is needed to make CB evaporate quickly from the solution system during solvent evaporation31. During the solvent evaporation stage, the percentage of DMF in the total solution rapidly increases. As shown in Fig. 3h, DMF binds to Pb2+ and I- in solution to form [PbImXn]2-m (X = DMF) complexes32, resulting in the inability to form a large number of PbI2 crystals. The ligand ions cannot be induced to insert into the octahedral layer. At the stable crystallization stage, DMF content decreases to supersaturation and the perovskites begin to nucleate. As for the long-chain ligands represented by DDA, due to the large negative value of the binding energy with CB, CB needs to absorb more heat to detach from the solution, the content of DMF does not undergo a large increase in the solvent evaporation stage. The overall solubility of the solution for PbI2 is low, a certain amount of PbI2 nanocrystals can be formed in the system. In later stage, with CB and DMF evaporation, the PbI2 nanocrystals underwent lamellar exfoliation33. The exfoliated PbI2 nanosheets continued to induce the crystallization of (DDA)2PbI434, as shown in Fig. 3I. The absorption signal at 425 nm varies with time as shown in Supplementary Fig. 11a. The absorption of PbI2 in (BA)2PbI4 increases rapidly in the later stage of Stage I. As the solvent evaporation progresses, the solution concentration increases rapidly, and PbI2 and PVK rapidly nucleate. In (DDA)2PbI4, the absorption of PbI2 steadily increases throughout Stage I. This further proves that the nucleation time of PbI2 in the long-chain ligand solution is earlier than that in the short-chain ligand solution. PbI2 undergoes a layered delamination process to form perovskite. In other perovskites, the signal at 425 nm follows a similar pattern (Supplementary Fig. 11b–d). Further, we selected four other anti-solvents of different polarities, which were toluene (Tol), o-xylene (o-Xy), dichloromethane (DCM), and o-dichlorobenzene (DCB), as the main components of the solution system for evaporation and crystallization. Supplementary Fig. 5c shows that in a lower polarity environment, the crystallization of some long-chain ligands presents a one-dimensional rod-like single crystal morphology, such as HA, OA, NonylA, and DDA in o-Xy. This is because the binding energy of the low-polarity anti-solvent with the ligand molecule is relatively weak, which reduces the heat energy that the anti-solvent molecule needs to absorb when leaving the system. Compared with the strongly polar anti-solvent molecule with a larger binding energy with the ligand molecule, its speed of leaving the system is accelerated. This leads to an increase in the proportion of DMF molecules in the long-chain ligand perovskite solution as in the short-chain ligand perovskite solution. Ultimately, long-chain ligands follow the same direct nucleation mechanism as those of short-chain ligands, forming 1D nanowire crystals. In DCB system with greater polarity, its crystalline morphology follows the same rule as that of the CB system, indicating that there are two different crystallization mechanisms in the solvent system. Influence of crystal structure on the growth process In order to further investigate the factors affecting the initiation of different crystalline morphologies, we established the cell models of perovskites formed by 14 different ligands by taking the cell parameters and the symmetry of the space group of (BA)2PbI4 single crystals as a reference35, the CIF files of 14 types of two-dimensional perovskite unit cells are included in Supplementary Data 1. In addition, we performed functional optimization on all cell models to make the semiconductor bandgap formed by the cells conform to the results measured in PL experiments (Supplementary Fig. 12, and Supplementary Table 7), to obtain accurate cell structures as much as possible. The cell parameters of each 2D perovskite cell are shown in Supplementary Table 4. Since the growth process of the crystals was under the kinetic condition of 70 °C, we appropriately expanded the crystal cell into a certain volume of supercells and simulated the kinetic state of the supercells under 343 K conditions (Supplementary Fig. 13, a to n). The interlayer spacing between the layers of 2D perovskite octahedra shows a gradual increasing trend as the size of the ligand molecules increases. However, with the growth of ligand main chains, ligand chains interspersed between layers. We tested the XRD pattens of 14 2D perovskite polycrystalline films, as shown in Supplementary Fig. 14. The (0 2 0) and (0 4 0) crystal planes of 14 2D perovskites were calibrated based on the standard cards and the moving trend of the crystal plane diffraction peaks varying with the B-axis length of the unit cell. The B-axis lengths were calculated, and the interlayer distances were further determined. For perovskites that cannot be calibrated, the functional calculation results are used finally. The interlayer distances of all component units and their ratios to the lengths of L-ligand molecules are shown in Supplementary Table 5. The ratio of layer spacing to ligand size of t-BA, i-PA, PA, BA, PhA, PMA and other short-chain ligands is close to 2, which means that there is no interchain interpenetration. For long-chain ligands such as PentA, HA, OA, NonylA, DDA, PEA, PPA, and PBA, the ratio of layer spacing to ligand size was closer to 1, indicating that ligand chains interspersed in the middle space of inorganic layers (Fig. 4a). Taking BA and DDA as an example, the short-chain ligand BA does not interweave between inorganic layers, and the methyl group at the end of the main chain between the two layers of butyl ammonium is relatively close, which forms an interaction between the layers and maintains a 2D structure (Fig. 4b)36. Whereas, for the long-chain ligand DDA, the carbon chains of the ligands interpenetrate between the inorganic layers, so the formation of interactions extends to the whole range of the organic layers that interpenetrate each other (Fig. 4c). Under the kinetic conditions of 343 K, the ligand chains undergo irregular movements between the layers. For (BA)2PbI4, the butyl chains show a more chaotic structure in the interlayer due to the formation of interactions only between the terminal methyl groups (Fig. 4d); whereas for (DDA)2PbI4, the dodecyl chains are more regular and organized in the interlayer as the range of interactions present is the whole range of organic layers (Fig. 4e). Molecular dynamics simulations of two-dimensional perovskite cells. a Layer spacing, ligand size and their ratio. The gray shaded areas represent the perovskites with non-overlapping interlayer ligands and those with completely overlapping interlayer ligands. The white part represents the partially overlapping ligands. b, c Schematic diagrams of the interlayer structures of b (BA)2PbI4 and c (DDA)2PbI4 at 0 K, with the shaded areas showing the extent of the interactions occurring. d, e Schematic diagrams of the supercell structures of d (BA)2PbI4 and e (DDA)2PbI4 at 343 K. f Interlayer binding energy of organic layers of 14 perovskite supercells at 343 K. The values of the binding energies increase with the size of the ligand molecules, indicating that the interlayer interactions are getting stronger. g, h Schematic diagrams of the organic layer configurations of g (BA)2PbI4 as well as h (DDA)2PbI4 (hydrogen atoms have been hidden). ∆d is the difference in bond lengths of the neighboring Pb-I bonds; ∠Pb-I-Pb represents the octahedral bond angle. The octahedral distortion of (BA)2PbI4 is larger than that of (DDA)2PbI4, which is due to the structural distortion of the inorganic layer caused by the weak interlayer interaction. We further calculated the interaction energy of the interlayer ligand molecules according to Eq. (5) (Fig. 4f). The binding energy between the two layers of ligands in the organic layer increases gradually with the growth of the main carbon chain of the ligand. For two ligands, t-BA and i-PA, because of the presence of branch chains, the number of methyl groups in contact with each other increases, so the binding energy between organic layers increases slightly. The interlayer binding energy of each (BA)2PbI4 and (DDA)2PbI4 unit cell organic layer of were calculated to be −1.85 × 103 and −2.43 × 103 kcal/mol, respectively (Supplementary Table 6). The greater the negative value of the binding energy, the greater the energy required to separate the two layers, so the more negative the binding energy, the stronger the interlayer binding force. According to the general law of binding energy, when the size of ligand molecules reaches a certain level, cross-overlaps occur between the octahedral layers of perovskite. Due to the limitation of lattice and ligand chain size, the space of molecular chain activity decreases, the degree of interlayer disorder decreases, the degree of order increases, and the degree of distortion to the inorganic octahedron is small. The range of the interlayer van der Waals forces is extended, the energy of each particle in the crystal is lower37, as shown in Fig. 4h. Whereas the effect of van der Waals forces on perovskite crystals formed by small steric ligands is only at the end of the ligand chain, not the whole organic layer, and the inorganic octahedron is distorted to a greater extent38, as shown in Fig. 4g. Crystals tend to grow towards a certain surface with lower energy, and different interlayer stacking configurations lead to different surface energies, resulting in 1D rod crystals and 2D lamellar crystals39. The kinetic simulation shows that this law is also applicable to the aromatic hydrocarbon system (Supplementary Fig. 13). We tested the Transition Electron Microscopy (TEM) images, High-Resolution Transition Electron Microscopy (HRTEM), selected area electron diffraction (SAED) patterns, and energy dispersive x-ray spectroscopy (EDX) of (BA)2PbI4 single crystals as well as (PEA)2PbI4 single crystals under an accelerating voltage of 200 kV to investigate the crystal orientation of solution-grown 2D perovskite single crystals. As shown in Fig. 5a, the continuous lattice stripes of the barred (BA)2PbI4 single crystals and the corresponding Fourier transform information (Fig. 5a, bottom right and Supplementary Table 8) indicate that the lattice spacing of the barred single crystals is 8.05 Å, which corresponds to the (1 0 0) crystallographic plane. EDX mapping (Supplementary Fig. 15a) and high-angle annular dark-field (HAADF) of (BA)2PbI4 (Supplementary Fig. 15b) as well as EDX spectra (Supplementary Fig. 15c) show that the 2D perovskite did not undergo decomposition or physical phase changes during the transfer from the Si substrate to the conducting carbon film. In the SAED pattern of the layered (PEA)2PbI4 single crystal, two sets of diffraction spots are observed (Fig. 5c, bottom right and Supplementary Table 9), including the diffraction spots of the perovskite and a set of hexagonal lattice diffraction spots. Through TEM characterization, it was proved that the hexagonal lattice diffraction spots belong to PbI2 (Supplementary Fig. 16 and Supplementary Table 10). Therefore, when calibrating the (PEA)2PbI4 single crystal, the diffraction spots of the non-hexagonal lattice were selected. The crystal face spacing of (PEA)2PbI4 single crystals in HRTEM (Fig. 5c) is 6.24 Å, which corresponds to the (1 1 1) crystal faces of (PEA)2PbI4. Its diffraction pattern (Fig. 5c, bottom right) can be labeled (11¯1) crystal diffraction spot, corresponding to the band axis [1 1 0]. EDX spectra shows that the major elements such as Pb and I are uniformly distributed in the nanosheets, forming 2D perovskite phase (Fig. 5d). The EDX mapping of element C could not show a clear profile due to the large amount of element C in the solvent CB (Fig. 5d), and EDX mapping, HAAEDF and EDX spectrum of each element are shown in Supplementary Fig. 15, e to f. As shown by the lattice information presented by TEM, the rod-like (BA)2PbI4 single crystals have the lowest surface energy at the (1 0 0) crystallographic plane, which extends along the (1 0 0) crystallographic plane when grown in solution and are stacked in the order of L+-I--L+, as shown in Fig. 5e. After nucleation, the area of the inorganic octahedral layer remains at the scale of nucleation without growth. The new ligand and inorganic layers continue to stack in accordance with (1 0 0), where the plane is parallel to the b-axis of the unit cell. The Pb-I is connected to the ligand molecules, and its extension direction is one-dimensional. Therefore, they ultimately form 1D NWs, as shown in Fig. 5f. For cubic phase chalcogenide cells, the surface energy of the (1 1 1) crystal surface is higher than that of the (1 0 0) crystal surface. Specifically, the surface energy of the (1 1 1) crystal faces of perovskite is 0.93 J/m2, while the surface energy of the (1 0 0) crystal faces is 0.89 J/m240,41. Sheeted (PEA)2PbI4 single crystals grow in solution with the lowest (1 0 0) crystal surface energy due to the stacking conformation of the interlayer organic ligands, which breaks the cubic phase conformation resulting in the (1 0 0) crystal surface energy. In contrast, (PEA)2PbI4 single crystals have the lowest (1 1 1) crystal surface energy. Thus, along the (1 1 1) crystal plane during crystal growth, the stacking follows the Pb2+-I--Pb2+ order, as shown in Fig. 5g. As a whole, starting from the nucleation site, the inorganic octahedron undergoes an extended growth with the consequent insertion of long-chain ligands into the inorganic layer, leading to the formation of 2D-scale lamellar single crystals as shown in Fig. 5h. TEM tests on other perovskites show that for single crystals presenting rod-like crystals, the dominant growth orientation is the crystal plane parallel to the B-axis, such as the (1 0 0) crystal plane (Supplementary Fig. 17 to 19, 24 to 25 and Supplementary Table 11 to 13, 18 to 19). In solution, the crystallization process of rod-shaped single crystals is formed by the continuous stacking of organic and inorganic layers. For plate-like crystals, the dominant orientation for their growth is the crystal planes that are not parallel to the B-axis, such as the (0 2 1) and (1 1 1) crystal planes (Supplementary Fig. 20 to 23, 26 and Supplementary Table 14 to 17, 20). In solution, the formation of lamellar single crystals is achieved by the extension of inorganic layers inward and outward in the plane and the insertion of organic molecules between the layers. TEM characterization of (BA)2PbI4 and (PEA)2PbI4 single crystals. a, c Lattice fringing and diffraction patterns of a (BA)2PbI4 and c (PEA)2PbI4 single crystals. b, d Energy dispersion spectra of b (BA)2PbI4 and d (PEA)2PbI4 single crystals, which provided information on single crystal orientation and lattice arrangement. The black part in the figure represents (BA)2PbI4 and (PEA)2Pb4, while the gray part is the conductive carbon film. e Schematic representation of the (1 0 0) crystal plane of a (BA)2PbI4 single crystal that extends as a stack in the ligand direction. f (BA)2PbI4 single crystals have the lowest surface energy at the (1 0 0) crystal face and grow along the c-axis of the cell after nucleation. g Schematic of the (1 1 1) crystal plane of (PEA)2PbI4 single crystal, which extends as a stack in the inorganic octahedral direction. h (PEA)2PbI4 single crystals have the lowest (1 1 1) crystalline surface energy, and the inorganic octahedral layer extends laterally after nucleation. By modeling the arrangement of organic layer ligands, we have gained a more accurate understanding of the interlayer interactions in layered 2D perovskites. The direction and magnitude of interlayer interactions determine the tendency of cell stacking, which further determines the crystalline form of 2D perovskites formed with different ligands. Our study explains the fundamental reasons affecting the morphology of 2D perovskite single crystals from a molecular perspective. Discussion In summary, we used 2D perovskite ligands with different molecular sizes and regulated the solution dynamics of precursors by the magnitude of their binding energies with solvent molecules to crystallize 2D perovskites by different mechanisms, finally realized the high-throughput preparation and characterization of 2D perovskite single crystals in the form of rods and flakes. We used in-situ UV-Vis absorption spectroscopy to investigate the crystallization mechanisms of 2D perovskites guided by ligands of different sizes. Contrary to previous reports, the process of 2D perovskite crystallization is not a pure solvent evaporation process. The binding energy between the ligand and the solvent is not negligible for the evaporation process. The fast rate at which solvent molecules in the precursor solution are detached from the overall solution system during evaporative crystallization leads to different changes in the solubility of the solution for PbI2 as well as perovskite during crystallization. This further leads to a different order of arrival of PbI2 as well as perovskites at supersaturation, and hence two different crystallization mechanisms. In the growth process after crystal nucleation, due to the different configurations of ligands of different sizes in the perovskite crystal lattice, with the increase of the ligand main chain length, the organic layers appear to be interspersed with each other in the configuration of ligand carbon chains, which further leads to a different range of interactions between the organic layers, resulting in different surface energy. The present study revisits the structure of 2D layered perovskite crystals and corrects the previous understanding of the arrangement of organic cations between layers. A conformational relationship between cell structure and crystal morphology is established, providing a basis for the modulation of layered 2D perovskite single crystals. Solution kinetic modulation of perovskite precursor solutions is an effective strategy universally applicable to the morphology control of various types of layered organic-inorganic 2D perovskites. In this work, we derived a universal law of solute-solvent binding energy regulation of 2D perovskite crystallization, which is applicable to alkane and aromatic hydrocarbon ammonium ligands. The regulation of solute-solvent intermolecular binding energy can be realized through molecular polarity, the introduction of special functional groups, etc. The diversity of 2D perovskite ligands signifies that ligand templates are the most effective method for modulating the optoelectronic properties of 2D perovskites. The present study provides universal laws applicable to most of the ligands. Mechanistic studies on the regulation of crystallization processes and crystal morphology have allowed us to regulate the growth of single crystals on a molecular scale, which contributes to further reducing the size of 2D perovskite optoelectronic devices to the micron level using the solution method, providing strong support for the large-scale application of 2D perovskites. Methods Materials Lead iodide (PbI2), tert-butylammonium iodide (t-BAI), iso-propylammonium iodide (i-PAI), propylammonium iodide (PAI), pentylammonium iodide (PentAI), hexylammonium iodide (HAI), octylammonium iodide (OAI), nonylammonium iodide (NonylAI), dodecylammonium iodide (DDAI), phenylammonium iodide (PhAI), benzylammonium iodide (PMAI), phenylpropylammonium iodide (PPAI), phenylbutylammonium iodide (PBAI) were purchased from Xi’an Yuri Solar Co., Ltd. N-butylammonium iodide (BAI) was purchased from TCI, and phenethylamine hydroiodide (PEAI) was purchased from Sigma-Aldrich. Chlorobenzene (CB) and o-dichlorobenzene (DCB) were purchased from Thermo Fisher Scientific Inc. Acetonitrile (ACN), N N-dimethylformamide (DMF) and Poly(methyl methacrylate) (PMMA, MW ~ 35000) were purchased from Aladdin. Single-sided polished high-purity monocrystalline silicon wafers (crystal direction [100], size 15×15 mm, flatness <1 μm) were purchased from LIGE SCIENCE. 15 mm×15 mm super white glass substrate purchased from Liaoning You Xuan New Energy Technology Co. Solution-phase synthesis of pure 2D halide perovskite single crystal In this work, 14 pure two-dimensional perovskite micro-sized single crystals have been synthesized using quaternary solvents. 10 μmol PbI2 was dissolved with 20 μmol LI (L = t-BA, i-PA, PA, BA, PentA, HA, OA, NonylA, DDA, PhA, PMA, PEA, PPA, PBA) in 2 mL of DMF/CB (1:1, v/v) co-solvent to obtain 5 mM of 2D perovskite stock solution. CB/ACN/DCB (9.5:0.1:0.01, v/v) co-solvent was prepared as a dilution solvent. In high-throughput experiments, the dilution factors of alkylamine ligands (t-BA, i-PA, PA, BA, PentA, HA, OA, NonylA, DDA) were set at 60, 120, 240, 480, 560, 640, and 960 times, while those of aromatic amine ligands (PhA, PMA, PEA, PPA, PBA) were set at 30, 60, 120, 180, 240, and 480 times. The Si substrates were preheated to 50, 70, 90, 110 °C and then 50 to 100 μL of perovskite dilute solution was dropped onto them. The single crystal growth time was approximately 1 to 3 min. The entire process was carried out in a nitrogen-filled glove box. The high-throughput platform operating procedures are described in the SI. Preparation of two-dimensional perovskite thin films A 1 M 2D perovskite precursor solution was prepared by dissolving PbI2 with LI in DMF at 1:2 (mol: mol). The dissolved solution was filtered with a 0.22 μm PTFE hydrophobic filter tip to remove impurities. The clean and dry glass substrate was placed in a plasma oxygen microwave apparatus for 10 min for plasma treatment, with the aim of cleaning the organic matter on the glass substrate and improving the infiltration between the perovskite and the glass substrate. 50 μL of precursor solution was taken and spin-coated on the glass substrate, the speed of the homogenizer was 400 rpm and the spin-coating time was 30 s. At the 20th s, 150 μL of CB was taken and added to the substrate in a rapid drop of 1 s or less. After spin-coating, the substrate was heated and annealed at 70 °C for 10 min. The entire process was carried out in a nitrogen-filled glove box. The high-throughput platform operating procedures are described in the SI. Preparation of TEM samples Prepare a CB solution of 10 mg/mL PMMA. Filter the PMMA solution using a 0.22 μm hydrophobic PTFE filter tip. The PMMA solution was spin-coated onto the Si substrate. The speed of the homogenizer was 4000 rpm for 30 s. After the spin-coating was finished, the Si substrate was annealed at 70 °C for 10 min. 2D perovskite was subsequently grown on the PMMA-covered Si substrate. After the growth was completed, CB was used to dissolve the PMMA on the surface of the Si substrate. The Si substrate containing PMMA and perovskite was placed in a centrifuge tube with CB added to the tube over the Si sheet. The centrifuge tube was placed in an ultrasonic machine for 1 h so that PMMA was dissolved by CB while perovskite single crystals were dispersed into CB. The dispersion was added several times dropwise onto a conductive carbon film (EMCN LG14-102 Transmission Electron Microscope Carbon Supported Membrane for Grid Carriers, 200 mesh). Eventually the perovskite single crystals were transferred to the conductive carbon film. Optical imaging Brightfield optical images were acquired using a Mshot metallurgical microscope MJ31. Brightness-adjustable white LEDs are used as light source with an Abbe spotting scope NA1.25 without filter. Brightfield image uniformity is obtained by the long working distance flat field objective L Plan 50X. Photoluminescence imaging Fluorescence field optical images were acquired by a Leica DM4000 automated microscope. The excitation light was modified to the blue band through the use of a filter, and all images were observed and photographed under a dark-field dry 100× objective. SEM imaging Backscatter SEM images were acquired using a Hitachi Regulus 8100 with accelerating voltages of 22 kV, 9 kV, 5 kV, and 15 kV. TEM imaging and spectrum acquisition Transmission electron microscopy (TEM) imaging, HRTEM, diffraction patterns (SAED), and energy spectra (EDX) were acquired by a Talos F200X G2 from Thermo Fisher, USA. The accelerating voltage was 200 kV. TEM imaging and chemical energy spectra of the samples were acquired automatically by Thermo Scientific Maps u’rftware (enabled by Thermo Scientific Velox software). The main analysis was performed using GATAN GMS 3 software after the acquisition was completed. CaRIne Grystallography 3.1 was used to assist in the calculation of the crystallographic indices facet spacing and angle. The calibrations of the diffraction patterns of the two-dimensional perovskite samples are shown in Supplementary Table 8 to 20. Absorption spectrum and In-situ absorption spectrum The absorption spectra of two-dimensional perovskite thin films were collected by Ocean Optics QE65 Pro scientific grade spectrometer, and the light source was a combination of deuterium-tungsten halogen lamp DT-MINI-2-GS, and the analysis software was BiaoQi Spectrum Analysis Software. The in-situ absorption spectra were collected by an in-house built in-situ test system, in which the spectrometer was an Ocean Optics QE65 Pro scientific grade spectrometer, and the light source was an LS-3 visible-near-infrared halogen lamp with a Y-type optical fiber and a hot bench. The test is performed by preheating the substrate to 70 °C. The white light fiber is incident perpendicular to the substrate and the solution is dripped from the side, and the test starts when the solution is dripped. The schematic diagram of in-situ absorption spectroscopy test is shown in Supplementary Fig. 9. Steady-state photoluminescence spectrum and Time-resolved photoluminescence spectroscopy The 2D perovskite film photoluminescence spectra and time-dependent photoluminescence spectra were collected by Pico Quant FluoTime300 fluorescence lifetime spectrometer. Excitation was performed using PDL 820 LDH-P-C-405 monochromatic picosecond laser diode head. The excitation intensity is 7.8. The UV-red detector used a Hybrid PMT with a 425 nm LP filter. The emission lens position is 9.5 mm and the emission attenuator position is 4%. The emission polarizer position is 54.7°. TrPL measurements were made using a single photon counter ps TCSPC TimeHarp 260 P. Tests were made using short range (SR) mode. Time channel resolution was 25 ps. Photoluminescence spectra of perovskite single crystals were obtained by focusing the excitation light with an Olympus BX43 microscope and focusing it onto individual single crystals. The PL spectral processing and TrPL data fitting software utilized was EasyTau, while the image capture and processing software was IDS Peak Cockpit. The fitting publics and procedures are illustrated in Supplementary Table 3. Molecular dynamics and DFT simulations For the 2D perovskite solution and cell dynamics and DFT simulations, the solution and cell models were constructed mainly using Material Studio, and geometry optimization was carried out using density functionals. Based on this, the kinetic energy, electrostatic potential and energy band structure were calculated. The solution and crystal models were optimized using the COMPASSIII force field in Forcite and the GGA-PBE general function in CASTEP module. The cell model was used as a reference with the crystal information of (BA)2PbI4, and the spacer cations were replaced to obtain the rest of the crystal model. The kinetic simulation temperature was set to the experimental temperature (343 K), and binding energy calculations were performed using a script where all kinetic energies were averaged from the output. The molecular electrostatic potential was calculated using the DMol3 module with the GGA-PBE generalization. The 2D perovskite cell energy band structure is calculated using CASTEP module, GGA-PBE generalized function, which ensures the reliability and consistency of the calculation results. The detailed process and parameters of the simulation calculations are shown in SI. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Transparent Peer Review file Reporting Summary Description of Additional Supplementary Files Supplementary Data 1 Source data Source data Acknowledgements This work was financially supported by the National Natural Science Foundation of China (52394273 and 52373179), the TCL Science and Technology Innovation Fund (20242065), the Fundamental Research Funds for the Central Universities (2025ZYGXZR024), and the Fund of Guangdong Provincial Key Laboratory of Luminescence from Molecular Aggregates (2023B1212060003). This work was partially supported by High Performance Computing Platform of South China University of Technology. We thank Dr. Yang at Shenzhen HUASUAN Tech. Co., Ltd., for insightful discussions and suggestions. Author contributions J.Q. synthesized and characterized two-dimensional perovskite single crystals and thin films. J.Q. and X.H performed the operation of the high-throughput platform. J.Q. and K.A performed the construction and use of the in-situ absorption test system. J.Q., L.L. and Y.M performed PL mapping testing and data collection. J.Q. performed molecular dynamics simulations, TEM characterization, and data analysis. N.L. conceived the idea and supervised the project. J.Q., W.M. and N.L. wrote the manuscript; all authors read and revised the manuscript. Data availability The data that support the findings of this study are available within the main text and its Supplementary Information. All data are available from the corresponding author upon request. Source data files are provided as a source data file. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41467-025-66505-1. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Reporting Summary Description of Additional Supplementary Files Supplementary Data 1 Source data Data Availability Statement The data that support the findings of this study are available within the main text and its Supplementary Information. All data are available from the corresponding author upon request. Source data files are provided as a source data file. Source data are provided with this paper.