In this study, we developed a synthetic strategy using Pb-deficient precursors to achieve enhanced control over halide perovskite NSs. We show that depletion of Pb ions in precursors permits improved control of the growth process governed by Pb2+/Cs+-mediated assembly of [PbI6]4- octahedron nuclei. Accordingly, we synthesized uniform CsPbI3 NSs with tunable thickness down to two octahedral layers, demonstrating striking emission at 563 nm with high stability against spectral drift. The versatility of the synthetic protocol was substantiated by achieving deliberate size control in other CsPbX3 NSs and CsPbX3 NCs. By taking advantage of these synthetic advances, we established full-color tuning in mono-halide perovskites.

Synthesis of CsPbI NSs with two octahedral layer thickness

Previously, Hoye and coworkers obtained thin 3-L CsPbI NSs using a Cs: Pb: I ratio of 1.875: 15: 30 in the precursor. However, the Cs-deficiency approach failed to further reduce the NS thickness due to insufficient kinetic control. To enhance the control over crystal growth, we propose further depleting Pb metals in the precursor. We reason that a Pb-deficient precursor could result in a diminished number of [PbI] nuclei along with a Pb-depleted environment after nucleation. Consequently, a relatively slow assembly of [PbI] polyhedrons and slow diffusion of Pb/Cs ions could be expected to prepare 2-L CsPbI NSs (Fig. 2a).

To test our hypothesis, we reduced Pb contents in the precursor while maintaining a regular I concentration to ensure rapid nucleation. Specifically, we employed cesium oleate (CsOA), PbI and PbI/ZnI as Cs, Pb, and I sources at concentrations of 2.5 mM, 11.3 mM and 75.2 mM (Cs: Pb: I = 1: 4.5: 30), respectively. In a typical synthesis, PbI/ZnI precursors were first dissolved in a solvent mixture of 1-octadecene (ODE, 5 mL), oleylamine (OAm, 0.5 mL) and oleic acid (OA, 0.2 mL) by heating at 135 °C for 1 h. The reaction was then conducted at 85 °C for 2 min by injecting CsOA dissolved in OA (0.5 mL).

Figure 2b-d and Supplementary Fig. 1 show transmission electron microscopy (TEM) images of the as-synthesized CsPbI, revealing a uniform NS morphology with a thickness of 4.46 nm, including ligands. Considering the ligand length of 1.60 nm and the octahedron size of 0.63 ± 0.01 nm, two [PbI] octahedral layers (2-L) in the thickness direction could be derived (Fig. 2d). The X-ray diffraction (XRD) pattern of the 2-L CsPbI NSs can be indexed in accord with β phase CsPbI crystals (Fig. 2e). The additional diffraction peaks at low angles were owing to the stacks of the 2D CsPbI NSs, from which an interplanar spacing of 4.46 nm could be derived (Fig. 2e, inset), in line with the TEM observation. To avoid misidentifying γ phase as β phase in XRD due to preferential (00 l) orientation, we performed two control experiments (Supplementary Fig. 2): (1) reducing sample thickness by half (5 → 2.5 mm), which maintained β phase peaks despite a lower signal-to-noise ratio, and (2) tilting the sample stage by 5°, causing the (210) peak to vanish due to orientation-dependent suppression. In both cases, no γ phase features emerged, conclusively supporting the β phase assignment. Spectroscopy investigations revealed a single photoluminescence (PL) peak centered at 563 nm and the first absorption maximum at 552 nm (Fig. 2f), in agreement with the theoretical values calculated for 2-L CsPbI NSs. These results validate the synthesis of 2-L CsPbI NSs with high purity.

To gain insights into the crystal growth process, we examined the size and morphology evolution of the CsPbI NSs during the synthesis. The results, as shown in Fig. 3a, Supplementary Figs. 3, and 4, indicate that the 2-L structure was quickly formed (5 s) after the reaction was initiated by injecting CsOA. As the reaction further proceeded, the lateral size of the NSs rapidly extended while the NS thickness remained constant (Fig. 3a and Supplementary Fig. 5). The observations suggest that the NS morphology in our synthesis is mediated by the oriented arrangement of OA and OAm ligands, which agrees with previous reports. In this connection, the competition between the assembly of perovskite octahedra and ligand capping at the early stage of the reaction is the key factor in determining the CsPbI NS thickness (Fig. 3b).

The capping rate of ligands is dictated by their diffusion coefficients, which are quantified by the Stokes-Einstein equation:

where k is the Boltzmann constant, T is the absolute temperature, η is the solvent's viscosity and r is the solvodynamic radius of the molecule. According to the above equation, the diffusion coefficients of OA and OAm remain basically unchanged at a given reaction temperature. Therefore, the formation of 2-L CsPbI NSs in our synthesis is dominantly enabled by the metal-deficient precursor that resulted in a slow octahedron assembly rate. Our control experiments revealed that an increase in either Pb or Cs concentration could trigger the growth of thicker NSs (Fig. 3c, d and Supplementary Table 1), confirming the critical role of metal deficiency in dictating the assembly rate of [PbI] octahedra.

To shed more light on the role of metal deficiency in synthesizing 2-L CsPbI NSs, we conducted control experiments using decreased amounts of I, which could mitigate Pb consumption due to the reduction of nucleation rate. As shown in Fig. 3e, 3-L CsPbI NSs emerged by decreasing the I concentration to 52.5 mM, and large δ-CsPbI nano-bundles were formed by further reducing the I concentration to 22.5 mM (Supplementary Fig. 6). These results demonstrate that the octahedron assembly is boosted by the accessibility of Pb/Cs ions, giving rise to oversized NCs. Notably, an appropriate overall precursor concentration is essential for synthesizing 2-L CsPbI NSs (Fig. 3f). The proportion increase of Cs, Pb and I concentrations was found to induce the formation of 3-L CsPbI NSs, stemming from the increased concentrations of [PbI] nuclei and Cs/Pb ions after nucleation.

The role of Zn in the NS synthesis

In our synthesis, ZnI was employed to maintain a regular I concentration due to the deficiency of PbI, which simultaneously served as Pb and I sources. To explore the possible effect of Zn ions on the synthesis, we conducted control experiments by replacing ZnI with HI and Zn(OA), respectively. TEM and spectroscopy characterizations reveal that HI could grant access to uniform 2-L CsPbI NSs as well (Fig. 4a and Supplementary Fig. 7), while synthesis using Zn(OA) only yielded large δ CsPbI nano-bundles under the same experimental conditions (Supplementary Fig. 8). These results demonstrate that Zn ions hardly contributed to the synthetic control of the CsPbI NSs.

Although uniform 2-L CsPbI NSs could be obtained using HI precursor, they displayed much inferior luminescence performance and stability compared to those prepared using ZnI precursor. Specifically, we determined a low photoluminescence quantum yield (PLQY) of 9.81% with a short luminescence lifetime of 6.4 ns (Supplementary Fig. 9) for HI-derived 2-L CsPbI NSs, which completely lost their emission after three days of storage at room temperature (Fig. 4b). In contrast, the 2-L CsPbI NSs prepared through ZnI displayed much higher PLQY (37.04%) and longer luminescence lifetime (9.8 ns). In addition, their emission properties were essentially preserved after five days of storage under the same conditions (Fig. 4c). These observations imply that Zn ions largely passivate defects and improve stability of the 2-L CsPbI NSs, likely through a surface-mediated interaction mechanism.

To probe the role of Zn ions in passivating and stabilizing the NSs, we performed multi-faceted characterizations. As shown in Fig. 4d, e, Supplementary Fig. 10, and Supplementary Table 2, X-ray absorption spectroscopy (XANES and EXAFS) at the Zn K-edge unequivocally confirms that Zn is situated at the surface rather than incorporated into the Pb lattice, exhibiting a local coordination environment comprising four O atoms and one I atom. This configuration suggests the formation of a distinct surface complex, wherein Zn is anchored directly to the 2-L CsPbI NSs via a Zn-I bond while simultaneously coordinating with OA (Fig. 4f). The nature of this organic coordination was further deciphered by Fourier Transform Infrared (FTIR) spectroscopy. As shown in Fig. 4g, HI-derived 2-L CsPbI NSs show no discernible carboxylate-related peaks in the 1700-1500 cm region, while 2-L CsPbI NSs using ZnI as precursor display two pronounced peaks at 1650 cm and 1550 cm. The separation of ≈ 100 cm between the asymmetric (ν) and symmetric (ν) stretches provides definitive evidence for bidentate chelation of carboxylate groups to Zn, confirming the formation of a stable Zn(oleate) complex chemically grafted to the surface. Furthermore, as shown in Fig. 4h, X-ray photoelectron spectroscopy (XPS) reveals a uniform shift of 0.2-0.4 eV to higher binding energies for the core levels of Cs 3 d, Pb 4 f, and I 3 d in 2-L CsPbI NSs prepared through ZnI. This systematic shift indicates increased binding energies of the constituent ions, directly correlating with an improved formation energy (E) of the crystal. Collectively, these results, combined with the data in Supplementary Table 3, demonstrate that Zn incorporation passivates surface defects through strong bidentate chelation and direct ionic anchoring, thereby enhancing the phase stability and suppressing non-radiative recombination in 2-L CsPbI NSs.

Similarly, we synthesized 2-L CsPbI NSs using SrI as a precursor and observed that, compared to 2-L CsPbI NSs prepared with HI, the SrI-derived samples exhibited improved stability. However, their performance remained inferior to that of ZnI-derived NSs. As shown in Supplementary Fig. 11, after 3 days of storage, the SrI-derived 2-L CsPbI NSs began to show a tendency toward agglomeration, accompanied by a broadening of the emission peak. Based on these observations, we propose that Sr may function through a mechanism similar to that of Zn, likely involving surface binding and partial defect passivation. Nevertheless, the stronger bond strength of Zn-I compared to Sr-I is likely a key factor responsible for the superior structural and optical stability achieved in ZnI-derived 2-L CsPbI NSs.

Versatility of the synthetic strategy

Conventional synthesis of thin CsPbI NSs (< 5-L) requires tight control of several experimental variables, including temperature, reaction time, and solvent polarity. As a result, a given synthetic protocol could hardly be adapted to prepare uniform NSs of variable thicknesses. The superior ability of the metal-depletion method to control crystal growth thus promoted us to establish fine-tuning of the CsPbI NS thickness. Specifically, we demonstrated that the thickness of CsPbI NSs could be continuously tuned by increasing Cs/Pb concentrations, in line with the theory of competition between perovskite octahedron assembly and ligand capping. Meanwhile, the reaction temperature could be harnessed as an additional parameter to achieve more exquisite reaction adjustment, preventing the formation of NSs with mixed thicknesses (Supplementary Figs. 12, 13).

After systematic optimization of Cs/Pb concentration and reaction temperature, we also synthesized uniform 3-L and 4-L CsPbI NSs (Fig. 5b and Supplementary Fig. 14) with high reproducibility. Specifically, 3-L CsPbI NSs were obtained by reaction at 75 °C using 7.5 mM Cs and 18.8 mM Pb (Cs: Pb: I = 3: 7.5: 30), following a similar growth pattern as the 2-L CsPbI NSs (Supplementary Fig. 15). In accord with the metal-dominated growth process, 4-L CsPbI NSs were obtained at even higher Cs/Pb concentrations of 12.5/37.6 mM (Cs: Pb: I = 5: 15: 30) using a reaction temperature of 90 °C. For comparison, the stepwise growth approach was attempted to realize uniform NSs with controllable thicknesses by adding precursors (Cs/Pb = 2/3 × 2.5 mM at 75 °C) to 2-L CsPbI NSs. The synthesis yielded mixed 2-/3-L structures, suggesting lateral growth or additional nucleation (Supplementary Fig. 16). Further attempts to grow 4-L from 3-L CsPbI NSs (Cs/Pb = 2/7.5 × 2.5 mM at 90 °C) instead destroyed the original structures, forming irregular NCs/nanoplates due to ligand desorption and redissolution (Supplementary Fig. 17). These results highlight the fundamental challenges of layer-by-layer growth in CsPbI systems, where fast nucleation/growth kinetics prevent precise thickness control through sequential precursor addition. Notably, our attempts to synthesize thicker CsPbI NSs by further increasing the metal concentration or ligand contents typically resulted in oversized CsPbI nano-bundles or uneven nanoplates, ascribed to substantially promoted reaction activity (Supplementary Fig. 18).

The metal-depletion strategy is equally effective for controlling the size of CsPbI NCs, which can be formed using high reaction temperatures. As shown in Fig. 5a, increasing the reaction temperature weakens van der Waals (vdW) interactions between hydrocarbon sidechains of the OA and OAm, resulting in a non-directional alignment of ligands that restricts crystal growth in all directions (Supplementary Fig. 19). Specifically, CsPbI NCs with a particle size of 12.64 nm were prepared by reacting 5 mM of CsOA, 18.8 mM of PbI and 18.8 mM of ZnI (Cs: Pb: I = 2: 7.5: 30) at 140 °C for 30 s (Fig. 5b). An effective size control from 12.64 to 18.50 nm was readily achieved by increasing the Cs and Pb concentration to 12.5 and 37.6 mM (Cs: Pb: I = 5: 15: 30) under otherwise identical conditions. The precise size control in these CsPbI NSs and NCs gives rise to tunable single-band emissions in the orange and red spectrum (Supplementary Fig. 20).

In a further set of experiments, we demonstrated the versatility of the metal-depletion strategy for synthesizing CsPbBr NSs and NCs. Notably, the method for preparing 2-L CsPbI NSs could be readily adapted to synthesize 2-L CsPbBr NSs by replacing PbI and ZnI with PbBr and ZnBr. The relatively low reactivity of Br-based perovskites could be well accommodated by adjusting the concentrations of Cs, Pb and Br sources to 2.5, 18.8, and 75.2 mM (Cs: Pb: Br = 1: 7.5: 30) and extending the reaction time to 30 min (Fig. 5c). Note that a mild reaction temperature of 70 °C was employed to avoid size defocusing during the prolonged reaction (Supplementary Fig. 21). A similar increase of Cs/Pb concentration coupled with a minor adjustment of ligand content could realize 3-L to 5-L CsPbBr NSs/nanoplates with uniform thickness (Fig. 5c and Supplementary Fig. 22). By regulating metal concentrations, CsPbBr NCs with a particle size of 8.42 nm (Cs: Pb: Br = 2: 7.5: 30) and 10.63 nm (Cs: Pb: Br = 5: 15: 30) were also obtained under reaction at 110 °C for 5 min using an OAm/OA content of 1.5/1.5 mL (Fig. 5c). These CsPbBr NSs/nanoplates and NCs render tunable single-band emissions in the blue and green spectrum (Supplementary Fig. 23).

Likewise, we also obtained 5-L (Cs: Pb: Cl = 1: 4.5: 30) and 6-L (Cs: Pb: Cl = 5: 7.5: 30) CsPbCl NSs by reaction at 80 °C for 120 min using 0.5/0.7 mL of OAm/OA (Fig. 5d). By increasing the OAm/OA content to 1.0/1.0 mL, we also prepared thicker CsPbCl nanoplates (Fig. 5d and Supplementary Fig. 24). Note that our attempts to synthesize thinner CsPbCl NSs by reducing the Cs/Pb concentration to 2.5/7.5 mM (Cs: Pb: Cl = 1: 3.0: 30) or decreasing the temperature to 60 °C were unsuccessful, likely due to the formation of CsPbCl (Supplementary Fig. 25). The preparation of CsPbCl NCs at elevated temperatures with further increasing the OAm/OA content was also ineffective, which resulted in uneven nanoparticle morphology due to the undesired lateral assembly of CsPbCl NSs (Supplementary Fig. 26). Nevertheless, the prepared CsPbCl NSs can cover the violet to blue spectra region (Supplementary Fig. 27). In alliance with the CsPbBr and CsPbI NSs/NCs, these mono-halide perovskites allow us to achieve full-color emission tuning without the issue of halide segregation (Fig. 6, Supplementary Figs. 28, 29 and Supplementary Table 4).

In conclusion, our study has revealed the substantial influence of Pb concentration on the growth kinetics of halide perovskites. Accordingly, we have established a synthetic strategy based on Pb-deficient precursors and achieved size control down to two octahedral layers in CsPbX NSs of various compositions. The synthetic advances allow us to realize full-color emission tuning in mono-halide perovskites without the issue of halide separation. These findings are expected to push forward the frontiers of CsPbX nanomaterials and promote advanced lighting and display applications.