The CsPbI3 perovskite NCs were synthesized using hot injection method involving a rapid injection of Cs-oleate solution into the solution of PbI2 and ligands (oleic acid and oleylamine) in octadecene at high temperature (170 °C). After this, the NCs were washed with methyl acetate (MeAc) to remove excess ligand and unreacted precursors as advised in previous reports [28, 29].
We verified the effects of MeAC antisolvent used in the washing on the properties of as-synthesized NCs by photoluminescence (PL) and FTIR measurement. The results show that the MeAC anti-solvent washing significantly deteriorated the optical properties of the NCs as evidenced in a reduced PL intensity even though the polarity of methyl acetate is relatively low. Accordingly, the photoluminescence quantum yield (PLQY) dropped from 73.0 to 59.2% after purification (Table S1). The decrease in PLQY is usually observed when the density of ligands on NCs surface is reduced [13]. This is also confirmed in the FTIR measurement (Fig. S1). The CsPbI3 NCs without MeAc washing exhibited clear vibration peaks of surface ligands of oleic acid and oleylamine. However, after being washed with MeAc, the intensity of these characteristic peaks is reduced significantly, implying a decrease in the ligand density [13].
By considering that the removal of ligands can leave behind defects and dangling bonds on the surface of the CsPbI3 NCs, we developed a strategy using GuI as the post-surface treatment agent. In this approach, we first made a solution of GuI in tetrahydrofuran (THF), which was subsequently added to the dispersion solution of the as-synthesized CsPbI3 NCs. Experimental details are included in the supporting information. We used the solution-phase ligand exchange strategy instead of solid state-ligand exchange (treat thin film of NCs) to reduce the effect of particles fusion and coarsening, which is not favorable for light-emitting applications [16, 30]. THF was used as the solvent for dissolving GuI due to its moderately low polarity. It means a small amount of the solvent should not significantly affect the crystal structure of the highly ionic crystals of CsPbI3 perovskite.
Interestingly, we found that the emission of the CsPbI3 NCs solution was enhanced significantly after the GuI solution was added (Fig. S2). Investigation of the evolution in the PL of the CsPbI3 NCs as a function of the volume of GuI solution added showed that the maximum PL emission was achieved with 20 \(\mathrm{\mu L}\) of GuI solution added (Fig. S3). Beyond this, the PL dropped dramatically, which is possibly due to the formation of a guanidinium containing non-perovskite phase as reported previously [15, 31]. Furthermore, we observed the appearance of a precipitate as the amount of GuI solution added was increased. We ascribe this to the excessive ligand exchange of oleylammonium with Gu+, resulting in the fusion of the NCs to form large particles [32].
Figure 1a shows the PL and UV-vis absorbance of the pristine (0 \({\mu L}\)) and treated NCs (20 \({\mu L}\)). It can be seen that the GuI treatment does not change the bandgap of the material, as the onset absorption wavelength and PL peak of the pristine and treated samples are in the same position. Both samples show the PL emission peak located at 686 nm. There is only small increase in the full width at half maximum (FWHM) of the PL peak from 34.2 to 36.3 nm after surface treatment, which can be assigned to larger size distribution of the NCs. Relative PLQYs of the NCs solution were determined by using Rhodamine 6G as the reference standard [33]. Detailed calculations for the PLQY values are shown in Table S1. The results show that the GuI treatment enhance the PLQY of the CsPbI3 NCs from 59 to 82% with the concentration of NCs solution in the PLQY measurement was approximately 4.5 × 10–5 mg mL−1. We performed time-resolved PL (TRPL) measurements to determine the PL lifetime of the pristine and GuI treated NCs (Fig. 1b). The parameter obtained from the TRPL fitting are shown in Table S2. The pristine perovskite shows a much faster decay compared to the GuI treated sample. Interestingly, the PL decay of the pristine sample needs to be fitted using a triexponential function rather than a biexponential function. The three-lifetime components of the pristine sample can be associated with Shockley–Read–Hall (SRH) recombination via defect trappings, radiative recombination of free electrons/holes, and Auger recombination [34, 35]. The lifetime decay components \(\tau_{1}\), \(\tau_{2}\), and \(\tau_{3}\) in the pristine sample were 3.34, 19.82, and 76.75 ns, respectively. While for the GuI treated CsPbI3 NCs, the PL lifetime decay can readily be fitted with a biexponential function where the fast decay \(\tau_{1}\) (25.42 ns) and long decay \(\tau_{2}\) (85.4 ns) components can be associated with radiative recombination process [35]. Clearly, there is significant number of traps associated defects presented in the pristine CsPbI3 NCs sample, which are successfully filled with GuI treatment. Most importantly, the GuI treated sample exhibits a much longer average lifetime (72.6 ns comparing to 64.7 ns in the pristine NCs), indicating improved quality of the perovskite NCs for efficient radiative recombination. This is probably because the GuI treatment efficiently suppressed the defects related charge trapping and promoted the radiative recombination in the CsPbI3 NCs.
The morphology of the perovskite NCs revealed by TEM (Fig. 1c–d) shows that the pristine CsPbI3 NCs exhibit a uniform cubic shape with an average size of ~ 11.8 nm. The high-resolution TEM measurement reveals a lattice spacing of 0.62 nm of the NCs, corresponding to the (100) plane of cubic phase CsPbI3 perovskite [27, 36]. The energy dispersive X-ray (EDX) elemental mapping shows clear distribution of Cs, Pb, and I over the NCs (Fig. S4). Upon GuI treatment, there is a slight increase in size of the particles from ~ 11.8 to ~ 14.0 nm. However, the NCs still maintained a uniform cubic shape with good crystallinity (Fig. 1d). The change in morphology of the treated NCs can be assigned to the dissolution and coarsening of NCs during ligands exchange process, which is commonly observed in perovskite NCs after surface treatment [37, 38]. Nevertheless, comparing to the solid-sate ligand exchange strategy, the effect of particle fusion and coarsening is less significant when this solution-phase ligand exchange was used. We have conducted SEM measurements to reveal the morphology of CsPbI3 NCs film prepared by using different ligand exchange strategy (Fig. S5). The result shows that both pristine CsPbI3 NCs film and the solution-phase treated CsPbI3 NCs film exhibited a uniform morphology with well-distributed small particles. While the CsPbI3 NCs films made by post-treatment with solid-state ligand exchange of GuI (0.5 mg mL−1 in ethyl acetate) have more compact morphology with closer packs of particles due to the reduced interdot interaction, which is consistent with the observation in other works using solid-state ligand exchange strategy [30, 32].
The XRD measurement was carried out to confirm the crystal structure of the as-synthesized NCs. As shown in Fig. 2a, both the pristine and the treated CsPbI3 NCs show similar XRD pattern, which belongs to cubic phase CsPbI3 (PDF 01-080-4039) with a preferential orientation along the (100) and (200) planes. No distinguishable XRD peak shift was detected with the GuI treated sample, implying that the GuI treatment does not alter the crystal structure of the CsPbI3 NCs. In other words, the Gu+ cation does not enter the crystal lattice of CsPbI3. The result is consistent with previous study on the Gu+ treatment of thin film of CsPbI3 perovskite [39].
X-ray photoelectron spectroscopy (XPS) was performed to characterize the chemical state of element on the NCs surface. The characteristic XPS signal for Cs 3d, Pb 4f, I 3d of CsPbI3 and O 1s, C 1s, N 1s associated with ligand bonding were detected as shown in the XPS scan spectrum of pristine and treated samples (Fig. 2b). A closer analysis of the high-resolution XPS spectrum of Cs 3d, Pb 4f, and I 3d shows that the GuI treatment does not affect the chemical state and bonding of these elements in CsPbI3 as all the peaks appear in similar shape and position (Fig. S6), which agree with the XRD result that the Gu+ cation does not incorporate into the lattice of CsPbI3 crystal. On the other hand, there is a noticeable change in the N 1s peak related to the surface ligands. In the pristine NCs, the N 1s peak can be fitted with a single peak at 402.8 eV, which corresponds to the protonated amine groups from oleyl ammonium ligand [40]. The treated NCs show a much stronger N 1s signal, which can be fitted with two peaks at 403.0 and 401.2 eV as shown in Fig. 2c. The N 1s peak at 403.0 eV can be assigned to the amine groups in oleyl ammonium similar to pristine NCs. The dominating N 1s peak at 401.2 eV is probably originates from the deprotonated guanidinium group with three amino groups [41]. This result implies the existence of Gu+ on the NCs surface, which provides extra amino group to help passivate the CsPbI3 NCs surface. This is consistent with previous research [15]. The quantitative analysis of the XPS spectra shows that there was a change in the relative atomic content of elements. Specifically, the atomic ratio of I/Pb was 2.64 in the pristine NCs. This ratio increased to 2.92 in the GuI treated NCs, suggesting that GuI treatment compensates the iodide loss in the washing step, heals the iodide vacancy on the surface of pristine CsPbI3 NCs. The result is consistent with the reduced defects confirmed in the above TRPL result. Furthermore, the significant increase of the integrated area of N 1s in the treated sample and increased atomic ratio of N/Pb from 0.18 (pristine sample) to 0.31 (treated sample) also confirm the presence of extra amino related to guanidinium cations on the surface of the NCs.
We further used the thermogravimetric analysis (TGA) to investigate the decomposition of organic ligands in the NCs (Fig. S7). For the pristine sample, a significant weight loss (~ 11 wt%) was observed when the temperature reach to roughly 200 °C, which corresponds to the loss of OA and OLA ligands [42]. The loss above 400 °C can be assigned to the thermal decomposition of CsPbI3 perovskite [43]. In contrast, the weight loss from OA and OLA ligand in treated NCs at 200 °C was significantly smaller (3 wt%). In addition, there is a sharp drop at around 300 °C which can be assigned to the melting of guanidinium iodide [44]. This implies that the Gu+ cation has been exchanged with the native OA and OLA ligand on the NCs.
Based on these findings, we propose the mechanism underlying the GuI post-treatment as illustrated in Fig. 2d. The CsPbI3 NCs have less ligands on the surface after being washed by methyl acetate (MeAc), resulting in exposure of halide vacancies on the surface which can act as charge carrier traps [45]. The GuI treatment provides an iodide source to fill the iodide vacancies on the perovskite NCs surface. Meanwhile, the Gu+ cation strongly couples to the surface of the NCs through extra hydrogen bond to further protect the NCs from aggregation. Considering the larger size of Gu+ (278 pm) comparing to the size of Cs+ cation (177 pm), Gu+ will preferentially locate to the surface of CsPbI3 instead of entering inside the crystal [15, 46], forming a passivation layer on the surface of CsPbI3 which can enhance the excitons confinement. This explains the enhanced PLQY (Fig. 1a).
We also measured the stability of the NCs by recording the PL emission of the NCs solution during storage in ambient air. The CsPbI3 NCs generally show a tendency to agglomerate due to the high surface energy of the small particles, making perovskite NCs solution unstable during storage [47]. In particular, when the ligand density on the surface is low after antisolvent treatment, the NCs often show degradation within 1 week of storage in ambient air [40]. As expected, the pristine CsPbI3 solution exhibited quite poor stability in ambient condition (relative humidity 40–65%). The PL of the solution was almost completely quenched after 10 days of storage. In contrast, the treated NCs exhibited no significant change after the first week and still maintained ~ 90% of the original PL emission even after 30 days (Fig. 3a). We see that the untreated NCs after aging time (7 days) have aggregated and fused into large particle and rods with hundreds of nanometre in size (Fig. 3b), which commonly occurs in CsPbI3 NCs after prolonged storage [48]. The XRD measurement also confirms the phase transformation of NCs from cubic α-CsPbI3 to orthorhombic δ-CsPbI3 (Fig. 3c). This explains the loss in photoluminescence. Clearly, the GuI treated NCs exhibit much-enhanced morphology and crystallinity stability. Under the same storage condition, the treated NCs remain dispersed cubic particles and retain their α-CsPbI3 phase (Fig. 3b–c). The enhanced stability can be ascribed to the passivating effect of the guanidinium cations on the surface of the NCs and the reduction of halide vacancy defects.
To demonstrate the benefit of GuI treated CsPbI3 in optoelectronic applications, we fabricated LED devices by using the pristine and treated CsPbI3 NCs respectively. The device configuration was indium tin oxide electrode/ poly(3,4-ethylene dioxythiophene):poly styrene sulfonate (PEDOT:PSS) (hole injection layer)/ CsPbI3 NCs (emissive layer)/ 2,2′,2″-(1,3,5-benzinetriyl)-tris(1-phenyl-1-H-benzimidazole) (TPBi) (electron injection layer) and Ag/LiF (counter electrode) as illustrated in Fig. 4a. Ultraviolet photoelectron spectroscopy (UPS) was used to measure the valance band of the NCs film. It shows that the GuI treatment slightly affects the electronic band structure of CsPbI3 NCs (Fig. S8) which can be due to the change in the electronic environment on the surface of NCs after ligand exchange. The band energy alignment of the materials in the device is illustrated in Fig. S8c. The LED devices exhibited stable and uniform electroluminescence (EL) (Fig. 4b-inset). Comparison of the EL spectra (Fig. 4b) shows that both devices show the same EL peak position at 695 nm and narrow full width at half maximum of 31.2 nm for the pristine and 32.0 nm for the treated NCs, implying high colour purity. The EL spectrum corresponds to the Commission Internationale del’Eclairage (CIE) color coordinates of (0.62, 0.28) (Fig. S9). There is a slight red shift of the EL peak comparing to the PL peak of the NCs solution in hexane. This is explained by the dielectric dispersion of the solvent and the interdot interaction in the thin films of NCs [49, 50]. At different driving voltage, the EL spectra of the devices remain stable with no significant spectral shift was detected (Fig. S10). It also can be seen that at the same applied voltage, the treated device exhibited higher EL emission (Fig. 4b). The I–V (current–voltage) characteristic in Fig. 4c shows that the treated device exhibits slight enhancement in the current density over the applied voltage range, which agrees with the result that GuI treatment does not alter the electronic structure of CsPbI3 NCs. However, the treated device exhibited much-improved luminance, implying higher current efficiency. The maximum luminance up to 7039 cd m−2 was achieved with the treated LED device at 9 V applied voltage (Fig. 4d), while the luminance of pristine device only reached to 5064 cd m−2. The reproducibility of the luminance of multiple devices (25 devices) is shown in histogram of maximum luminance in Fig. S11, indicating high reproducibility. Both the devices also show low turn-on voltage (at luminance of 1 cd m−2) of only ~ 2 V which is near to the optical bandgap (~ 1.72 eV refer Fig. S8) of the material. Further, there is significant improvement obtained in the current efficiency of the treated device (Fig. 4e) which shows a maximum current efficiency of 10.8 cd A−1 compared to 8.4 cd A−1 of the pristine device. The LEDs device with treated NCs showed a high EQE of 13.8%, which is ~ 3.6 time higher than device based on pristine CsPbI3 NCs (3.8%) (Fig. 4f).
We also have measured the operational stability of the device (Fig. 5a). The device was operated at constant current density of 25 mA cm−2 and the luminance of the device was recorded. The device fabricated by using the pristine CsPbI3 exhibited a much faster drop in the luminance with the half lifetime T50 was only around 6 min. While the device fabricated with the GuI treated CsPbI3 showed much-enhanced stability with the T50 ~ 20 min was achieved. The result clearly demonstrated the beneficial effect of GuI treatment method on performance of CsPbI3 NCs based LED devices. The performance of our devices is competing among the perovskite NCs LEDs employing different surface treatment method reported previously (Table S3). The results prove the beneficial effect of GuI treatment on enhancing the radiative recombination in the CsPbI3 emission layer.
In order to investigate the charge carrier mobility and trap density in the as-fabricated CsPbI3 film, we conducted the Space-charge-limited current (SCLC) measurement on the electron-only device (Fig. 5b). An obvious increase in the injection current of the device with treated CsPbI3 NCs comparing to the device fabricated with pristine CsPbI3 NCs. The result indicates a higher electron injection efficiency [13]. In addition, the device with treated NCs exhibited a trap-filling limited voltage (VTFL) of 1.81 V, which is significantly lower than that of the devices fabricated with pristine NCs (2.48 V). The reduction in VTFL obviously was resulted from the lower trap state density in the GuI treated CsPbI3 NC film, reflecting the effectiveness of this surface treatment method.