Barry
McKenna
a,
Abhinav
Shivkumar
a,
Bethan
Charles
b and
Rachel C.
Evans
*b
aSchool of Chemistry and CRANN, Trinity College, The University of Dublin, Dublin 2, Ireland
bDepartment of Materials Science & Metallurgy, University of Cambridge, UK. E-mail: rce26@cam.ac.uk
First published on 22nd May 2020
Lead halide perovskite nanocrystals (PNCs) have emerged as promising candidates for use in optoelectronic devices. Significant focus has been directed towards optimising synthetic conditions to obtain PNCs with tunable emission properties. However, the reproducible production of stable PNC dispersions is also crucial for fabrication and scale-up of these devices using liquid deposition methods. Here, the stability of methylammonium lead halide (MAPbX3 where X = Br, I) PNCs produced via the ligand-assisted reprecipitation process is explored. We have focussed on understanding how different combinations of specific synthetic factors – dilution, halide source and ratio as well as capping-ligand concentration – affect the stability of the resultant PNC dispersion. Photoluminescence spectroscopy, transmission electron microscopy and dynamic light scattering studies revealed that subtle changes in the reaction conditions lead to significant changes in the particle morphology and associated optical properties, often with catastrophic consequences on stability. This study highlights the importance of designing PNC dispersions in order to make more efficient and reliable optoelectronic devices.
Methylammonium lead halide PNCs (MA-PNCs) can be prepared using low-temperature, solution-based methods; the most common approach is ligand-assisted reprecipitation (LARP).17 Here, the perovskite precursor salts are dissolved in polar solvents, such as N,N-dimethylformamide (DMF) or dimethyl sulfoxide (DMSO). The resulting solution is injected rapidly into an antisolvent – often toluene or hexane – which typically contains long-chain capping ligands. These ligands have multiple functions, including improving the solubility of lead halide salt precursors, controlling the crystallisation kinetics and stabilising the final colloidal dispersion.18,19
The choice of capping ligand has a substantial effect on the behaviour of PNCs. Oleic acid (OA) and oleylamine (OY) are commonly used; however, these form an insulating layer and often need to be removed before application in an optoelectronic device such as an LED leaving the PNC susceptible to degradation.20 Previous work has overcome this issue through replacing OA and OY with a bidentate ligand,20 while use of branched capping ligands has also been shown to improve PNC stability.21
Varying the halide content, X, in MAPbX3 and CsPbX3 nanocrystals is commonly employed to achieve emission in the blue (X = Cl), green (X = Br) or red (X = I) spectral regions.22,23 However, emission can also be tuned across the entire visible spectrum through the use of mixed-halide PNCs, such as MAPbCl3−xBrx or MAPbBr3−xIx (0 ≤ x ≤ 1).24 Synthetically, these may be obtained by mixing the precursor halide salts in the desired ratio prior to injection into the antisolvent,23,25 mixing solutions of pure-halide PNCs,26 or via post-synthesis halide exchange.22,27,28 A tunable bandgap and high colour purity makes these mixed-halide PNC systems promising candidates for LEDs.11,29 However, these desirable properties also have the potential for applications in luminescent downshifting layers, where the emission could be optimised for the chosen solar cell.30
The emission colour also depends on the morphology of the PNCs, which is determined both by the composition and processing conditions.17 The first reported synthesis of MAPbBr3 nanocrystals yielded a mixture of 3D nanoparticles and 2D nanoplatelets.31 Subsequent studies have shown that extremely thin MAPbX3 nanoplatelets can be isolated directly through synthetic control25,32 or obtained indirectly through ligand-assisted exfoliation of the layered morphology.33,34 The capping ligands used in the LARP process significantly impact the crystal growth and dimensions, which can have a profound effect on their optical properties. For example, Sichert et al. showed that as the lateral dimension of 3D MAPbBr3 nanocrystals is gradually reduced to quasi-2D and 2D nanoplatelets with a thickness of a single unit cell, with the emission maximum shifting to higher energy due to quantum confinement effects.34 The precise composition and concentration of capping ligands and/or alkyl ammonium salt used in the LARP process may also have a profound effect on the emission colour and PLQY of the resultant PNCs.18,23
Interestingly, the dilution of PNC dispersions has an unpredictable effect on both nanocrystal size and emission properties. For example, Tong et al. showed that dilution of stock solutions of MAPbBr3 or MAPbI3 PNCs with the LARP antisolvent resulted in fragmentation and exfoliation into 2D nanoplatelets. The resulting emission spectrum was blue shifted, with a decrease in PLQY.35 A separate study investigating the effect of solvent environment on the nucleation and growth mechanism of CsPbX3 PNCs revealed that while the initial shape and size was controlled by the solvent polarity, subsequent dilution led to self-assembly and growth of the nanocrystals.36
As for all colloidal systems, the stability of the resultant PNC dispersion is highly sensitive to the solvent type and concentration.37 Instability may arise due to the retention of residual highly-coordinating polar solvents used in the LARP process to the particle surface.37 Recent studies have shown that a mixture of methylamine in acetonitrile is a good non-coordinating polar solvent for metal halide salts,38 which was later exploited to prepare red-emitting MAPbI3 nanocrystals with a PLQY of >93%.39
The size and morphology of MA-PNCs are significantly affected by synthetic factors such as reaction medium, capping ligands, solvent system and dilution. These factors have been discussed more generally in several excellent reviews.3,4,17,40 However, this sensitivity of MA-PNCs to synthetic conditions results in marked differences in both their stability and optical properties. These results are rarely reported in detail, perhaps because they are unlikely to produce a champion system. Yet, this information is crucial both for new researchers attempting to enter a competitive field, and also for the development of realistic pathways to scale-up. To address this gap, we systematically examine three key areas of PNC synthesis: (1) the effect of dilution of the parent dispersion, (2) choice of the halide source and ratio and (3) choice of capping ligand concentration. We examine the mechanism for particle nucleation and growth and provide insight into the dramatic changes in optical properties that may result upon introducing seemingly minor changes to the synthetic conditions.
PL lifetime measurements were carried out using a Horiba Fluorolog FL 3-22 equipped with a FluoroHub v2.0 single photon controller using the time-correlated single photon counting method (TCSPC), run in reverse mode. Samples were excited at 458 nm with a pulsed nanosecond light-emitting diode (NanoLED®). The instrument response function (IRF) was recorded prior to lifetime measurements using a scattering solution (Ludox, Aldrich) and the pulse width was 1.3 ns. Emission decay curves were analysed using IBH DAS6 software. The data were fit as a sum of exponentials and the errors in these fits are accounted for by considering non-linear least squares analysis for error minimisation.
Dynamic light scattering (DLS) measurements were performed on a Zetasizer Nano–ZS (Malvern Instruments) which uses a He–Ne laser (4.0 mW, 633 nm) light source and detection at a backscattering angle of 175°.
Transmission electron microscopy (TEM), scanning transmission electron microscopy (STEM), energy dispersive X-ray spectroscopy (EDX) and electron energy loss spectroscopy (EELS) were performed on a FEI Titan operating at 300 kV. A fixed volume (2 × 15 μL) of the PNC dispersion was drop-cast onto a carbon type B on 200 mesh Cu grid. Imaging was performed in TEM and STEM modes using the high-angle annular dark-field (HAADF) detector. EDX and EELS analysis was performed in STEM mode.
The PL spectrum (Fig. 1c) is characteristic of MAPbBr3 PNCs (λem ∼ 530 nm), suggesting negligible incorporation of iodide into the structure. TEM revealed the sample was composed of platelets with a lateral diameter of ca. 10 nm (Fig. 1d). EDX analysis showed the absence of iodide peaks, suggesting that, despite the inclusion of MAI in the reaction mixture, the resultant PNCs are predominantly all-bromide in composition. It is noted that the concentration of iodide present could be beyond the detection limit of the technique (Fig. S1†).
We believe that a combination of surface area, charge and ligand effects are responsible for the failed uptake of I− into the mixed halide MAPbBr3−xIx PNCs. PNC formation has been shown to proceed through a seed-mediated nucleation step, with seed Pb0 nanoparticles and the polarity of the environment determining the shape and size of resultant PNCs.36 The small particle size implies that the 1:3 ratio of Pb2+ to X− observed in bulk perovskites does not hold for the nanoparticles, due to the presence of an anion-rich surface, which has previously been shown for CdSe quantum dots41 and PNCs.18
Incorporation of the halide is also dependent on the halide source and ratios, since halide ions are important surface-active species which influence colloidal growth.42 Our precursor solution used a non-stoichiometric ratio of MAI and PbBr2 (0.8:1 molar ratio), it is possible this favoured incorporation of bromide into the structure, considering the structural distortion induced by the larger iodide ion.22 Moreover, the lower electronegativity of the iodide ion implies specific adsorption to the PNC surface is less likely than for the bromide,42 while hydrogen bonding between free I− ions and excess OY capping ligand may prevent competitive uptake.43
Nevertheless, the complete absence of iodide was unexpected. Previous studies have shown that components of PbX2 and MAX can be readily exchanged during crystallisation, both in solution and the solid state for bulk MAPbX3.27 The synthesis of mixed-halide MAPbBr3−xClx and MAPbBr3−xIx PNCs through reversible bromide exchange has also been shown.22 However, several key distinctions can be identified in the synthetic conditions of the latter study and ours, namely solvent, type of capping ligand, and sequential addition of the MAI precursor. In contrast, Zhang et al. reported the formation of MAPbBr3−xXx (X = Cl or I) quantum dots using very similar reaction conditions to those used here, differing only in the amine capping ligand and the halide source.18
These results highlight the sensitivity of the LARP procedure to subtleties in the reaction medium and demonstrate the importance of understanding the effect each step in the synthetic process can have on the final product. From this initial experiment, three key synthetic conditions (halide source, capping ligand concentration and dilution in a solvent) were identified as having significant effects on final PNC properties and were subsequently investigated. These initial findings motivated the work presented in the following sections which set out to assess the effect of (1) dilution of the parent dispersion (2) the choice of the halide source and ratio and (3) the choice of capping ligand concentration on PNC properties.
As expected, a corresponding increase in the PL lifetime was also observed upon dilution (Fig. S3†). Both samples exhibit biexponential decay curves, whose components are assigned to delocalised (τ1) and localised (τ2) exciton recombination;46 see Table S2† for a summary of the fitting parameters. Upon dilution, both lifetime components increased, from τ1 = 10.7 ns to 12.4 ns and τ2 = 24.2 ns to 28.8 ns. The associated amplitude (A2) of τ2 can provide an estimate for the quantity of defects in the perovskite lattice.46 Interestingly, upon dilution, A2 decreased, suggesting that as the PNCs aggregate, a loss of defect sites occurs.
In order to relate the PL properties upon dilution to changes in the nanocrystal dimensions, we next performed DLS studies on both the parent and diluted PNC samples over time. In the parent sample (D0), the mean hydrodynamic diameter, dh, was ∼14 nm (Fig. S2†). Upon dilution, dh, increased with time until values of >1 μm were obtained after about 16 hours (see Fig. 2b for D20). This behaviour indicates that, in our system, dilution results in particle growth rather than in the fragmentation and exfoliation of layers, as observed previously for diluted MAPbBr3 and MAPbI3 PNCs prepared via a very similar LARP procedure by Tong et al.35 We note that the only difference in reaction conditions used compared to the previous dilution study35 and ours is the use of mixed halide sources (i.e. MAI/PbBr2), hinting that the iodide ion may play a role in promoting particle growth. PL measurements were performed concurrently with the time-dependent DLS study (Fig. 2c). Interestingly, the red-shift in the PL maximum occurs very quickly after dilution (2 min), suggesting that once particle growth exceeds ∼500 nm, the optical properties are dominated by the bulk material.
Fig. 2d and e compare STEM images of MAPbBr3−xIx PNCs before and after dilution (D20). In the parent sample (see also Fig. 1d), square nanoplatelets are observed with a diameter of ∼10 nm (in good agreement with the dh ∼ 14 nm). Following dilution, the nanoplatelets aggregate into a larger cubic structure which reaches dimensions of ∼150 nm in size. EDX analysis of the D20 sample revealed detectable iodide peaks (Fig. S4†), confirming that particle growth results in incorporation of iodide ions from the reaction medium. The peak ratios in the EDX spectrum suggest a Br− to I− mass ratio of 4:1, which corresponds to a molar ratio of ca. 2.5:1. The experimental data suggest that dilution stimulates particle growth. It is also clear that iodide ions present in the dispersion become incorporated into the larger particles formed in the growth process. The combination of these factors leads to the observed redshift in the emission.
We will now consider the potential mechanism for particle growth following dilution. Under our reaction conditions, we consider that the formation of apparent single halide MAPbBr3 PNCs in the parent sample is preferred due to the presence of excess bromide ions and the preferred retention of the cubic phase. Unused iodide ions are thus present in the bulk solution, and/or associated with the surface capping ligands via hydrogen bonding. Dilution of the sample decreases the effective concentration of the capping ligand in the surrounding dispersion. We propose that this provides a driving force for partial dissolution of the capping ligand from the surface of the nanocrystal, reducing the colloidal stability, leading to rapid aggregation of the PNCs. As the particles agglomerate, iodide ions are incorporated into the aggregated structure, moving towards the stoichiometric ratio present in the reaction mixture. Particle growth will lead to a decrease in surface area, and therefore a reduction in the number of surface trap sites, which is supported by the increased in the amplitude of the τ2 component attributed to localised exciton recombination. Moreover, the decrease in the PLQY and increase in PL lifetime are indicative of a transition to more bulk-like behaviour; for example, bulk MAPbBr3−xIx films have shown lifetimes greater than 100 ns.47
To confirm that the dilution-induced changes in the optical properties can be explained by partial desorption of the capping ligands, a sample was prepared using chloroform as the antisolvent instead of toluene. The increased polarity of chloroform (dielectric constant, ε = 4.81, versus ε = 2.38 for toluene)48 should facilitate improved ionic dissociation, and thus prevent excessive agglomeration of particles on mixing with DMF (ε = 37.8). The Hansen solubility parameters of oleic acid (δt = 16.81 MPa1/2) indicates lower solubility in chloroform (δt = 18.94 MPa1/2) than toluene (δt = 17.95 MPa1/2).49 We note that the Hansen solubility parameter for OY should be comparable to that of OA due to the structural similarities.49 As such, if the proposed mechanism is valid, aggregation should be reduced upon dilution of the parent solution with chloroform. Fig. 3 shows the emission properties before and after dilution (D20). While the parent sample exhibits bright green emission (λem = 545 nm), following dilution an intense blue emission with λem = 475 nm is observed.
The blue-shifted emission of the diluted solution compared with the parent may be attributed to solvent exfoliation and thinning of the PNCs into 2D layers,34 or anion doping of chloride from the chloroform solvent molecules to the PNCs.35 However, the observed emission behaviour confirms the effect that the solvent can have on the dilution-induced behaviour of PNCs. Since toluene is a better solvent for the capping ligands, ligand dissociation and particle growth are favourable, whereas, chloroform, a poor ligand solvent, hinders this process.
All parent samples exhibited the characteristic green emission indicative of bromide-rich MAPbBr3−xIx PNCs (Fig. 4a), indicating lack of incorporation of iodide ions into the crystal structure. Upon dilution (D20) in toluene, the emission maximum red-shifted for all samples, although the extent of this shift depended on the specific mixture of halide sources used. For example, the largest emission shift (∼28 nm) was obtained using a 1:1 ratio of PbX2 and MAX, and from a combination of all four reagents. In comparison, a much smaller red-shift of ∼12 nm was observed for the sample prepared from a 2:1.6 ratio of PbX2 to MAX (Fig. 4c and d). The most significant change occurs when MAI is the primary iodide source (rather than PbI2) (Fig. 4d). Since MA cations are associated with the particle surface, this supports the hypothesis that iodide bound to surface ligands is favourably incorporated upon dilution and subsequent growth. These results demonstrate that the optical properties of PNCs can vary significantly when using different halide precursors.
Fig. 4 The halide source and ratio have a significant effect on the optical properties of MAPbBr3−xIx PNC dispersions. Photographs under UV illumination and corresponding PL spectra of PNC dispersions made with the standard Br:I ratio of ca. 2:1 using different halide sources (a, c) before and (b, d) after dilution D20. (e) and (f) show the PL spectra of PNC dispersions made using different halide ratios, before and after dilution D20, and as a function of time after dilution for sample R6. Full sample compositions given in Table S1, ESI† (λex = 400 nm). |
The effect of halide ratio in conjunction with different halide sources was also explored. PNC dispersions with different Br:I ratios (0.8:2 (R5), 1:1 (R6) and 6:1 (R7)) and different halide sources were also prepared (see Table S1†). After centrifugation, R5 was a transparent solution with a weak blue emission, whose spectrum suggested a composition of primarily PbBr2 seed particles. Upon dilution, the emission maximum changed from ca. 420 to 530 nm, and the sample exhibited the bright green emission characteristic of the standard parent MAPbBr3 PNC sample (∼2:1 Br− to I− ratio in precursor), see Fig. 4a. Sample R6 (1:1 Br− to I−) showed similar emission behaviour, with an emission red-shift of ∼25 nm upon dilution D20 (Fig. 4f); however, for R6, this occurred via a slow stepwise process. Initially upon dilution, two emission peaks of equal intensity were observed at 528 nm and 595 nm. A short time after (ca. 13 min), the relative intensity of the peak at 528 nm had decreased, while the peak at 595 nm had reverse blue shifted back to 555 nm, becoming the primary emission peak. The observed behaviour suggests that upon dilution (D20), two populations with different halide compositions are initially present. Subsequent agglomeration and/or growth results in a single population of micrometre-sized particles comprised primarily of MAPbBr3 with trace amounts of iodide impurities. Sample R7 (6:1 Br− to I−) gave a green emission and notably, upon dilution, no change in the optical properties were observed (Fig. S6†). Given the low iodide content of the precursor solution, this result further indicates that the change in the spectral properties upon dilution is partially due to iodide uptake as the particles grow.
Interestingly, for the samples containing the highest capping ligand concentration (OY100), the initial yellow solution was not observed, but rather a colourless, transparent solution was obtained immediately after centrifugation (Fig. 5a). This solution was only weakly emissive, with several additional peaks in the spectrum, indicating that a variety of particle sizes and compositions were present (Fig. 5b). Dilution (D20) of this sample in toluene results in an immediate red-shift in the emission maximum (λem ∼ 480 nm), corresponding to the blue–green emission observed (Fig. 5a and b).
STEM images of the samples before and after dilution reveal no change in particle size, with spherical dot-like structures with dimensions <10 nm retained (Fig. 5c and d). Similar behaviour has previously been attributed to the formation of one-dimensional PbBr2 nanoparticles upon solvation of excess MAI and Br− by the capping ligands, leading to a similarly structured emission spectrum as observed here for OY100.50 Since PbBr2 nanoparticles act as seeds for the growth of PNCs,18 it follows that upon dilution, the relative concentration of the capping ligand in the solution decreases, decreasing the solubility of any excess MAI and Br− and facilitating particle growth from these seeds. However, the emergence of a single emission peak upon dilution and the small particle dimensions (<10 nm), suggest further growth into a single population of quantum-confined MAPbBr3 PNCs.34 The dilute OY100 sample is unstable, as after 24 h a bright red fluorescence was observed (Fig. 5a). The corresponding emission spectrum contained an additional emission peak at 580 nm, of similar intensity to the primary peak at 480 nm (Fig. S8†). Upon centrifugation the red colour was lost, and the dispersion became non-photoluminescent, suggesting that large PNC particles (removed by centrifugation) were the origin of the red emission.
Footnote |
† Electronic supplementary information (ESI) available: Sample composition details, additional TEM, DLS, and photoluminescence spectra. See DOI: 10.1039/d0nr03227a |
This journal is © The Royal Society of Chemistry 2020 |