Colloidal CsPbX3 (X = Cl, Br, I) Nanocrystals 2.0: Zwitterionic Capping Ligands for Improved Durability and Stability.

Colloidal lead halide perovskite nanocrystals (NCs) have recently emerged as versatile photonic sources. Their processing and optoelectronic applications are hampered by the loss of colloidal stability and structural integrity due to the facile desorption of surface capping molecules during isolation and purification. To address this issue, herein, we propose a new ligand capping strategy utilizing common and inexpensive long-chain zwitterionic molecules such as 3-(N,N-dimethyloctadecylammonio)propanesulfonate, resulting in much improved chemical durability. In particular, this class of ligands allows for the isolation of clean NCs with high photoluminescence quantum yields (PL QYs) of above 90% after four rounds of precipitation/redispersion along with much higher overall reaction yields of uniform and colloidal dispersible NCs. Densely packed films of these NCs exhibit high PL QY values and effective charge transport. Consequently, they exhibit photoconductivity and low thresholds for amplified spontaneous emission of 2 μJ cm-2 under femtosecond optical excitation and are suited for efficient light-emitting diodes.

Semiconducting lead halides with perovskite crystal structure, recently known as photovoltaic materials showing power conversion efficiencies exceeding 22%,[1,2] also hold great promise as versatile photonic sources in the form of colloidal nanocrystals (NCs). Fully inorganic CsPbX3 (X = Cl, Br, or I, or a mixture thereof) have become popular choices owing to their chemical stability and broadly tunable photoluminescence (PL, 400–700 nm), small PL full width at half-maxima (fwhm, 12–40 nm for blue-to-red), and high PL quantum yields (QYs = 50–90%).[3,4] Their intrinsic defect tolerance,[5,6] i.e., the rather benign nature of surfaces with respect to PL efficiency, is a particularly important asset for employing these NCs in displays,[7] light-emitting diodes,[8−13] and potentially lasers.[14,15]

A highly pressing challenge related to organic–inorganic interfaces was identified in the early days of CsPbX3 NCs. Highly dynamic binding exists between the surface capping ligands, typically a pair consisting of an anion (Br– or oleate, OA–) and a cation (oleylammonium, OLAH+), and the oppositely charged NC surface ions (Scheme a).[3,16] Together with a mutual equilibrium between the ionized and molecular forms of these ligands (OA– + OLAH+ ⇋ OLA + OAH or OLAH+ + Br – ⇋ OLA + HBr, Scheme a), these dynamics cause rapid desorption of the protective ligand shell upon isolation and purification of colloids, which is practically observed as a loss of colloidal stability and a rapid decrease in PL QY. This eventually also leads to the loss of structural integrity, i.e., sintering of NCs into bulk polycrystalline materials. Thus far, the strategies to address such problems included embedding of NCs into a solid matrix[7,9,17−24] or using molecular additives in colloidal solutions of NCs.[25−29]

Scheme 1

(a) Depiction of Conventional Ligand Capping of Perovskite NCs Using Long-Chain Molecules with Single Head Groups, In the Ionized Form (OA– or Br–, OLAH+) and (b) a Novel Strategy wherein Cationic and Anionic Groups Are Combined in a Single Zwitterionic Molecule

The net effect of two possible sets of equilibria is facile ligand desorption during purification.

Examples of long-chain sulfobetaines, phosphocholines, and γ-amino acids tested in this work are depicted left to right (n = 1): 3-(N,N-dimethyloctadecylammonio)propanesulfonate, N-hexadecylphosphocholine, and N,N-dimethyldodecylammoniumbutyrate.

In this work, we present a general approach for the efficient surface ligand capping of CsPbX3 NCs using zwitterionic long-chain molecules, which are readily available commercially (i.e., sulfobetaines, phosphocholines, γ-amino acids, etc., Scheme b). For instance, 3-(N,N-dimethyloctadecylammonio)propanesulfonate is a long-chain sulfobetaine, broadly used as a low-cost detergent in, for example, shower gels, protein isolation, and antibacterial coatings. There are two major structural differences with respect to conventional carboxylate and ammonium capping ligands (e.g., OLA+OAH), both favoring stronger adhesion to the NC surface. First, the cationic and anionic groups have no possibility of mutual or external neutralization by Brønsted acid–base equilibria. Second, the binding to the NC surface is kinetically stabilized by the chelate effect.[30] In agreement with this argument, one can explain also the effective bicarboxylate binding reported recently by Bakr et al. for CsPbI3 NCs.[29]

In the proof-of-principle experiment, we fully replaced the OAH and OLA by a zwitterionic ligand (Figure a; b; see detailed methods in the Supporting Information and related Figures S1–S7). In a typical synthesis of 10 nm CsPbBr3 NCs, cesium 2-ethylhexanoate (0.2 mmol), lead(II) 2-ethylhexanoate (0.24 mmol), and 3-(N,N-dimethyloctadecylammonio)propanesulfonate (0.1 mmol, ligand) were combined in dried mesitylene (∼6 mL) and heated to 130 °C under inert gas. At this point, a trioctylphosphine-Br2 adduct (TOP-Br2, 0.3 mmol) dissolved in toluene (0.5 mL) was injected into the reaction mixture, which was then immediately cooled. The crude solution was centrifuged to remove insolubles (if any) and mixed with ethyl acetate (12 mL) to precipitate NCs. The NCs were isolated by centrifuging, redispersed in toluene (3 mL), and centrifuged again to remove a fraction of larger NCs (below 10% by weight; if any). Afterward, higher purity could be attained without the loss of structural integrity and a high PL QY could be retained by repeatedly (i.e., up to 3 more times) adding a nonsolvent (e.g., 6 mL of ethyl acetate, 3 mL of acetone, or 1 mL of acetonitrile), centrifuging the mixture, and redispersing the precipitate in toluene (3 mL). In contrast, conventional OA/OLA-capped CsPbBr3 NCs fully transform into poorly luminescent bulk material upon analogous washing. Zwitterionic-ligand-capped CsPbBr3 NCs can form much more concentrated colloids (up to 50–100 mg/mL) than their OA/OLA-capped counterparts. Furthermore, the typical synthesis yield of clean dispersible CsPbBr3 NCs is ca. 80% compared to only 10–20% for rather impure NCs prepared with the conventional OA/OLA capping. The obtained CsPbBr3 NCs showed a phase-pure orthorhombic crystal structure, identical to NCs received by OA/OLA synthesis[3,31,32] (Pnma space group; see powder X-ray diffraction patterns in Figure S4). The mass fraction of organic ligands was estimated to be ca. 11% using thermogravimetric analysis (Figure S5), corresponding to a ligand density of ca. 1.7 nm–2 (for 11 nm NCs).

Figure 1

Synthesis of zwitterionic-capped CsPbX3 NCs, exemplified for CsPbBr3: (a) reaction equation, (b) typical TEM images of CsPbBr3 NCs, (c) absorbance and emission spectra, (d) QY of NCs covered with the 3-(N,N-dimethyloctadecylammonio)propanesulfonate and OA/OLA after two steps of purification on day 1 and after storage for 28 days.

Solution nuclear magnetic resonance (NMR) spectra were acquired at various stages of the purification, confirming the formation of R3P(OOCR)2 [trioctyl-λ5-phosphanediyl bis(2-ethylhexanoate), Figure S6] and the complete removal of the free zwitterionic ligand, reagents, and reagent byproducts (Figure S7). Solution NMR fails to accurately resolve the resonances attributed to surface-immobilized ligand molecules because the slow tumbling of NCs in solution results in significant signal broadening. Therefore, the purified NCs were decomposed to liberate the ligands by their complete ionic dissolution in deuterated dimethyl sulfoxide (DMSO-d6). The NMR spectra of the resulting solution point to the zwitterionic ligand as the sole surface-bound species (Figure S7). Analogous findings on the preferential and exclusive binding of sulfobetaine and on colloidal durability were obtained when 2-ethylhexanoate was replaced with the oleate in the synthesis (Figure S8). We also tested halide sources such as oleylammonium bromide (OLAHBr) as alternatives to TOP-Br2. The OLA was found as a co-ligand at the surface (Figure S9), presumably in the form of OLAHBr. The sulfobetaine-to-OLA ratio was ca. 1.5. Diffusion-ordered NMR spectroscopy (DOSY NMR, Table S2), which probes the diffusion speed of the detected molecules, estimated that the diffusion coefficients for the broad resonances obtained from the zwitterionic ligand were nearly identical to the value independently calculated using the Stokes−Einstein equation for the actual size of the NCs (i.e., 5.17 vs 4.99 × 10–11 m2/s for the 11 nm NCs). The motion of a free ligand molecule, on the contrary, is 2 orders of magnitude faster (Table S2).The photophysical qualities (i.e., PL QYs, PL fwhm, and PL lifetimes) of the sulfobetaine-capped CsPbX3 NCs were commensurate with those of standard OLA/OA-capped NCs (see Figure c for CsPbBr3 NCs, Figure S10 for time-resolved PL, and Figures S11 and 12 for chlorides and iodides). The utility of other zwitterionic ligands—phosphocholines and γ-amino acids—is illustrated in Figure S13. The decisive role of zwitterionic surface capping for improving the chemical durability[33] of perovskite NCs can be illustrated by a comprehensive study relating the optical characteristics, foremost the PL QYs, to the variation in the number of washing steps, solvents, and aging period; see Figures d, S14, and 15 and additional discussions and details in the Supporting Information. The retention of PL QYs above 60% for the CsPbBr3 NCs was considered a benchmark for stability. Briefly, standard OLA/OA-capped NCs exhibited such PL QYs (∼80%) only when the number of washing steps did not exceed two and only for one antisolvent: ethyl acetate. Even in this best case, the PL QY dropped to ca. 20% after 28 days of storage under ambient conditions. On the contrary, the sulfobetaine-capped CsPbBr3 NCs, washed twice with ethyl acetate, acetone, or acetonitrile as antisolvents, retained PL QYs in the range of 70–90% for 28–50 days. These NCs could even moderately tolerate washing with alcohol (i.e., ethanol), showing a PL QY of 65% after two washings and ca. 40% after 28 days. The absorption and PL spectra as well as PL lifetimes of the zwitterionic-capped NCs remained largely unchanged during intense washing for up to four times (Figure S16).

By means of density functional theory (DFT), we analyzed the passivation of CsPbBr3 NCs capped with OLAH+Br–, OLAH+OA–, and C3-sulfobetaine (Figure ). The details of the methodology employed are provided in the Supporting Information. All of the relaxed species comfortably fit the perovskite crystal structure, with the ammonium group in the OLAH+Br– and OLAH+OA– engaging in hydrogen bond interactions with the corresponding anion.[34] Remarkably, the dimethylammonium group of the zwitterion, which can be expected as rather bulky, also can be easily accommodated in a cation site at the surface. For all species, the binding energy was computed to be ca. 40–45 kcal/mol, suggesting good affinity of all of the ion pairs to the surface. However, there is no substantial energetic difference between the conventional and zwitterionic passivation. This supports the theory proposed earlier in the introduction that the experimentally observed improvements are due to the chelate effect. We also analyzed the electronic structure to verify whether the different kinds of passivation could lead to the formation of localized surface states. For all cases, the bandgap of the perovskite remained intact and free of midgap states. The HOMO–LUMO levels of the ligands used were calculated and found to reside within the valence band and conduction band, respectively (Figure S17).

Figure 2

Top and side views of a binding site in a model CsPbBr3 NC (∼3 nm) computed at the DFT/PBE level of theory,[35] using the CP2K software package.[36] All structures have been fully relaxed. Cs atoms are drawn in gray, Pb in orange, Br in magenta, N in blue, C in light blue, O in red, S in yellow, and H in white. The binding site is circled in white for different ligands: (from left to right) conventional ligands OLAH+Br and OLAH+OA– and the zwitterionic C3-sulfobetaine. For computational advantage, the OLAH+ is replaced by methylammonium, the OA– by acetate, and the side chain in the zwitterion by a butyl group. At the bottom, the electronic structure of each NC is shown by depicting the molecular orbitals (MOs) close to the valence and conduction bands. The contribution of each atom type to a given MO is represented with a different color (Cs in gray, Pb in orange, and Br in magenta). In this plot, the contribution from the ligands is negligible compared to the full NC due the large number of MOs of the latter. In Figure S17 we, however, illustrate the relative energy alignment of the NC versus the frontier orbitals of the ligands.

Interestingly, the spacing between the cationic and anionic head groups of the sulfobetaines, namely, three or four carbon atoms, has observable experimental effects. The C3-sulfobetaines were better suited for the synthesis of the Cl- and Br-containing perovskites, while their C4 counterparts performed better for synthesizing the iodides, presumably owing to the larger cation–anion distances at the NC surface. This comparison held true for both the oleyl and octadecyl side chains; see the discussion in the Supporting Information and Figures S12 and 13.

A major and very typical issue for CsPbBr3 NCs is facile room-temperature sintering, which quickly renders the material polycrystalline and nonluminescent.[37] In contrast, zwitterionic-capped CsPbBr3 in the form of thin, densely packed films remains in the range of 70–80% of the solution PL QY values, and almost 60% of the initial QY in films stored under ambient conditions is retained even after 10 months. The retention of the quantum confinement in the thin films and the absence of sintering are apparent from the optical absorption spectra (Figure S18). Strong coupling between neighboring NCs facilitates exciton–exciton interactions, enabling multiexciton processes, which favor optical gain in the compact NC medium. When the optical pumping levels substantially exceed one exciton per NC, the population inversion of biexcitonic states is observed as emergence of an amplified spontaneous emission band (ASE, Figure a). The ASE threshold of 2 μJ cm2 (with 100 fs pulses, Figure b) is one of the lowest values reported for solution-processed NC films.[14,38−41]

Figure 3

(a) Amplified spontaneous emission (ASE) spectra showing evolution of the ASE band and (b) the threshold behavior for the intensity of the ASE band. (c) Photoconductivity spectrum inset: photo of a colloidal solution and drop-casted film of standard OA/OLA NCs (left) and C3-sulfobetaine-covered NCs (right). (d) Bias dependence of photoresponse, with the inset showing the scheme of a photodetector made from the substrate with an interdigitated electrode and a drop-casted film of NCs. (e) Corresponding work functions and HOMO–LUMO gaps and (f) current density and luminance vs applied voltage of a LED. Inset: electroluminescence spectrum measured at 3.5 V.

Dense NC packing and hence improved electronic coupling enable the observation of photoconductivity. In CsPbBr3 NC films, the photoresponsivity spectrum closely resembles the optical absorption spectrum, with typical responsivities (R) of about 0.5 A/W (Figure c). Hence, a photoconductive gain close to unity can be estimated (from G = R·hν·e–1, where hν and e are the photon energy and electron charge, respectively). This finding is corroborated by the observation of high PL QYs in these films. G > 1 can be expected only in the presence of secondary, i.e., trap-assisted, photocurrent. The photocurrent vs bias dependence shows saturation above 30–40 V (Figure d), indicating efficient charge collection. The apparently trap-free photoconductivity is also revealed in the linearity of the photocurrent vs incident light intensity plot (at least over 3 orders of magnitude in intensity) and in the relatively large bandwidth of about 90 Hz (Figure S19).

Efficient charge transport and high PL QYs are required characteristics for the eventual use of perovskite NC films in light-emitting diodes (LEDs). To assess the potential of the sulfobetaine-capped NCs, we used a device structure similar to that of Li et al. (Figure e).[12] The current density passing through the devices was rather high, limiting the peak external quantum efficiency (EQE) to 2.5% at 3.5 V (J = 21.7 mA/cm2, L = 1641 cd/m2, Figure f; see the statistics in Figure S20 and the plot of EQE/current efficiency vs voltage in Figure S21), trailing behind the most efficient CsPbBr3 NC LEDs in terms of EQEs (8.73 and 6.27%).[10,12] At the same time, the peak luminance of such devices exceeded 24 000 cd/m2 (Figure f), significantly brighter than the aforementioned efficient LEDs (1660 and 15 185 cd/m2, respectively)[10,12] but lagging behind the Cs/formamidinium mixed-cation bromide perovskite NC LED (55 005 cd/m2).[42] The electroluminescence wavelength and fwhm were 516 and 16 nm, respectively (at 3.5 V; inset in Figure f). Despite sulfobetaine being a long-chain ligand, the charge transport is not severely impeded, seen as high photoconductivity and high current densities in the LEDs. The current densities in our LEDs (current–voltage characteristics, Figures f and S22) are higher than those reported in LEDs from Li et al. (EQE = 6.27%)[12] but without concomitant increase in luminance, thus leading to a lower EQE (2.5%). This reduced efficiency may be due to imbalance between electrons and holes at higher current densities.

In conclusion, a novel class of capping ligands for perovskite NCs is proposed, wherein each ligand molecule is capable of coordinating simultaneously to the surface cations and anions. Colloidal perovskite NCs prepared with tightly bound ligands and without large quantities of excessive capping ligands will serve as an ideal platform for further engineering of these NCs. This may include the development of core–shell NC morphologies with enhanced thermal and environmental stability, as critically needed for applications in displays and lighting, or even for rendering perovskite NCs water-compatible for biomedical applications.