Designer phospholipid capping ligands for soft metal halide nanocrystals

Abstract

The success of colloidal semiconductor nanocrystals (NCs) in science and optoelectronics is inextricable from their surfaces. The functionalization of lead halide perovskite NCs^1-5 poses a formidable challenge because of their structural lability, unlike the well-established covalent ligand capping of conventional semiconductor NCs^6,7. We posited that the vast and facile molecular engineering of phospholipids as zwitterionic surfactants can deliver highly customized surface chemistries for metal halide NCs. Molecular dynamics simulations implied that ligand-NC surface affinity is primarily governed by the structure of the zwitterionic head group, particularly by the geometric fitness of the anionic and cationic moieties into the surface lattice sites, as corroborated by the nuclear magnetic resonance and Fourier-transform infrared spectroscopy data. Lattice-matched primary-ammonium phospholipids enhance the structural and colloidal integrity of hybrid organic-inorganic lead halide perovskites (FAPbBr₃ and MAPbBr₃ (FA, formamidinium; MA, methylammonium)) and lead-free metal halide NCs. The molecular structure of the organic ligand tail governs the long-term colloidal stability and compatibility with solvents of diverse polarity, from hydrocarbons to acetone and alcohols. These NCs exhibit photoluminescence quantum yield of more than 96% in solution and solids and minimal photoluminescence intermittency at the single particle level with an average ON fraction as high as 94%, as well as bright and high-purity (about 95%) single-photon emission.

Fig. 1. Surface chemistry of soft metal halide NCs.

a, Examples of ionic metal halides. b,c, Atomistic depiction of surface stabilization (b) and disintegration (c) of APbX₃ perovskite by common long-chain cationic (Cat⁺) and anionic (An⁻) ligands due to excess ligand quantity and low internal crystal energy. d, Zwitterionic molecules offer stronger (multidentate) binding and reduced reactivity owing to their neutral charge. e, Structural engineering of the head, bridge and tail groups unlocks their broad utility for stabilizing diverse metal halide NCs and dispersing them in various media.

Fig. 2. Binding of zwitterionic ligands to the FAPbBr₃ NC surface.

a, Schematics of different modes for binding of zwitterionic ligands, whose plausibility was assessed with replica-exchange MD simulations. b, Evolution of the BM populations, computed in a 50-ns-long replica-exchange MD simulation of a single PC ligand that was initially placed on the pristine FABr-rich (100) surface of FAPbBr₃. BM3 prevails in the ensemble in the late stages of the simulation. c, Results of a REDOR NMR experiment for ³¹P–²⁰⁷Pb coupling supports the theoretical prediction of surface anchoring with the phosphate group. Theoretical REDOR curves were calculated using conformations obtained from the MD simulations. d, FTIR spectroscopy analysis of the (P–O) stretching region, wherein the P–O bond weakening on ligand binding (that is, P–O–Pb bridge formation) shifts the signal to lower frequencies. ν_as, asymmetric stretching vibration.

Fig. 3. Ligand head-group engineering.

a,b, Geometries of the PC (a) and PEA (b) head groups on the (100) surface of FAPbBr₃, emphasizing the improved PEA fitness for the surface A-site. c–f, Snapshots from replica-exchange MD simulations of the FAPbBr₃ surfaces in which 50% (c,d) or 100% (e,f) of FABr pairs were substituted with the ligands. Ligand molecules are colour-coded according to their BM—BM1 (blue), BM2 (green), BM2' (orange) and BM3 (red). Although stable surfaces are observed at 50% substitution for both ligands, PC (c) and PEA (d), a noticeable number of PC ligands and FA and Br ions desorb from a PC-capped surface, leaving behind vacancies in the top-most surface layer (black dashed circles in c). At 100% substitution, the PC-capped surface starts to rupture (e), whereas it remains stable in the case of the PEA ligand (f). g,h, Colloids of purified hexadecyl-PEA-capped and hexadecyl-PC-capped purified FAPbBr₃ NCs (8.5 nm), as prepared (g) and after 7 days (h). i, Typical HAADF-STEM image of FAPbBr₃ NCs capped with the alkyl-PEA ligand. Scale bar, 20 nm.

Fig. 4. Examples of functional tail engineering.

a, Synthesis scheme for PEA ligands tested in this work, with 21 ligand tails shown in Extended Data Fig. 6. b–g, Different tails enable matching of the solvent polarity (b) with highly specific dispersibility, shown for FAPbBr₃ NCs (c–e) capped with aliphatic (c), aromatic (d), halogenated (e) tails and CsPbBr₃ NCs (f,g) with polyether tails. h, PPG-PEA-capped CsPbBr₃ NCs in PGMEA at a concentration of 2 g (CsPbBr₃) per ml of dispersion. i–k, Engineering inter-NC distance in monolayers of CsPbBr₃ NCs capped with PEA ligands with polystyrene tails by adjusting the ligand molecular weight: M_n = 900 Da (h), 1,200 Da (i) and 5,000 Da (j). Scale bars, 20 nm (i–k, main).

Fig. 5. Light emission from C8C12-PEA-capped FAPbBr₃ NCs on the ensemble and single dot levels.

a–d, Highlighted optical properties of FAPbBr₃ NCs in an ensemble. Films of FAPbBr₃ NCs of various sizes demonstrate equally high PL QY and tunable green emission (a). High PL QY is retained in thick films (b) and is not influenced by film storage in ambient conditions for at least 90 days (c). PL QY of greater than 90% is preserved on approximately 1,000-fold dilution of C8C12-PEA-capped NCs with octane, whereas a pronounced drop is observed for C8C12-PC- and OAm-capped NCs (d), indicative of dilution-induced surface degradation. e–i, Optical properties of FAPbBr₃ NCs at the single-particle level. Single OAm-capped NCs exhibit pronounced blinking (e), while single PEA-capped NCs exhibit high brightness and suppressed PL blinking (f), good PL stability (g), narrow PL linewidth (full-width at half-maximum (FWHM)) (h) and high single-photon purity (i). PL QY measurement error in (d) is ±1%. a.u., arbitrary unit. Scale bar, 10 mm.

Extended Data Fig. 1. Classification of ligand binding modes.

a, A typical map of configurations observed in replica-exchange MD simulations of the PC ligand on the (100) FAPbBr₃ surface. d_N-slab and d_P-slab define distances from the ligand head groups (nitrogen and phosphorus atoms, correspondingly) to the middle atomic plane of the slab. Four well-defined clusters of configurations correspond to the different binding modes of the ligand (Fig. 2). The diffuse region at large ligand-slab separations corresponds to free ligands. b–e, Snapshots of the observed binding modes - BM2 (b), BM3 (c), BM1 (d) and BM2′ (e).

Extended Data Fig. 2. Generality of the prediction that zwitterionic ligands tend to displace both A and X ions from the perovskite surfaces.

a–f, Evolution of binding mode populations in replica-exchange MD simulations of single PC (a,b,c) and PEA (d,e,f) ligand molecules on the FAPbBr₃ surface (a,d), (001) CsPbBr₃ surface (b,e), and (010) CsPbBr₃ surface (c,f). Binding with a displacement of both A and X ions (BM3) is thermodynamically preferred in all studied systems. *Crystallographic orientations refer to the primitive unit cell of CsPbBr₃ (see Supplementary Note 1 for more details).

Extended Data Fig. 3. FAPbBr₃ surfaces at various [FABr] and [Lig] concentrations.

a–f, Evolution of binding mode populations in systems with [Lig] = 50% and with a varying amount of surface FABr – 100% (a,b), 50% (c,d), and 0% (e,f). g,h, Evolution of binding mode populations in systems with [Lig] = 100% and [FABr] = 0%. In all scenarios, BM3 was identified as a dominant binding mode. The population of BM3 is also systematically higher for PEA ligand compared to PC, indicating a better fit of the former to the FAPbBr₃ surface. In addition, some new binding modes were discovered in systems where the rupture of the PbBr underlayer is observed — BM-B (ammonium in the PbBr layer), BM-B’ (phosphate in the PbBr layer), and BM-C (both ammonium and phosphate in the PbBr layer). However, these are marginal and are encountered only along the phase boundaries between the newly exposed FABr- and the original PbBr-terminated surfaces. i–p, Corresponding MD snapshots with ligands being color-coded according to their binding mode — BM1 (blue), BM2 (green), BM2’ (orange), BM3 (red), unbound (black), and other (gray).

Extended Data Fig. 4. Stoichiometry-dependent stability of (100) FAPbBr₃ surfaces.

a,b, Surface stability maps as a function of ligand and FABr concentrations computed for PC (a), and PEA (b) ligands. c–f, MD snapshots that illustrate three different regions observed on the stability maps. At high ligand+FABr concentrations, excess ions and ligands separate from the surface (blue area on the map) (c). As a result, the system acquires equilibrium stoichiometry corresponding to the green region on the map (d). At low ligand and/or FABr concentrations the surface becomes unstable again (a red region on the map) — segregation into two surfaces is observed for incomplete FABr passivation (f), whereas partial coverage solely with the ligand causes rupture of the PbBr underlayer (e). Ligand molecules in the snapshots are color-coded according to their binding mode — BM1 (blue), BM2 (green), BM2′ (orange), BM3 (red), unbound (black), and other (gray).

Extended Data Fig. 5. Synthesis of APbBr₃ NCs and testing of ligands.

a, In TOPO-DOPA synthesis, PbBr₂-trioctylphosphine (TOPO) complex reacts with diisooctylphosphinic acid (DOPA) salt of the corresponding A cation at room temperature, yielding monodisperse NCs (1). TOPO and DOPA are known as “bad ligands” and can be readily exchanged for zwitterion ligands (2a). If no capping ligands are added promptly after NCs are formed, the NCs rapidly lose their colloidal and structural integrity (2b). b, To purify phospholipid-capped NCs (see also Supplementary Note 6), a suited antisolvent is added, followed by centrifugation (3), whereas a supernatant containing unreacted precursors and free ligand molecules is discarded (4). If the ligand is’ good’, the resulting NCs pellet is redispersed in a suited solvent, yielding a stable colloid (5a). With a ‘bad’ ligand, NCs redisperse incompletely, do not redisperse (5b), or lose their colloidal integrity upon storage (6). Such a purification cycle (steps 3, 4, and 5a) can be repeated several times. c,d, In this experiment, two head groups and two tails were compared. Combining a ‘bad’ head group and a ‘bad’ tail (c) leads to the worst colloidal stability, while combining a ‘bad’ head group and a ‘good’ tail might still yield long-term stable colloids (d). Similar to FAPbBr₃, MAPbBr_3, and CsPbBr₃ (c) NCs with PC head-group and a ‘bad’ hexadecyl tail precipitate after three rounds of purification, with visible deposit highlighted with a white box on the photo. CsPbBr₃ NC colloids with both PEA or PC head-groups and a ‘good’ glycerodioleyl tail (d, commercially available) yield stable colloids; however, with different ligand coverage, in agreement with MD prediction. Furthermore, PC-based surface ligation is more labile, with both “good” and “bad” tails: NCs tend to increase their mean particle size (inset TEM images, scale bars 50 nm) and acquire more irregular NC morphology with storage or several purification cycles.

Extended Data Fig. 6. Survey of synthesized and commercial phospholipid ligands.

a, Tail groups of the synthesized PEA ligand library. b, Synthetic pathways for PPA, PEA, PC and N-alkyl substituted PEA ligands. b, Commercially available PC and PEA ligands. d, Multi-zwitterionic ligands.

Extended Data Fig. 7. Photocatalysis with poly(propylene glycole) PEA capped CsPbBr₃ NCs.

a, Photocatalytic C–C coupling reaction through activation of C–Br bond in benzyl bromide, catalyzed by CsPbBr₃ NCs. b, ¹H NMR of the final reaction mixture showing peaks from the starting material (around 4.4 ppm) and product (around 2.2 ppm) and how the reaction yield was calculated from integrating the two peaks. c, A table summarizing product yields in different solvents and with NCs capped with either commercial lecithin or poly(propylene glycole) (PPG)-PEA.

Extended Data Fig. 8. MAPbBr₃ and CsPbBr₃ single-dot emission at room temperature.

These NCs were capped with C8C12-PEA-ligands and purified three times, followed by dilution (×10⁵) and spin-coating onto a glass substrate. a,d, Blinking traces of MAPbBr₃ and CsPbBr₃ single dots. b,e, Pronounced antibunching evidence high purity of single photon emission. c,f, Stable and narrow emission from a single MAPbBr₃ NC (c) or CsPbBr₃ NC (f).

Extended Data Fig. 9. Head-group optimization by tuning the bridge length.

a–c, Lead iodide perovskites have a larger distance between A and X surface sites than bromides. Positive-to-negative moieties distance in the zwitterion thus has a pronounced effect on the ligand binding. Mixed halide CsPb(Br/I)₃ NCs (without anti-solvent purification, synthesis details in Supplementary Table 7) capped with phosphoalkylamine ligands featuring different distances between ammonium and phosphate functionalities (a): PEA, PPA, and PBA. After the first purification step with anti-solvent (b), PEA-capped NCs drop in PLQY from 95% to 42%, while longer-bridge PPA and PBA ligands retain high PLQY and NCs shape (c). d–g, In general, PEA ligands better suit Br-rich compositions, while PPA ligands make for a better choice for I-rich compositions. h,i, Mixed halide CsPb(Br/I)₃ NCs capped with 3-ammoniopropyl (2-octyl-1-dodecyl) phosphate (PPA-R6) display remarkable spectral stability during purification cycles with antisolvent (ethyl acetate:acetonitrile).

Extended Data Fig. 10. A compositional variety of metal halide NCs that can be stabilized with alkylphospholipid zwitterionic capping ligands.

a,b, Lead-free metal halides: low-dimensional (a) or double perovskites (b). c–e, Stable lead halide perovskite NCs with all three cations: Cs (c), FA (d) and MA (e), as well as of varying halide composition, can be prepared using zwitterionic ligands with phosphate and primary ammonium head-groups and varying bridge length (PEA, PPA, PBA) (additionally see Supplementary Figs. 15–17).