Colloidal CsPbX3 Nanocrystals with Thin Metal Oxide Gel Coatings

Lead halide perovskite (LHP) nanocrystals (NCs) have gathered much attention as light-emitting materials, particularly owing to their excellent color purity, band gap tunability, high photoluminescence quantum yield (PLQY), low cost, and scalable synthesis. To enhance the stability of LHP NCs, bulky strongly bound organic ligands are commonly employed, which counteract the extraction of charge carriers from the NCs and hinder their use as photoconductive materials and photocatalysts. Replacing these ligands with a thin coating is a complex challenge due to the highly dynamic ionic lattice, which is vulnerable to the commonly employed coating precursors and solvents. In this work, we demonstrate thin (<1 nm) metal oxide gel coatings through non-hydrolytic sol–gel reactions. The coated NCs are readily dispersible and highly stable in short-chain alcohols while remaining monodisperse and exhibiting high PLQY (70–90%). We show the successful coating of NCs in a wide range of sizes (5–14 nm) and halide compositions. Alumina-gel-coated NCs were chosen for an in-depth analysis, and the versatility of the approach is demonstrated by employing zirconia- and titania-based coatings. Compact films of the alumina-gel-coated NCs exhibit electronic and excitonic coupling between the NCs, leading to two orders of magnitude longer photoluminescence lifetimes (400–700 ns) compared to NCs in solution or their organically capped counterparts. This makes these NCs highly suited for applications where charge carrier delocalization or extraction is essential for performance.

TOP-Br2 (0.5 M): TOP (6 mL, 13.5 mmol) taken from the glovebox and reagent grade mesitylene (18 mL) were mixed in a capped 40 mL vial. Elemental bromine (0.6 mL, 11.6 mmol) was carefully added under vigorous stirring while controlling the temperature with a water bath.
Zwitterion capped CsPbBr3 NCs: A previously reported procedure was upscaled for this synthesis. 3 Pboleate (5 mL, 0.5 M, 2.5 mmol), Cs-oleate (4 mL, 0.4 M, 1.6 mmol), ASC18 (201 mg, 0.48 mmol) and ODE (10 mL) were added to a 50 mL three-necked-flask and heated to 100 °C under vacuum. Once the gas evolution stopped, the flask was put under an inert atmosphere and the temperature was adjusted to 130 °C. After reaching the reaction temperature, TOP-Br2 (5 mL, 0.5 M, 2.5 mmol) was injected and the reaction was cooled to room temperature immediately using an ice-bath.
To the crude solution 3.5 eq. ethyl acetate were added and the solution was centrifuged at 12.1k rpm (20130 g) for 1 min. The supernatant was discarded and the precipitate dispersed in 12 mL toluene. The NCs were washed two more times by precipitation with 36 mL ethyl acetate, centrifugation and dispersion in 12 mL toluene. After the last dispersion the solution was once more centrifuged at 12.1k rpm (20130 g) for 2 min to remove aggregated particles.
The concentration of the final solution was determined from the absorption spectra using the absorption coefficient reported by Maes et al. 2 Metal oxide sol-gel coatings: Alumina gel coated CsPbBr3 NCs: In a typical synthesis 0.12 mmol of the ASC18-capped NCs were mixed with ODE (12 mL) and the toluene was evaporated under vacuum. In the glovebox a previously prepared solution of Al(O s Bu)3 (0.36 mL, 0.5 M in mesitylene, 0.18 mmol, 1.5 eq.) was taken and mixed with a solution of AlBr3 (48 mg, 0.18 mmol, 1.5 eq.) in 0.4 mL anhydrous mesitylene. This results in roughly 0.8 mL of Al2Br3(O s Bu)3 precursor solution which is injected into the NC solution at room temperature under continuous stirring. Due to the reactivity of this precursor solution to ambient humidity, it was transferred to the reaction flask using a sealed syringe. The reaction was then heated to 120 °C as fast as possible using a heating mantle and kept at 120 °C for 10 min. After the reaction period, the flask was cooled back to room temperature using a water bath. The NCs were precipitated from the crude solution with acetone (12 mL). After precipitation the turbid solution was centrifuged at 12.1k rpm (20130 g) for 1 min and the supernatant discarded. The NCs were washed by dispersion in n-butanol (1 mL) and precipitation with diethyl ether (20-40 mL depending on the colloidal stability) followed by centrifugation at 12.1k rpm (20130 g) for 1 min. This washing step can be repeated any number of times and will give a clean colloid (insignificant leftover of the alumina gel in solution as evidenced by ICP-MS) after 3-5 steps. The product was finally dispersed in 2-6 mL of an alcohol (ethanol, isopropanol or n-butanol) and centrifuged once more at 12.1k rpm (20130 g) for 2 min to remove any aggregated particles.

Alumina gel coating of different compositions:
The protocol remains the same as described above, but any mixed Cl-Br or Br-I composition can be used instead of pure CsPbBr3 NCs and AlCl3 or AlI3 be used instead of AlBr3. The final composition (and therefore color) will be the stoichiometric mix of the used NCs and the aluminum halide. In order to obtain the intended emission color of the sol-gel coated NCs, the halide composition of the organically capped NCs had to be taken into account when calculating the aluminum halide ratios for the coating. This also means, that one type of halide cannot fully be exchanged for another halide.

Zirconia and Titania coatings:
The synthetic protocol as well as the washing procedure remained the same as for the alumina-gel-coated NCs. For zirconia coatings ZrBr4 and Zr(OBu)4 were used, for titania coatings TiBr4 and Ti(O i Pr)4 were used.

Characterization
Absorption spectra: Optical characterizations were performed at ambient conditions. UV-Vis absorption spectra of colloidal NCs were collected using a Jasco V670 spectrometer in transmission mode. The NC concentrations were determined from the absorption spectra using the absorption coefficient reported by Maes et al. 2 For the measurements NCs solutions were diluted down to 20-50 µg/mL. Zwitterion-capped NCs were dispersed in either hexane or toluene. Metal-oxogel-coated NCs were dispersed in butanol.

Photoluminescence (PL):
A Fluorolog iHR 320 Horiba Jobin Yvon spectrofluorometer equipped with a PMT detector was used to acquire steady-state PL spectra. NC solutions were measured in the same dilutions and solvents as the absorption measurements. NC films were obtained by spincoating concentrated solutions (≈10 mg/mL) and then measured in a 90° geometry.

Photoluminescence quantum yield (PLQY):
Absolute PL QYs of films and solutions were measured with a Hamamatsu C13534 Quantaurus-QY Plus UV-NIR absolute PL quantum yield spectrometer. The same solutions and films that were used to measure PL were also used to measure QY.

Time-resolved photoluminescence (TRPL):
Time-resolved PL traces were acquired in solution using a FluoTime300 spectrometer from PicoQuant.
Electron microscopy: TEM images were collected using a JEOL JEM-1400 Plus operated at 120 kV. SEM EDX was done on a FEI Quanta 200F equipped with an Octane Elect Super EDS system and was operated at 30 kV. NCs were deposited on carbon coated copper TEM grids from dilute solutions by drop casting.

X-Ray diffraction (XRD):
XRD patterns were collected with a STOE STADI P powder diffractometer operating in transmission mode. A germanium monochromator, Cu Kα irradiation (λ = 1.540598 Å) and a silicon strip detector (Dectris Mythen) were used. In order to measure an XRD pattern of NCs they were precipitated and dried to a powder which was then put between two stripes of Scotch tape.
Zeta potential (ZP): ZP measurements were performed on a Malvern Zetasizer Nano ZS. Ethanol absolute or anhydrous butanol were used as solvent, with crude NC solutions being diluted 100-300 times (to roughly 0.1 mg/mL) and the voltage being fixed to 10 V. The measured mobility data was automatically converted to zeta potential using Smoluchowski's theory and then fitted with a Gaussian function to obtain the average zeta potential.
NMR spectroscopy: 27 Al solid-state Magic Angle Spinning (MAS) NMR was measured on a 16.4 T Bruker Avance III HD spectrometer (Bruker Biospin, Fällanden, Switzerland). The instrument was equipped with a 2.5 mm double-channel solid-state probe head. The spinning frequency was set to 20 kHz. Chemical shifts were referenced to Al(NO3)3 in D2O (1.1 M). For the 1D spectra a full echo experiment was used with an excitation pulse of 3 µs. The echo delay was set to 40 cycles. 8192 transients were acquired with a recycle delay of 1 s. The 2D multi quantum MAS (MQMAS) spectra were acquired using a Bruker 3Q MAS pulse program for odd half integer spin nuclei, using 3 pulses with full echo acquisition (mp3qdfs). The 90-degree hard pulse was set to 3.5 µs and the selective refocusing pulse to 40 µs. 2048 transients were acquired. The echo build-up time was set to 15 rotor cycles to detect a full echo.

ICP-MS:
The elemental composition was measured on an ICP-QMS (Agilent 7500cs, United States). A microwave digestion approach including 20-30 mg sample, 1 ml HNO3, 0.2 ml cobalt solution (digestion recovery control) was used. Indium was used as an internal standard. Three subsets of each sample were digested. Each subset was measured 3 times. External calibrations including Al, Co, Cs, Pb and In as an internal standard were used. The confidence intervals were calculated based on nine measurements with a confidence level of 95%.

Alumina Gel:
Pure alumina gel was synthesized by reacting AlBr3 and Al(O s Bu)3 in ODE at 120 °C for 10 min. The product was then precipitated with acetone and dispersed in butanol. The product was deliberately not washed further in order to keep as much of the gel as possible.

PXRD pattern of alumina-gel-coated LHP NCs:
Besides being very thin the coating is also inherently amorphous due to being a gel. To confirm this, we measured powder X-ray diffraction of coated NCs which were not washed so that the most amount of alumina gel remains in the sample. The obtained diffractogram clearly shows the orthorhombic Pnma structure of the perovskite NCs 5 but no peaks appear for the alumina gel. 4 Figure S3: XRD pattern of a typical alumina-gel-coated CsPbBr3 NC sample, matching orthorhombic (Pnma) CsPbBr3 (red lines) 5 but showing no signs of any other crystalline component. 4

SEM EDX:
Another well-established tool for elemental analysis is energy-dispersive X-ray spectroscopy (EDX). SEM EDX was performed on a well washed sample in an attempt to determine the elemental composition of intact NCs. Unfortunately, the characteristic X-rays for the Al Kα overlap with the Br Lα so closely that they become indistinguishable (see Fig. S4). Comparing the integrated Br Lα and Br Kα ratios of pristine and coated NCs could in principle be used to determine the amount of Al. However, the 29 % increase in signal intensity of alumina gel coated compared to pristine NCs was not deemed significant enough over the inherent variance in signal intensity for low energy X-rays to be used as a proof for the presence of Al. Figure S4: SEM EDX comparison of ASC18 capped NCs (black) and alumina-gel-coated NCs (red). The spectra are normalized to the Br Kα peak and the inset shows a zoomed in cutout of the combined Br Lα and Al peak. Coating Thickness calculations:

ICP-MS:
The ICP-MS results gave us a ratio of Al to Pb (and Cs) and assuming that in the case of the well washed NCs all Al is located on the surface of the NCs in form of a coating, we can calculate the thickness of this coating. For the sake of this calculation we are additionally assuming that we are dealing with perfect cubes, although analogous calculation for the spherical NCs would lead to the exact same result. The calculation uses the volume of a cube and a coated cube as well as the density of the two materials to calculate the coating thickness. The ratio r of Al to Pb can be expressed as the molar ratio of the two compounds with a correction factor of 2 to accommodate for the 2 Al in Al2O3 and n being the molar amount.
The molar amounts can be substituted with the molecular weight M and the mass m using equation (2) which leads to equation (3).
The mass can be substituted with the volume V and the density δ using equation (4).
For the two volumes we can look at a cube with diameter d and a coating thickness s. The volume of the perovskite VP is given in equation (6) and the volume of the alumina gel coating VAl is given in equation (7). Putting these into equation (5) leads to equation (8) For clarity we pulled all constants together and made a factor = 2 × . The density of the coating was purposely not put into the factor so it could easily be changed afterwards for estimations on coatings with lower density. Using = 579.82 ⁄ , = 101.96 ⁄ and = 4.8

⁄
the factor a equals to = 2.369 3 ⁄ . Equation (8) was then rearranged for the coating thickness s giving (9).
Since the density of the alumina sol-gel is unknown, we used the numbers of Al2O3 (  ) as the lower and upper limit for the coating thickness (upper and lower limit for the coating density respectively). With a NC size of 10 nm, a ratio r of 0.5 would only give a coating thickness in the range of = 0.09 − 0.33 .

Zeta Potential measurements of alumina-gel-coated LHP NCs:
Electrophoretic mobility measurements were conducted on the solutions of NCs diluted to 0.1 mg/mL in ethanol or butanol, at 8-12 V. The reference measurement cannot be performed for the alumina gel itself due to the uncontrolled size of the gel species. The closest analogy to the alumina-gel-coated NCs would be alumina nanoparticles since the charges originate from the same surface-adsorbed hydroxo species in both cases. Reported zeta potentials for alumina nano-/microparticles at pH 7 range from +20 to +50 mV which matches with the measured data. [6][7][8] Figure S5: Zeta potential measurements plotted with a logarithmic scale, showing that over the measured range the zeta potential decreases linearly with the logarithm of the amount of base added. This is similar to the linear dependence of the zeta potential with pH that is often reported for aqueous solutions. The first point at 2.5 equivalents does not follow the trend, presumably because the amount added is so low, that its effect is basically quenched.

Sizing curve and calculation:
The average NC size s (in nm) of a sample was determined from the maximum of the first excitonic peak a (in eV) in the absorbance spectra using the sizing curve and formula reported by To get the correct maximum of the first excitonic peak the absorbance spectra were converted to energy scale and the second derivative was calculated. This eliminates the inherent underlying rise in absorbance and the possible scattering that would obscure the correct maximum.