Near-Unity Quantum Yields from Chloride Treated CdTe Colloidal Quantum Dots

Colloidal quantum dots (CQDs) are promising materials for novel light sources and solar energy conversion. However, trap states associated with the CQD surface can produce non-radiative charge recombination that significantly reduces device performance. Here a facile post-synthetic treatment of CdTe CQDs is demonstrated that uses chloride ions to achieve near-complete suppression of surface trapping, resulting in an increase of photoluminescence (PL) quantum yield (QY) from ca. 5% to up to 97.2 ± 2.5%. The effect of the treatment is characterised by absorption and PL spectroscopy, PL decay, scanning transmission electron microscopy, X-ray diffraction and X-ray photoelectron spectroscopy. This process also dramatically improves the air-stability of the CQDs: before treatment the PL is largely quenched after 1 hour of air-exposure, whilst the treated samples showed a PL QY of nearly 50% after more than 12 hours.

. Optimizing CdCl 2 reaction time. a, Absorption and b, photoluminescence spectra with respect to chloride treatment time. c, Ratio of photoluminescence intensity, to absorbance at 400 nm. (Some spectra have been omitted from a and b for clarity).

S3. Calculating the amount of CdCl 2 solution to add
The size of the CdTe CQDs was calculated using the following formula [3] D = (9.8127 x 10 -7 ) λ 3 -(1.7147 x 10 -3 ) λ 2 + 1.0064 λ -194.84 (1) where D is the average diameter of the CdTe CQDs ( in nm) and λ is the wavelength (in nm) corresponding to the absorption band edge.
The concentration of a CdTe CQD solution, C, ( in units of mol.dm -3 ) was calculated by [1] C = A/(10043D 2.12 )L (2) where A is the absorbance at the absorption band edge, and L is the path length through the solution.

a) b) c)
The total surface area of CdTe CQDs in solution, SA, is calculated from the product of the surface area of a single CQD and the concentration and volume, V, of the solution: Hence, the volume of 0.33 mold.m -3 CdCl 2 solution used in a chloride treatment, V Cl , so that it corresponds to 96 Clper nm 2 of CQD surface is calculated to be V Cl = 48SA/0.33 (4).

S4. Analysis of X-ray photoelectron spectroscopy data.
The sampling depth in X-ray photoelectron spectroscopy (XPS) is defined as 3 times the inelastic mean free path, 3; 95 % of the signal originates from this region. Given that the photoelectron flux emerging from the sample drops exponentially with distance below the surface, all the measurements described here primarily probe the first few nm of the nanoparticle surface, as 63% of the signal is received from within a distance of the surface. In a typical experiment, electron kinetic energies between 115 eV and 1000 eV were probed, corresponding to sampling depths of 4.1 -11.1 nm. This enabled us to quantify the surface composition of the QDs. Figure S3 shows the XPS data used to obtain the composition vs. depth plot in Figure 2c: the is omitted from the analysis in Figure 2b. Small depth-dependent signals due to P 2s were also detected, and are associated with residual TOP ligand and TDPA. Figure S3. a, Cd 3d (ca. 405, 412 eV binding energy) and N 1s (centered at ca. 400 eV binding energy) and, b,Cl 2p XPS spectra for 3.5 nm Cl-passivated CdTe QDs as a function of photoelectron kinetic energy (from 115 eV (red) to 1000 eV (black)) and hence sampling depth. The sample has been exposed to air after deposition for 20 min. The spectra are normalized to the Cd 3d 5/2 intensity at ca. 405 eV binding energy, and are corrected for photoionization cross section, except that of N 1s.
As the flux of photoelectrons is attenuated according to the Beer-Lambert law as it emerges from the sample, it is possible to obtain an estimate of the thickness of the ligand shell by a simple model incorporating the inelastic mean free paths. We adopt a 'two-layer model' that assumes a ligand shell overlies the core, and use the analytical model developed by Shard et al. which accounts for a surface made up of spherical particles with radii of the same order as the inelastic mean free path length [4][5][6] . This calculation requires us to estimate the number of organic ligands (oleylamine, TOP and TDPA) coordinated to the surface of the QD. Using the relative intensities of the C 1s, N 1s, P 2s, Cl 2p and Cd 3d signals as a guide, we carried out ligand shell thickness calculations for a number of reasonable Cd:C ratios (from 1:1 to lower coverages corresponding to 2:1, by analogy with previous work [7] ). For each, we calculated the ligand shell thickness for 6-8 different sampling depths; in principle the thickness obtained should be independent of the sampling depth used (within experimental error), and this test gives a useful indicator of the validity of the model chosen.
In order to obtain this condition, we found it was necessary to assume a Cd:C ratio of ca. 1:1, i.e.
roughly one organic ligand is bound to each surface CdTe unit. We obtain a ligand layer thickness of ca. 3.7 nm for a CdTe QD diameter of 4.5 nm. When this is compared with the length of the longest ligand (TDPA, ca. 2.1 nm), it can be seen that the calculated average organic layer thickness is more than one ligand layer. We assume that this corresponds to one layer of directly coordinated ligands with some excess ligand interdigitated into the ligand shell (consistent with the observation of signals from both protonated and unprotonated oleylamine), together with residual solvent molecules.

S5 Valence band photoemission studies.
In order to study the bonding of the chlorine at the CdTe nanoparticle surfaces, we have carried out valence band photoemission studies ( Figure S4). Unpassivated CdTe QDs react swiftly to form surface oxide in air, even during the short (few-minute) period of air exposure necessary to insert the samples into the UHV spectrometer (as evidenced by a strong Te 3d doublet lying 3.6 eV to higher binding energy than the CdTe signal in core level XPS, assigned to TeO 2 [9] ). For this reason, observations (see Figure 4 in themain text). The substantial increase in the valence band density of states at around 4 eV BE is consistent with DFT calculations for Cl-passivated PbS. [11] In the unpassivated QDs, surface states are thought to give rise to reactive partially-filled states in the band gap, which act as trapping sites. [7] In the chloride anion passivation process, we expect the Clion to donate electrons into these states, giving rise to new filled states (removing some empty states) in the band gap. These filled states are directly probed by the photoemission process. We note that DFT calculations for chloride-passivated Si nanocrystals also predict the observed band gap narrowing on Cl-passivation [12] We therefore propose that Cl bonds to the surface by the anion donating electrons into partially-filled reactive surface states, creating new completely filled states.
Although the noise level is significant, it appears that on ageing the CdTe/Cl, there may be a small decrease in the density of states at the VBM (at around 1-2 eV BE), which is consistent with the small blue shift observed over time in the PL of aged samples (see Figure 4 in the main text, and Figure S5 below). This is consistent with the onset of some initial oxidation resulting in a small decrease in size of the CdTe 'core' of the QDs as its surfaces become oxidized. This blue shift due to oxidation has been observed in a number of other chalcogenide systems. [13,14] Intensity (arb. units) The 'aged' samples have been exposed to air for ca. 66 h, while the fresh samples are exposed to air for 15 -90 min during sample mounting and pump-down of the spectrometer load-lock. The data are normalized to the Cd 4d intensity in all cases, and referenced to the Cd 4d BE. [8] S6. Effect of air-exposure on the absorption spectrum. Figure S5 shows the change in absorbance spectra on oxidation for the same sample used to produce Figure 4b in the main manuscript. The position of the first absorption peak decreases from 582 nm to 562 nm over 13 hours of air-exposure. If it is assumed that this shift in wavelength is caused by a change in CQD size then, using equation (1) above, this corresponds to a reduction in diameter of 0.16 nm. Figure S5: The change in the absorbance spectrum of a chloride-treated sample on air-exposure.

S8 High resolution (scanning) transmission electron microscopy
High resolution electron microscopy was used to probe the crystallinity and chemistry of the nanocrystals at the individual particle level. The area of the EELS spectrum image is shown by the green square in (a). As expected the particle contains cadmium and tellurium while oxygen (and carbon) are not localised within the nanoparticle but associated with the carbon support film and remaining surface ligands. Elemental maps were extracted after appropriate background subtraction for energy windows of (a) 444-528eV, (b) 588 -678eV and (c) 533-563 eV. This elemental analysis is in agreement with energy dispersive x-ray (EDX) spectroscopy data for the untreated sample which showed only cadmium and tellurium signals associated with the nanoparticles. All other elemental signals present in the EDX spectra (oxygen, carbon and copper) are known to originate either from the surface ligands or from the holey carbon/copper TEM support grid.

S7 Particle size analysis
In order to gain an accurate picture of the size and shape of the nanoparticles automated image analysis was performed on a large number of HAADF STEM images. The HAADF STEM imaging mode was chosen in preference to conventional bright field TEM images as these images give considerably higher contrast between the nanoparticles and the carbon support. Automated image analysis allows several hundred particles from each sample to be measured in a consistent manner, generating statistically significant data about the nanoparticle size and shape distributions. Fig. S9 shows the automated image processing procedure and Fig. S7 shows the resulting data, represented in histograms showing the spread of particle sizes and aspect ratios in both samples.

Fig. S9 Automated size analysis of HAADF STEM images.
As the line scan in (a) demonstrates, the intensity change going from the support to the particle is not a sharp step profile but a continuous gradient, this is due the spherical particles being thinnest at their edge and thickest in the centre. Defining the edge of a particle for size analysis is therefore not entirely straightforward and a consistent definition of the edge is required for automated image analysis. Automated size analysis was performed in the ImageJ software after the HAADF STEM Images were converted to 8-bit tiffs. The mean intensity for the carbon support was identified and subtracted as a constant background. The threshold intensity value could then be set as 10% of the maximum intensity to create thresholded images like that shown in (b). ImageJ particle analysis was then applied (c) with maximum and minimum areas defined as 30 nm 2 4 nm 2 respectively so as to avoid the detection of single heavy atoms as small particles or touching neighbours as large particles. This approach allowed the rapid analysis of several hundred particles from each sample.

Fig S10 Analysis of the size and morphology of nanoparticles from HAADF STEM images.
Based on the automated size analysis procedure outlined in Fig. S6, we were able to perform size analysis on 251 particles from the untreated sample (left hand column) and 542 particles from the Cl treated sample. In this analysis height and width were determined by fitting a bounding rectangle to the particle outlines. The untreated particles are found to have a mean width of 5.06 ± 0.52 nm and an mean height of 4.93 ± 0.55 nm; while the treated particles are found to have an mean width of 4.58 ± 0.76 nm and an mean height of 4.57 ± 0.70 nm. All errors quoted correspond to one standard deviation. Histograms of the resulting data show that both sets of nanoparticles have a relatively narrow size distribution and that the majority of particles are close to spherical in shape having aspect ratios close to 1 (mean aspect ratios of 1.03 and 1.01 for the untreated and the Cl treated respectively).