Uncovering active precursors in colloidal quantum dot synthesis

Studies of the fundamental physics and chemistry of colloidal semiconductor nanocrystal quantum dots (QDs) have been central to the field for over 30 years. Although the photophysics of QDs has been intensely studied, much less is understood about the underlying chemical reaction mechanism leading to monomer formation and subsequent QD growth. Here we investigate the reaction mechanism behind CdSe QD synthesis, the most widely studied QD system. Remarkably, we find that it is not necessary for chemical precursors used in the most common synthetic methods to directly react to form QD monomers, but rather they can generate in situ the same highly reactive Cd and Se precursors that were used in some of the original II-VI QD syntheses decades ago, i.e., hydrogen chalcogenide gas and alkyl cadmium. Appreciating this surprising finding may allow for directed manipulation of these reactive intermediates, leading to more controlled syntheses with improved reproducibility.


S1
arm of the flask was analysed to ascertain a background concentration of gasses. The Teflon pin of the flask was then opened allowing the headspace of the reaction to mix with the atmosphere in the arm of the flask. A second sample of gas was taken through the septum and the relative concentrations of gasses before and after mixing were determined. A calibration curve was created by varying the amounts of 13 CO 2 added to a schlenk flask filled with an equivalent volume of tetradecane as in the experimental unknowns with an argon internal standard that was added equally across all samples. The ratio of counts versus the internal standard was calculated by dividing the counts of 13 CO 2 (m/z=45) by those of Ar (m/z = 40). The average for each concentration was taken over five replicates.
NMR characterization Proton ( 1 H), carbon ( 13 C), phosphorus ( 31 P), and Selenium ( 77 Se) Nuclear magnetic resonance (NMR) spectra were recorded at ambient temperature on an Avance 500 (500 MHz) spectrometer (500. 1 Table S1. An external calibration standard (TPPSe in CD 2 Cl 2 ) was employed using a coaxial insert NMR tube. This provided a standard to compare relative intensities and relate concentrations without the possibility of the standard reacting with the species being studied. The external tube holds 490 µL of sample while the internal tube holding the reference holds 100 µL. The low concentrations resulted in low signal intensities coupled with the weak signal from 77 Se NMR, causing the 77 Se NMR experiments to be run for over 8 hours (overnight). The integrated peak intensities were compared to determine what percentage of TnBPSe was converted to H 2 Se by calculating how much H 2 Se was formed. The concentration of H 2 Se was back calculated from the known concentration of the TPPSe external standard.   Table  S1.

Supplementary
Supplementary Figure 6. 31 P (Left) and 77 Se NMR (Right) of TPPSe in CD 2 Cl 2 , used an external calibration standard in the quantitative 77 Se NMR Experiments. Chemical shifts are reported in Table S1.

Heat up method versus hot injection
Before optimizing the experiment to best mirror QD reaction conditions, a heat up method was used to observe the decomposition of TnBPSe to TnBPO. The TnBPSe, octanoic acid, and tetradecane were combined in the glove box and sealed. They were then heated to 250˚C for varying times. They were then brought back into the glovebox and transferred to NMR tubes for characterization by 1 H and 31 P NMR. This method resulted in significantly slower reaction than the hot injection approach more common to QD syntheses. Although slower, the experiment still provided relevant information regarding the decomposition of TnBPSe. Notably, in 1 H NMR, the acidic proton of the octanoic acid was seen to shift upfield and decrease in intensity over long reaction times, which supports our mechanism where these protons react with the phosphine selenide to form H 2 Se. Additionally, the sealed reaction vials become pressurized, indicating the formation of a gas. Almost complete conversion (96%) was observed after 19 hours.  Table S1.

Dependence on Carboxylic Acid
The dependence of the TnBPSe decomposition on the relative concentration of octanoic acid in the reaction flask was studied by varying the excess carboxylic acid in sealed flasks as mentioned above and measuring the 31 P NMR after 15 minutes of reaction at 250˚C after hot injection of the TnBPSe. The results show that increased octanoic acid results in increased conversion of TnBPSe to TnBPO. Variations in the conversion percentage were seen between trials but the general trend persists.  Figure 9. The conversion of TnBPSe to TnBPO in the presence of varied amounts of excess octanoic acid after 15 minutes of reaction at 250˚C. Excess octanoic acid is any amount greater than the two equivalents necessary to make Cd octanoate.

Reaction Mechanism
The evidence provided here and in the main text lead to our proposal of the reaction mechanism below. The mechanism accounts for all known products of this reaction reported here and in the literature. The high temperatures and complicated reaction matrix of common CdSe QD syntheses were considered while devising this mechanism.

Thermal decomposition of cadmium octanoate
Carboxylic acid and anhydride exchange When different carboxylates are used in the reaction -i.e. excess oleic acid and Cd acetate, we observe a mixed anhydride species through GCMS (Supplementary Figure 16. Mixed anhydrides that are formed through the reaction of acetic anhydride and oleic acid have a higher reactivity compared to symmetric anhydrides, 5 and are in a fast equilibrium with symmetric anhydrides and free carboxylic acid (reaching equilibrium in less than 10 minutes at 100˚C). 6 This observation provides evidence that these mixed anhydrides and free carboxylic acids are available to react, perpetuating our proposed mechanism. Water has been shown to increase the speed of QD formation and here it hydrolyses anhydrides into protonated and deprotonated carboxylic acids. This could contribute to the role of water in QD reactions, as it could dictate the relative concentrations of anhydride and carboxylic acid in solution. This will be a target of future studies.

C Labelled GCMS Experiment
An experiment was devised to test the hypothesis that Cd Carboxylate was also decomposing to the highly reactive alkyl cadmium. This decomposition would occur through decarboxylation of Cd carboxylate to form alkyl cadmium and carbon dioxide. Alkyl cadmium is extremely reactive, so to avoid attempting to isolate it we decided to look for the formation of carbon dioxide through a labelling study. Labelled 1-13 C-Octanoic acid was purchased and used to make labelled Cd octanoate, Supplementary Figure 17, line 1, which was subsequently reacted at 250˚C for 15 minutes and the headspace of the reaction was analysed by GCMS.

Alternative decomposition pathways
Another possible cadmium carboxylate decomposition at high temperatures (200-280°C) has been discussed by in the literature and also produces CO 2 as a decomposition product. 7 The proposed mechanism seen in Supplementary Figure 19 is based on the decomposition reaction of cadmium acetate in acetone where cadmium carbonate is formed and, upon further heating, evaporation of the solvent, and the release of 13 CO 2, cadmium oxide is recovered. 7 This mechanism has not been studied under temperatures typical of QD synthesis conditions and has only been studied for short chain carboxylates in low boiling point solvents. The original work describes that evidence for this decomposition reaction only is observed when the decomposition rate is very slow, 8 which would not be the case under QD reaction conditions. Further, we found no evidence of cadmium carbonate or cadmium oxide forming even after 16 hours of reaction. Cadmium carbonate (white solid, insoluble in tetradecane at high temperatures) and cadmium oxide (brown precipitate) should be easily observed in the decomposition reactions in this work if they were present. Though both cadmium carbonate and cadmium oxide have been used as QD precursors, both require excess carboxylic acid or phosphonic acid to produce QDs in solution or else both are unreactive, 9 indicating that the active cadmium precursor is neither of those species. The headspace of the reaction was analysed using GCMS, and the purged space in the arm of the flask was analysed to ascertain a background concentration of gasses. The Teflon pin of the flask was then opened allowing the headspace of the reaction to mix with the atmosphere in the arm of the flask. A second sample of gas was taken through the septum and the relative concentrations of gasses before and after mixing was determined. Proper technique is critical to ensure minimal contamination from atmospheric CO 2 . 13 C-labeled cadmium carboxylate was synthesized from 1-13 C-octanoic acid and cadmium oxide in tetradecane by heating to 210°C for 30 minutes in an schlenk flask (Supplementary Figure 19. The resulting clear colourless liquid was degassed and the arm of the flask was sealed with a septum and purged with nitrogen. The sealed flask was then heated to 280°C for 15 minutes. A calibration curve was created by varying the amounts of 13 CO 2 added to a schlenk flask filled with an equivalent volume of tetradecane as in the experimental unknowns with an argon internal standard that was added equally across all samples. The ratio of counts versus the internal standard was calculated by dividing the counts of 13 CO 2 (m/z=45) by those of Ar (m/z = 40). The average for each concentration was taken over five samples. Figure 20. GCMS Calibration curve of 13 CO 2 normalized to an argon external standard. Five replicates of each sample were taken and the linear fit was used to calculate the amount of 13 CO 2 detected. QD conditions include 12 times excess 1-13 C-Octanoic acid, standard to a QD synthesis where excess carboxylic acid is used as a ligand on the QD surface as well as to solubilize reactants. The QD synthesis included the addition of TnBPSe at the start of the reaction and QDs were formed through the heat up method in the same reaction flask (Supplementary Figure 18). The absorbance spectrum of the resulting QDs is depicted in Supplementary Figure 21. The low 13 CO 2 conversion efficiencies could be augmented by considering the percentage of labelling of the 1-13 C-octanoic acid starting material. A 13 C NMR of the 1-13 C-octanoic acid as purchased showed a 62% 13 C labelling of the carboxylic acid carbon. This would nearly double the percent conversion and could be an experimental source of error for these calculations.