Displacement of cadmium carboxylate from cadmium selenide nanocrystals (d = 3.3 - 3.7 nm) was monitored in situ using ¹H NMR spectroscopy. The vinylic hydrogens of surface-bound oleyl chains display a broad resonance in the range 5.3 – 5.6 ppm, well separated from other ligand-derived signals and therefore useful for monitoring displacement reactions (Figure 1). The NMR line-width provides a convenient method to distinguish bound ligands, which tumble slowly and have broad signals, from ligands moving freely in solution, which display sharp signals.^{17,32,33,42,43,82} In this way, NMR spectroscopy can be used to study nanocrystals in the presence of small molecule impurities, especially unconverted M(O₂CR)₂ remaining from the synthesis, as well as to determine the surface coverage of carboxylate ligands (see experimental). Using this approach we obtained isolated nanocrystals for this study with 3.3 - 3.7 carboxylates/nm², coverages similar to those reported previously.^17,32,45,83

Displacement of cadmium carboxylate from these nanocrystals was observed in the presence of several L-type Lewis bases including alcohols, amines, and phosphines. For example, adding N,N,N',N'-tetramethylethylene-1,2-diamine (TMEDA) to the nanocrystals displaces a carboxyl fragment with a sharp vinyl resonance that shifts up-field and increases in intensity if additional TMEDA is added (Figure 1). While in principle this result can be explained by simple displacement of the carboxylate ligands, this would require that the anionic charge of the carboxylate is seperated from the positively-charged surface-bound cation. Instead, we hypothesized that the “free” carboxyl fragment is derived from a TMEDA-bound cadmium carboxylate complex displaced from the surface of the nanocrystal. A similar reaction pathway has been previously suggested to explain the displacement of surfactant ligands with hydrazine, diamines,⁸⁴ phosphines, and primary amines⁸⁵ but never explicitely demonstrated.

To confirm our hypothesis the nanocrystals were separated by precipitation with methyl acetate and the supernatant analyzed (see experimental section). Upon drying under vacuum, a nearly colorless oil was obtained that showed sharp ¹H NMR signals from aliphatic and vinylic hydrogens characteristic of carboxyl fragments as well as cadmium-bound TMEDA ligands which are shifted slightly up-field from the frequencies of free TMEDA (~ 2:1 carboxyl:TMEDA). In addition, a broad feature of low intensity is visible at δ = 9.3 ppm that we assign to the acidic hydrogen of a carboxylic acid present in lower concentration (8 ± 3%) (Figure S1). A strong asymmetric stretching band from the cadmium-bound carboxylate (ν_assym(O₂CR) = 1560 cm^–1) is visible in the FT-IR spectrum that matches an independently prepared sample of (κ²-TMEDA)Cd(O₂CR)₂ and is readily distinguished from the signals of oleic acid mixed with TMEDA (ν (HO₂CR) = 1720 cm⁻¹) (Figure S2).

To provide further support for Cd(O₂CR)₂ removal, we studied Cd(O₂CR)₂ displacement in the presence of tri-n-butylphosphine (Bu₃P) with the help of ³¹P NMR spectroscopy. Much like TMEDA, adding Bu₃P to the nanocrystals causes new carboxyl resonances to appear in the ¹H NMR spectrum including one in the vinyl region that shifts up-field, sharpens, and increases in intensity when additional Bu₃P is added (Figure S3). (Bu₃P)Cd(O₂CR)₂ was isolated from the supernatant and identified with both ¹H and ³¹P NMR spectroscopies, including a diagnostic ³¹P NMR signal that matched an independently prepared sample (Figure 2, Figure S4). X-ray photoelectron spectroscopy of the isolated byproduct shows signals expected for (Bu₃P)Cd(O₂CR)₂ with no detectable signals in the range expected for Se (Figure 2), evidence that etching of the CdSe core is not the source of cadmium.^20,36,86-88

To investigate the generality of this reaction in other materials, a solution of zincblende cadmium sulfide nanocrystals with tetradecanoate ligands was mixed with TMEDA. Rapid displacement of (κ²-TMEDA)Cd(O₂CR)₂ (R = tridecyl) was observed by ¹H NMR spectroscopy (Figure S5). TMEDA also displaces a (κ²-TMEDA)Pb(O₂CR)₂ (R = oleyl) complex from PbS and PbSe nanocrystals without etching the nanocrystal cores, a result that was confirmed by comparing ¹H NMR, FT-IR, and electron-dispersive X-ray spectra of the soluble byproduct with those of an independently synthesized sample (Figures S6). In situ ¹H NMR spectra taken upon adding TMEDA to PbS and PbSe nanocrystals showed a similar appearance of free oleyl chains as observed for CdSe. However, TMEDA proved to be less effective for displacing M(O₂CR)₂ from PbS and PbSe nanocrystals than from CdSe nanocrystals, a property that likely derives from the partially occupied valence shell of the Pb²⁺ ion (Figures S7 and S8).

Having established that the removal of M(O₂CR)₂ using TMEDA is general for several classes of nanocrystals, we returned to cadmium selenide to measure the relative Z-type displacement potencies of L-type ligands in situ using ¹H NMR spectroscopy (Scheme 4). Among those studied, primary amines and TMEDA displace the greatest proportion of Cd(O₂CR)₂ (95 ± 10%, 2.0 M), while pyridine and Bu₃P displace a moderate amount (35 - 40 ± 5%, 2.0 M). Common anti-solvents used in nanocrystal purification, such as acetone, methyl acetate, and acetonitrile, do not displace significant amounts of L–Cd(O₂CR)₂, while high concentrations (2.0 M) of primary alcohols displace ~10% of the starting Cd(O₂CR)₂. This difference is not the result of a change in solvent dielectric; displacement with pyridine occurs to the same extent in d₆-benzene and d₂-methylene chloride (Figure S9). Related studies on the reaction of cadmium selenide nanocrystals with methanol propose that carboxylate ligands are displaced by proton-mediated X-type exchange without releasing Cd²⁺ (Scheme 2)^32,33,47 and leaving a surface bound methoxide.¹⁷ However, under our conditions 80 mg of nearly colorless (CH₃OH)_n•Cd(O₂CR)₂ was isolated from 1.86 g of nanocrystals (10% yield of total Cd(O₂CR)₂) upon precipitation with pure methanol, that was characterized unambiguously with ¹H NMR and FT-IR spectroscopies (Figure S10).

The relative potency of the Lewis bases shown in Scheme 4 illustrates several features of the Z-type displacement reaction. First, electronic effects are important as Bu₃P is more effective than tri-n-butylamine (Bu₃N), and primary amines are more effective than primary alcohols, presumably because they are more polarizable and electron rich, respectively, making them a better match for the soft, Lewis acidic cadmium ion. Second, chelation plays an important role, making TMEDA among the most effective displacement reagents while N,N-dimethyl-n-butylamine (Me₂NBu) is among the weakest. Chelation also explains why TMEDA is significantly more potent than N,N,N',N'-tetramethylbutylene-1,4-diamine, a result that is expected from the relative stability of the five membered ring formed in (κ²-TMEDA)Cd(O₂CR)₂ compared with the analogous seven membered ring in (κ²-N,N,N',N'-tetramethylbutylene-1,4-diamine)Cd(O₂CR)₂.⁸⁹ Finally, the steric profile of the incoming ligand has a significant impact on its displacement potency. Steric effects explain the large differences between Bu₃N, Me₂NBu, and n-butylamine, given their similar pK_bs.⁹⁰ Likewise, the relatively small steric profile of pyridine enables Cd(O₂CR)₂ removal, the potency of which is otherwise difficult to explain given its weaker basicity and hard/soft match with the cadmium center.

The sensitivity to sterics is intriguing and implies that displacement of Cd(O₂CR)₂ may be promoted by cooperative binding of the L-type ligand to nanocrystal surface where steric effects are known to strongly influence binding affinity.^18,45 In the case of primary amines, displacement is accompanied by binding of the L-type ligand to the surface of the nanocrystal, dramatically increasing the PLQY (see discussion below) and maintaining the nanocrystals solubility even after > 90% of the Cd(O₂CR)₂ has been displaced. On the other hand, displacing a similar amount of Cd(O₂CR)₂ using TMEDA causes precipitation. Presumably any bound TMEDA does not provide enough steric repulsion to maintain the solubility. On the basis of these observations, we conclude that the displacement reaction can be a cooperative process in cases where the added ligand binds both the nanocrystal surface and the displaced Cd(O₂CR)₂. In particular, ligands with a small steric profile, like pyridine and n-alkylamines, may drive the displacement equilibrium by binding the nanocrystal.

Having shown that a variety of ligands displace L–Cd(O₂CR)₂ from CdSe nanocrystals, we investigated the kinetics of the displacement reaction. In situ ¹H NMR measurements indicate that displacement is nearly complete (> 90%) within a few minutes of mixing TMEDA with nanocrystals in benzene. The initially rapid displacement is followed by a slower reaction that liberates a much smaller portion of (κ²-TMEDA)Cd(O₂CR)₂, however, the total change observed after the rapid initial reaction is small and near the estimated uncertainty of the NMR measurement (Figure 3). Surprisingly, the reaction rapidly reaches equilibrium, even under conditions where the majority of surface cadmium atoms are displaced. The fast kinetics indicate that nanocrystal stoichiometry is very sensitive to the presence of Lewis bases and is easily altered. Furthermore, these observations indicate that the relative extent of displacement shown in Scheme 4 reflects an equilibrium condition rather than differences in reaction rate. In support of this conclusion, we analyzed a mixture of pyridine and nanocrystals before and after heating to 100 °C for 6 hours and found no measurable change to the final ratio of bound and free Cd(O₂CR)₂ (Figure S11).

Knowing that TMEDA and primary amines are the most potent Cd(O₂CR)₂ displacement reagents and that the displacement reaction establishes equilibrium quickly, we attempted to isolate nanocrystals with a minimal Cd(O₂CR)₂ coverage. We hypothesized that the chelation ability of TMEDA coupled with the cooperative binding of primary amines might allow access to even lower coverages than either L-type ligand alone. Soluble nanocrystals with n-octylamine ligands (d = 3.5 nm, ~170 n-octylamine ligands, 4.5 ligands nm⁻², Table 1, Figure S12) could be obtained by displacing Cd(O₂CR)₂ with a 50:50 mixture of TMEDA and n-octylamine (see experimental section). However, ¹H NMR spectroscopy of nanocrystals isolated from this solution shows that 6% of the bound carboxyl ligands remain, a decrease from 160 per nanocrystal (4.2 nm⁻²) to 10 per nanocrystal (0.3 nm⁻²). No signals from TMEDA are visible in the ¹H NMR spectrum, presumably because primary amines preferentially bind the nanocrystal surface. The stoichiometry of the isolated nanocrystals was analyzed with Rutherford backscattering spectrometry, verifying that the Cd:Se ratio decreases from 1.10 ± 0.06 to 1.01 ± 0.05 after exposure to TMEDA (Figure S13). Repeated treatments with amine and TMEDA mixtures did not further reduce the carboxylate loading, and nanocrystals always retained a carboxylate coverage of at least 0.3 nm⁻². Similar difficulty was previously reported when removing native ligands using pyridine and other neutral donors, though the origins of this coverage dependent binding affinity remain are uncertain.^19,31,87,91 While it is natural to suppose that a mixture of surface sites with different binding affinities results in some ligands being bound more tightly, this behavior may also result from allosteric electronic effects associated with forming the complete polar facets of a stoichiometric nanocrystal.

The concentration dependence of the Cd(O₂CR)₂ coverage and the rapid displacement kinetics suggest that rebinding may be equally facile. To investigate the reversibility of displacement, a solution of unaggregated nanocrystals with a low surface Cd(O₂CR)₂ coverage (0.6 carboxylates/nm²) was stirred with anhydrous cadmium oleate at room temperature in toluene (see experimental section). After separating free Cd(O₂CR)₂ by repeated precipitation of the nanocrystals with methyl acetate, the carboxylate coverage had increased to 2.1 nm⁻², roughly 60% of the coverage prior to displacement with TMEDA (Figure S14). Similar results were obtained with CdS nanocrystals (see below). The partial rebinding at room temperature may result from slow organization of carboxylate ligands at high coverages or a slow surface reconstruction that must be reversed prior to rebinding. However, complete recovery of the original surface ligand density (3.1 nm⁻²) could be achieved by heating the nanocrystals to 240° C under nitrogen in the presence of cadmium oleate and oleic acid. Under these conditions the nanocrystals grow slightly (d = 3.6 to 3.8 nm) as measured by the small red-shift in their UV-visible absorption spectrum (7 nm), which we attribute to Ostwald ripening given that no selenium precursor was added (Figure 5). While the temperature dependence of binding deserves further study, the reversibility implies that surface-bound cadmium ions are in equilibrium with free cadmium complexes in solution, and thus nanocrystal stoichiometry is concentration dependent.

While nanocrystals with ligand coverages of ~3 carboxylates/nm² were chosen for this study on the basis of a ¹H NMR spectrum that lacked the sharp signals from free carboxyl fragments, a number of observations suggest that higher coverages are possible. First, the density of atoms on common facets of CdSe nanocrystals is greater than 3 nm^–2 (CdSe (111) = 6.2 atoms/nm² and (100) = 5.4 atoms/nm⁻²). Assuming that a maximum of one excess cadmium can bind each surface selenium atom of a cubeoctahedral nanocrystal with equal area (111) and (100) facets, cadmium coverages of 2.9 nm⁻² and thus carboxylate coverages of 5.8 nm⁻² are in principle possible. However, steric interactions between carboxylate ligands will prevent coverages from exceeding the density of crystalline alkane chains (4.9 chains/nm²) on any given facet.⁹² As a result, chain packing effects will prevent the coverage from reaching the upper limits defined by the underlying atomic surface density, and stable structures with a purely n-alkylcarboxylate ligand shell may never accommodate one cadmium for every surface selenium atom. Second, PLQY is known to drop during isolation of nanocrystals from crude synthesis mixtures, behavior that may derive from a change in the Cd²⁺ coverage.^52,63,93,94 Typically, an excess of unconverted M(O₂CR)₂ remains after synthesis, and its separation will reduce the solution concentration of M(O₂CR)₂ and can thereby change the coverage of weakly-bound metal cations.^50,52,95 A similar decrease in PLQY was observed in the present study where it is highest (~20%) prior to precipitation with methyl acetate and drops to ~10% as coverages of 3 carboxylates/nm² are reached. Both observations suggest that the ligand coverage obtained after isolation may be lower than that in the crude reaction mixtures produced by the nanocrystal synthesis.

Changes to stoichiometry caused by isolation and exposure to Lewis bases helps explain the sensitivity of nanocrystal PLQY. Ligation is known to influence surface derived mid-gap electronic states, and thus it is not surprising that changes to the coverage of Cd(O₂CR)₂ influence the PLQY.^{10,11,13,18,23,28,40,71,75,96,97} Previous studies have found that cadmium and zinc ions significantly improve PLQY and reduce photocharging when bound to nanocrystal surfaces. ^{21,27,41,53,68-76,98,99} Presumably these metal ions act as Z-type ligands, passivating mid gap states by binding surface Se sites and thereby preventing hole trapping.^10,25 With the exception of displacement reactions with primary n-alkylamines, removing Cd(O₂CR)₂ in this study decreases PLQY with the greatest change being caused by dissolution of nanocrystals in neat TMEDA (PLQY < 0.1%). Plotting the PLQY versus cadmium coverage for isolated samples as well as samples measured in situ shows a strong super-linear correlation (Figure 4). PLQY is very sensitive to coverage above 2.5 carboxylates/nm², while samples with lower coverages are weakly photoluminescent. Samples exposed to primary amines, however, display significantly increased PLQY despite removing the greatest proportion of Cd(O₂CR)₂. The increase is related to the high coverage of amine ligands rather than an effect derived from the cadmium coverage;^18,36,38 the PLQY of isolated nanocrystals with low Cd(O₂CR)₂ coverage is greatest at higher amine concentrations. Thus PLQY is complex and depends on the presence of L-type donor ligands as well as Z-type accepter ligands in a manner that is very sensitive to coverages near saturation.

In addition to changing the PLQY, decreasing the Cd(O₂CR)₂ coverage to 0.6 carboxylates/nm² causes a slight red-shift (1 nm) in the wavelength of the lowest energy absorption, an effect that might be explained by a change in the local dielectric, or a change in the confining potential caused by the surface MX₂ layer (Figure 5).²⁸ While this red-shift could be attributed to nanocrystal aggregation, dynamic light scattering and transmission electron microscopy measurements show that the nanocrystals do not significantly aggregate upon reducing the carboxylate coverage to 0.6 carboxylates/nm² (Figure S15 and S16). As a result, the insensitivity of the lowest energy absorption is surprising given that ~85% of the surface Cd(O₂CR)₂ in this sample has been displaced, a ~10% reduction in the total number of cadmium ions in the nanocrystal. When calculating the volume change using the ionic radii of Cd²⁺ (109 pm) and Se²⁸ (184 pm) and the void space of the unit cell,¹⁰⁰ this change in formula corresponds to a 3% decrease in volume and can be expected to produce a 4 nm blue-shift at this size.¹⁰¹ Similar effects were observed in spectra of CdSe nanocrystals across a range of sizes from d = 3.3-3.7 nm where slightly larger red shifts are observed for smaller sizes (Figure S17). While the lowest energy absorption is not altered with a decrease in the cadmium ion coverage, higher energy transitions are significantly influenced. Removal of Cd(O₂CR)₂ from cadmium selenide or cadmium sulfide nanocrystals quenches the band-edge PLQY and decreases the apparent intensity of the 1S_e-2S_3/2h absorption (Figure 5). The narrow absorption features of cadmium sulfide nanocrystals clearly show the apparent intensity of the 1S_e-2S_3/2h absorption is reduced by broadening rather than a change in extinction (Figure S18).

The insensitivity of the first absorption energy to changes in the coverage of surface Cd(O₂CR)₂ is surprising and suggests that its 5s functions do not contribute significantly to the conduction nor valence band edges, and instead contribute to states within the bands. A related model was proposed to explain the influence of a Cd²⁺ layer on the energy difference between the first two electronic absorptions.²⁸ This study also focused on the reaction of nanocrystals with alkylamines, where changes in the intensity and energy of the second electronic absorption were observed, effects that may also stem from removal of Cd(O₂CR)₂ in their case. While others have noted a relationship between surface structure and the line-width of these higher energy transitions, the origin of these changes remains unclear.^31,84

Changes to the nanocrystal absorbance and PLQY are reversed upon rebinding Cd(O₂CR)₂ at room temperature indicating that they are directly related to the coverage of Z-type ligands (Figure 5). While the absorbance and luminescence changes are only partially reversed upon stirring the nanocrystals with Cd(O₂CR)₂ at room temperature, they are completely reversed upon recovering the original surface ligand density at higher temperature (Figures 4 and 5C). Interestingly, similar changes are observed if nanocrystals with low carboxylate coverages are stirred with CdCl₂ rather than Cd(O₂CR)₂, corroborating a previous investigation of coordinatively unsaturated CdSe nanocrystals bound by tri-n-butylphosphine and cadmium chloride (Figure S19).⁴⁵ Together these observations support a general relationship between the coverage of cadmium ions and trap state passivation, as well as the optical spectrum of the nanocrystal.

Several recent studies have investigated the influence of excess metal ions on mid- gap electronic states and doping. While we have shown that Cd²⁺ bound by X-type carboxylate ligands passivate surfaces, the oxidation state of the surface-bound metal ion will influence whether metal enrichment passivates charge trap states or contributes new mid-gap states. For example, the electron concentration in lead chalcogenide nanocrystal thin films can be increased by vapor depositing elemental lead (Pb⁰).⁵⁷ Similarly, computational studies show that adsorption of Pb⁰ atoms to nanocrystal surfaces adds filled electronic states within the band gap resulting in metallic-like behavior.⁵⁴ In the same study, these mid-gap states are removed when the surface lead atoms are converted to Pb²⁺ and balanced by X-type ligands. Thus the metal-rich stoichiometry of nanocrystals does not appear to create trap states when all ions are in their 2+ oxidation state.

Instead, cadmium and lead ions can passivate the nanocrystal by acting as electron-accepting Z-type ligands that lower the chalcogen-derived filled states of surface atoms. However, ideal passivation schemes should manage both filled and empty mid-gap states. Both chalcogen-derived filled states as well as metal-derived empty states can lead to mid-gap levels, and passivation of these empty levels by a high coverage of amine donors can also increase PLQY. The benefits of this approach have been demonstrated in PbS nanocrystal photovoltaic cells with record-breaking efficiencies.^4,44,53,102 Identifying surface passivating layers that bind both metal and chalcogen sites is therefore an important avenue of research.

A detailed understanding of metal ion binding affinities is needed to better understand and control these materials. While the metal-enrichment of cadmium and lead chalcogenide nanocrystals has been shown to increase with decreasing nanocrystal diameter according to the ratio of its surface area and volume,^34,49,50 we have shown that nanocrystals have a wide range of surface metal coverages. Thus the stoichiometry obtained is sensitive to the reagents and concentrations used to separate nanocrystals from unreacted M(O₂CR)₂ precursors. Nanocrystal stoichiometries are therefore not fixed for a given size and depend on the isolation method. In the present study, a particular stoichiometry was obtained using NMR spectroscopy as a guide to monitor bound and free M(O₂CR)₂ during precipitation. However, a wider range of stoichiometries can be obtained by altering the isolation method including the solvents and concentrations used and the number of precipitation cycles as well as the binding affinity of the surface metal-surfactant layer.^16,17,52 In this sense, nanocrystals are not molecular; they do not have a single formula and nanocrystal purity is an indefinite concept. The method of synthesis can also influence the final stoichiometry.⁵⁰ For example, terminating a synthesis to obtain a desired size will influence the ratio of unconverted metal carboxylate and nanocrystals in the crude product.^63,103 As a consequence, the stoichiometry of the isolated product may not reflect the true size-dependence of the MX₂ binding, but instead the extent of precursor conversion and the effects of the isolation steps. Given the importance of stoichiometry to passivation, it is crucial that new synthetic methods precisely and systematically control stoichiometry in addition to optimizing a desired nanocrystal size and yield.

The effects of added Lewis bases on stoichiometry are also relevant to any post-synthesis modifications like exchange of native surfactant ligands. Many previous investigations of ligand exchange utilize photoluminescence intensity to monitor the extent of reaction unaware that stoichiometry may also change.^{6,18,24,35,40,96,104} Some of these studies report conflicting results including the observation that primary amines can both increase and decrease PLQY. Similarly, PLQY has been used to measure the concentration dependence of ligand binding. From these data relative ligand binding constants were estimated, though only by assuming that coverage and PLQY are proportional. However, other studies point out that PLQY and ligation have a nonlinear relationship, where the binding of even a single octanethiol molecule can quench the luminescence by at least 50%.²³ Our direct correlation of PLQY and Cd(O₂CR)₂ surface coverage demonstrates that there is a threshold of 2 carboxylates/nm² below which the PLQY is ≤ 1% and above which PLQY rapidly increases. Hence, PLQY and ligation are not simply related, both because of their nonlinear interdependence but also because displacement of Cd(O₂CR)₂ can occur concurrently with amine binding.

Stoichiometry is also important to the optoelectronic properties of nanocrystal thin films and methods must be designed that exchange X-type ligands while maintaining stoichiometry during film deposition. Traditional approaches to fabricate conductive nanocrystal films displace native surfactant ligands with pyridine,^4,91,105-110 primary alkylamines,^{15,73,111-116} hydrazine^{73,84,106,108,116-126} and alkane diamines,^{3,4,73,84,108,113,114,118,127-133} and are therefore likely to significantly reduce the metal-enrichment. Chelating alkane diamines have been used to displace native ligands and link nanocrystals, resulting in films where the inter-nanocrystal spacing depends on the diamine chain length.^{73,84,108,113,114,128-130,132,133} Our results show the M(O₂CR)₂ displacement efficiency depends on the diamine chain length because chelation stabilizes the displaced metal complex. The inter-nanocrystal distance measured in those studies may result, at least in part, from the relative displacement potency of the diamine. Indeed some of these studies report changes to the absorption and photoluminescence spectra, such as quenching of the PLQY and changes to the 1S_e-2S_3/2h absorption, similar to those described above.^31,84,114

Recent approaches to thin film fabrication that utilize thiocyanate, sulfide, halide, and metal-chalcogenide salts can, in principle, exchange X-type ligands without altering stoichiometry.^{4,41,44,53,75,102,134-136} However, the affinity of the new surface MX₂ complexes for the nanocrystal likely depends on the nature of both the cation and the anion, and therefore the resulting MX₂ complex may be more weakly bound than the starting metal-surfactant layer. In the case of chloride exchange, cadmium chloride appears to be a weaker Z-type ligand for the nanocrystal core than cadmium carboxylate,⁴⁵ perhaps because the carboxylate binds in a bidentate fashion.¹³⁷ Identifying ligand exchange methods that optimize inter-nanocrystal electronic coupling while maintaining a high coverage of MX₂ should produce conductive nanocrystal films with minimal trap state densities.⁴⁴

Cadmium nitrate tetra-aquo (99%), lead (II) oxide (reagent grade) sodium hydroxide, myristic acid (99%), selenium dioxide (99.8%), anhydrous oleic acid (99%), 1-tetradecanol (95%), eladic acid (99%), methanol (99.8%), hexamethyldisilathiane (synthesis grade), and 1-octadecene (90%) were purchased from Sigma Aldrich and used as received. Ferrocene (98%) was purchased from Sigma Aldrich and purified by sublimation before use. d₆-Benzene (99.6%), anhydrous acetonitrile (99.5%), and anhydrous methyl acetate (99.5%) were purchased from Sigma Aldrich, shaken with activated alumina, filtered, and stored over 4 Å molecular sieves in an inert atmosphere glove box at least 24 hours prior to use. Pentane, tetrahydrofuran, and toluene were dried over alumina columns, shaken with activated alumina, filtered and stored over 4 Å molecular sieves in an inert atmosphere glove box at least 24 hours prior to use. Diphenylphosphine (99%), N,N,'N,'N-tetramethylethylene-1,2-diamine (TMEDA) (98%), tri-n-octylphosphine (97%) and tri-n-butylphosphine (99%) were purchased from Strem and used without further purification. CdMe₂ was purchased from Strem and vacuum distilled prior to use. CAUTION: Dimethylcadmium is extremely toxic and because of its volatility and air-sensitivity should only be handled by a highly trained and skilled scientist. N,N-dimethylbutylamine (98%), N,N,N',N'-tetramethyl-1,4-butane diamine (97%), 1-nonanyl nitrile (98%), tri-n-butylamine (99%), n-octylamine (99%), and pyridine (99.5%) were purchased from Sigma Aldrich and dried over CaH₂, distilled, and stored in a nitrogen glove box. Cadmium tetradecanoate was synthesized from cadmium nitrate and tetradecanoic acid on 25 mmol scale following a procedure reported previously.⁹³ Cadmium oleate and cadmium eliadate were synthesized by adding the appropriate carboxylic acid in a 1:2 ratio to dimethyl cadmium dissolved in pentane. The resulting white powder was filtered and dried under vacuum.

All manipulations were performed under air-free conditions unless otherwise indicated using standard Schlenk techniques or within a nitrogen atmosphere glove box. NMR spectra were recorded on Bruker Avance III 500 MHz instruments. ¹H NMR spectra were acquired with sufficient relaxation delay to allow complete relaxation between pulses (30 seconds). UV-Visible data was obtained using a Perkin Elmer Lambda 650 or 950 spectrophotometers equipped with deuterium and tungsten halogen lamps. Photoluminescence spectra and quantum yield were measured using a FluoroMax-4 equipped with an Integrating Sphere from Horiba Scientific. FT-IR spectra were obtained on a Thermo Scientific Nicolet 6700 spectrometer equipped with a liquid N₂ cooled MCT-A detector. RBS measurements were conducted at the University of Western Ontario in the Interface Science Western facility.

Carboxylate terminated zincblende CdSe nanocrystals were synthesized using a previously reported procedure.⁹³ Typical reactions were run using 10 mmol of cadmium tetradecanoate and a Cd:Se ratio of 1. Isolated nanocrystals were dissolved in d₆-benzene and the vinyl region of the ¹H NMR spectrum was used to detect unbound ligands. Samples without signals from free oleyl chains were used to make d₆-benzene stock solutions for ligand exchange studies. Typical stock solutions ([^–O₂CR] = ~ 200 mM; [nanocrystal] = ~ 2 mM) were diluted for NMR spectroscopy ([^–O₂CR] = ~ 20 mM) and for UV-visible absorption spectroscopy and PLQY measurements ([^–O₂CR] = ~ 0.2 μM).

Tetradecanoate terminated zincblende CdS nanocrystals were synthesized and isolated at 2× the concentration and 5× the volume reported previously.⁶³ It is worth noting that the isolation procedure involves precipitating the nanocrystals from a TMEDA/toluene solution, rather than methyl acetate. Isolated nanocrystals were dissolved in d₆-benzene and the methyl region of the ¹H NMR spectrum was used to confirm that no free ligands were present. Typical stock solutions used for NMR spectroscopy ([^–O₂CR] = ~ 10 mM; [nanocrystal] = ~ 0.3 mM) were diluted for UV-visible absorption spectroscopy and PLQY measurements ([^–O₂CR] = ~ 0.3 μM).

PbS nanocrystals were synthesized using a previously reported procedure.¹³⁸ Typical reactions were run at 50 mL scale using of lead (II) oxide (0.12 molar), oleic acid (1.2 molar), and Pb:S ratio of 2:1. Isolated nanocrystals were dissolved in d₆-benzene and the vinyl region of the ¹H NMR spectrum was used to analyze the sample for free ligands. Samples without signals from free oleyl chains were used to make d₆-benzene stock solutions for ligand exchange studies. Typical stocks used for NMR spectroscopy ([^–O₂CR] = ~ 50 mM; [nanocrystal] = ~ 0.4 mM) were diluted for UV-Vis-NIR measurements ([nanocrystal] = ~ 0.4 μM).

PbSe nanocrystals were synthesized using a previously reported procedure.¹³⁹ Typical reactions were run at 50 ml scale using lead (II) oxide (0.1 molar), oleic acid (0.3 molar), and a Pb:Se:diphenylphosphine ratio of 1:2:0.2. Isolated nanocrystals were dissolved in d₆-benzene and the vinyl region of the ¹H NMR spectrum used to analyze the sample for free ligands. Samples without signals from free oleyl chains were used to make d₆-benzene stock solutions for ligand exchange studies. Typical stocks used for NMR spectroscopy ([^–O₂CR] = ~ 20 mM; [nanocrystal] = ~ 0.1 mM) were diluted for UV-Vis-NIR measurements ([nanocrystal] = ~ 0.1 μM).

The concentrations of nanocrystals and carboxylate or amine ligands in d₆-benzene stock solutions were determined using a combination of NMR and UV-Visible absorption spectroscopies. Ferrocene dissolved in d₆-benzene (10 μl, 0.05 M) was added to a known volume of the nanocrystal stock solution and used as an internal standard. The concentration of ligands was determined relative to the internal ferrocene standard by integrating the ligand methyl and ferrocene resonances and normalizing for the number of hydrogens respectively (3:10). Nanocrystals with n-octylamine ligands were precipitated from neat n-octylamine using methylacetate, isolated by centrifugation and twice reprecipitated from pentane solution with methylacetate to ensure that the remaining amine is all derived from surface bound ligands, rather than adventitious amine. ^45,140 The molar concentration of CdSe, CdS, PbS, and PbSe nanocrystals in these stock solutions was determined by diluting 10-30 μl to a known volume with toluene and measuring the absorbance at λ = 350 nm for CdSe and CdS nanocrystals and at λ = 400 nm for PbS and PbSe nanocrystals. At these wavelengths the extinction coefficient is independent of size.^{141,142,47,143}

The wavelength of the lowest absorption maximum was used to determine the average nanocrystal diameter^144,142 The diameter of PbS and PbSe nanocrystals were determined from the wavelength of the lowest energy absorption feature according to Hens.^47,143 From this diameter the number of ME (M = Cd, Pb; E = Se, S) units per nanocrystal were calculated by assuming a spherical shape and the molar volume of the bulk. The concentration of nanocrystals, the ratio of ligands per nanocrystal and the ligand surface density assuming a spherical shape were calculated from the number of ME units per nanocrystal, the molar concentration of ME, and ligands in the stock solution.

PLQY were measured using a Fluoromax-4 Fluorometer equipped with an integrating sphere. Samples were diluted to concentrations below 0.1 absorbance units at the 1Se-1Sh1/2 to minimize reabsorption. Data are not corrected for reabsorption and are therefore an underestimate. A blank sample of toluene was used to adjust the excitation and emission slits to avoid saturating the detector, and photoluminescence spectra, including the excitation wavelengths, were recorded for both the toluene blank and the nanocrystal solution. The number of photons absorbed by the sample was determined by measuring the difference between the blank and the sample at the excitation wavelength, and the photoluminescence spectrum of the nanocrystal solution was integrated from 500-650 nm to determine the photons emitted. PLQY was calculated from the ratio of photons emitted and photons absorbed. The validity of this approach was evaluated using freshly prepared solutions of Coumarin-153 in ethanol (PLQY = 53%).¹⁴⁵

CdSe nanocrystals (4 g) were dissolved in a d₆-benzene solution of Bu₃P or TMEDA (8 ml, 0.5 M) and stirred for 3 hours. The solution was diluted to a total volume of 30 ml, and methylacetate was added to a total volume to 90 ml causing precipitation of the nanocrystals. After separation of the nanocrystals by centrifugation (7000 RPM for 5 minutes), the clear, colorless supernatant was decanted and the nanocrystals dissolved in 15 ml of toluene and precipitated a second time with methyl acetate. The nanocrystals were separated by centrifugation and the clear, colorless supernatant decanted. This process was repeated a third time, after which the nanocrystals were discarded, and the combined supernatants dried under vacuum. A clear oil remained which had faint but perceptible pink hue, that was dissolved in toluene and diluted with methyl acetate to precipitate any remaining nanocrystals, which were separated by centrifugation. The volatiles were removed under vacuum and the residue again dissolved in toluene, diluted with methyl acetate and centrifuged to remove any remaining nanocrystals. The resulting clear solution was distilled to dryness under vacuum and dried for 3 hours, resulting in a clear, pale yellow oil that was analyzed with NMR, infrared absorption, and x-ray photoelectron spectroscopies. Removal with Bu₃P: ¹H NMR (C₆D₆, 500 MHz): δ = 0.93 (m, 15H, −CH₃), 1.1-1.6 (b, 62H, −CH₂, −CH₂P), 1.84 (m, 4H, β−CH₂), 2.08 (m, 8H, −C=CCH₂), 2.52 (t, 4H, α−CH₂), 5.48 (m, 4H, −CH=CH), 13.7 (b, −COOH). ³¹P{¹H} NMR (C₆D₆, 202.4 MHz): δ = −16.7 (b). FT-IR (Diamond ATR): ν = 1384 cm⁻¹ (s CO₂ assym), 1564 cm⁻¹ (s CO₂ sym), 1730 cm⁻¹ (w COOH), 2925 cm⁻¹ (s C-H). Removal with TMEDA: ¹H NMR (C₆D₆, 500 MHz): δ = 0.91 (t, 6H, −CH₃), 1.2-1.6 (b, 44H, −CH₂), 1.84 (m, 4H, β-CH₂), 1.94 (s, 4H, - NCH₂), 2.09 (m, 8H, −C=CCH₂), 2.24 (s, 12H, −N(CH₃)₂), 2.52 (t, 4H, α-CH₂), 5.49 (m, 4H, 8 CH=CH), 9.3 (b, -NH). FT-IR (Diamond ATR): ν = 1382 cm⁻¹ (s CO₂ assym), 1560 cm⁻¹ (s CO₂ sym), 1720 cm⁻¹ (w COOH), 2920 cm⁻¹ (s C-H).

CdSe nanocrystals (1.86 g) were dissolved in toluene (10 ml) and removed from the glove box in a centrifuge tube. In air, methanol was added to a total volume to 50 ml causing precipitation of the nanocrystals. After separation of the nanocrystals by centrifugation (7000 RPM for 5 minutes), the clear, colorless supernatant was decanted and saved. The nanocrystals were then dissolved in 10 ml of toluene, precipitated a second time with methanol, separated by centrifugation, and the clear, colorless supernatant decanted and saved. This process was repeated three additional times, after which the nanocrystals were discarded, and the combined supernatants distilled to dryness under vacuum. A clear faintly pink oil remains that was dissolved in toluene and diluted with methanol to precipitate any remaining nanocrystals. which were removed by centrifugation. This process was repeated once and the final clear supernatant was distilled to dryness under vacuum, resulting in a clear, pale yellow oil that was analyzed with NMR and infrared absorption spectroscopies. ¹H NMR (C₆D₆, 500 MHz): δ = 0.95 (m, 6H, −CH₃), 1.181.5 (b, 44H, −CH₂), 1.84 (m, 4H, β-CH₂), 2.1 (m, 8H, −C=CCH₂), 2.52 (t, 4H, α-CH₂), 5.49 (m, 4H, −CH=CH) FT-IR (Diamond ATR): ν = 1383 cm⁻¹ (s CO₂ assym), 1564 cm⁻¹ (s CO₂ sym), 2925 cm⁻¹ (s C-H) 3200 cm⁻¹ (b O-H).

PbS nanocrystals (0.5 g) were dissolved in d₆-benzene (2 mL) and TMEDA (3 mL) was added. After stirring for 10 minutes, acetonitrile (5 mL) was added and the nanocrystals separated by centrifugation. The clear supernatant was collected and dried under vacuum for 5 hours. d₆-benzene (0.6 ml) was added to dissolve the residue and the solution was analyzed with ¹H NMR spectroscopy. Infrared absorption spectroscopy was performed using diffuse reflectance geometry as a mixture with KBr. Energy dispersive X-ray spectroscopy was conducted on a film drop-cast on a highly ordered pyrolitic graphite substrate using a Cold Field Emission Hitachi 4700 Scanning Electron Microscope. ¹H NMR (C₆D₆, 500 MHz): δ = 0.91 (t, 6H, −CH₃), 1.281.6 (b, 44H, −CH₂), 1.84 (m, 4H, β-CH₂), 1.94 (s, 4H, −NCH₂), 2.09 (m, 8H, −C=CCH₂), 2.24 (s, 12H, −N(CH₃)₂), 2.52 (t, 4H, α-CH₂), 5.49 (m, 4H, −CH=CH), 9.3 (b, −NH). FT-IR (Diamond ATR): ν = 1384 cm⁻¹ (s CO₂ assym), 1560 cm⁻¹ (s CO₂ sym), 1720 cm⁻¹ (w COOH), 2920 cm⁻¹ (s C-H).

In a J-Young NMR tube, a 0.02 M solution of a desired ligand was prepared by adding the ligand (10 μl of a 1.1 M d₆-benzene stock solution) to a 590 μl stock solution of nanocrystals in d₆-benzene (0.02 M in - O₂CR). For higher concentrations (0.2 M and 2.0M), a second and third addition of ligand were added (1.1 × 10⁻⁴ mol and 1.1 × 10⁻³ mol respectively). ¹H NMR spectra were acquired within 10 minutes after ligand addition. The among of ligand displaced was determined by integrating the sharp resonance of free oleyl chains as a percentage of total vinyl region. Separate stock solutions of nanocrystals (2.4 × 10⁻⁵ M in ⁻O₂CR) and 0.02 M, 0.2 M, and 2.0 M in displacement ligand were prepared in quartz cuvettes under nitrogen for PLQY measurements.

Assuming a precision of a Cd-Se bond distance (~ 0.2 nm) in the diameter determination leads to the errors shown for the number of ME units in each nanocrystal. When calculating the number of ligands per nanocrystal, we estimate 10% error in the integration of the broad NMR signals and 5% error in the concentration of dilutions used to measure UV-Vis absorption. Volume additivity was assumed when calculating concentrations. RBS has a precision of 5%.

TMEDA (2 ml) was added to a stock solution of CdSe nanocrystals (2 ml, [^–O₂CR] = 0.2 M) and stirred for 1 hour. Nanocrystals were isolated by precipitation using methyl acetate and centrifugation. The red precipitate was dissolved in 5 ml of toluene and isolated by precipitation and centrifugation an additional 2 times. After the final centrifugation the nanocrystals were dissolved in pentane, dried under vacuum, and characterized by NMR and UV-visible absorption spectroscopies. Dynamic light scattering and transmission electron microscopy measurements show that nanocrystals with this ligand coverage are unaggreated (see supporting information).

For CdSe: In a three-neck round-bottom flask equipped with a thermocouple adapter, reflux condenser, and septum, 100 mg of cadmium elaidate, 25 mg of oleic acid, and 5 ml of a 0.08 M (in ⁻O₂CR) solution of CdSe nanocrystals in 18octadecene were degassed and then heated to 240 °C under argon for 1 hour. The solution was then transferred via cannula to a Teflon stoppered Schlenk flask and transported to a glove box. Nanocrystals were isolated using three cycles of dissolution in tetrahydrofuran and precipitation with methyl acetate. Tetrahydrofuran was necessary to separate the polymeric cadmium carboxylate. The isolated nanocrystals were dried under vacuum and dissolved in d₆8benzene for NMR and UV-vis analysis. For CdS: In a nitrogen-filled glove box, a 0.05 mM (in nanocrystals) solution of TMEDA-treated CdS nanocrystals (see CdS synthesis above) in C₆D₆ (2 mL) was combined with THF (5 mL) in a vial charged with a stir bar. Cadmium oleate powder (300 mg; 390 mmol) was added to this solution, which was left to stir for >3 hours. Nanocrystals were isolated using one cycle of dissolution in tetrahydrofuran and precipitation with methyl acetate, followed by two cycles of dissolution in pentane and precipitation with methyl acetate. The isolated nanocrystals were dried under vacuum and dissolved in d₆-benzene for NMR and UV-vis analysis. This sample was used for Figure S5.