Decoding structural transitions from CdSe nanoclusters to quantum dots through dynamic nuclear polarization NMR

Abstract

Achieving control over quantum dot size is essential for their performance, but remains challenging due to the limited atomic-level understanding of quantum dot growth. In this study, we employ signal-enhancing Dynamic Nuclear Polarization solid-state NMR to investigate the structural features of intermediate CdSe clusters and mature quantum dots to gain valuable insights. By integrating quantum mechanical calculations with experimental data, we leverage ¹¹³Cd chemical shift to elucidate local Cd environments at various positions, decoding the complex ligand distribution on cluster surfaces. Our findings reveal that ligand distribution is stabilizing through inter-ligand hydrogen bonds, while minimizing steric clashes during ligand packing on space-constrained planar facets. This study underscores the unique capability of ¹¹³Cd NMR to probe local Cd environments, offering a framework for monitoring structural transitions and improving size control during quantum dot growth.

Engineering of semiconductor nanocrystals requires a detailed understanding of their formation. Here, the authors use signal-enhanced NMR and theoretical calculations to decode the structural transition from metastable CdSe clusters to mature quantum dots.

Introduction

The emergence of quantum dots has inaugurated the nanotechnology era, igniting scientific exploration and technological advancement^{1, 2–3}. Among these, CdSe nanocrystals have emerged as exemplary materials, showing excellent optical and electronic properties poised to influence a multitude of fields including photonics, electronics, energy conversion/storage, nanomedicine, and biosensing^{4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14–15}. The pivotal quantum confinement effect, stemming from their nanoscopic dimensions, underscores the unique properties of quantum dots (QD)s but also makes precise control of nanocrystal size and distribution imperative for effective applications¹⁶. Engineering nanocrystals with defined properties thus requires a detailed understanding of the nanocrystal formation process.

Conventional methods for monitoring nanocrystal formation rely on ensemble properties to glean kinetic insights. Techniques such as absorption spectroscopy, manifesting shifted absorption peaks with increasing sizes, and dynamic light scattering, commonly used to monitor bulk properties such as size and homogeneity, provide limited compositional growth information regarding how the electronic structure and coordination environments of core and surface atoms evolve during nanocrystal formation^{17, 18–19}. A detailed, atomic-level understanding of this compositional growth would provide microscopic insights into the structural, dynamic, and electronic environment changes in nanocrystals during synthesis, potentially guiding the rational design of experimental strategies to manipulate QD growth.

While techniques like scanning electron microscopy, tunneling electron microscopy, and atomic force microscopy offer exceptional spatial resolution for inspecting inner structures in quantum dots, they often fall short in providing quantitative measurements of environmental changes in Cd and Se atoms, e.g., electronic structure or local structure^{18,20, 21, 22–23}. X-ray diffraction is powerful in providing atomic coordinates, but requires superior sample quality and careful statistical analysis, especially for small clusters that lack long-range periodicity due to their small size. Moreover, the periodicity requirements of diffraction techniques often blind them to heterogeneous surface interactions with ligands crucial for growth control and nanocrystal stabilization.

Nuclear magnetic resonance (NMR), widely employed for deriving atomic-resolution information across diverse systems^{24, 25–26}, has emerged as a powerful tool for studying quantum dots, and especially their surface ligands^{27, 28, 29, 30–31}. Information derived from NMR, such as chemical shift and chemical shift anisotropy (CSA), encapsulates rich details about chemical bonding, electronic/molecular structure, and lattice dynamics. Despite its high tolerance for sample inhomogeneity, the low sensitivity of conventional NMR has generally limited its utility in semiconductor nanocrystal studies, particularly for nuclei with low gyromagnetic ratios and at natural isotopic abundance, and long T₁ relaxation times, which limit the experimental repetition rate^{23, 24–25,32, 33–34}. Dynamic Nuclear Polarization (DNP) Solid-State NMR (SSNMR) circumvents this sensitivity issue by transferring electron polarization to nuclei, leveraging the larger Boltzmann population of electron spin. This signal enhancement, reaching up to 660-fold for ¹H³⁵, compresses months-long NMR experiments into days, vastly expanding its applicability across various systems^{28,36, 37, 38, 39–40}. The high sensitivity enables the optimization of NMR experimental conditions on the studied samples and opens up possibilities for conducting multidimensional NMR on naturally abundant samples. Additionally, the reduced T₁ relaxation times due to the presence of radicals allow for much faster acquisition rates. Recent advancements in DNP sample preparation and pulse sequence design have significantly enhanced NMR performance for materials samples. These improvements enable multidimensional experiments that probe nuclear connectivity and separate isotropic chemical shifts from chemical shift anisotropy (CSA), providing valuable site-specific structural insights in various materials, including CdS, CdSe, and CdTe nanocrystals and nanoplatelets.

Materials in the CdX (X = O, S, Se, Te) family have been studied by SSNMR using ¹¹³Cd chemical shifts to elucidate their ligation environments^{27,28,32,33,36,41, 42, 43, 44–45}. We compile recent ¹¹³Cd chemical shift data in Fig. 1. The selection of chalcogen ligands (X) significantly affects the chemical shift in CdX₄ compounds, with increasing shielding from S to Se, Te, and O. This counterintuitive trend has been theoretically explained by the paramagnetic deshielding effect counterbalancing the shielding from spin-orbital coupling³⁶. Cd isotropic chemical shifts have been widely used to differentiate Cd environments in various contexts, including distinguishing core and surface Cd in CdSe quantum dots and nanoplatelets^32,38,40, and even tracking the growth of different Cd sites during deposition procedures³⁶. Recent studies have revealed heterogenous ligand binding modes on QD and nanoplatelet surfaces, observable via ¹³C NMR and infrared spectroscopy⁴¹. Similarly, ¹¹³Cd chemical shifts reveal different Cd binding modes, for example, contrasting carboxylate ligands in chelating (CdSe₂O₂), bridging (CdSe₃O), and tilting (CdSeO₃) configurations²⁸. (Fig. 1) Non-chalcogenide ligands, like amines, have also been used in CdSe synthesis. ¹¹³Cd chemical shifts have been reported for (CdSe)₁₃ clusters passivated with amine in a recent study⁴². Given the emerging diversity in the ligation and structures of quantum dots and related materials, many moieties remain unexplored by NMR. Here, we investigate new moieties and demonstrate the utility of ¹¹³Cd chemical shifts in probing the local environment.

Fig. 1 A survey of ¹¹³Cd chemical shift data in various nanomaterials. [Images not available. See PDF.]

We analyzed several Cd-binding moieties, including those involving chalcogen and nitrogen-based ligands. The new moieties introduced in this study are highlighted with red squares. For example, the CdSe₂NO moiety represents a coordination of Cd with two Se atoms, one N atom, and one O atom. Notably, ¹¹³Cd chemical shifts are highly indicative of binding environments. Detailed experimental values, including chemical shifts, chemical shift anisotropy, and corresponding literature references, are provided in Supplementary Table 1.

CdSe atomically precise clusters (containing well-defined atoms)⁴⁶, also known as “magic size” clusters⁴⁷, are hypothesized to be kinetically trapped intermediates in nanocrystal formation (Fig. 2); they can further grow into nanocrystals or nanoplates under specific conditions^46,48,49. Meanwhile, etching with butylamine can convert nanocrystalline CdSe samples into smaller clusters⁵⁰. Specific changes in electronic and molecular structure over the transition between magic size clusters and mature nanocrystals are of great interest. Such insights will enable the discovery of new strategies to control size distribution and nanocrystal chemistry generally. Structural studies of atomically precise clusters are few in the literature^40,45,51, despite their immense potential. Motivated by the available detailed structures of these systems, and by their chemical reactivity, we attack the central question of identification of the NMR peaks for specific atoms in the cluster. Here, we demonstrate that Cd atoms with different ligands exhibit distinctive chemical shifts, enabling us to uncover the ligand distribution on the cluster surface. Our findings highlight the critical role of molecular forces, such as hydrogen bonding and steric clashes, in determining ligand distribution.

Fig. 2 Energy profile of CdSe QD growth. [Images not available. See PDF.]

An energy of formation profile is presented along number of atoms per crystal, with structural models of CdSe_350nm, CdSe_408nm nanoclusters as metastable intermediates, and the 2.8 nm mature QD as the final product.

Results and discussion

Structural models of the atomically precise CdSe clusters are depicted in Fig. 2 as we determined from single crystal X-ray diffraction and pair distribution function analysis as described earlier⁴⁷. With edge lengths from 1.7 to 2.6 nm, each cluster features a tetrahedral, Zinc Blende (ZB, a common crystal structure in QD with a cubic lattice) cadmium selenide core enriched in cadmium, surrounded by a layer of benzoate and N-butylamine ligands. The precise formulas derived from solution NMR and Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) and well-defined structures of these clusters significantly facilitate the analysis of NMR spectra⁴⁷. The clusters are denoted as CdSe_350nm and CdSe_408nm, respectively, based on their maximum absorption wavenumbers (i.e., 350 nm and 408 nm). Their formula, number of ligands, and coordination sites are summarized in Supplementary Table 2. Additionally, we present the structural model of a spherical nanocrystal with a diameter of 2.8 nm, representing the oleic acid-capped CdSe QD studied in this work (Supplementary Fig. 1).

An improved sample preparation method for successful material DNP studies

In DNP experiments, a radical, often a nitroxide biradical, is combined with the sample to provide an exogenous source of unpaired electrons. Continuous microwave radiation facilitates the transfer of electron polarization to nearby nuclei through the cross effect⁵². To achieve successful NMR signal enhancement, meticulous sample preparation is crucial. Direct polarization transfer from polarizing radicals to nearby nuclei is limited to a short range, typically within a few nanometers, due to dipolar coupling; polarization then propagates via spin diffusion, usually through the solvent, which can extend over much greater distances⁵³. Nevertheless, the diffusion length is constrained by the proton T₁ (a few seconds in many samples) and the solvent’s deuteration level. Homogeneous mixing of the sample with the radical ensures uniform polarization transfer to the sample, thereby improving the overall enhancement.

A common radical impregnation procedure for preparing DNP samples involves soaking solid samples with a radical solution to impregnate the samples with radicals. However, this impregnation procedure often results in heterogeneity due to insufficient mixing, and separation may occur during cooling, as DNP experiments are typically conducted at around 100 K. This challenge is frequently tackled by utilizing host materials such as silicon dioxide or hexagonal boron nitride for the radical-sample mixture, preventing separation during cooling and improving microwave efficiency to achieve significant enhancements^37,38. However, the presence of host materials reduces the effective sample volume within the NMR rotor³⁸. After optimizing the ratio of sample to host material and refining preparation procedures, a balance has been achieved between DNP enhancement and sample loading, improving overall sensitivity³⁷.

Initially, we used the radical impregnation procedure by preparing an AMUPol solution in a 1:9 protonated/deuterated toluene mixture and adding it to powder samples without host materials. However, we observed a low enhancement factor, which we attribute to insufficient and inhomogeneous mixing of the added radicals. The bulky structure of radicals may cause limited diffusion of radicals through the samples. To overcome this, we implemented an alternative preparation method (Fig. 3a). Using the same protonated/deuterated toluene (ratio 1:9) as a solvent for AMUPol, we dissolved the radical and then added samples in small increment into the radical solution to attain a well-mixed, homogenous solution-like consistency. This approach was particularly effective for CdSe materials, as toluene is a commonly used solvent for these samples. The 90% deuterated toluene provided an optimal balance between facilitating spin diffusion and minimizing proton relaxation effects from protonated methyl groups, which could otherwise reduce DNP enhancement. Additionally, using toluene as a “native” solvent, instead of alternatives like tetrachloroethane, allowed us to study CdSe materials in a more representative environment.

Fig. 3 An optimized DNP-NMR sample preparation method for studying CdSe materials. [Images not available. See PDF.]

a The general sample preparation procedure used in this study to prepare DNP-NMR samples. b–d Examples of significant signal enhancements on ¹H (both 16 scans), ¹³C (both 16 scans, CP Experiments) and ¹¹³Cd (both 4096 scans, CP Experiments) using DNP-NMR, comparing spectra with the microwave source on and off. Source data are provided as a Source Data file.

The DNP enhancements achieved using this protocol are shown in Fig. 3b–d; they are comparable to or slightly lower than those reported for host-based methods. However, the lower enhancement can be compensated for by high sample loading in the rotor; the high loading of studied materials is evidenced by the observation of ¹¹³Cd signals with the gyrotron off (Fig. 3d); the overall high sensitivity enables us to easily optimize experimental parameters (using 8-128 scans) and acquire high signal-to-noise (S/N) spectra. (For example, in Fig. 3d, with 2.8 h acquisition with 2.5 s pulse delay, the ¹¹³Cd peak at −20 ppm has S/N~300.) The high S/N facilitated confident spectrum fitting to extract NMR information. It is also worth noting that the advantage of not using host materials is even more prominent in Fast MAS studies, where even smaller rotor diameters are used.

NMR reveals symmetric and asymmetric bonding in the core and on the surface, respectively, in the mature QD

We initiated our measurements by examining ¹¹³Cd nuclei on the mature oleate-capped QDs. Opting for ZB-quantum dots with a mean diameter of approximately 2.8 nm ensured comparability in size to the largest cluster, CdSe_408nm, with an estimated tetrahedral edge length of 2.567 nm. Similar CdX QD with different sizes or shapes or chalcogen element X have been studied through ¹¹³Cd NMR, identifying Cd atoms in the core or on the surface through ¹¹³Cd-¹³C, ¹¹³Cd-⁷⁷Se and ¹¹³Cd spin diffusion experiments^28,37,38,40. Similarly, we identified two major Cd environments in the mature QD sample, corresponding to core and surface Cd atoms (Fig. 4).

Fig. 4 Spectral assignments for core and surface Cd atoms in QD. [Images not available. See PDF.]

a Comparison of Direct Excitation (DE, red) and ¹H-¹¹³Cd cross-polarization (CP, blue) spectra for the 2.8 nm QD. Peaks corresponding to core Cd, surface Cd, and starting material Cd oleate are labeled as #1, #2, and #3, respectively. Spinning sidebands are indicated with asterisks. b Illustration showing how Cd atoms near ligands exhibit higher CP efficiency in the ¹H-¹¹³Cd CP experiment. c The spectrum was fitted to extract the chemical shift and chemical shift anisotropy (CSA). Full fitting parameters and results can be found in Supplementary Fig. 2. Source data are provided as a Source Data file.

To further confirm the assignments of these two environments, we applied a common solid-state NMR technique by comparing the ¹¹³Cd spectra obtained using direct excitation (DE) on ¹¹³Cd nuclei (in red) and through cross-polarization (CP) transfer from nearby ¹H to ¹¹³Cd (in blue). The surface Cd, located near ligands containing ¹H, exhibited high ¹H −¹¹³Cd cross-polarization efficacy, resulting in amplified surface signal as compared to the core signal in ¹H-¹¹³Cd CP spectra (Fig. 4a, blue), making use of the large gyromagnetic ratio and polarization of ¹H. Similar strategies have proven effective in many non-DNP SSNMR studies^{32, 33–34}. Direct enhancement of ¹¹³Cd, on the other hand, results in spectra with a dominant signal from the core, due simply to the larger proportion of core sites in the sample. The effective depth for direct polarization transfer from the solvent has been shown by previous DNP studies to approach 40 Å (4 nm)⁵⁴, which, for the sizes of QD used in the current study, allows effective polarization of the entire core. In the CP spectrum, we observed surface Cd, characterized by the isotropic peak and spinning sidebands, as well as the core Cd, assumed to be layers underneath the surface due to short-ranged ¹H −¹¹³Cd polarization transfer. In the DE spectrum (Fig. 4a, red), we observed an intense peak for core Cd and very weak signals for surface Cd. We noted a slightly shifted center and broader linewidth for the core Cd peak compared to the CP spectrum.(Fig. 4a, peak #1 expansion) We attribute this to the fact that CP probes core atoms with only a limited penetration depth, while DE can probe the full ensemble of core Cd atoms. Despite being highly similar in structure, i.e., both representing a CdSe₄ moiety, the interior core Cd do show a slight chemical shift perturbation compared to core Cd close to the surface.

The core Cd peak exhibited no spinning sidebands at 8 kHz magic angle spinning (MAS), indicating minimal chemical shift anisotropy (CSA). CSA, intrinsic to chemical bonding, suggests that the chemical bond environment of the Cd atoms is highly symmetric for the core Cd atoms, consistent with the tetrahedral bonding geometry with Se atoms. Conversely, the surface Cd displayed a large CSA, evidenced by the presence of many spinning sidebands. Experiments were conducted at MAS rates of 8, 11.111, and 13.333 kHz in order to distinguish the isotropic peak from the spinning sidebands. Since the isotropic peak position is invariant with respect to the MAS rate, but spinning sidebands are spaced at integer multiples of the MAS rate, the latter can be easily identified by their apparent motion in the spectrum as the MAS rate is altered. (Supplementary Fig. 2). Quantitative spectral fitting revealed the following parameters: for core Cd, δ(iso) = −52.2 ppm, | Δ | <10 ppm, η < 0.05; for surface Cd, δ(iso) = −312.2 ppm, Δ = −255.6 ppm, η = 0.79. (Fig. 4c, Supplementary Table 3, all ¹¹³Cd shifts referenced to Me₂Cd). Values for both sites are very consistent with previous studies, including those where different sizes or shapes of CdSe materials were examined (Fig. 1)^19,32,38,41. Such a result suggests that ¹¹³Cd chemical shifts are highly sensitive to their local environment, i.e. distinguishing core and surface, but less sensitive to global structure. The large values of CSA parameters, Δ and η of surface Cd indicate that their local electronic environment is notably asymmetric. These surface Cd atoms are hypothesized to be located at the dominant {100} facet, bonding with two Se atoms and two oxygen atoms from the carboxylate group in the ligand^28,41.

Additionally, we measured ⁷⁷Se spectra by DE and ¹H-⁷⁷Se CP. Similarly, we discerned the surface and core Se atoms, with chemical shifts of −524 ppm and −614 ppm, respectively (Supplementary Fig. 3; all ⁷⁷Se shifts are referenced to Me₂Se). Such assignments agree with previous reports made based on ⁷⁷Se spin diffusion experiments to distinguish core and surface Se²⁸. The observed Se chemical shifts are indicative of the ZB structure rather than the Wurtzite structure³⁸ as expected for the QDs used in this study. In addition to the two major peaks, we observed another broad up-field peak, which can be decomposed into two peaks centered at −656 and −701 ppm (Supplementary Table 4). We cannot confidently assign these minor peaks; we tentatively attribute them to heterogenous surface Se at other types of facets. Se atoms in Site 4 show a considerably large CSA value, indicating that they might be uncoordinated.

Spectral assignments in Cluster_408nm reveals ligand distribution

Next, we investigated the larger CdSe cluster, CdSe_408nm, which has an edge length of 2.6 Å and a slightly smaller volume than the spherical 2.8 nm QD. The formula for CdSe_408nm is Cd₈₄Se₅₆X₅₆L₅₆, where X represents benzoate and L represents N-butylamine. We recorded both DE and ¹H-¹¹³Cd CP spectra for ¹¹³Cd nuclei. (Fig. 5a–d) We again collected spectra at different MAS frequencies to exclude spinning sidebands and determine three major isotropic shifts, at −22, −164, and −236 ppm (Fig. 5a–c). Assigning these peaks was possible by analyzing the structural model of CdSe_408nm (Fig. 2), which has 84 cadmium atoms in four types of Cd environments: vertex (4), edge (30), face (40), and interior (10). The diversity of atom types is further complicated by the presence of two ligand types, benzoate and N-butylamine, that coordinate surface Cd atoms using oxygen and nitrogen atoms, respectively.

Fig. 5 Spectral assignments for distinctive sites in CdSe_408nm. [Images not available. See PDF.]

a–c¹H-¹¹³Cd CP spectra recorded at various magic angle spinning (MAS) speeds, along with fitted spectra for four distinct Cd environments. Root-Mean-Square (RMS) values were calculated to quantify the differences between the experimental and fitted spectra. The solid red line represents the experimental spectrum, while the dashed black line corresponds to the synthesized spectrum, combining the fits for various sites. The difference spectrum between the experimental and synthesized spectra is shown as a red solid curve in the lower panel with a dashed blue line as the baseline at 0. d¹¹³Cd DE spectra recorded at 13.333 kHz, along with the corresponding fitted spectra. Experimental, synthesized, and difference spectra are displayed in the same format. Details on the fitting procedure are provided in the experimental section, and the fitting results are summarized in Supplementary Table 5. Source data are provided as a Source Data file.

Despite this complexity, we assigned the first isotropic peak at −22 ppm to the interior and face Cd atoms bound to N-butylamine. We base this assignment on the following observations:

1. The interior atoms have a highly similar chemical bonding environment to the core Cd atoms in mature QD. We observed significant signals from interior Cd in the QD sample in both DE and CP experiments (Fig. 4a). Similarly, we expect interior Cd atoms in the clusters to be polarized and contribute to the observed signal. As the majority of interior Cd atoms in the clusters are located one layer beneath the surface, they fall within the expected range for direct polarization transfer from radicals in DE experiments and ¹H −¹¹³Cd cross-polarization in the CP experiment⁵⁴. The isotropic chemical shift for the core Cd in the 2.8 nm QD is −52.2 ppm, which is closest to −22 ppm. Therefore, we assign this peak to the core Cd atoms.

2. The interior Cd atoms alone cannot account for the intensity of this peak since they constitute only 11.9% of the total Cd atoms. Peak intensities on the DE spectrum should semi-quantitatively represent Cd atom stoichiometry. Considering the depth from the surface and distance from the ligands and polarizing radicals, the relative signal intensity of interior Cd atoms should be lower than the stoichiometric ratio of interior and surface atom types. However, the peak in question comprises approximately 56.4% of the total signal, as calculated by peak integration on the DE spectrum, suggesting that it must contain intensity from surface atoms. The face Cd atoms have the closest chemical environment to the core Cd, with three sites bonded with Se and one site coordinated with a ligand. The combined percentage of face and center Cd atoms (59.5%) closely matches the peak intensity (56.5%). We later discovered that only face Cd ligated with N-butylamine exhibit matching chemical shifts, and these are the dominant type of face Cd atoms (see below).

To support our hypothesis, we conducted density functional theory (DFT) calculations on the NMR parameters of smaller pyramidal cluster that is missing corner cadmium atoms and only contains face and edge Cd atoms (Fig. 6a). Previous studies have shown that calculated ¹¹³Cd chemical shifts are in excellent agreement with experimental values for various Cd compounds, including CdSeO₃, CdSe₂O₂, and CdSe₃O moieties on the CdSe QD surface (Fig. 1)^55,56. In this model cluster, the face Cd atoms were ligated with N-butylamine, while the edge ones were ligated with one N-butylamine and one benzoate. Chemical shifts are calculated for individual Cd sites as shown by the solid bars (Fig. 6b); Gaussian distributions were fitted to core and surface sites to extract averages and standard deviations. The average of the face Cd (−38 ppm) closely matches the experimentally observed resonance (−22 ppm), supporting our assignment of this peak to both face and interior Cd atoms. Previous studies of a similar CdSe₃N moiety, present in (CdSe)₁₃ clusters, reported chemical shift between −60 and −70 ppm; considering the nearly 1000 ppm-wide chemical shift range of ¹¹³Cd and the variation in coordination geometry, the similarity is consistent with our observed values⁴².

Fig. 6 DFT calculations-assisted Cd chemical shift assignments. [Images not available. See PDF.]

a The cluster model used for DFT calculations of NMR parameters, where face Cd atoms are shown in purple and are ligated with N-butylamine, while edge Cd atoms are shown in yellow and are ligated with both N-butylamine and benzoate. Se atoms are omitted for clarity. b DFT calculation results for ¹¹³Cd chemical shifts of the face (purple) and edge Cd (yellow) atoms, fitted to Gaussian distributions with corresponding mean and standard deviation. Source data are provided as a Source Data file.

We conducted further quantitative analysis of the CP and DE spectra by fitting them with four Cd sites, two of which are the interior Cd and face Cd with N-butylamine, as discussed above. Utilizing spectra acquired at multiple spinning speeds, we restrained the fittings by employing the same NMR parameters to fit all spectra in order to prevent overfitting (Fig. 5a–d). The fitted peak for the interior Cd is highlighted in gray, and the fitting results are as follows: δ(iso) = −22 ± 2 ppm, | Δ|<10 ppm, |η|<0.05. The upbound of | Δ | and |η| values here are chosen arbitrarily to reflect a minimal chemical shift anisotropy due to the symmetric bonding environment in interior atoms.

The fitted peak for the face Cd is highlighted in blue, and according fitting parameters are as follows: δ(iso) = −22 ± 2 ppm, Δ = 156 ± 5 ppm, η = 0.4 + 0.1/−0.2. The calculated Δ is 189 ± 27 ppm, and η is 0.4 ± 0.1, both of which are consistent with experimental values. Despite similarity in isotropic shift for face and interior Cd, the face Cd shows considerably larger CSA as it has an asymmetric bonding environment due to coordination with the nitrogen atom in N-butylamine.

The third fitted peak, highlighted in red, has an accurately determined isotropic chemical shift; however, due to weak intensity and overlap with nearby peaks, its CSA parameters cannot be confidently fitted: δ(iso) = −164 ± 5 ppm, | Δ|<150 ppm, |η|<1. The fourth fitted peak, highlighted in green, has the following fitting parameters: δ(iso) = −236 ± 5 ppm, Δ = −158 ± 15 ppm, |η|<0.6.

We assigned the peak at −164 ppm to the face Cd atoms coordinated with benzoate. The CdSe₃O moiety is similar to previously reported surface Cd atoms with a bridging carboxylate group (Fig. 1), whose reported chemical shift (~−150 ppm) closely matches the observed −164 ppm. We also used DFT to calculate the chemical shifts of this environment based on the simplified model but with benzoate on the face; however, the calculated chemical shifts for the four face sites show a large deviation, indicating that our model is either trapped in a local minimum during geometry minimization or intrinsically unstable. However, a recent DFT calculation based on a more simplified model of the CdSe₃O moiety, employing spin-orbital levels and the ZORA Hamiltonian (consistent with our theoretical approach), yielded a value of −129 ppm, close to the observed −164 ppm⁵⁵.

We assigned the peak at −236 ppm to edge Cd atoms coordinated with one benzoate and one N-butylamine. The chemical shifts of such Cd atoms were calculated to be −222 ± 35 ppm (Fig. 6b), consistent with the observed fourth peak at −236 ppm. We ruled out other models, such as edge atoms with two N-butylamine or two benzoate ligands.(Supplementary Discussion) There is no published experimental data for the Cd₂N₂ moiety, but, for the Cd₂O₂ moiety, Cd atoms with a similar configuration shown in Fig. 1 have a chemical shift between −310 and −322 ppm.

We also measured ⁷⁷Se spectra and fitted them for the cluster (Supplementary Fig. 4, Supplementary Table 6). Overall, the spectra are quite congested; site-specific assignments of these peaks have not been performed due to extensive peak overlap and limited experimental data available for referencing.

Validating chemical shift assignments with Cluster_350nm

We then investigated the smaller cluster, Cluster_350nm, with an estimated edge length of 1.7 nm. The formula for this system is Cd₃₅Se₂₀X₃₀L₃₀, where X represents benzoate and L represents N-butylamine. The structure model shows that it has a vertex (4), an edge (18), a face (12), and an interior (1). The ¹H-¹¹³Cd CPMAS spectrum shows two major peaks at −20 ppm and −260 ppm. (Fig. 7) The first peak is very close to the −22 ppm peak for the interior and face Cd in the cluster_408nm. Therefore, we assign this peak also to the interior and face Cd with N-butylamine. Since there is only one interior Cd in Cluster_350nm, this validates our previous assignment of this peak to both interior and face Cd to account for this most intense peak.

Fig. 7 Spectral assignments for distinctive sites in CdSe_350nm. [Images not available. See PDF.]

¹H-¹¹³Cd CP spectra of CdSe_350nm at 8.4 kHz (red solid line) and 11 kHz (blue solid line). The 8.4 kHz spectrum is fitted with four distinct Cd environments, represented as blue (face Cd), gray (interior Cd), green (edge Cd), and red (potential vertex Cd). The overall fit is shown by the black dashed line. The line at −164 ppm emphasizes the absence or minimal presence of a peak at this position in the 11 kHz spectrum, in contrast to CdSe_408nm. Detailed fitting results are provided in Supplementary Table 7. Source data are provided as a Source Data file.

The other peak is at −260 ppm, close to the peak (−236 ppm) assigned to edge Cd coordinated with one benzoate and one amine in Cluster_408nm. Though it is somewhat shifted, we still assign this peak to the edge Cd coordinated with benzoate and amine. We attribute the slight chemical shift perturbation to different ligand packing as cluster size decreases, leading to variation in bond length or angle between Cd and N/O atoms, especially considering the bulky side chain in both types of ligands.

In Cluster_350nm, we observe minimal or no peak at/around −164 ppm, which would correspond to face Cd coordinated with benzoate. This is especially evident in the 11 kHz MAS spectrum; at 8 kHz, an overlap of spinning sidebands falls in this region (Fig. 7). This suggests that there is little or no Cd in such an environment in Cluster_350nm. Notably, Cluster_350nm has only 3 face Cd compared to 10 in Cluster_408nm, and the space available on each face is much smaller. The rigid, bulky phenyl side chain of benzoate may cause steric clashes due to the limited space, making such ligation less likely. This ligand arrangement on the facets is consistent with a model displaying the lowest energy in a quantum mechanics (QM) calculation⁵¹. (To be discussed later)

Quantitative fitting of the spectrum shows that the CSA parameters of the face and edge Cd are quite similar to those seen in CdSe_408nm (Supplementary Table 7), indicating that these two types of Cd share a similar electronic environment with their counterpart in Cluster_408nm (Supplementary Table 5).

To further validate our proposed model, we performed a ¹H-¹⁵N CP experiment using room-temperature NMR to probe the ¹⁵N chemical shift of the N-butylamine ligand bound at various sites (face vs. edge).(Supplementary Fig. 5) The spectrum reveals a single broad peak (Full-Width at Half Height: 8.65 ppm) at 30.5 ppm. This result indicates that the ¹⁵N chemical shift is insufficient to differentiate slight variations in the Cd environments to which the nitrogen is bound. This observation is consistent with the behavior of the Cd(benzoate)₂(N-butylamine)₂ compound, where the Cd atom has six coordination sites instead of four, yet the ¹⁵N chemical shift changes by only ~2 ppm. We anticipate that multinuclear multidimensional NMR experiments, capable of correlating signals from different ligands or their bound Cd atoms, could provide valuable validation of our model. However, such experiments require isotopically labeled ligands, which we will attempt to produce for future investigations.

Ligand distribution models and HB network

Based on our experimental findings, we propose ligand distributions for both clusters. For the CdSe_350nm cluster, all face sites are coordinated with flexible N-butylamine ligands to prevent steric clashes. The edge sites are coordinated with one N-butylamine and one benzoate each, while the vertex sites are coordinated with three benzoate ligands (Fig. 8a). We also show that such a distribution is the only model consistent with our experimental and computational results.(Supplementary Discussion: Ligand distribution model validation) The ligand distribution in the CdSe_408nm cluster is similar except for face Cd. NMR analysis suggests that some face Cd atoms coordinate with benzoate ligands. We distributed the benzoate ligands across the face sites to maximize the number of hydrogen bonds formed between benzoate and N-butylamine (Fig. 8b, details discussed below). Although we have limited experimental data on ligand coordination at the vertex sites, modeling of the vertex Cd atoms results in sufficient space for them to be able to coordinate with three benzoate ligands without inducing steric clashes.

Fig. 8 Ligand distribution models. [Images not available. See PDF.]

Proposed ligand distribution models for CdSe_350nm (a) and CdSe_408nm (b) based on NMR-derived restraints, are shown. A close-up view highlights the potential hydrogen bond (HB), with the corresponding distance between H and O atoms labeled, between X (benzoate) and L (N-butylamine) on the surface of the CdSe_408nm cluster.

Previous QM computation studies have shown that HB networks significantly enhance the thermodynamic stability of clusters⁵¹. However, performing QM geometry optimizations to accurately evaluate HB networks in large clusters is challenging due to the high flexibility of ligands and the complexity of the energy landscape. As a result, using QM-calculated HB numbers as an energy function to explore ligand distribution space and identify configurations that maximize HB formation is computationally prohibitive. To efficiently estimate the number of HBs, we developed an algorithm that counts: (1) intra-HBs (where X and L ligands are bound to the same Cd atom) and (2) inter-HBs (between X and L ligands on adjacent Cd atoms). For inter-HBs, we restricted the count to one HB per adjacent Cd pair. Using this algorithm to compute the number of HBs as an energy function, we performed a Monte Carlo simulation to assign ligands on the CdSe cluster surface, maximizing HB formation without experimental restraints. The maximum HB configurations yielded 114 and 226 HBs for CdSe_350nm and CdSe_408nm, respectively. Applying the same algorithm to calculate HBs in our proposed models (Fig. 8a, b), we found 102 HBs for the CdSe_350nm model and 220 HBs for the CdSe_408nm model. These results suggest that while our proposed models do not achieve the maximum possible HB formation, they nonetheless support extensive HB networks, particularly in the case of CdSe_408nm, where space constraints are reduced.

Our model for the CdSe_350nm cluster aligns closely with a previously proposed model based on energy minimization (EM) among a few candidates⁵¹. In the EM study, the most stable isoform of Cd₃₅Se₂₀X₃₀L₃₀ (X = formate, L = ammonia) shows a configuration with three L ligands on the face, seven X ligands and five L ligands on the edge, and two X ligands and one L ligand on the vertex. Our model differs by swapping one X ligand from the edge with one L ligand on the vertex. This difference could be attributed to the smaller ligands used in the EM calculations. Our assignment is supported by the absence of a characteristic peak around −310 ppm for the CdSe₂O₂ moiety, which would be expected in the EM model but not in our NMR-derived model.

In conclusion, in this study, we demonstrated that the ¹¹³Cd chemical shift serves as a unique probe for distinguishing different sites in CdSe nanomaterials. We successfully addressed a key challenge in assigning experimental ¹¹³Cd chemical shifts to well-defined Cd environments in atomically precise CdSe clusters with the help of quantum mechanical calculations, supplemented by previous experimental results in similar environments. The newly obtained chemical shift data for these novel moieties will aid in the structural assignment of related systems in future research. Furthermore, we used ¹¹³Cd chemical shifts to tackle the complex problem of determining ligand distribution on the surface of CdSe clusters. Our findings indicate that hydrogen bonding and steric clashes are key drivers of ligand arrangement, with flexible amine ligands being favored in the facets when spatial constraints are present in small clusters. Our work paves the way for exploring structural transitions from clusters to mature QDs, as we provide chemical shift information for both states. We anticipate that future studies will leverage ¹¹³Cd chemical shifts to dynamically monitor structural changes and investigate how growth conditions influence growth.

Methods

Cluster and QD synthesis

Atomically precise CdSe_350nm, CdSe_408nm tetrahedral clusters were synthesized from cadmium benzoate and bis(trimethylsilyl)selenide according to Beecher et al. ⁴⁷ Following Hamachi et al. ⁵⁷, 2.8 nm CdSe nanocrystals were synthesized from 1,3-diphenyl-2-imidazolidineselone (0.150 mmol, 1.0 equiv), cadmium oleate (0.180 mmol, 1.2 equiv,) and oleic acid (0.360 mmol, 2.4 equiv.). The morphology of QDs is shown in TEM image in Supplementary Fig. 1. The absence of precursors, such as cadmium benzoate or Cd(benzoate)₂(N-butylamine)₂, in the cluster samples can be confirmed by the lack of peaks at the corresponding chemical shifts in the ¹¹³Cd NMR spectra. Chemical shifts for cadmium benzoate and Cd(benzoate)₂(N-butylamine)₂ can be inferred from Fig. 4a and Supplementary Fig. 6, respectively.

DNP sample preparation

All samples were packed into 3.2 mm sapphire rotors (Bruker Corp.), sealed using silicone rubber disks specifically adapted for the rotor size, and finally capped with zirconia drive caps optimized for cryogenic operation. Samples were packed and sealed in a glove box under a nitrogen atmosphere to prevent sample oxidation. AMUpol dissolved in 90% deuterated toluene served as the DNP polarization reagent. 30 μL of 20 mM AMUPol radical solution was added to a mortar, and 20–30 mg of powder-form CdSe samples were gradually introduced in small increments to ensure full dissolution. The resulting mixture was highly homogeneous and solution-like. Subsequently, the mixed solutions were carefully pipetted into 3.2 mm sapphire Bruker rotors. The total sample weight ranges from 20 mg to 30 mg, with radical concentrations ranging from 8.8 mM to 17.8 mM across various samples, determined using an EMX Nano Benchtop EPR spectrometer(Bruker). These prepared samples were stored at 4 °C prior to DNP-NMR measurements.

DNP-NMR experiments

DNP experiments were conducted on a Bruker 9.4 T (equivalent to 400 MHz for ¹H) spectrometer equipped with a Neo console. A 3.2 mm HXY triple-channel probe was configured in HX mode and tuned accordingly to facilitate ¹H, ¹³C, ¹¹³Cd, and ⁷⁷Se experiments. The sample temperature was maintained at approximately 100 K, with samples spun at various spinning speeds (i.e., 8 kHz, 11.111 kHz, 13.333 kHz). The typical pulse powers were 100 kHz (50 W) for ¹H, 62.5 kHz (37 W) for ¹³C, 62.5 kHz (55 W) for ¹¹³Cd, and 62.5 kHz (65 W) for ⁷⁷Se. Direct excitation experiments were performed using a 90° excitation pulse followed by echo detection with a rotor-synchronized echo delay (two rotor periods). All cross-polarization (CP) spectra were acquired through polarization transfer from ¹H to the observed nuclei, with the CP conditions optimized at various magic-angle spinning speeds. Proton decoupling was applied using the SPINAL-64 scheme with a 100 kHz ¹H RF field during all spectra acquisitions, except for ¹H spectra. For more detailed information on the NMR experimental parameters for each spectrum presented, please refer to the Supporting Information (Supplementary Table 8). ¹¹³Cd T₁ is measured to be approximately 1 s for cluster samples (Supplementary Fig. 7); in the direct excitation experiment, 10 s pulse delay was used to ensure magnetization fully recovered to equilibrium to improve the accuracy of stoichiometry measurements. To demonstrate sample integrity following DNP-NMR experiments, we measured UV-Vis spectra of a CdSe_350nm sample before and after the NMR experiments, which showed minimal changes. (Supplementary Fig. 8)

Chemical shifts for ¹³C nuclei were referenced externally to the downfield adamantane line at 40.48 ppm, while chemical shifts for other nuclei were referenced accordingly by calculation to correspond to 0 ppm for DSS in D₂O (¹H), Me₂Cd (¹¹³Cd), and Me₂Se (⁷⁷Se). Spectra were processed using Topspin 3.6.0 and plotted using the nmrglue software⁵⁸. The orientation-dependent chemical shift anisotropy (CSA) was determined by the inequality |δzz−δ(iso)|≥ |δxx−δ(iso)|≥ |δyy−δ(iso) |. Based on this expression, δ(iso), Δ, and η were calculated as follows: δ(iso) = (δxx + δyy + δzz)/3, Δ = δzz−δ(iso), and η = (δyy - δxx)/Δ.

Spectrum fitting

The fitting of the ¹¹³Cd spectra was conducted in Topspin utilizing the sola module and verified with SIMPSON spectrum simulation⁵⁹. To determine the number of sites to be included in the fitting, we compared ¹¹³Cd spectra taken at various spinning speeds to identify the isotropic chemical shift, as spinning sidebands move with changes in spinning speed. For the peak centered at −22 ppm, two isotropic chemical shifts corresponding to the core and face Cd atoms, as detailed in the main text, were used to fit. Once the isotropic chemical shifts were fixed, spectra taken at 8 kHz were utilized to fit the chemical shift anisotropy (CSA) information, as they contained more spinning sidebands than spectra acquired at higher MAS rates and thus better preserved the CSA information. The initial values of the CSA parameters, peak intensity, and line broadening parameters were manually adjusted to roughly match the experimental spectra. Subsequently, the Simplex algorithm was applied to optimize the fitting. Both Lorentzian and Gaussian broadening were applied to fit lineshapes. The derived CSA parameters were then employed to fit spectra taken at 11.111 and 13.333 kHz, with variables only in peak intensities and line broadening parameters. All fittings were evaluated by calculating the root-mean-square (rms) value of the normalized fitting and experimental spectra. The successful fitting of spectra taken at various spinning speeds with the same set of isotropic and CSA parameters significantly bolstered confidence in the extracted NMR parameters. Fitting errors were estimated to have a 2% influence on the overall fitting by varying each parameter individually.

DFT calculation

DFT calculations of ¹¹³Cd NMR parameters were conducted using ORCA (5.0.3)⁶⁰. Given the computational challenges associated with heavy atoms like Cd and Se, direct calculations on the clusters, even the smaller ones, are computationally expensive. As an alternative, we constructed smaller CdSe structure models comprising only the face and edge Cd atoms.(Fig. 6a) These models were employed to investigate the effects of different ligands on Cd chemical shifts by generating models with varying ligand distributions. Subsequently, all structure models underwent QM geometry optimization. DFT calculations were then performed using the PBE0 functional with ZORA relativistic correction⁵⁵ and SARC/J method for NMR calculations. The basis sets employed for different elements were as follows: hydrogen, carbon, nitrogen, oxygen: “def2-SVP, cadmium: “SARC-ZORA-TZVP”, selenium: “def2-TZVP”. The calculated nuclei shielding shift was then referenced against the calculated shielding values for model compounds Me₂Cd and Me₂Se before being compared to experimental chemical shifts.

Hydrogen Bond calculation and Monte Carlo Simulation

The code for hydrogen bond (HB) calculations and Monte Carlo simulations used in ligand assignment is available on GitHub (https://github.com/Xyunyao/CdSe_nmr_DFT). Both intra- and intermolecular HBs between butylamine and benzoate are counted. For intermolecular HBs between adjacent Cd atoms, only one HB is considered per butylamine-benzoate pair, since correct molecular orientation is required. We applied this HB counting function as a scoring metric, using Monte Carlo simulations and Metropolis criteria to maximize HB formation. Given the simplified HB counting algorithm, this method estimates the effectiveness of HB formation rather than identifying conformers with the maximum number of HBs.

Acknowledgements

A.E.M. discloses support for the research of this work from NSF MCB 24-27038. J.S.O. discloses support for publication of this work from NSF CHE-2004008. We thank Kirk Baughman, Dr. Chengming He, Dr. Lakshimi Bhai from for their assistance with manuscript preparation, sample preparation, and instrument operation.

Author contributions

Y.X., I.V.S., and A.E.M. designed and conducted the solid-state NMR characterization, analyzed NMR results. Y.X., L.Z., and R.A.F. designed and conducted QM calculations and analyzed the results. M.W.G., N.S., A.B., and D.C. prepared the materials; Y. X. drafted the manuscript. A. E.M., J.S.O. supervised the project. All authors contributed to scientific discoveries and the writing of the manuscript.

Peer review

Peer review information

Nature Communications thanks Xueqian Kong, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available to download at FigShare (https://doi.org/10.6084/m9.figshare.28898612).  are provided with this paper.

Code availability

The custom code used for analyzing NMR data, DFT data and Monte Carlo simulation can be found at GitHub (https://github.com/Xyunyao/CdSe_nmr_DFT) as well as in Zenodo (https://doi.org/10.5281/zenodo.15626365).

Competing interests

The authors declare the following competing interests: R.A.F. is a cofounder and consultant for Schrodinger Inc. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-025-62724-8.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Kovalenko, MV et al. Prospects of nanoscience with nanocrystals. ACS Nano; 2015; 9, pp. 1012-1057.1:CAS:528:DC%2BC2MXhtlyrsL4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25608730][DOI: https://dx.doi.org/10.1021/nn506223h]

2. Steigerwald, ML; Brus, LE. Semiconductor crystallites: a class of large molecules. Acc. Chem. Res.; 1990; 23, pp. 183-188.1:CAS:528:DyaK3cXisVels7Y%3D [DOI: https://dx.doi.org/10.1021/ar00174a003]

3. Owen, JS; Chan, EM; Liu, H; Alivisatos, AP. Precursor conversion kinetics and the nucleation of cadmium selenide nanocrystals. J. Am. Chem. Soc.; 2010; 132, pp. 18206-18213.2010JAChS.13218206O1:CAS:528:DC%2BC3cXhsFSqtr7E [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21128655][DOI: https://dx.doi.org/10.1021/ja106777j]

4. Dai, X., Deng, Y., Peng, X. & Jin, Y. Quantum-dot light-emitting diodes for large-area displays: towards the dawn of commercialization. Adv. Mater. 29, 1607022 (2017).

5. Nomura, M; Arakawa, Y. Light sources: Shaking quantum dots. Nat. Photonics; 2012; 6, pp. 9-10.2012NaPho..6..9N1:CAS:528:DC%2BC3MXhs1Gks7vK [DOI: https://dx.doi.org/10.1038/nphoton.2011.323]

6. Bruns, O. T. et al. Next-generation in vivo optical imaging with short-wave infrared quantum dots. Nat. Biomed. Eng. 1, 0056 (2017).

7. Kim, TH; Jun, S; Cho, KS; Choi, BL; Jang, E. Bright and stable quantum dots and their applications in full-color displays. MRS Bull.; 2013; 38, pp. 712-720.2013MRSBu.38.712K1:CAS:528:DC%2BC3sXhsVOgsLbF [DOI: https://dx.doi.org/10.1557/mrs.2013.184]

8. Buckley, S., Rivoire, K. & Vučković, J. Engineered quantum dot single-photon sources. Reports Prog. Phys. 75, 126503 (2012).

9. Talapin, DV; Steckel, J. Quantum dot light-emitting devices. MRS Bull.; 2013; 38, pp. 685-691.2013MRSBu.38.685T1:CAS:528:DC%2BC3sXhsVOgsLbP [DOI: https://dx.doi.org/10.1557/mrs.2013.204]

10. Shirasaki, Y; Supran, GJ; Bawendi, MG; Bulović, V. Emergence of colloidal quantum-dot light-emitting technologies. Nat. Photonics; 2013; 7, pp. 13-23.2013NaPho..7..13S1:CAS:528:DC%2BC38XhvV2ntLfI [DOI: https://dx.doi.org/10.1038/nphoton.2012.328]

11. Rafailov, EU; Cataluna, MA; Sibbett, W. Mode-locked quantum-dot lasers. Nat. Photonics; 2007; 1, pp. 395-401.2007NaPho..1.395R1:CAS:528:DC%2BD2sXnslymtrg%3D [DOI: https://dx.doi.org/10.1038/nphoton.2007.120]

12. Li, H., Wu, K., Lim, J., Song, H. J. & Klimov, V. I. Doctor-blade deposition of quantum dots onto standard window glass for low-loss large-area luminescent solar concentrators. Nat. Energy1, 16157 (2016).

13. Weiss, EA. Designing the surfaces of semiconductor quantum dots for colloidal photocatalysis. ACS Energy Lett.; 2017; 2, pp. 1005-1013.1:CAS:528:DC%2BC2sXlslSksbo%3D [DOI: https://dx.doi.org/10.1021/acsenergylett.7b00061]

14. Kagan, CR; Lifshitz, E; Sargent, EH; Talapin, DV. Building devices from colloidal quantum dots. Science; 2016; 353, aac5523. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27563099][DOI: https://dx.doi.org/10.1126/science.aac5523]

15. Liu, W et al. Compact biocompatible quantum dots functionalized for cellular imaging. J. Am. Chem. Soc.; 2008; 130, pp. 1274-1284.2008JAChS.130.1274L1:CAS:528:DC%2BD1cXivF2guw%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18177042][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2665712][DOI: https://dx.doi.org/10.1021/ja076069p]

16. Rossetti, R; Nakahara, S; Brus, LE. Quantum size effects in the redox potentials, resonance Raman spectra, and electronic spectra of CdS crystallites in aqueous solution. J. Chem. Phys.; 1983; 79, pp. 1086-1088.1983JChPh.79.1086R1:CAS:528:DyaL3sXks1Witr4%3D [DOI: https://dx.doi.org/10.1063/1.445834]

17. Abécassis, B et al. Real-Time in Situ Probing of High-Temperature Quantum Dots Solution Synthesis. Nano Lett.; 2015; 15, pp. 2620-2626.2015NanoL.15.2620A [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25815414][DOI: https://dx.doi.org/10.1021/acs.nanolett.5b00199]

18. Lee, J., Yang, J., Kwon, S. G. & Hyeon, T. Nonclassical nucleation and growth of inorganic nanoparticles. Nat. Rev. Mater. 1, 16034 (2016).

19. Palencia, C et al. An in situ and real time study of the formation of CdSe NCs. Nanoscale; 2020; 12, pp. 22928-22934.1:CAS:528:DC%2BB3cXitFKltLbP [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33206104][DOI: https://dx.doi.org/10.1039/D0NR05879C]

20. Georgiev, P et al. Implementing atomic force microscopy (AFM) for studying kinetics of gold nanoparticle’s growth. Colloids Surf. A Physicochem. Eng. Asp.; 2013; 434, pp. 154-163.1:CAS:528:DC%2BC3sXhtVKmsLbM [DOI: https://dx.doi.org/10.1016/j.colsurfa.2013.05.064]

21. Park, J et al. 3D structure of individual nanocrystals in solution by electron microscopy. Science; 2015; 349, pp. 290-295.2015Sci..349.290P1:CAS:528:DC%2BC2MXhtFKkt7fP [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26185247][DOI: https://dx.doi.org/10.1126/science.aab1343]

22. Billinge, SJL; Levin, I. The problem with determining atomic structure at the nanoscale. Science; 2007; 316, pp. 561-565.2007Sci..316.561B1:CAS:528:DC%2BD2sXksFektr8%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17463280][DOI: https://dx.doi.org/10.1126/science.1135080]

23. Steigerwald, ML et al. Surface derivatization and isolation of semiconductor cluster molecules. J. Am. Chem. Soc.; 1988; 110, pp. 3046-3050.1988JAChS.110.3046S1:CAS:528:DyaL1cXhvFehurg%3D [DOI: https://dx.doi.org/10.1021/ja00218a008]

24. Thayer, AM; Steigerwald, ML; Duncan, TM; Douglass, DC. NMR Study of Semiconductor Molecular Clusters. Phys. Rev. Lett.; 1988; 60, 2673.1988PhRvL.60.2673T1:CAS:528:DyaL1cXltVentr4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10038419][DOI: https://dx.doi.org/10.1103/PhysRevLett.60.2673]

25. Yesinowski, JP. Solid-state NMR of inorganic semiconductors. Top. Curr. Chem.; 2012; 306, pp. 229-312.1:CAS:528:DC%2BC38XhsVeiurvJ [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21898208][DOI: https://dx.doi.org/10.1007/128_2011_208]

26. Xu, Y et al. Electron and Spin Delocalization in [Co6Se8(PEt3)6]0/+1 Superatoms. ChemPhysChem; 2024; 25, 1:CAS:528:DC%2BB3sXisFGhu77J [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38057144][DOI: https://dx.doi.org/10.1002/cphc.202300064] e202300064.

27. Kurihara, T; Noda, Y; Takegoshi, K. Quantitative solid-state NMR study on ligand-surface interaction in Cysteine-capped CdSe magic-sized clusters. J. Phys. Chem. Lett.; 2017; 8, pp. 2555-2559.1:CAS:528:DC%2BC2sXot1yns7w%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28534619][DOI: https://dx.doi.org/10.1021/acs.jpclett.7b00909]

28. Chen, Y et al. Revealing the surface structure of CdSe nanocrystals by dynamic nuclear polarization-enhanced 77Se and 113Cd solid-state NMR Spectroscopy. J. Am. Chem. Soc.; 2021; 143, pp. 8747-8760.2021JAChS.143.8747C1:CAS:528:DC%2BB3MXht1aisrzM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34085812][DOI: https://dx.doi.org/10.1021/jacs.1c03162]

29. Knauf, RR; Lennox, JC; Dempsey, JL. Quantifying ligand exchange reactions at CdSe nanocrystal surfaces. Chem. Mater.; 2016; 28, pp. 4762-4770.1:CAS:528:DC%2BC28XhtVKmt77O [DOI: https://dx.doi.org/10.1021/acs.chemmater.6b01827]

30. Hens, Z; Martins, JC. A solution NMR toolbox for characterizing the surface chemistry of colloidal nanocrystals. Chem. Mater.; 2013; 25, pp. 1211-1221.1:CAS:528:DC%2BC3sXisV2hsLo%3D [DOI: https://dx.doi.org/10.1021/cm303361s]

31. Anderson, NC; Hendricks, MP; Choi, JJ; Owen, JS. Ligand exchange and the stoichiometry of metal chalcogenide nanocrystals: Spectroscopic observation of facile metal-carboxylate displacement and binding. J. Am. Chem. Soc.; 2013; 135, pp. 18536-18548.2013JAChS.13518536A1:CAS:528:DC%2BC3sXhslCrsr%2FI [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24199846][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4102385][DOI: https://dx.doi.org/10.1021/ja4086758]

32. Ratcliffe, CI et al. Solid state NMR studies of photoluminescent cadmium chalcogenide nanoparticles. Phys. Chem. Chem. Phys.; 2006; 8, pp. 3510-3519.1:CAS:528:DC%2BD28Xntl2ktbs%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16871340][DOI: https://dx.doi.org/10.1039/b606507b]

33. Berrettini, MG; Braun, G; Hu, JG; Strouse, GF. NMR analysis of surfaces and interfaces in 2-nm CdSe. J. Am. Chem. Soc.; 2004; 126, pp. 7063-7070.2004JAChS.126.7063B1:CAS:528:DC%2BD2cXjvVShu70%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15174877][DOI: https://dx.doi.org/10.1021/ja037228h]

34. Lovingood, DD; Achey, R; Paravastu, AK; Strouse, GF. Size- and site-dependent reconstruction in CdSe QDs evidenced by 77Se{1H} CP-MAS NMR spectroscopy. J. Am. Chem. Soc.; 2010; 132, pp. 3344-3354.2010JAChS.132.3344L1:CAS:528:DC%2BC3cXitV2hsbw%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20163186][DOI: https://dx.doi.org/10.1021/ja907511r]

35. Lilly Thankamony, AS; Wittmann, JJ; Kaushik, M; Corzilius, B. Dynamic nuclear polarization for sensitivity enhancement in modern solid-state NMR. Prog. Nucl. Magn. Reson. Spectrosc.; 2017; 102–103, pp. 120-195. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29157490][DOI: https://dx.doi.org/10.1016/j.pnmrs.2017.06.002]

36. Piveteau, L et al. Colloidal-ALD-Grown Core/Shell CdSe/CdS Nanoplatelets as Seen by DNP Enhanced PASS-PIETA NMR Spectroscopy. Nano Lett.; 2020; 20, pp. 3003-3018.2020NanoL.20.3003P1:CAS:528:DC%2BB3cXjsV2nsr0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32078332][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7227022][DOI: https://dx.doi.org/10.1021/acs.nanolett.9b04870]

37. Hanrahan, MP et al. Probing the Surface Structure of Semiconductor Nanoparticles by DNP SENS with Dielectric Support Materials. J. Am. Chem. Soc.; 2019; 141, pp. 15532-15546.2019JAChS.14115532H1:CAS:528:DC%2BC1MXhs1Kks7bO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31456398][DOI: https://dx.doi.org/10.1021/jacs.9b05509]

38. Piveteau, L et al. Structure of colloidal quantum dots from dynamic nuclear polarization surface enhanced NMR Spectroscopy. J. Am. Chem. Soc.; 2015; 137, pp. 13964-13971.2015JAChS.13713964P1:CAS:528:DC%2BC2MXhs1Ohur%2FP [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26473384][DOI: https://dx.doi.org/10.1021/jacs.5b09248]

39. Rossini, AJ et al. Dynamic nuclear polarization surface enhanced NMR Spectroscopy. Acc. Chem. Res.; 2013; 46, 1942.1:CAS:528:DC%2BC3sXksVyis7o%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23517009][DOI: https://dx.doi.org/10.1021/ar300322x]

40. Piveteau, L et al. Resolving the core and the surface of CdSe quantum dots and nanoplatelets using dynamic nuclear polarization enhanced PASS-PIETA NMR Spectroscopy. ACS Cent. Sci.; 2018; 4, pp. 1113-1125.1:CAS:528:DC%2BC1cXhtFyisrvE [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30276244][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6161058][DOI: https://dx.doi.org/10.1021/acscentsci.8b00196]

41. Zhang, J et al. Identification of facet-dependent coordination structures of carboxylate ligands on CdSe nanocrystals. J. Am. Chem. Soc.; 2019; 141, pp. 15675-15683.2019JAChS.14115675Z1:CAS:528:DC%2BC1MXhslGnsrvF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31503473][DOI: https://dx.doi.org/10.1021/jacs.9b07836]

42. Hsieh, TE et al. Unraveling the structure of magic-size (CdSe)13 cluster pairs. Chem. Mater.; 2018; 30, pp. 5468-5477.1:CAS:528:DC%2BC1cXhtlCnsbbM [DOI: https://dx.doi.org/10.1021/acs.chemmater.8b02468]

43. Boles, MA; Ling, D; Hyeon, T; Talapin, DV. The surface science of nanocrystals. Nat. Mater.; 2016; 15, pp. 141-153.2016NatMa.15.141B1:CAS:528:DC%2BC28Xht1Cksrs%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26796733][DOI: https://dx.doi.org/10.1038/nmat4526]

44. Liu, H; Owen, JS; Alivisatos, AP. Mechanistic study of precursor evolution in colloidal group II-VI semiconductor nanocrystal synthesis. J. Am. Chem. Soc.; 2007; 129, pp. 305-312.2007JAChS.129.305L1:CAS:528:DC%2BD28Xhtleiu7nF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17212409][DOI: https://dx.doi.org/10.1021/ja0656696]

45. Levchenko, TI; Lucier, BEG; Corrigan, JF; Huang, Y. Crystalline superlattices of nanoscopic CdS molecular clusters: An X-ray Crystallography and 111Cd SSNMR Spectroscopy study. Inorg. Chem.; 2018; 57, pp. 204-217.1:CAS:528:DC%2BC2sXhvFOhu7bF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29243923][DOI: https://dx.doi.org/10.1021/acs.inorgchem.7b02403]

46. Cossairt, BM; Owen, JS. CdSe clusters: At the interface of small molecules and quantum dots. Chem. Mater.; 2011; 23, pp. 3114-3119.1:CAS:528:DC%2BC3MXmsFCrur4%3D [DOI: https://dx.doi.org/10.1021/cm2008686]

47. Beecher, AN et al. Atomic structures and gram scale synthesis of three tetrahedral quantum dots. J. Am. Chem. Soc.; 2014; 136, pp. 10645-10653.2014JAChS.13610645B1:CAS:528:DC%2BC2cXhtFajsbjN [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25003618][DOI: https://dx.doi.org/10.1021/ja503590h]

48. Jiang, ZJ; Kelley, DF. Role of magic-sized clusters in the synthesis of CdSe nanorods. ACS Nano; 2010; 4, pp. 1561-1572.1:CAS:528:DC%2BC3cXisFGlsrg%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20192241][DOI: https://dx.doi.org/10.1021/nn100076f]

49. Cunningham, PD; Coropceanu, I; Mulloy, K; Cho, W; Talapin, DV. Quantized reaction pathways for solution synthesis of colloidal ZnSe nanostructures: a connection between clusters, nanowires, and two-dimensional nanoplatelets. ACS Nano; 2020; 14, pp. 3847-3857.1:CAS:528:DC%2BB3cXjvVGmu70%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32105062][DOI: https://dx.doi.org/10.1021/acsnano.9b09051]

50. Landes, C; Braun, M; Burda, C; El-Sayed, MA. Observation of large changes in the band gap absorption energy of small CdSe nanoparticles induced by the adsorption of a strong hole acceptor. Nano Lett.; 2001; 1, pp. 667-670.2001NanoL..1.667L1:CAS:528:DC%2BD3MXntlWqtLg%3D [DOI: https://dx.doi.org/10.1021/nl015619t]

51. Nguyen, KA; Pachter, R; Jiang, J; Day, PN. Systematic study of structure, stability, and electronic absorption of Tetrahedral CdSe clusters with carboxylate and amine ligands. J. Phys. Chem. A; 2018; 122, pp. 6704-6712.1:CAS:528:DC%2BC1cXhtlyktLbI [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30040899][DOI: https://dx.doi.org/10.1021/acs.jpca.8b02813]

52. Maly, T et al. Dynamic nuclear polarization at high magnetic fields. J. Chem. Phys.; 2008; 128, 052211.2008JChPh.128e2211M [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18266416][DOI: https://dx.doi.org/10.1063/1.2833582]

53. Barnes, AB et al. High-field dynamic nuclear polarization for solid and solution biological NMR. Appl. Magn. Reson.; 2008; 34, pp. 237-263.1:CAS:528:DC%2BD1cXhtFent7%2FN [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19194532][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2634864][DOI: https://dx.doi.org/10.1007/s00723-008-0129-1]

54. Akbey, Ü et al. Dynamic nuclear polarization of spherical nanoparticles. Phys. Chem. Chem. Phys.; 2013; 15, pp. 20706-20716.1:CAS:528:DC%2BC3sXhslOmtr3O [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24192797][DOI: https://dx.doi.org/10.1039/c3cp53095g]

55. Biswas, R; Chen, Y; Vela, J; Rossini, AJ. Relativistic DFT calculations of Cadmium and Selenium solid-state NMR Spectra of CdSe nanocrystal surfaces. ACS Omega; 2023; 8, pp. 44362-44371.1:CAS:528:DC%2BB3sXit1OmsrvM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38027327][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10666156][DOI: https://dx.doi.org/10.1021/acsomega.3c07680]

56. Holmes, ST; Schurko, RW. A DFT/ZORA Study of Cadmium magnetic shielding tensors: analysis of relativistic effects and electronic-state approximations. J. Chem. Theory Comput.; 2019; 15, pp. 1785-1797.1:CAS:528:DC%2BC1MXisFyiu7w%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30721042][DOI: https://dx.doi.org/10.1021/acs.jctc.8b01296]

57. Hamachi, LS et al. Precursor reaction kinetics control compositional grading and size of CdSe_1-x S_x nanocrystal heterostructures. Chem. Sci.; 2019; 10, pp. 6539-6552. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31367306][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6615248][DOI: https://dx.doi.org/10.1039/C9SC00989B]

58. Helmus, JJ; Jaroniec, CP. Nmrglue: An open source Python package for the analysis of multidimensional NMR data. J. Biomol. NMR; 2013; 55, pp. 355-367.1:CAS:528:DC%2BC3sXmtlCht74%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23456039][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3636164][DOI: https://dx.doi.org/10.1007/s10858-013-9718-x]

59. Bak, M; Rasmussen, JT; Nielsen, NC. SIMPSON: A general simulation program for solid-state NMR Spectroscopy. J. Magn. Reson.; 2000; 147, pp. 296-330.2000JMagR.147.296B1:CAS:528:DC%2BD3cXotlKlurk%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11097821][DOI: https://dx.doi.org/10.1006/jmre.2000.2179]

60. Neese, F. Software update: the ORCA program system, version 4.0. WIREs Comput. Mol. Sci.; 2018; 8, pp. e1327-e1327. [DOI: https://dx.doi.org/10.1002/wcms.1327]