We studied the interaction of five typical ALD precursors for the growth of metal oxides by c-ALD on carboxylate passivated NC surfaces.2,19 We opted for alkyl-metals and metal-amides as they are soluble in non-polar solvents and their expected byproducts should not interfere with shell formation. We selected trimethylgallium (TM-Ga), dimethyl zinc (DM-Zn), trimethyl aluminium (TM-Al), tris(dimethylamido)gallium (TDMA-Ga) and tetrakis(dimethylamido)titanium (TDMA-Ti), as a representative set to contrast the interaction strength of each precursor (Supplementary Fig. 1).19,20
For the purpose of this study, we elected to focus mainly on PbS NCs, as they are valuable for a variety of optoelectronic applications and their surface chemistry is well-established.21,22 We also opted to grow metal oxide shells on InP NCs, CdSe nanoplatelets (NPLs) and NaGdF4 NCs as they are recognized alternatives in display, lasing, and biological imaging but face persistent issues in terms of stability.23–26 We synthesized PbS NCs (spheres and cubes), InP NCs, CdSe NPLs and NaGdF4 NCs all capped by oleate ligands through previously established protocols (Supplementary Figs. 2–12).26–30
Figure 1a-e reports 1H NMR spectra which track changes in the alkene resonance of bound oleates on PbS NCs while titrating the c-ALD precursor.31 During each titration, a new resonance upfield to the alkene resonance emerges, which indicates an interaction of the precursors with the oleate ligands. Titrations were arrested when precipitation of the NCs occurred. We compared the amount of displaced/transformed oleates from the surface per equivalents of added precursor (Supplementary Figs. 13–22), which we used as a metric to quantify the strength of interaction between the ALD precursors and the oleate ligands. Based on such quantification, we established a reactivity trend among the precursors: TM-Al\(>\)DM-Zn\(\ge\)TM-Ga\(<\)TDMA-Ga\(<\)TDMA-Ti (Fig. 1, Supplementary Note 1).
Drawing parallels with gas phase ALD on polymers32–35, we consider that an adduct forms between the Lewis acidic metal precursors and the Lewis basic carbonyls of bound oleates. TM-Al is a harder Lewis acid than TM-Ga and DM-Zn.36 Thus, TM-Al is expected to interact more strongly with the bound oleates compared to the other two methyl-based precursors. Through the same principle, it is anticipated that the harder Lewis acidity of TDMA-Ti compared to TDMA-Ga rationalizes the larger release of oleates for the former. Finally, the larger displacement observed for TDMA-Ga compared to TM-Ga correlates well with the harder character of the amide ligands compared to the methyl groups, effectively hardening the precursor and making it more reactive towards the bound oleates.
We then sought to identify the chemical species generating the new 1H NMR resonances. Based on a suite of 1D and 2D NMR characterization, alkyl-metals result in the formation of cluster-like species (Supplementary Figs. 23–31), which do not interact with the NC surface. These cluster-like species are known to assemble from such precursors and metal-carboxylates.37–40 Instead, when using metal amide precursors, we found sharp resonances (Fig. 1f, Supplementary Figs. 32–40), that correspond to dimethyl oleamides.41,42 The presence of this species is consistent with an amidation reaction between the precursor and the surface oleate, which results in the formation of dimethyl oleamide and of a Pb-O-M (M = Ti or Ga) bond on the surface of the PbS NCs (Fig. 1g).32 Thus, metal-amide precursors react with surface oleate to directly anchor the precursor to the NC surface by transforming the starting ligands into species which do not interact with the NC surface.
Having learned that all the metal precursors do interact with the surface of the oleate-capped PbS NCs, we proceeded in growing the oxide shell according to the previously reported c-ALD method, which yields hybrid oxide shells with embedded oleate ligands.14,16 Each precursor lead to the formation of such shells on the NCs, with the exception of TM-Ga (Supplementary Figs. 41–43), which is consistent with the reported reactivity scale. This latter issue can be circumvented by having an alumina layer initially present, exploiting the higher reactivity of TM-Al to seed gallium oxide formation (Supplementary Figs. 44–45). Further, only dilute and sub-stoichiometric amounts of the precursors can be employed in this method without causing loss of colloidal stability of the NCs. This limitation results in a partial coverage of the NC surface by the oxide layer for each cycle. To expand the scope of this previous methodology and yield a fully inorganic interface between the NC and the oxide shell, we reasoned that both the precursor and oxygen source needed to be colloidally solubilizing ligands.
To this end, we developed a methodology that relies upon the observation of dimethyl oleamide and Pb-O-M bond formation during the addition of metal-amide precursors. We first sought to endow the metal precursor with colloidally solubilizing functionalities (i.e make it a good ligand). Thus, we performed a transamidation reaction to partially replace the short destabilizing dimethylamide with solubilizing didodecylamides (Fig. 2a, Supplementary Figs. 46–47, Supplemental Note 2).44 Further, we noted that the addition of oleic acid to metal-amide precursors results in the formation of metal oxides and dimethyl oleamides (Fig. 2b Supplementary Fig. 48).
We combined these two reactions to propose a scheme for the growth of metal oxides that displaces all the native ligands and generates a fully inorganic interface between the NC and the oxide shell without loss of colloidal stability (Fig. 2c). In the first half-reaction, the transamidated precursor efficiently replaces all the native oleate ligands through an amidation reaction to form oleamide and bound gallium-amides. Then, oleic acid is added to induce the release of secondary amines by forming gallium oleate through a protic exchange with the bound amides, effectively regenerating a bound oleate surface while completing the second half reaction and representing a full c-ALD cycle. In the following c-ALD cycles, the newly passivating oleates are converted into oleamide by the transaminated precursors, and a second metal oxide bond forms, permitting the growth of a second metal oxide layer. These steps can be repeated arbitrarily as each addition involves the positioning of a long chained aliphatic tail on the surface of the NC. Further, the evolved byproducts are alkyl oleamides and secondary amines, two species that interact weakly with the NC surface,41 and therefore should not interfere with the process.
We tracked the complete process through a suite of solution NMR techniques (Fig. 3, Supplementary Figs. 49–54) to confirm the proposed mechanism. As a first step, the transamidated gallium precursor was added to a suspension of PbS NCs. 1H NMR revealed the partial transformation of native oleates to alkyl oleamide through the emergence of narrow alkene resonances (Supplementary Fig. 49). A variable-temperature (VT) NMR experiment was performed to promote the full replacement of oleates and revealed that all the oleates are replaced after reaching 70\(℃\) (Fig. 3a). Concurrently, a very broad resonance arises (Fig. 3b), which is associated with bound Ga-didodecylamides. Diffusion ordered NMR spectroscopy (DOSY) and nuclear Overhauser effect spectroscopy (NOESY) further confirmed the anchoring of gallium amides to the NC surface and the completion of the first half reaction (Supplementary Fig. 49). Presence of excess precursor does not result in further reactions, suggesting the reaction is self-limiting (Supplementary Fig. 50). Importantly, the NCs remained colloidally stable despite all the native oleates having been replaced by a Ga-amide terminated surface.
Upon introduction of oleic acid, the signal arising from bound gallium-didodecylamide is lost and a bound oleate resonance is reestablished in the 1H NMR spectrum (Figs. 3a-b, Supplementary Fig. 51). We note that this resonance is upfield compared to the original PbS bound oleates, and is consistent with oleates being bound to a metal oxide.41 DOSY and NOESY further confirm this assignment. These steps were repeated twice and tracked, showing a similar behavior (Figs. 3a-b, Supplementary Figs. 52–55). This demonstration confirms that our methodology proceeds as intended, i.e. by alternating the surface ligands between bound oleates and bound amides (Fig. 3c). Thus, this approach should be growing metal oxides on the surface of PbS NCs.
Figure 4 displays high-angle annular dark field- scanning transmission electron microscopy (HAADF-STEM) images with the corresponding energy dispersive X-Ray spectroscopy (EDXS) elemental maps of different NCs upon the growth of gallium or titanium oxide shells. After shell growth of both metal oxides, PbS nanocubes retain their shape and size, while EDXS elemental maps confirm the spatial co-localization of both Pb and the gallium or titanium, and the presence of a shell (Fig. 4a-f; PbS spheres are presented in Supplementary Fig. 56). We applied the same approach to oleate passivated NaGdF4 NCs, InP NCs and CdSe NPLs to generalize the growth procedure (Fig. 4g,h, Supplementary Fig. 57–59). Further, optical characterization confirmed that the InP and CdSe cores are unaffected by the growth of metal oxides (Supplementary Fig. 60–61). These results evidence the generalizability of this procedure to a variety of oleate passivated nanocrystalline compositions.
Using PbS nanocubes as our benchmark, we then sought to investigate the structure of the gallium oxide shell and its evolution during the growth. High-resolution X-ray diffraction (HR-XRD) analysis revealed no significant alteration to the PbS lattice (Supplementary Fig. 63). Further, the absence of additional diffraction peaks suggested that the gallium oxide shell is amorphous. We then performed XAS at the Ga K, Pb L2 and Pb L3 edges of PbS nanocubes having undergone one, three and five cycles of growth. The Pb L2 and L3 X-ray absorption near edge structure (XANES) revealed no significant alteration throughout any step in the growth process (Supplementary Fig. 64). Together with the XRD analysis, these data suggest that the grown gallium oxide is amorphous and does not induce surface oxidation or any measurable geometric strain to the PbS lattice.
Looking at the Ga K-edge XANES, the spectral line shape evolves towards gallium oxide nucleated by directly mixing the transamidated TDMA-Ga and oleic acid as more cycles are performed (Fig. 5a, Supplementary Fig. 48, 65). Extended X-ray absorption fine structure (EXAFS) analysis revealed clear changes during the growth of the shell (Fig. 5b, Supplementary Figs. 66–67, Table 1). First, the EXAFS of PbS nanocubes that have undergone only one cycle is consistent with the presence of O (first shell, r: 1.96Å), Ga (second shell, r: 3.01Å) and Pb (third shell, r: 3.91Å) scattering, but shows no evidence for a direct Ga-S bond. The absence of interactions with sulfur indicates that the anchoring of the first layer of gallium oxide is through the surface ligands rather than through undercoordinated sulfur sites. Furthermore, the third shell in the EXAFS is dominated by Pb-Ga scattering rather than by Ga-Ga. This Pb-Ga scattering fades as more cycles are performed and becomes better described by Ga-Ga scattering. Concomitantly, principle component analysis (PCA, Supplementary Figs. 68–69) indicated that the Ga K-edge XAS of PbS nanocubes after one cycle is consistent with a single component. Based on the EXAFS, we associate this component to gallium being a monolayer away to Pb. This component decreases from near unity to \(\sim\)30% and \(\le\)5% for the three and five cycles, respectively, while a second spectroscopic component associated with Ga-Ga scattering emerges. The EXAFS analysis also suggests a transition in the configuration of gallium atoms being closer to octahedral in the first cycle and leaning more towards tetrahedral as the shell grows (Supplementary Table 1). This result additionally emphasizes that the first layer is distinct and strongly influenced by the underlying PbS. Structurally the grown metal-oxide appears to be fully amorphous, yet its local structure might find similarities with the disordered \(\gamma\)-Ga2.67O4 phase.45–48
On the basis of all the above observations, we propose a plausible mechanism for the formation of the metal oxide shells initiated by ligand-modified precursors (Fig. 6). Starting with a carboxylate-functionalized surface (Fig. 6a), the first cycle results in the nucleation of a monolayer of gallium-oxygen species through the amidation of bound oleates by the metal amide precursor, with each metal likely anchoring with one or two surface sites (Fig. 6b). The formation of such monolayers is consistent with the self-limited nature of the proposed c-ALD process (Supplemental Note 2). During the following cycles, the growth likely proceeds by a continuous linear extension of gallium-oxygen species which merge to form the extended three-dimensional highly disordered gallium oxide network that propagates away from the surface (Fig. 6c).
Overall, the self-limiting character of the process allows for the layer-by-layer growth of fully inorganic metal oxide coatings of NCs while preserving their colloidal stability. Consequently, the presented c-ALD methodology is unique compared to the alternative metal oxide coating strategies (Supplementary Note 3).