Influence of Surface Chemistry on Metal Deposition Outcomes in Copper Selenide-Based Nanoheterostructure Synthesis

The use of nanoparticle surface chemistry to direct metal deposition has been well-studied in the modification of metal nanoparticle substrates but is not yet well-established for metal chalcogenide particle substrates, although integration of these particles into nanoheterostructures is of high interest. In this report, we investigate the effect of Cu2–xSe surface chemistry on the morphology of metal deposition on these plasmonic semiconductor nanoparticles. Specifically, we functionalize Cu2–xSe nanoparticles with a suite of 12 different ligands and investigate how different aspects of the ligand structure do or do not impact the morphology and extent of subsequent metal deposition on the Cu2–xSe surface. Surprisingly, our results indicate that the morphology of the resulting metal deposits and the extent of metal deposition onto the existing Cu2–xSe particle substrate are indistinguishable for the majority of ligands tested. An exception to these findings is observed for particles functionalized by quaternary alkylammonium bromides, which exhibit statistically distinct metal deposition patterns compared to all other ligands tested. We hypothesize that this unique behavior is due to a cooperative binding mechanism of the quaternary alkylammonium bromides to the surface of copper selenide. Taken together, these results yield both new strategies for controlling postsynthetic modification of copper selenide nanoparticles and also reveal limitations of surface chemistry-based approaches for this system.


Characterization of the Ligand Shell
In this report, we investigate the effects of a ligand shell on the post-synthetic deposition morphology of Au and Pt on colloidal Cu2-xSe NPs.Of course, to make the most robust structure-synthesis correlations, we aimed to characterize the ligand shell with the detail necessary to make those correlations.In particular, we sought to determine the zeta-potential, thickness, and chemical composition of each ligand shell tested.Below, we elaborate on challenges encountered while conducting these analyses as a footnote to our discussions in the main text.
To determine the density and identity of small molecule ligands on the NP surface quantitatively, our preferred method uses a 1D, solution phase 1 H-NMR based approach which has the advantage of not only giving the number of ligands per particle, but also being able to distinguish the variety of ligands present in most cases. 1 Unfortunately, the Cu 2+ content of Cu2-xSe NPs makes this analysis unfeasible due to peak broadening from the paramagnetic cation. 2 Thermogravimetric analysis (TGA) is another possible technique to quantify the ligand shell for metallic nanoparticles, at least in terms of mass.However, this method is limited in the characterization of multicomponent ligand shells (i.e.particle ligand shells that contain more than one ligand type), since often it is not possible to distinguish between ligand removal temperatures.We attempted to use TGA to analyze the ligand shell compositions but found high variation from sample to sample.The method also required large amounts of material for analysis (> 5-10 mg of conjugated Cu2-xSe NPs per experiment), and this barrier combined with the limited insight of the results, led us to explore alternative methods.
Zeta potential measurements support ligand exchange from CTAB to other ligands, where measured changes in zeta potential are consistent with each ligand charge (Table S1).These results are further supported by X-ray photoelectron spectroscopy (XPS) analysis.XPS was useful in determining both what was and was not present on the Cu2-xSe NP surfaces.In particular, it was important to note that we did not observe evidence of surface-bound bromide (or chloride or iodide) on any of the NP surfaces including those NPs functionalized with CTAB.Nitrogen signals indicating surface bound quaternary ammonium species observed in the CTAB-functionalized Cu2-xSe NP samples were not present after ligand exchange with either PEGSH, PVP, or SDS (Figure S3-6).XPS also indicates the presence of S-Cu(II) interactions in the case of PEGSH, and O-Cu(II) interactions in the case of PVP.
Taken together, the data support ligand exchange from CTAB to other ligand chemistries, with low or no levels of detectable CTAB after exchange.However, a quantitative picture of the ligand shell chemistries could not be obtained within the scope of a single report.

Notes on Ligand Concentration Selection
Ligand concentrations used for the ligand exchange were guided by three factors: 1) that ligand exchange occurs via mass action ligand exchange, and therefore, 2) the concentration should be in vast excess of theoretically available nanoparticle surface area as determined by particle concentration and theoretically determined minimum ligand footprint; as shown by several works concerning ligand functionalization on many types of nanoparticle surfaces, 1, 3-10 and 3) the concentration of ligand needed to obtain the most reproducible particle properties (e.g.XPS spectra, zeta potential, and extinction spectra) across multiple, independent trials.
In our experiments, the calculation of theoretical footprint is straightforward for the small molecule ligands tested and we have reported this approach in previous work. 1,11 wever, the calculation of theoretical maximum ligand loading in the case of polymeric ligands is, by definition, more complex because multiple ligand-surface interactions are possible per polymer chain.Therefore, we benchmarked our ligand excess to 100 times available surface area for our smallest footprint ligand.
In addition to these considerations, we took into consideration the varying affinities of the ligands for the NP surface, which impact the efficiency of the mass action ligand exchange.For example, a ligand with a weaker binding energy to the Cu2-xSe surface will need a larger ligand excess to achieve the same extent of ligand exchange when compared to a high binding affinity ligand.Therefore, ligands such as SDS required a larger excess compared to thiol-containing ligands.
For polymeric ligands, one can assess concentration comparison to the small molecule ligands either by benchmarking their concentration with respect to polymer chain concentration or to monomer unit concentration.Using either metric, there is some deviation from direct comparison to the ligand exchange conditions for small molecule ligands.
Taken together, these factors represent a wide set of conditions that make using the same concentration of each ligand during ligand exchange a factor that is likely to introduce more variation than it prevents.Instead, the concentrations listed in our experimental section represent a vast excess of ligand with respect to surface area in all cases, and then are additionally modified to concentrations empirically determined to give both the most reproducible particle properties (e.g.zeta potential, X-ray photoelectron spectra, and extinction spectra) and final NP morphologies.Table S1.Representative zeta potential measurements of the as-synthesized CTAB capped Cu2-xSe NPs and ligand exchanged Cu2-xSe NPs capped with 3.5 kDa PVP, 1 kDa PEGSH and SDS.Literature suggests that potentials of less than or equal to |10 mV| should be considered neutral. 12

X-ray Photoelectron Spectroscopy (XPS) analysis
XPS spectra of CTAB-Cu2-xSe NPs (Figure S3), shows no presence of halide on the surface of the Cu2-xSe NPs.The N 1s spectra shows two peaks at 399.3 eV and 402.7 eV.4][15] The Se 3d spectra of bare Cu2-xSe materials typically exhibits Se 4-peaks at ~53.8 eV. 16For our CTAB-capped Cu2-xSe, the Se 3d5/2 is shifted to slightly higher binding energy (~54.2eV), which corresponds to the Se-N interaction from the CTA + moiety.For the XPS spectra of PEGSH-Cu2-xSe NPs, S 2p3/2 spectra show two distinct peaks at 162 eV and 166.4 eV (Figure S4).The peak at 162 eV corresponds to a metal-thiol interaction that indicates that PEGSH binds through the S atom. 17The higher binding energy peak at 166.4 eV corresponds to the presence of oxidized thiol species.N and Br regions did not show resolvable peaks suggesting a thorough ligand exchange to PEGSH.From the XPS spectra of PVP-Cu2-xSe we can conclude that the interaction with Cu2-xSe surface likely occurs through the oxygen and not through the nitrogen in the pyrrolidone repeat unit (Figure S5).The O 1s spectrum shows two distinct binding energies at 530.9 eV and 532.6 eV.Literature reports suggest that unbound carbonyl in the pyrrolidone of PVP exhibits binding energies in the 530-531 eV range. 18,19 f the carbonyl O interacts with the Cu surface one would expect a shift to higher binding energies, and we observe a shift at higher binding energy (532.6 eV) consistent with that interaction.The N 1s spectrum also shows only one species, with a chemical shift that matches references to free PVP(C-N) as opposed to residual CTAB. 19Se 3d peaks for PVP-Cu2-xSe only show one binding environment at 54 eV, which suggests that there is only one type of Se 4-environment and that there is no CTAB-Se 4-interaction (N.B.The peak at ~59 eV corresponds to an oxidized Se 2+ species which likely arises from degraded NPs).Taken together, these XPS results are consistent with thorough ligand exchange of CTAB to PVP.The N 1s and Br 3d spectra of the SDS-capped Cu2-xSe NPs show that there is no evidence of CTAB on the surface indicating a complete exchange (Figure S6).The Cu 2p spectra shows the presence of both Cu(I) and Cu(II) species, but the exact binding moiety for SDS is not clear.The S 2p spectrum contains peaks that correspond to free SDS (Peak A at ~168.4 eV) 20 and some reduced sulfur species (peaks B and C).If SDS was binding through the S group to the Cu2-xSe surface, we would expect a S species at a higher binding energy than free SDS (>169 eV).Since we do not observe any higher binding energy species, we hypothesize that SDS binds through the O group to the surface of Cu2-xSe.The O1s spectrum shows only one species which is within the range of SDS's sulfate group but we cannot definitively assign this interaction. 20, 21

Statistical Analysis of Single vs. Multiple Islands on Cu2-xSe NPs
Single factor analysis of variance (ANOVA) was performed using Microsoft Excel 2017 Analysis ToolPak.This method is used to assess whether multiple populations have the same mean by comparing the variation between the means of multiple samples to the variation within each sample. 22The null hypothesis for this experiment is that measured Au island distributions are statistically the same.Here, we use two figures of merit to determine if our sample populations are statistically the same.First, if the p value is larger than a given confidence level, a (in this case, 95%, a = 0.05), then the null hypothesis cannot be rejected.Next, if the value (denoted F) produced by this sample is below a critical value (denoted Fcritical) for a given confidence level (a = 0.05), the means of each population are statistically the same.
The ANOVA derived parameters for Au depositions on PVP, PEGSH and SDS capped NP are statistically the same with one another.However, CTAB is statistically distinct from the other three ligands.

Calculating the Percent Surface Coverage
We have calculated the average percentage surface area of an individual Cu2-xSe NP that is covered by Au islands.To calculate this value, we estimated each Au island as a hemisphere as an approximation of the actual pseudospherical morphologies observed by TEM in each deposition case.(Scheme S1) Scheme S1.Schematic representation of surface coverage calculation.
We assume that the length (h) of the Au island is the radius of the hemisphere, which is measured from the TEM images using ImageJ software.Using the area of a circle and this value as the radius, we get the average surface area of a Cu2-xSe NP covered by each Au island.The next parameter we measure from the TEM images is the average number of Au deposits on each Cu2-xSe NP (N).The total surface area of the Cu2-xSe NPs covered by the Au islandic deposits is given by: ℎ 2 .
To better compare this value between different surface chemistries of Cu2-xSe, we express surface coverage as a percentage of the total surface area of a Cu2-xSe NP.
The average total surface area of a Cu2-xSe NP is calculated from the TEM size distributions as: 4 2 , where R is the average radius of the NP core.Putting these values together, the percentage of surface coverage is given by: (

Size Distributions of Particles and Deposition Islands
The size distributions of the core particle and metallic island deposits were analyzed using over 200 particles for each sample.Sizes are represented below both in histogram and table form for reader convenience.
Table S3.Size distributions of core nanoparticles (diameter) and deposition islands (length) (first deposition step).

Figure S9 .
Figure S9.(A,C) Representative TEM images of homogeneous Au and Pt nucleation on increasing the Au/Pt to Cu2-xSe ratio; (B,D) % modification and % surface coverage of Cu2-xSe NPs.

Figure S10 .
Figure S10.Representative TEM images of the Cu2-xSe (A) and first (B), second (C) and fourth (D) sequential gold deposition on Cu2-xSe NPs capped with CTAB.

Figure S20 .
Figure S20.Histograms of size distributions of gold islandic deposition length from sequential deposition on Cu2-xSe capped with (A,B,C) MUA (D,E,F) MDA and (G,H,I) MBA.

Figure S22 .
Figure S22.Percent surface coverage and percent modification of gold metal deposition on Cu2-xSe NPs capped with (A,B) PVP and (C,D) PEGSH.

Figure S27 .
Figure S27.Representative TEM images of the Cu2-xSe (A) and first (B), and second (C) sequential gold deposition on Cu2-xSe NPs ligand exchanged with TMOAB.

Figure S28 .
Figure S28.Representative TEM images of the Cu2-xSe (A) and first (B), and second (C) sequential gold deposition on Cu2-xSe NPs ligand exchanged with TOAB.

Figure S29 .
Figure S29.Representative TEM images of the Cu2-xSe (A) and first (B), second (C), and fourth (D) sequential gold deposition on Cu2-xSe NPs capped with MBA.

Table S2 .
Statistical analysis of Au island morphology using ANOVA.