Direct Excitation Transfer in Plasmonic Metal-Chalcopyrite-Hybrids: Insights from Single Particle Lineshape Analysis

Abstract

Hybrid nanomaterials containing both noble metal and semiconductor building blocks provide an engineerable platform for realizing direct and/or indirect charge and energy transfer for enhanced plasmonic photoconversion and photocatalysis. In this work, silver nanoparticles (AgNPs) and chalcopyrite (CuFeS₂) nanocrystals (NCs) are combined into a AgNP@CuFeS₂ hybrid structure comprising NCs embedded in a self-assembled lipid coating around the AgNP core. In AgNP@CuFeS₂ hybrid structures, both metallic and semiconductor NCs support quasistatic resonances. To characterize the interactions between these resonances and their effect on potential charge and energy transfer, direct interfacial excitation transfer between the AgNP core and surrounding CuFeS₂ NCs is probed through single particle lineshape analysis and supporting electromagnetic simulations. These studies reveal that CuFeS₂ NCs localized in the evanescent field of the central AgNP induce a broadening of the metal NP lineshape that peaks when an energetic match between the AgNP and CuFeS₂ NC resonances maximizes direct energy transfer. Dimers of AgNPs whose resonances exhibit poor energetic overlap with the CuFeS₂ NC quasistatic resonance yield negligible lineshape broadening in a control experiment, corroborating the existence of resonant energy transfer in the AgNP@CuFeS₂ hybrid. Resonant coupling between the metallic and semiconductor building blocks in the investigated hybrid architecture provides a mechanism for utilizing the large optical cross-section of the central AgNP to enhance the generation of reactive charge carriers in the surrounding semiconductor NCs for potential applications in photocatalysis.

Graphical Abstract

Introduction

The interaction of light with conduction band electrons in nanoparticles (NPs) with diameters smaller than the wavelength of the incident light can result in the excitation of localized surface plasmon resonances (LSPRs).¹ The resonance frequency of these coherent collective electron oscillations depends on the material, shape, and size of the NPs. LSPRs are most often associated with noble metal NPs,^2–5 but they are also observed in doped semiconductors.^6,7 The optical properties of plasmonic nanomaterials are determined by their dielectric function (ε), the real and imaginary components of which dictate the wavelength-dependent scattering and absorption cross-sections, respectively.⁸ The dipolar plasmon resonance wavelength for a spherical NP in the quasistatic limit is given by the Fröhlich resonance condition, which relates the real part of the dielectric function of the metal, Re(ε), to the dielectric constant of the ambient medium, ε_m:⁹ Re(ε) = −2ε_m. Due to their strong optical cross-sections, plasmonic NPs have found many applications in diverse scientific fields, including biosensing^10,11, medical diagnostics¹², nano-optics¹³, and photocatalysis^14–16, to name only a few. These applications commonly rely on colloidal Au or Ag NPs as these are easily accessible in defined sizes and shapes through established synthesis strategies and support tunable plasmon resonances throughout the visible and into the infrared (IR) region of the electromagnetic spectrum.^17–23 Recently, semiconductor nanocrystals (NCs) that sustain LSPRs^6,7,24 have attracted significant attention as alternative plasmonic materials. One appeal of these alternative plasmonic NCs is that their elemental building blocks show much greater natural abundance than noble metals. Other potential advantages include improved refractory properties²⁵ and compatibility with CMOS fabrication processes.^26,27 Advances in colloidal synthesis and a deeper understanding of free carrier properties have propelled research into materials like metal oxides^28,29, chalcogenides^15,30–33, and even silicon³⁴ for plasmonic applications.^27,28

A particular interesting class of heavy-metal-free “plasmonic” semiconductor nanomaterials are chalcopyrite (CuFeS₂) NCs. Notably, the real part of this intermediate band semiconductor is negative between 2.10 eV and 2.55 eV, and pure CuFeS₂ NCs were shown to fulfill the Fröhlich resonance condition at ~490 nm, giving rise to a distinct quasistatic resonance.³⁰ Similar to the plasmonic resonances observed in small metal NPs, decay of the CuFeS₂ NC-generated resonance has been associated with hot-electron generation.^30,37,38 In general, hybrid structures containing both noble metal and semiconductor components are of interest for 1) hot electron extraction from metal NPs and 2) stabilization of the hot electrons in the semiconductor NCs through a Schottky barrier.³⁹ However, “indirect” charge transfer mechanisms, like hot electron transfer (HET),^40–43 that require the decay of plasmons into excited electron-hole pairs in the metal prior to a charge transfer to the semiconductor in a subsequent step, suffer from a rapid thermalization of the electrons that competes with the transfer.^44,45Alternative “direct” charge and energy transfer mechanisms, such as plasmon-induced interfacial charge transfer transition (PICTT)^43,46 or plasmon-induced resonant energy transfer (PIRET)^47,48, are driven by plasmons without a need of prior generation of hot charge carriers and thus have some potential for overcoming challenges associated with HET.^49–52 Hybrid structures consisting of a central AgNP core surrounded by CuFeS₂ NCs (AgNP@CuFeS₂) represent a class of metal-semiconductor hybrids in which both building blocks exhibit collective resonances in the visible range of the electromagnetic spectrum.^53–55 This design provides opportunities for implementing direct charge and energy transfer mechanisms, including under conditions whereby the resonances of both metal and semiconductor components overlap to facilitate an efficient energy transfer.

Direct excitation transfer mechanisms differ from indirect processes in that they represent plasmon dephasing processes. Importantly, as the plasmon lifetime determines the width of the scattering spectra of individual NPs, the presence of these plasmon decay processes is experimentally tractable through single particle spectroscopy. In this work, we utilize single particle spectroscopy and combine it with electromagnetic simulations to investigate AgNP@CuFeS₂ hybrids containing AgNPs with nominal diameters of 40 nm and 60 nm, as well as AgNP-dimer(60)@CuFeS₂ hybrids containing dimers of 60 nm AgNPs. We present evidence of a CuFeS₂ NC-induced broadening of the AgNP resonance line-shape in AgNP@CuFeS₂ hybrids that depends on the energetic overlap between the resonances of the two building blocks. These experimental observations corroborate the existence of a direct resonant energy transfer between AgNP and CuFeS₂ NCs in AgNP@CuFeS₂ hybrids.

Results and Discussion

Assembly and Characterization of AgNP@CuFeS₂ Hybrid Plasmonic Structures

Previous investigations of noble metal NP - CuFeS₂ NCs hybrid structures focused on Au nanorods as central core with a self-assembled lipid layer containing CuFeS₂ NCs.⁵⁶ The lipid coating in this design serves two purposes: 1) it places oleylamine-passivated CuFeS₂ NCs in close vicinity to the NP, and 2) provides a hydrophobic environment that protects the water-sensitive CuFeS₂ NCs from the aqueous environment. The vertical and especially the strong longitudinal plasmon resonance of AuNRs interact, however, only with the low energy tail of the CuFeS₂ NC resonance. To achieve a stronger and more tunable spectral overlap between the resonances of the central noble metal NP core and surrounding CuFeS₂ NCs, we replaced the anisotropic AuNR as core of the hybrid structure with a spherical AgNP of different diameters (Figure 1A). This experimental strategy was chosen to characterize how the coupling between the resonances in AgNP and CuFeS₂ NCs affects the dephasing of the noble metal NP plasmon through single particle spectroscopy.

Figure 1. Characterizing the components of the AgNP@CuFeS₂ NC hybrid structure.

A) Schematic of the AgNP@CuFeS₂ NC hybrid structure. B) UV-Vis spectra of the CuFeS₂ NCs in toluene. The absorbance is plotted as function of energy (blue) or wavelength (red). C) UV-Vis spectra of the CuFeS₂ NCs and lipid-coated 40 nm and 60 nm AgNP. D) HR-TEM of CuFeS₂ NCs. E) TEM image of assembled AgNP(60)@CuFeS₂ hybrid nanostructures that shows small CuFeS₂ NCs around the metal NP core. The scale bars for C and D are 5 nm. F) TEM image of a AgNP-dimer(60)@CuFeS₂. G) The MP-AES measurement for Ag, Cu, Fe concentration (see legend) of AgNP(60)@CuFeS₂, AgNP(60), and CuFeS₂ NCs containing liposomes.

The CuFeS₂ NCs had an average diameter of 3.62 ± 0.65 nm. UV-Vis spectra, a high resolution transmission electron microscopy (HR-TEM) image, and the X-Ray diffraction (XRD) spectrum of CuFeS₂ NCs are provided in Figure 1B–C, Figure 1D, and Figure S1, respectively. In good agreement with prior literature, the CuFeS₂ NC UV-Vis spectrum shows a broad band with a peak at 2.616 eV assigned to the quasistatic resonance of CuFeS₂ NCs (Figure 1B).^30,31 Importantly, the plasmon resonances of lipid-coated 40 nm and 60 nm diameter AgNPs overlap with the CuFeS₂ NC spectrum and lie close to its peak (Figure 1C), confirming the feasibility of hybrid structures with resonantly coupled metal and semiconductor building blocks.

AgNP@CuFeS₂ hybrid nanostructures were assembled in a “one-pot” process by combining a lipid mix containing the CuFeS₂ NCs with colloids of AgNPs with diameters of 40 nm or 60 nm in the presence of 1-octadecane thiol.⁵⁶ The lipid mix comprised dipalmitoyl phosphatidylcholine (DPPC), 1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS), and cholesterol in a molar ratio of 55 : 5 : 40. This strategy led to lipid coated AgNPs (Figure 1E) that contained CuFeS₂ NCs in the lipid coating (Figure 1F). In the case of the hybrid structure with a 60 nm NP core (AgNP(60)@CuFeS₂) in Figure 1F the measured Ag - NC edge-to-edge separations lie between 1.6 nm to 2.2 nm, indicating that the CuFeS₂ NCs were localized close but not in direct contact with the AgNP core. Energy-dispersive X-ray spectroscopy (EDAX) elemental mapping confirmed the enrichment of Fe, Cu, and S elements around AgNPs in AgNP@CuFeS₂ (Figure S4). We further validated successful assembly of AgNP(60)@CuFeS₂ by measuring the Ag, Cu, Fe contents of the hybrid material as well as of controls of the separate building blocks by Microwave Plasma Atomic Emission Spectroscopy (MP-AES). The two controls were AgNP(60) and CuFeS₂ NC containing liposomes at the same concentration as used for the assembly of AgNP(60)@CuFeS₂ and cleaned by centrifugation like the hybrid material. The higher Cu and Fe content in the hybrid assemblies confirms the successful integration of the two components into one nanostructure. We generated AgNP@CuFeS₂ with four different CuFeS₂ NCs input concentrations, resulting in average loadings between 29 and 284 CuFeS₂ NCs per NP for a 60 nm AgNP core and between 26 and 51 CuFeS₂ NCs per NP for a 40nm AgNP core, as determined by MP-AES (Table 1). The AgNP@CuFeS₂ hybrid nanostructures were stable in aqueous solution for at least a week when stored at 4°C.

Table 1.

Summary of CuFeS₂ NCs bound per NP, surface area per NC, resonance frequency (${\omega }_{\mathit{\text{res}}}$) and excitation transfer damping term, ${\Gamma }_{\mathit{\text{ET}}}$, for all investigated samples. Standard deviations were determined from at least n = 50 measurements.

Sample	#NCs	${\mathit{\text{nm}}}^2/\mathit{\text{NC}}$	ωres (eV)	${\Gamma }_{\mathit{\text{ET}}}\left(\text{eV}\right)$
AgNP-Lipid (40)	N.A.	N.A.	2.677±0.057	0.132±0.045
AgNP(40)@CuFeS2	26	196.97	2.557±0.041	0.409±0.087
AgNP(40)@CuFeS2	51	97.83	2.533±0.029	0.455±0.059
AgNP-Lipid (60)	N.A.	N.A.	2.583±0.050	0.134±0.034
AgNP(60)@CuFeS2	29	386.74	2.539±0.045	0.193±0.033
AgNP(60)@CuFeS2	79	143.14	2.412±0.045	0.287±0.039
AgNP(60)@CuFeS2	168	67.40	2.395±0.045	0.331±0.055
AgNP(60)@CuFeS2	284	39.79	2.353±0.051	0.359±0.084

Single Particle Spectroscopy and Excitation Transfer Analysis

Excitation transfer between the AgNP core and surrounding CuFeS2 NCs in the investigated AgNP@CuFeS2 hybrid structures through plasmon relaxation mechanisms such as PIRET and PICTT is related to chemical interface damping (CID), which was previously studied through linewidth analysis of single particle spectra.^57,58 When modeling single particle spectral linewidths, the contribution of different plasmon decay processes to the total linewidth $\Gamma$ need to be considered, and our analysis follows here the strategy outlined by Foerster et al.⁵⁹ The bulk damping, ${\Gamma }_0={\gamma }_b$, captures plasmon damping due to collisions of oscillating electrons with core electrons, ions or phonons in the metal and can be obtained by fitting a free electron Drude model to the Johnson and Christy Ag dielectric function in the IR spectral range, where damping through interband transitions is negligible.⁶⁰ This approach yields γb = 0.0212 eV.⁶⁰ Once the bulk damping has been determined, the contribution of interband transitions to the imaginary component of the metal dielectric function, ${\epsilon }_{\mathit{\text{IB}}}$, can be obtained by subtracting the Drude dielectric function from the measured Johnson Christy dielectric function (${\epsilon }_{\mathit{\text{JC}}}$):^59,60

$${\epsilon }_{\mathit{\text{IB}}}=\mathit{\text{Im}}\left({\epsilon }_{\mathit{\text{JC}}}\right)-\frac{{\gamma }_b{{\omega }_p}^2}{{\omega }^3+\omega {{\gamma }_b}^2}$$ (1)

where ω is the frequency and ${\omega }_p$ the plasma frequency. After determining ${\epsilon }_{\mathit{\text{IB}}}$ the contribution from interband damping is obtained as:⁵⁹

$${\Gamma }_{\mathit{\text{IB}}}={\epsilon }_{\mathit{\text{IB}}}\times \frac{\mathrm{ħ}{\omega }^3}{{{\omega }_p}^2}$$ (2)

${\Gamma }_{\mathit{\text{IB}}}\left(\omega \right)$ is plotted in Figure S7. Due to the small size of the investigated NP scatterers, electron scattering at the NP surface, ${\Gamma }_{\mathit{\text{surf}}}$, is another relevant plasmon damping mechanism that needs to be considered:^61,62

$${\Gamma }_{\mathit{\text{surf}}}=\alpha \frac{\mathrm{ħ}{v}_f}{R}$$ (3)

Here, $\alpha$ is surface scattering parameter whose lower limit for silver NPs has been determined as 0.7 by Kreibig and co-workers,^62,63 $v_f$ is the Fermi velocity and R is the NP radius. The surface scattering damping values for 40 nm and 60 nm diameter AgNPs are ${\Gamma }_{\mathit{\text{surf}}}=$ 0.032 eV and 0.022 eV, respectively. Another contribution to plasmon dephasing in NPs is the volume V dependent radiation damping:⁵⁹

$${\Gamma }_{\mathit{\text{rad}}}=h{\kappa }_{\mathit{\text{Rad}}}V$$ (4)

We determined the radiation damping constant, ${\kappa }_{\mathit{\text{rad}}}$,^63,64 from $\Gamma$ measurements of AgNPs of different sizes under consideration of their respective ${\Gamma }_{\mathit{\text{surf}}}$ values as 2.847×10⁻⁷ fs⁻¹nm⁻¹. In AgNP@CuFeS₂ hybrids, excitation transfer between AgNP and CuFeS₂ NC provides a fifth contribution to the total damping, which we assign here to the excitation transfer term ${\Gamma }_{\mathit{\text{ET}}}$. Due to the spacing between NCs and metal NPs in AgNP@CuFeS₂ hybrids, we anticipate that the contribution from direct energy transfer to ${\Gamma }_{\mathit{\text{ET}}}$ outweighs that from direct charge transfer. Furthermore, as the AgNP in the investigated hybrid materials are surrounded both by NCs and a self-assembled molecular shell, it is possible that ${\Gamma }_{\mathit{\text{ET}}}$ is not exclusively determined by resonant interactions between AgNP and NCs but that CID induced by the molecular shell may contribute as well. The relative contributions of both mechanisms can, however, be distinguished by comparing ${\Gamma }_{\mathit{\text{ET}}}$ for lipid-coated AgNPs with and without CuFeS₂ NCs.

${\Gamma }_{\mathit{\text{ET}}}$ is a priori unknown but can be determined from the experimentally derived natural linewidth $\Gamma$ of the scattering spectra of individual AgNP@CuFeS₂ particles with the outlined estimates of ${\Gamma }_0,{\Gamma }_{\mathit{\text{IB}}},{\Gamma }_{\mathit{\text{rad}}}$ and ${\Gamma }_{\mathit{\text{surf}}}$ through the relation:

$${\Gamma =\Gamma }_0+{\Gamma }_{\mathit{\text{IB}}}+{\Gamma }_{\mathit{\text{rad}}}+{\Gamma }_{\mathit{\text{surf}}}+{\Gamma }_{\mathit{\text{ET}}}$$ (5)

Figure 2A–D shows representative single particle scattering spectra $I\left(\omega \right)$ of AgNP@CuFeS₂ hybrids. $\Gamma$was determined from Lorentzian fits to these spectra:

$$I\left(\omega \right)=\frac{I_{\mathit{\text{max}}}*{\Gamma }^2}{4{\left(\omega -{\omega }_{\mathit{\text{res}}}\right)}^2+{\Gamma }^2}$$ (6)

Figure 2.

Single particle scattering spectra (blue dots) and Lorentzian fits (red line) for A) lipid-coated 60 nm AgNP, B) lipid-coated 40 nm AgNP, C) AgNP(60)@CuFeS₂ (#NCs=168), D) AgNP(40)@CuFeS₂ (#NCs=51).

The Lorentzian fits, which are included as solid lines in the spectra, provide an accurate description of the lineshape in the range between approx. 2.0 eV to 2.8 eV. Discrepancies between experimental data and fit at higher energies arise from a decrease in the diffraction efficiency of the grating whose center wavelength was fixed at 550 nm. Spectra with complex line shapes that could not be fit by a single Lorentzian, indicative of larger NPs or clusters, and energetic outliers were removed from the analysis. ${\Gamma }_{\mathit{\text{ET}}}$ was determined for all remaining individual NPs as outlined above. Figure 3 A, B summarizes this data and plots ${\Gamma }_{\mathit{\text{ET}}}$ as function of peak plasmon energy (ω_res) for the investigated AgNP@CuFeS₂ hybrids with 40 nm or 60 nm AgNP cores and specified CuFeS₂ NC loading (CuFeS₂ NCs per AgNP). AgNPs with lipid coating but without CuFeS₂ NCs were included, as well.

Figure 3.

A) Excitation transfer damping (${\Gamma }_{\mathit{\text{ET}}})$ for AgNP@CuFeS₂ with 60 nm AgNP core and for lipid-coated 60 nm AgNP as function of resonance frequency ω_res. B) ${\Gamma }_{\mathit{\text{ET}}}$ for AgNP@CuFeS₂ with 40 nm AgNP core and lipid-coated 40 nm AgNP as function of resonance frequency ω_res. The CuFeS₂ NC concentration per AgNP are specified. Error bars represent standard deviations of at least n = 50 measurements.

Although the single NP data show some spread, two general trends are distinguishable in the data: for both AgNP sizes the addition of CuFeS₂ NCs results in a red-shift of the plasmon resonance and an increase of ${\Gamma }_{\mathit{\text{ET}}}$. The average ${\Gamma }_{\mathit{\text{ET}}}$ values and the corresponding resonance energies for the different experimental conditions are summarized in Table 1. The summary reveals that the Γ_ET values of AgNP@CuFeS₂ are systematically higher than those of the lipid-coated AgNP controls, which confirms that damping related to excitation transfer between AgNP and CuFeS₂ NCs outweighs the contribution of CID associated with the molecular shell under the chosen experimental conditions. Only for AgNP(60)@CuFeS₂ (#NCs = 29) the measured Γ_ET value approaches that of lipid-coated AgNP(60), presumably due to the low concentration of NCs in this hybrid architecture.

The localization of resonant CuFeS₂ NCs in the near-field of the AgNP provides opportunities for energy transfer from the AgNP to the semiconductor NCs, but one may expect a potential broadening of the resonance even if non-resonant dielectric NPs are localized in the evanescent field of the AgNP due to refractive index changes. To evaluate the relative magnitude of the effects, we performed T matrix calculations for hybrid structures that contain either CuFeS₂ NCs or hypothetical dielectric NCs of identical size localized around a central 60 nm diameter AgNP core. In these simulations, we assumed an increasing number (4, 8, 12) of 10 nm diameter CuFeS₂ or dielectric (n_r = 3.000) NCs around a central NP core with an edge-to-edge separation (NP to NC) of 2 nm. In case of the CuFeS₂, we observe a stepwise increase in total linewidth ($\Gamma$) with the addition of 4, 8, and 12 NCs, reaching $\Gamma$ values of 0.3888 eV, 0.4044 eV, and 0.4199 eV, respectively. Replacing the CuFeS₂ NCs with dielectric NCs yielded nearly constant $\Gamma$ values (0.3727 eV, 0.3725 eV and 0.3715 eV). Even a large change in the ambient refractive index from n_r = 1.330 to n_r = 2.000, which shifts the peak of the plasmon resonance by 0.750 eV, had only a relatively small effect on the width of the plasmon resonance of $\Delta \Gamma =0.0810\mathit{\text{eV}}$. The spectral broadening observed for CuFeS₂ NCs is much larger than typical refractive index mediated effects.

For AgNP@CuFeS₂ hybrids with either 40 nm and 60 nm AgNP core, ${\Gamma }_{\mathit{\text{ET}}}$ increase with decreasing resonance energy in Figure 3. However, the slope differs for the two different NP sizes. Intriguingly, a plot of all ${\Gamma }_{\mathit{\text{ET}}}\left(\omega \right)$ data in Figure 4 indicates a maximum in the ${\Gamma }_{\mathit{\text{ET}}}\left(\omega \right)$ distribution at an intermediate resonance energy of approx. 2.469 eV, which coincides with the peak of the simulated extinction spectrum of a 5 nm CuFeS₂ NC in a membrane that is included as red line in Figure 4. The ${\Gamma }_{\mathit{\text{ET}}}\left(\omega \right)$ single particle data follows the general shape of the NC extinction spectrum with a Pearson correlation coefficient of 0.80. The strong positive correlation between the measured ${\Gamma }_{\mathit{\text{ET}}}\left(\omega \right)$ distribution and the envelope of the CuFeS₂ NC extinction spectrum indicates coupling between the overlapping resonances of the AgNP core and the surrounding CuFeS₂ NCs and is consistent with the PIRET energy transfer mechanism. The plasmonic field of the optically excited AgNPs excites the quasistatic resonance of the NCs in the near-field and dissipates energy in this process, resulting in a broadening of the AgNP spectra.

Figure 4.

Γ_ET (blue points) for all AgNP@CuFeS₂ conditions is plotted as function of the resonance frequency ω_res. A polynomial interpolation is included as black line. The red line shows the simulated scattering spectrum of an individual CuFeS₂ NC (n_r = 1.5000).

To further validate the presence of a resonant energy transfer between AgNP and CuFeS₂ NCs, we next determined ${\Gamma }_{\mathit{\text{ET}}}$ for silver nanostructures with plasmon resonances that show little to no overlap with the CuFeS₂ NC resonance. Since the spectral tunability of spherical AgNPs is limited due to the relatively weak size-dependence of the LSPR, we determined ${\Gamma }_{\mathit{\text{ET}}}$ of dimers of AgNPs that are decorated with CuFeS₂ NCs through correlated single particle scattering spectroscopy – scanning electron microscopy (SEM) measurements (Figure 5A). To that end, a silicon-patterned glass substrate was used to facilitate a precise identification of dimers observed in darkfield microscopy in the SEM. This approach allowed for a selective measurement of the scattering spectra of CuFeS₂ NCs decorated AgNP dimers (AgNP-dimer(60)@CuFeS₂). The examination of dimers through this combined spectroscopic and microscopic approach revealed an average longitudinal plasmon resonance energy of 1.989 eV ± 0.036 (n=25) for the CuFeS₂ NC coated AgNP dimers (Figure 5B and S8). The emergence of a longitudinal plasmon resonance in the AgNP dimer that lies lower in energy than the CuFeS₂ NC resonance at 2.580 eV is accompanied by a significant reduction in ${\Gamma }_{\mathit{\text{ET}}}$. Our lineshape analysis reveals a drop of ${\Gamma }_{\mathit{\text{ET}}}$ to 0.032 eV ± 0.023 (Figures 5C and S10). In this analysis we evaluated ${\Gamma }_{\mathit{\text{rad}}}$for the 60 nm AgNP dimers based on eq. 4 with ${\kappa }_{\mathit{\text{rad}}}=1.373\times {10}^{-7}{\text{fs}}^{\text{-1}}{\text{nm}}^{\text{-1}}$ determined from electromagnetic simulations of 1 nm separation dimers with different NP sizes, and we used the dimer volume to evaluate ${\Gamma }_{\mathit{\text{rad}}}$ and ${\Gamma }_{\mathit{\text{surf}}}$. The striking difference in ${\Gamma }_{\mathit{\text{ET}}}$ between hybrid structures with a AgNP or a AgNP dimer core corroborates the hypothesis that coupling between energetically overlapping resonances in AgNP monomers and CuFeS₂ NCs accounts for the maximum in ${\Gamma }_{\mathit{\text{ET}}}\left(\omega \right)$ observed for monomeric AgNP@CuFeS₂ hybrids at ω = 2.470 eV. This conclusion is also supported by another set of control experiments, in which we replaced CuFeS₂ NCs in AgNP(60)@CuFeS₂ with 5 nm AgNPs whose resonance is shifted to a higher energy than that of the 60 nm AgNP (Figure S9A and B). Also in this case the spectral detuning resulted in a much smaller Γ_ET (Figure S9C). Overall, the spectral lineshape analysis of assemblies with different degrees of energetic overlap of the individual components supports a spectral broadening in AgNP@CuFeS₂ hybrids that is dominated by a resonant energy transfer from the central AgNP to surrounding CuFeS₂ NCs.

Figure 5.

A) An Optical darkfield microscope image (inset) and a low-resolution SEM micrograph were correlated to identify individual scatterers and localize hybrid structures with a AgNP dimer core. Ⅰ-Ⅴ mark areas for which magnified SEM images were included. Both red and green lines highlight the correlation between optical and SEM images. B) Scattering spectrum for dimer shown in inset. Blue: experimental spectrum; pink and green dash: deconvoluted Lorentzians of the plasmon resonances at 1.96eV and 2.48eV; red line: sum of the two Lorentzians. The scale bar in the inset is 50 nm. C) ${\Gamma }_{\mathit{\text{ET}}}$ of lipid coated 60 nm AgNP and AgNP-dimer(60)@CuFeS₂ hybrids as function of resonance frequency ω_res.

Near-field Simulations of AgNP@CuFeS₂ Hybrids

We simulated the E-field intensity for a dimer of a 60 nm AgNP and 10 nm CuFeS₂ NC as function of interparticle separation. The dimer axis was oriented along the x-axis, and the E-field intensity was calculated in the xy plane for a x-polarized plane wave propagating along z. Figure 6 contains E-field intensity maps at λ = 440 nm for separations between 1 nm and 5 nm in steps of 1 nm, as well as for the separation of 10 nm. The maps show a strong localization of E-field intensity in the gap between the NP and the NC that decreases with separation. Interestingly, the peak of the E-field intensity is localized closer to the smaller CuFeS₂ NC than to the AgNP surface at short separations. This behavior suggests that the smaller NC acts as a nanolens for the E-field provided by the larger AgNP. Similar effects have previously been observed for self-similar all-metallic plasmonic antennas.^65,66 One distinct feature of the AgNP@CuFeS₂ hybrid is that despite significant differences in the size of the 60 nm AgNPs and the few nm sized CuFeS₂ NCs the resonances of both components lie close in energy, allowing for an efficient excitation of the CuFeS₂ NC resonance through the near-field of the AgNP, ultimately resulting in the formation of an electromagnetic hot-spot on the CuFeS₂ NC surface. At larger separations, the coupling between AgNP and CuFeS₂ NC is weak, and the maximum E-field intensity is localized close to the AgNP surface.

Figure 6.

E-field intensity maps for AgNP-dimer(60)@CuFeS₂ as function of interparticle separation: A) - E) show separations between 1 nm - 5 nm in 1 nm increments. In F) the gap width is 10 nm.

Linear Sweep Voltammetry Confirms Enhanced Photocatalytic Efficacy of AgNP@CuFeS₂ Hybrids

A resonant energy transfer from AgNP to CuFeS₂ NC as indicated by our single particle spectroscopy would result in the formation of excited electron-hole pairs in the semiconductor that are available for photocatalysis. To validate the formation of reactive charge carriers and to probe for a gain in catalytic efficacy by incorporating 60 nm AgNPs and CuFeS₂ NCs into AgNP@CuFeS₂ hybrids, we performed linear sweep voltammetry for the hydrogen evolution reaction using glassy carbon electrodes functionalized with A) CuFeS₂ NCs, B) AgNP, and C) AgNP@CuFeS₂ hybrids (Figure 7) both with and without illumination at 470 nm. The comparison of the different conditions shows a clear increase in photocurrent for AgNP@CuFeS₂ hybrids, confirming an increase in reactive charge carrier formation and indicating a synergistic gain in photocatalysis. The localization of the CuFeS₂ NCs in the lipid coating around the NP can affect the accessibility of the reactive sites. While the thin lipid coating does not impede a measurable catalytic gain for the hydrogen evolution reaction, for other reactions the barrier may be more significant. It is even conceivable that the lipid coating could be used to increase the selectivity of the photocatalyst.

Figure 7.

Linear Sweep Voltammetry (LSV) curves (4^th scan) obtained with and without 470nm LED illumination for (A) CuFeS₂ NCs, (B) AgNP(60)@lipid, (C) AgNPs(60)@CuFeS₂ in acidic aqueous solution.

Conclusion

In summary, we have characterized the interfacial excitation transfer between metallic and semiconductor building blocks in AgNP@CuFeS₂ hybrids through single particle spectroscopy and subsequent lineshape analysis in combination with electromagnetic simulations. This approach allowed for quantitative analysis of the effect that CuFeS₂ NCs localized in the evanescent field of the AgNP have on the damping of the metal plasmon. The damping was found to increase with the number of NCs in the hybrid structure as well as with the energetic overlap between CuFeS₂ NC and AgNP resonances, indicative of a resonant, direct energy transfer from AgNP to CuFeS₂ NCs. The resonant coupling and associated energy transfer in AgNP@CuFeS₂ hybrids could potentially lead to an increase in long-lived electron-hole pair generation when compared to the separated metallic and semiconductor building blocks and has thus potential to overcome a longstanding bottleneck in conventional plasmonic photocatalysis. Its applicability to enhance the efficiency of plasmonic photocatalysis warrants further systematic investigation.

Experimental section

Synthesis of CuFeS₂.

CuFeS₂ nanocrystals were synthesized following previously described procedures through a hot injection method in an air-free environment.^31,56 0.5 mmol Copper(II) acetylacetonate (Cu(acac)₂, trace metal grade), 0.5 mmol of Iron(III) acetylacetonate (Fe(acac)₃, 97%) and 6.65 mL of Oleic Acid (OA, 90%, technical grade) were first added to a 100 mL round bottom flask in an argon filled glovebox. The mixture was attached to a Schlenk line and heated to 120°C under vacuum for 30 minutes until all powder had been dissolved. The flask was then filled with argon and heated to 180°C. In a separate flask, 0.2 M sulfur (99.8%, trace metal grade) in oleylamine (OLAM, 70%, technical grade) (S/OLAM) was dissolved at 80°C under argon for 30 minutes. Once the reaction flask reached 180°C, 2 mL of 1-dodecanethiol was rapidly injected into the flask, following the initial injection, 5 mL of S/OLAM mixture was injected over ~30 sec. The reaction was allowed to proceed for 3 minutes at 180°C before cooling to room temperature. The resulting solution was transferred to argon-filled glove box for long-term storage. All chemicals were purchased from Millipore Sigma.

Characterization of CuFeS₂ NCs.

HRTEM images of CuFeS₂ NCs were recorded with a Tecnai Osiris TEM (FEI, USA) using a 300 kV electron beam. To prepare the samples, CuFeS₂ NCs were washed with hexane/ethanol mixture multiple times before resuspension in hexane to remove excess ligands. Then, the NCs in hexane were dropcast onto ultrathin carbon film-coated copper TEM grids. After complete evaporation of hexane, the grids were washed with drops of acetone and then ethanol before overnight storage in a dry box.

X-ray diffraction (XRD) measurements were performed by dropcasting the CuFeS₂ NCs in hexane onto a zero-background silicon sample holder for X-ray diffraction (XRD) measurements. A Burker D2 Phaser analyzer with a coupled theta-2 theta scan was used to record the diffraction profiles of the particles. XRD data were analyzed with an open-source software Fityk and the reference lines of chalcopyrite (JCPDS# 00–035-0752) were obtained from the International Centre for Diffraction Data (ICDD) database.

Preparation of Hybrid Plasmonic Structures.

AgNP@CuFeS₂ nanostructures were prepared through a modified one-pot self-assembly method described previously.⁵⁶ 50 mol % DPPC, 5 mol % DOPS, and 45 mol % cholesterol solutions in chloroform (Avanti Polar Lipids) were mixed with NCs chloroform suspension. The mixture was then rotary evaporated at 32°C and desiccated overnight. 2 mL of water was added to the round bottom flask containing the dehydrated lipid film and then bath sonicated for 5 min to form CuFeS₂ NCs containing liposomes. Subsequently, 160 μL of 2 mg/mL octadecanethiol (ODT, Sigma-Aldrich) ethanol solution and 1 mL of citrate stabilized AgNPs (10¹⁰ particle/mL) (nanoComposix, NanoXact Silver Nanospheres) were added. The solution was incubated overnight to allow the formation of AgNP@CuFeS₂ hybrid nanostructures, washed by repeated centrifugation and resuspension, and finally stored at 4 °C.

Si Patterned Glass.

High-precision microscope cover glasses (Deckgläser) were plasma cleaned in a Plasma Asher (PVA TePla America M4L). Then, an 80nm thick Si layer was deposited using a CHA Industries Solution Process Development System and silicon (Si (P-type)) Pieces Evaporation Materials (Kurt J. Lesker Company). Three slot TEM grids (Tedpella Inc.) were placed on the coated glass slide, followed by CF₄/O₂ etching (Plasma-Therm 790).

Sample Preparation for Single Particle Spectroscopy.

Si patterned glass or conventional glass slides were coated with Poly-L-lysine 0.1 % (w/v) in H₂O (Sigma-Aldrich) for 15 minutes. After washing with H₂O, 200 μL of hybrid plasmonic structures solution was dropcast and incubated for 5 minutes. A PDMS spacer, made by heating PDMS (Sigma-Aldrich) and sylgard 184 silicone elastomer (Fisher Scientific) to 80°C, was placed on the glass slide to generate a simple chamber. 40μL of water was pipetted onto the glass slide before the chamber was sealed with a second glass slide.

Darkfield Measurements.

Optical measurements were performed with an Olympus Inverted Microscope under darkfield illumination using a 100 W Tungsten lamp. The scattered light was collected with a 60x oil objective (NA = 0.65) and spectrally analyzed with a spectrometer (Andor SR-500I-A) and a back-illuminated CCD camera (Andor DU401-BR-DD). The setup is shown in Figure S6. To correlate the optical measurements with SEM analysis, the top glass slides and polydimethylsiloxane (PDMS) spacers were removed carefully from the bottom glass slides. After removal of residual water with a nitrogen stream, a thin Au/Pd film was sputter coated onto the sample (Cressington 108). SEM images were obtained on the Zeiss Supra 55 VP SEM at 5 kV.

Structural Characterization.

Transmission Electron Microscopy (TEM) was performed on a JEOL F200 TEM at 200kV. A high-angle annular dark-field detector was used in Scanning Transmission Electron Microscopy (STEM) mode, and a JEOL EDAX detector with Oxford instruments Aztec software was used for Energy Dispersive X-ray spectroscopy (EDAX). Samples were prepared on nickel/carbon TEM grids (Tedpella Inc.). Solutions of AgNP-NCs hybrid structures were dropcast onto the grids and incubated for 30 minutes in a H₂O-saturated environment before the solution was removed with a Kimwipe.

Microwave Plasma Atomic Emission Spectroscopy (MP-AES) Measurements.

For the MP-AES, the particles were washed twice with deionized water and finally redispersed in 1 mL of deionized water. All of the samples were then added to a 12 well plate and incubated overnight at 60 C in aqua regia. After the incubation, the ions were redissolved in 1 mL of 2% HNO₃ and the sample were measured in Agilent 4210 MP-AES. All the standards from 10 ppb to 10000 ppb are also made in 2% HNO₃.

Photoelectrochemical Characterization.

Linear Sweep Voltammetry (LSV) curves were recorded with a potentiostat (Gamry Instruments) using a three-electrode system: a 3 mm diameter Glassy Carbon (GC) working electrode, a standard Ag/AgCl reference electrode and a platinum wire counter electrode. All electrodes were purchased from BASI Analytical Instruments. 40 μL water suspension from 4 °C storage of the nanocomposites or control groups were dropcast onto the GC electrode and dried overnight in the dark. The LSV curves were collected in an acid electrolyte with 1 M HCl. A scan rate of 50 mV/s and a collection rate of 10 Hz was used. During the measurement, a collimated 470 nm LED (Thorlabs) with a power density of 6.6 mW mm⁻² was focused onto the GC working electrode.

Electromagnetic Simulations.

Unless otherwise noted, simulations of the far-field scattering spectra of AgNPs, AgNP@CuFeS₂ hybrids, AuNP and AuNP@CuFeS₂ hybrids were performed with the T-matrix for Electromagnetic Radiation with Multiple Scatterers (TERMS) package⁶⁷ using an effective refractive index of n_r = 1.3891 for the glass / water interface.^68,69 The dielectric functions for Ag and Au were used as implemented and for CuFeS₂ NCs the dielectric function was taken from Gaspari et al.³⁰ (Table S3). Finite-Difference Time-Domain (FDTD) simulation of CuFeS₂ were performed with the Ansys Lumerical software package.⁷⁰ A total-field scattered-field source with a 0.5 nm mesh was used. Near-field simulations of 60 nm AgNPs and 10 nm CuFeS₂ were performed with the TERMS package⁶⁷ at 440 nm with linear polarization with an effective refractive index of n_r = 1.3891 for the glass/water interface. The wavelength is determined by the resonance peak from far-field scattering spectra.

Supplementary Material

Additional methods for data analysis and outlier removal, further characterization of single CuFeS₂ NCs and hybrid structure, simulation data for AgNP@CuFeS₂ nanostructures, simulation spectrum for AgNP(40/60) and CuFeS₂, description of the darkfield set-up, interband damping plot for AgNP, more example for dimer Lorentzian fitting, insight of AgNP(60)@AgNP(5) hybrid structures, overview for all the datapoints in the experiment and dielectric data for CuFeS₂ NCs.