Andreas
Wolf
,
Torben
Kodanek
and
Dirk
Dorfs
*
Institute of Physical Chemistry and Electrochemistry, Leibniz Universität Hannover, Callinstr. 3A, 30167 Hannover, Germany. E-mail: dirk.dorfs@pci.uni-hannover.de
First published on 29th October 2015
Localized surface plasmon resonances (LSPRs) of degenerately doped copper chalcogenide nanoparticles (NPs) (Cu2−xSe berzelianite and Cu1.1S covellite) have been modified applying different methods. The comparison of the cation exchange (Cu2−xSe) and intercalation (Cu1.1S) of Ag(I) and Cu(I) has shown that Ag(I) causes a non reversible, air stable shift of the LSPR. This was compared to the influence of Au(I) cation exchange into Cu1.1S platelets under the formation of Cu1.1S–Au2S mixed nanoplatelets. Furthermore, we show the growth of Au domains on Cu2−xSe, and discuss the interaction of the two plasmonic parts of the obtained dual plasmonic Cu2−xSe–Au hybrid particles.
In this article we analyze different approaches to modify and stabilize the LSPR of copper chalcogenide NCs in an attempt to increase their controlled flexibility for future applications: ion intercalation, ion exchange and metal growth. First, we explore the LSPR response of Cu1.1S vs. Cu2−xSe NCs to the addition of univalent ions. We compare the different behavior of the Cu2−xSe LSPR and the Cu1.1S LSPR, specifically the different shifting, damping and temporal stability of the LSPR after the integration of Cu(I) ions vs. Ag(I) ions (Scheme 1). Second, we present to the best of our knowledge for the first time growth of Au domains on pre-synthesized Cu2−xSe NPs (Scheme 1). The few previous reports on Au–Cu2−xSe28 and Au–Cu2−xS22,29–34 heterodimers have been either growing the copper chalcogenide onto a readily existing gold seed22,32–34 or by first synthesizing Au–Cu29–31,35 alloy NPs, which were subsequently converted by sulfidation to Au–Cu2−xS. The here presented synthesis approach enables us to grow Au on readily existing berzelianite Cu2−xSe NPs resulting in dual-plasmonic hybrid NPs. This synthesis approach enables us to analyze the change of the chalcogenide LSPR through the interaction with differently sized metal domains on their surface. The same approach, applied to Cu1.1S NPs, results in a cation exchange to Au2S–Cu1.1S platelet shaped hybrid particles (Scheme 1).
![]() | ||
Scheme 1 Overview of the presented approaches for tuning the LSPR in spherical Cu2−xSe NPs and Cu1.1S nanoplatelets. |
Previous reports have shown that for Cu2−xSe and Cu1.1S a bathochromic shift accompanied by a decrease of the oscillator strength of the LSPR can be detected after the addition of Cu(I), that finally leads to a complete disappearance of the LSPR for higher Cu(I) amounts.1,6 This was shown to be reversible for Cu2−xSe.1 It was also shown that during the oxidation of Cu2Se to Cu2−xSe the cubic berzelianite crystal structure does not change.1 For the Cu1.1S covellite system instead, a structural change due to the Cu(I) intercalation was reported; first to a probably metastable phase and for Cu2S to either a metastable phase or a mixture of phases.6
Adding Cu(I) ions to both systems leads to the same optical response (bathochromic shift and intensity decrease of the LSPR) and ultimately to the complete disappearance of the plasmon band (Fig. 1A/C). Instead, the systems behave contrary to each other when adding an Ag(I) containing solution (Fig. 1B/D). The Cu1.1S NPs, whose LSPR have a maximum at 1230 nm, show a decrease of the oscillator strength of the LSPR, which is accompanied by a bathochromic shift until it completely disappears when adding higher Ag(I) amounts. This is similar to the addition of Cu(I). When exposed to air, however, the Ag(I) induced shift is almost irreversible (Fig. 1D) which might be interesting for applications that require specific and stable LSPRs in the NIR. This is remarkable as a Cu(I) induced shift of the LSPR maximum position is reversible (Fig. 1C) when the sample is exposed to air. Cu2−xSe NPs treated with Ag(I) show also a damping of the LSPR intensity, which is, however, not accompanied by a shift to longer wavelength. In this system we rather observe a slight hypsochromic shift of the LSPR maximum (Fig. 1B). Adding 25% Ag(I) ions in relation to the determined copper contents of the NPs leads to a LSPR shift from 1155 nm to 1130 nm after 2 h reaction time. The LSPR shifts and damping stabilizes after a few hours at around 1090 nm. Additionally this shift and the damping are non-reversible under air exposure, which is in contrast to the shift after Cu(I) addition (Fig. 1A). The bathochromic shift for Cu2−xSe after the Cu(I) addition is based on the reduction of the copper deficiencies to zero, resulting in a non plasmonic Cu2Se system.
Literature lately discusses the stoichiometric changes during reduction as an incorporation of Cu(I) combined with a reduction of the chalcogenide, while a fraction of the Cu(I) is oxidized in solution to Cu(II) in order to provide the required electrons. In fact, the Manna group showed that no Cu(II) can be detected inside copper chalcogenide NPs before or after the reduction.6,11
The LSPR frequency of both self-doped systems can be described by the Drude model.4 From eqn (S5) in the ESI† it can be seen, that at a given plasma frequency the increase of the damping factor (γ) leads to an increase of the imaginary part of the dielectric function and hence, according to eqn (S3) in the ESI† to a decrease of the extinction cross section. Following this we suggest, that the reduction of the LSPR band in Cu2−xSe through the Ag(I) addition is caused by a continuous increase of the damping in the system. This damping could be caused by Ag(I) ions that serve as charge carrier scattering centers. An increase of the silver content would hence cause an increased damping. This would cause a decrease of the extinction cross section. As the carrier density does not change significantly it is likely that Ag(I) preferentially exchanges Cu(I) instead of intercalating, as this would not significantly influence the hole density. This is supported by the significant increase of the amount of small Cu2O NPs that can be seen in the TEM (see Fig. 2A) and HAADF-STEM (see Fig. S1 in the ESI†) images in comparison to samples of the seed particles.
![]() | ||
Fig. 2 TEM images and EDX mappings of Ag, Se respectively S and Cu averaged with the corresponding HAADF-STEM images, of Cu2−xSe (A) and Cu1.1S (B) NPs after addition of 25% Ag(I). |
Comparing γ of the Cu2−xSe system, extracted as the full width half maximum from the absorbance plotted against the energy (eV) (see Fig. S2 in the ESI†), it can be seen that γ increases from 0.43 eV (0% Ag) to 0.53 (12.5% Ag) to 0.57 eV (25% Ag). In contrast to this γ stays constant within the margin of error when Cu(I) ions are added, from 0.43 eV (0% Cu–I–) to 0.42 eV (12.5% & 25% Ag) to 0.41 eV (50% Ag). The increase in γ supports the theory that incorporated Ag(I) ions act as charge carrier scattering centers for the plasmon resonance.
STEM-EDX analysis of a Cu2−xSe sample treated with 25% Ag(I) ions relative to the copper content shows that silver is evenly distributed over a whole NP and does not form Ag2Se islands (Fig. 2A). This behavior is supported by literature that shows ternary mixed silver–copper–selenide compounds.38 Additionally we can also exclude the formation of a Ag2Se shell, as this would have led to a strong bathochromic shift of the plasmon band, due to the high frequency dielectric constant of Ag2Se (ε∞ = 11 ± 1 (ref. 39)). This finding is different to the formation of Janus-like NPs that was recently reported for the exchange with divalent cations in Cu2−xSe.40 However some EDX mappings of the same sample showed variations in the silver content between different particles (see Fig. S1 in the ESI†). Combining these findings two different mechanisms for the LSPR damping in Cu2−xSe are possible. One theory is that Ag(I) is forming charge carrier scattering centers in the NPs and hence would increase γ, which would lead to a decrease in the plasmon band intensity. The second theory is that Cu(I) is exchanged to Ag(I) preferentially in some particles and in such a high ratio that they lose their LSPR completely. Hence, in that case the weakened LSPR would be caused by Cu2−xSe NPs that do not show any Ag(I) incorporation after a total addition of 25% Ag(I).
The Cu1.1S system shows a behavior very different to the above described Cu2−xSe when Ag(I) is added (Fig. 1D). While the LSPR intensity is decreasing it additionally shifts strongly to longer wavelengths. This behavior is quite similar to the one upon addition of Cu(I) (Fig. 1C), however not reversible upon O2 exposure. This correlates with a decrease in the charge carrier density (hole density) in the NPs. γ does not increase with the addition of Ag(I) and Cu(I). Hence neither of these ions incorporation does create new charge carrier scattering centers.
Elemental mapping of the Cu1.1S NPs treated with Ag(I), does not show island or shell formation (Fig. 2B). Again the silver seems to be homogeneously distributed over each particle, and also no significant variations of the silver content between different particles can be detected.
In the above mentioned analysis it could be shown that the two copper chalcogenide systems behave optically quite opposed. The reason for this could be the different crystal systems. Cu1.1S is a hexagonal system with layers, where the reduction of the sulfur in the NCs is accompanied by the intercalation of Cu(I) between the layers of the original covellite lattice under braking of S–S bonds. At high intercalation levels a structural change to a metastable phase or a mixture between an orthorhombic and a monoclinic phase occurs.6 We suggest a similar intercalation of Ag(I) ions into the pristine covellite (00-006-0464) starting NPs. The XRD patterns show the continuous formation of a second phase due to the intercalation, the orthorhombic AgCuS (00-044-1436) (Fig. 3B). Upon further Ag(I) addition a third body-centered tetragonal Ag3CuS2 (00-012-0207) phase starts to form. This is in line with the optical observations, as this would cause a continuous shift due to an increase of the cation content in the NPs and hence a decrease of the charge carrier density. The better LSPR stability when exposed to oxygen further shows that the incorporation of the Ag(I) is not reversible under oxygen exposure, in contrast to added Cu(I) ions. This is most likely due to more oxygen stable Cu(I)–Ag(I)–S phases in comparison to the metastable Cu(I)–S phases that occur during the Cu(I) intercalation.
The XRD patterns of the cubic berzelianite Cu2−xSe (01-072-7490) NPs and 25% Ag(I) sample show the formation of a orthorhombic CuAgSe (00-010-0451) phase (Fig. 3A). If the Ag(I) content is increased further (to 100%) a full exchange can be observed, as the orthorhombic Ag2Se phase (01-089-2591) is forming. This shows a clear difference to the Cu1.1S system.
To recap the results in this paragraph, it can be seen that due to a controlled exchange (Cu2−xSe) or intercalation (Cu1.1S) with univalent ions it is possible to damp and shift the LSPR maximum of copper chalcogenides. While both systems show the same, under O2 exposure reversible, optical trend for the addition of Cu(I), their behavior varies significantly for the addition of Ag(I). The Cu2−xSe NPs show a strong, non reversible damping after Ag(I) addition. The LSPR of the Cu1.1S however can be almost non-reversibly bathochromically shifted by the Ag(I) addition. This allows a new way to permanently tune the position of the LSPR maximum for degenerately doped chalcogenides. The use in applications that require a stable and specific LSPR under ambient conditions is hence enabled.
The growth of Au on Cu2−xSe can be controlled by changing the amount of Au-precursor added to the same amount of Cu2−xSe NP solution. Cu2Se:
Au ratios between 1
:
0.24 and 1
:
1.89 have been achieved. The TEM bright field images (Fig. 4) show that high contrast Au domains have been grown on the Cu2−xSe seed NPs (lower contrast domains). For the 1
:
0.24 sample 76% of the Cu2−xSe seed particles have been covered with a 5.6 nm ± 1.3 nm Au domain. For the 1
:
0.56 sample 8.0 nm ± 1.6 nm domains have been grown on 73% of the seed particles (no separately nucleated gold particles were observed). This corresponds roughly to the calculated size of the Au domain per particle assuming 100% reaction yield, which is 6 nm for the 1
:
0.24 sample and 8 nm for the 1
:
0.56 sample. The size of the sample with the largest Au amount (1
:
1.89) could not be determined due to its strong tendency to agglomerate. The XRD patterns (Fig. 5) show a continuous intensity growth of the reflexes that can be assigned to fcc Au (PDF card #: 00-004-0784) while the relative intensity of the berzelianite reflexes is decreasing at the same time until they can hardly be detected for the 1
:
1.86 sample. The relative decrease can be explained by the higher X-ray atomic scattering factor for gold.
With increasing Au content, the absorption spectra (measured with an integrating sphere) of the as-prepared samples show a more pronounced absorption peak centered at 550–600 nm which originates from an LSPR in the Au domains of the hybrid NPs (Fig. 4). The intensity of the absorption band of the Cu2−xSe LSPR on the other hand decreases with an increasing amount of Au and shows a small hypsochromic shift of about 100 nm from the 1:
0.24 to 1
:
0.56 sample. A further increase of the Au amount (1
:
1.89 sample) leads to a broad absorption band where no distinct Cu2−xSe LSPR absorption can be distinguished. The broad absorption band rather shows a maximum around 600 nm, however with an absorption onset above 2000 nm. It cannot be excluded that the difference in absorption behavior in comparison to the other samples originates in the strong agglomeration that occurred for this high Au contents.
The as-prepared samples show a LSPR in the NIR region due to the copper deficiency in the Cu2−xSe part. By the addition of Cu(I) ions to this hybrid system, additional Cu(I) is built into the Cu2−xSe, and partially in solution oxidized to Cu(II) (in order to reduce the selenium in the NPs). Hence, the free charge carrier concentration (holes) in the semiconductor domain is reduced. This, as has been discussed in literature1,4,6 and previously in this publication, leads to a bathochromic shift and decreased intensity of the copper chalcogenide LSPR as can be seen in Fig. 4.
In contrast to the two distinct and separate LSPR maxima that can be observed for our 1:
0.24 and 1
:
0.56 sample (Fig. 4), Liu et al. described for Cu2−xSe grown on Au seed particles a broad LSPR across visible and NIR wavelengths. They qualitatively analyzed this using a so called quantum model.28 As Cu2−xSe NPs are strongly p-doped, the Fermi level is close to the valence band. In order to achieve thermal equilibrium at the interface, electrons of the metal will diffuse into the semiconductor. The consequence will be an electron density reduction in the Au part, and hence a bathochromic shift of its LSPR. The semiconductor LSPR should show a broadening and reduced intensity, due to a lower charge carrier density.28 Following this model a decrease of p-doping would lead to less electrons being able to diffuse from the metal to the semiconductor part, as to the decrease of p-doping. This, however, would be expected to result in a hypsochromic shift for the Au LSPR for higher Cu(I) amounts being added. No such shift can be observed (Fig. 4). In fact the metal LSPR shifts bathochromically for about 20–30 nm for the 1
:
0.24 and the 1
:
0.56 sample and around 50–60 nm for the 1
:
0.89 sample. One explanation for this shift is the bathochromic shift of the Cu2−xSe band gap absorption, due to its reduction according to the reverse Burstein–Moss effect.
It is worth to note that the position of the plasmon maximum above 550 nm would require a pure Au NP to have a size well above 50 nm. As no such Au particles could be found in any sample the LSPR maximum must hence be explained differently. Another reason for the absorption maximum at relatively long wavelengths is the high frequency dielectric constant of Cu2−xSe (ε∞ = 11.0–11.6)41 compared to toluene. This approach also explains why no hypsochromic shift of the Au related LSPR is observed during the addition of Cu(I).
Adding Ag(I) ions to the Au–Cu2−xSe hybrid NPs the semiconductor LSPR can be non reversible, bathochromically shifted as we showed for the pristine Cu2−xSe particles (see Fig. S3 in the ESI†).
We showed here by Au growth experiments that with our synthesis strategy we received dual-plasmonic hybrid Au–Cu2−xSe NPs which exhibit two surprisingly little interacting LSPRs. This is, to the best of our knowledge, the first report of weakly coupled LSPRs in Au–copper chalcogenide hybrid systems. The difference to previous reported hybrid systems might be the reverse growth strategy that allows the coupling of the two domains at room temperature.
The absorbance spectra show a LSPR maximum shift from 1100 nm to 1365 nm from no Au added to a Cu1.1S:
Au ratio of 1
:
11.2 (Fig. 6). The shift of the LSPR is highly stable under air exposure over several days and only shows slight variations in the oscillator strength. The LSPR shift behaves very similar to the one observed for the addition of Ag(I), however due to the slower reaction kinetics an even better control of the exact maximum position can be achieved. The high stability of the LSPR shift under oxygen exposure can be explained due to the formation of the stable Au2S phase. The increasing formation of Au2S can be clearly seen in the provided TEM images in Fig. 6. In comparison to pure covellite NPs (see Fig. S4 in the ESI†) these samples show a high contrast crown, due to the higher contrast of elements with a higher atomic number in bright field imaging, which is increasing with higher Au amounts added. By controlling the amount of Au added this technique gives an additional method to shift and stabilize the LSPR maximum at a desired wavelength.
A possible reason for the different behavior of the two systems are their different redox potentials. These result in the reduction of the Au(III) precursor to Au(0) for the Cu2−xSe system but only to Au(I) for the Cu1.1S system.
In summary the growth of Au domains on Cu2−xSe with size control can be achieved with the here presented method. This leads to Au–Cu2−xSe dual-plasmonic hybrid particles with two LSPRs originating from the two domains of the particles. It was shown for the first time that the LSPRs of such Au–copper chalcogenide hybrids influence each other little. Applying the same strategy to grow Au on the Cu1.1S NPs does not result in the growth of Au domains but rather in a cation exchange and the formation of Au2S–Cu1.1S NPs (see Fig. 6). As with the presented method the exchange is not complete, the LSPR can still be detected; even though it is strongly bathochromically shifted. Additionally the shift is stable also under air exposure for several days.
Overall, the post-synthetic treatment of plasmonic copper chalcogenides gives, despite of its complexity, a manifold way to tune the LSPR of these particles and especially to stabilize the LSPR position also under ambient conditions.
Pre-degassed OLA and ODE were prepared by placing each solvent in a separate flask and heating them under reflux (≤1 × 10−3 mbar, 115 °C) for 6 h, while purging them multiple times with argon. Afterwards they are transferred and stored inside the glove box under inert gas.
All other chemicals were used without further purification.
If not otherwise mentioned all experiments have been done under argon using standard Schlenk line techniques or were done inside a nitrogen filled glove box.
A typical Cu2−xSe NP solution in toluene contains 0.053 mmol per ml Cu2−xSe, this corresponds to roughly 2.5 × 10−6 mmol NPs per ml.
An Au-precursor stock solution was freshly prepared with an Au:
DDAB
:
DDA ratio of 1
:
5.5
:
22 and an Au concentration of 0.006 mmol per ml toluene. For that AuCl3 (72.9 mg; 0.24 mmol), DDAB (611.7 mg; 1.32 mmol) and DDA (980.3 mg; 5.29 mmol) were dissolved in 40 ml toluene. 1.7 ml of a diluted Cu2−xSe solution (cCu2−xSe = 0.033 mmol ml−1) and 25 μl OLA are placed in a sample vial and under stirring calculated amounts of the Au-precursor solution are added slowly. The solution is left stirring for 30 min. The particles are separated by adding MeOH and centrifugation at 3700g. The NPs are redispersed in 4 ml toluene.
For Cu1.1S 2 ml of a diluted solution (cCu1.1S = 0.009 mmol ml−1) were placed in a sample vial. All further steps have been done like for the Au growth on Cu2−xSe.
Footnote |
† Electronic supplementary information (ESI) available: Further HAADF-STEM images and EDX mappings, further absorbance spectra and the theoretical dependence of the extinction cross section against γ. See DOI: 10.1039/c5nr05425g |
This journal is © The Royal Society of Chemistry 2015 |