Ternary Chalcogenide-Based Quantum Dots and Carbon Nanotubes: Establishing a Toolbox for Controlled Formation of Nanocomposites.

A general procedure based on electrostatic self-assembly for preparing nanocomposites based on carbon nanotubes (CNTs) and ternary chalcogenide semiconductor nanoparticles is shown. This was achieved by surface functionalization of the single components through well-established protocols, for CNTs, and a transferable general strategy for the nanoparticles. Heterostructures were then synthesized through electrostatic interaction between oppositely charged components. Structural, colloidal, and optical properties were characterized by transmission electron microscopy, X-ray diffraction, infrared spectroscopy, dynamic light scattering, ζ-potential, and absorption- and (time-resolved) photoluminescence measurements. Interestingly, the nanocomposites showed a blue shift in their excitation and emission spectra when compared to the pure nanoparticles but only when analyzed in powder form. Further investigations in the form of density functional theory (DFT) calculations were performed to evaluate the origin of the change in the optical properties.

Introduction

Carbon nanotubes (CNTs), either in their single-walled (SWCNT) or multiwalled form (MWCNT), have gained a lot of attention in the past decades and established themselves in various scientific disciplines as well as in the fields of nanoengineering and nanotechnology,[1,2] such as nanomedicine,[3] electrode materials,[4−6] and optoelectronics.[7−9] The wide scope of applications is not surprising due to their exceptional characteristics. To name a few, some of the most prominent properties are their excellent thermal and electrical conductivity (owing to their delocalized π-electron system),[10] high Young’s modulus[11] and tensile strength,[12] good resilience,[13] stability, and a high surface area.[14] Therefore, numerous studies have been conducted to modify and functionalize the sidewalls of the CNTs to expand their potential applications.[15−20]

Due to the high attention and research activity, the chemistry behind the functionalization of CNTs is well understood, which makes them together with their unique properties ideal candidates as templates for the assembly with other functional nanomaterials. It has been shown that the surface decoration of CNTs with various organic or inorganic species can increase the tunability of their physical, electrical, and optical properties.[21−23] Thus, as the demand for smaller, faster, and more efficient electronic devices increases, nanocomposites and functional CNTs with new or tuned properties may play an important role in the miniaturization process of nanoengineering and nanotechnology, including new catalyst materials,[24] sensitization of solar cells,[25] nanobatteries,[26] etc.

Within such applications, nanocomposites consisting of semiconductor quantum dots (QDs) that are attached to CNTs have emerged as an active field.[27−29] The unique possibility to tune important optical and electronic properties of the QDs by varying their size is of great interest in many applications.[30] Multinary chalcogenide materials, like CuInS2 (CIS), Cu(In,Ga)S2 (CIGS), and Cu(In,Ga)Se2 (CIGSe) are herein of considerable interest as promising semiconductors for optoelectronic devices[31−33] due to their outstanding electronic and optical properties (owing to their properties as a direct band-gap semiconductor) and precise control of these by uncomplicated modification of the preparation conditions (size, surface chemistry, composition), as well as cost-efficient production out of commercially available precursors.

However, before nanomaterials can be efficiently utilized, there is the necessity to understand the synthesis, surface engineering, and internal effects, which can arise from the functionalization and combination of single nanomaterials into nanocomposites for their posterior integration and application. The aim of this work is, therefore, to investigate how, generally, QDs can chemically be attached to CNTs, what influence this has on the optoelectronic properties, and how they can be tuned through surface engineering. Based on our previous work that allows for versatile surface engineering of ternary chalcogenide-based nanocrystals[34] and well-established methods to functionalize the sidewalls of MWCNTs, we synthesized QD-CNT nanocomposites by simple electrostatic interaction for the present report. By modifying the surface with ligands of different alkyl chain lengths, the separation distance between QDs and CNT was varied and crucial properties such as the band gap and the electron transfer between the nanomaterials were studied intensively by experimental and theoretical ab initio methods.

Methods

Chemicals and Reagents

MWCNTs (95% carbon assay, 6–9 nm × 5 μm (outer diameter × length)), ethylenediamine (EDA), N,N′-dicyclohexylcarbodiimid (DCC), copper(I) acetate (Cu(OAc)), indium(III) acetate (In(OAc)3), zinc acetate (Zn(OAc)2), stearic acid (SA), sulfur (S), 1-dodecanethiol (DDT), 1-octadecene (ODE), oleylamine (OAm), 8-amino-1-octanethiol-hydrochloride (AOT), 2-(dimethylamino)-ethanethiol hydrochloride (DMA), 2-(diethylamino)-ethanethiol hydrochloride (DEA), 3-mercaptopropionic acid (MPA), 11-mercaptoundecanoic acid (MUA), l-glutathione (l-GLU), and tetramethylammonium hydroxide (TMAH) were all purchased from Sigma Aldrich. Cysteamine hydrochloride (CYS), conc. nitric acid (HNO3, w = 65%), chloroform (CHCl3), n-hexane, methanol (MeOH), and toluene were obtained from Carl Roth. All reagents were used as received without any prior purification step.

Carboxyl Functionalization of Carbon Nanotubes (c-MWCNTs)

Covalent carboxyl functionalized MWCNTs (c-MWCNTs) were obtained by partial oxidation with concentrated nitric acid similar to previous reports.[35] In a typical synthesis, 200 mg of pristine MWCNTs was dispersed in 60 mL of HNO3 (w = 65%) and ultrasonicated for 30 min. Afterward, the mixture was heated up and refluxed for different oxidation times (2, 4, 6, and 8 h). Subsequently, the reaction was cooled to room temperature and the c-MWCNTs were purified by repeated centrifugation at 10,000 rpm for 15 min and washing with bidest. H2O until the pH of the supernatant was neutral. The resulting solid was dried at 60 °C for 24 h. Stock solutions of c-MWCNT were prepared by dissolving them at concentrations of 0.5 mg/mL in bidest. H2O by means of probe ultrasonication (at 20,000 kHz and a 48% amplitude for 3 min).

Amine Functionalization of Carbon Nanotubes (a-MWCNTs)

Amine-functionalized MWCNTs (a-MWCNTs) were synthesized by covalent bonding of a diamine to the beforehand prepared c-MWCNTs with DCC, as described by Oytun et al.[36] In a typical synthesis, 10 mg of c-MWCNTs was dispersed in 10 mL of EDA and ultrasonicated for 30 min. Then, 350 mg of DCC was added and the mixture was sonicated for an additional 15 min. Afterward, the dispersion was stirred for 96 h at 100 °C temperature. For purification, the solution was centrifuged at 10,000 rpm for 15 min and washed with EtOH at least three times, and subsequent dried at 60 °C for 24 h. Stock solutions of a-MWCNT were prepared by dissolving them at concentrations of 0.5 mg/mL in bidest. H2O by means of probe ultrasonication (at 20,000 kHz and a 48% amplitude for 3 min).

Synthesis of CuInS2, CuGaS2, and CuInSe2 Core Quantum Dots

The synthetic approach was adopted from our previously published paper[34] and transferred from CuInS2 (CIS) to synthesize CuGaS2 (CGS) and CuInSe2 (CISe). In a typical procedure, 73.5 mg (0.6 mmol) of Cu(OAc), 175.2 mg (0.6 mmol) of In(OAc)3 (or 220.2 mg (0.6 mmol) Ga(acac)3), 1023.9 mg (3.6 mmol) of SA, 15.0 mL of ODE, and 15.0 mL of DDT were loaded in a 50 mL three-neck flask and vacuumed for 30 min at 120 °C to remove any water or oxygen that was left in the reaction mixture. Afterward, it was flushed with Ar (by means of Schlenck-line technology) and heated with a heating mantle under vigorous stirring to 200 °C. Then, 2.4 mL of a solution prepared by dissolving 128.3 mg (4.0 mmol) of S in 10.0 mL of ODE at 120 °C (or 2.4 mL of a solution prepared by dissolving 315.8 mg (4.0 mmol) of Se powder in 5.0 mL of OAm and 5.0 mL of DDT at room temperature under a N2 atmosphere) was injected quickly into the reaction mixture. The solution was further heated to 230 °C (240 °C for the synthesis of CGS) and kept at that temperature for 30 min to allow the growth of the core QDs. For the purification of the prepared nanocrystals, the reaction mixture was cooled to 60 °C with a water bath. Then, 10 mL of toluene was injected into the reaction mixture to stabilize the QDs. By adding MeOH, the QDs were precipitated and centrifuged at 10,000 rpm for 10 min. The supernatant was discarded and the underlying QDs were washed at least three more times with EtOH (T = 60 °C) to purify them. STEM photographs and the temporal evolution of the optical properties are shown in Figure S1a–f.

Synthesis of CuInS2/ZnS, CuGaS2/ZnS, and CuInSe2/ZnS Core/Shell Quantum Dots

The growth of the first ZnS shell was carried out in situ directly after the completion of the QD core growth without any purification step. The reaction was quenched in an ice bath to temperatures of 60 °C and water-free Zn(OAc)2 was added as a shell precursor. For details and the exact amount of Zn(OAc)2 corresponding to a certain shell thickness, see the Supporting Information (Figure S2 and Table S1). The mixture was heated to 240 °C and held at that temperature for 2 h to let the ZnS shell grow around the nanoparticles. The purification of these core/shell QDs is similar to that of the core QDs.

Synthesis of CuInS2/ZnS/ZnS, CuGaS2/ZnS/ZnS, and CuInSe2/ZnS/ZnS Core/Shell/Shell Quantum Dots

For the overgrowth of a second ZnS shell around the core/shell QDs to prepare core/shell/shell QDs, 0.3 mmol purified CIS/ZnS QDs were dissolved in 3.0 mL of CHCl3 and placed together with 6.0 mL of OAm, 6.0 mL of ODE, and the amount of Zn(OAc)2 needed for two more ZnS monolayers (ML) in a 50 mL three-neck flask. The flask was vacuumed at room temperature for 30 min. Afterward, it was flushed with Ar and heated with a heating mantle under vigorous magnetic stirring to 180 °C. Then, a stoichiometric amount of a solution prepared by dissolving 641.4 mg (20.0 mmol) of S in 20.0 mL of ODE at 120 °C was injected dropwise at a rate of 0.6 mL/min into the reaction system. After the addition of the S precursor, the second ZnS shell was allowed to grow for 2 h. Note: for CGS/ZnS/ZnS, an agglomerated turbid yellow/greenish solution was obtained instead of a clear and stable nanodispersion. Before purification, this mixture was centrifuged at 10,000 rpm for 5 min and the resulting clear yellow (luminescent) solution was processed further, while the green solid residue was discarded (for a more detailed analysis of this observation, see Figure S4). Next, the reaction mixture was cooled to room temperature in a water bath and the CIS/ZnS/ZnS core/shell/shell QDs were precipitated by the addition of EtOH and centrifuged at 10,000 rpm for 5 min. The supernatant was discarded and the precipitate was redispersed in 5.0 mL of n-hexane and precipitated again by adding 10.0 mL of EtOH. This washing procedure was repeated at least three times.

Ligand Exchange

Further, 0.1 mmol core/shell/shell QDs were dissolved in 5.0 mL of CHCl3. Under strong magnetic stirring, a solution of 2.0 mmol thiol (CYS, AOT, MPA or MUA) and 729.2 mg (4.0 mmol) of TMAH dissolved in 1.0 mL of MeOH was added dropwise. Afterward, the dispersion was stirred for 2 h. The mixture was precipitated at 10,000 rpm for 10 min and washed three times with 5.0 mL of bidest. H2O and 30.0–40.0 mL of acetone. To ensure good long-term colloidal stability through a dense surface ligand coverage, the precipitate was dispersed in 5.0 mL of bidest. H2O and 2.0 mmol corresponding thiol and refluxed for 2 h (for the MUA samples, this step was omitted because it led to irreversible agglomeration). The mixture was purified again as described above and dried at 50 °C for 24 h. Note: After drying, some samples could not be redispersed in an aqueous solution and formed turbid solutions. To yield clear and stable dispersions, 100 mg of core/shell/shell QDs and 100 mg of the corresponding thiol were refluxed once again for 1 h. The mixture was purified again as described above. During these clean-up steps, the colloidal stability increased with every washing step. Stock solutions of the QDs were prepared by dissolving a certain amount of nanocrystals in water to yield a 5.0 mg/mL concentration, adding the corresponding thiol as a stabilizer at concentrations of 1.0 mg/mL and adjusting the pH value of the solution with 1.0 mol/L HCl or NaOH depending on the nature of the stabilizing ligand (see Figure ). Due to the nature of thiol ligands that oxidize quite rapidly, the stock solutions were stored in a fridge at 4 °C until further use. If agglomeration occurred after prolonged storage time, further addition of stabilizing ligands could redisperse the QDs, yielding clear solutions again.

Figure 3

Photographs of the phase transfer from organic to aqueous media under daylight and UV light (a), ζ-potential and DLS size dependency of the pH value of CYS- (b), AOT- (c), MPA- (d), and MUA- (e) capped CISe, CIS, and CGS QDs.

QD-CNT Nanocomposite Synthesis

For the synthesis of QD-CNT composites, 50 mg core/shell/shell QDs (10.0 mL of the QD stock solution) and a certain amount of c-MWCNTs or a-MWCNT stock solution were added together (different nanocomposites with varying amounts of MWCNTs were prepared and the respective weight percentage (wt %) is with respect to the 50 mg of QDs) and filled up to 15.0 mL with bidest. H2O. The pH of the solution is adjusted with 0.1–1.0 mol/L HCl or NaOH and afterward stirred for at least 3 h at room temperature. The resulting nanocomposites were analyzed in solution as well as in powder form after precipitation with EtOH and centrifugation at 10,000 rpm for 10 min and drying at room temperature for 24 h.

X-ray Diffraction (XRD) Analysis

The crystallographic structure of the prepared nanomaterials was studied by XRD techniques using a Rigaku MiniFlex II Desktop diffractometer equipped with a radiation source from an X-ray tube with Cu Kα radiation (λ = 1.54 Å). Measurements were performed in a 2θ range of 20–80° with a scan rate of 2°/min and 0.02° increases.

Size and ζ-Potential Measurements

A Malvern Instruments Zetasizer Nano ZS was used for particle size determination by dynamic light scattering (DLS) and ζ-potential from electrophoretic mobility measurements, together with a standard pH meter (SevenMulti, Mettler Toledo). Measurements were performed from sample dispersions of 0.1–5 mg/mL. The pH value was varied using hydrochloric acid (0.1–1.0 M) and sodium hydroxide (0.1–1.0 M). For DLS measurements, a 633 nm laser in combination with 173° backscatter detection was used. Each measurement was averaged from 10 runs, each lasting 10 s. The measurements were repeated three times and the averages were taken. For the ζ-potential, each measurement was collected from 15 runs; all experiments were repeated three times and the averages were recorded.

Photoluminescence (PL), PL Decay, and Absorption Measurements

PL and absorption spectra as well as PL decay curves at room temperature were recorded to characterize the optical properties. Emission spectra were recorded using a Spectrofluorophotometer FS5 from Edinburgh Instruments. UV–vis absorbance measurements were performed using an Analytik Jena Specord 200Plus. PL decay curves were measured on a fluorescence spectrometer FLS980 from Edinburg Instruments, equipped with an EPL-450 ps laser (λex = 445.6 nm, pulse width ≤ 70 ps). For the measurements of the decay curves, the monitoring wavelength was set to the wavelength of the PL at its maximum intensity.

Cyclic Voltammetry (CV) and Differential Pulse Voltammetry (DPV)

A ModulLabXM ECS from Solartron analytical was used for the electrochemical characterization of the samples. Measurements were performed in a three-electrode cell using a 1 cm2 platinum sheet counter electrode (CE), a silver wire pseudo reference electrode (RE), and a 3 mm (diameter) glassy carbon (GC) working electrode (WE). The RE is calibrated against the ferrocene/ferrocenium (Fc/Fc+) couple.[37] LiClO4 (0.1 M) in acetonitrile (CH3CN) was used as an electrolyte. CV tests were performed in the electrolyte with a potential of +1.5 to −2.0 V (relative to the ferrocene (II/III) standard reduction potential) at a scan rate of 100 mV/s. DPV measurements were carried out with an increasing potential of 3 mV, a pulse amplitude of 60 mV, a pulse width of 50 ms, and a pulse period of 500 ms. To perform measurements on the QD/MWCNT composites, the GC working electrode was dip-coated with a thin film of the sample material. To increase the affinity of the QDs to the electrode and to improve the accuracy of the coating, the GC substrate underwent a number of cleaning and preparation steps: polishing with a 0.3 μm alumina slurry on paper cloth for 2 min, rinsing with deionized water, cleaning in an ultrasonic bath with deionized water for 15 min followed by pure ethanol for 10 min, and then drying in air. The clean and pretreated substrate was dipped into a water-based dispersion of the QD/MWCNT for 20 min, then air-dried. The procedure is repeated twice for each sample.

Fourier Transform Infrared (FTIR) Spectroscopy

The adsorption of organic ligands and capping of the nanoparticles were studied by FTIR spectroscopy using a Nicolet iS 5 with a diamond attenuated total reflectance (ATR) controller from Thermo Fischer Scientific. Measurements were performed from 4000 to 600 cm–1.

Computational Details

Calculations were performed in the framework of a density functional theory (DFT) computational method,[38,39] as implemented in the Quantum ESPRESSO package.[40,41] Energies were calculated using a plane-wave basis set, scalar relativistic ultrasoft pseudopotentials (USPP), and the generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional to describe the exchange–correlation potential.[42] To improve the accuracy of the calculations, a Hubbard U correction[43] was applied on the Cu 3d, Ga 3d, In 4d, Zn 3d, S 2p, and Se 3 3p states to correct the well-known problem that LDA and GGA functionals underestimate the band gap and are not able to reproduce the correct position of d or f states of transition or rare earth metals.[44] Since the linear-response theory (used to calculate ab initio values for the Hubbard U parameters[45]) has its limitations and fails for closed-shell systems,[46] the U parameters were optimized simultaneously to reproduce the experimental lattice parameter, band gap, and the relative position of d states of ZnS and CuInS2, similar to Mattioli et al. and Khan et al.[47,48] Satisfactorily converged results were achieved using 60 Ry as the kinetic energy cutoff for wave functions and 480 Ry as the kinetic energy cutoff for the charge density and potential (see Figure S11). The convergence thresholds for selfconsistency, the total energy, the forces for ionic minimization, and the pressure for variable cell relaxations were set at 1.0 × 10–8 Ry, 1.0 × 10–4 Ry, 1.0 × 10–3 Ry/Bohr, and 5.0 × 10–4 GPa, respectively. A 6 × 6 × 6 Monkhorst–Pack k-point grid including the Γ point was used for sampling the 1st Brillouin zone of bulk ZnS, CuInSe2, CuInS2, and CuGaS2, while a Γ-point-only sampling was used for the calculation of the stand-alone QDs and CNT-QD structures. The supercell with the QD located at the center contains a vacuum region of at least 10 Å to avoid periodic interactions. Furthermore, to compensate for surface defect states arising from dangling bonds on surface ions, pseudoatoms, with fractional nuclear charges (1.75 for Cu+, 1.25 for In3+ and Ga3+, 1.5 for Zn2+, and 0.5 for S2– and Se2–) as suggested by Huang et al.,[49] were used to passivate these. Electronic band and state positions with respect to the vacuum level were obtained by calculating the planar average electrostatic potentials along the z-axis and the vacuum region was used as the absolute energy reference. The QDs were constructed by cutting out a spherical shape from the optimized underlying bulk structure by selecting all atoms within a radius of r < x2 + y2 + z2 around a sulfur atom. All systems were optimized with respect to their geometry before any electronic structure calculation was performed.

Results and Discussion

Surface Modification of MWCNTs

The synthetic strategy, illustrated in Figure , for the construction of nanocomposites consisting of QDs and MWCNTs follows electrostatic self-assembly of the single components. This can be achieved by electrostatic interaction between charged functional groups (i.e., protonated amine groups and deprotonated carboxyl groups). The main advantage of this approach is that important properties like, e.g., band gaps and band edges can be fine-tuned before the building blocks of the composite are united. Thus, already established and well-controlled synthesis protocols can be used. To yield charged CNTs, the pristine MWCNTs were functionalized and carboxyl and amino groups were introduced to the sidewalls (c-MWCNT and a-MWCNT). In Figure a, the pH-dependent ζ-potential and DLS size curves of pristine MWCNTs are shown and what happens after carboxylation and amidation (c-MWCNT and a-MWCNT). Pristine MWCNT without any functional groups on the surface are not colloidally stable, as indicated by high hydrodynamic diameters (agglomerates). After oxidative treatment by boiling the pristine MWCNT in concentrated nitric acid, the partial oxidation from pristine CNTs to c-MWCNT leads to carboxyl groups on the nanotube surface; these can easily be deprotonated, resulting in a negative surface charge and repulsion between single nanotubes[50] thus yielding stable dispersions in water over a large pH range. By increasing the oxidation time, more functional groups are introduced, and simultaneously more of the structure gets altered and destroyed.[51] As it can be seen, the ζ-potential as well as the hydrodynamic diameter are decreasing (see Figure b). Also, visible damages on the sidewalls were observed for prolonged oxidation times (see Figure S7). After amidation of the c-MWCNT to a-MWCNT, mostly amine groups are the functional groups on the outer sidewalls. These can be protonated in acidic conditions resulting in positive surface charges, leading again to repulsion between the nanotubes and stable dispersions. But different from c-MWCNT, the colloidal stability of a-MWCNT is given in a smaller pH range. Above pH ∼9.5, the surface charges are not high enough to compensate for the intermolecular forces and agglomeration occurs, which is in accordance with increasing hydrodynamic diameters (see Figure c). Using c-MWCNT with more functional carboxyl groups to synthesize a-MWCNT (longer oxidation times) should also lead to more amine groups and thus to higher positive surface charges. This is why the trend of the ζ-potential is reversed with oxidation time, as indicated by the arrows. However, the influence on the ζ-potential and DLS sizes is not that pronounced as it is for c-MWCNTs, but still a trend toward higher ζ-potentials and a wider pH range of colloidal stability is observed.

Figure 1

Illustration of the preparation of QD/MWCNT heterostructures by electrostatic self-assembly.

Figure 2

ζ-Potential and DLS size dependency on the pH value of pristine MWCNTs (a), c-MWCNTs (b), and a-MWCNTs (c).

Synthesis and Surface Engineering of CIS-, CGS-, and CISe-Based QDs

To obtain charged nanocrystals based on CuInS2, CuGaS2, and CuInSe2, our previously published work for the versatile surface engineering of QDs[34] was used as a fundament and further developed. As shown in Figure S5a, the XRD diffraction peaks reveal that the synthesis protocol can be easily modified to synthesize different chalcopyrite nanometer-sized materials. All diffraction peaks are broadened due to the finite particle size,[52] which is in accordance with STEM photographs of core/shell nanoparticles with different shell thicknesses (see Figure S3a–c). Figure S5a,b shows the absorption and emission spectra of these three chalcopyrite materials as well as photographs. Due to differences in the band gaps of the materials, the emission peaks and absorption bands are blue-shifted from CuInSe2 over CuInS2 toward CuGaS2, which is in accordance with the literature,[53] emphasizing again that the modified synthesis protocol works. Also shown is the influence of the passivating ZnS shell. With an increase of ZnS monolayers (ML) around the QD cores, the photoluminescence intensity increases substantially and blue shifts toward higher energies. The intensity increase can be attributed to the quenching of surface and trap states and the blue shift is a result of the QD cores shrinking[54] and the formation of ZnS alloyed cores through diffusion and cation exchange processes.[55]

The results of the ligand exchange are shown in Figure . Photographs show the success of the phase transfer from organic to aqueous media with only minor influences on the optical properties (see Figure a). Confirmation of the replacement of oleylamine as a capping ligand with different thiol-based ligands is presented via the pH-dependent ζ-potentials and DLS sizes (Figure b–e). For CYS- and AOT-capped nanoparticles, high positive ζ-potentials were measured in acidic media, yielding stable dispersions as indicated by low hydrodynamic diameters. CYS and AOT molecules possess a functional amine group, which can be easily protonated (and positively charged) at low pH values, explaining the measured positive ζ-potential. The charged particles repulse each other through Coulombic forces, which is in accordance with the observed stable dispersions and low measured particle sizes. With an increase in the pH value, the ζ-potential decreases (−NH3+ deprotonates and loses the charge), the dispersions become unstable, and the QDs start to agglomerate as it is pointed out with measured increasing DLS sizes (electrostatic repulsion between particles gets weakened so much that attracting intermolecular forces like dipole–dipole, hydrogen bonds, and van der Waals interactions are dominating). For MPA- and MUA-capped QDs, the trend is reversed, indicating that high negative ζ-potentials were measured in basic media leading to stable dispersions and small hydrodynamic diameters. On decreasing the pH (and therefore the absolute value of the ζ-potential), agglomerates were formed, again indicated by increased measured particle sizes. The functional group in MPA and MUA molecules is a carboxyl group, which can be easily deprotonated in basic media leading to a negative charge and electrostatic repulsion between particles capped with these ligands. Similar to described above, when the carboxyl group loses its charge (e.g., gets protonated at lower pH values), the dispersion loses stability and agglomerates are observed. A further indication of a successful ligand exchange is the FTIR spectra presented in Figure S6 with the most characteristic absorption bands (like the stretching vibration of COOH, C=O, C–O, N–H, and C–N) together with adsorbed water.

To our best knowledge, this is the first report of CuInSe2- and CuGaS2-based nanocrystals capped with cysteamine and CuInSe2, CuInS2, and CuGaS2 capped with 8-amino-1-octanethiol as a ligand, once again highlighting the possibility of universally changing the surface properties of chalcopyrite nanomaterials, and thus generating a toolbox for the construction and design of new heterostructures in the nanometer region.

Electrostatic Interaction between QDs and MWCNTs to Form Nanocomposites

A prerequisite for the formation of QD/MWCNT nanocomposites is that the single building blocks are colloidally stable and sufficiently charged. The above described and characterized surface functionalization of both MWCNTs and QDs reveals that c-MWCNTs are negatively charged and colloidally stable over the whole measured pH range (see Figure b, from pH ∼2.5–12.0). Oppositely charged QDs (functionalized with CYS or AOT) were stable from about pH ∼2.5–8 (see Figure b,c). Thus, for the synthesis of heterostructures consisting of negatively charged MWCNTs and positively charged QDs, a wide pH range could be utilized. Unlike c-MWCNTs, a-MWCNTs, which are positively charged were only stable till a pH of ∼9.5, above which agglomeration occurred (see Figure c). Negatively charged QDs (functionalized with MPA or MUA) were only stable above pH ∼8.5 (see Figure d,e). This together gives a very small pH window for the preparation of nanocomposites consisting of positively charged MWCNTs and negatively charged QDs.

In Figure , this is illustrated with STEM photographs of nanocomposites prepared under different conditions. It can be seen that if equally charged QDs and MWCNTs in stable conditions or oppositely charged binary components but in unstable conditions (in terms of the above-discussed pH value) were used to prepare QD/MWCNT heterostructures a bad, up to nonexistent surface coverage of the carbon nanotube sidewalls with nanoparticles was observed (as it can be seen in Figure a,b). In contrast, if oppositely charged QDs and MWCNTs in stable conditions were used to prepare the nanocomposites, a fine homogeneous and even decoration and coverage of the MWCNTs surface with QDs were achieved (Figure c). This highlights the above-mentioned necessity that the single components used for the synthesis need to be colloidally stable and also underlines the precise control of the electrostatic interaction via the pH of the solution. For an overview and zoomed-in STEM photographs, see Figure S9. In areas where multiple nanotubes overlap each other, a clustering of the nanoparticles is observed, which is in accordance with previously observed bridging adsorption between multiple CNTs by the charged nanoparticles.[56] Differences for composites prepared with low or high concentrations of MWCNTs or with core/shell nanoparticles with thin or thick shells can be seen in Figure S10. It should be noted, however, that the reversing of the self-assembly could not be achieved by varying the pH value. For that to happen, repulsive forces between the components would be necessary, which means one of the components has to switch its surface charge from negative to positive or vice versa. As was indicated via the pH-dependent ζ-potentials and particle sizes above, the components are only colloidally stable in acidic or basic media (not in both, except the c-MWCNT, but they possess a negative ζ-potential across all measured pH values), indicating that agglomerates would form when one tries to invert the surface charge via the pH. When agglomerates are observed, it is a sign that the attractive forces are stronger than the repulsive ones. However, in theory, such a composite with reversible self-assembly could be produced with bifunctional (zwitterionic) components and might be an interesting topic for further research.

Figure 4

STEM photographs of differently prepared nanocomposites: oppositely charged QDs and MWCNTs but an unstable pH region (a), equally charged binary components and a stable pH region (b), and oppositely charged building blocks and a stable pH region (c).

Further investigations were carried out to examine the optical properties of the synthesized nanocomposites by photoluminescence emission and excitation spectroscopy (for simplicity and clarity, only the emission spectra are shown here; the excitation spectra are shown in Figure S8a). Figure a,b shows the optical appearance of cysteamine-capped CuInSe2, CuInS2, and CuGaS2 (from top to bottom)-based nanoparticles without and with 1, 2, 3, 5, and 10 wt % c-MWCNTs (from left to right) as stable dispersions. Figure c shows the corresponding emission spectra, showing the characteristic broad peaks in the near-infrared region for CISe, around 650 nm for CIS and around 550 nm for CGS. Luminescence quenching is observed with the addition of the nanotubes. This has two possible reasons. As carbon nanotubes are an absolute black material, they obviously absorb light in the visible up to the near-infrared region of the electromagnetic spectrum.[57] Furthermore, a nonradiative decay pathway, resulting from a charge transfer of photoexcited electrons or holes from the conduction or valence band of the nanocrystals (donor) to the empty electronic states of the CNTs (acceptor) is possible.[58] Such a charge transfer should only be possible when the separation distance between MWCNTs and QDs is small, indicating a strong attachment between the components (such behavior was observed above, as shown in Figure c). Further investigations of the charge transfer were performed in terms of Stern–Volmer analysis.[59] As shown in Figure d, the Stern–Volmer plot of CIS-based nanoparticles capped with different ligands gave nonlinear behavior. A possible reason for these nanocomposites to give nonlinear Stern–Volmer plots is the presence of both static and dynamic quenching.[60] This would mean that the standard expression of the Stern–Volmer equation is given as followswhere I0 and I are the intensity maximum of the unquenched and quenched emitter, [Q] is the concentration of the quencher, and Kd is the Stern–Volmer constant that in a purely dynamic fluorescence quenching process is the product of the bimolecular quenching rate constant and the fluorescence lifetime in the absence of an added quencher, which gets modified by a second term responsible for the static fluorescence quenching process to give the following expression

Figure 5

Optical appearance of the composites with different amounts of CNTs under daylight (a) and UV light (b) (345 nm excitation); corresponding emission spectra (c); Stern–Volmer plot for only dynamic quenching of CIS-CNT composites (d); and Stern–Volmer plots for both dynamic and static quenching of CISe-CNT (e), CIS-CNT (f), and CGS-CNT (g) heterostructures.

where Ks is the Stern–Volmer constant in a purely static fluorescence quenching process. Figure e–g shows the Stern–Volmer plots for simultaneous static and dynamic fluorescence quenching. It can be seen that after linear fitting and regression analysis, the data points and calculated R2 values almost show a linear behavior. As luminescence quenching through charge transfer is a way more complex topic as we represented it here, there are a lot more reasons for nonlinear Stern–Volmer plots, explaining the still not ideal R2 values. Nevertheless, the simplified representation is sufficient for our intents and characterization. The quenching rates for longer chain ligands decrease, as can be seen when comparing MPA- and MUA-capped (or CYS and AOT capped) quantum dots, which can be attributed to hindered charge transfer of the photoexcited exciton from the semiconductor material to the MWCNTs due to elongated separation through the ligand. This is in accordance with previous reports.[61] However, the length of the ligand is not the only variable responsible for the charge transfer, as negatively charged QDs have much higher quenching rates than positively charged QDs (despite, e.g., CYS being a shorter ligand than MUA). This suggests an overall stronger interaction between negatively charged QDs and a-MWCNTs than between positively charged QDs and c-MWCNTs, as it is also described by Pan et al.[62] It is further indicated that next to static quenching, dynamic quenching is present through PL decay lifetime measurement.[63] If the lifetime values remain unchanged in the presence of MWCNTs, it might be due to static quenching. However, as shown in Figure , there was a decrease in lifetime with increasing amount of nanotubes, which supports our assumption of both static and dynamic luminescence quenching being present.

Figure 6

PL decay times of cysteamine-capped CIS-based QDs with increasing amounts of c-MWCNTs.

Another interesting finding was that when the dispersed nanocomposites were precipitated, dried, and analyzed in their powder form, a slight blue shift of ∼10–20 nm of the emission peak was observed when heterostructures were compared to the pure QDs (also after precipitation in powder form), as shown in Figure . This behavior was observed for all used semiconductor materials, and in Figure a–d, it can be seen that it is also independent of the capping ligand. Interesting to mention is that the emission maximum of the composite materials is almost at the same wavelength as the emission maximum of the quantum dots in solution. Also, it should be noted that no blue shift was observed for the dispersed nanocomposites in solution or for powder nanocomposites prepared from unstable conditions, as shown in Figure S8b,c. Further, no shifts were observed from electrochemical analysis, as shown in Figure . The cathodic DPV peak voltage increases toward higher negative potentials from CISe over CIS toward CGS (Figure a) and upon increasing the thickness of the ZnS shell (Figure b). Only the cathodic reduction peak was observable, as the anodic oxidation of the nanoparticles is at higher potentials than that of the GC working electrode (see the CV scan). Nevertheless, the cathodic peak contains information about the conduction band minimum (CBM),[64] which upon increasing the band gap also has to shift toward higher energies. For the composite material, the intensity of the peak voltage increases upon increasing the amount of MWCNTs, which is the result of higher conductivities in those samples, but a shift of the peak voltage, as was observed for different semiconductor materials or thicker shells, did not occur. This indicates that the observed blue shift is an optical phenomenon rather than an influence on the electronic structure of the semiconductor material. However, optical observations are reported for nanoparticles attached to the sidewalls of carbon nanotubes or other substrates[56,65−67] and explained by electrostatic interactions between the binary components and the influence of electric fields originating from the substrate onto the band gap of the semiconductor material (quantum-confined stark effect).

Figure 7

Emission spectra of cysteamine-capped CISe-, CIS-, and CGS-based QDs with increasing amounts of c-MWCNTs, characterized in their powder form.

Figure 8

Normalized emission spectra of MPA- (a), MUA- (b), CYS- (c), and AOT- (d) capped nanoparticles based on CISe, CIS, and CGS, characterized in their powder form and compared to emitters in solution.

Figure 9

Comparison of cathodic DPV scans of different chalcogenide core/shell nanoparticles (a) (inset shows the CV scan), DPV scans of pristine cysteamine-capped CuInS2/ZnS core/shell nanoparticles and attached to c-MWCNTs (b), and DPV peak voltage of CISe, CIS, and CGS QD/MWCNT structures with increasing amounts of ZnS monolayer shells (c).

To test this assumption, we compared nanocomposites constructed from cysteamine-capped CIS/ZnS QDs and c-MWCNTs that were functionalized for different times by partial oxidation through boiling in concentrated nitric acid. As it was discussed above, this led to more functional groups and thus to more charges and stronger electric fields (see Figure b). If the electric fields would be responsible for the observed blue shift, we would expect a trend with increasing the electric field. However, this is not what we measured, as shown in Figure a,b. The emission maximum of the heterostructures is at almost the exact wavelength, slightly shifted by ∼20 nm compared to the pure QDs. This observation suggests that the electrical field exerted from the substrate might not be the reason for this shift. Further investigations and discussions about the origin of this shift will follow in the next section (see Section ). Figure c shows the corresponding Stern–Volmer plots showing that an increase in oxidation time leads to reduced photoluminescence quenching. This can be attributed to the partial destruction of the conjugated π-electron system when oxidizing the scaffold of the nanotubes to introduce −OH, −CHO, and −COOH groups, and by destroying this delocalized electron system, the ability to act as a charge acceptor gets additionally weakened.[68] Furthermore, the influence of the shell thickness on quenching rates and charge transfer between semiconductor nanoparticles and carbon nanotubes was investigated. Figure shows the emission spectra of 3 + 2 ML (Figure a,b) and 7 + 2 ML (Figure c,d) thick core/shell quantum dots that are attached to MWCNTs presented. Once again, the above-mentioned blue shift is also observed in these nanocomposites. Counterintuitively, we found that the photoluminescence quenching is enhanced with an increase in the passivating ZnS shell thickness, as shown in Figure e–g. This is in contrast to the minimization of quenching rates that are normally observed in core/shell QDs as a function of shell thickness.[69,70] Typically, one would assume that larger shells act as a physical barrier (like longer chain ligands), and with an increased distance between an emitter and a quencher, the rates for charge transfers get more improbable as tunneling and other energy transfers (e.g., Förster resonance energy transfer (FRET)) decrease with the distance. However, such behavior for core/shell particles was observed before in the literature and attributed to prolonged lifetimes of photogenerated charge carriers in core/shell structures with thicker shells.[71,72] Given that the Stern–Volmer constant is directly proportional to the lifetime, it appears that the enhanced probability of the quencher to diffuse toward and interact with a photoexcited emitter nanoparticle due to its long lifetime is largely responsible for the observed increased quenching rates. On top of that, it is worth mentioning that CISe/ZnS, CIS/ZnS, and CGS/ZnS are so-called type I core/shell structures, indicating that the valence band maximum (VBM) and the conduction band minimum (CBM) are lower and higher for the shell material than for the core material on a total energy scale, thus confining photogenerated electrons and holes mostly to the core region. Due to the quantum confinement effect, however, the band gap shrinks (VBM and CBM get higher and lower on a total energy scale) for larger particles. Transferred to core/shell structures with increasing shell thickness, this means that the energy barrier between the core and shell material shrinks, as it is described by Liu et al.[73] Such an effect would transform CuInS2/ZnS from a type I to a quasi-type II structure upon increasing the amount of ZnS (which was confirmed via DFT calculations, see Figure S15), making charge transfer more probable.

Figure 10

Normalized emission spectra of heterostructures constructed from cysteamine-capped CuInS2/ZnS nanoparticles and c-MWCNT with different oxidation times (a, b) and the corresponding Stern–Volmer plots (c).

Figure 11

Normalized emission spectra of CISe, CIS, and CGS with different shell thicknesses of 3 + 2 ML (a, b) and 7 + 2 ML (c, d) of ZnS (in powder form). Comparison of the Stern–Volmer plots of CISe (e), CIS (f), and CGS (g) nanoparticles with different shell thicknesses (capped with CYS and quenched with c-MWCNT in solution).

DFT Calculations

To further investigate the influence of electric fields originating from the charged substrates onto the nanoparticles and to understand what influence the binding between QDs and CNTs has on the electronic structure of the semiconductor material, we performed DFT calculations. Since theoretical studies in the form of DFT calculations on quantum-confined QDs consisting of the investigated chalcopyrite materials are scarce and to our best knowledge no such investigations have been performed on CISe/ZnS and CGS/ZnS core/shell nanostructures, we intended to improve the accuracy of these calculations to make them as comparable to experimental results as possible. Therefore, at a relatively low computational cost, the strong on-site Coulombic interaction of localized electrons, which is insufficiently described by LDA or GGA is corrected by an additional Hubbard-like term (Hubbard U parameter). Analogous to our previous studies,[74] the values for the U parameters were obtained semiempirically (see Figure S12) and were 4.8 eV for S, 4.5 eV for Se, 9.1 eV for Zn, 6.8 eV for Cu, 8.1 eV for In, and 9.8 eV for Ga. The improvement from the standard GGA calculations toward more accurate results with GGA + U calculations is represented in the Supporting Information (see Table S2 and also see Figures S13 and S14). Figure a–c shows the calculated band gaps for differently sized QDs and materials as a function of the applied electrical field. The relaxed structures that were the basis of these calculations can be seen in Figures S13 and S14. The electric field was applied along the z-axis and it was simulated by adding a saw-like potential to the bare ionic potential, as implemented in the Quantum ESPRESSO package. It should be mentioned that electric fields induced by point charges, lines of charges, and planes of charges can also be simulated with the help of the Environ plug-in for Quantum ESPRESSO.[75] Since the results follow the same trend, we decided to use the computational less demanding procedure via application of the saw-like potential. When the band gaps of the different materials are compared, they follow the same trend as experimentally determined (Eg: CuGaS2 > CuInS2 > CuInSe2). Also, the band gap increases with decreasing particle size as one would expect according to the quantum confinement effect.[76] Upon suspending the nanostructures to a positive external electric field, the band gap decreases for all materials and different sizes of the structures. For negative electric fields, the band gap increases slightly to a certain maximum after which it decreases again, and this effect is more pronounced in small structures and loses intenseness upon increasing the particle size. This observation is different from our previous studies on electric fields on ZnS QDs, and we would expect that the maximum of the band gap for perfectly passivated nanostructures is centered at E⃗(z) = 0.[74] This however is in contrast to the computational results, which means that the surfaces of the calculated nanostructures have a nonvanishing dipole density (like in our mentioned previous studies, ZnS with surface defects). Probably this is the result of the unpreventable asymmetric distribution of surface atoms in ternary materials when two opposing surfaces are compared. This nonvanishing dipole density decreases for larger structures as the differences between two opposing surfaces also become smaller, simply due to statistical reasons, and could even be completely compensated by constructing core/shell structures with a binary material like ZnS as a shell material (see Figure d). Figure e shows the calculated absorption spectra, with and without applied electric fields, for the smallest CIS QD and basically confirms what was discussed above. It should be mentioned that the time-dependent DFT calculations (TDDFT) for the absorption spectra are performed without the Hubbard U correction, as this possibility is not implanted so far in the Quantum ESPRESSO package. Therefore, the energy of the first excitonic peak (which corresponds to the band-gap energy) is shifted compared to the other calculated band gaps. Nevertheless, the trend is the same. To answer the question now whether the observed blue shift is a result of the electrostatic interaction between the single components, one has to look at both theoretical and experimental results. The simulations suggest an increase of the band gap for cases with nonvanishing dipoles (like the ternary core structures or core–shell structures with surface defects) and negative applied electric fields and vice versa, a decrease of the band gap with positive electric fields. From these results alone it would be plausible to blue shift the emission wavelength by coupling quantum dots to a charged substrate. However, the simulations also suggest that this blue shift should increase upon increasing the field strength and also a red shift should be observed for an oppositely charged substrate. Experimentally, this is not what we found, as shown in Figure a,b; for increased surface charges of the MWCNTs, the blue shift did not increase further. Second, using oppositely charged MWCNTs, as shown in Figure a–d, did not result in a red shift.

Figure 12

Electric field dependency of the band gap for different sized nanoparticles based on CISe (a), CIS (b), CGS (c), as well as core–shell/structures (d) and calculated absorption spectra of CIS QDs subjected to different field strengths (e).

Furthermore, the actual physical binding between the nanoparticles and the carbon support was also investigated, which is shown in Figure . Three structures are therefore simulated. First, a CIS quantum dot capped with cysteamine as a ligand (Figure a). Second, a c-CNT that binds via the functional −COO– group directly to the core quantum dot (Figure b). Third, a c-CNT where the −COO– group binds to the −NH3+ group of the cysteamine ligand (Figure c). No significant change in the density of states (DOS) for the nanocrystal core is observed, as shown in Figure d (red line corresponds to the core DOS; the influence on larger quantum dots or core/shell structures would be even smaller, thus was not considered here). This means that also other effects arising from the coupling of the binary components of the nanocomposites do not seem to have an influence on the electronic structure. Thus, we reach the conclusion that the observed change in the optical spectra is not related to induced electric fields alone. An explanation in accordance with all experimental and theoretical data would be that the blue shift is caused due to Förster resonance energy transfer (FRET). FRET is the nonradiative transfer of energy from a donor to an acceptor.[77] This energy transfer is sensitively dependent on the distance (see eq ).In eq , kFRET is the rate of the FRET, kD is the emission rate of the donor, R0 is the Förster distance, and r is the distance between the donor and the acceptor. In this study, in the case of the investigated solid samples (aggregated QDs), the emission wavelength was red-shifted due to a very close contact between single particles and possible energy transfer between QDs. In solution or fixed on the surface of MWCNTs, the interparticle distance is much larger (see again Figure c); thus, it seems reasonable that energy transfer between semiconductor nanoparticles is inhibited in these cases and therefore not red-shifted (or as we interpreted it, is observed as a blue shift). This would also explain why we do not observe this blue shift in nanocomposites formed from unstable dispersions or with the same surface charge of the single components, as the QDs agglomerate there to yield small particle distances (see again Figure a,b). Similar behavior was observed in the literature when QDs in their powder form were compared to QDs encapsulated in a polystyrene matrix.[78] These findings highlight once again the necessity to precisely control the aggregation behavior through surface functionalization to be able to produce nanocomposites with defined and controlled structures.

Figure 13

Relaxed structures of cysteamine-functionalized CuInS2 QDs (a) and carboxylic-functionalized CNT binding to those nanostructures (orange: Cu, brown: In, yellow: S, white: Hx or H, black: C, red: O, blue: N) directly as a passivating ligand (b) or electrostatically to the amine group of the cysteamine (c) and the corresponding DOS (d).

Conclusions

In summary, the construction of nanocomposites consisting of MWCNTs and ternary chalcogenide QDs is demonstrated via electrostatic self-assembly. Through easy and reproducible surface modification of CISe/ZnS, CIS/ZnS, and CGS/ZnS core/shell nanoparticles, together with already well-established functionalization protocols for CNTs, a toolbox for the synthesis of various heterostructures is assembled. A toolbox means access to the single components with positive or negative surface charges and the possibility to influence the magnitude of those, as well as the option to modify the structures (e.g., increase shell thicknesses or use longer chain ligands as a spacer). Thus, different variants of QDs and MWCNTs were obtained, which depending on the surface properties adjusted before were able to self-assemble into nanocomposites. We herein report the synthesis of high-quality CuInSe2-, CuInS2-, and CuGaS2-based QDs capped with ligands like AOT and CYS and the electrostatically driven construction of nanocomposites from these together with oppositely charged MWCNTs, which to our best knowledge has not been reported before. ζ-Potential and DLS size analysis were performed to specify the surface charge and colloidal stability of the single components, which were crucial to construct nanocomposites where the nanotubes show a homogeneous surface coverage with nanocrystals. Steady-state and time-resolved photoluminescence studies were examined to investigate the quenching behavior and kinetics. During these characterizations, we found a blue shift of emission and excitation maxima, which was reported before for other nanocomposites. The origin of this shift was not well researched and there were assumptions that it could be caused by the electrostatic interaction between the binary components of the composites. From the experimental site with the established toolbox and from the theoretical site supported through DFT calculations, we were able to exclude that theory and concluded that the observed shift in energy is more likely due to FRET. Nevertheless, this change in the optical properties apart from its origin indicates a good contact between the single components, which is usually preferred in heterostructures tuned toward optoelectronic applications.