Grzegorz
Gabka
a,
Piotr
Bujak
*a,
Jan
Żukrowski
b,
Damian
Zabost
a,
Kamil
Kotwica
a,
Karolina
Malinowska
c,
Andrzej
Ostrowski
a,
Ireneusz
Wielgus
a,
Wojciech
Lisowski
d,
Janusz W.
Sobczak
d,
Marek
Przybylski
be and
Adam
Pron
a
aFaculty of Chemistry, Warsaw University of Technology, Noakowskiego 3, 00-664 Warsaw, Poland. E-mail: piotrbujakchem@poczta.onet.pl
bAcademic Centre for Materials and Nanotechnology, AGH University of Science and Technology, 30-059 Kraków, Poland
cFaculty of Chemistry, University of Warsaw, Pasteura 1, 02-093 Warsaw, Poland
dInstitute of Physical Chemistry, Polish Academy of Science, Kasprzaka 44/52, 01-224 Warsaw, Poland
eFaculty of Physics and Applied Computer Science, AGH University of Science and Technology, 30-059 Kraków, Poland
First published on 5th May 2016
It is demonstrated that ternary Cu–Fe–S nanocrystals differing in composition (from Cu-rich to Fe-rich), structure (chalcopyrite or high bornite) and size can be obtained from a mixture of CuCl, FeCl3, thiourea and oleic acid (OA) in oleylamine (OLA) using the heating up procedure. This new preparation method yields the smallest Cu–Fe–S nanocrystals ever reported to date (1.5 nm for the high bornite structure and 2.7 nm for the chalcopyrite structure). A comparative study of nanocrystals of the same composition (Cu1.6Fe1.0S2.0) but different in size (2.7 nm and 9.3 nm) revealed a pronounced quantum confinement effect, confirmed by three different techniques: UV-vis spectroscopy, cyclic voltammetry and Mössbauer spectroscopy. The optical band gap increased from 0.60 eV in the bulk material to 0.69 eV in the nanocrystals of 9.3 nm size and to 1.39 eV in nanocrystals of 2.7 nm size. The same trend was observed in the electrochemical band gaps, derived from cyclic voltammetry studies (band gaps of 0.74 eV and 1.54 eV). The quantum effect was also manifested in Mössbauer spectroscopy by an abrupt change in the spectrum from a quadrupole doublet to a Zeeman sextet below 10 K, which could be interpreted in terms of the well defined energy states in these nanoparticles, resulting from quantum confinement. The Mössbauer spectroscopic data confirmed, in addition to the results of XPS spectroscopy, the co-existence of Fe(III) and Fe(II) in the synthesized nanocrystals. The organic shell composition was investigated by NMR (after dissolution of the inorganic core) and IR spectroscopy. Both methods identified oleylamine (OLA) and 1-octadecene (ODE) as surfacial ligands, the latter being formed in situ via an elimination–hydrogenation reaction occurring between OLA and the nanocrystal surface.
Over the past five years, research on tin-containing quaternary nanocrystals such as quaternary semiconductors Cu2ZnSnS4 has intensified as possible replacements for indium-containing materials and also due to their interesting physical properties. Nanocrystals of kesterite (band gap of bulk material = 1.5 eV) show high absorption coefficients in the solar spectrum range and for this reason have been tested with success in photovoltaic cells.13,14 Their high Seebeck coefficient values should also be underlined, which lead to thermoelectric applications.15
The crystal structure of CuFeS2 was first reported by Burdick and Ellis16 as an ZnS-type but later Pauling and Brockway17 demonstrated the existence of the chalcopyrite structure (space group I2d). CuFeS2 is an antiferromagnetic semiconductor with a very small band gap (0.6–0.7 eV) and for this reason exhibits interesting electrical, optical and magnetic properties.18,19 A combination of magnetism and electronic transport properties makes Fe–Cu–S nanocrystals suitable materials for spintronics applications.20,21 Nanocrystals of a ternary small band gap semiconductor CuFeS2 also show enhanced thermoelectric properties when compared to their bulk form.22 Recently, ternary Cu–Fe–S nanocrystals with different compositions have been discussed as new generation plasmonic materials.23
Surprisingly, only few methods enabling the preparation of reduced size CuFeS2 nanocrystals have been reported. To date, the smallest nanoparticles (6.4 ± 0.5 nm) were obtained by injection of sodium diethyldithiocarbamate to a mixture of CuCl2, FeCl3 and oleic acid (OA) in 1-dodecanethiol (DDT). The optical band gap determined for these nanocrystals was 1.2 eV, significantly smaller than the band gap of the bulk semiconductor (ca. 0.6 eV), which could be considered as a manifestation of the quantum confinement effect.22 Larger size nanocrystals with spherical (12 ± 4 nm) or pyramidal (30 ± 5 nm) shapes were obtained by hot-injection of a sulfur solution in trioctylphosphine (TOP) to a mixture of copper(I) diethyldithiocarbamate and iron(III) diethyldithiocarbamte in a mixture of OA and dichlorobenzene (DCB).24 Two more recent studies described the preparation of much larger nanoparticles, which could be considered as being on the borderline of nano and bulk materials.25,26
In this study, we present a new, simple heating up method leading to the smallest Cu–Fe–S nanocrystals ever reported (from 2 to 3 nm). We demonstrate that by changing the reaction mixture composition, we are able to controllably change the composition of the resulting nanocrystals from Cu-rich to Fe-rich. Moreover, by applying the recently developed method of initial ligands recovery, we unequivocally identify them by NMR spectroscopy.27 Finally, for nanocrystals of the same composition and different sizes, we demonstrate a clear quantum confinement effect through a set of spectroscopic and electrochemical measurements.
Sample | Cu/Fe/S/OAa | Composition | Structure | Size (nm) |
---|---|---|---|---|
a Molar ratio of the precursors. | ||||
B1 | 1.0/1.0/2.0/3.9 | Cu1.92Fe1.00S2.05 | Chalcopyrite | 2.9 ± 0.4 |
B2 | 1.0/1.0/2.0/2.1 | Cu1.62Fe1.00S2.01 | Chalcopyrite | 2.7 ± 0.3 |
B3 | 1.0/1.0/2.0/1.2 | Cu1.64Fe1.00S2.04 | Chalcopyrite | 9.3 ± 1.7 |
B4 | 4.0/1.0/3.5/14.3 | Cu4.20Fe1.00S3.20 | High bornite | 1.5 ± 0.4 |
B5 | 1.0/4.0/6.5/26.0 | Cu1.00Fe1.79S2.27 | Chalcopyrite | 2.3 ± 0.4 |
In the successful preparation of Cu–Fe–S nanocrystals, the selection of ligands (OA) and a highly coordinating solvent (OLA) in combination with appropriate reaction conditions turned out to be crucial. In particular, by application of the heating-up method, chalcopyrite-type Cu–Fe–S nanocrystals could be obtained at 180 °C from a mixture of CuCl, FeCl3, thiourea and OA in OLA.
B1 nanocrystals (Cu1.92Fe1.00S2.01) were spherical in shape and very small (diameter of 2.9 ± 0.4 nm). B2 and B3 nanocrystals of very similar composition showed however a different morphology. The former, spherical- or cubic-type in shape were of slightly smaller size than the B1 nanocrystals (diameter of 2.7 ± 0.3 nm), the latter were much larger and tetrahedral in shape (edge size of 9.3 ± 1.7 nm) (Table 1, Fig. 2). Both shapes were isotropic yielding the aspect ratio of 1.
Thus, the concentration of ligand molecules in the reaction mixture has a profound effect on the nanocrystals size. For higher concentrations, this effect was very small, below a certain OA to precursor ratio, an abrupt increase in the nanocrystals size was observed.
By varying the metal to precursor ratio in the reaction mixture, it was possible to obtain either Cu-rich or Fe-rich nanocrystals (see Fig. S1, ESI† for the corresponding Cu, Fe and S EDS spectra and Table 1 for nanocrystals compositions derived from these spectra). For the nanocrystals prepared with a Cu:
Fe ratio = 4.0
:
1.0, the increase in the copper content was very pronounced (B4, composition: Cu4.20Fe1.00S3.20). A weaker effect was observed for the inversed precursor ratio (Cu
:
Fe = 1.0
:
4.0), i.e. in the resulting nanocrystals, a smaller increase in the iron content was observed (B5, composition: Cu1.00Fe1.79S2.27). This finding indicates the higher reactivity of CuCl when compared to FeCl3 under the applied conditions of nanocrystal nucleation and growth.
Similarly as all the Cu–Fe–S nanoparticles prepared with high OA to precursor ratios, the B4 and B5 nanocrystals were spherical in shape and small (diameters of 1.5 ± 0.4 nm and 2.3 ± 0.4 nm for B4 and B5 nanocrystals, respectively), as evidenced from their TEM images presented in Fig. 3 and 4. In the same figures, their X-ray diffractograms are shown. B4 nanocrystals revealed a different crystal structure than the other four batches studied with a set of reflections at 27.0°, 28.6°, 32.8°, 34.2°, 47.1° and 55.9° corresponding to the planes indexed as (311), (222), (400), (331), (440) and (622) in the high bornite structure of Cu5FeS4 (JCPDS 34-0135). B5 nanocrystals, similarly as those of B1–B3 batches showed the structure of chalcopyrite (JCPDS 37-0471).
![]() | ||
Fig. 3 X-ray diffractogram (high bornite structure, JCPDS 34-0135) and TEM image (and enlarged TEM image) of the B4 nanocrystals. |
![]() | ||
Fig. 4 X-ray diffractogram (chalcopyrite structure, JCPDS 37-0471) and TEM image (and enlarged TEM image) of the B5 nanocrystals. |
![]() | ||
Fig. 5 High-resolution Cu2p, Fe2p and S2p XPS spectra of B2 (blue) and B3 (red) Cu–Fe–S nanocrystals. |
The S2p spectra of the B2 and B3 nanocrystals differ in a more pronounced manner. In both cases, two peaks can be distinguished at 162.2 and 168.7 eV, attributable to the sulfide form (S2−)31 and oxidized forms of sulfur (SOx2−, x = 3, 4),33 respectively. For the B2 nanocrystals, which are smaller in size than the B3 ones, the relative intensity of the peak ascribed to the oxidized forms of sulfur is higher. Partial oxidation of surfacial sulfur atoms is a common phenomenon in metals sulfide nanocrystals33–35 and in some cases can be desirable. For example, the presence of oxidized forms of sulfur on the surface of PbS nanocrystals used for the fabrication of photodetectors resulted in an improvement in device performance.33
As already stated, the Fe2p XPS spectra seem to indicate the co-existence of Fe(III) and Fe(II) in the B2 and B3 nanocrystals. It is tempting to verify this conclusion by 57Fe Mössbauer spectroscopy since this is the most sensitive technique used for the detection of non-equivalent forms of iron.
![]() | ||
Fig. 6 57Fe Mössbauer spectra of the Cu–Fe–S nanocrystals (batches B2 and B3), registered at varying temperatures (6, 10, 16 and 78 K). |
At sufficiently low temperature, i.e., at 6 K, the magnetic component of the broadened lines dominates the spectra and the quadrupole doublet practically disappears. Actually, there are two possible interpretations of the observed phenomenon:
(a) B2 and B3 nanoparticles are ferromagnetic/ferrimagnetic, however, they show very strong superparamagnetism, i.e. very low blocking temperature, which is not surprising for so small nanocrystals (2.7–9.3 nm),
(b) B2 and B3 nanoparticles are antiferromagnetic showing a macroscopic quantum effect for the slowly relaxing macrospins of the magnetic sublattices.44
Basically, both the B2 and B3 nanoparticles could be magnetic (ferro-, ferri, antiferro- or antiferri-) of very low Curie/Neel temperature and therefore known as paramagnetic (i.e., never measured below liquid nitrogen temperature). Line broadening in this case could be a result of different Fe sublattices and substituting Cu, and thus forming different chemical surroundings contributing differently to the measured Mössbauer spectra. However, the spectra shown in Fig. 6 do not show any continuous transition from the Zeeman sextet to the quadrupole doublet via decreased splitting with increasing temperature. Therefore, such interpretation does not seem to be valid. Even then it is difficult to propose a final interpretation since the spectra measured for both ferromagnetic (β-CuFeS2) and antiferromagnetic (α-CuFeS2) nanoparticles are expected to be qualitatively similar. Note that ferro/ferri and antiferro/antiferri states cannot be distinguished from the Mössbauer spectra even for the bulk material if the spectra are measured with no external magnetic field applied.
More precisely, in the case of ferromagnetic nanoparticles, the collapse of a low temperature well-resolved hyperfine magnetic structure in the spectrum into a quadrupole doublet should be accompanied by a line broadening at intermediate temperatures. Moreover, superparamagnetic spectra obtained at the same temperature should be strongly dependent on the size of the ferromagnetic nanoparticles. This is not the case (see Fig. 6): almost the same spectra are measured at 10 K for both the 2.7 and 9.3 nm nanoparticles. The spectral lines remain “narrow” and only the relative contribution of the well-resolved magnetic structure and the quadrupole doublet/doublets changes rapidly with increasing temperature. This may suggest the presence of antiferromagnetic α-CuFeS2 (chalcopyrite) in these extremely small size nanocrystals, which leads to a macroscopic quantum effect. The quantum nature of the effect is concluded from the well-defined energy states (resulting from the quantum confinement phenomena for electrons in nanoparticles) of populations, which are temperature dependent.44 In the case of ferromagnetic particles, the ground state represents a quasi-continuous spectrum independent of the anisotropy constant, whereas it is strongly anisotropy dependent in the case of the antiferromagnetic nanoparticles (small anisotropy allows the coupling between sublattice magnetizations to be strong).
Considering the antiferromagnetic state is observed only at very low temperatures, one should also remember that the Neel temperature is usually drastically suppressed with decreasing size of the antiferromagnetic particles (the finite-size effect).45
In Fig. 7, the Mössbauer spectra obtained at 78 K for B2 and B3 are compared. The doublet of the broadened spectral lines can be fitted using four components of equal contributions exhibiting slightly different splitting parameters (see Fig. 7). This type of fit helps to obtain the hyperfine parameters precisely. The measured isomer shift (δ) is the same for the nanocrystals of both batches (0.442 mm s−1vs. α-Fe), whereas the quadrupole splitting (Q.S.) values slightly differ, being 0.78 mm s−1 and 0.82 mm s−1 for the B2 and B3 nanocrystals, respectively. This is not surprising since the δ reflects the chemical composition, which is independent of the particle diameter, whereas the Q.S. reflects symmetry of charge distribution, which could vary for the different size of nanoparticles.
The analysis of δ and Q.S. can be helpful in phase identification. Isomer shift values determined from the Mössbauer spectra of the B2 and B3 nanocrystals are very high for Fe(III) and very low for Fe(II),41 implying an intermediate iron oxidation state, which can be achieved though the fast electron exchange between these two form of iron.40 Very similar Mössbauer parameters were recently reported for nanocrystals of isocubanite (known as non-magnetic but never magnetically tested at very low temperatures).46 The chemical composition of Cu-rich B2 and B3 nanocrystals (Cu1.62Fe1.00S2.01 and Cu1.64Fe1.00S2.04, respectively) is however very different from that of a Fe-rich cubanite (CuFe2S3).
From the temperature dependence of the Mössbauer spectrum parameters, using eqn (1) and (2), it is possible to calculate the Debye temperature, θD, characterizing the lattice dynamics of the nanocrystalline materials studied.
δ(T) = δ0 + δSOD(T) | (1) |
![]() | (2) |
Fig. 8 shows the temperature dependence of δ, determined from the Mössbauer spectra of the B2 and B3 nanocrystals registered at different temperatures. The obtained Debye temperature values (432 K and 419 K for B2 and B3 nanocrystals, respectively) are higher by 70 to 150 K than those determined experimentally for bulk Cu–Fe–S crystals,47 including the results obtained by Mössbauer spectroscopy.38 Theoretical calculations also lead to values below 300 K.48 Note that consistent with these findings, the value of θ obtained for the smaller B2 nanocrystals is higher than that found for the larger B3 nanocrystals. An increase in the Debye temperature with decreasing nanocrystals size was also observed for binary semiconducting CdS49 and CdTe50 nanocrystals. In contrast, for PbS nanocrystals, a different behavior was reported, namely, the bimodal dependence of the Debye temperature on the nanocrystals size.51 With decreasing nanocrystal size, the Debye temperature initially increased, as in the case of CdS and CdTe, but below a given “critical” size, it suddenly started to decrease.
![]() | ||
Fig. 8 Temperature dependences of δ measured for B2 and B3 nanocrystals together with the best fit to eqn (1) and (2). |
In the identification of the initial surfacial ligands in ternary Cu–Fe–S nanocrystals, we applied the method previously elaborated in our group.27 It consists of the recovery of surfacial ligands through selective dissolution of the inorganic core with HCl. Fig. 9 shows the 1H NMR spectra of the ligands recovered from the B2 and B3 nanocrystals together with the spectra of pure 1-octadecene (ODE), oleic acid (OA) and oleylamine hydrochloride (OLA-HCl), which are added for comparative reasons. In the spectrum of the organic part of the B2 batch, a clear signal ascribed to OLA-HCl can be distinguished at 5.4 ppm and 2.9 ppm, the former originating from the vinyl protons (–CHCH–) and latter from protons of the methylene group adjacent to the protonated amino group (–C
NH3+). No evidence of the presence of oleic acid as a ligands can be found in the spectrum since its diagnostic signal, originating from protons of methylene group adjacent to the carboxylic one (C
COOH), cannot be found at 2.3 ppm. In addition, in the spectral range of 4.5–6.0 ppm, very weak signals attributed to ODE can be identified. In the spectrum of the organic part of the B3 nanocrystals, the same set of signals is present; however, the contribution of the ODE signals is more significant.
The IR spectra of the recovered organic shell of the synthesized nanocrystals (Fig. S3 of ESI†) are perfectly consistent with the 1H NMR results. In addition to the bands at 2921 and 2861 cm−1 attributed to the C–H stretching vibrations in the methylene groups of the aliphatic chains bands, characteristic of OLA (at 1628 cm−1 and 1575 cm−1) and ODE (1462 cm−1, 1401 cm−1 and 905 cm−1) can be distinguished.
From the abovementioned results, it can be concluded that between the two components of the reaction mixture, which contain coordinating groups (OA and OLA), only OLA is bound to the nanocrystal's surface as an initial ligand. The presence of ODE requires some comments since in the elaborated preparation procedure, this chemical is not added to the reaction mixture. Therefore, it has to be formed in situ from OLA via an elimination–hydrogenation reaction occurring between OLA and the nanocrystal surface. This reaction is strongly dependent on the concentration of oleic acid (OA). In the preparation of the B3 nanocrystals, whose spectrum shows a higher amount of ODE than that of B2, a significantly smaller concentration of OA was used (see Table 1 and Table S1, ESI†).
In situ generation of ODE has already been observed in the preparation of cobalt nanocrystals involving OLA as a solvent53,54 and in the case of the synthesis of kesterite-type Cu2ZnSnS4 nanoparticles.55
The hydrodynamic diameter is another parameter characterizing nanocrystals capped with organic ligands. It describes the effective diameter of a nanocrystal, which includes the inorganic core and ligands coordinated to it. By comparison with the nanoparticle diameter determined from the TEM images, it provides information concerning the thickness of the organic shell. Moreover, its evolution in time can be considered as a test of the colloidal stability of the investigated nanocrystals. Dynamic light scattering (DLS), widely used in the studies of hydrodynamic diameters of macromolecular compounds can also be applied in the case of nanocrystals stabilized with organic ligands.56,57 In Fig. 10 the histograms of hydrodynamic diameters determined by DLS are shown for the B2 and B3 nanocrystals dispersions in chloroform. For comparison, their average diameters obtained by statistical treatment of the TEM images are indicated. In the case of the B2 nanocrystals, the obtained DLS results are very consistent with the TEM images; their hydrodynamic diameter (6–9 nm) is more than two times larger than the average inorganic core diameter (2.7 ± 0.3 nm). This is caused by the fact that in the dispersion in chloroform, not only ligands but also some solvent molecules, penetrating the organic shell of the nanocrystals, constitute additional components in the coordination sphere. The obtained hydrodynamic diameter values of the B2 nanocrystals are also consistent with the values determined by DLS for other families of nanocrystals of comparable core and ligand size.56,57 The colloidal solutions were stable as demonstrated by repeated DLS measurements, carried out within few days of each other, which did not show an increase in their hydrodynamic diameter.
For the B3 nanocrystals, the measured hydrodynamic diameter is unusually large, 10 to 20 nm larger than the diameter of the inorganic core. Such a big difference cannot be justified by the thickness of the organic shell, even if the penetration of the solvent to it is considered. It is therefore highly probable that in chloroform solution, the B3 nanocrystals form small agglomerates composed of a few nanoparticles.
The absorption edge, as determined from the (Ahν)0.5vs. hν dependence (see Fig. 12), yields an optical band gap of 0.69 eV for the B3 nanocrystals (9.3 ± 1.7 nm), close to the value measured for bulk CuFeS2.32 For the smaller B2 nanocrystals (2.7 ± 0.3 nm) the optical band gap increases to 1.39 eV, indicating the quantum confinement effect.
The band gaps, Eg, were calculated using the (Ahν)0.5vs. hν relationship (where A = absorbance, h = Planck's constant and ν = frequency). In the majority of publications devoted to inorganic semiconductors, a direct optical band gap is estimated from the (Ahν)2vs. hν relationship. This method is less precise and ambiguous. Detailed justification for the preferable use of (Ahν)0.5vs. hν relationship can be found in the work of Markus Meinert and Günter Reiss.63
EHOMO = −(Eox + 5.1) (eV) | (3) |
ELUMO = −(Ered + 5.1) (eV) | (4) |
The electrochemical data are collected in Table 2, together with the determined HOMO and LUMO energies and the electrochemical and optical band gaps. The electrochemical band gaps (0.74 eV and 1.54 eV for B3 and B2, respectively) are slightly higher than the optical ones (0.69 eV and 1.39 eV). This is caused by the additional Coulombic repulsion energy of the exciton electron and hole, which has to be taken into account during the electrochemical measurements.65
In the case of chalcopyrite Cu–Fe–S nanocrystals, the HOMO can be considered as a combination of Cu3d–S2p, whereas the LUMO can be consider as the Fe3d–S2p orbitals.32 The electrochemical data show that the HOMO in the B2 nanocrystals is lowered by ΔE = 0.15 eV when compared to the case of the B3 nanocrystals due to the quantum confinement effect, whereas the LUMO is increased by ΔE = 0.65 eV. The more pronounced effect of quantum confinement on the LUMO energy has already been reported for chalcopyrite-type CuInS2 nanocrystals.67,68 This effect can be rationalized by the differences in the effective mass of electrons (me*) and holes (mh*). Considering the 1S(e) (LUMO) and 1S(h) (HOMO) levels as functions of 1/me* and 1/mh* (where me* < mh*), the LUMO energy has to be affected in a more pronounced manner by the diminishing nanocrystal size when compared to the HOMO.
Footnote |
† Electronic supplementary information (ESI) available: Detailed information on nanocrystals preparation and characterization, EDS, XPS, IR spectra and cyclic voltammograms. See DOI: 10.1039/c6cp01887d |
This journal is © the Owner Societies 2016 |