Langmuir 1997, 13, 827-832
827
Sphalerite-Wurtzite Intermediates in Nanocrystalline CdS Walter Vogel* and Joachim Urban Fritz-Haber-Institut der Max-Planck-Gesellschaft, D-14195 Berlin, Germany
Manisha Kundu and S. K. Kulkarni Department of Physics, Center for Advanced Studies in Material Science and Solid State Physics, University of Pune, Pune 411007, India Received April 30, 1996. In Final Form: November 7, 1996X Cadmium sulfide nanoparticles have been synthesized using mercaptoethanol as a capping agent. A refined analysis of the powder X-ray diffraction by Debye function analysis shows that the as synthesized material consists of both ∼50 wt % hexagonal wurtzite-type nanoparticles of about 2 nm and larger cubic sphalerite-type particles of about 2.6 nm, respectively. Both species have also been identified by highresolution electron microscopy, some of them show multiply twinning with 5-fold decahedral symmetry. The thermal stability of these nanoparticles is surprisingly high under He. The total mass-averaged size increases from 2.3 nm (as synthesized) to 3.2 nm after exposure to 315 °C in helium for 12 min. However, both polymorphs are still present. Heavy sintering occurs only above ∼380 °C, leading to wurtzite particles of 41 nm on average. These particles are strongly faulted by hexagonal (001) stacking faults (faulting probability ) 1.1%) but show negligible twin defects. After synthesis, the particles show the optical absorption peaks at ∼400 nm. The related optical band gap of 3.1 eV is in excellent agreement with the theoretical tight-binding approximation of Lippens and Lannoo for CdS particles of this size.
Introduction For quite some time it has been known that the particle size affects the electronic structure of materials below a critical size, which depends mainly on the type of the material. Size effects may be caused by the surface itself or are due to the reduced number of atoms in a particle. In semiconductor materials, the size effect is observable when the particle diameter is near or less than the Bohr diameter of exitons in the corresponding bulk material. An increase in the energy gap can be observed when the particle size decreases.1,2 In most cases, however, the size quantization effect in semiconductors is readily observable when the particle size is less than 10 nm. In II-VI semiconductors, the energy gap is direct. The assumption is being made that the smallest particles of semiconductors like CdS and ZnS are of cubic structure and form the basis for band structure calculations.1,3,4 Nanometer-sized “quantum dot” II-VI semiconductors have been precipitated in continuous glass matrices used as optical filters and characterized by high-resolution electron microscopy (HREM), electron diffraction, and X-ray diffraction (XRD).5-7 A straightforward analysis XRD using the Scherrer formula, common for micrometersized particles, however, becomes questionable for nanometer-sized particles as the number of atoms per cluster becomes exceedingly small (e103 atoms/cluster). It is therefore necessary to analyze nanoparticles of semiconductors by using more rigorous techniques. Synthesis of monodisperse self-assembling CdE (E ) S, Se, Te) nanocrystallites was recently achieved by the MIT X Abstract published in Advance ACS Abstracts, February 1, 1997.
(1) Brus, L. E. J. Chem. Phys. 1984, 80, 4403. (2) Wang, Y.; Herron, N. J. Phys. Chem. 1991, 95, 523. (3) Lippens, P. E.; Lannoo, M. Phys. Rev. 1989, B39, 10935. (4) Nair, S. V.; Sinha, S.; Rustogi, K. C. Phys. Rev. 1987, B35, 4098. (5) Liu, L. C.; Kim, M. J.; Risbud, S. H.; Carpenter, R. W. Philos. Mag. 1991, B63, 769. (6) Liu, L. C.; Risbud, S. H. J. Appl. Phys. 1990, 68, 28. (7) Borelli, N. F.; Hall, D. W.; Holland, H. J.; Smith, D. W. J. Appl. Phys. 1987, 61, 5399.
S0743-7463(96)00426-X CCC: $14.00
group of Bawendi and co-workers.8,9 They used numerical computer simulations of the X-ray powder diffraction to characterize the nanocrystallite structural features. In the present study the CdS nanocrystals are non-monodisperse. This makes numerical simulation techniques more complex. The so-called Debye function analysis (DFA) has shown both, i.e., the determination of the size distributions and the structural features of highly dispersed metal catalysts10-12 or ligand stabilized organometallics,13,14 respectively. In the present study this technique was employed for the first time to characterize polydispersed semiconducting CdS nanocrystals and to examine their thermal stability with the help of in-situ XRD. HREM will be applied as a complementary technique. Experimental Section Synthesis. CdS nanoparticles have been synthesized by aqueous chemical methods described elsewhere.15-17 In brief, the procedure involves capping of CdS nanoparticles by an organic material during chemical synthesis. In this case mercaptoethanol (C2H5OSH) was used as a capping agent for synthesis. It is assumed that the hydrogen is removed in the reaction and mercaptoethanol bonds with Cd. Synthesis involves reduction of CdCl2 (0.01 M) with mercaptoethanol and rigorous stirring while mixing the two solutions. Na2S (0.01 M) is added to this solution dropwise. All the solutions are made in water, and reactions are carried out under nitrogen atmosphere to prevent oxidation. After the reaction is completed, the colloid (8) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706. (9) Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Science 1995, 270, 1335. (10) Gnutzmann, V.; Vogel, W. J. Phys.Chem. 1990, 94, 4991. (11) Vogel, W.; Sachtler, W. M. H.; Zhang, Z. Ber. Bunsenges. Phys. Chem. 1993, 97, 280. (12) Hartmann, N.; Imbiehl, R.; Vogel, W. Catal. Lett. 1994, 28, 373. (13) Vogel, W.; Rosner, B.; Tesche, B. J. Phys. Chem. 1993, 97, 11611. (14) Vogel, W.; Duff, D. G.; Baiker, A. Langmuir 1995, 11, 401. (15) Kundu, M.; Khosravi, A. A.; Singh, P.; Kulkarni, S. K. J. Mater. Sci., in press. (16) Khosravi, A. A.; Kundu, M.; Kuruvilla, B. A.; Shekhavat, G. S.; Gupta, R. P.; Sharma, A. K.; Vyas, P. D.; Kulkarni, K. S. Appl. Phys. Lett. 1995, 67, 2506. (17) Nosaka, Y.; Yamagushi, Miyama, H.; Hayashi, H. Chem. Lett. 1988, 605.
© 1997 American Chemical Society
828
Langmuir, Vol. 13, No. 4, 1997
Vogel et al.
Figure 1. Optical absorption spectrum of the as synthesized nano-CdS. The maximum at λ ) 400 nm corresponds to a band gap of Eg ∼ 3.1 eV.
Figure 3. Corrected intensity profile of as synthesized nanoCdS, compared with best fits using Debye functions of a sequence of cubic clusters (dotted), hexagonal clusters (dash-dotted), and a combination of both (solid line).
Figure 2. Debye functions of a CdS nanocrystal consisting of 216 molecules in the cubic and hexagonal arrangement. The figure demonstrates the diffraction effects caused by one and two (001)-twin planes introduced into the wurtzite-type model cluster. is centrifuged and the precipitate is repeatedly washed in distilled water. The thermal stability of the colloid was tested by differential thermogravimetry (DTG). Optical absorption spectra are taken using a Hitachi 303 double beam spectrophotometer. It shows an excitonic peak at ∼400 nm corresponding to an optical band gap of 3.1 eV and is obviously blue shifted (Figure 1). Bulk CdS absorption peaks at 530 nm and should be broader. It has also been noticed that the nanoparticles synthesized by chemical methods showed redshifted luminescence. It was therefore assumed that defects or crystal faults are responsible for this observation. Due to the overlap of the diffraction peaks, it was so far impossible to distinguish between cubic and hexagonal structures. X-ray Diffraction. The X-ray diffraction patterns were measured with a Guinier powder diffractometer (Huber) using Cu KR1 radiation in 45° transmission. For the high temperature in situ measurements the sample was fixed between two 0.1 mm Be platelets in a reaction cell described elsewhere.12 The patterns were corrected for background scattering and the usual angular factors. For the DFA simulation, a set of Debye functions of model clusters (CdS)N, was used which either showed cubic sphalerite or hexagonal wurtzite structure, respectively. In a least-squares fit of nonlinear parameters (Marquardt algorithm) these Debye functions are added, using the respective number fractions as free parameters. Additional free parameters are the DebyeWaller factors of both cluster types and the lattice spacing. A detailed description of the DFA method is given elsewhere.10-14 Figure 2 gives an example of the spherically averaged diffracted intensity (Debye function) of a CdS nanocrystal with N ) 216, arranged as cubic ZnS type, and as hexagonal wurtzite type, the latter containing none, one, and two (001)-twin planes, respectively.
In this study the intensity, contributed by the mercaptoethanol ligands was not taken into account for several reasons: (i) The exact ligand positions are unknown, which is why their contribution cannot be determined. (ii) Their total intensity contribution amounts to 15% in relation to the total intensity of the CdS nucleus. This number holds provided that every Cd atom is linked to a C2H5OS group. However, only a certain fraction of Cd atoms, depending on the cluster size, is linked to a C2H5OS group. (iii) The ligand contribution is expected to be rather unstructured; i.e., it can be treated as a diffuse background. However, for even smaller particles of this type the ligand contribution will no more be negligible. Electron Microscopy. In addition to the X-ray diffraction experiments, a structural characterization of the samples was performed by means of HREM. Three different samples were investigated: (a) CdS as prepared (sample 1); (b) CdS annealed at 225 °C under He (sample 2); (c) CdS annealed at 600 °C under He (sample 3). All samples were dispersed in methanol and deposited on an amorphous carbon film of about 5 nm thickness. A Philips CM200 FEG microscope, 200 kV, equipped with a field emission gun was used. The coefficient of spherical aberration was Cs ) 1.35 mm. The magnification was 389000×, which was calibrated by an Au cluster sample (PLANO). The images were digitized in sizes of 256 × 256 pixels with a pixel size of 0.039 94 nm. After being digitized, the images were stored in a computer for further image processing. To determine the structures and the netplane distances, the power spectra (square of the modulus of the Fourier transform of the images) were calculated.18
Results X-ray Diffraction. In Figure 3, the corrected XRD pattern of as synthesized nano-CdS (circles) is shown, compared with the simulated pattern of a set of (i) cubic sphalerite type clusters (dotted curve) and (ii) hexagonal wurtzite type clusters (dash-dotted line). Both simulations deviate from the experimental curve, especially in the range of the second maximum. The related R-values (R ) x(∑(∆yi)2)/∑yi) of these fits are 1.6% and 1.1%, respectively. For a combination of both cluster types the simulation is strongly improved with R ) 0.8% (solid line). The real situation is probably different from this idealized picture. A simulation with a homogeneous assembly of clusters that contain a certain amount of planar defects could further improve the agreement but has not yet been conducted. This is supported by a detailed high-resolution line profile analysis of the sintered mate(18) Urban, J.; Sack-Kongehl, H.; Weiss, K. Z. Phys. 1993, D28, 245.
Sphalerite-Wurtzite Intermediates
Figure 4. (top) Hexagonal wurtzite diffraction pattern of the sample treated at 460 °C in He. (bottom) Integral width of Bragg peaks corresponding to the top figure. No correction is made for the instrumental broadening (dbinst ∼ 0.9 × 10-3 Å-1). They are compared with calculated line broadening using the parameters given in the figure.
rial, heat treated to 460 °C in helium (Figure 4). This material, approximately 40 nm in size, is wurtzite-type. The hexagonal hkl powder profiles are systematically broadened (i) by finite average crystallite size L and (ii) by stacking faults (probability R) and twin faults (probability β) for all reflections with h - k ) 3n ( 1.19 The integral line width db in reciprocal b-units (b ) 2 sin θ/λ) is
db ) 1/L + (e/bc2)(3R + iβ) and i ) 3 for l even; i ) 1 for l odd Interestingly, the stacking fault probability is found to be R ) 1.1%, while twin fault defects can be excluded (β ) 0). The former defects can be understood as small sphalerite-type domains, consisting of three to four net planes. Figure 5 shows the DFA simulations (left) and the corresponding size distributions (right) for as synthesized CdS and after annealing in He at 225 and 315 °C, respectively. The related mass mean diameters are 2.28, 2.68, and 3.22 nm. The initial ∼50% fraction of hexagonal CdS remains nearly constant during heating but is shifted to larger cluster sizes, as is the case for the cubic particles. Despite the fact that the weights of individual clusters in the bar-graph should not be taken too seriously, the overall tendency of this early stage of growth is apparently the formation of a bimodal distribution of particle sizes. Diffraction patterns taken in situ at different temperatures are shown in Figure 6. At 235 and 345 °C the first diffuse peak narrows and shifts to smaller angles. The large shift (∆b/b ) -3.1 × 10-2 between 22 and 345 °C) cannot be explained by the thermal expansion coefficient Rlin ) 5.7 × 10-6 K-1 of CdS. In fact, the room temperature DFA simulation gives an average contraction with respect (19) Warren, B. E. X-Ray Diffraction; Addison-Wesley Publishing Co.: Reading, MA, 1969; p 303 and p 253.
Langmuir, Vol. 13, No. 4, 1997 829
to the bulk CdS of 1.4(4)% for the as synthesized CdS nanocrystals. This number is being reduced to 0.4% after annealing at 225 °C, and no measurable contraction is found after annealing at 315 °C in He. Wang and Herron have reported a contraction of ∼3% for 1.5 nm CdS nanoparticles by simply comparing the position of the first maximum of the XRD patterns.20 At 460 °C sintering is evident from the splitting into the three hexagonal wurtzite lines. Above ∼380 °C a sudden jump in the intensity, the counter being placed stationary at the angular position of the hkl ) 101 peak, is indicative for heavy sintering of the nanoparticles. This is supported by the thermogravimetric measurement shown in Figure 7. According to the DTG trace most weight losses occur between 140 and 380 °C indicating that the ligand shell is fully decomposed above this temperature. Electron Microscopy. Morphology and Size Distribution. With regard to the solubility in methanol the three samples showed different behaviors. Sample 1 could easily be dispersed which resulted in the separation of single nanoparticles, which at times, however, showed coagulation. For electron microscopy experiments the poor contrast of the CdS particles on the carbon substrate was such that only larger particles, i.e., larger than 3 nm diameter, could be measured with atomic resolution and thus characterized. Also a large amount of smaller clusters was observed, i.e., as small as 2 nm, whose exact sizes, however, could not be determined. Sample 2 showed a slightly different behavior. The solubility was less than that in sample 1, but the mean particle size was increased even though a correct characterization of the size distribution of the smaller clusters was likewise not possible due to the lack of contrast of these species. However, the increase in particle size for sample 2 is in good agreement with the X-ray results. Sample 3 could not be dispersed at all. Very large domains (larger than 30 nm) were observed which were fully crystallized showing occasionally small distortions. Structure. Sample 1. Most particles show heavy distortions which often do not allow a correct determination of their structures. Examples of the hexagonal and the cubic modification of CdS were found, but also decahedral particles were observed. An example of the decahedral structure is shown in Figure 8a together with the corresponding power spectrum. The particle has a diameter of 4.8 nm and is viewed closely along the 5-fold axis. The 5-fold axis is slightly tilted with respect to the substrate which can be deduced from the power spectrum. Also, from the power spectrum the corresponding lattice spacings of the (111) planes in the deformed tetrahedral subunits and one (220) lattice plane are obtained: d111 ) 0.338 nm, d220 ) 0.202 nm, which is to be compared with the cubic lattice spacings of d111 ) 0.336 nm and d220 ) 0.206 nm. An example for a 3.6 nm diameter face-centered cubic cuboctahedron in the [110]-orientation is shown in Figure 8b. The corresponding lattice spacings are d111 ) 0.347 nm, d-111 ) 0.341 nm, and d200 ) 0.294 nm. The angles between the identified planes are obtained as 70°, 55°, and 55°. Here also a small deviation from the bulk values for the angles and the lattice constants is observed which indicate a certain amount of distortion. A hexagonal particle of about 2.9 nm in the [001]-orientation is displayed in Figure 8c. There are also distortions present which are mirrored in the different lattice spacings, 0.353, 0.330, and 0.315 nm, which should be equal for undistorted hexagonal particles. Sample 2. After the sample was annealed at 225 °C under He, the cluster size increased as discussed above. (20) Wang, Y.; Herron, N. Phys. Rev. 1990, B42, 7253.
830
Langmuir, Vol. 13, No. 4, 1997
Vogel et al.
Figure 5. (a) DFA simulations for as received nano-CdS and for samples treated at 225 and 315 °C in He, respectively. (b) The corresponding size distributions of both hexagonal and cubic model clusters. Numbers in the top figure are the number of CdS molecules per model cluster.
Figure 7. Thermogravimetry in Ar, 5 K/min, of nanocrystalline CdS as prepared.
Figure 6. Diffraction patterns of nano-CdS, measured in situ in an He atmosphere at four different temperatures.
However, most analyzed particles showed hexagonal symmetry in different orientations. About 30% of the evaluated nanoparticles showed cubic structure. Two examples for decahedral structures close to the 5-fold orientation were observed. One of them is displayed in
Figure 9a. The particle is also slightly deformed, which can be seen from the different 111 reflections of the corresponding subunits (0.320-0.351 nm). Figure 9b shows a particle of cubic symmetry. An example for a hexagonal particle is shown in Figure 9c. Sample 3. After being annealed at 600 °C in He, the sample is fully recrystallized and shows mainly the hexagonal symmetry. An example thereof is shown in Figure 10. Discussion The observed optical band gap of 3.1 eV and the mass mean diameter of 2.3 nm are in excellent agreement with
Sphalerite-Wurtzite Intermediates
Langmuir, Vol. 13, No. 4, 1997 831
Figure 8. Electron micrographs together with their power spectra below: (a) decahedron of CdS as prepared, close to the 5-fold orientation, particle diameter 2R ) 4.8 nm, d111 ) 0.338 nm, d220 ) 0.202 nm; (b) CdS as prepared, cubeoctahedron in the [110]orientation, particle diameter 2R ) 3.6 nm, d111 ) 0.347 nm, d-111 ) 0.341 nm, d200 ) 0.294 nm; (c) CdS as prepared, hexagonal particle in the [001]-orientation, particle diameter 2R ) 3.6 nm, d010 ) 0.353 nm, d100 ) 0.330 nm, d1-10 ) 0.315 nm, distorted.
Figure 9. Electron micrographs together with their power spectra below: (a) CdS annealed at 225 °C under He, decahedron with orientation close to the 5-fold axis, 2R ) 4.8 nm, d111 ) 0.320-0.351, distorted; (b) CdS annealed at 225 °C under He, cubic structure close to the [110]-orientation with distortions, 2R ) 3.6 nm, d111 ) 0.335 nm, d200 ) 0.274 nm; (c) CdS annealed at 225 °C under He in the [001]-orientation, hexagonal structure, 2R ) 9.5 nm, d100 ) d010 ) d1-10 ) 0.356 nm, d110 ) 0.206 nm, d200 ) 0.179 nm.
the tight binding band gap calculations of Lippens and Lannoo.3 For their calculations the authors assumed the
CdS clusters to take the cubic Td space group. Stampfl et al. have recently shown by angle-resolved photoelectron
832
Langmuir, Vol. 13, No. 4, 1997
Figure 10. Electron micrograph together with its power spectrum below: CdS annealed at 600 °C under He; hexagonal structure; orientation [001]; d100, d010, d1-10 ) 0.353, 0.344, 0.330 nm, distorted; large domain.
spectroscopy that the valence band structure of cubic CdS agrees with previous results on the wurtzite structure at equivalent symmetry points.21 It is known that CdS cluster molecules forming single crystals take the structure of super-tetrahedral fragments of the cubic (sphalerite) lattice.22-24 However, the polymorphism tendency of CdS is well-known. Few examples of hexagonal (wurtzite) structure of CdS nanocrystals are reported in the literature. Vossmeyer et al.25 have shown by TEM and XRD that the crystal structure depends sensitively on the chemical root leading to both cubic and hexagonal nanocrystals. Recently, Zelaya-Angel et al. observed a transition from cubic to hexagonal structure in thin CdS films on heating in Ar.26,27 Herron et al. have prepared thiophenolate-capped CdS particles and characterized them among other techniques
Vogel et al.
by XRD.28,29 They assigned the observed peaks to cubic sphalerite and applied the Scherrer formula for the size determination. However, this assignment cannot be made unequivocally as shown in Figure 2. Application of the Scherrer equation db1/2 ) 1.88/L (db1/2 ) full width half maximum in units of b, L ) edge length of a cube) demands an isolated Bragg peak and a well-defined base line.19 The first maximum in Figure 2 is, however, composed of two (sphalerite) or three (wurtzite) Bragg peaks, respectively, and the baseline is ill-defined. Application of the Scherrer formula gives L ) 3.2 nm (cube edge length) and D ) 3.9 nm for the diameter of a sphere of equal volume, respectively. On the other hand, by the DFA simulations of Figure 2 sphere-equivalent diameters of 1.92 nm (sphalerite), 2.15 nm (wurtzite), and 2.28 nm (sphalerite + wurtzite) are obtained. Similarily, overestimated lattice contractions may be obtained if one simply relates them to the peak shifts of the XRD pattern. Murray et al.9 observed no deviation from the bulk bond length in monodisperse CdSe nanocrystals using computer simulation of the X-ray powder diffraction. Goldstein et al.30 have studied the melting of CdS nanocrystals by TEM and selected area electron diffraction (SAD). Melting was reported to occur at ∼300 °C for the smallest particles of 2.4 nm diameter. The organic capping agents desorb at 250 and 190 °C for thiophenol and mercaptoacetic acid under high vacuum conditions, respectively. In our study, performed in an helium ambient, no signs of melting could be observed and residues of the mercaptoethanol ligands are still present at 345 °C, preventing the nanocrystals from heavy sintering. However, our data are obtained from an assembly of particles showing a relatively broad size distribution which restricts physical conclusions related to the particle size. Froment and Lincot studied the phase formation process of chemical bath deposited chalcogenides by HREM.31 In this study stacking faults in the hexagonal CdS films are frequently observed, the density being d ) 1011 to 1012 cm-3. We compared this number to our sintered CdS nanocrystals. The volume density is related to stacking fault probability R by d ) R(Lc/d002)/(LaLc2), where La and Lc are the lateral and normal crystallite dimensions, respectively. With La ≈ Lc ) 41 nm, R ) 0.011 we obtain d ) 1016 cm-3. The higher value can be related to quite different growth mechanisms, i.e., not atom by atom, but coalescence of individual nanoparticles. In conclusion, we have shown that a refined analysis of the powder diffraction can give a detailed insight of the structure as well as of the size distribution of nanocrystalline semiconductors; the latter are required as a reliable basis to relate the optical properties to theoretical band gap calculations. Acknowledgment. We thank T. Ru¨hle for performing the DTA/TGA measurements. K. Weiss and H. SackKongehl took part in evaluating the TEM images and the power spectra. M.K. thanks the Council of Scientific and Industrial Research, India, and S.K.K. thanks the University Grants Commission, India, for financial support. LA960426K
(21) Stampfl, A. P. J.; Hofmann, Ph.; Schaff, O.; Bradshaw, A. M. Submitted to Phys. Rev. B. (22) Dance, I. G.; Choy, A.; Scudder, M. L. J. Am. Chem. Soc. 1984, 106, 6285. (23) Herron, N.; Calabrese, J. C.; Farneth, W. E.; Wang, Y. Science 1993, 259, 1426. (24) Vossmeyer, T.; Reck, G.; Katsikas, L.; Haupt, E. T. H.; Schulz, B.; Weller, H. Science 1995, 267, 1476. (25) Vossmeyer, T.; Kastsikas, L.; Giersig, M.; Popovic, I. G.; Diesner, K.; Chemseddine, A.; Eychmu¨ller, A.; Weller, H. J. Phys. Chem. 1994, 98, 7665.
(26) Zelaya-Angel, O.; Esparza-Garcia, A. E.; Falconi, C.; LozadaMorales, R. Solid State Commun. 1995, 94, 81. (27) Zelaya-Angel; Hernandez, L.; de Melo, O.; Alvarado-Gil, J. J.; Lozada-Morales, R.; Falconi, C.; Vargas, H. Vaccum 1995, 46, 1083. (28) Herron, N.; Wang, Y.; Eckert, H. J. Am. Chem. Soc. 1990, 112, 1322. (29) Wang, Y.; Herron, N. Phys. Rev. 1990, B42, 7253. (30) Goldstein, A. N.; Echer, C. M.; Alivisatos, A. P. Science 1992, 256, 1425. (31) Froment, M.; Lincot, D. Electrochim. Acta 1995, 40, 1293.