Colloidal PbTe–Au nanocrystal heterostructures

PAPER

www.rsc.org/materials | Journal of Materials Chemistry

Colloidal PbTe–Au nanocrystal heterostructures†‡ Isabella R. Franchini,ab Giovanni Bertoni,a Andrea Falqui,a Cinzia Giannini,c Lin Wang Wangd and Liberato Manna*a Received 31st July 2009, Accepted 21st September 2009 First published as an Advance Article on the web 20th October 2009 DOI: 10.1039/b915687a Colloidal PbTe nanocrystals are reacted with AuCl3 in the presence of dodecylamine and tetraoctylammonium bromide in a toluene solution. At room temperature, only homogenous nucleation of isolated Au nanocrystals in solution is observed. At higher temperatures (i.e. 60  C or higher) the gold ions/atoms are able to diffuse through the PbTe nanocrystals and to form one or more metallic gold regions inside them, while most of the remaining volume of each nanocrystal becomes amorphous. Longer reaction times lead to the growth of a single balloon-shaped Au domain attached via its apex to the surface of each nanocrystal. The structural and compositional quantification of the starting PbTe nanocrystals and of the various reaction products is often complicated by several reactive processes occurring during in situ analysis by electron microscopy. Evidences of the formation of a metastable Au3Te compound are presented.

1. Introduction Colloidal inorganic nanocrystals represent suitable samples with which to investigate chemical-physical phenomena occurring in nanoscale solids or at their surface, like for instance ion exchange, impurity diffusion, doping, galvanic replacement reactions, heterogeneous nucleation processes, surface chemisorption of organic species, and twinning.1–15 Indeed the synthesis of colloidal nanocrystals is now well advanced, and additionally by being free-standing surfactant-stabilized nanoparticles in solution they are readily accessible to reactive species. The ability to control the chemi-sorption of organic stabilizers on the surface of growing nanocrystals,8,11 the formation of twinning defects,9,11 and the oriented attachment of building units16 has allowed for instance the fabrication of colloidal nanocrystals with tailored sizes and shapes. Additionally, the possibility to epitaxially grow a second material, or more generally to control the heterogeneous nucleation of a second material on top of a preformed nanocrystal has led to the fabrication of a wide variety of nanocrystal architectures.9,10,17 Examples are nanoparticles in which two or more materials are either grown uniformly on the top of each other in onion-like structures (i.e. core–shell)8,18,19 or are organized in separate

a Fondazione Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genova, Italy. E-mail: [email protected]; Tel: +39 010 71781502 b National Nanotechnology Laboratory of CNR-INFM, Via per Arnesano km 5, I-73100 Lecce, Italy c CNR-Istituto di Cristallografia (IC), Via Amendola 122/O, I-70126 Bari, Italy d Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, California, USA 94720 † This paper is part of a Journal of Materials Chemistry issue highlighting the work of emerging investigators in materials chemistry. ‡ Electronic supplementary information (ESI) available: TEM images of dimeric structures occasionally found in the ‘‘close to cubic-shaped’’ PbTe nanocrystals that were formed when pumping was applied, PbTe nanocrystals reacted with the AuCl3 solution at room temperature, and fits to the XRD spectra shown in the work. See DOI: 10.1039/b915687a

This journal is ª The Royal Society of Chemistry 2010

domains that share only a small portion of their surfaces, like for example colloidal heterodimers made of two joined spherical nanocrystal domains,14 or nanorods decorated at one or both tips with a domain of another material.14,17 Partial or complete anion/ cation exchange or even galvanic replacement reactions on colloidal nanocrystals have been exploited to create various types of nanocrystals, including core–shells,2,20 to prepare nanocrystals of given materials (i.e. Ag2Se) in shapes that were not accessible by direct chemical synthesis for such materials (i.e. nanorods, tetrapods),21 or to create unique linear nanocrystal superlattices (i.e. striped CdS–Ag2S nanorods).22 Also, differences in the diffusion rate of two species across the interface between two materials, which can lead to the formation of voids (the so-called Kirkendall effect), were found to be responsible for the formation of certain types of hollow nanocrystals.23–25 In this contribution, we would like to highlight the formation of various types of nanostructures arising from the reaction of colloidal PbTe nanocrystals with AuCl3 in the presence of dodecylamine and tetraoctylammonium bromide. PbTe is a low band gap semiconductor with a remarkably large Bohr exciton radius and high thermoelectric figure of merit.26,27 The latter is generally higher in the corresponding low-dimensional nanostructures (especially 1-dimensional nanostructures), due to increased density of states near the Fermi level.26–29 The possibility of selectively nucleating Au domains on specific regions on the surface of PbTe nanocrystals could be exploited then as a means for contacting them and/or easily extracting carriers out of them, with potential applications in detectors and thermoelectric devices. So far, only a few studies have been reported on Au–PbTe interfaces, mainly prepared by electrodeposition of lead telluride films/large crystals on gold substrates, and aiming at assessing the epitaxial relationships between the two materials and their optical properties.30,31 Our initial drive was therefore to explore the possibility to grow metallic gold domains on top of the PbTe nanocrystals, in analogy with the several reported cases of noble metal nucleation on selected regions of cadmium chalcogenides nanocrystals32–37 J. Mater. Chem., 2010, 20, 1357–1366 | 1357

and also on PbS38,39 (it is worth to point out that Au–PbS and Au–PbSe nanocrystal heterostructures have been prepared also starting from Au nanoparticles as catalysts40). However, a recent work on the reaction of solutions containing Au(III) species with PbS nanocrystals has highlighted that lead chalcogenides might often react differently towards Au(III) than the corresponding cadmium chalcogenides.41 In that work the formation of goldhollow PbSx hybrid nanoparticles was reported, which has not been observed so far in the corresponding CdS nanocrystals.41 In the case of PbTe nanocrystals, the reactivity towards Au(III) or in general towards oxidation reactions should be even more remarkable, mainly for the stronger tendency of Te2 atoms to be oxidized with respect to both S2 and Se2 . We found that when a toluene solution of PbTe nanocrystals was reacted with a toluene solution containing AuCl3, dodecylamine and an alkylammonium salt as stabilizer, various reaction products could be recovered, depending on the reaction conditions and on the way the starting PbTe nanocrystals had been synthesized. At room temperature, mainly small isolated gold nanoparticles were formed, with practically no trace of gold domains nucleated heterogeneously on the surface of PbTe nanocrystals. At higher temperatures, on the other hand (60  C or higher), gold atoms/ions could easily diffuse inside the PbTe nanocrystal lattice and form one or more crystalline Au domains mainly inside the nanoparticles, leading to a crystalline-Au core– amorphous PbxTeyAuz shell structure. Further reaction of the PbTe nanocrystals with the AuCl3 solution led to the growth of a single, large balloon-shaped (or mushroom-shaped) Au domain attached via its apex to the surface of each nanocrystal. This peculiar growth of the gold domain enables it to share only a small interfacial area with the nanocrystal surface. The present system represents an example of the complex reactive behavior that inorganic nanocrystals exhibit often, and which should be taken into account when designing and characterizing elaborate nanostructures.

2. Experimental All syntheses were carried out in a standard Schlenk line set-up under nitrogen flow, and all following reactions and manipulations were carried out under inert nitrogen in a glove box. Also, all samples were stored in the glove box. 2.1.

Synthesis of PbTe nanocrystals

The synthesis of PbTe nanocrystals was procedures described in the literature42,43 tions. Briefly, a solution of precursors following way: in a three-neck flask, 1.9 1358 | J. Mater. Chem., 2010, 20, 1357–1366

2.3.

carried out following with minor modificawas prepared in the mmol of lead acetate

Reaction of PbTe nanocrystals with the AuCl3 solution

The ‘‘AuCl3 solution’’ was prepared by mixing two separate solutions. For solution A, 24 mg of AuCl3 (0.08 mmol) and 44 mg of TOAB (0.08 mmol) were dissolved in 5 mL of toluene. For solution B, 140 mg (0.75 mmol) of DDA were dissolved in 5 mL of toluene. Small fractions of solutions A and B (750 mL each) were mixed prior to their use. The resulting yellowish ‘‘gold precursor solution’’ was quickly injected into 1.5 ml of the solution containing PbTe nanocrystals dissolved in toluene (the concentration of PbTe nanocrystals is this solution was 10 8 M). For each sample of PbTe nanocrystals, we carried three series of experiments, the first one in which the reaction temperature was 25  C, the second one in which it was 60  C, and the third one in which it was 90  C. In each series of experiments the reaction was allowed to run for various time lapses. In order to remove the excess of unreacted gold species and the small Au nanoparticles that had nucleated separately in solution, the nanocrystals were precipitated by addition of methanol and were then redissolved in 1 ml of toluene. 2.4.

Chemicals

Lead acetate trihydrate (Pb(C2H3O2)2$3H2O, 99.99%), gold trichloride (AuCl3, 98%), tetraoctylammonium bromide (TOAB or C32H68BrN, 98%), dodecylamine (DDA or C12H27N, 98%), oleic acid (OA or C18H34O2, 90%), tellurium powder (99.999%), trioctylphosphine (TOP or C24H51OP, 90%), and phenyl ether (Ph2O or C12H10O, 99%) were purchased from Sigma-Aldrich. All solvents used were anhydrous and were purchased from Sigma-Aldrich. 2.2.

trihydrate was dissolved in 2 ml of oleic acid (OA) in the presence of 10 ml of phenyl ether (Ph2O). Two approaches were then followed. In one case the resulting solution was pumped to vacuum at 70  C for 2 h. By this procedure we could eliminate all the acetic acid and water from the solution, as well as other low molecular weight impurities. After pumping to vacuum, the solution was stirred for 20 min at 150  C, and then it was cooled at room temperature and 3.8 ml of a Te–TOP solution (0.5 M in Te, prepared in advance) was rapidly injected into it. The second approach differed from the one just described in that no pumping step was applied to the solution. In both approaches then the freshly prepared solution of precursors was injected into 10 ml of phenyl ether solution heated at 200  C, and the reaction was allowed to proceed for 15 min at 200  C. After the synthesis, the mixture was cooled to room temperature. A size-selective precipitation was carried out in order to improve the size distribution of the particles. Hexane was added to the crude solution (1 : 1 v/v) and the nanocrystals were precipitated by adding ethanol. They were then re-dissolved in 10 ml of toluene and stored under inert atmosphere.

Characterization and modeling

2.4.1. Transmission electron microscopy (TEM). Samples for TEM were prepared by dropping dilute solutions of nanocrystals onto carbon coated copper grids and letting the solvent evaporate. Low magnification TEM images were recorded on a JEOL JEM-1011 microscope operating at 100 kV. Cs-corrected High Resolution and Scanning Transmission (STEM) images were acquired on a JEOL JEM-2200FS, the latter using a high angle annular detector (HAADF) and a small probe size (0.2 nm). Spherical Aberration (Cs) correction greatly improved point resolution in TEM, the latter being 0.9 A (in our instrument) versus the typical 1.9 A for a not Cs-corrected microscope with same acceleration voltage and objective lens type. The chemical composition of the nanostructures was investigated on the JEOL JEM-2200FS using a JED-2300 Energy Dispersive X-ray Analyzer. The beam (1 nm probe size) was scanned on the sample This journal is ª The Royal Society of Chemistry 2010

(STEM) and the X-ray spectrum (EDX) was acquired at each point of the scan. The chemical quantification was performed using the Cliff-Lorimer (or Ratio) method, which is considered as a good approximation for thin films.44 The Te, Au, and Pb La peaks where fitted using the EELSMODEL program,45 using the same approach that gave optimal results in the quantification of the EELS spectra,46 and extended here to EDX, by using a model consisting of Gaussian peaks on an exponential background. 2.4.2. Powder X-ray diffraction (XRD). The powder XRD measurements were performed with a RIGAKU-INEL powder diffractometer with 12 KW rotating anode (RINT200 series) and copper anode, a (111) germanium monochromator and a curved position sensitive detector CPS-120 INEL. Concentrated solutions of the purified nanocrystals were spread on top of a silicon miscut substrate. The sample was allowed to dry and was then measured in reflection geometry. The spectra were fitted with the program FULLPROF,47 which combines the Rietveld method with an anisotropic peak shape description (with spherical harmonics) of the nanocrystal shape anisotropy. 2.4.3. Elemental analysis and determination of the nanocrystal concentration. Elemental analysis to determine the total concentration of Pb in the solution of nanocrystals was performed via inductively coupled plasma atomic emission spectroscopy (ICP-AES) with a Varian Vista AX spectrometer. The samples were digested in HCl–HNO3 3 : 1 (v/v). The nanocrystal concentration was carried out in the following way: first the average nanocrystal sizes were assessed via statistical analysis on TEM images of several hundreds of nanocrystals. Then the average number of Pb atoms per nanocrystal was determined by building a structural model of the nanocrystal. Knowledge of the average number of Pb atoms per nanocrystal and of its total concentration in solution allows for the determination of the concentration of nanocrystals. This estimate was clearly affected by significant errors, especially for the nanocrystals coated with a thick amorphous shell (see below). 2.4.4. Density functional calculations. These were carried out under local density approximations with the PEtot code.48 Plane wave norm-conserving pseudopotential methods were used, with 65 Ryd energy cut off for the plane wave basis. The lattice constants for face centered cubic (fcc) Au and for fcc Au3Te (which is a metastable alloy that was formed during in situ TEM analysis, see below) were estimated from the minimum of the energy as a function of the lattice constant. The energies of fcc, hcp and trigonal Te were calculated. In each of the calculations the lattice parameters were varied until the minimum energy was found. For the trigonal crystal structure, both the c axis and the hexagonal lattice constants were varied independently in order to find the minimum energy. The formation energy per atom for fcc Au3Te was calculated as: (E(Te) + 3E(Au)-E(Au3Te))/4. To test the structural stability of fcc Au3Te, a large supercell was used, which contained several units of Au3Te, and small random displacements from the ideal positions were introduced in the initial atomic positions. The system was then relaxed according to ab initio atomic forces. In this approach, if the system goes back to the ideal positions of Au3Te, then the system is stable (a local minimum), otherwise it is unstable. This journal is ª The Royal Society of Chemistry 2010

3. Results 3.1.

Synthesis and characterization of PbTe nanocrystals

Different nanocrystal morphologies were formed depending on whether or not a vacuum pumping step was applied to the solution of lead acetate trihydrate in oleic acid and phenyl ether. If no pumping was applied, nearly spherical shaped particles were formed (Fig. 1a). They often showed a pseudo-cubic habit, that can be seen as the morphological coexistence of two different crystalline forms, a {100} cube together with a {110} rhombic dodecahedron, very likely due to the fact that the growth along the [220] directions was slower than that along the [200] ones. On the contrary, when pumping was applied the shapes of particles were closer to cubes (Fig. 2a and b), with developed 100 facets. A representative low magnification TEM image of sizeselected pseudo-cubic PbTe nanocrystals is shown in Fig. 1a. The sample was rather monodisperse in size (the average size was equal to 16 nm), except for the presence of a few smaller nanocrystals. At high resolution (Fig. 1b and 1c) each individual nanocrystal appeared as formed by a crystalline core (with lattice spacings characteristic of rock-salt PbTe) and by an amorphous shell. The thickness of this shell could be up to 4 nm for the present sample. Under the convergent beam of the STEM FEG the amorphous shell (Fig. 1d), tended to disappear in a few seconds (Fig. 1e), and the process was accompanied by an increase in density of about 30% in the area in which the shell was located. Such change could be correlated with an in situ crystallization of the shell. PbTe nanocrystals are prone to rapid surface oxidation49–52 and the presence of a layer of amorphous oxide of Pb and Te was reported already on the surface of PbTe nanocrystals.53 It therefore seems plausible that these PbTe nanocrystals where coated with an oxide shell. By spatially resolved EDX analysis (Fig. 1c) we found in the shell region a Pb : Te ratio equal to 1.01  0.05 (almost identical to the crystalline core), while the oxygen stoichiometry was not far from 3 (3.3), although the quantification of this element is affected by the contribution of oxygen from other sources (hydrocarbons on the TEM grid, organic solvents from sample preparation, copper oxide from the holder, and so on). Among the possible known oxides of Pb and Te,49 PbTeO3 is likely therefore to be the major chemical species present on the surface of the nanocrystals studied here. These data should, however, be taken with caution, since the nanoparticles were also modified significantly during the EDX scan and the shell region eventually contracted and crystallized. It is also known from literature that oxides like PbTeO3 can be converted to PbTe upon heating under a reducing atmosphere.54 Hence the disappearance of the shell under the intense electron beam in the TEM could be possibly correlated to a local crystallization of amorphous PbTeO3 and probably also to a partial reductive annealing of PbTeO3 to PbTe caused by the beam. Similar processes (local reduction and crystallization) induced by electron beam irradiation are well documented.55–58 The main X-ray powder diffraction peaks from a dried sample of these nanocrystals (Fig. 1g) were easily attributable to crystalline PbTe, and their widths were compatible with an average size of the crystalline PbTe domain equal to 8.6 nm (see Fig. S3 of J. Mater. Chem., 2010, 20, 1357–1366 | 1359

Fig. 2 (a) Wide-field low magnification image of ‘‘close to cubic’’ PbTe nanocrystals prepared after pumping to vacuum the solution of lead acetate trihydrate in oleic acid and phenyl ether. (b) A HRTEM image of a representative area in the same grid, indicating the absence of thick amorphous shells around these particles.

Fig. 1 (a) Wide-field low magnification image of pseudo-cubic PbTe nanocrystals prepared without pumping to vacuum the solution of lead acetate trihydrate in oleic acid and phenyl ether. (b) and (c) HRTEM images of individual PbTe nanocrystals, revealing a crystalline core and often an amorphous shell. (d) A PbTe nanocrystal coated with an amorphous shell is imaged under the convergent beam of the STEM FEG. (e) In the same nanocrystal a few seconds after the amorphous shell had disappeared. (f) A spatially resolved linear EDX scan along the diameter of a single nanocrystal (i.e. along the red line) shows that the Pb : Te ratio is close to 1 : 1 both in the core and in the shell regions. The image in the background is a HAADF image of the PbTe nanocrystal, revealing a noticeable difference in scattering power between the core and the shell regions. (g) Wide angle X-ray powder diffraction (XRD) spectrum from a dried sample of these PbTe nanocrystals (the peak positions of bulk rock salt PbTe are also shown).

the ESI‡), hence much smaller than the average nanocrystal size as estimated by low resolution TEM analysis. The XRD data therefore confirmed the presence of a thick amorphous shell. Spurious peaks present in the spectrum could be correlated to 1360 | J. Mater. Chem., 2010, 20, 1357–1366

organic contaminants, which apparently we were not able to eliminate completely by repeated washing and re-dissolution of the sample. The ‘‘close to cubic-shaped’’ PbTe nanocrystals that were formed when pumping was applied (Fig. 2a, b) where rarely characterized by a thick amorphous shell. Most HRTEM images of nanocrystals of this sample (see for instance Fig. 2b) exhibited lattice fringes throughout the nanoparticles. Most of the particles were oriented in [001] zone axis and for a chosen particle lattice sets, facets directions and zone axis are clearly reported in the figure. Occasionally, a few particles exhibited two types of dimeric structures, as reported in the supporting information (Fig. S1‡). In these ‘‘close-to-cubic-shaped’’ nanocrystals the oxide shell, if present, was remarkably thin as it could not be detected by TEM. The comparison between the two PbTe nanocrystal samples indicates therefore that the impurities present in the solution of lead acetate trihydrate in oleic acid and phenyl ether (among them residual acetic acid and water), when they are not pumped away, are both able to influence the morphology of growth of the nanocrystals and to cause substantial oxidation of the outermost PbTe layers. Houtepen et al.59 found that residual acetic acid and water molecules could alter significantly the growth of PbSe nanocrystals, whose morphology was star-shaped in the presence of trace amounts of these contaminants (and explained as arising from oriented attachment), while it was pseudo-cubic when the contaminants had been removed by prolonged pumping to vacuum. In our case the morphologies found for PbTe, and the additional presence of the amorphous shell, differ considerably from the PbSe case studied by Houtepen et al.,59 but nevertheless they highlight once again the important role of impurities in the synthesis of colloidal nanocrystals. 3.2. Reaction of PbTe nanocrystals with the AuCl3 solution at room temperature When both samples of PbTe nanocrystals were reacted with the gold solution at room temperature, only the nucleation of small isolated gold nanoparticles in solution was observed, and no substantial variation in the size/shape of nanocrystals was detected, even if the reaction was allowed to run for several hours (see Fig. S2 of the ESI‡). Heterogeneous nucleation of Au(0) This journal is ª The Royal Society of Chemistry 2010

‘‘patches’’ on the nanocrystals, with the dodecylamine being the reducing agent for Au(III), has been reported already for various types of colloidal nanocrystals acting as substrates/catalysts for this type of redox reaction.32,34,41,60 The difficulty with which Au(0) patches grow on these PbTe nanocrystals in the present case could be related to the presence of a surface oxide layer (and a thin layer is probably present even on the cube-shaped nanoparticles), which does not seem to be a good catalyst for the reduction of the gold species present in solution to Au(0) at the expenses of the dodecylamine. Alternatively it could be due to some other sort of activation barrier to be overcome in these nanocrystal samples (i.e. the removal of surfactant molecules adsorbed to their surface). It is also known that for the many nanocrystalline materials on which growth of Au (0) is documented, those regions on the nanocrystal surface (especially corners/tips) where Au patches had actually nucleated were in part acting as sacrificial material, in other words as co-reducing agents together with the amine,32,34,41 This could be deduced by the shrinkage of these regions upon Au(0) growth on them. In the present case however, the likely presence of the surface oxide layer (no matter whether this was PbTeO3 or a different oxide, like for instance Pb2TeO4 or many other possible Pb and Te containing oxides) would not allow galvanic replacement reactions. All these oxides are lead salts of oxo-acids of tellurium, in which the oxidation state of Te is either IV, VI, or both.61 Therefore, galvanic replacement reactions should not actually take place in this layer as Te is already oxidized. 3.3. Reaction of PbTe nanocrystals with the AuCl3 solution at 60–90  C: core–shell nanostructures 

At higher temperatures (i.e. 60 C or higher), in addition to the nucleation of small (1–2 nm) isolated Au nanoparticles in solution, the PbTe nanocrystals started to exhibit different reactivities towards the AuCl3 solution, depending on whether they were coated or not by a thick amorphous shell. The results discussed here are refer to the experiments carried out at 90  C for 5 min, but comparable results were found for reactions carried out at 60  C for 30 min. In Fig. 3a and b are reported low magnifications TEM images of a typical sample recovered after the pseudo-cubic PbTe nanocrystals coated with a thick amorphous shell were reacted with the gold precursor solution for 5 min, and after the excess unreacted precursors and most of the small Au nanoparticles present in solution had been removed via size-selective precipitation. In this sample the two peculiar features were that the average nanoparticles diameters had contracted by about 30%, which could be due to etching of the thick amorphous shell present on the starting nanocrystals, and higher contrast domains were found in the core region of many nanocrystals. The latter was also confirmed by STEM HAADF analysis (Fig. 3c), according to which one or more domains of higher atomic number were always located in the inner regions of the nanocrystals, indicating the presence of mainly core-shell structures. Au-related regions were also visible on regions further from the nanocrystal centre, which could point to the attainment of heterodimer-like AuPbTe heterostructures. However, if Au domains were present on the surface of the PbTe nanocrystals in a significant amount of This journal is ª The Royal Society of Chemistry 2010

Fig. 3 (a) and (b) Wide-field low magnification images of the PbTe nanocrystals, initially coated with a thick amorphous shell, after they were reacted with the AuCl3 solution for 5 min at 90  C. High contrast domains are located inside the nanocrystals. (c) A HAADF image of the same sample, showing strongly scattering domains located in the core of the nanocrystals. (d) A HRTEM image of representative Au–PbTe core– shell nanocrystals. (e) Wide angle XRD spectrum of a dried sample of these nanocrystals. The spectrum is dominated by diffraction from the inner Au domains (the peak positions of bulk fcc Au are also shown).

particles, then in a non-negligible subset of particles a heterodimer-like structure should have been observed under TEM, given the many possible orientations of the various particles deposited on the supporting carbon field and imaged by TEM. Instead, we could only find particles in which the higher contrast Au domains were either located at their center or away from the center, however always in within the lower contrast region of the overall particle (see for instance the HAADF image of Fig. 3c). While this does not represent a complete proof of solely core– shell formation, we can safely conclude that the presence of Au domains on the surface of nanocrystals, hence the formation of heterodimer-like Au–PbTe heterostructures, was a rare event. Based on HRTEM analysis of this sample (Fig. 3d) the higher contrast regions were made of metallic fcc Au, although the lattice spacings were often larger than those of bulk Au by up to 4–5%. Additionally, there was almost no trace of other crystalline domains in these core–shell nanostructures, i.e. most of the other regions of the starting nanocrystals had become amorphous. The amorphization was likely a side effect of the diffusion J. Mater. Chem., 2010, 20, 1357–1366 | 1361

of Au species through the lattice, as found by Mokari et al. for InAs nanocrystals.62 The whole volume of the starting nanocrystals couldn’t be converted to or replaced by metallic Au, even if the reaction was allowed to run for a longer time. This case will be discussed in detail in section 3.4. The powder X-ray diffraction spectrum of a dried sample of these nanocrystals (Fig. 3e) contained only the peaks of fcc gold, and it could be fitted by considering as diffracting material a spherical Au nanocrystal with average size of 7.1 nm (see Fig. S4 of the ESI‡), which was consistent with the average size of the Au domains located inside the particles as estimated from HRTEM. These diffraction peaks were too narrow to arise from the residual isolated Au nanoparticles of 1–3 nm diameters that still contaminated the sample, which would yield much broader peaks. The inter-planar spacings found by XRD (Fig. 3e) were fully consistent with those of bulk Au, as opposed to the slightly lager values of inter-planar spacings found by TEM. In section 4.2 we will give a plausible explanation of this discrepancy. A markedly different behavior was observed when the ‘‘closeto-cube-shaped’’ PbTe nanocrystals (which lacked a thick amorphous shell) were reacted with the AuCl3 solution at 90  C. Here, after the first 5 min of reaction, Au patches had nucleated on the surface of the nanocrystals, mainly at the corners (i.e. the 111 facets), which are indeed the most reactive sites (see Fig. 4a). Similar results were found for reactions carried out at 60  C for 30 min. At later reaction times (10 min at 90  C), or if the PbTe– Au nanocrystals were cleaned and left in toluene at room temperature for a few days, these Au patched eventually diffused inside the nanocrystals (Fig. 4b–d), yielding again core–shell

nanostructures in which the PbTe regions had been almost completely amorphized. In this case therefore our findings were in agreement with those of Mokari et al. on the two-step reactivity of InAs nanocrystals towards the same type of gold precursor.62 Additionally, no appreciable shrinkage in size nor change in shape was detected on these PbTe nanocrystals, which strengthened our previous hypothesis that in the pseudo-cubic PbTe nanocrystals coated with a thick amorphous shell the considerable decrease in particle size was due to etching of this initially present shell. 3.4 Further reaction of PbTe nanocrystals with the AuCl3 solution The products resulting from the reaction of both PbTe nanocrystal samples with the AuCl3 solution at 90  C for 15 min (or at 60  C for 1 h) were core–shell Au–PbTe nanocrystals which additionally carried a large Au particle attached to them (see Fig. 5a and e). Each of these large Au particles had a shape that was reminiscent of a ‘‘balloon’’ or of a ‘‘mushroom’’, and which probably allowed them to minimize both their interfacial area shared with the core–shell Au–PbTe nanocrystals and their own surface to volume ratio. These balloons appeared to grow only on top of nanocrystals that had already developed a core–shell structure, i.e. apparently only after the PbTe nanocrystals had become somehow saturated by Au inside. The inner structure of these core–shell nanocrystals was also peculiar. Instead of carrying a single large Au domain inside and which should be expected given the large cohesive energy of gold,63 each nanocrystal contained many small (1–4 nm) high contrast domains, and in some cases even 10 of them or more (see for instance the inset of Fig. 5a). Again these inner domains, as revealed by HRTEM, had lattice parameters that could be up to 4–5% larger than that of fcc gold. The lattice parameters of the large balloons on the other hand were fully consistent with those of fcc gold. The diffraction spectra from these nanocrystal samples were overwhelmingly dominated by the contribution from the large Au balloons, and no trace of other nanocrystalline material was detectable.

4. Discussion 4.1.

Fig. 4 (a) Wide-field low magnification image of the ‘‘close-to-cubic’’ shaped PbTe nanocrystals, after they were reacted with the AuCl3 solution for 5 min at 90  C. High contrast domains are located at the edge and at the corners of the nanocrystals. (b) Low magnification and (c) and (d) HRTEM images of representative Au–PbTe core–shell nanocrystals after 10 min reaction at 90  C. This time the Au domains were mostly located inside the nanocrystals.

1362 | J. Mater. Chem., 2010, 20, 1357–1366

Mechanism of formation of Au–PbTe core–shells

The formation of Au–PbTe core-shell nanostructures can be explained by a fast diffusion of gold species inside the PbTe lattice, similarly to what observed by Mokari et al. on InAs nanocrystals.62 PbTe has a small band gap and a considerably large dielectric constant,43,64 and based on previous empirical literature data indicating that diffusion of Au is faster in semiconductor materials with both small band gap and large dielectric constant,65 it seems plausible that InAs and PbTe should equally promote fast Au diffusion. This proceeds very likely via a combination of substitutional and interstitial diffusion mechanisms.66,67 Also in the present case it is likely that Au(0) species form at the surface of nanocrystals upon reduction of Au(III) with dodecylamine and then they quickly diffuse from there to the inner regions of the nanocrystals. It is unclear whether also Au+ and Au3+ ions are able to diffuse through the lattice without reacting with the lattice itself, as both This journal is ª The Royal Society of Chemistry 2010

species can easily oxidize Te2 ions. This type of diffusion could be expected to be more significant or even predominant through the thick oxide shell of the pseudo-cubic PbTe nanocrystals (as discussed earlier), and which could possibly account for our failure to observe the intermediate nucleation of Au(0) patches at the surface of these nanocrystals, at least for most nanocrystals. Unfortunately the oxide shell itself appeared to be etched away during the reaction with the AuCl3 solution, so that no sound conclusions can be drawn. We can also safely exclude the predominance of Au3+ species in this diffusion process, since test reactions of PbTe nanocrystals with AuCl3 dissolved in toluene in the absence of dodecylamine as reducing agent (therefore only using tetraoctyl ammonium bromide as stabilizer) led to immediate and complete dissolution of PbTe and to the concomitant in situ formation of large Au nanocrystals even at room temperature. Hence dodecylamine is indeed acting as the reducing agent of Au(III) species to Au(0) and possibly even to Au(I) species at the surface of the PbTe nanocrystals. Overall, the actual formation of these core–shell nanostructures could be described therefore as the result of diffusion of mainly Au(0) species inward, balanced by Pb2+ and Te2 diffusion outward, although we cannot exclude as side reactions redox processes involving Te2 and Au(I) and Au(III) species. Additionally we cannot completely rule out some degree of cation exchange of Pb2+ ions with Au(I) and Au(III) species (see section 4.2), mainly because we cannot determine the oxidation states of the various Au species in the shell region and how these oxidation states are distributed among the Au species. 4.2. In situ formation of Au–Te alloys during TEM observations

Fig. 5 (a) A wide-field low magnification image of the PbTe nanocrystals initially coated with a thick amorphous shell, after they were reacted with the AuCl3 solution for 15 min at 90  C. Several small highcontrast domains are located inside the nanocrystals and additionally almost each nanocrystal carries a large balloon-shaped Au domain attached to it. The inset reports a detail of the sample at twice the magnification. If the tiny Au domains were located on the surface of the PbTe nanocrystals and not at their interior, then on some nanocrystals they would appear to be located on one side of the PbTe particles, since the various nanocrystals lay on the supporting carbon film with all possible spatial orientations. However, we have never observed such cases. (b) A HRTEM image of an area of the same sample. (c) A HAADF image of a region of the same grid in which several typologies of nanocrystals can be found. (d) Magnified views of some selected nanocrystals, taken from (c), in which it is possible to identify a cavity in each nanocrystal, likely due to the detachment of the large balloon-shaped Au domain previously attached to it. (e) Low magnification image of the ‘‘close-to-cubic’’ shaped PbTe nanocrystals, after they were reacted with the AuCl3 solution for 15 min at 90  C.

This journal is ª The Royal Society of Chemistry 2010

The discrepancy between lattice parameter values in the crystalline Au domains located inside the nanocrystals found by HRTEM and those found by XRD was somehow puzzling. To get insights into possible causes of this discordance we performed spatial resolved EDX analysis to quantify the chemical concentrations of Pb, Te and Au in the nanocrystals. As test sample we chose the nanocrystals whose low magnification TEM image is displayed in Fig. 5a. Two different approaches were followed in the EDX analysis. In the first approach (short exposure) the beam (1 nm wide) was scanned on areas about 200 nm wide (as in Fig. 5c), such that the beam irradiation on the particles was highly reduced and thus there was lower probability of sample modification during irradiation. The spectra from the cores and from the shells of the structures (such as the particles denoted by ‘‘a’’ in Fig. 5c) were recorded by integrating the corresponding points from the acquired map. In a second approach (long exposure) the beam was scanned on the cores and on the shells of individual nanocrystals (therefore the scanned areas were a few nanometres wide), which resulted in a higher dose of electrons per each nanocrystal with respect to the previous case (approximately 100 times higher if considering also the time integration). The data collected on several nanocrystals were then averaged to improve statistics. To estimate the concentrations in the highcontrast cores, we needed to subtract the contribution from the shell surrounding the top and bottom part of the nanocrystal (always present when a projection of the sample is viewed). In order to do so, we subtracted the shell spectrum from the J. Mater. Chem., 2010, 20, 1357–1366 | 1363

acquired spectrum from the core regions, by normalizing it to the Pb intensity. Consistency was checked by looking at the residual intensity of Te (negative counts for Te would point to a wrong assumption). The results of the chemical quantification of Te, Pb, and Au are in these two approaches are shown in Table 1, and the fits are shown in Fig. 6. It is clear from the figure and from Table 1 (first row) that under low dose the spectrum from the high contrast core regions consists of Au only. The remaining amorphous regions (the ‘‘shell’’), still under low dose, have roughly a Pb35Te57Au8 stoichiometry (Table 1 second row), although the Au characteristic X-ray photons could additionally originate from the ubiquitous ultra-small Au particles that contaminated this sample. Other possible reasons for Au contamination in these regions could be cation exchange reactions involving Pb2+ with Au+ or Au3+, or simply the presence of Au(0) species on their way towards the inner Au crystalline domains. The high Te : Pb molar ratio could be explained actually as due to a partial redox reaction of Te2 ions in the lattice by Au(I) or Au(III) species, leading to Te species with oxidation number equal to 0 or higher

Table 1 Chemical composition of the high contrast core regions and of the shell regions, as determined from spatially resolved EDX analysis on the nanocrystals displayed in Fig. 5c. The reported data represent an average over measurements on 10 isolated nanocrystals. The amount of Pb in the core is not reported since this was always around 1%, i.e. below the standard deviation

Core (short exposure) Shell (short exposure) Core (long exposure) Shell (long exposure)

Te

Pb

Au

Error