- Email: [email protected]
Encapsulation of copper and zinc oxide nanoparticles inside small diameter carbon nanotubes
Accepted Manuscript Encapsulation of copper and zinc oxide nanoparticles inside small diameter carbon nanotubes Dennis Großmann, Axel Dreier, Christian W. Lehmann, Wolfgang Grünert PII: DOI: Reference:
S1387-1811(14)00571-X http://dx.doi.org/10.1016/j.micromeso.2014.09.057 MICMAT 6802
To appear in:
Microporous and Mesoporous Materials
Received Date: Revised Date: Accepted Date:
6 June 2014 22 September 2014 24 September 2014
Please cite this article as: D. Großmann, A. Dreier, C.W. Lehmann, W. Grünert, Encapsulation of copper and zinc oxide nanoparticles inside small diameter carbon nanotubes, Microporous and Mesoporous Materials (2014), doi: http://dx.doi.org/10.1016/j.micromeso.2014.09.057
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ENCAPSULATION OF COPPER AND ZINC OXIDE NANOPARTICLES INSIDE SMALL DIAMETER CARBON NANOTUBES Dennis Großmann, Axel Dreier1,+, Christian W. Lehmann1, Wolfgang Grünert*
Lehrstuhl für Technische Chemie, Ruhr-Universität Bochum, P.O. Box 102148, D-44780 Bochum, Germany 1
Max-Planck-Institut für Kohlenforschung, D-45470 Mülheim a. d. Ruhr, Germany
*Corresponding Author: Tel. +49 234 322 2088, email. [email protected] +
deceased August 11, 2013
Keywords:
Carbon nanotubes, filling, nanoparticles, Cu, ZnO,
1
Abstract Copper and zinc oxide nanoparticles have been reproducibly deposited into carbon nanotubes (CNT) of 6-7 nm internal diameter via simple impregnation techniques with different metal salts followed by thermal decomposition of the precursors and reduction in H2 in case of Cu. Oxygen functionalization via a gas-phase method involving thermal shocks was a critical step while traditional functionalization with nitric acid resulted in failures. Intra-CNT location of CuO particles could be proven by STEM images, and was examined by TEM for materials prepared by various routes. It was found that Cu and Zn oxide nanoparticles could be deposited throughout the whole interior CNT space. The filling capacity depended on the preparation conditions, on conditions of subsequent precursor decomposition, and on the inner diameter of the CNTs. After the reduction of the CuO nanoparticles, XRD, XAFS, and N2O reactive frontal chromatography indicated a bimodal particle size distribution due to the presence of agglomerates outside the CNTs. To enhance selectivity for endohedral location, a washing step with HNO3 with the inner CNT space blocked by xylene was applied to selectively remove aggregates in the outer space. Based on the best procedures for introduction of CuO and ZnO, a bimetallic CuZnO@CNT sample was prepared via a consecutive preparation route.
2
1.
Introduction
Since the upsurge of research on carbon nanotubes (CNTs) following Iijima’s report on their formation during fullerene synthesis [1], these materials have been also extensively studied as catalysts or catalyst supports. This effort has recently created evidence on a dependence of catalytic and redox properties of catalytic components on their location. Thus, Fe2O3 turned out to be reduced at lower temperatures when deposited in the interior space than on the outside of small CNTs [2], catalytic activities of Fe (in Fischer-Tropsch synthesis [3]) or Rh (for the formation of ethanol from synthesis gas [4]) within the hollows were superior to those exhibited by the same metals when deposited onto the CNTs. This was ascribed to an “electronic confinement effect”, which causes a shift in the π-electron density to occur when the inner diameter (i.d.) decreases below 10 nm [5, 6]. These results have raised the interest in filling CNTs of very small i.d. with metals or metal oxides also among catalytic scientists. The deposition of metal oxide nanoparticles within CNTs was described already a decade ago [7]. Most studies undertaken since then employ impregnation techniques supported by ultrasound, performed with cut nanotubes to facilitate the filling of the inner space by the solutions employed [8]. Such impregnation can be combined with selective washing steps to remove species simultaneously deposited on the exterior surfaces [9]. Impregnation with different solvents favorably interacting with the interior or the (functionalized) exterior CNT surfaces has been shown to result in selective deposition of the particles in the interior or on the outside [10]. Choice of ligands for the precursors to be deposited [11] or chemical modification of exterior or interior surfaces [12] are strategies for increasing deposition selectivity on the basis of molecular recognition. Volatile compounds may be introduced into CNTs via the gas phase [13], and even the use of supercritical CO2 has already been reported [14]. For CNTs with >15 nm i.d., there is a rich literature on deposition of various compounds into the interior space (e.g. [15-17]). Filling of CNTs becomes, however, more difficult with smaller diame-
3
ters [18]. Deposition of oxide clusters into CNTs of <10 nm i.d. has succeeded so far only by ultra-sound supported impregnation [8]. In the following, we report the deposition of copper and zinc oxide particles in the interior space of CNTs with 3-5 nm nominal i.d. This effort was driven by the interest to create model catalysts for the investigation of Cu-ZnO interactions in industrial methanol synthesis catalysts. While there are numerous reports on the introduction of Fe, Co, Rh, Ru, or Pd into CNTs of similar sizes [2, 8, 10, 19-21], literature on the deposition of Cu and Zn oxide is scarce. Problems may be envisaged for Cu due to decreasing interactions between the graphene surface and transition metal atoms at increasing number of d-electrons [22, 23]. While this may interfere with the stabilization of small Cu metal particles, already the synthesis of Cu oxide nanoparticles is complicated by the existence of highly mobile hydroxo nitrate intermediates, which have to be carefully avoided to obtain satisfactory dispersions (e.g., in nanoporous SiO2 [24]). There has been a report on successful deposition of Cu oxide particles into CNTs of 4-10 nm nominal i.d. [25], however, the route did not provide the desired results with some of our materials and may therefore be not general. Literature on modification of CNTs with ZnO appears to be limited to deposition on the exterior surface by now [26]. We will now describe new, reproducible routes for the selective encapsulation of copper and zinc oxide nanoparticles in the channels of CNTs with small diameters. The particles were studied by electron microscopy (STEM, TEM), XRD, and XAFS. In the case of copper, the reduction behavior of the oxide species was investigated by temperature-programmed reduction (TPR). After reduction, the size of the metal particles was assessed by N2O reactive frontal chromatography (RFC), by XAFS, and by XRD.
4
2. Experimental 2.1. Material Synthesis Raw CNTs (CheapTubes Inc., Brattleboro, U. S. A., nominal i.d. 3-5 nm, length 10-50 µm) were functionalized by thermal stress following a procedure adapted from literature [27]. By moving a quartz tube into and out of a tubular furnace, the raw material was instantaneously heated to 873 K in a flow of 150 ml/min dry air and after 10 minutes rapidly cooled to room temperature where it was kept for 12 min. This sequence was repeated twice. The procedure removes the caps of the nanotubes as well as amorphous carbon and it is able to shorten long CNTs [27]. It is very severe, further repetition leads to complete loss of the CNTs. CNTs functionalized by this route will be denoted as O-CNT. From these, N-CNTs were made by treatment in 10 % NH3/He at 673 K according to a procedure outlined in [28] (for details see [29]). Alternatively, CNTs were also functionalized by refluxing in concentrated HNO3 (68 %) for 14 h, subsequent washing and drying (→ CNT-A). For some experiments, short commercial CNTs were used (CheapTubes Inc., nominal i.d. 3-5 nm, length 0.5-2 µm). In addition, wide CNTs (Pyrograf, Applied Science, i.d. around 50 nm) were employed for a reference preparation. Both CNT materials were functionalized via the gas-phase route, the wide CNTs will by designated as CNT-Py. The following route based on O-CNTs will be described as the “standard route” for deposition of CuO nanoparticles in the CNT interior: O-CNTs were added into a solution of Cu(NO3)2 in tetrahydrofurane (THF) (0.0096 mol/l) followed by tip sonication for 1 h while cooling with a water bath. “Tip sonication” denotes a procedure where the ultrasonic power is directly introduced to the solution by a Sonotrode (UIS250LK tip sonicator, Hielscher Ultraschall-Technologie, maximum power 100W). Subsequently, the solvent was evaporated slowly under ambient conditions prior to drying the solid at 323 K for 10 h. For precursor decomposition, 5
the sample was heated to 523 K in He at a rate of 5 K/min and the temperature was kept for 2 h. The decomposition temperature was specified on the basis of a thermogravimetric measurement (see Fig. S1, supporting information). The lowest possible temperature was chosen because further temperature increase was found to result in significant CuO particle sintering and clustering [29]. The material produced via this route will be denoted as CuOts/O-CNT. A copper loading of 10 wt-% was targeted in the preparations. Parameters of the preparation, e.g. Cu source, solvent, type of CNT functionalization, were varied for optimization purposes. In addition, a number of preparation techniques described in literature for the deposition of other metals species in CNTs [3, 5, 20] were adopted for Cu by changing the metal precursor to copper nitrate. This work is described in detail in [29], only conclusions thereof will be given below. A washing procedure to remove agglomerates outside the CNTs, the formation of which could not be avoided even in the best preparations, was developed following ideas outlined in [5, 9, 10]. First, the inner channels of the calcined sample were blocked with xylene (1.25 g per 50 mg sample) by subjecting the slurry of CNTs in xylene to an ultrasonic bath for 4 h. The solid was filtered off, added to concentrated HNO3 (68 wt-%) and stirred for 30 min. Finally, the sample was filtered off again, washed with water and calcined under the same conditions as before. ZnO nanoparticles were encapsulated in the CNT interior by impregnating O-CNTs with a solution of zinc citrate (Zn3C12 H10O14 * 3 H2O) in aqueous ammonia (0.126 mol/l) to incipient wetness. The resulting solid was dried at 323 K for 10 h before it was heated in He to 623 K at a rate of 5 K/min where the temperature was kept for 2h. It should be noted that this decomposition temperature, which was again chosen on the basis of TG data (Fig. S1) is higher than that employed for CuO/CNT preparations. The sample with a nominal zinc loading of 10 wt% will be labeled as ZnOiwi/O-CNT. An alternative preparation paralleled the one described
6
above for copper: the functionalized CNTs were impregnated under tip sonication with a solution of Zn(NO3)2 in THF (0.0094 mol/l). After impregnation, the same drying and thermal treatment at 623 K was applied. Based on the previously described routes for the preparation of monometallic samples, a consecutive preparation route was developed for the bimetallic samples. First, zinc oxide nanoparticles were encapsulated in the CNT interior following the route for ZnOiwi/O-CNT before the sample was impregnated with copper as described for CuOts/O-CNT. After both impregnations, the washing step was applied. For comparison, Cu and Zn were also introduced into CNT-Py. Due to the large diameter facilitating liquid impregnation, a simultaneous route was employed here: copper and zinc nitrate were dissolved in THF and impregnated under tip sonication. 2.2. Characterization Properties of the CNT supports were studied by nitrogen physisorption and by XPS. BET surface areas and pore volumes were determined with a modified Autosorb 1C instrument (Quantachrome) after outgassing the samples at 523 K in He. XPS spectra were measured with an ultrahigh vacuum setup equipped with a high-resolution Gammadata-Scienta SES 2002 analyzer using monochromatized Al Kα as incident radiation. C 1s, O 1s and N 1s were integrated over Shirley-type backgrounds with the Casa-XPS software, for calculation of atomic compositions, the sensitivity factors provided by this software were used. Elemental analysis was made by atomic absorption spectroscopy. Samples were melted in sodium peroxide and then dissolved in aqueous hydrochloric acid. Afterwards the atomic absorption was measured with a Varian SpectrAA 220 spectrometer. Transmission electron microscopy (TEM) images were obtained on a Hitachi H-7500 instrument operated at an accelerating voltage of 100 kV. The samples were dispersed in ethanol under ultrasonication and placed onto a carbon film supported over a copper grid. The micro-
7
scope was also equipped with an energy-dispersive X-Ray detector. Scanning electron microscopy (SEM) measurements were performed on a Hitachi S-5500 Ultra-High Resolution FE-SEM equipped with a BF/DF Duo-STEM detector. The microscope was operated at an accelerating voltage of 30 kV. The samples were placed on a 3 nm carbon-film, placed on 400 mesh copper grid. X-Ray diffractograms were measured with a Panalytical MPD theta-theta diffractometer using CuKα radiation (λ = 1.54056 Å). The signals were recorded over an angular range of 2 Θ = 10-70° (step size - 0.035°, dwell - 25 s/step). The reduction properties were investigated by temperature-programmed reduction (TPR) in 5% H2 / He, ramping the temperature to 513 K at 2 K/min and monitoring the hydrogen concentration with a thermal conductivity based detector (Rosemount, Hydros 100). For most samples, the specific copper surface area was measured subsequently by N2O reactive frontal chromatography as described in [30, 31]. In-situ XAS spectra (Cu K edge, 8.979 keV; Zn K edge, 9.659 keV) were measured in transmission mode at Hasylab, Hamburg (Station C). A Si(111) double-crystal monochromator was used, detuned to 50% of the maximum intensity to exclude higher harmonics from the XRay beam. The hand-pressed sample pellets were reduced in an in-situ EXAFS cell [32] in a flow of 5 % H2/He at a temperature ramp of 2 K/min and a maximum temperature of 513 K for 1 h (H2_513K). XAFS spectra were measured at liquid nitrogen temperature (LNT), and the spectrum of a Cu (or Zn) foil was recorded at the same time for energy calibration. Some samples were then re-heated in He to 513 K and subsequently reduced at 873 K either in dilute hydrogen or in dilute CO (10 % in He). Finally, the XAFS spectra were measured at liquid nitrogen temperature again. The data was analyzed using the VIPER software package [33]. For background subtraction, a Victoreen polynomial was fitted to the pre-edge region. The smooth atomic background, µ 0, was estimated using a smoothing cubic spline. The Fourier analysis of the k²-weighted experimental function was performed with a Kaiser Window.
8
From the CuK EXAFS, the Cu(0)-Cu path was analyzed on the basis of a FEFF8.20 calculation for cubic copper metal [34] in order to determine structural parameters, in particular the coordination number (C. N.), which permits estimating the primary particle sizes based on a spherical particle model [35]. Thermogravimetric analyses (see Fig. S1) were performed in a CAHN TG-2131 microbalance (Raczek Analysentechnik). The sample was initially heated to 323 K with 2 K/min. After 2 h at 323 K, the temperature was increased to 623 K with 5 K/min. All steps were done in flowing helium (113 ml/min).
3. Results 3.1. Carbon nanotubes Relevant physicochemical data of the CNTs employed are collected in Table 1. The increased BET surface areas and pore volumes after oxygen functionalization via gas-phase or HNO3 routes show that the CNT interior is accessible after both procedures. The final N functionalization reproducibly increases both pore volume and BET surface area of the O-CNTs employed. The short CNTs exhibit larger surface areas and pore volumes than the original ones. While the surface areas obtained after liquid and gas-phase functionalization are similar, the pore volume of CNT-A is clearly larger than that of O-CNT. The difference of ca. 20 % is too large to arise exclusively from the growth catalyst remaining in O-CNT. A contribution may come from the less selective action of oxygen, which may destroy or crack tubes leaving some non-porous debris behind, which still contributes to the BET surface area. The surface compositions derived by XPS shows higher oxygen contents for the liquid functionalization – ca. 14 at-% in CNT-A (Table 1). The smaller surface oxygen content achieved with the gas-phase method is due to the volatile character of the most likely reaction products CO and CO2. N functionalization results in significant loss of oxygen. The amount of N intro-
9
duced is small but reproducible. The short CNTs contain already a significant amount of oxygen, which was only marginally increased by the gas-phase functionalization employed. 3.2. Copper into CNTs - Impregnation Fig. 1 shows images taken from the CuOts/O-CNT sample. By switching from scanning (Fig 1a) to scanning transmission mode (Fig. 1b) the presence of particles in the CNT interior can be demonstrated. Apparently they have entered deeply into the channels and are not concentrated at opened CNT tips. The rough morphology of the CNT surface is caused by the severe functionalization. Fig. 2 shows the XRD patterns of the functionalized CNTs and the copper containing sample (9.1 wt-% Cu according to AAS analysis). Two strong reflections around 26.1° and 43.0° arise from the CNTs, two weak reflections at 37.31° and 63.08° can be ascribed to magnetite (ICSD #01-075-0449) formed from the residual growth catalyst. The XRD pattern of the Cu containing sample clearly shows diffraction lines arising from the monoclinic copper (II) oxide phase (tenorite, ICSD #01-089-2529, strongest reflections at 35.58° and 38.74°). Estimation by the Scherrer equation resulted in an average CuO crystallite size of 11 nm, somewhat in excess of the typical inner diameter, which was estimated to be 6-7 nm from the TEM images shown below. In Fig. 3, the XAS spectra (XANES and EXAFS) of the initial CuOts/OCNT are shown. One can clearly see that copper was exclusively present as Cu (II) oxide. The shapes of the XANES as well as of its first derivative and even the EXAFS spectrum are close to those of the CuO reference. In TPR, the Cu(II) oxide was reduced to metallic copper, which caused an asymmetric signal peaking at 449 K (Fig. 4). The asymmetry of the peak, with a small shoulder at low temperatures, might indicate the presence of very small particles, and hence, a bimodal particle size distribution. Under the conditions of our TPR experiment, bulk CuO physically mixed with SiO2 at comparable concentration was reduced in a single peak at 503 K. In Fig. 5, TEM im-
10
ages of an unreduced and a reduced sample are compared. In the reduced sample, particles can be seen in the CNTs as well which were identified as Cu particles by EDX. Some of them seem to fill out the whole inner diameter like a plug. The particles, did, however, not grow along the x-axis of the CNT channels, and hence, the extent of sintering remained moderate. The diffractogram obtained after this reduction is depicted in Fig. 2. Together with signals of magnetite, a reflection of Cu metal at 50.5 ° (the one at 43.4 ° is obscured by a carbon signal) can be clearly seen. The particle size estimated from the line width was 17.5 nm. The EXAFS spectrum of the sample reduced at 473 K in hydrogen is shown in Figure 3 b. It resembles the spectrum of the metal foil. The first shell is weaker than in the spectrum of the reference foil, which indicates a small particle size. From the model fit, a Cu-Cu C. N. of 9.8 was obtained (Cu-Cu distance - 2.54 Å), which results in an estimated average particle size of 2.1 nm. This is smaller than the size seen in the TEM images and far below the size of the Cu particles obtained from XRD. On the other hand, the higher scattering events are all intense up to 7 Å (uncorrected) in the spectrum of the reduced sample (Fig. 3b). Although their intensity is weaker than in the Cu foil, there is hardly any amplitude decay along the distance, which would be expected for particles of 2 nm size. Therefore, the spectrum indicates the presence of large particles coexisting with very small ones. One may suggest from this that the intra-CNT particles may be polycrystalline and made up from very small primary crystallites. The large particles are probably in extra-CNT aggregates. From the N2O RFC measured after TPR, an exposed specific copper surface area of 14.3 m²/g Cu was obtained. On a purely geometric basis, this corresponds to a particle size of 47 nm, far beyond the size of Cu particles seen in the TEM and assessed by EXAFS and even by XRD. Apparently, a large part of the Cu surface is inaccessible. This may be due to contact with the CNT walls and with other copper particles in polycrystalline aggregates.
11
N-functionalization of CNTs has been recently described to favor intra-CNT deposition of particles [36]. Application of the standard tip sonication method to N-CNTs proved successful, visual inspection of the TEM images did, however, not show improvements that would justify the additional step from oxygen to nitrogen functionalization [29] (see also Fig. S2). As mentioned in the introduction, impregnation of Cu species into CNTs functionalized by treatment with nitric acid (CNT-A) as reported in [25] failed in our hands. A typical TEM image taken from a sample made by our standard route with CNT-A is shown in Fig. 6. Cu oxide particles are deposited in high dispersion, but almost exclusively on the external surface. While in some other frames there are a few particles possibly in endohedral locations, it is clear that the vast majority of copper is outside the CNTs, in obvious contrast to the result obtained with the gas-phase functionalization. Several routes described in literature for the introduction of other elements into CNTs functionalized with HNO3 (e.g. aqueous impregnation, urea deposition precipitation, incipient wetness impregnation with both water and ethanol solvents [3, 5, 10, 37]) failed when modified for the use with copper nitrate [29]. As an example, Figure S3 shows TEM images of samples prepared along the route of Wang et al. for the deposition of Fe oxide into CNTs [38] using aqueous nitrate solution and a standard ultrasonic bath (40 W). In Figure S3a, Fe oxide particles can be seen in the interior space while Cu ended up in large aggregates as seen in Figure S3b. On the other hand, gas-phase functionalization opened the way to CuO nanoparticles encapsulated in CNTs via different routes, e.g. incipient wetness impregnation with different solvents (THF, aqueous ammonia, water) and varying copper precursors (nitrate and citrate). However, from visual inspection of the TEM images, the amount of confined particles was less compared to CuOts/O-CNT in these cases (for details see [29]; as an exception, incipient impregnation with an aqueous copper nitrate solution failed with N-CNTs for unknown reasons). The standard tip sonication route was also successfully applied to the short commercial CNTs (cf.
12
section 2.1.) where we expected a larger particle concentration due to the smaller aspect ratio. While the experiments were successful, the expectation of larger particle loadings was not
a)
fulfilled [29] (see example in Fig. S4). CuO was inserted into large-pore CNT-Py via the standard tip sonication route without problems. A TEM image of this material is shown in Figure S5. The X-ray diffractogram showed the presence of CuO, but also of hematite and of α-Fe [29]. 3.3. Copper into CNTs – Impregnation combined with washing With narrow CNTs, the average CuO particle sizes determined by XRD show that a signifi-
a)
b)
cant amount of copper should have been deposited on extrahedral locations. It was therefore attempted to further increase the ratio between particles in interior and exterior space by a selective washing step in concentrated nitric acid. During the washing, the CNT interior space was blocked by xylene as proposed in [38] for a different purpose. In Fig. 7, SEM images in back-scattering mode of the sample before and after washing are shown. It is quite obvious that the larger agglomerates (shown as bright spots) were almost completely removed. Due to the washing procedure, the copper content of the sample decreased from 9.1 to 4.8 wt-%. The TPR profile shows a maximum reduction temperature in the same range as before (443 K) but also a pronounced shoulder at lower temperatures (426 K), which can be attributed to very small copper particles (Fig. 4). Notably, the exposed Cu surface determined by N2O RFC increased from 14.3 to 47.9 m²/g Cu by the washing step. The latter corresponds to a particle size of 14 nm. While this is still larger than the inner pore diameter, the accessibility of the Cu surface is apparently strongly increased by removal of the big aggregates. Thus, by combining the standard tip sonication procedure with selective washing, materials with Cu particles predominantly within the CNTs can be reproducibly prepared. Despite this success, our results warrant a note of caution with respect to the use of xylene for pore blocking. Extension of the acidic washing step to 2 h could lead to complete loss of cop-
13
per. In electron micrographs, particles within the CNTs were rare after washing. Some nanoparticles found after the washing step are depicted in Figure 8a. They were initially considered as CuO particles because of obvious differences to growth catalyst residues with respect to morphology and contrast. Later, a high-resolution EDX investigation revealed that the observed particles contained iron and nickel, apparently from the growth catalyst. Hence, the sample was investigated in more detail by STEM, and an example is shown in Figure 8b. The images show the presence of CuO as small disordered patches and dots in the range of 1-3 nm, highly dispersed within the complete CNT surface, even within the walls, in which the severe functionalization treatment had apparently created cracks and pores. These results suggest that the xylene had not well blocked the interior of our CNTs or had been gradually displaced by the acid, which is in some conflict with the high endo/exohedral selectivity obtained with this method in preparations shown in [38]. 3.4. Zinc into CNTs Routes for deposition of zinc oxide into CNTs were developed based on the experience made with copper. Therefore, the tip sonication route with zinc nitrate in THF solution (ZnOts/OCNT) was used as a reference point. Fig. 9a shows a TEM image of ZnOts/O-CNT. It can be seen that small nanoparticles were encapsulated in the CNT interior. The particles are located both at the opened CNT tip and deeply in the channel and their dimensions are slightly smaller than the inner diameter although some larger agglomerates were detected as well. This was also confirmed by the X-Ray diffraction pattern, which is shown in Fig. 2. The strong and sharp reflection lines at 31.86°, 35.53°, and 36.36° and a number of additional peaks can be ascribed to the hexagonal zincite structure of ZnO (ICSD #01-073-8765). An average crystallite size of 16.5 nm resulted from the application of the Scherrer equation to the line widths, which is considerably larger than the inner diameter of the CNTs. By testing other preparation techniques described above for the confinement of Cu oxide nanoparticles, incipient wetness impregnation of O-CNTs with zinc citrate in ammonia solution 14
has been found to give even better results regarding the amount and the dimensions of the confined nanoparticles. This was not only observed in TEM images (see example in Fig. 9b) where no larger agglomerates could be detected, but also in the X-Ray diffractogram (Fig. 2). The reflections were significantly broadened, and a crystallite size 6.5-7.0 nm was estimated applying the Scherrer equation to the reflections at 31.86° and 56.42°, respectively. This was the first example where the crystallite size was in the range of the inner CNT channel diameter. On the other hand, application of the same incipient wetness impregnation route with zinc citrate to CNT-A failed completely: ZnO particles were deposited only on the external surfaces (Fig. S6). The reducibility of the confined zinc oxide species was investigated because it had been found that the reduction of confined CuO nanoparticles occurred at lower temperatures compared to extra-CNT species (Fig. 4). However, reduction of the confined ZnO nanoparticles could not be observed in an in-situ XAS study (Figure 10). Neither the XANES (or its first derivative) nor the Fourier-transformed EXAFS spectra showed any indication for the presence of zerovalent zinc species even after reduction in CO at 873 K. The severe conditions of the latter experiment rather caused a better ordering of the ZnO aggregates as indicated by a slight increase of the Zn-Zn shell. However, the distances for the Zn-X (X = O and Zn) shells were found to remain unchanged after reduction. The presence of totally oxidized zinc was confirmed by an X-ray diffractogram of the sample used for the EXAFS investigations (Fig. S7). In summary, zinc oxide nanoparticles were successfully and with high dispersion deposited in the CNT interior. It should be noted that the best preparation routes turned out to be different for copper and zinc. In particular, the temperature tolerated for the decomposition of the copper precursor (523 K) is lower than that required for the decomposition of the zinc precursor (623 K). Therefore, the bimetallic sample was prepared via a consecutive route.
15
3.5. Copper and Zinc into CNTs For the deposition of the Cu/ZnO catalyst system into CNTs, zinc oxide nanoparticles were encapsulated in the O-CNTs interior first by following the route described for ZnOiwi/O-CNT, including precursor decomposition at 623 K and washing in HNO3. Subsequently, copper was added using the procedure for CuOts/O-CNT, including precursor decomposition now at 523 K, and a second washing step with HNO3. In TEM images taken in this state of the material, no particles or agglomerates could be observed at all (Figure 11a). Likewise, the XRD pattern (Fig. 2) did not show any indication for copper or zinc oxide crystalline phases although both elements were detected in the sample by AAS (Cu – 3.8 wt-%, Zn – 1.6 wt-%) and, though with very low intensity, also by XPS [29]. The EXAFS spectra of this sample, which are displayed in Figures 3 (CuK) and 10 (ZnK), contain only the first coordination sphere (oxygen) around Cu and Zn, no higher spheres are visible. Apparently, the preparation, though successive, had resulted in a highly disordered oxide material inside the CNTs. In TPR, the reduction peak occurred at 487 K, significantly higher than in the monometallic samples (Fig. 4), and TEM images taken after this reduction step showed metal particles in the CNT interior (Fig. 11b). For this sample, no XRD was taken after reduction at 513 K due to limited availability because even more severe reduction conditions (10% CO/H2, 673 K, 30 min) did not result in any signals assignable to a Cu-containing phase (Fig. 2). The EXAFS spectra after reduction in dilute hydrogen at 513 K are presented in Figures 3 (CuK) and 10 (ZnK). No changes caused by the reduction can be found in the ZnK spectrum. The CuK spectrum again resembles that of the Cu foil, but with very low intensity of the first shell and almost vanishing signals around 7 Å. From the C. N. of the first Cu-Cu shell (6.9), an average particle size of below 1 nm was estimated. The observation of some intensity around 7 Å indicates the presence of larger particles here as well. In sharp contrast to these data, the ex-
16
posed specific copper surface area as determined by N2O RFC was just 5.3 m²/g Cu, which would correspond to particle sizes of >100 nm. Obviously, some of the problems observed with Cu/O-CNT occurred in this sample as well: The contradictions between the primary particle sizes (EXAFS) and the particles observed in the TEM suggest that the latter may be aggregated from very small crystallites. The contradiction between the sizes just mentioned and the size derived from N2O RFC indicates that most of the Cu surface is covered – by ZnO, by CNT walls, or cut off by pore plugging. The very small primary particle size of the Cu crystallites (<1 nm, as compared to >2 nm in the monometallic sample) suggests an interaction between Cu and Zn species during preparation favoring high dispersion of primary particles. The same trend was obtained also in preparations with silica-based mesoporous host matrices where the bimetallic samples always exhibited a finer Cu dispersion than the monometallic ones [29, 39]. As mentioned in section 2.1., Cu and Zn were introduced into CNT-Py simultaneously, which did not cause any problems. A TEM image of the sample is shown in Fig. S8.
4. Discussion In the results presented above, we have shown that CuO and ZnO nanoparticles can be reproducibly deposited into CNTs of 6-7 nm i.d. when these have been previously functionalized by a gas-phase route that involves thermal stress. The selectivity for the inner space is significant, but needs to be improved by a washing step in HNO3 during which the CNT interior should be blocked, e.g. with xylene. After reduction of the oxide clusters, the copper remains in the CNTs. Its accessibility from the gas-phase was demonstrated by interaction with N2O, where, however, a relatively small accessible surface area was obtained although XAFS indicates primary particle sizes on the order of only 2 nm. This contradiction suggests that much of the Cu surface area is blocked by contact with other Cu particles, with CNT walls, or by occasional plugging of whole CNT cross sections by (polycrystalline) Cu particles.
17
Sequential introduction of ZnO and CuO with subsequent reduction resulted in a CuZnO/CNT sample, where the accessible Cu surface area was even smaller than in the monometallic sample despite significantly smaller primary particle size (XAFS). This may be ascribed to an enhanced degree of surface blocking, e.g. by interaction with ZnO. Routes for introduction of Cu and Zn oxide into CNTs previously functionalized by HNO3, as earlier described in literature [25], failed in our study for ZnO, and in particular for CuO. Particles were predominantly or exclusively deposited onto the external CNT surfaces when the procedures best for O-CNTs (i.e., after gas-phase functionalization) were applied to CNT-A (Fig. 6, S6). The BET surface areas and pore volumes (Table 1) show that our CNTs were open after HNO3 treatment, and the successful deposition of Fe oxide particles (Fig. S3) suggests that the interior of CNT-A could be wetted. A possible reason for this failure has been recently indicated by Machado et al. [40] who found that concentrated HNO3 functionalized the exterior surface of CNTs of comparable inner diameters but left the interior surfaces unchanged. Thus, introduction of Cu and Zn oxo species, which apparently requires strong interaction with the surface, was unsuccessful while deposition of Fe oxide, which probably occurs by precipitation rather than by ion exchange, was possible. Still, this does not explain the contradiction of our observations with those in [25] where Cu oxide particles were introduced in CNTs of similar inner diameter after HNO3 functionalization, with comparable magnitudes of accessible Cu surface areas being achieved. It has been shown recently that the result of functionalization with HNO3 depends critical on properties of the CNT surfaces, e.g. on graphitization degree and defect content [41]. Thus, differences in the properties of CNTs employed may have caused the contradictions mentioned. The apparent failure of our attempt to block the CNT interior by xylene may be another indirect indication for an oxygen functionalization of the interior surfaces. Excellent exoendohedral selectivity has been reported when the CNT interior was blocked with xylene
18
while Ru was deposited on the external surface from an aqueous solution [38]. If the interior of our CNTs were hydrophobic there would be no driving force for the xylene to give the acid access to the interior space, which apparently has happened during our washing procedure: after long extension, all Cu was lost, after our standard duration, the CuO particles were dissolved. The copper was, however, not removed from the interior space. Growth catalyst residuals were apparently also dissolved and formed a few particles during drying while the Cu and (the Zn where present) was recovered in highly disordered layers (Figures 8 and 11, EXAFS spectra of initial CuZnO/O-CNT in Figures 3 and 10b). Activity data measured with the Cu-ZnO/CNT samples mentioned in this paper will be reported elsewhere together with related characterization data showing the response of the samples to relevant pretreatments (for basic information see [29]).
5. Conclusions Copper and zinc oxide nanoparticles were introduced into carbon nanotubes of 3-5 nm nominal interior diameter (actually 6-7 nm), and the CuO particles were reduced to obtain intraCNT Cu nanoparticles. The preparation routes were optimized by systematic variation of the preparation conditions leading to a new standard synthesis route for each oxide. CNTs were previously functionalized by thermal stress in air because attempts to deposit Cu oxide particles into CNTs functionalized by the conventional HNO3 route failed. Wetness impregnation of functionalized CNTs with copper nitrate in THF solution under tip sonication was the optimum route leading to copper containing CNTs with a large amount of confined particles. After reduction, a bimodal particle size distribution was concluded from characterization data (e.g. very low specific copper surface area). Extra-CNT aggregates could be largely removed by a selective washing with HNO3 after blocking the CNT interior with xylene, which increased the accessible Cu surface dramatically. There were, however, indications that the xylene was gradually replaced by acid in the CNT interior during the washing step.
19
For ZnO in CNTs, incipient impregnation with zinc citrate in ammonia solution was the best preparation route. Already before the washing step a sample with an estimated ZnO crystallite size in the range of the inner CNT diameter was obtained, where extra-CNT particles could not be detected. Based on these results, a bimetallic sample containing Cu and ZnO in CNTs was prepared via a consecutive route including two washing steps. While Cu and Zn oxide species formed very small and amorphous aggregates in the as-prepared material, which could not be observed by TEM and XRD, the aggregates (Cu nanoparticles) encapsulated in the CNTs became visible for TEM after reduction.
Acknowledgements We kindly thank Mr. H.-J. Bongard for the SEM/STEM analysis, Dr. Th. Reinecke for XRD measurements, Mrs. S. Plischke for TG and Mrs. N. Arshadin for the TPR and N2O-RFC experiments. The work has been funded by the German Science Foundation (DFG) in the framework of the Collaborative Research Center “Metal-Substrate Interactions in Heterogeneous Catalysis” (SFB 558), which is gratefully acknowledged.
Appendix A. Supplementary Data Supplementary data associated with this article (Figures S1 through S8) can be found in the online version, at …
20
References [1] S. Iijima, Nature 354 (1991) 56-58. [2] W. Chen, X.L. Pan, X.H. Bao, J. Am. Chem. Soc. 127 (2008) 7421-7426. [3] R.M.M. Abbaslou, A. Tavassoli, J. Soltan, A.K. Dalai, Appl. Catal. A 367 (2009) 4752. [4] X.L. Pan, Z.L. Fan, W. Chen, Y.L. Ding, H.Y. Luo, X.H. Bao, Nature Materials 6 (2007) 507-511. [5] X. Pan, X. Bao, Chem. Commun. (2008) 6271-6281. [6] X. Blase, L.X. Benedict, E.L. Shirley, S.G. Louie, Phys. Rev. Lett. 72 (1994) 18781881. [7] S.C. Tsang, Y.K. Chen, P.J.F. Harris, M.L.H. Green, Nature 372 (1994) 159-162. [8] T. Trang Nguyen, P. Serp, ChemCatChem 5 (2013) 3595-3603. [9] Q. Fu, G. Weinberg, D.S. Su, New Carbon Materials 23 (2008) 17-20. [10] J.-P. Tessonnier, O. Ersen, G. Weinberg, C. Pham-Huu, D.S. Su, R. Schlögl, ACS Nano 3 (2009) 2081-2089. [11] E. Castillejos, P.-J. Debouttière, L. Roiban, A. Solhy, V. Martinez, Y. Kihn, O. Ersen, K. Philippot, B. Chaudret, P. Serp, Angew Chem. Int. Ed. 48 (2009) 2529-2533. [12] L. Jiang, L. Gao, Carbon 41 (2003) 2923-2929. [13] K. Schulte, J.C. Swarbrick, N.A. Smith, F. Bondino, E. Magnano, A.N. Khlobystov, Adv. Mater. 19 (2007) 3312-3316. [14] W.A. Solomonsz, G.A. Rance, M. Suyetin, A. La Torre, E. Bichoutskaia, A.N. Khlobystov, Chem. Eur. J. 18 (2012) 13180-13187. [15] W. Baaziz, X. Liu, I. Florea, S. Begin-Colin, B.P. Pichon, C. Ulhaq, O. Ersen, M. SoriaSanchez, S. Zafeiratos, I. Janowska, D. Begin, C. Pham-Huu, J.Mater. Chem. A 1 (2013) 13853-13861. [16] J. Zhang, J.-O. Müller, W. Zheng, D. Wang, D. S. Su, R. Schlögl, Nano Letters 8 (2008) 2738-2743. [17] E. Castillejos, R. Chico, R. Bacsa, S. Coco, P. Espinet, M. Pérez-Cadenas, A. GuerreroRuiz, I. Rodríguez-Ramos, P. Serp, Eur. J. Inorg. Chem. (2010) 5096-5102. [18] D. Ugarte, A. Chatelain, W.A. deHeer, Science 274 (1996) 1897-1899. [19] C. Wang, S. Guo, X. Pan, W. Chen, X. Bao, J.Mater. Chem. 18 (2008) 5782-5786. [20] W. Chen, Z. Fan, X. Pan, X. Bao, J.Am. Chem. Soc. 130 (2008) 9414-9419. [21] H. Friedrich, S. Guo, P.E. de Jongh, X. Pan, X. Bao, K.P. de Jong, ChemSusChem 4 (2011) 957-963. [22] H.L. Zhuang, G.P. Zheng, A.K. Soh, Comp. Mater. Sci. 43 (2008) 823-828. 21
[23] K. Nakada, A. Ishii, Solid State Commun.151 (2011) 13-16. [24] P. Munnik, M. Wolters, A. Gabrielsson, S.D. Pollington, G. Headdock, J.H. Bitter, P.E. de Jongh, K.P. de Jong, J.Phys. Chem. C 115 (2011) 14698-14706. [25] D. Wang, G. Yang, Q. Ma, M. Wu, Y. Tan, Y. Yoneyama, N. Tsubaki, ACS Catal. 2 (2012) 1958-1966. [26] C.S. Chen, T.G. Liu, L.W. Lin, X.D. Xie, X.H. Chen, Q.C. Liu, B. Liang, W.W. Yu, C.Y. Qiu, J.Nanoparticle Res. C7 - 1295 15 (2012) 1-9. [27] M.Q. Tran, C. Tridech, A. Alfrey, A. Bismarck, M.S.P. Shaffer, Carbon 45 (2007) 2341-2350. [28] S. Kundu, W. Xia, W. Busser, M. Becker, D.A. Schmidt, M. Havenith, M. Muhler, Phys. Chem. Chem. Phys. 12 (2010) 4351-4359. [29] D. Grossmann, PhD Thesis, Bochum 2013. [30] O. Hinrichsen, T. Genger, M. Muhler, Chem. Eng. Technol. 11 (2000) 956. [31] G.C. Chinchen, C.M. Hay, H.D. Vandervell, K.C. Waugh, J. Catal. 103 (1987) 79-86. [32] F.W.H. Kampers, T.M.J. Maas, J. van Grondelle, D.C. Brinkgreve, D.C. Koningsberger, Rev. Sci. Instr. 60 (1989) 2635-2638. [33] K.V. Klementiev. VIPER software for Windows (Visual Processing in EXAFS Researches), freeware www.cells.es/Beamlines/CLAESS/software/viper.html [34] J.J. Rehr, R.C. Albers, Rev. Mod. Phys. 72 (2000) 621-654. [35] M. Borovski, J. Phys. IV 7 (1997) C2-259-260. [36] H.J. Schulte, PhD Thesis, Bochum 2011. [37] V.K. Kaushik, C. Sivaraj, P.K. Rao, Appl. Surf. Sci. 51 (1991) 27-33. [38] C.F. Wang, S.J. Guo, X.L. Pan, W. Chen, X.H. Bao, J. Mater. Chem. 18 (2008) 57825786. [39] M.W.E. van den Berg, S. Polarz, O.P. Tkachenko, K.V. Klementiev, M. Bandyopadhyay, L. Khodeir, H. Gies, M. Muhler, W. Grünert, J. Catal. 241 (2006) 446-455. [40] B.F. Machado, M. Oubenali, M. Rosa Axet, T. Trang Nguyen, M. Tunckol, M. Girleanu, O. Ersen, I.C. Gerber, P. Serp, J. Catal. 309 (2014) 185-198. [41] J.-P. Tessonnier, D. Rosenthal, F. Girgsdies, J. Amadou, D. Begin, C. Pham-Huu, D. S. Su, R. Schlögl, Chem. Commun. (2009) 7158-7160.
22
Table 1 – Textural properties and surface compositions of CNTs employed for encapsulation of Cu and ZnO nanoparticles. Sample
Surface area, m2/g
Pore volume, ml/g
raw CNT
78
0.14
CNT-A
188
O-CNT
Composition of near-surface layer, by XPS C, at-%
O, at-%
N at-%
97.7
2.3
-
0.45
85.8
14.2
-
184
0.35
91.5
8.5
-
N-CNT
214
0.37
96.3
3.7
0.3
raw sh-CNT
224
0.51
95.9
4.1
-
O-sh-CNT
279
0.53
94.6
6.2
-
N-sh-CNT
288
0.58
95.5
4.5
0.3
23
Legends Figure 1 – Electron micrograph of the calcined copper-containing sample CuOts/O-CNT, simultaneously recorded in (a) SEM mode and in (b) bright-field STEM mode. Figure 2 – XRD patterns of functionalized O-CNTS (black), calcined copper and zinc oxide containing samples (red) and of the bimetallic sample after severe reduction in CO/H2 (blue). Figure 3 – CuK XAS spectra of Cu species and of Cu-ZnO enclosed in CNTs after different treatments, compared with references; a) XANES and first derivatives, b) Fouriertransformed CuK spectra (k2-weighted, absolute values) at liquid nitrogen temperature. Figure 4 – TPR profiles of different Cu oxide containing samples, compared with CuO mixed with SiO2. Figure 5 – TEM images of copper-containing sample CuOts/O-CNT a) before and b) after standard reduction (H2_513K). Figure 6 – TEM image of CuOts/CNT-A oxide containing sample, with CuO species predominantly deposited on the external surface. Figure 7 – SEM images of sample CuOts/O-CNT recorded with backscattered electrons (BSD mode), a) before and b) after selective washing with concentrated nitric acid. Figure 8 – Electron micrographs of Cu oxide containing sample CuOts/O-CNT after selective washing with concentrated nitric acid, in transmission (a) and STEM mode (b). The particles in (a) were identified as originating from the growth catalyst. In (b), the black patches and dots were identified as Cu compounds by EDX. EDX results at different analysis points (marked with circles): 1) 0 wt% Fe, 0 wt% Ni, 100 wt% Cu; 2) 3.9 wt% Fe, 6.7 wt% Ni, 89.4 wt% Cu; 3) 0 wt% Fe, 0.6 wt% Ni, 99.4 wt% Cu; 4) 3.1 wt% Fe, 12.5 wt% Ni, 84.4 wt% Cu; 5) 0 wt% Fe, 8.2 wt% Ni, 91.8 wt% Cu. Figure 9 – TEM images of Zn oxide containing samples, a) ZnOts/O-CNT and b) ZnOiwi/OCNT. 24
Figure 10 – ZnK XAS spectra of initial (red) monometallic zinc and bimetallic sample, after standard reduction H2_513K (green), and after severe reduction H2 or CO at 873K (blue): a) XANES and first derivatives, b) Fourier-transformed ZnK spectra (k2-weighted, absolute values) measured at liquid nitrogen temperature. Figure 11 – TEM images of bimetallic sample CuZnO/O-CNT a) before and b) after standard reduction.
25
Figure 1 26
Figure 2
27
Figure 3
28
Figure 4
29
a)
Figure 5
30
Figure 6
31
a)
b)
Figure 7
32
a) b)
Figure 8
33
a)
b)
Figure 9
34
Figure 10
35
a)
b)
Figure 11
36
Research Highlights • • • • •
CuO and ZnO deposited in CNTs of 6-7 nm i.d. by ultrasound-assisted impregnation CNT to be functionalized by thermal shock in air rather than by HNO3 Selectivity for intra-CNT location improved by selective washing procedure Reduction of intra-CNT CuO gives polycrystalline Cu NPs with limited accessibility Cu/ZnO aggregates intended to mimic the methanol synthesis catalyst prepared in CNTs
37
Graphical abstract
38
S1387-1811(14)00571-X http://dx.doi.org/10.1016/j.micromeso.2014.09.057 MICMAT 6802
To appear in:
Microporous and Mesoporous Materials
Received Date: Revised Date: Accepted Date:
6 June 2014 22 September 2014 24 September 2014
Please cite this article as: D. Großmann, A. Dreier, C.W. Lehmann, W. Grünert, Encapsulation of copper and zinc oxide nanoparticles inside small diameter carbon nanotubes, Microporous and Mesoporous Materials (2014), doi: http://dx.doi.org/10.1016/j.micromeso.2014.09.057
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ENCAPSULATION OF COPPER AND ZINC OXIDE NANOPARTICLES INSIDE SMALL DIAMETER CARBON NANOTUBES Dennis Großmann, Axel Dreier1,+, Christian W. Lehmann1, Wolfgang Grünert*
Lehrstuhl für Technische Chemie, Ruhr-Universität Bochum, P.O. Box 102148, D-44780 Bochum, Germany 1
Max-Planck-Institut für Kohlenforschung, D-45470 Mülheim a. d. Ruhr, Germany
*Corresponding Author: Tel. +49 234 322 2088, email. [email protected] +
deceased August 11, 2013
Keywords:
Carbon nanotubes, filling, nanoparticles, Cu, ZnO,
1
Abstract Copper and zinc oxide nanoparticles have been reproducibly deposited into carbon nanotubes (CNT) of 6-7 nm internal diameter via simple impregnation techniques with different metal salts followed by thermal decomposition of the precursors and reduction in H2 in case of Cu. Oxygen functionalization via a gas-phase method involving thermal shocks was a critical step while traditional functionalization with nitric acid resulted in failures. Intra-CNT location of CuO particles could be proven by STEM images, and was examined by TEM for materials prepared by various routes. It was found that Cu and Zn oxide nanoparticles could be deposited throughout the whole interior CNT space. The filling capacity depended on the preparation conditions, on conditions of subsequent precursor decomposition, and on the inner diameter of the CNTs. After the reduction of the CuO nanoparticles, XRD, XAFS, and N2O reactive frontal chromatography indicated a bimodal particle size distribution due to the presence of agglomerates outside the CNTs. To enhance selectivity for endohedral location, a washing step with HNO3 with the inner CNT space blocked by xylene was applied to selectively remove aggregates in the outer space. Based on the best procedures for introduction of CuO and ZnO, a bimetallic CuZnO@CNT sample was prepared via a consecutive preparation route.
2
1.
Introduction
Since the upsurge of research on carbon nanotubes (CNTs) following Iijima’s report on their formation during fullerene synthesis [1], these materials have been also extensively studied as catalysts or catalyst supports. This effort has recently created evidence on a dependence of catalytic and redox properties of catalytic components on their location. Thus, Fe2O3 turned out to be reduced at lower temperatures when deposited in the interior space than on the outside of small CNTs [2], catalytic activities of Fe (in Fischer-Tropsch synthesis [3]) or Rh (for the formation of ethanol from synthesis gas [4]) within the hollows were superior to those exhibited by the same metals when deposited onto the CNTs. This was ascribed to an “electronic confinement effect”, which causes a shift in the π-electron density to occur when the inner diameter (i.d.) decreases below 10 nm [5, 6]. These results have raised the interest in filling CNTs of very small i.d. with metals or metal oxides also among catalytic scientists. The deposition of metal oxide nanoparticles within CNTs was described already a decade ago [7]. Most studies undertaken since then employ impregnation techniques supported by ultrasound, performed with cut nanotubes to facilitate the filling of the inner space by the solutions employed [8]. Such impregnation can be combined with selective washing steps to remove species simultaneously deposited on the exterior surfaces [9]. Impregnation with different solvents favorably interacting with the interior or the (functionalized) exterior CNT surfaces has been shown to result in selective deposition of the particles in the interior or on the outside [10]. Choice of ligands for the precursors to be deposited [11] or chemical modification of exterior or interior surfaces [12] are strategies for increasing deposition selectivity on the basis of molecular recognition. Volatile compounds may be introduced into CNTs via the gas phase [13], and even the use of supercritical CO2 has already been reported [14]. For CNTs with >15 nm i.d., there is a rich literature on deposition of various compounds into the interior space (e.g. [15-17]). Filling of CNTs becomes, however, more difficult with smaller diame-
3
ters [18]. Deposition of oxide clusters into CNTs of <10 nm i.d. has succeeded so far only by ultra-sound supported impregnation [8]. In the following, we report the deposition of copper and zinc oxide particles in the interior space of CNTs with 3-5 nm nominal i.d. This effort was driven by the interest to create model catalysts for the investigation of Cu-ZnO interactions in industrial methanol synthesis catalysts. While there are numerous reports on the introduction of Fe, Co, Rh, Ru, or Pd into CNTs of similar sizes [2, 8, 10, 19-21], literature on the deposition of Cu and Zn oxide is scarce. Problems may be envisaged for Cu due to decreasing interactions between the graphene surface and transition metal atoms at increasing number of d-electrons [22, 23]. While this may interfere with the stabilization of small Cu metal particles, already the synthesis of Cu oxide nanoparticles is complicated by the existence of highly mobile hydroxo nitrate intermediates, which have to be carefully avoided to obtain satisfactory dispersions (e.g., in nanoporous SiO2 [24]). There has been a report on successful deposition of Cu oxide particles into CNTs of 4-10 nm nominal i.d. [25], however, the route did not provide the desired results with some of our materials and may therefore be not general. Literature on modification of CNTs with ZnO appears to be limited to deposition on the exterior surface by now [26]. We will now describe new, reproducible routes for the selective encapsulation of copper and zinc oxide nanoparticles in the channels of CNTs with small diameters. The particles were studied by electron microscopy (STEM, TEM), XRD, and XAFS. In the case of copper, the reduction behavior of the oxide species was investigated by temperature-programmed reduction (TPR). After reduction, the size of the metal particles was assessed by N2O reactive frontal chromatography (RFC), by XAFS, and by XRD.
4
2. Experimental 2.1. Material Synthesis Raw CNTs (CheapTubes Inc., Brattleboro, U. S. A., nominal i.d. 3-5 nm, length 10-50 µm) were functionalized by thermal stress following a procedure adapted from literature [27]. By moving a quartz tube into and out of a tubular furnace, the raw material was instantaneously heated to 873 K in a flow of 150 ml/min dry air and after 10 minutes rapidly cooled to room temperature where it was kept for 12 min. This sequence was repeated twice. The procedure removes the caps of the nanotubes as well as amorphous carbon and it is able to shorten long CNTs [27]. It is very severe, further repetition leads to complete loss of the CNTs. CNTs functionalized by this route will be denoted as O-CNT. From these, N-CNTs were made by treatment in 10 % NH3/He at 673 K according to a procedure outlined in [28] (for details see [29]). Alternatively, CNTs were also functionalized by refluxing in concentrated HNO3 (68 %) for 14 h, subsequent washing and drying (→ CNT-A). For some experiments, short commercial CNTs were used (CheapTubes Inc., nominal i.d. 3-5 nm, length 0.5-2 µm). In addition, wide CNTs (Pyrograf, Applied Science, i.d. around 50 nm) were employed for a reference preparation. Both CNT materials were functionalized via the gas-phase route, the wide CNTs will by designated as CNT-Py. The following route based on O-CNTs will be described as the “standard route” for deposition of CuO nanoparticles in the CNT interior: O-CNTs were added into a solution of Cu(NO3)2 in tetrahydrofurane (THF) (0.0096 mol/l) followed by tip sonication for 1 h while cooling with a water bath. “Tip sonication” denotes a procedure where the ultrasonic power is directly introduced to the solution by a Sonotrode (UIS250LK tip sonicator, Hielscher Ultraschall-Technologie, maximum power 100W). Subsequently, the solvent was evaporated slowly under ambient conditions prior to drying the solid at 323 K for 10 h. For precursor decomposition, 5
the sample was heated to 523 K in He at a rate of 5 K/min and the temperature was kept for 2 h. The decomposition temperature was specified on the basis of a thermogravimetric measurement (see Fig. S1, supporting information). The lowest possible temperature was chosen because further temperature increase was found to result in significant CuO particle sintering and clustering [29]. The material produced via this route will be denoted as CuOts/O-CNT. A copper loading of 10 wt-% was targeted in the preparations. Parameters of the preparation, e.g. Cu source, solvent, type of CNT functionalization, were varied for optimization purposes. In addition, a number of preparation techniques described in literature for the deposition of other metals species in CNTs [3, 5, 20] were adopted for Cu by changing the metal precursor to copper nitrate. This work is described in detail in [29], only conclusions thereof will be given below. A washing procedure to remove agglomerates outside the CNTs, the formation of which could not be avoided even in the best preparations, was developed following ideas outlined in [5, 9, 10]. First, the inner channels of the calcined sample were blocked with xylene (1.25 g per 50 mg sample) by subjecting the slurry of CNTs in xylene to an ultrasonic bath for 4 h. The solid was filtered off, added to concentrated HNO3 (68 wt-%) and stirred for 30 min. Finally, the sample was filtered off again, washed with water and calcined under the same conditions as before. ZnO nanoparticles were encapsulated in the CNT interior by impregnating O-CNTs with a solution of zinc citrate (Zn3C12 H10O14 * 3 H2O) in aqueous ammonia (0.126 mol/l) to incipient wetness. The resulting solid was dried at 323 K for 10 h before it was heated in He to 623 K at a rate of 5 K/min where the temperature was kept for 2h. It should be noted that this decomposition temperature, which was again chosen on the basis of TG data (Fig. S1) is higher than that employed for CuO/CNT preparations. The sample with a nominal zinc loading of 10 wt% will be labeled as ZnOiwi/O-CNT. An alternative preparation paralleled the one described
6
above for copper: the functionalized CNTs were impregnated under tip sonication with a solution of Zn(NO3)2 in THF (0.0094 mol/l). After impregnation, the same drying and thermal treatment at 623 K was applied. Based on the previously described routes for the preparation of monometallic samples, a consecutive preparation route was developed for the bimetallic samples. First, zinc oxide nanoparticles were encapsulated in the CNT interior following the route for ZnOiwi/O-CNT before the sample was impregnated with copper as described for CuOts/O-CNT. After both impregnations, the washing step was applied. For comparison, Cu and Zn were also introduced into CNT-Py. Due to the large diameter facilitating liquid impregnation, a simultaneous route was employed here: copper and zinc nitrate were dissolved in THF and impregnated under tip sonication. 2.2. Characterization Properties of the CNT supports were studied by nitrogen physisorption and by XPS. BET surface areas and pore volumes were determined with a modified Autosorb 1C instrument (Quantachrome) after outgassing the samples at 523 K in He. XPS spectra were measured with an ultrahigh vacuum setup equipped with a high-resolution Gammadata-Scienta SES 2002 analyzer using monochromatized Al Kα as incident radiation. C 1s, O 1s and N 1s were integrated over Shirley-type backgrounds with the Casa-XPS software, for calculation of atomic compositions, the sensitivity factors provided by this software were used. Elemental analysis was made by atomic absorption spectroscopy. Samples were melted in sodium peroxide and then dissolved in aqueous hydrochloric acid. Afterwards the atomic absorption was measured with a Varian SpectrAA 220 spectrometer. Transmission electron microscopy (TEM) images were obtained on a Hitachi H-7500 instrument operated at an accelerating voltage of 100 kV. The samples were dispersed in ethanol under ultrasonication and placed onto a carbon film supported over a copper grid. The micro-
7
scope was also equipped with an energy-dispersive X-Ray detector. Scanning electron microscopy (SEM) measurements were performed on a Hitachi S-5500 Ultra-High Resolution FE-SEM equipped with a BF/DF Duo-STEM detector. The microscope was operated at an accelerating voltage of 30 kV. The samples were placed on a 3 nm carbon-film, placed on 400 mesh copper grid. X-Ray diffractograms were measured with a Panalytical MPD theta-theta diffractometer using CuKα radiation (λ = 1.54056 Å). The signals were recorded over an angular range of 2 Θ = 10-70° (step size - 0.035°, dwell - 25 s/step). The reduction properties were investigated by temperature-programmed reduction (TPR) in 5% H2 / He, ramping the temperature to 513 K at 2 K/min and monitoring the hydrogen concentration with a thermal conductivity based detector (Rosemount, Hydros 100). For most samples, the specific copper surface area was measured subsequently by N2O reactive frontal chromatography as described in [30, 31]. In-situ XAS spectra (Cu K edge, 8.979 keV; Zn K edge, 9.659 keV) were measured in transmission mode at Hasylab, Hamburg (Station C). A Si(111) double-crystal monochromator was used, detuned to 50% of the maximum intensity to exclude higher harmonics from the XRay beam. The hand-pressed sample pellets were reduced in an in-situ EXAFS cell [32] in a flow of 5 % H2/He at a temperature ramp of 2 K/min and a maximum temperature of 513 K for 1 h (H2_513K). XAFS spectra were measured at liquid nitrogen temperature (LNT), and the spectrum of a Cu (or Zn) foil was recorded at the same time for energy calibration. Some samples were then re-heated in He to 513 K and subsequently reduced at 873 K either in dilute hydrogen or in dilute CO (10 % in He). Finally, the XAFS spectra were measured at liquid nitrogen temperature again. The data was analyzed using the VIPER software package [33]. For background subtraction, a Victoreen polynomial was fitted to the pre-edge region. The smooth atomic background, µ 0, was estimated using a smoothing cubic spline. The Fourier analysis of the k²-weighted experimental function was performed with a Kaiser Window.
8
From the CuK EXAFS, the Cu(0)-Cu path was analyzed on the basis of a FEFF8.20 calculation for cubic copper metal [34] in order to determine structural parameters, in particular the coordination number (C. N.), which permits estimating the primary particle sizes based on a spherical particle model [35]. Thermogravimetric analyses (see Fig. S1) were performed in a CAHN TG-2131 microbalance (Raczek Analysentechnik). The sample was initially heated to 323 K with 2 K/min. After 2 h at 323 K, the temperature was increased to 623 K with 5 K/min. All steps were done in flowing helium (113 ml/min).
3. Results 3.1. Carbon nanotubes Relevant physicochemical data of the CNTs employed are collected in Table 1. The increased BET surface areas and pore volumes after oxygen functionalization via gas-phase or HNO3 routes show that the CNT interior is accessible after both procedures. The final N functionalization reproducibly increases both pore volume and BET surface area of the O-CNTs employed. The short CNTs exhibit larger surface areas and pore volumes than the original ones. While the surface areas obtained after liquid and gas-phase functionalization are similar, the pore volume of CNT-A is clearly larger than that of O-CNT. The difference of ca. 20 % is too large to arise exclusively from the growth catalyst remaining in O-CNT. A contribution may come from the less selective action of oxygen, which may destroy or crack tubes leaving some non-porous debris behind, which still contributes to the BET surface area. The surface compositions derived by XPS shows higher oxygen contents for the liquid functionalization – ca. 14 at-% in CNT-A (Table 1). The smaller surface oxygen content achieved with the gas-phase method is due to the volatile character of the most likely reaction products CO and CO2. N functionalization results in significant loss of oxygen. The amount of N intro-
9
duced is small but reproducible. The short CNTs contain already a significant amount of oxygen, which was only marginally increased by the gas-phase functionalization employed. 3.2. Copper into CNTs - Impregnation Fig. 1 shows images taken from the CuOts/O-CNT sample. By switching from scanning (Fig 1a) to scanning transmission mode (Fig. 1b) the presence of particles in the CNT interior can be demonstrated. Apparently they have entered deeply into the channels and are not concentrated at opened CNT tips. The rough morphology of the CNT surface is caused by the severe functionalization. Fig. 2 shows the XRD patterns of the functionalized CNTs and the copper containing sample (9.1 wt-% Cu according to AAS analysis). Two strong reflections around 26.1° and 43.0° arise from the CNTs, two weak reflections at 37.31° and 63.08° can be ascribed to magnetite (ICSD #01-075-0449) formed from the residual growth catalyst. The XRD pattern of the Cu containing sample clearly shows diffraction lines arising from the monoclinic copper (II) oxide phase (tenorite, ICSD #01-089-2529, strongest reflections at 35.58° and 38.74°). Estimation by the Scherrer equation resulted in an average CuO crystallite size of 11 nm, somewhat in excess of the typical inner diameter, which was estimated to be 6-7 nm from the TEM images shown below. In Fig. 3, the XAS spectra (XANES and EXAFS) of the initial CuOts/OCNT are shown. One can clearly see that copper was exclusively present as Cu (II) oxide. The shapes of the XANES as well as of its first derivative and even the EXAFS spectrum are close to those of the CuO reference. In TPR, the Cu(II) oxide was reduced to metallic copper, which caused an asymmetric signal peaking at 449 K (Fig. 4). The asymmetry of the peak, with a small shoulder at low temperatures, might indicate the presence of very small particles, and hence, a bimodal particle size distribution. Under the conditions of our TPR experiment, bulk CuO physically mixed with SiO2 at comparable concentration was reduced in a single peak at 503 K. In Fig. 5, TEM im-
10
ages of an unreduced and a reduced sample are compared. In the reduced sample, particles can be seen in the CNTs as well which were identified as Cu particles by EDX. Some of them seem to fill out the whole inner diameter like a plug. The particles, did, however, not grow along the x-axis of the CNT channels, and hence, the extent of sintering remained moderate. The diffractogram obtained after this reduction is depicted in Fig. 2. Together with signals of magnetite, a reflection of Cu metal at 50.5 ° (the one at 43.4 ° is obscured by a carbon signal) can be clearly seen. The particle size estimated from the line width was 17.5 nm. The EXAFS spectrum of the sample reduced at 473 K in hydrogen is shown in Figure 3 b. It resembles the spectrum of the metal foil. The first shell is weaker than in the spectrum of the reference foil, which indicates a small particle size. From the model fit, a Cu-Cu C. N. of 9.8 was obtained (Cu-Cu distance - 2.54 Å), which results in an estimated average particle size of 2.1 nm. This is smaller than the size seen in the TEM images and far below the size of the Cu particles obtained from XRD. On the other hand, the higher scattering events are all intense up to 7 Å (uncorrected) in the spectrum of the reduced sample (Fig. 3b). Although their intensity is weaker than in the Cu foil, there is hardly any amplitude decay along the distance, which would be expected for particles of 2 nm size. Therefore, the spectrum indicates the presence of large particles coexisting with very small ones. One may suggest from this that the intra-CNT particles may be polycrystalline and made up from very small primary crystallites. The large particles are probably in extra-CNT aggregates. From the N2O RFC measured after TPR, an exposed specific copper surface area of 14.3 m²/g Cu was obtained. On a purely geometric basis, this corresponds to a particle size of 47 nm, far beyond the size of Cu particles seen in the TEM and assessed by EXAFS and even by XRD. Apparently, a large part of the Cu surface is inaccessible. This may be due to contact with the CNT walls and with other copper particles in polycrystalline aggregates.
11
N-functionalization of CNTs has been recently described to favor intra-CNT deposition of particles [36]. Application of the standard tip sonication method to N-CNTs proved successful, visual inspection of the TEM images did, however, not show improvements that would justify the additional step from oxygen to nitrogen functionalization [29] (see also Fig. S2). As mentioned in the introduction, impregnation of Cu species into CNTs functionalized by treatment with nitric acid (CNT-A) as reported in [25] failed in our hands. A typical TEM image taken from a sample made by our standard route with CNT-A is shown in Fig. 6. Cu oxide particles are deposited in high dispersion, but almost exclusively on the external surface. While in some other frames there are a few particles possibly in endohedral locations, it is clear that the vast majority of copper is outside the CNTs, in obvious contrast to the result obtained with the gas-phase functionalization. Several routes described in literature for the introduction of other elements into CNTs functionalized with HNO3 (e.g. aqueous impregnation, urea deposition precipitation, incipient wetness impregnation with both water and ethanol solvents [3, 5, 10, 37]) failed when modified for the use with copper nitrate [29]. As an example, Figure S3 shows TEM images of samples prepared along the route of Wang et al. for the deposition of Fe oxide into CNTs [38] using aqueous nitrate solution and a standard ultrasonic bath (40 W). In Figure S3a, Fe oxide particles can be seen in the interior space while Cu ended up in large aggregates as seen in Figure S3b. On the other hand, gas-phase functionalization opened the way to CuO nanoparticles encapsulated in CNTs via different routes, e.g. incipient wetness impregnation with different solvents (THF, aqueous ammonia, water) and varying copper precursors (nitrate and citrate). However, from visual inspection of the TEM images, the amount of confined particles was less compared to CuOts/O-CNT in these cases (for details see [29]; as an exception, incipient impregnation with an aqueous copper nitrate solution failed with N-CNTs for unknown reasons). The standard tip sonication route was also successfully applied to the short commercial CNTs (cf.
12
section 2.1.) where we expected a larger particle concentration due to the smaller aspect ratio. While the experiments were successful, the expectation of larger particle loadings was not
a)
fulfilled [29] (see example in Fig. S4). CuO was inserted into large-pore CNT-Py via the standard tip sonication route without problems. A TEM image of this material is shown in Figure S5. The X-ray diffractogram showed the presence of CuO, but also of hematite and of α-Fe [29]. 3.3. Copper into CNTs – Impregnation combined with washing With narrow CNTs, the average CuO particle sizes determined by XRD show that a signifi-
a)
b)
cant amount of copper should have been deposited on extrahedral locations. It was therefore attempted to further increase the ratio between particles in interior and exterior space by a selective washing step in concentrated nitric acid. During the washing, the CNT interior space was blocked by xylene as proposed in [38] for a different purpose. In Fig. 7, SEM images in back-scattering mode of the sample before and after washing are shown. It is quite obvious that the larger agglomerates (shown as bright spots) were almost completely removed. Due to the washing procedure, the copper content of the sample decreased from 9.1 to 4.8 wt-%. The TPR profile shows a maximum reduction temperature in the same range as before (443 K) but also a pronounced shoulder at lower temperatures (426 K), which can be attributed to very small copper particles (Fig. 4). Notably, the exposed Cu surface determined by N2O RFC increased from 14.3 to 47.9 m²/g Cu by the washing step. The latter corresponds to a particle size of 14 nm. While this is still larger than the inner pore diameter, the accessibility of the Cu surface is apparently strongly increased by removal of the big aggregates. Thus, by combining the standard tip sonication procedure with selective washing, materials with Cu particles predominantly within the CNTs can be reproducibly prepared. Despite this success, our results warrant a note of caution with respect to the use of xylene for pore blocking. Extension of the acidic washing step to 2 h could lead to complete loss of cop-
13
per. In electron micrographs, particles within the CNTs were rare after washing. Some nanoparticles found after the washing step are depicted in Figure 8a. They were initially considered as CuO particles because of obvious differences to growth catalyst residues with respect to morphology and contrast. Later, a high-resolution EDX investigation revealed that the observed particles contained iron and nickel, apparently from the growth catalyst. Hence, the sample was investigated in more detail by STEM, and an example is shown in Figure 8b. The images show the presence of CuO as small disordered patches and dots in the range of 1-3 nm, highly dispersed within the complete CNT surface, even within the walls, in which the severe functionalization treatment had apparently created cracks and pores. These results suggest that the xylene had not well blocked the interior of our CNTs or had been gradually displaced by the acid, which is in some conflict with the high endo/exohedral selectivity obtained with this method in preparations shown in [38]. 3.4. Zinc into CNTs Routes for deposition of zinc oxide into CNTs were developed based on the experience made with copper. Therefore, the tip sonication route with zinc nitrate in THF solution (ZnOts/OCNT) was used as a reference point. Fig. 9a shows a TEM image of ZnOts/O-CNT. It can be seen that small nanoparticles were encapsulated in the CNT interior. The particles are located both at the opened CNT tip and deeply in the channel and their dimensions are slightly smaller than the inner diameter although some larger agglomerates were detected as well. This was also confirmed by the X-Ray diffraction pattern, which is shown in Fig. 2. The strong and sharp reflection lines at 31.86°, 35.53°, and 36.36° and a number of additional peaks can be ascribed to the hexagonal zincite structure of ZnO (ICSD #01-073-8765). An average crystallite size of 16.5 nm resulted from the application of the Scherrer equation to the line widths, which is considerably larger than the inner diameter of the CNTs. By testing other preparation techniques described above for the confinement of Cu oxide nanoparticles, incipient wetness impregnation of O-CNTs with zinc citrate in ammonia solution 14
has been found to give even better results regarding the amount and the dimensions of the confined nanoparticles. This was not only observed in TEM images (see example in Fig. 9b) where no larger agglomerates could be detected, but also in the X-Ray diffractogram (Fig. 2). The reflections were significantly broadened, and a crystallite size 6.5-7.0 nm was estimated applying the Scherrer equation to the reflections at 31.86° and 56.42°, respectively. This was the first example where the crystallite size was in the range of the inner CNT channel diameter. On the other hand, application of the same incipient wetness impregnation route with zinc citrate to CNT-A failed completely: ZnO particles were deposited only on the external surfaces (Fig. S6). The reducibility of the confined zinc oxide species was investigated because it had been found that the reduction of confined CuO nanoparticles occurred at lower temperatures compared to extra-CNT species (Fig. 4). However, reduction of the confined ZnO nanoparticles could not be observed in an in-situ XAS study (Figure 10). Neither the XANES (or its first derivative) nor the Fourier-transformed EXAFS spectra showed any indication for the presence of zerovalent zinc species even after reduction in CO at 873 K. The severe conditions of the latter experiment rather caused a better ordering of the ZnO aggregates as indicated by a slight increase of the Zn-Zn shell. However, the distances for the Zn-X (X = O and Zn) shells were found to remain unchanged after reduction. The presence of totally oxidized zinc was confirmed by an X-ray diffractogram of the sample used for the EXAFS investigations (Fig. S7). In summary, zinc oxide nanoparticles were successfully and with high dispersion deposited in the CNT interior. It should be noted that the best preparation routes turned out to be different for copper and zinc. In particular, the temperature tolerated for the decomposition of the copper precursor (523 K) is lower than that required for the decomposition of the zinc precursor (623 K). Therefore, the bimetallic sample was prepared via a consecutive route.
15
3.5. Copper and Zinc into CNTs For the deposition of the Cu/ZnO catalyst system into CNTs, zinc oxide nanoparticles were encapsulated in the O-CNTs interior first by following the route described for ZnOiwi/O-CNT, including precursor decomposition at 623 K and washing in HNO3. Subsequently, copper was added using the procedure for CuOts/O-CNT, including precursor decomposition now at 523 K, and a second washing step with HNO3. In TEM images taken in this state of the material, no particles or agglomerates could be observed at all (Figure 11a). Likewise, the XRD pattern (Fig. 2) did not show any indication for copper or zinc oxide crystalline phases although both elements were detected in the sample by AAS (Cu – 3.8 wt-%, Zn – 1.6 wt-%) and, though with very low intensity, also by XPS [29]. The EXAFS spectra of this sample, which are displayed in Figures 3 (CuK) and 10 (ZnK), contain only the first coordination sphere (oxygen) around Cu and Zn, no higher spheres are visible. Apparently, the preparation, though successive, had resulted in a highly disordered oxide material inside the CNTs. In TPR, the reduction peak occurred at 487 K, significantly higher than in the monometallic samples (Fig. 4), and TEM images taken after this reduction step showed metal particles in the CNT interior (Fig. 11b). For this sample, no XRD was taken after reduction at 513 K due to limited availability because even more severe reduction conditions (10% CO/H2, 673 K, 30 min) did not result in any signals assignable to a Cu-containing phase (Fig. 2). The EXAFS spectra after reduction in dilute hydrogen at 513 K are presented in Figures 3 (CuK) and 10 (ZnK). No changes caused by the reduction can be found in the ZnK spectrum. The CuK spectrum again resembles that of the Cu foil, but with very low intensity of the first shell and almost vanishing signals around 7 Å. From the C. N. of the first Cu-Cu shell (6.9), an average particle size of below 1 nm was estimated. The observation of some intensity around 7 Å indicates the presence of larger particles here as well. In sharp contrast to these data, the ex-
16
posed specific copper surface area as determined by N2O RFC was just 5.3 m²/g Cu, which would correspond to particle sizes of >100 nm. Obviously, some of the problems observed with Cu/O-CNT occurred in this sample as well: The contradictions between the primary particle sizes (EXAFS) and the particles observed in the TEM suggest that the latter may be aggregated from very small crystallites. The contradiction between the sizes just mentioned and the size derived from N2O RFC indicates that most of the Cu surface is covered – by ZnO, by CNT walls, or cut off by pore plugging. The very small primary particle size of the Cu crystallites (<1 nm, as compared to >2 nm in the monometallic sample) suggests an interaction between Cu and Zn species during preparation favoring high dispersion of primary particles. The same trend was obtained also in preparations with silica-based mesoporous host matrices where the bimetallic samples always exhibited a finer Cu dispersion than the monometallic ones [29, 39]. As mentioned in section 2.1., Cu and Zn were introduced into CNT-Py simultaneously, which did not cause any problems. A TEM image of the sample is shown in Fig. S8.
4. Discussion In the results presented above, we have shown that CuO and ZnO nanoparticles can be reproducibly deposited into CNTs of 6-7 nm i.d. when these have been previously functionalized by a gas-phase route that involves thermal stress. The selectivity for the inner space is significant, but needs to be improved by a washing step in HNO3 during which the CNT interior should be blocked, e.g. with xylene. After reduction of the oxide clusters, the copper remains in the CNTs. Its accessibility from the gas-phase was demonstrated by interaction with N2O, where, however, a relatively small accessible surface area was obtained although XAFS indicates primary particle sizes on the order of only 2 nm. This contradiction suggests that much of the Cu surface area is blocked by contact with other Cu particles, with CNT walls, or by occasional plugging of whole CNT cross sections by (polycrystalline) Cu particles.
17
Sequential introduction of ZnO and CuO with subsequent reduction resulted in a CuZnO/CNT sample, where the accessible Cu surface area was even smaller than in the monometallic sample despite significantly smaller primary particle size (XAFS). This may be ascribed to an enhanced degree of surface blocking, e.g. by interaction with ZnO. Routes for introduction of Cu and Zn oxide into CNTs previously functionalized by HNO3, as earlier described in literature [25], failed in our study for ZnO, and in particular for CuO. Particles were predominantly or exclusively deposited onto the external CNT surfaces when the procedures best for O-CNTs (i.e., after gas-phase functionalization) were applied to CNT-A (Fig. 6, S6). The BET surface areas and pore volumes (Table 1) show that our CNTs were open after HNO3 treatment, and the successful deposition of Fe oxide particles (Fig. S3) suggests that the interior of CNT-A could be wetted. A possible reason for this failure has been recently indicated by Machado et al. [40] who found that concentrated HNO3 functionalized the exterior surface of CNTs of comparable inner diameters but left the interior surfaces unchanged. Thus, introduction of Cu and Zn oxo species, which apparently requires strong interaction with the surface, was unsuccessful while deposition of Fe oxide, which probably occurs by precipitation rather than by ion exchange, was possible. Still, this does not explain the contradiction of our observations with those in [25] where Cu oxide particles were introduced in CNTs of similar inner diameter after HNO3 functionalization, with comparable magnitudes of accessible Cu surface areas being achieved. It has been shown recently that the result of functionalization with HNO3 depends critical on properties of the CNT surfaces, e.g. on graphitization degree and defect content [41]. Thus, differences in the properties of CNTs employed may have caused the contradictions mentioned. The apparent failure of our attempt to block the CNT interior by xylene may be another indirect indication for an oxygen functionalization of the interior surfaces. Excellent exoendohedral selectivity has been reported when the CNT interior was blocked with xylene
18
while Ru was deposited on the external surface from an aqueous solution [38]. If the interior of our CNTs were hydrophobic there would be no driving force for the xylene to give the acid access to the interior space, which apparently has happened during our washing procedure: after long extension, all Cu was lost, after our standard duration, the CuO particles were dissolved. The copper was, however, not removed from the interior space. Growth catalyst residuals were apparently also dissolved and formed a few particles during drying while the Cu and (the Zn where present) was recovered in highly disordered layers (Figures 8 and 11, EXAFS spectra of initial CuZnO/O-CNT in Figures 3 and 10b). Activity data measured with the Cu-ZnO/CNT samples mentioned in this paper will be reported elsewhere together with related characterization data showing the response of the samples to relevant pretreatments (for basic information see [29]).
5. Conclusions Copper and zinc oxide nanoparticles were introduced into carbon nanotubes of 3-5 nm nominal interior diameter (actually 6-7 nm), and the CuO particles were reduced to obtain intraCNT Cu nanoparticles. The preparation routes were optimized by systematic variation of the preparation conditions leading to a new standard synthesis route for each oxide. CNTs were previously functionalized by thermal stress in air because attempts to deposit Cu oxide particles into CNTs functionalized by the conventional HNO3 route failed. Wetness impregnation of functionalized CNTs with copper nitrate in THF solution under tip sonication was the optimum route leading to copper containing CNTs with a large amount of confined particles. After reduction, a bimodal particle size distribution was concluded from characterization data (e.g. very low specific copper surface area). Extra-CNT aggregates could be largely removed by a selective washing with HNO3 after blocking the CNT interior with xylene, which increased the accessible Cu surface dramatically. There were, however, indications that the xylene was gradually replaced by acid in the CNT interior during the washing step.
19
For ZnO in CNTs, incipient impregnation with zinc citrate in ammonia solution was the best preparation route. Already before the washing step a sample with an estimated ZnO crystallite size in the range of the inner CNT diameter was obtained, where extra-CNT particles could not be detected. Based on these results, a bimetallic sample containing Cu and ZnO in CNTs was prepared via a consecutive route including two washing steps. While Cu and Zn oxide species formed very small and amorphous aggregates in the as-prepared material, which could not be observed by TEM and XRD, the aggregates (Cu nanoparticles) encapsulated in the CNTs became visible for TEM after reduction.
Acknowledgements We kindly thank Mr. H.-J. Bongard for the SEM/STEM analysis, Dr. Th. Reinecke for XRD measurements, Mrs. S. Plischke for TG and Mrs. N. Arshadin for the TPR and N2O-RFC experiments. The work has been funded by the German Science Foundation (DFG) in the framework of the Collaborative Research Center “Metal-Substrate Interactions in Heterogeneous Catalysis” (SFB 558), which is gratefully acknowledged.
Appendix A. Supplementary Data Supplementary data associated with this article (Figures S1 through S8) can be found in the online version, at …
20
References [1] S. Iijima, Nature 354 (1991) 56-58. [2] W. Chen, X.L. Pan, X.H. Bao, J. Am. Chem. Soc. 127 (2008) 7421-7426. [3] R.M.M. Abbaslou, A. Tavassoli, J. Soltan, A.K. Dalai, Appl. Catal. A 367 (2009) 4752. [4] X.L. Pan, Z.L. Fan, W. Chen, Y.L. Ding, H.Y. Luo, X.H. Bao, Nature Materials 6 (2007) 507-511. [5] X. Pan, X. Bao, Chem. Commun. (2008) 6271-6281. [6] X. Blase, L.X. Benedict, E.L. Shirley, S.G. Louie, Phys. Rev. Lett. 72 (1994) 18781881. [7] S.C. Tsang, Y.K. Chen, P.J.F. Harris, M.L.H. Green, Nature 372 (1994) 159-162. [8] T. Trang Nguyen, P. Serp, ChemCatChem 5 (2013) 3595-3603. [9] Q. Fu, G. Weinberg, D.S. Su, New Carbon Materials 23 (2008) 17-20. [10] J.-P. Tessonnier, O. Ersen, G. Weinberg, C. Pham-Huu, D.S. Su, R. Schlögl, ACS Nano 3 (2009) 2081-2089. [11] E. Castillejos, P.-J. Debouttière, L. Roiban, A. Solhy, V. Martinez, Y. Kihn, O. Ersen, K. Philippot, B. Chaudret, P. Serp, Angew Chem. Int. Ed. 48 (2009) 2529-2533. [12] L. Jiang, L. Gao, Carbon 41 (2003) 2923-2929. [13] K. Schulte, J.C. Swarbrick, N.A. Smith, F. Bondino, E. Magnano, A.N. Khlobystov, Adv. Mater. 19 (2007) 3312-3316. [14] W.A. Solomonsz, G.A. Rance, M. Suyetin, A. La Torre, E. Bichoutskaia, A.N. Khlobystov, Chem. Eur. J. 18 (2012) 13180-13187. [15] W. Baaziz, X. Liu, I. Florea, S. Begin-Colin, B.P. Pichon, C. Ulhaq, O. Ersen, M. SoriaSanchez, S. Zafeiratos, I. Janowska, D. Begin, C. Pham-Huu, J.Mater. Chem. A 1 (2013) 13853-13861. [16] J. Zhang, J.-O. Müller, W. Zheng, D. Wang, D. S. Su, R. Schlögl, Nano Letters 8 (2008) 2738-2743. [17] E. Castillejos, R. Chico, R. Bacsa, S. Coco, P. Espinet, M. Pérez-Cadenas, A. GuerreroRuiz, I. Rodríguez-Ramos, P. Serp, Eur. J. Inorg. Chem. (2010) 5096-5102. [18] D. Ugarte, A. Chatelain, W.A. deHeer, Science 274 (1996) 1897-1899. [19] C. Wang, S. Guo, X. Pan, W. Chen, X. Bao, J.Mater. Chem. 18 (2008) 5782-5786. [20] W. Chen, Z. Fan, X. Pan, X. Bao, J.Am. Chem. Soc. 130 (2008) 9414-9419. [21] H. Friedrich, S. Guo, P.E. de Jongh, X. Pan, X. Bao, K.P. de Jong, ChemSusChem 4 (2011) 957-963. [22] H.L. Zhuang, G.P. Zheng, A.K. Soh, Comp. Mater. Sci. 43 (2008) 823-828. 21
[23] K. Nakada, A. Ishii, Solid State Commun.151 (2011) 13-16. [24] P. Munnik, M. Wolters, A. Gabrielsson, S.D. Pollington, G. Headdock, J.H. Bitter, P.E. de Jongh, K.P. de Jong, J.Phys. Chem. C 115 (2011) 14698-14706. [25] D. Wang, G. Yang, Q. Ma, M. Wu, Y. Tan, Y. Yoneyama, N. Tsubaki, ACS Catal. 2 (2012) 1958-1966. [26] C.S. Chen, T.G. Liu, L.W. Lin, X.D. Xie, X.H. Chen, Q.C. Liu, B. Liang, W.W. Yu, C.Y. Qiu, J.Nanoparticle Res. C7 - 1295 15 (2012) 1-9. [27] M.Q. Tran, C. Tridech, A. Alfrey, A. Bismarck, M.S.P. Shaffer, Carbon 45 (2007) 2341-2350. [28] S. Kundu, W. Xia, W. Busser, M. Becker, D.A. Schmidt, M. Havenith, M. Muhler, Phys. Chem. Chem. Phys. 12 (2010) 4351-4359. [29] D. Grossmann, PhD Thesis, Bochum 2013. [30] O. Hinrichsen, T. Genger, M. Muhler, Chem. Eng. Technol. 11 (2000) 956. [31] G.C. Chinchen, C.M. Hay, H.D. Vandervell, K.C. Waugh, J. Catal. 103 (1987) 79-86. [32] F.W.H. Kampers, T.M.J. Maas, J. van Grondelle, D.C. Brinkgreve, D.C. Koningsberger, Rev. Sci. Instr. 60 (1989) 2635-2638. [33] K.V. Klementiev. VIPER software for Windows (Visual Processing in EXAFS Researches), freeware www.cells.es/Beamlines/CLAESS/software/viper.html [34] J.J. Rehr, R.C. Albers, Rev. Mod. Phys. 72 (2000) 621-654. [35] M. Borovski, J. Phys. IV 7 (1997) C2-259-260. [36] H.J. Schulte, PhD Thesis, Bochum 2011. [37] V.K. Kaushik, C. Sivaraj, P.K. Rao, Appl. Surf. Sci. 51 (1991) 27-33. [38] C.F. Wang, S.J. Guo, X.L. Pan, W. Chen, X.H. Bao, J. Mater. Chem. 18 (2008) 57825786. [39] M.W.E. van den Berg, S. Polarz, O.P. Tkachenko, K.V. Klementiev, M. Bandyopadhyay, L. Khodeir, H. Gies, M. Muhler, W. Grünert, J. Catal. 241 (2006) 446-455. [40] B.F. Machado, M. Oubenali, M. Rosa Axet, T. Trang Nguyen, M. Tunckol, M. Girleanu, O. Ersen, I.C. Gerber, P. Serp, J. Catal. 309 (2014) 185-198. [41] J.-P. Tessonnier, D. Rosenthal, F. Girgsdies, J. Amadou, D. Begin, C. Pham-Huu, D. S. Su, R. Schlögl, Chem. Commun. (2009) 7158-7160.
22
Table 1 – Textural properties and surface compositions of CNTs employed for encapsulation of Cu and ZnO nanoparticles. Sample
Surface area, m2/g
Pore volume, ml/g
raw CNT
78
0.14
CNT-A
188
O-CNT
Composition of near-surface layer, by XPS C, at-%
O, at-%
N at-%
97.7
2.3
-
0.45
85.8
14.2
-
184
0.35
91.5
8.5
-
N-CNT
214
0.37
96.3
3.7
0.3
raw sh-CNT
224
0.51
95.9
4.1
-
O-sh-CNT
279
0.53
94.6
6.2
-
N-sh-CNT
288
0.58
95.5
4.5
0.3
23
Legends Figure 1 – Electron micrograph of the calcined copper-containing sample CuOts/O-CNT, simultaneously recorded in (a) SEM mode and in (b) bright-field STEM mode. Figure 2 – XRD patterns of functionalized O-CNTS (black), calcined copper and zinc oxide containing samples (red) and of the bimetallic sample after severe reduction in CO/H2 (blue). Figure 3 – CuK XAS spectra of Cu species and of Cu-ZnO enclosed in CNTs after different treatments, compared with references; a) XANES and first derivatives, b) Fouriertransformed CuK spectra (k2-weighted, absolute values) at liquid nitrogen temperature. Figure 4 – TPR profiles of different Cu oxide containing samples, compared with CuO mixed with SiO2. Figure 5 – TEM images of copper-containing sample CuOts/O-CNT a) before and b) after standard reduction (H2_513K). Figure 6 – TEM image of CuOts/CNT-A oxide containing sample, with CuO species predominantly deposited on the external surface. Figure 7 – SEM images of sample CuOts/O-CNT recorded with backscattered electrons (BSD mode), a) before and b) after selective washing with concentrated nitric acid. Figure 8 – Electron micrographs of Cu oxide containing sample CuOts/O-CNT after selective washing with concentrated nitric acid, in transmission (a) and STEM mode (b). The particles in (a) were identified as originating from the growth catalyst. In (b), the black patches and dots were identified as Cu compounds by EDX. EDX results at different analysis points (marked with circles): 1) 0 wt% Fe, 0 wt% Ni, 100 wt% Cu; 2) 3.9 wt% Fe, 6.7 wt% Ni, 89.4 wt% Cu; 3) 0 wt% Fe, 0.6 wt% Ni, 99.4 wt% Cu; 4) 3.1 wt% Fe, 12.5 wt% Ni, 84.4 wt% Cu; 5) 0 wt% Fe, 8.2 wt% Ni, 91.8 wt% Cu. Figure 9 – TEM images of Zn oxide containing samples, a) ZnOts/O-CNT and b) ZnOiwi/OCNT. 24
Figure 10 – ZnK XAS spectra of initial (red) monometallic zinc and bimetallic sample, after standard reduction H2_513K (green), and after severe reduction H2 or CO at 873K (blue): a) XANES and first derivatives, b) Fourier-transformed ZnK spectra (k2-weighted, absolute values) measured at liquid nitrogen temperature. Figure 11 – TEM images of bimetallic sample CuZnO/O-CNT a) before and b) after standard reduction.
25
Figure 1 26
Figure 2
27
Figure 3
28
Figure 4
29
a)
Figure 5
30
Figure 6
31
a)
b)
Figure 7
32
a) b)
Figure 8
33
a)
b)
Figure 9
34
Figure 10
35
a)
b)
Figure 11
36
Research Highlights • • • • •
CuO and ZnO deposited in CNTs of 6-7 nm i.d. by ultrasound-assisted impregnation CNT to be functionalized by thermal shock in air rather than by HNO3 Selectivity for intra-CNT location improved by selective washing procedure Reduction of intra-CNT CuO gives polycrystalline Cu NPs with limited accessibility Cu/ZnO aggregates intended to mimic the methanol synthesis catalyst prepared in CNTs
37
Graphical abstract
38