- Email: [email protected]
Plasmon imaging: An efficient TEM-based method for locating noble metal particles dispersed on oxide catalysts at very low densities
Micron 39 (2008) 344–347 www.elsevier.com/locate/micron
Short communication
Plasmon imaging: An efficient TEM-based method for locating noble metal particles dispersed on oxide catalysts at very low densities D.R.G. Mitchell a,*, Xingdong Wang b, Rachel A. Caruso b a
Institute of Materials and Engineering Science, ANSTO, PMB 1, Menai, NSW 2234, Australia b PFPC, School of Chemistry, The University of Melbourne, Melbourne Vic. 3010, Australia Received 13 February 2007; received in revised form 5 March 2007; accepted 5 March 2007
Abstract We have used transmission electron microscopy to study catalysts comprising nanoparticulate gold dispersed on a highly porous nanoparticulate TiO2 (anatase) support. The similarity of the morphology of the two phases, and the low number density of gold particles (1 in 65,000) makes this challenging. Diffraction contrast imaging could not differentiate the two phases, since TiO2 oriented at strong Bragg conditions, produced similar contrast to the Bragg/mass-thickness contrast of the gold. Mass-thickness contrast imaging allowed gold to be differentiated from TiO2 only in the thinnest regions, where the mass-thickness of TiO2 was low. Plasmon imaging, using an energy loss of 24 eV and an energy window width of 5 eV, was very effective at locating gold. Both the TiO2 and impregnating resin produced a strong plasmon signal, while the much weaker signal from the gold made it appear dark. This permitted the gold particles to be readily located, irrespective of whether they were located in the thin or thick regions of the TiO2 support. Crown Copyright # 2007 Published by Elsevier Ltd. All rights reserved. Keywords: TiO2; Gold; Catalyst; TEM; Plasmon imaging
1. Introduction Many heterogeneous catalyst systems are based on nanometre scale dispersions of noble metals, such as Au, Pt, Pd, etc., on metal oxide supports, such as TiO2 and Al2O3. Detailed microstructural understanding of how the noble metal is distributed, its form and its interaction with the support is a critical part of developing highly efficient catalyst systems. The nanoscale of these metal and support systems means that transmission electron microscopy is an indispensable tool in their study. Many noble metals exhibit greatly enhanced catalytic activity when dispersed on oxides such as TiO2, due to strong metal-support interactions (Hadjiivanov and Klissurski, 1996). Catalysts comprising gold dispersed on TiO2 have been shown to have excellent photocatalytic activity for the removal of organic dyes from waste water (Kamat, 2002; Li and Li, 2002; Arabatzis et al., 2003; Li et al., 2005; Sonawane and Dongare, 2006). The high cost of noble metals means that metal loadings on such catalysts is generally extremely low
* Corresponding author. Tel.: +61 2 9717 3456; fax: +61 2 9543 7179. E-mail address: [email protected] (D.R.G. Mitchell).
(<1 wt.%). Therefore, locating noble metal particles, which may be of similar shape and size to particles of the oxide support, can be very challenging, especially when only perhaps 1 in 105 particles is the noble metal of interest. In this brief note we describe the use of plasmon imaging as a tool to aid particle location. Plasmon imaging utilises the low energy loss signal on an energy filtered TEM. This signal is sufficiently intense to enable live imaging. Therefore, very large areas of material can be rapidly surveyed to locate the particles of interest. 2. Experimental The catalysts examined consisted of a dispersion of various amounts of Au (0.05–0.15 wt.%) on a TiO2 (anatase) support. In this study, the gold nanoparticles were incorporated into an agarose gel using three different synthesis routes, as described elsewhere (Wang et al., 2007). Titanium dioxide was then coated onto the agarose scaffold before the samples were heated to 450 8C to remove the agarose and crystallise the TiO2 to the anatase phase (Sonawane and Dongare, 2006). In order to preserve the open porous structure for TEM examination, the material was impregnated with an LR-white resin. After curing,
0968-4328/$ – see front matter. Crown Copyright # 2007 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2007.03.002
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
it was ultramicrotomed into sections approximately 90 nm thick. Specimens were examined using a JEOL 2010F transmission electron microscope (TEM) operating at 200 kV. This was fitted with a Gatan Imaging Filter (GIF) and an Oxford Instruments ISIS energy dispersive X-ray spectroscopy (EDS) system, for qualitative analysis. Catalyst powders were examined using X-ray diffraction (XRD) and inductively coupled plasma mass spectrometry (ICPMS). 3. Results and discussion XRD measurements indicated that the average sizes of the anatase crystals were in the range 15–18 nm, and TEM examination (see later) supported this. Depending on the preparation route, gold particles were found to vary in diameter between 10 and 40 nm. Assuming a typical anatase particle size of 20 nm, a gold particle size range of 10–40 nm (TEM measurement), and a gold loading of 0.03–0.15 wt.% (ICPMS measurement) the range of possible Au:TiO2 ratios is 1 Au particle in 1000–65,000 TiO2 particles. The presence of the agarose template created a highly open and porous Au/TiO2 structure. The surface area was 50 5 m2 g 1 for all the materials. Fig. 1 shows a typical diffraction contrast image of the highly open structure, with pore diameters of several hundred nanometres. The networks are comprised of agglomerated polyhedral anatase particles with a mean diameter of 20 nm. The pile-up of features in projection which occurs in the structure (Fig. 1 top right) leads to very dark regions in this imaging mode. Gold particles (identity confirmed with EDS), such as that labelled in Fig. 1, are randomly distributed in the material. Any gold particles occurring in regions of high TiO2 particle density would not be apparent in this imaging mode.
Fig. 1. Bright field diffraction contrast image of a typical 10 nm gold particle sitting within the titania network. Typical pore diameters are around 100 nm. The network structures are composed of agglomerated particles of anatase 10– 20 nm in diameter.
345
Spectroscopic methods, such as x-ray energy dispersive spectroscopy (EDS) and electron energy loss spectroscopy (EELS) can readily differentiate noble metal from TiO2. However, the extremely low particle loadings and the time required to focus a probe on the particle and initiate an analysis, mean that this method is impractical for identifying particles on the basis of a point by point analysis in the first instance. Similarly, techniques such as energy filtered TEM (EFTEM) can be used to map elemental distributions. However, this is not a live technique, requiring up to several minutes per frame, and is far too slow to be used for surveying large areas of specimen. However, EFTEM is useful for high contrast mapping of the gold distribution, once the particles have been located. High angle annular dark field (HAADF) microscopy in scanning transmission electron microscopy (STEM) mode has been shown to be particularly effective in studying these types of materials (Pennycook, 1981; Akita et al., 2001). In HAADF, the incoherent scattering of electrons though high angles shows an approximate dependence on the square of the atomic number. It therefore provides excellent sensitivity for heavy noble metal dispersions on much lighter metal oxides and provides sufficient intensity for live imaging. However, the use of STEM mode is not without its drawbacks. On our instrument, lens reconfiguration on switching from TEM to STEM mode involves a 1 h downtime while the instrument stabilizes. To aid more rapid throughput of materials, a technique operating solely in TEM mode was desirable, since other TEM-based methods, such as high resolution TEM and EFTEM imaging, were also to be applied to these materials. The results of these characterisation studies will be reported elsewhere (Wang et al., 2007). The method selected for Au particle location was plasmon imaging. Plasmons are low energy loss events (<50 eV) which occur due to longitudinal wave-like oscillations of weakly bound electrons. Most materials produce quite broad (20 eV) plasmon peaks, and this low loss region of the energy loss spectrum is sensitive to specimen thickness. With increasing thickness, plural scattering results in multiples of plasmon peaks. In most instances plasmon imaging in achieved by inserting a narrow energy selecting slit (1–5 eV) and qualitatively imaging at various energy losses. This is an effective way of quickly differentiating phases with reasonably well resolved plasmon peaks (Sigle et al., 2003; Mitchell et al., 2005). Most importantly, the plasmon signal is sufficiently intense to permit live imaging (Sasaki et al., 2001). Fig. 2 shows a series of images of the same region captured using various imaging modes. In Fig. 2a, which is a diffraction contrast image (objective aperture inserted), contrast between particles is predominantly the result of orientation differences. Those particles close to strong Bragg diffraction conditions produce intense diffracted beams. The objective aperture below the specimen blocks these diffracted beams, causing the particles to appear dark. Random orientation amongst crystalline particles then produces a range of intensities. Electron dense, crystalline materials like Au, scatter electrons strongly, and produce both strong Bragg diffraction contrast (when suitably oriented) as well as strong mass-thickness scattering
346
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
Fig. 2. The effect of various imaging modes on the contrast between gold and TiO2: (a) Diffraction contrast image showing that gold and TiO2 close to Bragg conditions are indistinguishable; (b) mass-thickness contrast image of (a) showing that gold is differentiated from TiO2 at strong Bragg conditions. However, thick TiO2 regions appear dark in this mode, and gold in such regions cannot be observed; (c) 24 eV (5 eV slit width) plasmon image of (a). Gold appears dark while TiO2 and resin are bright. Thick regions of TiO2 are brightest, revealing any gold particles within them.
through high angles. In this material, where gold particles (labelled in Fig. 2a) are of similar size and shape to the TiO2 particles, diffraction contrast imaging makes it very difficult to locate the gold directly from cursory inspection, due to the similarity of contrast from the two phases. Mass-thickness contrast imaging was evaluated (Fig. 2b). Here, the objective aperture is removed, and so the contrast reflects the net mass-thickness of the specimen. Whilst the gold particle was darker than the surrounding TiO2 particles, regions of specimen where the net thickness of TiO2 was high also produced similar (dark) contrast. The gold was uniformly distributed with the TiO2, and therefore the gold particles were present at the pore surfaces and embedded within the framework of TiO2 particles. Gold particles located in the
pore walls where TiO2 particle density was highest, would not be visible using mass-thickness imaging. Therefore, this imaging mode was inefficient, since it could only locate a small subset of gold particles in the thinnest regions of the specimen. For plasmon contrast imaging (Fig. 2c), we used an energy selecting slit width of 5 eV to create a window centred on an energy loss of 24 eV. TiO2 and the mounting resin (carbonaceous) exhibit a strong plasmon excitation at this energy, but gold far less so. Catalytically important metals such as Au, Pt, Ag and Pd all exhibit relatively weak plasmon peaks over this energy range, and so should be amenable to study with this method. No objective aperture was present, to minimise diffraction contrast effects. In this imaging mode, both the resin
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
and TiO2 were illuminated strongly, while the gold appeared dark. Unlike mass-thickness imaging, the contrast is reversed in plasmon imaging, and thick regions of TiO2 were brighter than thin regions. Therefore, it was possible to quickly scan large areas of specimen and identify the gold particles, even where they were located within thick clusters of TiO2 forming the pore walls, which they invariably were. In the absence of an objective aperture, and in view of the significant thickness of the TEM specimen (90 nm), not all regions of the specimen were at focus. As a result, TiO2 grains close to strong Bragg conditions did produce strong diffraction contrast when they were defocused, due to the separation of bright and dark field images. However, such candidate particles could be quickly eliminated from the search, since the bright field/dark field contrast could be quickly nulled by refocusing. The contrast from gold particles, was not significantly affected by focus. Identification of gold particles via plasmon imaging was confirmed using focused probe EDS. This showed the plasmon imaging method to be extremely effective in locating the gold particles. Having identified a method for locating the gold, large areas of foil could be quickly scanned with a live plasmon image displayed. This enabled gold particles, with number densities as low as 1 in 65,000 to be located in just a few minutes per particle. 4. Summary Diffraction contrast, mass-thickness and plasmon imaging have been compared as methods for locating low number densities of gold nanoparticles dispersed on a porous nanoparticulate TiO2 (anatase) support. The similar shape and morphology of the gold and TiO2 meant that where TiO2 particles were located close to strong Bragg diffraction conditions, diffraction contrast imaging was poor at differentiating the two phases. Mass-thickness imaging highlighted gold particles in the thinnest regions of the specimens. However, in the pore walls the high mass-thickness contrast of the TiO2 made it impossible to locate the majority of the gold particles which were located in these regions. Plasmon imaging, using an energy window width of 5 eV at an energy loss of 24 eV was highly effective, since both the TiO2 and the resin used for TEM specimen impregnation produced a strong plasmon signal, especially in the pore walls, and this contrasted strongly with the weak plasmon emission of the gold. The
347
plasmon signal is sufficiently intense to permit live imaging and rapid surveys of large areas to locate the very low density of particles (as low as 1 in 65,000) are readily achieved. The similar plasmon emission characteristics of many catalytically important noble metals, suggests this method might be applicable to a wide range of catalyst systems. Acknowledgements This work was funded by the Australian Research Council and the Australian Institute of Nuclear Science and Engineering (AINSE). RAC acknowledges the Australian Research Council for an Australian Research Fellowship. Dr Simon Crawford is thanked for TEM specimen preparation. References Akita, T., Tanaka, K., Okuma, K., Koyanagi, T., Haruta, M., 2001. TEM and HAADF-STEM study of a Au catalyst supported on a TiO2 nano-rod. J. Electron Microsc. (Tokyo) 50, 473–477. Arabatzis, I.M., Stergiopoulos, T., Andreeva, D., Kitova, S., Neophytides, S.G., Falaras, P., 2003. Characterization and photocatalytic activity of Au/TiO2 thin films for azo-dye degradation. J. Catal. 220, 127–135. Hadjiivanov, K.I., Klissurski, D.G., 1996. Surface chemistry of titania (anatase) and titania-supported catalysts. Chem. Soc. Rev. 25, 61–69. Kamat, P.V., 2002. Photoinduced transformations in semiconductor-metal nanocomposite assemblies. Pure Appl. Chem. 74, 1693–1706. Li, F.B., Li, X.Z., 2002. Photocatalytic properties of gold/gold ion-modified titanium dioxide for wastewater treatment. Appl. Catal. A: Gen. 228, 15–27. Li, X.Z., He, C., Graham, N., Xiong, Y., 2005. Photoelectrocatalytic degradation of bisphenol A in aqueous solution using a Au:TiO2/ITO film. J. Appl. Electrochem. 35, 741–750. Mitchell, D.R.G., Attard, D.J., Finnie, K.S., Triani, G., Barbe´, C.J., Depagne, C., Bartlett, J.R., 2005. TEM and ellipsometry studies of nanolaminate oxide films prepared using atomic layer deposition. Appl. Surf. Sci. 243, 265–277. Pennycook, S.J., 1981. Study of supported ruthenium catalysts by STEM. J. Microsc. 124, 15–22. Sasaki, K., Tsukimoto, S., Konno, M., Kamino, T., Saka, H., 2001. Dynamic chemical mapping near a Si/SiO2 interface at elevated temperatures using plasmon-loss images. J. Microsc. 203, 12–16. Sigle, W., Kra¨mer, S., Varshney, V., Zern, A., Eigenthaler, U., Ru¨hle, M., 2003. Plasmon energy mapping in energy-filtering transmission electron microscopy. Ultramicroscopy 96, 565–571. Sonawane, R.S., Dongare, M.K., 2006. Sol-gel synthesis of Au/TiO2 thin films for photocatalytic degradation of phenol in sunlight. J. Mol. Catal. A, Chem. 243, 68–76. Wang, X.D., Egan, C., Zhou, J.F., Zhou, M.F., Prince, K., Mitchell, D.R.G., Caruso, R.A., 2007. Preparation of porous Au/TiO2 nanocomposites by templating with agarose gel, in preparation.
Short communication
Plasmon imaging: An efficient TEM-based method for locating noble metal particles dispersed on oxide catalysts at very low densities D.R.G. Mitchell a,*, Xingdong Wang b, Rachel A. Caruso b a
Institute of Materials and Engineering Science, ANSTO, PMB 1, Menai, NSW 2234, Australia b PFPC, School of Chemistry, The University of Melbourne, Melbourne Vic. 3010, Australia Received 13 February 2007; received in revised form 5 March 2007; accepted 5 March 2007
Abstract We have used transmission electron microscopy to study catalysts comprising nanoparticulate gold dispersed on a highly porous nanoparticulate TiO2 (anatase) support. The similarity of the morphology of the two phases, and the low number density of gold particles (1 in 65,000) makes this challenging. Diffraction contrast imaging could not differentiate the two phases, since TiO2 oriented at strong Bragg conditions, produced similar contrast to the Bragg/mass-thickness contrast of the gold. Mass-thickness contrast imaging allowed gold to be differentiated from TiO2 only in the thinnest regions, where the mass-thickness of TiO2 was low. Plasmon imaging, using an energy loss of 24 eV and an energy window width of 5 eV, was very effective at locating gold. Both the TiO2 and impregnating resin produced a strong plasmon signal, while the much weaker signal from the gold made it appear dark. This permitted the gold particles to be readily located, irrespective of whether they were located in the thin or thick regions of the TiO2 support. Crown Copyright # 2007 Published by Elsevier Ltd. All rights reserved. Keywords: TiO2; Gold; Catalyst; TEM; Plasmon imaging
1. Introduction Many heterogeneous catalyst systems are based on nanometre scale dispersions of noble metals, such as Au, Pt, Pd, etc., on metal oxide supports, such as TiO2 and Al2O3. Detailed microstructural understanding of how the noble metal is distributed, its form and its interaction with the support is a critical part of developing highly efficient catalyst systems. The nanoscale of these metal and support systems means that transmission electron microscopy is an indispensable tool in their study. Many noble metals exhibit greatly enhanced catalytic activity when dispersed on oxides such as TiO2, due to strong metal-support interactions (Hadjiivanov and Klissurski, 1996). Catalysts comprising gold dispersed on TiO2 have been shown to have excellent photocatalytic activity for the removal of organic dyes from waste water (Kamat, 2002; Li and Li, 2002; Arabatzis et al., 2003; Li et al., 2005; Sonawane and Dongare, 2006). The high cost of noble metals means that metal loadings on such catalysts is generally extremely low
* Corresponding author. Tel.: +61 2 9717 3456; fax: +61 2 9543 7179. E-mail address: [email protected] (D.R.G. Mitchell).
(<1 wt.%). Therefore, locating noble metal particles, which may be of similar shape and size to particles of the oxide support, can be very challenging, especially when only perhaps 1 in 105 particles is the noble metal of interest. In this brief note we describe the use of plasmon imaging as a tool to aid particle location. Plasmon imaging utilises the low energy loss signal on an energy filtered TEM. This signal is sufficiently intense to enable live imaging. Therefore, very large areas of material can be rapidly surveyed to locate the particles of interest. 2. Experimental The catalysts examined consisted of a dispersion of various amounts of Au (0.05–0.15 wt.%) on a TiO2 (anatase) support. In this study, the gold nanoparticles were incorporated into an agarose gel using three different synthesis routes, as described elsewhere (Wang et al., 2007). Titanium dioxide was then coated onto the agarose scaffold before the samples were heated to 450 8C to remove the agarose and crystallise the TiO2 to the anatase phase (Sonawane and Dongare, 2006). In order to preserve the open porous structure for TEM examination, the material was impregnated with an LR-white resin. After curing,
0968-4328/$ – see front matter. Crown Copyright # 2007 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2007.03.002
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
it was ultramicrotomed into sections approximately 90 nm thick. Specimens were examined using a JEOL 2010F transmission electron microscope (TEM) operating at 200 kV. This was fitted with a Gatan Imaging Filter (GIF) and an Oxford Instruments ISIS energy dispersive X-ray spectroscopy (EDS) system, for qualitative analysis. Catalyst powders were examined using X-ray diffraction (XRD) and inductively coupled plasma mass spectrometry (ICPMS). 3. Results and discussion XRD measurements indicated that the average sizes of the anatase crystals were in the range 15–18 nm, and TEM examination (see later) supported this. Depending on the preparation route, gold particles were found to vary in diameter between 10 and 40 nm. Assuming a typical anatase particle size of 20 nm, a gold particle size range of 10–40 nm (TEM measurement), and a gold loading of 0.03–0.15 wt.% (ICPMS measurement) the range of possible Au:TiO2 ratios is 1 Au particle in 1000–65,000 TiO2 particles. The presence of the agarose template created a highly open and porous Au/TiO2 structure. The surface area was 50 5 m2 g 1 for all the materials. Fig. 1 shows a typical diffraction contrast image of the highly open structure, with pore diameters of several hundred nanometres. The networks are comprised of agglomerated polyhedral anatase particles with a mean diameter of 20 nm. The pile-up of features in projection which occurs in the structure (Fig. 1 top right) leads to very dark regions in this imaging mode. Gold particles (identity confirmed with EDS), such as that labelled in Fig. 1, are randomly distributed in the material. Any gold particles occurring in regions of high TiO2 particle density would not be apparent in this imaging mode.
Fig. 1. Bright field diffraction contrast image of a typical 10 nm gold particle sitting within the titania network. Typical pore diameters are around 100 nm. The network structures are composed of agglomerated particles of anatase 10– 20 nm in diameter.
345
Spectroscopic methods, such as x-ray energy dispersive spectroscopy (EDS) and electron energy loss spectroscopy (EELS) can readily differentiate noble metal from TiO2. However, the extremely low particle loadings and the time required to focus a probe on the particle and initiate an analysis, mean that this method is impractical for identifying particles on the basis of a point by point analysis in the first instance. Similarly, techniques such as energy filtered TEM (EFTEM) can be used to map elemental distributions. However, this is not a live technique, requiring up to several minutes per frame, and is far too slow to be used for surveying large areas of specimen. However, EFTEM is useful for high contrast mapping of the gold distribution, once the particles have been located. High angle annular dark field (HAADF) microscopy in scanning transmission electron microscopy (STEM) mode has been shown to be particularly effective in studying these types of materials (Pennycook, 1981; Akita et al., 2001). In HAADF, the incoherent scattering of electrons though high angles shows an approximate dependence on the square of the atomic number. It therefore provides excellent sensitivity for heavy noble metal dispersions on much lighter metal oxides and provides sufficient intensity for live imaging. However, the use of STEM mode is not without its drawbacks. On our instrument, lens reconfiguration on switching from TEM to STEM mode involves a 1 h downtime while the instrument stabilizes. To aid more rapid throughput of materials, a technique operating solely in TEM mode was desirable, since other TEM-based methods, such as high resolution TEM and EFTEM imaging, were also to be applied to these materials. The results of these characterisation studies will be reported elsewhere (Wang et al., 2007). The method selected for Au particle location was plasmon imaging. Plasmons are low energy loss events (<50 eV) which occur due to longitudinal wave-like oscillations of weakly bound electrons. Most materials produce quite broad (20 eV) plasmon peaks, and this low loss region of the energy loss spectrum is sensitive to specimen thickness. With increasing thickness, plural scattering results in multiples of plasmon peaks. In most instances plasmon imaging in achieved by inserting a narrow energy selecting slit (1–5 eV) and qualitatively imaging at various energy losses. This is an effective way of quickly differentiating phases with reasonably well resolved plasmon peaks (Sigle et al., 2003; Mitchell et al., 2005). Most importantly, the plasmon signal is sufficiently intense to permit live imaging (Sasaki et al., 2001). Fig. 2 shows a series of images of the same region captured using various imaging modes. In Fig. 2a, which is a diffraction contrast image (objective aperture inserted), contrast between particles is predominantly the result of orientation differences. Those particles close to strong Bragg diffraction conditions produce intense diffracted beams. The objective aperture below the specimen blocks these diffracted beams, causing the particles to appear dark. Random orientation amongst crystalline particles then produces a range of intensities. Electron dense, crystalline materials like Au, scatter electrons strongly, and produce both strong Bragg diffraction contrast (when suitably oriented) as well as strong mass-thickness scattering
346
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
Fig. 2. The effect of various imaging modes on the contrast between gold and TiO2: (a) Diffraction contrast image showing that gold and TiO2 close to Bragg conditions are indistinguishable; (b) mass-thickness contrast image of (a) showing that gold is differentiated from TiO2 at strong Bragg conditions. However, thick TiO2 regions appear dark in this mode, and gold in such regions cannot be observed; (c) 24 eV (5 eV slit width) plasmon image of (a). Gold appears dark while TiO2 and resin are bright. Thick regions of TiO2 are brightest, revealing any gold particles within them.
through high angles. In this material, where gold particles (labelled in Fig. 2a) are of similar size and shape to the TiO2 particles, diffraction contrast imaging makes it very difficult to locate the gold directly from cursory inspection, due to the similarity of contrast from the two phases. Mass-thickness contrast imaging was evaluated (Fig. 2b). Here, the objective aperture is removed, and so the contrast reflects the net mass-thickness of the specimen. Whilst the gold particle was darker than the surrounding TiO2 particles, regions of specimen where the net thickness of TiO2 was high also produced similar (dark) contrast. The gold was uniformly distributed with the TiO2, and therefore the gold particles were present at the pore surfaces and embedded within the framework of TiO2 particles. Gold particles located in the
pore walls where TiO2 particle density was highest, would not be visible using mass-thickness imaging. Therefore, this imaging mode was inefficient, since it could only locate a small subset of gold particles in the thinnest regions of the specimen. For plasmon contrast imaging (Fig. 2c), we used an energy selecting slit width of 5 eV to create a window centred on an energy loss of 24 eV. TiO2 and the mounting resin (carbonaceous) exhibit a strong plasmon excitation at this energy, but gold far less so. Catalytically important metals such as Au, Pt, Ag and Pd all exhibit relatively weak plasmon peaks over this energy range, and so should be amenable to study with this method. No objective aperture was present, to minimise diffraction contrast effects. In this imaging mode, both the resin
D.R.G. Mitchell et al. / Micron 39 (2008) 344–347
and TiO2 were illuminated strongly, while the gold appeared dark. Unlike mass-thickness imaging, the contrast is reversed in plasmon imaging, and thick regions of TiO2 were brighter than thin regions. Therefore, it was possible to quickly scan large areas of specimen and identify the gold particles, even where they were located within thick clusters of TiO2 forming the pore walls, which they invariably were. In the absence of an objective aperture, and in view of the significant thickness of the TEM specimen (90 nm), not all regions of the specimen were at focus. As a result, TiO2 grains close to strong Bragg conditions did produce strong diffraction contrast when they were defocused, due to the separation of bright and dark field images. However, such candidate particles could be quickly eliminated from the search, since the bright field/dark field contrast could be quickly nulled by refocusing. The contrast from gold particles, was not significantly affected by focus. Identification of gold particles via plasmon imaging was confirmed using focused probe EDS. This showed the plasmon imaging method to be extremely effective in locating the gold particles. Having identified a method for locating the gold, large areas of foil could be quickly scanned with a live plasmon image displayed. This enabled gold particles, with number densities as low as 1 in 65,000 to be located in just a few minutes per particle. 4. Summary Diffraction contrast, mass-thickness and plasmon imaging have been compared as methods for locating low number densities of gold nanoparticles dispersed on a porous nanoparticulate TiO2 (anatase) support. The similar shape and morphology of the gold and TiO2 meant that where TiO2 particles were located close to strong Bragg diffraction conditions, diffraction contrast imaging was poor at differentiating the two phases. Mass-thickness imaging highlighted gold particles in the thinnest regions of the specimens. However, in the pore walls the high mass-thickness contrast of the TiO2 made it impossible to locate the majority of the gold particles which were located in these regions. Plasmon imaging, using an energy window width of 5 eV at an energy loss of 24 eV was highly effective, since both the TiO2 and the resin used for TEM specimen impregnation produced a strong plasmon signal, especially in the pore walls, and this contrasted strongly with the weak plasmon emission of the gold. The
347
plasmon signal is sufficiently intense to permit live imaging and rapid surveys of large areas to locate the very low density of particles (as low as 1 in 65,000) are readily achieved. The similar plasmon emission characteristics of many catalytically important noble metals, suggests this method might be applicable to a wide range of catalyst systems. Acknowledgements This work was funded by the Australian Research Council and the Australian Institute of Nuclear Science and Engineering (AINSE). RAC acknowledges the Australian Research Council for an Australian Research Fellowship. Dr Simon Crawford is thanked for TEM specimen preparation. References Akita, T., Tanaka, K., Okuma, K., Koyanagi, T., Haruta, M., 2001. TEM and HAADF-STEM study of a Au catalyst supported on a TiO2 nano-rod. J. Electron Microsc. (Tokyo) 50, 473–477. Arabatzis, I.M., Stergiopoulos, T., Andreeva, D., Kitova, S., Neophytides, S.G., Falaras, P., 2003. Characterization and photocatalytic activity of Au/TiO2 thin films for azo-dye degradation. J. Catal. 220, 127–135. Hadjiivanov, K.I., Klissurski, D.G., 1996. Surface chemistry of titania (anatase) and titania-supported catalysts. Chem. Soc. Rev. 25, 61–69. Kamat, P.V., 2002. Photoinduced transformations in semiconductor-metal nanocomposite assemblies. Pure Appl. Chem. 74, 1693–1706. Li, F.B., Li, X.Z., 2002. Photocatalytic properties of gold/gold ion-modified titanium dioxide for wastewater treatment. Appl. Catal. A: Gen. 228, 15–27. Li, X.Z., He, C., Graham, N., Xiong, Y., 2005. Photoelectrocatalytic degradation of bisphenol A in aqueous solution using a Au:TiO2/ITO film. J. Appl. Electrochem. 35, 741–750. Mitchell, D.R.G., Attard, D.J., Finnie, K.S., Triani, G., Barbe´, C.J., Depagne, C., Bartlett, J.R., 2005. TEM and ellipsometry studies of nanolaminate oxide films prepared using atomic layer deposition. Appl. Surf. Sci. 243, 265–277. Pennycook, S.J., 1981. Study of supported ruthenium catalysts by STEM. J. Microsc. 124, 15–22. Sasaki, K., Tsukimoto, S., Konno, M., Kamino, T., Saka, H., 2001. Dynamic chemical mapping near a Si/SiO2 interface at elevated temperatures using plasmon-loss images. J. Microsc. 203, 12–16. Sigle, W., Kra¨mer, S., Varshney, V., Zern, A., Eigenthaler, U., Ru¨hle, M., 2003. Plasmon energy mapping in energy-filtering transmission electron microscopy. Ultramicroscopy 96, 565–571. Sonawane, R.S., Dongare, M.K., 2006. Sol-gel synthesis of Au/TiO2 thin films for photocatalytic degradation of phenol in sunlight. J. Mol. Catal. A, Chem. 243, 68–76. Wang, X.D., Egan, C., Zhou, J.F., Zhou, M.F., Prince, K., Mitchell, D.R.G., Caruso, R.A., 2007. Preparation of porous Au/TiO2 nanocomposites by templating with agarose gel, in preparation.