Caroline E.
Blackmore
ab,
Neil V.
Rees
b and
Richard E.
Palmer
*a
aNanoscale Physics Research Laboratory, School of Physics and Astronomy, University of Birmingham, Birmingham, B15 2TT, UK. E-mail: R.E.Palmer@bham.ac.uk
bSchool of Chemical Engineering, University of Birmingham, Birmingham, B15 2TT, UK
First published on 4th March 2015
Size-selected binary platinum–titanium dioxide (Pt–TiO2) clusters have been generated using a magnetron sputtering gas condensation cluster source and imaged using a Scanning Transmission Electron Microscope (STEM) in High Angle Annular Dark Field (HAADF) mode. The core–shell clusters exhibit a Pt core of preferred size 30 ± 6 atoms (1 nm), embedded in an oxidised Ti shell, independent of the overall cluster size (varied between 2 nm and 5 nm). Smaller clusters, with mass ≤50000 Daltons, show a single Pt core while larger clusters, ≥55000 Daltons, feature multiple Pt cores, either isolated or aggregated within the TiO2 shell. These clusters may have applications in solar hydrogen production; preliminary work indicates catalytic active in the hydrogen evolution reaction.
Platinum–titanium (Pt–Ti) nanoparticles have been predicted theoretically to offer better reaction kinetics and lower poisoning levels than pure platinum nanoparticles within Polymer Electrolyte Membrane (PEM) fuel cells.8,9 The target is a core of Ti atoms within a shell, one to two atoms thick, of Pt atoms. By contrast, for solar hydrogen production (water splitting) TiO2 shell nanoparticles may be preferred.10,11 Chemical synthesis of Pt–TiO2 core–shell clusters has been reported.12–14 Amongst research into TiO2 photo catalysis,15,16ref. 17 shows that the addition of Pt improves the efficiency of visible light absorption, thus improving the water splitting capacity.
The aim of this work is to research the multi-core structure of Pt core–TiO2 shell clusters, which, ultimately may have applications within solar hydrogen production. Within this paper we report the production of size-selected Pt core–TiO2 shell clusters by gas-phase cluster beam methods. This proof of concept work focuses on fine mass control of the clusters, producing enough clusters for electrochemical characterization. The clusters are deposited and imaged on holey carbon supports by aberration-corrected STEM in HAADF mode, established as a powerful probe of the atomic structure of size-selected clusters.18–22 Small clusters display a single Pt core, whilst larger clusters exhibit multiple Pt cores; in both cases the shell is TiO2. Multiple core nanoclusters have previously been produced via chemical synthesis23–25 and, recently, by magnetron sputtering-based techniques.26 These have been predominantly targeted towards biomedical applications, exploiting their magnetic properties. We identify a preferred core size of 30 ± 6 Pt atoms, which seems to represent a critical nucleus for the condensation of Ti atoms in the source. Larger clusters are generated by the modular assembly of this ‘simple’ core–shell cluster by gas-phase aggregation.
To analyse the STEM images a macro was written in FIJI. The macro works by taking the value of each pixel and averaging it with the value of its nearest four neighbours. Regarding the background the median value from the image is taken as a good estimation of the background intensity, thus the criteria for thresholding is a comparison of this median and the corrected pixel value. This method reduces the amount of single pixel noise (which is the same intensity as the Pt atoms) filtered as Pt. The FIJI built in analyse particles function is then used to identify and number these areas as cores.
The growth mode of the clusters, as deduced from the STEM data, is initial Pt condensation followed by Ti condensation onto the Pt cores. The multiple core structure of the clusters shows that smaller clusters of Pt–Ti, of mass ∼30 kDa are aggregating in the source to form the larger clusters, i.e., what we have is a kind of modular construction. From the images it can be seen that some of the Pt cores in the larger clusters have a similar size to those in the small “building block” clusters. Larger cores, we also observed, are presumably formed by aggregation of the cores during the coalescence of the “modules”.
To illustrate our method of quantitative analysis of the STEM images Fig. 2A and C show one nanocluster (60 kDa) before and after thresholding, as described earlier. Discarding particles less than 0.25 nm in diameter as noise and single atoms, the area within the particle identified in Fig. 2C is used to calculate the radius of an equivalent circle, and thence the volume of a Pt sphere. Using the bulk density of Pt, the sphere's volume is then converted to the number of Pt atoms in each core. From Fig. 2B and D it can be seen that for a two-cored cluster our macro identifies two particles as expected, and outputs the Pt core nuclearity for each. This analysis has been performed on a minimum of 50 clusters for each mass of cluster deposited.
The method of analysis employed is very reproducible and fast, but of course errors can be introduced. The automatic thresholding can lead to some Pt atoms being missed or TiO2 atoms being included within the threshold. To minimise such errors we checked all images manually, adjusting as necessary. Assuming that the cores are spherical leads to an error that cannot be corrected for. Using the bulk density of Pt to calculate the number of Pt atoms within the calculated volume will also lead to an error, as the atomic structure of the embedded is not resolved sufficiently to correct for the bond lengths. To reflect the accumulated errors, the histogram bar widths in Fig. 3–5 have been chosen to be 12 atoms so that the number of Pt atoms is quoted as the centre of the bar ±6 atoms. Thus the relationships between the number of Pt atoms in different cores are sound, but the precision in the absolute numbers of Pt atoms might be improved according to the analysis method deployed.
Fig. 3 Histogram to display the number of cores within each cluster with relation to each core size. |
Fig. 4 Analysis of the number of Pt atoms within the single cored clusters between 20 kDa and 50 kDa. |
Repeated imaging with STEM of clusters has shown the clusters to be stable over a period of months.
To further investigate the phenomenon of a preferred Pt core size, Fig. 4 shows the size distribution of the Pt cores from clusters with only one core and overall cluster mass less than 50 kDa. 92% of the clusters up to 50 kDa were found to have single cores. These are the “modules” from which we believe larger clusters are assembled in the gas condensation growth process. The figure shows a single peak at 30 ± 6 Pt atoms for the nuclearity of the core. Within this range of cluster mass there is no change in average core size as the deposited cluster size varies. Instead the data support the idea that a preferred size of Pt core functions as a critical nucleus upon which Ti atoms are condensed. Theoretical work on the structure of small Pt particles predicts that for Pt clusters between 24 and 38 atoms, cubic isomers have the lowest energy, with 27 atoms being especially stable.32
Fig. 5A and B show Pt core-size for Pt–Ti clusters of mass larger than 55 kDa. They have been split into single core and two core plots. 64% of the clusters above 55 kDa have a single core, while 25% have two cores. The remaining 11% contain 3 or more cores. The single core graph shows a peak at 126 ± 6 Pt atoms, with very few cores containing less than 60 Pt atoms. This observation is again consistent with the picture in which single cores of a size ∼30 Pt atoms coalesce during cluster aggregation during the growth process. This interpretation is confirmed by looking at the double core data, Fig. 5B which does show a peak near 30 Pt atoms, specifically, at 42 ± 6 Pt atoms; we believe these are the cores that have not coalesced during cluster aggregation.
It is tempting to speculate that the probability that two Pt cores, in two Pt–Ti core–shell clusters which aggregate during gas phase growth, coalesce into a single larger core may depend on the relative location of the Pt cores with respect to the edge of the cluster, as well as the relative orientation of the clusters upon collision. Our data suggests that the Pt cores are often found off-centre, i.e. towards the cluster edge (e.g.Fig. 1D) but this notion needs further experimentation and analysis. Eccentric cluster cores in core–shell systems were recently predicted.33
Control experiments were performed by recording cyclic voltammograms at a voltage scan rate of 25 mV s−1 using first, the bare GC electrode, and secondly the GC electrode modified with TiO2 clusters. For this work the clusters of Pt–TiO2 and pure TiO2 were soft landed on to the GC electrode. These TiO2 clusters were produced in the same manner and mass range as the Pt–TiO2 clusters. Fig. 6 clearly shows that there is no perceptible change in the position of the reduction peak, confirming that TiO2 does not behave catalytically compared to bare GC.
Next, experiments were conducted with the Pt–TiO2 clusters deposited onto the GC electrode. Fig. 6 shows a clear shift to lower overpotentials of both the onset of the reduction wave and the peak. In this case, we are justified in interpreting this shift as an indication of improved catalytic ability due to the low electrode coverage (ca. 5%), precluding the likelihood of the mixed mass transport regimes observed in multilayer deposit systems which also causes shifts in the wave to lower overpotentials.34
Work is currently in progress to unambiguously elucidate the kinetic parameters, since initial modelling suggests that the deposit may not be purely Case IV but is equally likely to be Case II or III,35 which is consistent with the approximate coverage of the GC substrate electrode. We note the lower magnitude of current compared to the GC control experiment is consistent with Case II or III behaviour.
Using a very approximate Case IV assumption, modelling indicates that the presence of the Pt within the Pt–TiO2 cluster has an accelerative effect on the kinetics of ca. 2 orders of magnitude. Detailed studies involving variation of coverage and simulating a Case III model are expected to confirm the rate constant and transfer coefficient.
This journal is © the Owner Societies 2015 |