Article pubs.acs.org/JPCC
Graphene Edges Dictate the Morphology of Nanoparticles during Catalytic Channeling Filippo Pizzocchero,† Marco Vanin,‡ Jens Kling,§ Thomas W. Hansen,§ Karsten W. Jacobsen,‡ Peter Bøggild,† and Timothy J. Booth*,† †
DTU Nanotech, Technical University of Denmark, Ørsteds Plads, 345E, Kgs, Lyngby 2800, Denmark DTU Physics, Technical University of Denmark, Fysikvej 311, Kgs, Lyngby 2800, Denmark § DTU CEN, Technical University of Denmark, Fysikvej 307, Kgs, Lyngby 2800, Denmark ‡
S Supporting Information *
ABSTRACT: We perform in-situ transmission electron microscopy (TEM) experiments of silver nanoparticles channeling on mono-, bi-, and few-layer graphene and discover that the interactions in the one-dimensional particle−graphene contact line are sufficiently strong so as to dictate the three-dimensional shape of the nanoparticles. We find a characteristic faceted shape in particles channeling along graphene ⟨100⟩ directions that is lost during turning and thus represents a dynamic equilibrium state of the graphene−particle system. We propose a model for the mechanism of zigzag edge formation and an explanation of the rate-limiting step for this process, supported by density functional theory (DFT) calculations, and obtain a good agreement between the DFT-predicted and experimentally obtained activation energies of 0.39 and 0.56 eV, respectively. Understanding the origin of the channels' orientation and the strong influence of the graphene lattice on the dynamic behavior of the particle morphology could be crucial for obtaining deterministic nanopatterning on the atomic scale.
■
INTRODUCTION The structure and smoothness of graphene edges, substrateinduced effects, and surface contamination by e.g. lithographic resists used in production are determining factors for the behavior and performance of a wide range of graphene-based devices.1−3 Catalytic channeling by oxidation or reduction is currently the only technique that produces channels withcrystallographically oriented suspended graphene edges with a roughness of less than 1 nm4−13 and a uniform width in the nanometer to micrometer scale determined by the diameter of the nanoparticle, and without the use of resists.4,7,11 For any possibility of control of this phenomenon, the atomistic processes need to be better understood. The metal−graphene interface is currently under study;12,13 however, an understanding of the interactions that govern nanoparticle channeling is still lacking despite prior study of this process in graphite for over 6 decades.15−24 It is our hypothesis that the predominance of graphene (100) “zigzag” edges observed in the catalytic oxidation of suspended graphene by silver nanoparticles11 is a result of the higher reactivity of graphene armchair versus zigzag edges during oxidative channeling and that the interaction of the graphene zigzag edge and the surface of the silver nanoparticle is responsible for the long, straight, and smooth channels produced in this way. Here we are able to determine and explain the structure of silver nanoparticles during channeling and the alignment of the channels along the ⟨100⟩ directions in graphene. Moreover, we show that the configuration of the graphene edge atoms, a one© 2014 American Chemical Society
dimensional interface, is able to dictate the three-dimensional shape of the silver nanoparticles. We do this through a combination of in-situ observations of the channeling behavior of silver nanoparticles on suspended mono- to few-layer graphene (Figure 1) at high temperature in a controlled atmosphere inside an environmental transmission electron microscope (ETEM) and simulation of the interface and the energy barrier for carbon atom removal from graphene by density functional theory (DFT). Furthermore, our DFT calculations of the binding energies of carbon atoms to the zigzag and armchair edges of graphene support the experimentally observed lack of dependence of channeling velocity on silver nanoparticle size.11 We conclude that the carbon is directly gasified from the silver−graphene interface15 as opposed to first dissociating into the silver bulk and later being gasified from the surface of the silver particle.19
■
EXPERIMENTAL METHODS Sample Preparation. Graphene was prepared on 90 nm PMMA-coated silicon substrates and transferred to either Quantifoil TEM grids47 or a TEM grid produced in place by electroplating.48 These procedures result in multiple freestanding graphene regions with a size in excess of 2 μm supported at the periphery. The samples were inserted into a TEM sample holder with an Inconel heating element (Gatan Received: January 23, 2014 Revised: February 5, 2014 Published: February 5, 2014 4296
dx.doi.org/10.1021/jp500800n | J. Phys. Chem. C 2014, 118, 4296−4302
The Journal of Physical Chemistry C
Article
0.1−1 Å layer is sputtered onto the entire sample surface ex situ. Electron Microscopy. The samples are imaged at 80 keV in an FEI Titan ETEM equipped with a monochromator and spherical aberration corrector of the objective lens. Images and videos are recorded on a CCD camera. To initiate the channeling behavior, we introduce oxygen at a flow rate of 2− 10 mL/min, resulting in a pressure of 4.5−13 Pa in the immediate vicinity of the sample. We use magnifications of 10K−87K×, corresponding to 180−650 pm/pixel. The beam current is kept constant at 20 nA, but the intensity of the beam is varied to ensure adequate signal-to-noise ratio at different magnifications. Drift correction, where necessary, is performed on the recorded videos using automated image registration based on the SIFT technique:49 briefly, key features are extracted from an image based on contrast and then identified in the next image in the sequence, with the relative image translation used to compensate for the drift. DFT Simulations. Density functional theory calculations were performed with the GPAW35 code, which implements the Projector Augmented Wave (PAW)53 method for treating the core electrons. The RPBE54 exchange-correlation functional was employed. In all calculations, the wave functions were expanded onto a real space grid, with a grid spacing of 0.2 Å. The graphene/nanoparticle interface was modeled as a Ag(211) stepped surface with a wide graphene nanoribbon (GNR) perpendicular to it, in contact with the silver atoms at the step. The opposite edge is saturated with hydrogen atoms. A 4 × 4 × 1 Monkhorst−Pack grid of k-points was used, and the structures were relaxed until the remaining forces were
