Nanoscale Structure, Dynamics, and Aging Behavior of Metallic Glass Thin Films

Scanning tunnelling microscopy observations resolve the structure and dynamics of metallic glass Cu100−xHfx films and demonstrate scanning tunnelling microscopy control of aging at a metallic glass surface. Surface clusters exhibit heterogeneous hopping dynamics. Low Hf concentration films feature an aged surface of larger, slower clusters. Argon ion-sputtering destroys the aged configuration, yielding a surface in constant fluctuation. Scanning tunnelling microscopy can locally restore the relaxed state, allowing for nanoscale lithographic definition of aged sections.

such as enlarging clusters during scanning or the larger spontaneous formation of very large surface clusters. Hopping clusters proved to be too small for clear observation with SEM. In order to verify hopping, the use of TEM is required, however the observations necessitate a change in substrate and film thickness.
Use of scanning electron microscopy offers the opportunity to observe films identical to those observed in the STM. A sample composed of Cu 80 Hf 20 , 50 nm thick was transfered into a Hitachi S-5500 high resolution SEM immediately following deposition onto an oxidized silicon wafer. The film structure reveals a very flat surface as expected, but also shows some sections where the topography exhibits larger smooth features protruding above the rest of the surface. The more variable topography in these sections makes the film structure less stable, meaning that these areas are ideal locations to investigate for dynamics.
Indeed, in these areas, multiple types of surface changes were seen. Upon scanning these features, the surface was found to planarize, and to convert from a smooth appearance into a pebbled texture. This corroborates the STM observations of aging. The SEM cannot image the extremely small amorphous clusters initially, however as scanning drives the unstable sections of the surface to rearrange, clusters grow and stabilize into the surface reconstruction observed with STM (Fig. 4). The clusters are lithographically enlarged with the beam, until they become large enough to be resolved by the SEM. This matches very well with the aging process observed following sputter cleaning of the STM samples.
Additional shifts of material in varying quantities between locations were observed ( Fig.   5). Holes enlarged or filled in the film surface, large cluster like objects appeared and disappeared. Periodically, a preferential site would begin to accumulate material rapidly, forming very large surface features (Fig. 5). These events match the large surface changes found in STM, including the observation of material surrounding the large feature disappearing in conjunction with the feature growing.
The SEM observations did not have sufficient resolution to observe clusters sufficiently small that they exhibited stable hopping. Observations of films on TEM substrates (silicon nitride, graphene, or carbon membranes) also proved to be challenging, with poor contrast when looking directly through the film. Success was found using an unconventional technique of depositing the sample films onto low density arrays of carbon nanotubes. Films were sputtered onto carbon nanotube substrates, coating the tubes with a very thin (5-15 nm) layer of Cu 80 Hf 20 . The samples were immediately transferred to a JOEL-2100 TEM for observation. The coating of a 3D substrate affords the opportunity to look at sections of the film in profile, where fluctuations in thickness of the film are much more obvious. A caveat to this observational technique is the fact that the film is on a different substrate, is a different thickness, and has very different film geometry. However, this approach did allow TEM based verification of hopping clusters as shown in figure 6 with hopping events highlighted by colored arrows.      . At the edges of the tube, the film can be observed in profile. Jumps are rare, and difficult to spot. Fast TEM scans were acquired at a frequency of one per 60 s and compared to identify cluster jumps from image to image. In the four sequential panels shown, several clusters jump. Color coded arrows indicate transitions between frames.