In this instance, anin situtesting setup is one in which the quantitative response of a sam- ple is gathered simultaneously with a morphological response, in this case micrographs. The Greer groupin situinstrument, SEMentor [34], performs this task through a combina- tion scanning electron microscope (SEM) and nanomechanical testing arm (nanoindenter). From it we can learn, by observation, what deformation events correspond to unique signals
in the data, thus it is our ‘mentor.’
Samples to be tested are mounted on a 90 degree SEM stub and loaded into the SEM chamber at a tilt of 4 degrees from vertical. The mechanical testing arm is mounted on a side port of the SEM (FEI Quanta 200) at an angle of 4 degrees from the horizon. The geometry is illustrated in the schematic in Figure 2.14 with photos of the actual system given in Figure 2.15. The mechanical arm is a derivative of the technology in the DCM head of the Agilent Nanoindenter G200 and is therefore subject to the same limitations. The first limitation is in the raw displacement which ranges from−15to +15µm. Also, maximum attainable load is 10 mN with a resolution of 50 nN. The maximum load is sufficient for our tests, but because of the limited raw displacement, in situ testing of 60
µm tall VACNT pillars is limited to nominal strains of only 50%.
el ec tr on b ea m c ol um n mechanical testing arm sample stage 4º tilt vacuum chamber mounted sample (via 90º SEM stub)
indenter head
Figure 2.14: Schematic of thein situtesting instrument, SEMentor.
Load versus displacement data obtained in the SEMentor is analyzed slightly differently from that gathered in the G200. This is for two reasons. First, characterization of the DCM head is more difficult due to the proprietary nature of its exact geometry. Second, an accurate quantitative measurement is unnecessary as SEMentor is most beneficial as a visualization instrument. Therefore, raw load and displacement are corrected using the
Figure 2.15: Photo of the in situ testing instrument, SEMentor [34]. The image on the left is the complete system with the mechanical testing arm shielded from environmental influences. The image on the upper right is a few inside the SEM chamber. The image on the lower right is the nanomechanical testing arm before shielding has been installed.
‘Support Spring Stiffness’ channel, which is a table of leaf spring stiffness values as a function of raw displacement, ks,table(uraw), obtained during a calibration run. This data
is quite noisy, so the channel must be collected during thein situcompression, smoothed, then removed from the applied load yielding an approximate load on sample,
pcorr =praw(uraw)−praw,surf−(uraw−uraw,surf)ks,table(uraw), (2.8)
where the subscript ‘surf’ refers to value at the point of surface contact. This correction is necessary because of the large position dependence of ks,table, which would otherwise
Chapter 3
Characterization of VACNT Morphology
3.1
Introduction
The specific microstructure of VACNT structures plays a key role in their mechanical re- sponse. This has been found qualitatively for a range of structures obtained through slight variations in growth by McCarter et al. [16] as well as being evident in the range of mod- uli reported in the literature (see Table 4.1). This chapter overviews some characteristic morphological aspects of the CNT bundles tested as well as the methodologies used to obtain these properties. Several key features set VACNTs apart from other materials. As stated earlier, the structures are hierarchical. That is, under lower magnification (∼1,000×), the tubes in the bundles appear vertically aligned, i.e., perpendicular to the substrate (Fig. 1.2, left). However, under larger magnifications (> 30,000×), it becomes evident that the CNTs are randomly oriented in a very porous network, forming a fibrous, interconnected web of support structures, where individual tubes interact with one another (Fig. 3.1, left and right). The presence of these interactions is the distinguishing attribute of this type of VACNT material in contrast to vertically aligned CNT ‘forests,’ in which individual tubes are far enough apart (and short enough) to grow perpendicularly to the substrate without interacting with their neighbors. Second, there exists a height-dependent inhomogeneity in the bundle structure due to its growth mechanism [35]. Visual inspection indicates there is a lower density of tubes with less vertical alignment at the bottom. This, in turn, results in fewer and weaker load bearing members at the bottom and therefore a more compliant material in this region of the pillar. These microstructural gradients are illustrated in the
high magnification SEM images taken of the pillar surface at evenly spaced heights along the pillar axis in Fig. 3.1. Quantification of this gradient has only been reported for high energy synchrotron measurements of bulk CNT films as they grow [35], a method that is both expensive and inapplicable to our sample geometry. For this reason, in Section 3.2 I discuss the image analysis techniques developed to analyze this relative density gradient directly from the SEM micrographs. These techniques are developed in collaboration with Peter Capak of the Spitzer Science Center at Caltech.
500 nm 10 µm
Figure 3.1: Cylindrical pillar with32,000×magnification insets, revealing the highly inho- mogeneous CNT microstructure from bottom to top. The lower leftmost image corresponds to the bottom of the pillar and illustrates the sparser (less dense) and less vertically aligned CNTs when compared to the top of the pillar (upper rightmost image). Note that the surface tubes appear brightest because they return more signal to the electron detector, but these tubes are not indicative of the internal pillar microstructure and should be looked beyond in order to observe the density and alignment variation discussed.
Other important microstructural elements are the makeup of the individual CNTs them- selves and the average density or porosity of the structure. Presently, there appears to be some link between tube diameter (number of walls) and/or the surface roughness of CNTs on the presence of irrecoverable versus recoverable deformation behavior [36], but this connection remains to be rigorously proven. In the present work, individual CNTs
were characterized by transmission electron microscope (TEM) (see Fig. 3.2). Diameters are found to vary between 15 and 30 nm with 22 nm being the average value. The tubes themselves are multiwalled, typically comprised of 4–5 walls per tube. Average density is clearly an important feature in comparing the mechanical response of any foam-like mate- rial. For example, the elastic modulus of a foam scales with the relative density squared for foamed metals and polymers [23], where relative density is the foam density divided by the density of a single, monolithic strut. Similar relationships exist for energy dissipation and plateau stress. Unfortunately, determining either the average tube number density or mass density has proven challenging due to the small size of individual pillars (lack of material for bulk measurement) and the large amount of open space (small surface area per gram of sample). Attempts to determine the density of the samples tested are discussed in Section 3.3.
20 nm 10 nm
Figure 3.2: TEM images illustrating the typical multiwall CNTs making up the VACNT bundles tested. There are typically 4–5 walls per tube. Images taken by A. T. Jennings.