Target preparation of an inclined nanocolumn

This section treats the most demanding FIB specimen preparation for a tomographic inves-tigation. The low-dimensional GaN based heterostructures which are described in section A.2, have to be isolated and oriented with their preferential growth direction to the tilt axis of the goniometer. The indispensability of a tomographic approach is highlighted in figure 3.7.

An idealized highly symmetric shape of a hexagonal nanocolumn is shown in the sketches.

The inclusion of In on different facets can be studied by standard cross-sectional samples as long as a highly symmetric surface justifies the consideration of projections from one low-indexed direction (cf., e.g., [151–153]). The schemes in figure 3.7(a) and (b) illustrate such a situation for core-shell nanocolumns with a hexagonal symmetry that grow epitaxially on a substrate with a high symmetry surface orientation. A plan-view specimen reveals the layer structure on facets in the [0001] zone which is outlined in the right of figure 3.7(a). The section perpendicular to a side facet allows to analyse the layer sequence grown on the side walls as well as on the top and pyramidal facets which is demonstrated in figure 3.7(b). Here, this case corresponds to one of the

11¯20

zone axes. In contrast, this does not work for the perpendicular orientation which provides a

1¯100

viewing direction. The layers then lie inclined toward the imaging direction. The presented sample exhibits the less symmetric (11¯22) surface and nanocolumns that grow along the [0001] direction, i.e. inclined toward the surface normal. This arrangement is schematically illustrated in figure 3.7(c). Standard cross-sectional samples result in specimen foils perpendicular to the surface. Accessible zone axes are the unfavourable [1¯100] and the high indexed [¯1¯123] direction. The actual morphology of the nanocolumns is unknown and the analysis of projection images requires assumptions on the habit similar to the idealized representations in figure 3.7.

As a consequence of this geometrical situation, conventional TEM projections are insuffi-cient to characterize the abundance of In in (In,Ga)N layers on different facets. Therefore, a method is described in the following that enables the analysis of the nanocolumns’

morphol-ogy and the In content on different facets. The presented strategy comprises the orientation and isolation of the target nanocolumn which is described in this section, and the data acqui-sition for electron tomography which is discussed in the subsequent sections. The strategy commences with the adaptation of sample preparation by FIB with respect to the require-ments of electron tomography. The faced challenges are grouped in requirerequire-ments imposed by the tomography experiment and such imposed by the treatment of the material system with Ga⁺-ions. At this point, it has to be emphasized that the presented procedure of sample preparation works at the limit of the used Ga⁺-ion based dual beam device.

The first group defines the aspired orientation of the sample at the tomography pin. The rotation around the surface normal~n for the acquisition of the tilt series is not feasible. The thickness that electrons have to penetrate, would be too high if the nanocolumns (diameter of circa 180 nm) are imaged along the [¯1¯123] direction. Beyond, the highest material con-trast between (In,Ga)N layers and GaN is realized by looking approximately parallel to the {1¯100} planes. The distinct contrast improves the tomographic results by means of SNR.

Consequently, the nanocolumn axis, i.e. the [0001] direction, is ideally aligned with the tilt axis to view all

11¯20 and

1¯100

directions.

The second group deals with the interaction characteristics of the charged Ga⁺-ion beam with GaN and Al₂O₃. The latter material is an isolator and charges up which impedes the position control of the milling area or disables imaging at all. The former material is prone to heavily recrystallize and/or redeposit when etched by the Ga⁺-ion beam (cf. section 2.2.1).

In the following, the different steps of the successful preparation are described.

In a first step, the whole sample is coated with thermally evaporated carbon to protect the surface against ion beam damage like in the former examples of antimonide samples.

In addition, the carbon layer improves the conductivity of the surface. Afterwards, carbon is deposited by GIS-FIB in two distinct directions. The inset in figure 3.8(a) schematically illustrates the two deposited carbon stripes. The one labelled "A" offers the common protec-tion against ion beam damage during the lift-out procedure. The one labelled "B" is aligned with the preferential growth direction of the nanocolumns. Later on, it serves as protection during machining the sample along the direction that is adapted to the sample geometry.

Figure 3.8(a) shows the SEM view onto the sample surface after both steps of carbon depo-sition. A major optimization of the lift-out procedure concerns the processing of GaN with

(1122)

Figure 3.7.The schematic depicts hexagonal GaN nanocolumns that grow along the high symmetric [0001]

direction. If this direction is aligned with the sample surface normal, conventional TEM plan-view (a) and cross-sectional specimens (b) provide information about layers inserted on the bounding facets (orange).

(c) The inclined growth of nanocolumns complicates the access to this information.

5 µm (a)

1 µm (c)

5 µm (b)

A B A

B

Al₂O₃ GaN C

Figure 3.8.(a) The SEM image shows two protective carbon layers A and B. The inset schematically demon-strates the arrangement of the layers with respect to the nanocolumn orientation. (b) SEM image: The lamella glued to the manipulator needle/tip is cut free except for two spots (arrows) where GaN is rede-posited. (c) The FIB scan over the ion beam polished cross-section of the lamella reveals the chemical structure of the sample with the Al2O3substrate, the GaN template with the GaN based nanocolumns on top and the protective carbon layer. The occurrence of redeposited GaN is described in the text.

an Ga⁺-ion beam. Figure 3.8(b) presents a more than 2 µm thick lamella immediately before the removal from the sample. The large thickness is owed to the size of nanocolumns pro-jected along the [¯1¯123] direction. The manipulator needle is attached and slits to detach the lamella are cut. Two arrows mark regrown GaN at the lower trench which has been cut first.

This reconnection impedes the lift-out. The pronounced tendency of GaN redeposition or recrystallization in the vicinity of a milling area is underlined by figure 3.8(c) and by figure 2.7. The former image is recorded from the polished lamella’s cross-section with the FIB.

It highlights the chemical structure. In contrast to the polished area, the remaining lamella surface (oblique view, right image part) is covered with a rough layer. The insulating Al₂O₃ appears black due to the positive charging by Ga⁺-ions and the consequent reduction of the secondary electron signal. The carbon deposition appears in dark grey and is embedded in GaN represented in light grey. The left rim of the Al₂O₃ substrate evidences a coverage with GaN, too. Obviously, the left side of the lamella has been opposite to the side from which the final trench has been applied. Hence, the reconnection of the lamella with the substrate has to be prevented. Here, the generous and "iterative" cutting of trenches solves the problem. Iterative means, the sample has to be rotated by 180^◦ for several times in or-der to successively establish the bottom cut to free the lamella from both sides unor-der 52^◦. In that way, the symmetric Al₂O₃ wedge at the bottom of figure 3.8(c) is established. The initial trenches to define the lamella are created equally deep for this procedure. It is noticed that the low sputter yield of Al2O3limits the attainable depth of the trenches or the lamella height, respectively. Finally, the GaN redeposition gives a further reason to thoroughly em-bed the GaN nanocolumns in carbon. It prevents the random growth of GaN crystallites at the nanocolumns surface. In figure 3.8(c), the inclined nanocolumns on top of the GaN tem-plate are visible. Voids have remained during carbon embedment. The successful coverage of the low-dimensional objects only works where the distance between nanocolumns is

suf-5 µm

1 µm

Al₂O₃ 2 µm C GaN

(a) (b) (d)

(c) specimen post

lamella

GaN Al₂O₃

target axistilt

tiltaxis n

Figure 3.9.(a) SEM BSE image of the attached raw lamella and (b) a magnified extract showing the aligned nanocolumns in polished cross-section of the lamella. The BSE signal is sensitive to the chemical com-position. (c) The sketch of the lamella cross-section with the correct nanocolumn alignment points to the need for a further carbon deposition as filling material. (d) The final raw lamella with the carbon filling is imaged by secondary electron SEM.

ficiently large. Later experiments have been carried out with an embedment of nanocolumns in liquid epoxy and the subsequent bake-out. This way is promising for ordered arrays of nanocolumns with a homogeneous morphology. Of course, the visibility of nanocolumns in the FIB or electron microscope is impeded. The consequent loss of local control is unac-ceptable for the presented sample of inclined nanocolumns.

The lamella is attached to the tomography sample holder post (figure 3.9(a)) under the maximal stage tilt angle of 54^◦. The chemically sensitive BSE SEM view in figure 3.9(a) and (b) show the lamella after tilting back to 0^◦. The nanocolumns are nearly aligned with the axis of the specimen post as shown in the enlarged cross-section of the lamella in figure 3.9(b). In fact, a tilt of 4.4^◦is missing. This aspect is discussed later in section 5.1. The SEM views in figures 3.9(a) and (b) are actually bird’s eye views, i.e. only the projection of the specimen post axis is found vertically in the images. The schematic in figure 3.9(c) depicts the situation with the vertical axis orientation within the image plane. The illustration of the chemical composition demonstrates a further step before isolating the target object in the final specimen shape. The hollow wedge below the region of interest in the direction of the later tilt axis has to be filled to support the final specimen. The post is turnedex situ by 90^◦ to access this wedge for carbon deposition. The result is shown in figure 3.9(d). This step will be unnecessary if a lamella is cut in the final direction from the very beginning of the preparation. The corresponding attempt has not been successful due to the redeposition of GaN. In contrast, this strategy works well for low "correction" angles. For instance, the alignment of the low-indexed [001] axis with the tilt axis for tomography is carried out for the vicinal Si(001) substrates (4^◦ miscut) which are faced in the preceding sections.

The raw lamella with its carbon support is reduced to a 280 nm thick and 1.3 µm wide lamella. The isolation of a target nanocolumn in a needle has not been attempted although it might be very helpful. The simple reason is the balance between success rate and the limited possibility to reach the necessary position refinement in the FIB-SEM to separate

the too close nanocolumns. On the other hand, the periodic arrangement of SAG objects facilitates the targeting because positions of nanocolumns can be guessed. At this point it has to be remarked that a FEG-SEM integrated in the dual-beam system would offer enhanced capabilities to localize target objects.

This last example of specimen preparation highlights the versatile and unique capabilities of the FIB-SEM instrument. It provides the access to the characterization of more and more complex 3D structures at the nanoscale. Probably, a dedicated operator will be needed in future to assemble necessary experience and to be equipped with an adequate amount of time.