TEM imaging provided information on the nc size and depth distributions through the cross-section of the samples. TEM was performed on a total of eight samples; all four doses, as-implanted and 1100°C anneal.
Figure 6-6 displays the central region of 1e16, as-imp. Nc sizes are of the order ~2 nm, uniformly distributed through the implanted region. However, a large fraction of the implanted Cu is believed to be bound with O from the amorphous SiO2 film (as
will be discussed in section 6.2). With TEM oxidised Cu is not differentiable from the surrounding SiO2, and not observable on the micrographs. A quantitative
determination of the oxidised fraction could be obtained from detailed TEM analysis, calculating the fraction of Cu atoms apparent in nc and subtracting from the implanted dose determined by RBS. Micrographs of 1e16, 1100°C indicated no nc-Cu, as was to
be expected from the RBS results were the sample appears devoid of Cu.
Individual bands for each implant energy were observed for the higher dose samples, 3e16, 1e17 and 3e17, with five, three and one band, respectively, according to multiple energy implants, apparent in e.g. Figure 6-7. It appears technically difficult to achieve a totally uniform size distribution, however the deviation in size distribution seems less apparent than for single energy implants [15]. Furthermore it substantially increases the yield in EXAFS measurements.
Figures 6-8 and 6-9 are micrographs from the deepest band of 3e16, as-implanted and 1100°C, respectively. A fairly uniform size distribution can be observed. The fraction
Figure 6-6: Micrograph of 1e16, as-imp. Red circles indicate a selection of nanocrystal of diameter ~2 nm.
of small nc appears to decrease with annealing, suggesting these have coalesced into larger nc. The band-structure is obvious from Figure 6-7.
These images indicate that relatively low implant doses (e.g 3e16), and no annealing (e.g. as-implanted), are adequate conditions for the formation of Cu nc. However, annealing enhances the nc growth by Ostwald ripening, consuming the smaller nc. The highest energy implant (Figure 6-7) appears separated from the rest, a feature currently not understood. Further it also exhibits the largest nc of the bands. However, note that the deep region is also the thickest part of the sample27.
27 See the TEM sample processing technique, Section 5.2.3.
1 mm
~2.6 mm
Shallow region, SiO2 surface
Figure 6-7: Micrograph of 3e16, 1100°C. The full depth of implanted area is shown. The deepest band corresponds to the 5000 keV implant, with projected range of 2.59 mm. The shallower bands are smeared into each other. Note that the deepest band appears to contain the overall largest nanocrystals.
100 nm
Figure 6-8: Micrograph of 3e16, as-imp. Red circles point out a selection of large nanocrystals of diameter ~20 nm.
Figure 6-9: Micrograph of 3e16, 1 1 0 0°C. Red circles point out a selection of large nanocrystals of diameter ~20 nm.
Figure 6-10 displays the full depth of the implanted region of the 1e17, 1100°C. The
sample was implanted at three energies, and we can observe three bands in the micrograph. A shallow region with small nc, a deep region with relatively larger nc, and an overlapping thin region with the largest nc. Figures 6-11 and 6-12 are higher
resolution micrographs of the deeper band in the 3e17, as-implanted and 1100°C
sample, respectively. Notice several nc, e.g. in Figure 6-12, where the large ~25 nm nc overlaps with one half its size. At this dose the size distribution before and after annealing was similar.
3e17, as-imp and 1100°C is shown in Figure 6-13, to the left is the perforation. The
sample surfaces are separated by glue, and on each side the total thickness of SiO2
(~4.5 mm) is visible with the implanted region, which is found at depth 1 – 2 mm,
confirming results from TRIM. Significantly larger nc are found in the deeper region indicating preferential nucleation and/or growth in this area. The reason for this is not fully understood. It is interesting also to note how abrupt the implanted region ends, Figure 6-14. Figure 6-15 is a close-up of the deepest region of 3e17, 1100°C and
shows the band of large nc (> 30 nm), with the deeper region almost depleted (to the right). Figure 6-16 is extracted from a shallower region of the same sample, the mean diameter of nc here is found to be < 20 nm.
500 nm
Approximate position of SiO2 surface
Direction of implantation Figure 6-10: Micrograph of 1e17, 1100°C. A three band structure is seen corresponding to a depth of ~1 – 2 mm, confirming the theoretical range as calculated by TRIM. See Table 3-1 for reference on straggle length.
Figure 6-11: Micrograph of 1e17, as-imp. Red circles show nanocrystals of diameter ~15 nm. Figure 6-12: Micrograph of 1e17, 1100°C ~12 nm ~12 nm ~14 nm ~25 nm
For all samples, the nc are close-to-spherical and non-facetted, which can be expected with an amorphous matrix. Figure 6-17 shows a high-resolution image of a nc in the 3e17, 1100°C sample of size ~20 nm. Lattice fringes can be observed. These arise as a
consequence of diffraction and interaction of the electron beam penetrating the nc lattice, and is a direct measure of the lattice spacing.
Figure 6-13: Micrograph of 3e17, as- imp and 1100°C. Each color on the scale corresponds to 1 mm. The implanted region is found between 1 – 2 mm in depth, in accordance with TRIM results.
Deep implant region Shallow implant region SiO2/glue interface
2 mm SiO2/Si interface Si SiO2 as-imp Si SiO2 1100°C
Ion beam milled x-sectional hole
Figure 6-14: Micrograph of 3e17, as-imp. The whole implanted region is shown, with the shallow region at the top left and the deeper region at lower right. Note the general increase in nanocrystal size with depth.
~ 50 nm
~ 50 nm ~ 30 nm
~ 50 nm
Figure 6-15: Micrograph of 3e17, 1100°C. The deepest region shows an end-band of very large nanocrystals, up to 50 nm in diameter. The shallow region is to the left and contains significantly smaller nanocrystals.
~18 nm ~15 nm ~18 nm ~18 nm ~18 nm ~15 nm
Figure 6-16: Micrograph of 3e17, 1100°C. Extract from the middle region shows nanocrystals of sizes < 20 nm.
The lattice spacing was found to be d = (2.08 ± 0.05) Å, corresponding to the nominal bulk value of 2.088 Å for (111) Cu. Note that this nc is observed to be a single crystal, no twinning is apparent.
In conclusion TEM further confirmed the Cu phase was bulk like. TEM provided a size distribution of the nc as opposed to mean sizes estimated by XRD. In addition micrographs confirmed a lack of nc in the 1e16, 1100°C sample, consistent with a lack of Cu apparent from RBS
measurements. A general increase in mean nc size was observed with increasing dose and temperature, with the exception of the 3e17 samples.