The MC simulation experiments were designed to confirm in silico the hypothesis that there is an energy difference of the BSE signal from the AuNPs on the bottom side of either the 70 or 100 nm ultrathin section and the ones on the upper side. From the simulation results, we took the normalized distributions of the BSE energies from the Au layer, representing the AuNPs on the top or bottom, against the relative number of hits. The energy difference between the BSE signals of the top and bottom AuNPs was clearly visible from the peak representing the energy with the highest hit count (the most likely number of interactions of electrons with the resin); however, this accounted for less than 1% of the “detected energies”. Therefore, for statistical relevance, we decided to show the differences between the ranges of the highest hit count instead. For each accelerating voltage, we took the top 60% of signals with the highest hit counts, calculated their median, and used that to visualize the difference between the BSE signals of the top and bottom NPs. The full range of the distribution could not be used, as the signals were spread too much. The results of the simulations showed clear differences between the BSE energies of the signals of the top and bottom AuNPs and how they are related to the accelerating voltage used. The energy difference between the signals of the top and bottom AuNPs increased with decreasing accelerating voltage. The low difference at 1 and 2 kV was due to the beam not penetrating the section at all, and the BSE signal originated in the resin only and not in the gold below it. The hypothesis inferred from the results was that the optimal accelerating voltage lies between 3 and 10 kV (Fig. 1 ab). At accelerating voltages larger than 10 kV, the energy difference between the signals of the top and bottom nanoparticles is too low to provide a sufficient visual difference (Fig. 1c).
This hypothesis was confirmed experimentally. The single accelerating voltage was sufficient to simultaneously visualize and distinguish the gold markers placed on both sides of the sections together with the cell structure. First, we used 100 nm sections with 10 nm AuNPs on both sides. At one selected area, multiple images at different accelerating voltages in the range of 3–15 kV (3, 5, 10, and 15 kV) were recorded and compared both with STEM and BSE imaging (Fig. 1d).
The accelerating voltages yielding the largest differences between the signals of the top and bottom nanoparticles were found to lie between 5 and 10 kV. The visual difference in the BSE images was visible at 5 kV; however, we also observed differences in the STEM images. The largest visual difference in the STEM images was at 10 kV, with the detector set to DF mode. In the BSE images, the top AuNPs are brighter and exhibit a sharp edge, while the bottom AuNPs have less contrast and have a blurry edge. In the STEM images, the top NPs have a darker shade of grey than the bottom ones. These visual differences are sufficient for the recognition of AuNPs lying on the top and bottom together with the specimen ultrastructure, which is visible in the STEM images (Fig. 1d).
Figure 1: Selection of the optimal accelerating voltage for BSE and STEM imaging at one energy. (a-b) MC simulations of the difference between the BSE signals of the top and bottom AuNPs. The optimal range of accelerating voltages, according to the energy difference in the median, is highlighted in green. (c) Exact values of the differences in the median energies. The values in yellow belong to the accelerating voltages that do not penetrate the section, and the values in green represent the hypothetical optimal range for single-energy BSE imaging. The values in orange exhibit insignificant energy differences and therefore are not usable. (d) Rough visualization of NP (10 nm) visibility and distinguishability at different accelerating voltages on 100 nm sections with different settings of the STEM detector. Common parameters: resolution: 2560x1920, dwell time: 15.625 µs, scalebar at the overview image: 500 nm. STEM settings: 3 kV – B F + HAADF, 5 kV – BF + DF, 10 kV – DF, 15 kV – BF + DF.
More accurate measurements (1 kV steps) determined that the most suitable accelerating voltage for the 100 nm sections is 7–8 kV. At this energy, the nanoparticles were well distinguishable in both BSE and STEM images, and the cellular structure was visible as well (Fig. 2). The STEM detector was operated with a combination of BF and DF.
When the recorded data were combined with those from the MC simulation experiments, the most suitable accelerating voltages produced an energy difference larger than 0.67 (Fig. 1c) and were powerful enough to penetrate the section. Therefore, the accelerating voltages selected for the 70 nm sections were those that fulfilled the same conditions: 5 kV and less.
The most suitable accelerating voltage for the 70 nm sections is 4–5 kV (Fig. 2.). The STEM detector operated with the BF + DF combination at 5 kV; however, at 4 kV, the STEM operated with the BF + HAADF combination, as it provided much better contrast and signal.
Figure 2: (a) Optimal parameters for the simultaneous imaging of the top and bottom markers using 10 and 15 nm AuNPs on both sides. Resolution: 6144x4096 px, dwell time: 10 µs, scale bar: 100 nm. Arrow colour coding: blue – 10 nm NP on top, yellow – 15 nm NP on the top, red – 10 nm NP on the bottom, green – 15 nm NP on the bottom; (b-c) software increased difference between the top and bottom 10 nm AuNPs in ImageJ. Colour coding: red arrow – bottom NP, blue arrow – top NP; (b) original BSE image. Scalebar: 250 nm resolution: 2560x1920, dwell time: 15.625 µs, working distance: 4.8 mm, 100 nm section, accelerating voltage: 5 kV. (c) ImageJ processed image – bottom NP – sphere shape, top NP – donut shape;
The detection and differentiation of four markers using just 2 AuNP sizes with the same imaging parameters (accelerating voltage corresponding to the section thickness) were also demonstrated. When 10 and 15 nm AuNPs were used as markers from both sides, we could differentiate 4 different markers, e.g., small and bright with a sharp edge (top 10 nm), small, less bright and blurry (bottom 10 nm), big and bright with a sharp edge (top 15 nm), big, less bright and blurry (bottom 15 nm) on the BSE detector at the focus point (Fig. 3). As mentioned above, the annular STEM detector was used to detect either the combination of BF (bright field) + DF (dark field) signals at higher voltages or BF + HAADF (high angle annular dark field) signals at lower voltages (Table 1).
Table 1
Best working configurations of the STEM detector for a given section thickness and accelerating voltage
Section thickness | Accelerating voltage | STEM detector settings |
70 nm | 4 kV | BF + HAADF |
70 nm | 5 kV | BF + DF |
100 nm | 7 kV | BF + DF |
100 nm | 8 kV | BF + DF |
The visual difference between the top and bottom AuNPs in the BSE image could be further enhanced with image processing software, such as FIJI/ImageJ. By adjusting the maxima and minima of the B/W threshold, the top nanoparticles changed from spherical to donut shaped, while the bottom nanoparticles remained spherical. This shape difference could be large enough to be used for the automatic registration and differentiation of the markers, which would speed up the process of marker location.