Typical current-voltage (I-V) characteristics in Schottky junctions of the two different types measured at 297 K are shown in Figure 1a, and those measured at 77 K are shown in Figure 1b. The series resistance dominates the forward, and leakage resistance dominates the reverse I-V characteristics in samples where the NCs migrated near the silicon surface. The typical leakage resistance is about 1 kΩ at 297 K and increases to 56 kΩ at 77 K. The cited resistance values are not related to the figures. The figures demonstrate the different types of I-Vs. The leakage resistance is thermally activated, indicating that the Fermi level in the cap is pinned by point defects at about 160 meV from the conduction band. The thermal activation of the leakage resistance was evaluated by linear fits to the reverse I-V and by plotting the fitted resistance values as a function of reciprocal temperature. The capacitance-voltage (C-V) characteristics of the junctions measured at room temperature are shown in Figure 1c, and those measured at 77 K are shown in Figure 1d. Schottky junction capacitance 260 pF--indicated as a line in Figure 1c--corresponds to 50-nm depleted layer thickness, equal to the nominal cap thickness. Below -1 V reverse bias at low temperature, the doping calculated from the 1/C ²-voltage plot is appropriate for the semiconductor substrate in both types of samples. The doping concentration profile was calculated from the C-V characteristics measured at different temperatures (not shown in the figures). The calculated doping profiles in the two type of samples are shown in Figure 2. The samples with NCs migrating near the surface show high concentration of donors, while in samples with NCs remaining near 50-nm deposition depth, the donor concentration is low.

The DLTS spectra were measured at -1 V reverse bias, and 20 μs, 0 V filling pulses. The energy position of the deep level calculated from the DLTS Arrhenius plot is about 0.25 eV, appropriate for the Cr level at E _c--0.27 eV in n-type silicon 9 . The large concentration of doping depicted in Figure 2 in samples where the NCs migrated near the surface is explained by large concentration of Cr-related point defects in the cap. In the samples with NCs near the deposition depth, the low donor concentration depicted in Figure 2 is explained by the low concentration of Cr-related deep-level defects. The markedly different concentrations of Cr-related point defect in the two types of samples indicate that these defects may be related to the observed migration of the NCs during the cap growth. To enable us understand the role of the Cr-related defects in migration of NCs, we require further experiments.

Tapping-mode AFM amplitude and phase images of the samples with NCs near the surface are shown in Figure 3a,b, respectively. The tapping-mode AFM amplitude (Figure 3a) is not sensitive to the CrSi₂ NCs. Several bigger NCs appear in the phase image (Figure 3b). We suppose that it is due to the electrical interaction of NCs with the vibrating scanning tip. The phase of the vibration changes, but does not cause energy dissipation, interpreted as height in the amplitude image. Some spherical protrusions appear with a diameter of about 90 nm and a height of 12 nm in the amplitude. The morphology measured in contact-mode does not show these large protrusions. The difference can be an effect of the pressure of the tip in contact-mode and the possible wear of the sample, since repeated scans over the imaged areas has shown visible degradation of the surface. However, we suppose that these protrusions are mainly due to areas with large NC density in the cap, resulting in virtual height increase in tapping-mode amplitude image.

In samples with NCs 50 nm deep below the surface, the morphology and the phase of the tapping-mode AFM of the silicon surface measured are shown in Figure 4a,b, respectively. The NCs are hardly visible in both amplitude and phase images; the interaction of the vibrating tip with NCs embedded 50 nm deep in silicon is weak.

The sample with NCs close to the surface exhibited large leakage at room temperature in macroscopic junctions. To understand the reason of leakage this sample was analysed by conductive AFM. The conductive tip is scanned on the surface, and the current at a given voltage is recorded and mapped. Conductive AFM reveals the local conductivity differences in the vicinity of NCs. Across most of the surface, the current was nearly constant and even independent of the polarity of the bias. Locally, evidence of rectification could be observed. The tip-wafer junction can be easily degraded by the local current load, and so the reliability of repeated measurements using other bias conditions is questionable. The results show that primarily the large concentration of Cr-related point defect in this sample is responsible for the leakage. The large local electric field around the NCs may also act as local short-circuit path; however, this kind of leakage was not strongly dependent on temperature, as observed in large-area Schottky junctions.

The contact-mode AFM amplitude and SCM images recorded simultaneously in sample with redistributed NCs are shown in Figure 5a,b, respectively. The SCM shows definitely better contrast and spatial resolution than AFM, indicating that the detection of NCs is improved when electrical interaction is involved in the image. The size of the observed objects is appropriate for NC sizes seen earlier in XTEM and in AFM images 6 8 .

The contact-mode AFM and SCM images of the sample with NCs at 50-nm depth are shown in Figure 6a,b, respectively. The NCs are hardly visible in AFM, as the sample surface is flat. The NCs are apparent in SCM. The higher conductivity of inclusions increases the locally sensed capacitance, and thus, the difference of electrical properties of silicon and CrSi₂ gives a better contrast for the detection of the NCs by scanning tip capacitance sensing. NCs deep below the surface are revealed in SCM images, and are not shown in the morphology measured simultaneously. Deep NC features are generally expected to appear somewhat smeared 10 . A cross section of the SCM image across a NC is shown in Figure 7. The half-width of the peak agrees with the size of the NCs. It shows that the interaction of the charged NC and the measuring tip is strong, which controls the transport, and we assume that the measured capacitance is dominated by the NC-host junction. The measurements indicate that the NCs embedded deep--having electrical characteristics different from the defect-free host--can be detected using the SCM with high resolution.