presents the PC-AFM topography and current maps for all samples, along with the surface topography measured by PF-AFM. PC-AFM without illumination resulted in nearly no measurable current flow, so all PC-AFM data in this figure were obtained under illumination. In addition, topography from PC-AFM shows low spatial resolution since contact mode with a higher radius tip was required; PF-AFM was therefore used to obtain higher quality topography images.
Focussing on the PC-AFM topography of the In0.12
N film (Figure 1
c), hexagonal V-pits and trench defects are visible, along with elevated ridges around both types of defect (example indicated by the red arrow). PF-AFM confirms these observations (Figure 1
k). The raised plateau (blue arrow) corresponds to a layer formed by a previous PC-AFM scan over this smaller area; comparing with the relevant current map in Figure 1
g, the layer appears to be insulating. This suggests that oxidation is still occurring at the surface of InGaN as it did for pure GaN—such a layer formed on all InGaN samples during PC-AFM scans. Also noticeable is high current flow around the edges of defects; however, this is very similar to the effect observed by Kim et al. [18
] which was likely due to tip-sample contact area changes, and as such should be treated with suspicion. A particularly noteworthy feature is the current drop which occurs around each defect (e.g., see red arrow), corresponding very well to the raised ridges in the topography. The ridges exhibit plateaus up to several 10s of nm wide, and across these flat raised regions the tip-sample contact area is expected to be constant. Figure 2
presents height and photocurrent cross sections of the insulating layer, along with average radial height/current cross sections of a ridge around a typical V-pit. The positions from which the layer and ridge cross sections were taken are indicated in Figure 1
c,g by blue dashed lines and red circles respectively.
PC-AFM results gained from the In0.05
N and In0.09
N samples (Figure 1
a,b,e,f) are similar to those from the In0.12
N film; however, the ridges around the V-pits cannot be made out in the PC-AFM topography. On the other hand, they are visible in the higher spacial resolution PF-AFM images (Figure 1
i,j), and the corresponding photocurrent maps again indicate reduced current flow in these regions. (It is worth noting that the streaks after V-pits in Figure 1
b are artefactual, arising from feedback gain being too high during image acquisition).
Given that the photocurrent drops cannot be directly correlated with the ridges in this case, it was necessary to analyse the measured photocurrent drop width distribution obtained from PC-AFM and compare it to the ridge width distribution from PF-AFM topography. These distributions were found by taking the average radial height cross section around 54 pits in PF-AFM, and the average radial current cross section of 32 dislocations in the PC-AFM current results.
The measured distributions from this analysis are presented in Table 2
; for both samples, the width distributions of the ridges and the current drops match well. This strongly suggests that the photocurrent drops still correspond to the ridges around each defect for the In0.05
N and In0.09
N samples. PF-AFM was also used to extract average ridge height from 180 V-pits across all three low indium content samples, with the result of 3.33 ± 0.05 Å—i.e., on the order of one atomic monolayer.
Finally, it’s also worth noting the lower current measured in regions with no pits in both Figure 1
e,f. We speculate that this is due to the tip becoming contaminated and leading to a contact resistance increase, before sudden height changes at a V-pit place more force on the tip and clear some of the contamination, causing a current increase on the first line after encountering a V-pit.
Results from the In0.15
N sample (Figure 1
d,h,l) make it clear that this film is substantially different from those previously analysed. The surface is significantly rougher, with noticeable hills and valleys over a 12 nm height scale. Ridges are not visible in either the PC-AFM or PF-AFM topography, and no photocurrent drops are observed around the pits in the current map. The current map also shows connected regions of low current which correspond very well to the valleys observed in the topography. Current readings within the pits are different from before, with one side of the pit exhibiting higher current than the other side. This is likely due to the feedback system not keeping up with rapid height changes, leading to lower contact force on one side (and hence lower current) and higher contact force on the other side (increased current). This effect may be present only in the In0.15
N film results due to different feedback gains or a differently shaped tip compared to similar measurements on the other samples. It is worth noting that to facilitate observation of the features associated with the valleys between the dislocations, Figure 1
d,h are presented at a different lateral scale to the other parts of the figure.
I-V curves obtained away from any defects exhibit diode-like behaviour, likely because the tip forms a rectifying contact with the surface. They are also quite noisy (see Figure 3
a inset), so analysing them individually would lead to significant error. Instead, an effective method is to take ten I-V curves in the region of interest with the same sampling rate and voltage range, and then average the measured current at each voltage point over all the I-V curves. As such the standard error in each current value can also be found, which allows for more meaningful comparison of current-voltage behaviour on different samples with/without illumination. The linear portion of the curve can then be fitted with a straight line, and extrapolated back to obtain a “turn-on” voltage,
; the gradient of this line also gives an indication of the conductance of the system, G
. All of the above leads to quite a cluttered I-V plot, so to simplify interpretation error bars are represented by lines; the final result can be seen in Figure 3
provide a means of quantitatively comparing the current-voltage characteristics of different samples away from any V-pit ridges, as shown in Figure 3
All films were analysed in one session with the same AFM tip, in the testing order In0.12
N → In0.09
N → In0.05
N → In0.12
N → In0.15
N → In0.12
N. The In0.12
N film was repeated throughout to track the tip status; any change in
for this sample between the different measurements would indicate a change in the tip. The inset of Figure 3
b shows how
for the In0.12
N film varied over the course of the session, with the error bars corresponding to the standard error of extrapolating the linear fit to
in each case. It can be seen there is no overall drift in
as the test is repeated, but the standard error of extrapolation is much lower than the change in
each time, which indicates that tip changes dominate the error in these measurements. As such the standard deviation of these three measurements is used to roughly gauge the error in all the measurements, and is plotted as error bars in the main plot of Figure 3
b. The value for the In0.12
N film in the main plot is the average of the three repeated measurements.
was always much lower when light was incident on the film; this is expected, since incoming photons will generate electron-hole pairs in the InGaN, increasing carrier density at the Schottky contact with the PC-AFM tip and thereby reducing the Schottky barrier height. Hence the applied voltage required to overcome the Schottky barrier () is lower under illumination. The trend between different samples is less clear; generally, seems to decrease with increasing In content, which could be expected from the corresponding decrease in bandgap.
c shows data for conductance, G
, obtained from the same experiment; in this case the variance of the repeat measurements is large (inset), so no comparison can clearly be made between samples. Carrying out more repeats would be difficult due to the trade-off between identifying tip changes and further damaging the tip to do so. Interestingly, G
was always lower when the light was on, which is unexpected. This could be explained by the insulating layer that builds up on the surface once current starts to flow; higher initial current may lead to more insulating layer being generated, which would reduce the measured conductance. Conversely,
should be unaffected since no current has been generated at this point in the I-V curve and an electrochemical reaction should not be able to occur without the movement of charge.
Measuring I-V curves also provides a way to quantify nanoscale features such as the defect ridges, since the location at which the I-V curve is taken can be specified. A key consideration is that the ridges are not much larger than the tip radius, so any drift in the sample after the initial topography map is taken could lead to the tip missing the ridge and mistakenly taking an I-V curve of background material. Figure 4
a,b illustrate how the formation of insulating layer can actually be helpful in identifying when this occurs, allowing for the resultant curves to be excluded from the analysis. The topography map taken after I-V measurements (with the light on) in Figure 4
a shows that if the tip misses the ridge, the insulating layer also forms away from the ridge (pink circle). On the other hand, if the tip correctly contacts the ridge, we see insulating layer there (green circle). Figure 4
b shows the individual I-V curves obtained from the pink and green circle measurements—there is a large difference between the two. Hence the “pink” result should be excluded from our ridge current-voltage analysis. Figure 4
a was acquired after extensive use of the AFM tip, so the light “speckles” over the surface can be attributed to tip damage. However, these artefactual speckles must not be confused with insulating regions which were formed in regular arrays on the surface by I-V curve measurements (e.g., red arrows).
c presents the average I-V curves obtained with the light on and off, away from defect ridges and on them, using this method on the In0.12
N layer. Measured turn-on voltage was much higher on the ridges; the increases were
V with the light on and off respectively. Interestingly, conductance was also higher on the ridges once current did start flowing, with an increase of
) for the light on and off; but as previously stated, conductance measurements are made unreliable due to the insulating layer.
presents TEM results for the In0.09
N and In0.15
N layers, focussing on a single V-pit cross-section in each. Comparison between these two samples is particularly apt since In0.09
N is representative of the samples with ridges around pits and photocurrent drops (low indium content samples), while In0.15
N had no ridges or drops.
Looking at the annular dark field (ADF) scanning TEM (STEM) image of the In0.09
N film in Figure 5
a, there is clear contrast between the pure GaN buffer layer and the InGaN thick layer, with the GaN appearing darker due to its lower average atomic mass. Of most interest is the faintly visible darker region around the V-pit extending down to the GaN interface, which suggests lower indium concentration in this area. This is confirmed by the STEM energy-dispersive X-ray spectroscopy (EDS) map of indium concentration (Figure 5
d), once the drift in the specimen has been accounted for. Figure 5
b,c present bright field TEM images with g
= 0002 and
respectively, and demonstrate that the only dislocation present is the threading dislocation at the V-pit apex. Faint contrast is also visible at the boundary of the indium-deficient region.
e–h present results from the same methods applied to the In0.15
N V-pit. Focussing on Figure 5
e, ADF-STEM no longer shows clear contrast between the GaN buffer layer and the InGaN, suggesting strain is affecting the contrast. More interestingly, the bright field TEM images suggest there are multiple dislocations branching out around the V-pit. STEM-EDS shows no reduction in indium concentration around the V-pit, and perhaps even indicates a slightly increased concentration close to the facets.
The STEM-EDS maps provide indium content as atomic percent rather than as x
N. Hence, the values on the scale bars in Figure 5
should be doubled to allow comparison with the XRD results in Table 1
. Such a comparison, however, suggests a lower average indium content in the STEM-EDS measurement than in XRD. This is likely due to a systematic error in the EDS quantification method, but does not effect the above observations concerning relative indium contents.