Analysis of compositional gradients in cu(In,ga)(s,se)2 solar cell absorbers using energy dispersive x-ray analysis with different acceleration energies

Figure 4 gives the typical appearance of the absorber layers in a cross section prepared by breaking. All three absorbers exhibit an average thickness of ca. 1.6 µm with a thickness varying between 1.2 and 2 µm. Even though several line scans were performed for each sample, this variation in thickness does not allow for averaging the line scans. This is a major disadvantage because a single line scan can hardly be representative for the whole sample, especially since lateral deviations in composition and pores above the back contact are present to some degree. Above the pores, the remaining absorber thickness is substantially reduced, resulting in a locally changed Ga/In gradient.

Another problem is that a flat fracture surface, as required for an ideal sample for EDX, is the exception and not the rule because different grains break differently. The low energy Se(L) signal is especially influenced by the uneven surface. Depressions emit a reduced Se (L) signal because of reabsorption at the edges while the higher energy Cu(K) and Ga(K) signals are not as much affected. Because of this, no [Ga]/([In]+[Ga]) ratios were calculated from the line scans. Nevertheless, the EDX line scans shown in Figure 5 clearly indicate that the Ga gradient is strong in sample A with high Se supply, and the Ga content is almost homogeneous for samples B and C with a reduced Se supply.

To overcome the disadvantages of line scans perpendicular to the back contact mentioned in Section 3.2, additional line scans that were ca. 6 µm long were taken parallel to the back contact at four different distances. The average composition and standard deviation (shown as shaded area) were calculated for each line scan and plotted against the distance of the line scan to the front of the absorber. The results for Ga and In are given in Figure 6.

The more lateral inhomogeneities are present along the line scan, the larger the standard deviation will be. Close to the back contact, where pores are present, the Se, Ga, and Cu standard deviations are therefore increased.

Even though this type of measurement introduces averaging, it still depends on where on the sample the operator chooses to perform the line scan. Long range inhomogeneities are not detectable. Furthermore, the result of the measurement is influenced by the number of pores present and the flatness of the surface of the cross section. Since the width of the line scan is given by D_x, which is 0.9 to 1 µm, the line scans on the 1.6 µm thick absorber layer overlap strongly.

Again, it is clearly visible from Figure 6 that sample A with the highest Se supply shows the strongest In and Ga gradients. The difference between the Ga content at the front and at the back is the largest for A, followed by B, and small for C with the lowest Se supply. The difference between samples B and C is most likely just a local coincidence since it cannot be seen in the line scans taken perpendicular to the back contact (Figure 5) or in the EDX measurements with different acceleration energies taken at the front and the back of the delaminated absorbers (Figure 7).

EDX measurements over 25 × 25 µm² were conducted at 12 different locations on both sides of each sample at energies ranging from 15 to 30 keV. The average composition and standard deviation were calculated for each energy. In Figure 7 they are shown plotted against the average depth of maximum excitation R_m for this energy. The connecting lines between individual points serve as guide for the eye only. The average R_m for the chosen lines is 0.3 µm at 15 keV, 0.5 µm at 20 keV, 0.8 µm at 25 keV, and 1.1 µm at 30 keV (see Table 3). Positive values of R_m indicate that the measurement was performed from the front side, while negative values of R_m indicate that the measurement was performed from the back side. For the same absolute value R_m at the front and back, the location of the X-ray production volume in the sample is different.

The EDX measurements performed with different acceleration energies exhibit a very small standard deviation between different locations, indicating that the long-range homogeneity is very good. Samples B and C show no significant differences in their Ga and In profiles, even though they were processed with different Se supplies. Only the initial reduction of the Se supply from sample A to B has a great impact on the Ga/In gradients. Sample A shows the well-known strong enrichment of Ga close to the back contact corresponding with an In-depletion, while samples B and C exhibit an almost homogeneous Ga and In content.

From the results shown in Figure 7, the [Ga]/([In] + [Ga]) ratios were calculated and plotted against the average maximum of X-ray production R_m. In Figure 8 they are compared to GDOES measurements on samples from the same production run. The depth resolution of GDOES measurements is determined by the curvature of the bottom of the sputtered crater and sputter induced roughening of the samples [8] and is typically 5 to 10% of the layer thickness [4].

The connecting lines between individual [Ga]/([In] + [Ga]) ratios calculated from EDX with different acceleration energies serve as guide for the eye only. EDX shows the same trend as GDOES does for all samples with the exception of sample B close to the surface. Sample A has a steep Ga/In gradient with a very low Ga content at the front and strong Ga enrichment at the back (see Table 4 and Table 5). Sample C with the lowest Se supply has a high GGI close to the front which rises slightly and almost linearly to the back. The same applies for sample B outside a 300 nm wide surface region where a reduced GGI is seen only in GDOES. This was further analyzed with additional EDX measurements at 10 kV and is discussed below. Figure 8 shows that both methods depict the progression of the GGI between surface and backside very similarly for sample C and sample B away from the surface. Sample A with its strong Ga/In gradient shows a different course of the curve in GDOES and EDX. GDOES with its better depth resolution measured an S-shaped increase of the GGI with a convex curvature in the upper part and a concave curvature in the lower part of the absorber and an inflection point in between. EDX with different acceleration energies is not able to reproduce this S shape because it measures the average composition inside a volume that increases with increasing depth.