Cause for soft breakdown

These findings suggest that probably only one specific property of electrically active de-fects allows for current conducting channels under reverse bias conditions, excluding BD in recombination active defects not meeting the specific property.

A qualitative explanation of some of these observations can be made based on a model that Kveder et al. [188, 189] employed to explain different experimental observations related to recombination at clean and decorated dislocation networks. Following their argumentation, dislocation luminescence is generated by radiative recombination be-tween energy levels (or bands) E_DL,C and E_DL,V which are highly localized and situated close to the conduction band and the valence band, respectively, both about 0.80 eV apart.

In this picture, strong and weak dislocation luminescence (anti-correlating with the soft breakdown intensity) corresponds to a high and low transition probability between EDL,C

and EDL,V, respectively. The transition can for example be suppressed in the presence of additional trap levels within the band gap [189].

In literature, non-radiative carrier recombination at crystal defects is therefore attributed to metallic impurities which are either bound to the dislocation core, are stuck at grain boundaries or accumulated in “impurity clouds” around dislocations [190]. In addition, metals tend to form precipitates, see also chapter 5, which also can be highly electrically active [191].

If soft breakdown were caused by an accrual of single metal atoms at dislocation cores or by the surrounding atom clouds, the question arises why the reverse current flows selec-tively through particular BD channels. One would expect that the atoms are well dis-persed along the crystal defects due to simple thermodynamics, thus the differences be-tween the defect configurations should be very small and BD should occur in more sites.

Therefore, it is more likely that soft BD is related to metal precipitates of a sufficient size, since their density is relatively small and each local configuration is unique depending on the impurity element of the cluster, the precipitate size and position relative to the space charge region etc.

While the electrical activity of the system of impurity clouds within the silicon can be de-scribed by the Shockley-Read-Hall formula of single energy states around the center of the silicon band gap (see chapter 3.2.3), thus leading to a high recombination activity due to their large total electrically active surface, the system of a metal precipitate in silicon is a completely different issue. As soon as the precipitate contains a critical num-ber of metal atoms, it possesses a metal Fermi energy and its behavior turns metal-like, which has significant implications for the physical properties which are discussed in sec-tion 6.5.8.

In literature, breakdown has already been attributed to metal precipitates a long time ago. Goetzberger et al. [167] for example examined the diode behavior of simple silicon pn-junctions, some of which were intentionally contaminated with Cu. The latter tended to show a rather soft reverse I-V characteristic. With the help of a potential mapping technique, they found that the applied reverse voltage was distributed inhomogeneously across the junction. The lowest voltage, signifying the BD site, was concentrated in a small spot which the authors assumed to be a Cu precipitate in the space charge region without providing further proof. In another study by Katz and Cullis [192, 193], the poor reverse characteristics of a silicon p⁺n junction was attributed to rod-like FeSi₂ precipi-tates.

To further substantiate the above hypotheses and in order to connect soft breakdown sites with the presence of metal precipitates, the following experiments were carried out [194]:

A damage-etched standard industrial solar cell made from standard feedstock material was characterized via bias-dependent EL measurements. In forward bias, it shows a typi-cal distribution of recombination active regions, while soft BD light emission is detected in reverse bias at many singular spots, correlating to the recombination active defects. Due to the damage etch no hard breakdown (type III) is observed.

For the microscopic investigations, a sample of approx. 10 x 20 mm² was cut out contain-ing many soft breakdown sites. Figure 6.19 (a) displays the EL measurement in forward bias (+600 mV) of a part of the sample in which recombination active defects appear as dark regions. At -10 V reverse bias, soft BD light is emitted from the spots marked by the blue circles. Some breakdown sites are located along grain boundaries, which are charac-terized by dark continuous lines in the EL image, while others are found in areas of rather homogeneously decreased band-to-band luminescence which is typical for high disloca-tion density regions. Similar regions of breakdown sites correlating with recombinadisloca-tion active defects are found in all standard mc-Si solar cells, therefore this sample represents soft breakdown regions well. However, due to the complexity of the microscopic meas-urements and the restricted experiment time at the synchrotron facility, the following study could be performed only on two exemplary breakdown sites.

Figure 6.19: (a) Forward EL image (600 mV) of the solar cell piece containing many re-combination active defects. Soft BD light emission is detected at the sites marked by the blue circles. (b) SEM image of the grain indicated by the dashed rectangle in (a). BD light emission (see insets; the white bars correspond to 10 µm) is located at the sites 1 and 2 along two grain boundaries.

In order to locate the breakdown sites with a high precision, microscopic investigations were carried out with the EL- / PL-spectroscopy mapping tool⁵¹. The two breakdown sites found in the macroscopic EL images are located at two different boundaries of one grain (marked by the white circles in the SEM image), see insets in Figure 6.19 (b). Character-istic features of the wafer surface around these coordinates, found in the microscopic images (resolution ~1 µm), are taken as a marker.

Micro-X-ray fluorescence (µ-XRF) measurements of these two breakdown sites were per-formed at the beamline ID22NI at the European Synchrotron Radiation Facility (ESRF).

The nano-imaging stage features a spot size of ca. 100 x 100 nm together with a high photon flux (~10¹² ph/sec between 6.5 and 18 keV). The combination of small spot size and high photon flux allows for the detection of transition metal precipitates with a di-ameter in the order of some ten nanometers [121] when the clusters are located within some ten micrometers beneath the wafer surface.

Figure 6.20: Synchrotron µ-XRF measurements (ESRF, ID22NI) of the Fe Kα-line (a) at BD site 1 in Figure 6.19 and (b) at BD site 2.

51 Measurement performed in collaboration with P. Gundel.

Using the markers determined in the EL maps, the µ-XRF mappings were carried out in an area of 20 x 20 µm² centered at the pre-breakdown spots. The results are shown in Figure 6.20.

Iron precipitate colonies are detected at both pre-breakdown sites. They are distributed along single lines which correspond to the grain boundaries. In region 1, also one copper precipitate is found. The distance between this copper cluster, however, and the position of the pre-breakdown is determined to be approx. 5 µm. The location of the iron precipi-tate colonies, on the other hand, coincides well with the breakdown spots taking into ac-count the limits of the spatial resolution provided by the EL- / PL-mapping tool (1 µm) and the accuracy of the positioning of the X-ray spot on the sample, which is estimated to be around 1 to 2 µm.

In addition, several µ-XRF mappings distant to the pre-breakdown sites along the grain boundary labeled A in Figure 6.19 were performed. However, in spite of the high recom-bination activity of this grain boundary no other precipitates within the detection limit of around 30 nm in diameter are discovered, underlining the evident concurrence of Fe-containing large clusters and pre-breakdown spots.

This result is further maintained by investigations performed at MPI Halle [195]. By com-bining high resolution EL microscopy with FIB, TEM and EDX, metal containing precipi-tates of various chemical compositions were found at type II BD sites. Besides Fe, Cu and Mn oxides were detected.

In summary, there is strong evidence that metal containing precipitates cause soft BD.

Since metal precipitates are highly recombination active, the dislocation luminescence is suppressed at soft breakdown sites, explaining previous findings.