RTP and CFP P-Al co-diffused solar cells from neighboring wafers

For EFG, no neighboring wafers exist like in the case of block-cast mc-Si. Each wafer is unique. So, if two processes are to be compared with their influence on the material properties, one either has to process large numbers of wafers to get a statistically meaningful result or has to use EFG wafers that are vertically adjacent with respect to the crystal growth. In the latter case, the initial properties of a transition region of a few centimeters are at least comparable.

This offers the possibility to compare different processes, e.g. CFP with RTP, by subjecting one wafer to CFP and the other to RTP. Evaluation of the transition region by the use of mapping techniques yields the desired information. This was done in the following experiment for P-Al co-diffused solar cells.

Experimental

For the experiment, vertically adjacent state-of-the-art EFG wafers of 5 5 cm²were prepared by laser cutting of a standard 10 10 cm²wafers. After a RCA surface clean and deposition of 1 m of Al on the back, one of the 5 5 cm wafers was P-Al co-diffused by RTP (1000 C, 25 s) and the adjacent one by CFP (820 C, 1 h). In both cases, the target sheet resistance was 80 /sq. After P-glass removal in HF the wafers were cut into 2.5 2.5 cm² pieces.

After this cut four stripes of wafers were available with each stripe containing four adjacent 2.5 2.5 cm² pieces of which the upper two were diffused by RTP and the lower two by CFP. Then, the front contacts were prepared by photolithography, evaporation of Ti/Pd/Ag and subsequent Ag-plating. Finally, the wafer edges were isolated by laser cutting, and a forming gas anneal step at 350 C for 30 min was applied.

Results and solar cell analysis

Several of the 2.5 2.5 cm²pieces broke during or after laser cutting, which indicates the high internal stresses introduced during growth of the EFG crystal. Nevertheless, for each of the four stripes, at least one RTP and one CFP processed solar cell remained undamaged. Table 4.11 contains the IV parameters under illumination of these solar cells.

Many of the solar cells suffer from poor fill factors, in particular the CFP solar cells. It is not clear to what extent this can be attributed to technological problems (e.g. edge isolation) or to CFP-induced changes in the material properties (e.g. poor second diode quality due to creation of defects in the space charge region). This has to be taken into account when comparing the conversion efficiencies of the CFP solar cells with those of the RTP solar cells. However, it comes out clearly that the RTP solar cells exhibit significantly higher and than the respective adjacent CFP cells. This observation fully agrees with the experimental results presented for the cells made of non-consecutive sections of the ribbon. Remarkably, the best RTP cell has an efficiency of 8.9 % without hydrogen passivation.

As already mentioned, slight differences in the emitter or the back-surface-field properties cannot explain the differences in and because solar cells from FZ reference samples subjected to the same RTP and CFP processes, respectively, agree well in their IV parameters.

Tab. 4.11: Parameters of solar cells fabricated from adjacent pieces of state-of-the-art EFG. The P-Al co-diffusion was carried out either by RTP or by CFP. Cell size is 2 2 cm .

Stripe Cell P-Al " " (

# # co-diffusion [mV] [mAcm^-2] [%] [%]

I 1 RTP 541 21.1 74.7 8.5

I 3 CFP 485 16.4 68.7 5.5

II 1 RTP 554 22.0 73.0 8.9

II 3 CFP 532 20.2 71.3 7.6

III 1 RTP 541 21.1 64.0 7.3

III 4 CFP 516 18.5 75.3 7.2

IV 1 RTP 529 20.4 74.9 8.1

IV 4 CFP 496 16.4 74.8 6.1

Rather, the cause must be higher carrier lifetimes (diffusion lengths) within the EFG bulk after RTP than after CFP P-Al co-diffusion, respectively. In order to substantiate this supposition, local SR-LBIC measurements were carried out and mappings of the effective diffusion length

# were deduced. For instance, Fig. 4.28 shows the # mappings of the RTP cell III-1 and of its CFP counterpart III-4. Please note that the solar cells are made from neighboring sections of the same 10 10 cm² wafer. Both cells contain a lot of thin stripes where a high density of twin grain boundaries prevails. These stripes are known to represent the areas of the highest diffusion length in the as-grown stage. Therefore, these cells are made of wafer sections representing the good areas of EFG. Obviously, they exhibit significantly higher # values after RTP than after CFP. The # distribution, also shown in Fig. 4.28, fully supports this statement. It is stretched towards# values well in the 200 m range in case of RTP whereas

# of the CFP cell does not exceed 130 m. Since different back surface recombination velocities cannot possibly account for # differences at such a level, this is a clear indication of a higher bulk diffusion length after RTP. Hence the improved and of the RTP cell compared to the CFP cell. It is difficult to see whether there is a difference in the intra-stripe regions where no twin grain boundaries exist or their density is rather low. There are definitely regions which show the same performance after RTP as after CFP. Interestingly, these seem to be the regions which exhibit low# of 40 to 50 m at the maximum. One might speculate that these regions are all of the same crystal nature and contain defects similar to those observed in the regions which were shown earlier not to respond to hydrogen passivation. Obviously, they also do not benefit from the RTP P-Al co-diffusion process. It seems appropriate to say that these areas are the problematic areas of the EFG material.

EFG ribbons do not always feature stripes of twin grain boundary lamellas. Instead, they show regions containing other sorts of grain boundaries, high dislocation densities or sometime inclusions or even cracks. In general, EFG of this kind features low initial bulk diffusion lengths.

For example, the solar cells IV-1 and IV-4 from Tab. 4.11 are made of wafer sections featuring almost no twins. As can be seen in the# mappings in Fig. 4.29, the RTP solar cell is superior to the CFP solar cell made of the consecutive section. In the very left part of the CFP cell,#

0 >180 CFPIII-4

RTPIII-1

0 50 100 150 200 250

2 4

6 CFP

RTP

Counts [x100]

L_eff [µm]

Fig. 4.28: Mapping and distribution of for 2 2 cm solar cells from vertically adjacent EFG wafers cut out of crystall region showing numerous stripes of twin grain boundary lamellas. The upper wafer was P-Al co-diffused by RTP and the lower one by CFP.

0 >120

CFPIV-4 RTPIV-1

0 50 100 150 200 250

2 4 6 8 10

CFP

RTP

Counts [x100]

L_eff [µm]

Fig. 4.29: Mapping and distribution of for 2 2 cm² solar cells from vertically adjacent EFG wafers cut out of crystal region exhibiting almost no twin grain boundaries. The upper wafer was P-Al co-diffused by RTP and the lower one by CFP.

is in the range of 30 to 90 m whereas in the RTP processed counterpart it exceeds 100 m and reaches up to 175 m. The right half of the CFP cell is of extremely low quality with

# below 30 m. At first look, it seems that there are many of those already mentioned regions which, for example, do not even respond to hydrogen passivation. However, the RTP counterpart proves that most of these poor regions can be improved up to 60 m by the RTP P-Al co-diffusion process. Only a small proportion of the area remains as bad as in case of the CFP cell. Those are supposedly the EFG regions which hardly respond to any standard gettering or hydrogen passivation due to the specific defects present. Very likely, the poor performance regions of EFG feature high dislocation densities. However, this characteristic alone cannot explain the low performance. They have to be decorated by certain impurities (e.g. metals) which can hardly be removed or inactivated during standard solar cell processes. Future work should concentrate on these poor performance regions as they surely limit solar cell efficiency.