Comparison of Process I and Process II

For a comparison of the two processes, solar cells were manufactured from neighbouring Baysix mc-Si according to the sequences given in Figure 2.2 (Process I) and Figure 2.20 (Process II) without mechanical V-texturing. The solar cell manufactured by Process I led to Voc=605 mV and Jsc=32.1 mA/cm² compared to Voc=615 mV and Jsc=33.0 mA/cm² for Process II (cell area 24 cm²). For a detailed analysis about the origin of these improvements, local LBIC and reflectivity measurements were carried out. The results are depicted as an Leff-mapping in Figure 2.21. The mean value of Leff increased from 181 µm to 220 µm.

The mapping shows the two beneficial effects of Process II. These are:

• Improvements in back surface passivation by a reduced SB in Process II

• Improved hydrogen passivation in Process II due to simplified in-diffusion through the rear side

Figure 2.21: Spatially resolved mappings of the effective diffusion length Leff obtained from SR-LBIC measurements. The left mapping shows a BCSC manufactured by Process I, the right by Process II.

A histogram of Leff for the two solar cells is given in Figure 2.22. Two effects can be observed. For the peak values corresponding to the grains with a high diffusion length, Leff

is 222 µm for Process I and 282 µm for Process II. The corresponding bulk diffusion lengths according to equation 2.2 are LB=241 µm taking SB=4000 cm/s in Process I and LB=271 µm taking SB=700 cm/s in Process II. Hence the enhanced bulk passivation due to facilitated hydrogen in-diffusion through the rear surface improved LB in the “good” grains by 30 µm.

Also, the relative frequency of diffusion lengths below 150 µm is reduced for Process II. In this case, the improved rear surface passivation has no influence of Leff, and therefore this is also caused by improved hydrogenation. In principle, the Al/Si eutectic can also be removed in Process I to facilitate hydrogen in-diffusion. However, this process step was not considered for this process in order to maintain its simplicity, also with regard to a possible transfer of this process in industrial production lines.

50 100 150 200 250 300 350 0

5 10 15 20 25 30 35

Process I Process II

counts [a.u.]

L_eff [µm]

Figure 2.22: Histogram of L_eff obtained by SR-LBIC-mappings for solar cells manufactured by Process I and Process II.

2.8.4 Determination of optimum process temperature for MIR hydrogen passivation in Process II

Experiment

The experiment was carried out on Baysix from Bayer and on Eurosil from Eurosolare (wafer size 12.5x12.5 cm², w=350 µm, ρ=1.8 Ωcm). For Baysix (wafer size 12.5x12.5 cm², w=330-340 µm, ρ=1 Ωcm) neighbouring wafers were taken which were also used for the determination of the optimum process temperature and duration of MIR hydrogen passivation within Process I (see section 2.5.5). Again the wafers were cut to a wafer size of 5x5 cm² before MIR hydrogen passivation leading to four solar cells with an area of 24 cm² after edge isolation. As for the investigations in Process I the temperatures for the MIR hydrogen passivation were varied between 350 °C and 500 °C with the process time fixed at 120 min.

Results Baysix

The results of the illuminated IV-measurement are illustrated in Figure 2.23. The cells were measured untabbed with one probe per busbar explaining the rather moderate fill factor. For the average efficiency the following can be concluded:

• MIR hydrogen passivation led to improvements in Voc, Jsc, FF and η independent of the applied process temperature between 350 °C and 500 °C

• Optimum process temperature is 450 °C

• Increase at 450 °C amounts to ∆Voc=13 mV (2.2% rel.), ∆Jsc=1.0 mA/cm² (3.1% rel.),

∆FF=1.1%abs. (1.5% rel.) and ∆η=1.0%abs. (7.0% rel.)

no 350°C 400°C 450°C 500°C

Figure 2.23: Influence of process temperature for MIR hydrogen passivation in Process II on Baysix mc-Si.

Discussion

The increase in Voc and Jsc was significantly higher in Process II compared to Process I on neighbouring wafers. This further indicates that hydrogen passivation is facilitated in Process II due to the removal of the rear contact. Whereas in Process I, process temperatures between 400 °C and 500 °C led to an almost identical η, in Process II the temperature of 450 °C reached the best results. The decrease observed at 500 °C is discussed in the next section.

For a more detailed characterisation of hydrogen passivation, LBIC measurements were carried out with a high resolution of 25 µm. The scans were done on a size of 2x2 cm² for an unpassivated cell and for MIR hydrogenation at 450 °C. The mappings of the IQE at λ=980 nm are given in Figure 2.24 for both solar cells, the histogram of the IQE for the two cells is shown in Figure 2.25.

Figure 2.24: Spatially resolved measurements of the Internal Quantum Efficiency (IQE) at λ=980 nm determined by local LBIC and reflectivity measurements. (left) Unpassivated and (right) MIR hydrogen passivation at 450 °C for 120 min. A picture of the solar cell is shown in the middle indicating the crystal grain structure.

The LBIC-mapping shows that regions with an IQE below 0.4, corresponding to low lifetime regions in the vicinity of crystal defects, are reduced for the passivated cell compared to the unpassivated one. Hence, the recombination activity of crystal defects is lowered. From the LBIC-mappings it can also be concluded that hydrogen diffusion through the contact groove is of minor importance. If a strong effect of in-diffusion through the contact grooves would occur, one expects, that the recombination of grain boundaries in the vicinity of the grooves is significantly reduced. However, this effect can not be observed.

Besides the passivation of the low lifetime regions, also the peak value of the IQE corresponding to high lifetime regions is reduced for the passivated cell. One possible explanation could be an inhomogenous Al-alloying with a higher SB. This seems rather unlikely, since the same process parameters for screen printing and firing were applied.

Fischer observed [Fis02c] that τB is reduced, if a test sample with screen printed Al-BSF and removed Al rear contact is subjected to an additional treatment at 450 °C for 1 h. He suggests, that fast diffusing metallic impurities are inserted into the wafer and/or into the highly doped p⁺-region during contact firing which diffuse deeply into the wafer during subsequent annealing steps, even at moderate process temperatures of 450 °C. This effect could also explain the lower gain for hydrogen passivation at a process temperature of 500 °C.

However, still an average gain in η of 1%abs. was achieved due to hydrogen passivation at 450 °C. The lifetime reduction in the good grains due to metallic poisoning can be reduced by lowering the passivation time. A time of 90 min was sufficient for Process I, and a further reduction seems feasible for Process II due to the facilitated diffusion of hydrogen through the rear surface.

A picture taken from the corresponding section of the LBIC mappings shows a large quantity of crystal defects, whereas in the LBIC-scans only a reduced number is visible.

This indicates that crystal defects have a different recombination strength, depending on the decoration with impurities like oxygen and metals [Law00].

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0

100 200 300 400 500

with MIRHP hydrogenation without hydrogenation

counts [a.u]

IQE

Figure 2.25: Histogram of the IQE from the mappings in Figure 2.24.

Eurosil

The results of the illuminated IV-measurements on Eurosil are illustrated in Figure 2.26.

For the average values the following can be observed:

• Optimum process temperature is 450 °C. At this temperature, the best values of Voc, Jsc

and η are obtained.

• The gain at this temperature is around ∆Voc=6 mV (1.0% rel.), ∆Jsc=0.4 mA/cm² (1.2% rel.), and ∆η=0.4%abs. (2.8% rel.).

The increase in η due to hydrogenation was lower for Eurosil than for Baysix. Also Spiegel observed, that the efficiency of MIR hydrogenation is reduced for Eurosil compared to Baysix [Spi98]. Therefore the two mc-Si materials respond differently to bulk passivation treatments. Whereas bulk passivation by P-gettering was very effective for Eurosil, hydrogenation was more effective for Baysix. Hence, the bulk lifetime is limited by different recombination centres. For Baysix, the defects can be passivated by hydrogen (e.g.

decorated defects at dislocations and grain boundaries) whereas the removal of metallic impurities by gettering is the most successful technique on Eurosil.

2.8.5 Comparison of hydrogenation by MIR passivation and PECVD silicon nitride