Lifetime Measurements

Beside optical and electrical losses, recombination of electron hole pairs limits the effi-ciency of the solar cell.

As mentioned in Section 3.4, the lower spectral response under rear side illumination is a result of the non-infinite effective lifetime. The effective lifetime is a result of all re-combination processes in the device, which includes rere-combination in the bulk as well as at the front and rear surfaces and in the emitter. The bulk lifetime in silicon is mainly determined by recombination through defects (Shockley-Read-Hall (SRH) recombination) and especially in heavily doped regions, such as the emitter and BSF, by Auger recom-bination, while radiative recombination can be neglected due to the indirect band gap of silicon¹⁰. At surfaces, the recombination via the defects generated by the abrupt ending of the crystal lattice dominates and is described by the surface recombination velocity (SRV).

In a solar cell the individual contributions of the different recombination processes to the effective lifetime can not be isolated. But the investigation of symmetric test struc-tures allows the characterization of certain aspects, for example the quality of the surface passivation and the development of the bulk lifetime through the processing. These in-vestigations were carried out on three types of symmetric test structures - planar p-type float zone silicon wafer¹¹with a (i) thermal oxide, (ii) 110 Ω/sq emitter and thermal oxide and (iii) 60 Ω/sq diffused boron BSF and thermal oxide on each side.

The effective lifetime was determined using quasi steady state photoconductance decay (QSSPC) [144] measurements. In this contactless method the sample is illuminated with a light pulse, whose intensity varies slowly compared to the effective lifetime of the sam-ple (quasi-steady state). This light pulse generates excess carriers in the samsam-ple, which change the conductivity and can therefore be measured inductively. From this the effec-tive lifetime for a wide range of injection levels can be measured.

In Figure 3.5 the injection level dependent results of the QSSPC measurements for the symmetric test structures are shown before and after alnealing.

The effective lifetime increased due to the alneal for all samples. Alnealing has a pas-sivating effect, since the interface state densities at the Si-SiO₂ interface are reduced.

Compared to the conventional anneal (without evaporated aluminium) the content of atomic hydrogen in the oxide is believed to be larger due to a corrosive reaction between the aluminium and residual water in the oxide. The enhanced concentration of hydrogen leads to a more effective saturation of the dangling bonds at the silicon surface and there-fore a reduced density of states within the gap and therethere-fore a reduced recombination due to these traps [141].

The run of the curve of the symmetric sample with SiO₂ on each side changes due to the alneal. Before alnealing the effective lifetime is increasing steadily in the double logarithmic depiction with a maximum of 22 µs at the highest injection level measurable (1.75×10¹⁶cm⁻³). This suggests that SRH recombination is dominant, which means that the oxide does not passivate the surface at this stage. After alnealing a nearly injection

10The lifetime of electron hole pairs if only radiative recombination was present would be in the range of 10 ms and is therefore not the limiting process for these cells.

11The wafers used for the test structures are the same as used for the solar cell processing.

10¹¹ 10¹² 10¹³ 10¹⁴ 10¹⁵ 10¹⁶ 10¹⁷

Excess Carrier Density [cm^-3]

Oxide before alnealing

Excess Carrier Density [cm^-3]

Excess Carrier Density [cm^-3]

Oxide before alnealing

Figure 3.5: Injection level dependent results of QSSPC measurements for three symmetric test structures (thermal oxide on planar 0.5 Ωcm p-type wafer, thermal oxide on planar wafer with 110 Ω/sq emitter and thermal oxide on planar wafer with 60 Ω/sq BSF) before and after alnealing respectively.

level independent effective lifetime is found up to about 8 × 10¹⁵cm⁻³ with the highest measured effective lifetime of about 300 µs. It can be concluded that the passivation properties must be enhanced due to the alneal, but is not distinguishable, whether the bulk or still the surface is limiting the effective lifetime.

For injection levels higher than 2 × 10¹⁶cm⁻³ a strong decrease in effective lifetime is observed, which is due to Coulomb enhanced Auger recombination [145].

For both the oxide covered emitter and BSF samples, nearly no injection level dependence up to 10¹⁶cm⁻³ is found. Also in these samples, the effective lifetime decreases for higher injection levels due to Coulomb enhanced Auger recombination [145].

As expected, the passivation of the n-type surface (emitter) is much more effective than the passivation of the back surface field, which can be seen in the surface recombination velocities (see also Table 3.3). For the same injection level the SRV found for boron diffused samples is much higher than for the test structure with emitter. It is known that n-type silicon is more effectively passivated by silicon oxide than p-type due to much

Sample Treatment τ_effmax S_eff

in µs in cm/s

(∆n_av=5×10¹⁵cm⁻³) (∆n_av=5×10¹⁵cm⁻³)

oxide no alneal 15.2 ± 0.5 36.5 ± 0.5

oxide alneal 300 ± 40 1.8 ± 0.2

emitter + oxide no alneal 49 ± 1 11.2 ± 0.2

emitter + oxide alneal 200 ± 20 2.5 ± 0.5

BSF + oxide no alneal 6.5 ± 1 81 ± 2

BSF + oxide alneal 14.9 ± 0.5 37 ± 1

Table 3.3: Overview of the results of the lifetime measurements on the symmetric test samples.

higher capture cross section for electrons than for holes¹²[146].

From the depiction of 1/τ_eff−1/τ_Augerversus excess carrier density the emitter saturation current is extractable, provided high injection to ensure τ_SRH is constant (slope method).

For the test samples with emitter and thermal oxide this lead to J0e = 29-36 fA before the alneal and 8 fA after the alneal. The implied V_OC was found to increase due to the alneal from 665 mV to 704 mV.

Therefore the surface passivation of these cells is good and does not limit the cell perfor-mance.

12The effective surface recombination velocity Seff depends on the specific recombination velocity of the minority carrier Sp0 and Sn0 respectively, which is proportional to the capture cross section of the corresponding type of carrier. Therefore a higher capture cross section for electrons leads to higher Seff

in the BSF.