The 1 Ω cm FZ samples with diffused emitters on both sides received a forming gas anneal for 25 min at 425 °C. Afterwards the effective carrier lifetime was measured with the QssPC technique at an excess carrier density of ∆n = 1⋅1015 cm-3
W is the wafer thickness, De is the diffusion constant for electrons, Sfront and Srear
are the surface recombination velocity at the front and at the back surface respectively, and γm is the smallest found eigenvalue of the upper equation [87,96].
For the planar samples both surfaces contributed to the same extent to surface recombination and Sfront and Srear are equal. This value was then used as the S-value of the non-textured side of the plasma-textured sample, thus the influence of the texture on emitter recombination could be deduced. The measurements of the
effective minority carrier lifetime were converted to emitter dark saturation current densities j0e via (see also equation 2-41)
eff
Table 5.2 summarises the results and the emitter profile parameters.
Table 5.2: Summary of emitter profile parameters measured by SIMS. The measured dark emitter saturation current density for planar and plasma textured surfaces were calculated from the minority carrier lifetime measurements.
No.
The limit imposed by j0e on the open-circuit voltage VOC of a solar cell was calculated via the equation
where kT/q is 25.9 mV at 25 °C (compare equation 2-44). The short-circuit current density was assumed to be jSC = 38 mA/cm2 and the results are presented on the right-hand axis in the graphs. For the planar samples (Fig. 5.3, solid symbols) a
8 The diffusion at 825 °C was performed in another tube furnace with a different control system for the phosphorus layer deposition. Thus the combination of diffusion temperature and emitter profile is not directly comparable to the others.
strong correlation between emitter sheet resistivity and saturation current density could be established.
Ω
Fig. 5.3: Dark saturation current density joe and imposed limit on VOC of emitters passivated by a thin thermal oxide as a function of emitter sheet resistivity. Planar (solid symbols) and plasma-textured samples (hollow symbols) were investigated. The lines are guides-to-the-eye (solid:
data of Cuevas [97], dash: this work planar, dots: this work plasma textured).
Fig. 5.4: Dark saturation current density joe and imposed limit on VOC of emitters passivated by a thin thermal oxide as a function of phosphorus surface concentration NS. Planar (solid symbols) and plasma-textured samples (hollow symbols) were investigated. The lines are guides-to-the-eye.
This was in good agreement with data published by Cuevas et al. [97] which was derived from a very similar experiment. For the planar samples, the emitter saturation current density decreased linearly with increasing sheet resistivity. The emitters without drive-in oxidation exhibit slightly higher recombination values than the driven-in counterparts. The reason could be found in the dependence of recombination on the phosphorus surface concentration as shown in Fig. 5.4. For the plasma-textured samples (Fig. 5.3, hollow symbols) the recombination was much higher than for the planar samples, which was partly due to the increased surface of the texture. The textured samples did not benefit from higher sheet resistivities, i.e. j0e was hardly affected by the differences in the emitter properties.
This is also evident in Fig. 5.4 where the emitter recombination did not benefit from a reduced surface concentration of NS≤ 1020 cm-3. For the plasma-textured samples a relation between joe and the depth of the emitter xj (measured on planar samples where the phosphorus concentration reached the bulk boron doping level of NA = 1.5⋅1016 cm-3) could not be established either. This led to the conclusion that the geometry of the texture (see Fig. 5.5) with its many sharp peaks, had a major impact on emitter recombination. Those peaks were probably always highly doped, even for the lowly doped planar emitters, since the phosphorus pentoxide layer delivered the phosphorus atoms from all sides. Consequently, the plasma-textured samples were limited by this geometrical issue. The choice of the best emitter diffusion process can take other factors into account, e.g. the gettering of impurities which can occur during phosphorus diffusions or the thermal degradation of multicrystalline silicon at high temperatures.
Fig. 5.5: Schematic drawing of emitter diffusion for a plasma textured surface. The very sharp peaks are highly doped after any emitter diffusion and determine j0e.
5.4 Effect of emitter diffusion on minority carrier lifetime in multicrystalline silicon
One important effect of the emitter diffusion is the change of minority carrier lifetime in multicrystalline silicon caused by thermal degradation and/or gettering of impurities (see chapter 3). For the emitter diffusions of the previous chapter, plasma-textured and planar neighbouring wafers of the same brick were processed in the same run with the FZ-samples. The effective minority carrier lifetime was measured with the QssPC method and evaluated at an excess carrier density of
∆n = 1⋅1015 cm-3. The textured samples with emitter showed lower effective minority carrier lifetimes than the planar samples. The measurements of the multicrystalline wafers reflect the properties of the emitter and the texture. To separate recombination in the bulk from surface recombination, the emitter was etched away from the planar samples and the wafers were covered with a silicon nitride, which effectively suppressed surface recombination (S ≈ 10 cm/s).
Fig. 5.6: Measurement of minority carrier lifetime in multicrystalline silicon. The measurements are arithmetic average values of QssPC measurements on every wafer evaluated at an excess carrier density of ∆n = 1⋅1015 cm-3. All emitters were passivated by a thin thermal oxide layer.
The filled bars denote the effective carrier lifetime τeff of plasma-textured samples, the hollow bars show the bulk lifetime of silicon nitride passivated planar samples after the emitter was etched away.
Thus, the measured lifetimes represent the bulk lifetime τbulk in very good approximation. The result is displayed in Fig. 5.6. For the measurements including the emitter on the surface, the emitter saturation current densities were re-calculated. But this led to much higher j0e-values than for the FZ samples discussed in the previous section. The reason is probably found in an increment of τbulk due to a hydrogen passivation during silicon nitride layer deposition, a topic being under discussion at the moment (see [98-100] and section 7.2). This would lead to an overestimate of the recombination in the emitter. Nevertheless, the determination of the bulk lifetime from the samples passivated with silicon nitride allowed for an assessment of the deteriorating effect caused by the oxidation and the beneficial gettering effect of the emitter diffusion. The obtained data shows, that the wet drive-in oxidations for fours hours at temperatures between 800 °C and 900 °C did not degrade the minority carrier lifetime. At least not irreversibly as the silicon nitride deposition could have caused a hydrogenation of the bulk, which improved the carrier lifetime. Thus the wet oxidation could be used for emitter drive-in. The oxidation for one hour at 1050 °C significantly reduced the material quality and should be excluded from processing of multicrystalline silicon solar cells.
Compared to the reference sample an increased carrier lifetime was only detected for the single-step emitters which were not driven-in. This was probably caused by the beneficial effect of the gettering. These processes could be used for solar cells where the rear surface passivation took place before emitter diffusion or when it is a low-temperature process which does not alter the doping profile.
Now the last question remaining to be answered before choosing the best emitter diffusion for the solar cells, is the contact formation on the front. This is the topic of the next section.