Determination of and reflection of light at the SiAl interface

In order to determine the effective back surface recombination velocity , the solar cells

#15 and #19 from the last section were modelled by PC1D simulation, trying to reproduce simultaneously the behavior of the IQE and that of the total front surface reflection . In the long wavelength region, the measured is the fraction of light reflected at the front surface (primary reflection) plus the fraction of light which first enters the solar cell but leaves it in the end after internal reflections (secondary reflection). In PC1D, this is accounted for by specifying the front reflectance and the internal reflection probabilities at the back and at the front surface. For the same surface, the internal reflectance can be enterd for the first bounce of light with that surface and for the second bounce. Also, for each surface the internal reflection can be specular (polished) or diffuse (Lambertian). In order to enter the correct primary reflection for the simulation, the measured was cut off at wavelengths greater than 950 nm, and extrapolated up to 1200 nm from its linear behavior in the 850 to 950 nm region. After importing this front surface reflection into PC1D, the measured IQE and the measured were simulated simultaneously by adjusting the internal reflection parameters and . For the simulations a high bulk lifetime of 1 ms was assumed in order to get an upper limit of . Furthermore, the measured emitter profile of Fig. 3.8 was used and , and 02 were set to the values determined from the dark IV analysis according to Tab. 3.8.

Results

In Fig. 3.16, the measured and the PC1D-fitted values of the IQE and in the long wavelength region are plotted. The parameters corresponding to the PC1D fit are listed in Tab. 3.9. For the thin co-diffused Al-BSF (Seq. A), an of 1200 to 1400 cm/s is found. An lower by a factor of three of 400 to 500 cm/s is found for the screen-printed one (Seq. B), which is due to the increased BSF thickness. Narasimha et al. [121] have also fabricated Al-BSFs by screen-printing and rapid alloying using Al paste Ferro FX-53-038. For comparison, on planar 2.3 cm FZ-Si they have obtained of 200 cm/s which corresponds to almost 400 cm/s if they had used 1.25 cm Si like in this work.

It is worth calculating the base saturation current belonging to according to [45, 48]

2

#

! !

! ! , (3.9)

where)is the base thickness (i.e. cell thickness minus thickness of the BSF),# is the electron diffusion length, is the electron diffusion constant and is the base doping. At 25 C,

800 900 1000 1100 1200 0

20 40 60 80 100

Seq. B) SP Al, RTF measured PC-1D fit

Seq. A) Evap. Al, RTD+RTO measured

PC-1D fit

IQ E , R

[% ]

λ [nm]

Fig. 3.16: Plot of the measured and the PC1D-fitted long wavelength IQE and total front reflection of the RTP solar cells #15 (Seq. A) and #19 (Seq. B).

= 8.6 10⁹cm^-3 [173]. With) of approximately 200 m, = 27.9 cm/s [18] and# of approximately 1600 m⁵ we obtain a of 3 to 4 10^-13 Acm^-2 for the 1.25 cm FZ solar cells with the screen-printed Al-BSF. Together with the measured = 1 10^-13 Acm^-2 (see Fig. 3.10) this matches well to the 01 of 4 to 4.5 10^-13 Acm^-2 determined from dark IV analysis of the cells (see Tab. 3.8). This means, that of these cells is still limited by rather than by . Further reduction of is necessary to fully exploit the potential of the RTO passivated RTD emitter.

With respect to the internal reflection properties, the best PC1D fits are obtained if specular reflection is assumed for the evaporated Al-BSF (Seq. A) but diffuse reflection for the screen-printed Al-BSF (Seq. B). The diffusive nature of an Al-BSF formed by screen-printing and alloying has been observed already by other authors (e.g. see [142]). Regarding the internal reflection probabilities at the BSF, Seq. A is superior to Seq. B. However, as the front surface of the cells is not textured, the fundamental difference in the reflection of light, i.e. specular versus diffuse, has a strong influence on the overall light conversion.

As mentioned, the BSF formed from evaporated Al has a predominantly specular nature.

Consequently, the perpendicularly impinging light hits the back surface, is reflected mainly perpendicularly and thus hits the front surface mainly perpendicularly as well. This explains the rather low reflection probability for the first bounce of just 44 %. Therefore, the probability that light leaves the solar cell after the first run through the solar cell is high, leading to a high

5This value is equivalent to a carrier lifetime of 1 ms representing the best case scenario.

Tab. 3.9: Values for the internal reflection parameters and the effective back surface recombination velocity as obtained from PC1D fitting of the measured IQE and of the total front surface reflection.

Process sequence of Al-BSF

A) B)

Internal reflection properties

Front surface first bounce [%] 44 89

second bounce [%] 92 93

type specular specular

Back surface (Al-BSF) first bounce [%] 86 65

second bounce [%] 86 65

type specular diffuse

Back surface passivation [cm/s] 1200-1400 400-500

. However, the nature of this BSF is not pure specular. Apparently, after a second run through the solar cell, the remaining light impinges mainly non-perpendicularly on the front surface, resulting in a drastically increased reflection probability for the second bounce of 92 %. As the front surface is planar, Seq. A has in fact a negative impact on the overall light conversion capabilities due to its predominantly specular nature. To fully benefit from its good reflection capabilities, surface texturing at least of the front surface has to be applied leading to non-perpendicular light pathes through the solar cell.

On the contrary, the predominantly diffusive nature of the screen-printed BSF leads to a largely increased internal reflection probability for the first bounce at the front surface of approximately 89 %. This indicates mainly non-perpendicular incidence of light after the diffusive reflection at the back. In contrast to Seq. A, this value increases only slightly after a second run through the solar cell. Although the reflection probability is smaller for Seq. B than for Seq. A, the overall light conversion capabilities are superior when planar wafers are used, as is still the case in most multicrystalline solar cell productions.

3.2.5 Summary

RTP-alloyed Al-BSFs were formed form evaporated and screen-printed Al layers, respectively.

In the first case, a 2 m thick layer of ultra-pure Al is evaporated on the back and BSF formation is achieved simultaneously with P emitter diffusion. In the second approach, a thick layer of Al paste is screen-printed on the back and the BSF is formed in a separate RTP alloying step with a peak time of just 1 s. Solar cell were fabricated on 1.25 cm planar Fz-Si wafers using both methods. An effective back surface recombination velocity of 400 to 500 cm/s has been found for the screen-printed Al, while the thin evaporated Al ayer resulted in a BSF with of 1200 to 1400 cm/s. We will show in the next section that of an Al-BSF does not depend on the carrier injection level. Regarding the pure reflection probabilities, the BSF formed from the evaporated Al is superior to the screen-printed one. However, there is a major difference in the nature of reflection. The evaporated Al-BSF exhibits rather mirror-like, i.e. specular

reflection, whereas the screen-printed one obeys diffuse reflection. When the front surface of a solar cells is not textured, the fundamental difference in the reflection of light, i.e. specular vs.

diffuse, has a strong influence on the overall light conversion performance. As a consequence, the screen-printed Al-BSF leads to enhanced overall light conversion capabilities. Of course, this statement holds only for planar front surfaces. Using the screen-printed Al-BSF, an all-RTP-processed 18.7 % efficient solar cell has been fabricated from 1.25 cm planar Fz-Si. The overall process time was below 2 minutes.