Efficiency Losses

In addition to the described recombination losses within the emitter, more efficiency losses are likely to occur during the operation of a solar cell. Fig.2.5shows a schematic view of a lab-type solar cell with its optical and electrical losses. The incident light is only partly converted within the solar cell. First, a fraction of the incident light dissipates by processes involving phonons especially due to the fact that Si is an indirect semiconductor. Also sketched is the free carrier absorption which becomes more important with higher carrier concentration and longer wavelengths [85]. In accordance to the absorption coefficient in Si, mainly short wavelengths are reflected or slightly absorbed by the front ARC layer. Additionally, metallization leads to shadowing losses at the front. Parts of the cell area are covered by the Ti/Pd/Ag front grid. The Ag top layer reflects the whole light spectrum which therefore is lost for light conversion.

Note that the front grid is defined by photolithography. After plating, the achieved grid finger width is approximately 20 µm (see microscopic image in Fig.2.6a) which minimizes the metal coverage below 3% compared with approximately 7% of a conventional screen printed solar cell [94].

The amount of reflected light at the SiNx:H covered cell surface might be reduced by applying a sur-face texture. This is also sketched for the cell in Fig.2.5. On mc-Si it is commonly done by acidic textur-ization consisting of HF and HNO₃[95] since the standard alkaline texturization used for monocrystalline Si is only effective on (100) crystal orientation. The etching of Si atoms is hindered due to their higher density along the (111) orientation. Hence, the intersecting planes with (111) orientation are left over forming random pyramids [96]. This means that the alkaline texture on mc-Si would only be successful on single grains with proper orientation. An alternative plasma texture is applied on a string ribbon Si sample and is shown in Fig.2.7a. The structure resembles a sponge-like structure with hole sizes of approximately 200 nm. In Fig.2.7bits lowered reflectance (sample B) is compared to a non-textured sample A of similar grain structure. A reduction of reflectance by 33%rel is reported at a wavelength of 700 nm. The remote plasma is ignited in a sulfur-hexafluoride SF6, nitrogen N2 and oxygen O2 gas ambient and reaches sufficient homogeneity on four 5×5 cm²samples which is shown elsewhere [97].

In Fig.2.6aa microscopic image of a plated Ti/Pd/Ag finger on plasma textured string ribbon is shown.

Note that the cell results of Chapter7are produced without any kind of texture.

Predominantly long wavelengths reach the backside and are either absorbed or reflected back into the cell. As an effective backside reflector evaporated Al is used in the present cell concept leading to enhanced light conversion. A further reflection at the front is sketched in Fig.2.5for the light originating

Al BSF

Figure 2.5: Schematic view of a lab-type solar cell with its optical and electrical losses. [98]

from the first backside reflection. This might be continued by further internal reflections and might additionally contribute to light conversion. Mainly due to its described wavelength dependency, light conversion in p-type silicon is most efficient within the range of 500 nm to 900 nm (compare theIQEof a FZ cell depicted in Fig.2.3).

The photogenerated charge carriers do not necessarily contribute to the net current flow. They under-lie different recombination mechanisms like the above mentioned Auger recombination in the emitter.

Also carrier loss occurs within the depletion region, the bulk and at the backside. The depletion re-gion is highly sensitive to, e.g., metal impurities which might diffuse in from the contacts and induce shunts at the p-n junction. Bulk recombination in mc-Si mainly originates from crystal defects like grain boundaries and dislocations as well as from impurities like, e.g., Fe atoms partly incorporated in precipi-tates but also dissolved interstitially. Optimizing for example firing parameters like the peak temperature and/or the composition of the Al paste used for BSF formation might reduce carrier loss at the rear.

Electrochemical capacitance-voltage (ECV) measurements of the Al-BSF profiles with different peak temperatures are shown in Fig. 2.6b. The ECV method will be described in Sec.2.3.3. The sample temperature¹ needs to exceed 577^◦C for BSF formation. At this point the eutectic reaction of Al and Si takes place and beyond that temperature a melt is formed with more silicon solved than within the eutectic composition (Al-rich composition with 12.2 at.% Si [99]). Below the near-surface region of the eutectic the Al is incorporated into the Si host as acceptor. The higher the set peak temperature is, the deeper is the Al⁻concentration profile of the respective mc-Si sample. The lowest set peak temperature of 801^◦C is not sufficient to form a wide overcompensating BSF.

Note the excess surface concentrations of both profiles with higher peak temperatures compared to slightly deeper concentrations at≈1 µm. This was previously observed by Huster and Schubert [100] and

1In Fig.2.6b, the set peak temperature of the belt furnace is given instead of the sample temperature.

2.2 Operating Principle and Production of Lab-type Solar Cells 31

(a)

19.41µm

(b)

Figure 2.6: (a) Microscopic image of a plated Ti/Pd/Ag finger on plasma textured string ribbon silicon with a width of 19.4 µm. (b) Electrochemical capacitance-voltage (ECV) measurements of Al-BSF formation on mc-Si at three different firing peak temperatures.

(a) (b)

Figure 2.7: SEM images of the sponge-like plasma texture on string ribbon silicon with its lowered reflectance compared to a flat sample of similar grain structure.

does not agree with the solid solubility of Al in Si [101]. R. Bock explained the observation by a surface enlargement of≈60 % due to Al-rich eutectic residuals [102]. These residuals are identified by islands filled with the eutectic and covered by a thin Si layer. The silicon layer around the aluminum protects it from being etched off by HCl. This is used to remove the Al paste and to prepare the surface for the electrochemical capacitance-voltage (ECV) measurement. The ECV method is described in Sec.2.3.3.

In general, metallization, surface passivation and the emitter are critical cell components being per-manently under technological change. In this work the solar cell process is kept constant and the focus will be on the characterization of the bulk material. In particular, lifetime characterization is performed before and after phosphorus diffusion gettering (PDG). The effect of PDG results in a significant im-provement of the material quality and will be described in Sec.2.3.2.

(a)

J_PH J₀₁ J₀₂ R_SH R_S

V,J Incident

photons

(b)

Figure 2.8: (a) Equivalent circuit diagram of an illuminated solar cell: double-diode model. (b) Illuminated J-V curve of a 2×2 cm²FZ cell.