Solar cells convert sunlight into electrical energy using the photo effect. With a mar-ket share of 60%, multicrystalline silicon (mc-Si) is the most frequently used absorber material. Standard mc-Si ingots are directionally solidified in a fused silica (SiO2) crucible, which exhibits a silicon nitride (Si3N4) inner coating. After the entire raw material has been melted, the nucleation takes place on the Si3N4 inner coating at the bottom of the crucible. This results in an inhomogeneous initial grain structure and an increased fraction of dislocation clusters in the upper part of the ingot, which decrease the quality of standard mc-Si. Therefore, the global goal is the development of a cost-effective technology that reduces the formation of clusters and enhances the quality of mc-Si ingots. One way of achieving that goal is to produce the so-called \"high performance multi crystalline silicon\" (HPM-Si). During the directional solidification silicon raw material remains unmelted at the bottom of the SiO2 crucible, whereby crystallization does start on the silicon feedstock a few millimeters above the crucible bottom. Compared to standard mc-Si, a finer grained structure with many small grains is formed, which are separated by so-called random grain boundaries. Since the movement of dislocations across this grain boundary type has rarely been observed, the risk of formation of dislocation clusters, which have a negative impact on the efficiency of solar cells, is greatly reduced for HPM-Si. However, the disadvantage of the HPM-Si compared to the mc-Si is the yield loss resulting from the unmelted raw material at the crucible bottom. Hence, the aim of the present work is to produce mc-Si with a fine-grained structure in combination with a high fraction of random grain boundaries without the disad-vantage of yield loss. In order to investigate the grain structure in dependence of the nucleation conditions G1 ingots having a mass of 14.5 kg and dimensions of 220x220x130 mm³ were directional solidified in a furnace. The analysis of the grain structure with respect to the grain size, grain orientation and the random grain boundary length fraction and the comparison with the HPM-Si reference crystal took place on horizontal wafers with a thickness of 3mm. One possibility to influence the grain structure of mc-Si could be the variation of the cooling conditions before the start of crystallization at the crucible bottom. In a first series of experiments, a gas-flowed cooling plate, positioned below the crucible, was used. An increased gas flow increases the axial heat flow downwards and the cooling rate below the crucible bottom in the same direction. The detected cooling rate, measured by a thermocouple in the silicon melt 5 mm above the crucible bottom, varied in a range between 0.06-1.5 K/min. An increased cooling rate increases the supercooling, with a maximum of 2K. The analysis of the grain structure shows that a reduction in the cooling rate in combination with the lowest supercooling minimizes the average grain size and increases the fraction of random grain boundaries. However, an HPM-Si like grain structure (grain size and fraction of random grain boundaries comparable to HPM-Si) could not completely produced. Furthermore, due to the extended process time, the wafer yield is reduced, whereby the reduction of the cooling rate is not a preferable method for the industrial process. In a second experimental series, which took place under constant cooling rates, the influence of an additional nucleation layer on the initial grain structure was investigated. For this purpose, the additional nucleation layer was applied on the already existing Si3N4 inner coating on the crucible bottom. In order to adjust a HPM-Si like grain structure, the contact angle of the silicon melt on the additional nucleation layer should be lower than on the Si3N4 inner coating. The theoretical basis for this hypothesis is the relationship between the contact angle and the nucleation energy, which states that a reduced contact angle lowers the nucleation energy and can ultimately lead to more nuclei. Furthermore, in order to avoid melting, the additional nucleation layer must have a higher melting point than silicon. Suitable materials for the application as a foreign seed sample are SiC, SiO2 and Al2O3, which are used in the form of particles with different sizes. The production of the additional nucleation layer was carried out by a spraying as well as by an embedding procedure. These layers exhibit different thermal conductivity as well as surface roughness. Embedded nucleation layers generate higher roughness values than sprayed nucleation layers. The analysis of the grain structure identified the surface roughness as the main influencing factor on the initial grain size. While an increased surface roughness (Rq>100μm) results in a fine-grained structure (average grain size: <2mm²) comparable to HPM-Si, the average grain size increases (>2 mm²) with a reduced surface roughness (Rq<100μm). However, the analysis of the grain boundary relationship shows that the fraction of random grain boundaries does not correlate with the average grain size. Only a ma-terial dependency was detected. All SiO2 nucleation layers generate an increased fraction of random grain boundaries, comparable to the HPM-Si material. In contrast, the fraction of random grain boundaries was reduced for all SiC nucleation layers. This result is probably established with the different thermal conductivities of the used materials. The increased thermal conductivity of the sample with the SiC nucleation layers increases the cooling rate, promoting dendritic growth. In contrast the lower thermal conductivity of the SiO2 nucleation layers reduces the cooling rate and dendritic growth is suppressed. Since dendrites exhibit a Σ3 grain boundary relationship in the center, the fraction of this grain boundary type increases for SiC nucleation layers and the fraction of random grain boundaries decreases. In this thesis, various possibilities for influencing the grain structure have been pre-sented. A SiO2 nucleation layer with a roughness value Rq> 200μm represents an industrially relevant solution for the production of mc-Si with comparable properties to the HPM-Si without the disadvantages of yield loss. Hence, it was possible to in-crease the yield with comparable material quality, whereby the production costs could be reduced. Some first crucible manufacturers have already transferred the use of the SiO2 nucleation layers on top of the already existing Si3N4 inner coating at the crucible bottom to production.

Solar cells convert sunlight into electrical energy using the photo effect. With a mar-ket share of 60%, multicrystalline silicon (mc-Si) is the most frequently used absorber material. Standard mc-Si ingots are directionally solidified in a fused silica (SiO2) crucible, which exhibits a silicon nitride (Si3N4) inner coating. After the entire raw material has been melted, the nucleation takes place on the Si3N4 inner coating at the bottom of the crucible. This results in an inhomogeneous initial grain structure and an increased fraction of dislocation clusters in the upper part of the ingot, which decrease the quality of standard mc-Si. Therefore, the global goal is the development of a cost-effective technology that reduces the formation of clusters and enhances the quality of mc-Si ingots. One way of achieving that goal is to produce the so-called \"high performance multi crystalline silicon\" (HPM-Si). During the directional solidification silicon raw material remains unmelted at the bottom of the SiO2 crucible, whereby crystallization does start on the silicon feedstock a few millimeters above the crucible bottom. Compared to standard mc-Si, a finer grained structure with many small grains is formed, which are separated by so-called random grain boundaries. Since the movement of dislocations across this grain boundary type has rarely been observed, the risk of formation of dislocation clusters, which have a negative impact on the efficiency of solar cells, is greatly reduced for HPM-Si. However, the disadvantage of the HPM-Si compared to the mc-Si is the yield loss resulting from the unmelted raw material at the crucible bottom. Hence, the aim of the present work is to produce mc-Si with a fine-grained structure in combination with a high fraction of random grain boundaries without the disad-vantage of yield loss. In order to investigate the grain structure in dependence of the nucleation conditions G1 ingots having a mass of 14.5 kg and dimensions of 220x220x130 mm³ were directional solidified in a furnace. The analysis of the grain structure with respect to the grain size, grain orientation and the random grain boundary length fraction and the comparison with the HPM-Si reference crystal took place on horizontal wafers with a thickness of 3mm. One possibility to influence the grain structure of mc-Si could be the variation of the cooling conditions before the start of crystallization at the crucible bottom. In a first series of experiments, a gas-flowed cooling plate, positioned below the crucible, was used. An increased gas flow increases the axial heat flow downwards and the cooling rate below the crucible bottom in the same direction. The detected cooling rate, measured by a thermocouple in the silicon melt 5 mm above the crucible bottom, varied in a range between 0.06-1.5 K/min. An increased cooling rate increases the supercooling, with a maximum of 2K. The analysis of the grain structure shows that a reduction in the cooling rate in combination with the lowest supercooling minimizes the average grain size and increases the fraction of random grain boundaries. However, an HPM-Si like grain structure (grain size and fraction of random grain boundaries comparable to HPM-Si) could not completely produced. Furthermore, due to the extended process time, the wafer yield is reduced, whereby the reduction of the cooling rate is not a preferable method for the industrial process. In a second experimental series, which took place under constant cooling rates, the influence of an additional nucleation layer on the initial grain structure was investigated. For this purpose, the additional nucleation layer was applied on the already existing Si3N4 inner coating on the crucible bottom. In order to adjust a HPM-Si like grain structure, the contact angle of the silicon melt on the additional nucleation layer should be lower than on the Si3N4 inner coating. The theoretical basis for this hypothesis is the relationship between the contact angle and the nucleation energy, which states that a reduced contact angle lowers the nucleation energy and can ultimately lead to more nuclei. Furthermore, in order to avoid melting, the additional nucleation layer must have a higher melting point than silicon. Suitable materials for the application as a foreign seed sample are SiC, SiO2 and Al2O3, which are used in the form of particles with different sizes. The production of the additional nucleation layer was carried out by a spraying as well as by an embedding procedure. These layers exhibit different thermal conductivity as well as surface roughness. Embedded nucleation layers generate higher roughness values than sprayed nucleation layers. The analysis of the grain structure identified the surface roughness as the main influencing factor on the initial grain size. While an increased surface roughness (Rq>100μm) results in a fine-grained structure (average grain size: <2mm²) comparable to HPM-Si, the average grain size increases (>2 mm²) with a reduced surface roughness (Rq<100μm). However, the analysis of the grain boundary relationship shows that the fraction of random grain boundaries does not correlate with the average grain size. Only a ma-terial dependency was detected. All SiO2 nucleation layers generate an increased fraction of random grain boundaries, comparable to the HPM-Si material. In contrast, the fraction of random grain boundaries was reduced for all SiC nucleation layers. This result is probably established with the different thermal conductivities of the used materials. The increased thermal conductivity of the sample with the SiC nucleation layers increases the cooling rate, promoting dendritic growth. In contrast the lower thermal conductivity of the SiO2 nucleation layers reduces the cooling rate and dendritic growth is suppressed. Since dendrites exhibit a Σ3 grain boundary relationship in the center, the fraction of this grain boundary type increases for SiC nucleation layers and the fraction of random grain boundaries decreases. In this thesis, various possibilities for influencing the grain structure have been pre-sented. A SiO2 nucleation layer with a roughness value Rq> 200μm represents an industrially relevant solution for the production of mc-Si with comparable properties to the HPM-Si without the disadvantages of yield loss. Hence, it was possible to in-crease the yield with comparable material quality, whereby the production costs could be reduced. Some first crucible manufacturers have already transferred the use of the SiO2 nucleation layers on top of the already existing Si3N4 inner coating at the crucible bottom to production.