Innerhalb dieser Arbeit werden die Wechselwirkungen von interstitiellem Eisen mit Defekten, wie Versetzungen und Ausscheidungen, untersucht. Dazu wurde zunächst das Löslichkeitsverhalten von Eisen im p-dotierten mc-Silizium im Temperaturbereich von 550°C bis 800°C gemessen. An Versetzungsstrukturen konnte in einigen Bereichen eine deutlich erhöhte Konzentration an gelöstem Eisen im as cut Zustand und nach dem Tempern der Probe festgestellt werden. Diese wurde mit der Struktur der Versetzungsanordnung erklärt. An vorhandenen Ausscheidungen im Silizium konnte eine Abnahme der gelösten Eisenkonzentration beobachtet werden. Die Dichte und Morphologie der Ausscheidungen haben dabei einen großen Einfluss. Der Eintrag des Eisens in das Silizium aus Tiegeln mit unterschiedlicher Reinheit während der Kristallisation wurde untersucht. Dabei zeigte sich ein deutlicher Einfluss der Tiegelreinheit auf die Konzentration am Rand des Blockes, jedoch nur einen geringen Einfluss auf die Eisenkonzentration im Blockinneren.

Die vorliegende Arbeit befasst sich mit den Untersuchungen zum Einfluss der Züchtungsbedingungen auf die Eigenschaften von multikristallinen (mc) Silizium-Kristallen. Im Mittelpunkt stehen Züchtungsexperimente mit einer gezielten Variation der Züchtungsaufbauten und Züchtungsgeschwindigkeiten. Die gezüchteten Kristalle wurden umfassend charakterisiert im Hinblick auf die Kohlenstoffkonzentration, die Kornstruktur, die Vesetzungsdichte, Verteilung der Ausscheidungen und Ladungsträgerlebensdauer. Zusätzlich wurde die Versetzungsanordnung in Abhängigkeit von der Wachstumsrate bzw. Abkühlrate systematisch untersucht. Als Ergebnis wurde gezeigt, dass die Züchtungsbedingungen die Kohlenstoffkonzentration, die Versetzungsdichte, die Bildung von den Ausscheidungen sowie die Ladungsträgerlebensdauer beeinflussen können, jedoch nicht die Korngröße. Es wurde ein direkter Zusammenhang zwischen Ausscheidungsgebieten und erhöhte Versetzungsdichte beobachtet. Im Rahmen der Arbeit wurde festgestellt, dass die endgültige Versetzungsstruktur sich als Resultat von Gleit- und Erholungsprozessen darstellt.

The efficiency of multicrystalline (mc) silicon solar cells depends strongly on the fraction of recombination active crystal defects. This work focuses on a systematic analysis of how the area fraction of recombination active crystal defects and thus the solar cell efficiency is af-fected by the grain structure of mc-silicon wafers, i.e. grain size, grain orientation and type of the grain boundaries between adjacent grains. For that purpose a new characterization method was developed which allows the measurement of the grain orientation and grain boundary type of full 156x156 mm² mc-silicon wafers. The results of the grain structure analysis were correlated with the etch pit density, the recombination active area fraction measured by photo-luminescence imaging, and the solar cell efficiency in order to quantify the most important features of the grain structure, which were relevant to obtain high quality mc-silicon wafer material. For the determination of the grain orientation and grain boundary type two metrology sys-tems were combined. The so-called grain detector determines the geometrical data of each grain (size and form) by a reflectivity measurement. Afterwards the wafer with the geomet-rical information of all grains is transferred into the so-called Laue Scanner. This system irra-diates each grain larger 3 mm² with white x-rays and creates a backscatter diffraction pattern (Laue pattern) for each grain. From this Laue pattern the grain orientation and the grain boundary type of neighboured grains is calculated and statistically analysed in combination with the geometrical data of the grain detector. In this work the grain structure of twelve industrially grown mc-silicon bricks, which were produced by different manufacturers, and two laboratory grown bricks were investigated. Seven of these bricks show a fine grain structure. This material named class F is considered to be typical for so-called High Performance Multi (HPM) silicon. The other bricks show a coarse-grained structure. This grain structure was called class G and corresponds to the con-ventional mc-silicon material. The results show that the grain structures of the start of the crystallization process differ sig-nificantly between class F and class G. The class F mc-silicon wafers have a uniform initial grain size (characterized by coefficient of variation CV¬KG < 2.5) and grain orientation (charac-terized by coefficient of variation CVKO < 1.5) distribution with a small mean grain size (< 4 mm²) and a high length fraction of random grain boundaries (> 60 %) in comparison to the class G wafers. Despite the totally different initial grain structure for the class F and class G bricks, the grain structure of the wafers which represent the end of the crystallization process is more or less comparable. It can be concluded that the development of the grain structure along the crystal height of the class F bricks is driven by an energy minimization due to the surface energy and the grain boundary energy, that means that the share of (111) oriented grains having the lowest surface energy and the share of ∑3 grain boundaries having the lowest interface energy increase from the start of crystallization to the end. This phenomenon could not be observed for the class G bricks, which show a decreasing ∑3 length fraction and a decreasing area fraction of {111} oriented grains. This energetically unfavourable grain structure development is not clear so far but it means another kind of energy minimization effect must exist within class G. This could be for instance the formation of dislocations. The grain structure investigations show clearly that especially the initially fine-grained struc-ture of the class F bricks, i.e. at the start of crystallization, influences beneficially the area fraction of recombination active defects and the solar cell efficiency subsequently. This ob-servation can be explained as follows. Reduced dislocation cluster formation: • The small grain sizes in combination with the low length fraction of ∑3 grain bounda-ries capture the dislocations within a grain. Dislocations are not able to move across the grain boundaries which have not the ∑3-type within moderate stress and tempera-ture fields. This prohibits the formation and expansion of large dislocation cluster. • The previously described energetically driven grain selection and the continuously in-creasing grain size from bottom to top leads to an overgrowth of grains. This means that also dislocated grains will disappear which also prohibits the formation of large dislocation cluster. Reduced possibility of dislocation formation: • Compared to the class G bricks the area fraction of {111} oriented grains is reduced. Therefore, the possibility of the formation of dislocations is reduced, because they would be activated first in {111} oriented grains taking the Schmidt factor in account which is lowest for {111} oriented grains. After the dislocation generation within a {111} oriented grain, the dislocation can move forward on 3 of 4 possible {111} slip planes which have an angle of 19.5° with regard to the growth direction. No other ori-entation has more slip planes for the dislocation movement which have an angle smaller 20° with regard to the growth direction. These arguments in combination with the high reproducibility of the characteristic initial class F structure can explain the observed low recombination active area fraction from start to end of crystallization which was smaller 5 % and especially the low variation of 2 % of the electrical active wafer area in between the class F bricks. One can also easily explain the higher recombination active area fraction up to 14 % and the large variation of 10 % between the class G bricks due to the obtained grain structure data. These differences in the recombination active area fractions are reflected in the solar cell efficiency which is 0.4 % higher for the class F bricks compared to the class G bricks. In consideration of the above mentioned reasons it is not beneficial for the industrial ingot production technology to increase the ingot height further, due to the fact that the advanta-geous initial grain structure properties of class F bricks disappear with increasing crystal height.

The usage of more inexpensive silicon feedstock for the crystallization of multicrystalline silicon blocks promises cost reduction for the photovoltaic industry. Less expensive substrates made out of metallurgical silicon (MG-Si) are used as a mechanical support for the epitaxial solar cell. Moreover, conventional inert solar cells can be produced from up-graded metallurgical silicon (UMG-Si). This feedstock has higher content of impurities which influences cell performance and mechanical strength of the wafers. Thus, it is of importance to know these effects in order to know which impurities should be preferentially removed or prevented during the crystallization process. Solar cell processing steps can also exert a change in the values of mechanical strength of processed multicrystalline silicon wafers until the fabrication of a solar cell. Bending tests, fracture toughness and dynamic elastic modulus measurements are performed in this work in order to research the mechanical behavior of multicrystalline silicon crystallized with different qualities of silicon feedstock. Bending tests and residual stress measurements allows the quantification of the mechanical strength of the wafers after every solar cell processing step. The experimental results are compared with theoretical models found in the classical literature about the mechanical properties of ceramics. The influence of second phase particles and thermal processes on the mechanical strength of silicon wafers can be predicted and analyzed with the theoretical models. Metals like Al and Cu can decrease the mechanical strength due to micro-cracking of the silicon matrix and introduction of high values of thermal residual stress. Additionally, amorphous silicon oxide particles (SiOx) lower the mechanical strength of multicrystalline silicon due to thermal residual stresses and elastic mismatch with silicon. Silicon nitride particles (Si3N4) reduce fracture toughness and cause failure by radial cracking in its surroundings due to its thermal mismatch with silicon. Finally, silicon carbide (SiC) and crystalline silicon oxide (SiOx) introduce thermal residual stresses but can have a toughening effect on the silicon matrix and hence, increase the mechanical strength of silicon wafers if the particles are smaller than a certain size. The surface of as-cut wafers after multi-wire sawing presents sharp micro-cracks that control their mechanical behavior. Subsequent removal of these micro-cracks by texture or damage etching approximately doubles the mechanical strength of silicon wafers. The mechanical behavior of the wafers is then governed by defects like cracks and particles formed during the crystallization of multicrystalline silicon blocks. Further thermal processing steps have a minor impact on the mechanical strength of the wafers compared to as-cut wafers. Finally, the mechanical strength of final solar cells is comparable to the mechanical strength of as-cut wafers due to the high residual thermal stress introduced after the formation of the metallic contacts which makes silicon prone to crack.

Innerhalb dieser Arbeit werden die Wechselwirkungen von interstitiellem Eisen mit Defekten, wie Versetzungen und Ausscheidungen, untersucht. Dazu wurde zunächst das Löslichkeitsverhalten von Eisen im p-dotierten mc-Silizium im Temperaturbereich von 550°C bis 800°C gemessen. An Versetzungsstrukturen konnte in einigen Bereichen eine deutlich erhöhte Konzentration an gelöstem Eisen im as cut Zustand und nach dem Tempern der Probe festgestellt werden. Diese wurde mit der Struktur der Versetzungsanordnung erklärt. An vorhandenen Ausscheidungen im Silizium konnte eine Abnahme der gelösten Eisenkonzentration beobachtet werden. Die Dichte und Morphologie der Ausscheidungen haben dabei einen großen Einfluss. Der Eintrag des Eisens in das Silizium aus Tiegeln mit unterschiedlicher Reinheit während der Kristallisation wurde untersucht. Dabei zeigte sich ein deutlicher Einfluss der Tiegelreinheit auf die Konzentration am Rand des Blockes, jedoch nur einen geringen Einfluss auf die Eisenkonzentration im Blockinneren.

Die vorliegende Arbeit befasst sich mit den Untersuchungen zum Einfluss der Züchtungsbedingungen auf die Eigenschaften von multikristallinen (mc) Silizium-Kristallen. Im Mittelpunkt stehen Züchtungsexperimente mit einer gezielten Variation der Züchtungsaufbauten und Züchtungsgeschwindigkeiten. Die gezüchteten Kristalle wurden umfassend charakterisiert im Hinblick auf die Kohlenstoffkonzentration, die Kornstruktur, die Vesetzungsdichte, Verteilung der Ausscheidungen und Ladungsträgerlebensdauer. Zusätzlich wurde die Versetzungsanordnung in Abhängigkeit von der Wachstumsrate bzw. Abkühlrate systematisch untersucht. Als Ergebnis wurde gezeigt, dass die Züchtungsbedingungen die Kohlenstoffkonzentration, die Versetzungsdichte, die Bildung von den Ausscheidungen sowie die Ladungsträgerlebensdauer beeinflussen können, jedoch nicht die Korngröße. Es wurde ein direkter Zusammenhang zwischen Ausscheidungsgebieten und erhöhte Versetzungsdichte beobachtet. Im Rahmen der Arbeit wurde festgestellt, dass die endgültige Versetzungsstruktur sich als Resultat von Gleit- und Erholungsprozessen darstellt.

Die Rekombinationsaktivität von Versetzungen und Korngrenzen in multikristallinem Silicium wird von Kupfer und anderen metallischen Verunreinigungen wie Eisen mitbestimmt. Das Hauptziel der Arbeit war es, die Verteilung von Kupfer und dessen Wirkung auf die Rekombinationsaktivität von Versetzungen und Korngrenzen genauer zu untersuchen. Dazu wurden optische und elektrische Untersuchungen an gezielt mit Metallen verunreinigten Modellmaterialien durchgeführt. Nicht alle Versetzungen sind rekombinationsaktiv. Es konnte gezeigt werden, dass der Anteil rekombinationsaktiver Versetzungen am Gesamtinventar und die Hintergrunddiffusionslänge von der Verunreinigung mit Metallen abhängig sind. Ergebnisse von Untersuchungen an Proben, die Diffusionsexperimenten unterzogen wurden, deuten auf unterschiedliches Ausscheidungsverhalten von Kupfer und Eisen hin sowie auf Wechselwirkungen mit Versetzungen und Korngrenzen, die mit der Diffusionstemperatur und den Abkühlbedingungen in Zusammenhang stehen.

The usage of more inexpensive silicon feedstock for the crystallization of multicrystalline silicon blocks promises cost reduction for the photovoltaic industry. Less expensive substrates made out of metallurgical silicon (MG-Si) are used as a mechanical support for the epitaxial solar cell. Moreover, conventional inert solar cells can be produced from up-graded metallurgical silicon (UMG-Si). This feedstock has higher content of impurities which influences cell performance and mechanical strength of the wafers. Thus, it is of importance to know these effects in order to know which impurities should be preferentially removed or prevented during the crystallization process. Solar cell processing steps can also exert a change in the values of mechanical strength of processed multicrystalline silicon wafers until the fabrication of a solar cell. Bending tests, fracture toughness and dynamic elastic modulus measurements are performed in this work in order to research the mechanical behavior of multicrystalline silicon crystallized with different qualities of silicon feedstock. Bending tests and residual stress measurements allows the quantification of the mechanical strength of the wafers after every solar cell processing step. The experimental results are compared with theoretical models found in the classical literature about the mechanical properties of ceramics. The influence of second phase particles and thermal processes on the mechanical strength of silicon wafers can be predicted and analyzed with the theoretical models. Metals like Al and Cu can decrease the mechanical strength due to micro-cracking of the silicon matrix and introduction of high values of thermal residual stress. Additionally, amorphous silicon oxide particles (SiOx) lower the mechanical strength of multicrystalline silicon due to thermal residual stresses and elastic mismatch with silicon. Silicon nitride particles (Si3N4) reduce fracture toughness and cause failure by radial cracking in its surroundings due to its thermal mismatch with silicon. Finally, silicon carbide (SiC) and crystalline silicon oxide (SiOx) introduce thermal residual stresses but can have a toughening effect on the silicon matrix and hence, increase the mechanical strength of silicon wafers if the particles are smaller than a certain size. The surface of as-cut wafers after multi-wire sawing presents sharp micro-cracks that control their mechanical behavior. Subsequent removal of these micro-cracks by texture or damage etching approximately doubles the mechanical strength of silicon wafers. The mechanical behavior of the wafers is then governed by defects like cracks and particles formed during the crystallization of multicrystalline silicon blocks. Further thermal processing steps have a minor impact on the mechanical strength of the wafers compared to as-cut wafers. Finally, the mechanical strength of final solar cells is comparable to the mechanical strength of as-cut wafers due to the high residual thermal stress introduced after the formation of the metallic contacts which makes silicon prone to crack.

The efficiency of multicrystalline (mc) silicon solar cells depends strongly on the fraction of recombination active crystal defects. This work focuses on a systematic analysis of how the area fraction of recombination active crystal defects and thus the solar cell efficiency is af-fected by the grain structure of mc-silicon wafers, i.e. grain size, grain orientation and type of the grain boundaries between adjacent grains. For that purpose a new characterization method was developed which allows the measurement of the grain orientation and grain boundary type of full 156x156 mm² mc-silicon wafers. The results of the grain structure analysis were correlated with the etch pit density, the recombination active area fraction measured by photo-luminescence imaging, and the solar cell efficiency in order to quantify the most important features of the grain structure, which were relevant to obtain high quality mc-silicon wafer material. For the determination of the grain orientation and grain boundary type two metrology sys-tems were combined. The so-called grain detector determines the geometrical data of each grain (size and form) by a reflectivity measurement. Afterwards the wafer with the geomet-rical information of all grains is transferred into the so-called Laue Scanner. This system irra-diates each grain larger 3 mm² with white x-rays and creates a backscatter diffraction pattern (Laue pattern) for each grain. From this Laue pattern the grain orientation and the grain boundary type of neighboured grains is calculated and statistically analysed in combination with the geometrical data of the grain detector. In this work the grain structure of twelve industrially grown mc-silicon bricks, which were produced by different manufacturers, and two laboratory grown bricks were investigated. Seven of these bricks show a fine grain structure. This material named class F is considered to be typical for so-called High Performance Multi (HPM) silicon. The other bricks show a coarse-grained structure. This grain structure was called class G and corresponds to the con-ventional mc-silicon material. The results show that the grain structures of the start of the crystallization process differ sig-nificantly between class F and class G. The class F mc-silicon wafers have a uniform initial grain size (characterized by coefficient of variation CV¬KG < 2.5) and grain orientation (charac-terized by coefficient of variation CVKO < 1.5) distribution with a small mean grain size (< 4 mm²) and a high length fraction of random grain boundaries (> 60 %) in comparison to the class G wafers. Despite the totally different initial grain structure for the class F and class G bricks, the grain structure of the wafers which represent the end of the crystallization process is more or less comparable. It can be concluded that the development of the grain structure along the crystal height of the class F bricks is driven by an energy minimization due to the surface energy and the grain boundary energy, that means that the share of (111) oriented grains having the lowest surface energy and the share of ∑3 grain boundaries having the lowest interface energy increase from the start of crystallization to the end. This phenomenon could not be observed for the class G bricks, which show a decreasing ∑3 length fraction and a decreasing area fraction of {111} oriented grains. This energetically unfavourable grain structure development is not clear so far but it means another kind of energy minimization effect must exist within class G. This could be for instance the formation of dislocations. The grain structure investigations show clearly that especially the initially fine-grained struc-ture of the class F bricks, i.e. at the start of crystallization, influences beneficially the area fraction of recombination active defects and the solar cell efficiency subsequently. This ob-servation can be explained as follows. Reduced dislocation cluster formation: • The small grain sizes in combination with the low length fraction of ∑3 grain bounda-ries capture the dislocations within a grain. Dislocations are not able to move across the grain boundaries which have not the ∑3-type within moderate stress and tempera-ture fields. This prohibits the formation and expansion of large dislocation cluster. • The previously described energetically driven grain selection and the continuously in-creasing grain size from bottom to top leads to an overgrowth of grains. This means that also dislocated grains will disappear which also prohibits the formation of large dislocation cluster. Reduced possibility of dislocation formation: • Compared to the class G bricks the area fraction of {111} oriented grains is reduced. Therefore, the possibility of the formation of dislocations is reduced, because they would be activated first in {111} oriented grains taking the Schmidt factor in account which is lowest for {111} oriented grains. After the dislocation generation within a {111} oriented grain, the dislocation can move forward on 3 of 4 possible {111} slip planes which have an angle of 19.5° with regard to the growth direction. No other ori-entation has more slip planes for the dislocation movement which have an angle smaller 20° with regard to the growth direction. These arguments in combination with the high reproducibility of the characteristic initial class F structure can explain the observed low recombination active area fraction from start to end of crystallization which was smaller 5 % and especially the low variation of 2 % of the electrical active wafer area in between the class F bricks. One can also easily explain the higher recombination active area fraction up to 14 % and the large variation of 10 % between the class G bricks due to the obtained grain structure data. These differences in the recombination active area fractions are reflected in the solar cell efficiency which is 0.4 % higher for the class F bricks compared to the class G bricks. In consideration of the above mentioned reasons it is not beneficial for the industrial ingot production technology to increase the ingot height further, due to the fact that the advanta-geous initial grain structure properties of class F bricks disappear with increasing crystal height.:Abstract 1. Einleitung 1.1 Photovoltaik 1.2 Stand der Technik 1.2.1 Blockerstarrung von multikristallinem Silizium 1.2.2 Kornorientierungsbestimmung 1.3 Zielsetzung und Gliederung der Arbeit 2. Grundlagen 2.1 Silizium 2.1.1 Elektrische Eigenschaften 2.1.2 Oberflächenenergien des Siliziums 2.2 Kristalldefekte in multikristallinem Silizium 2.2.1 Versetzungen 2.2.2 Korngrenzen 2.2.3 Wechselwirkung zwischen Versetzungen und Korngrenzen 3. Mess- und Auswertemethodik 3.1 Detektion der Körner 3.1.1 Aufbau und Funktionsweise 3.1.2 Definition der Kenngrößen 3.1.3 Fehlerbetrachtung 3.2 Detektion der Kornorientierungen und Korngrenztypen 3.2.1 Theoretische Betrachtung 3.2.2 Aufbau und Funktionsweise 3.2.3 Definition der Kenngrößen 3.2.4 Fehlerbetrachtung 3.3 Detektion der Ätzgrubendichte 3.3.1 Aufbau und Funktionsweise 3.3.2 Definition der Kenngrößen 3.3.3 Fehlerbetrachtung 3.4 Detektion des rekombinationsaktiven Flächenanteils 3.4.1 Aufbau und Funktionsweise 3.4.2 Definition der Kenngrößen 3.4.3 Fehlerbetrachtung 3.5 Korrelation der rekombinationsaktiven Kristalldefekte mit der Kornorientierung 4. Probeninformation 5. Ergebnisteil 5.1 Korngrößenverteilung 5.1.1 Säulenklassifizierung 5.1.2 Klasse F Säulen 5.1.3 Klasse G Säulen 5.2 Kornorientierungsverteilung 5.2.1 Klasse F Säulen 5.2.2 Klasse G Säulen 5.3 Korngrenztypverteilung 5.3.1 Klasse F Säulen 5.3.2 Klasse G Säulen 5.4 Ätzgrubendichte 5.4.1 Klasse F Säulen 5.4.2 Klasse G Säulen 5.5 Rekombinationsaktiver Flächenanteil 5.5.1 Klasse F Säulen 5.5.2 Klasse G Säulen 5.6 Korrelation der Ergebnisse 5.6.1 Mittlere Korngröße und Variationskoeffizient vs. rekombinationsaktiver Flächenanteil 5.6.2 Korngrenztyplängenanteil vs. rekombinationsaktiver Flächenanteil 5.6.3 Kornorientierung vs. rekombinationsaktiver Flächenanteil 5.6.4 Ätzgrubendichte vs. rekombinationsaktiver Flächenanteil 6. Diskussion der Ergebnisse 6.1 Einfluss des Kristallzüchtungsprozesses auf die Korngrößen-, die Kornorientierungs- und Korngrenztypverteilung 6.2 Einfluss der Kornstruktur auf den elektrisch aktiven Defektanteil 6.3 Einfluss der Kornorientierung auf den elektrisch aktiven Defektanteil 6.4 Einfluss der Kornstruktur auf die elektrische Aktivierung von Versetzungsclustern 6.5 Einfluss der Verunreinigungen auf die Solarzelleneffizienz 7. Zusammenfassung und Ausblick Verwendete Abkürzungen und Symbole Literaturverzeichnis Veröffentlichungen Betreute studentische Arbeiten Danksagung

Die Rekombinationsaktivität von Versetzungen und Korngrenzen in multikristallinem Silicium wird von Kupfer und anderen metallischen Verunreinigungen wie Eisen mitbestimmt. Das Hauptziel der Arbeit war es, die Verteilung von Kupfer und dessen Wirkung auf die Rekombinationsaktivität von Versetzungen und Korngrenzen genauer zu untersuchen. Dazu wurden optische und elektrische Untersuchungen an gezielt mit Metallen verunreinigten Modellmaterialien durchgeführt. Nicht alle Versetzungen sind rekombinationsaktiv. Es konnte gezeigt werden, dass der Anteil rekombinationsaktiver Versetzungen am Gesamtinventar und die Hintergrunddiffusionslänge von der Verunreinigung mit Metallen abhängig sind. Ergebnisse von Untersuchungen an Proben, die Diffusionsexperimenten unterzogen wurden, deuten auf unterschiedliches Ausscheidungsverhalten von Kupfer und Eisen hin sowie auf Wechselwirkungen mit Versetzungen und Korngrenzen, die mit der Diffusionstemperatur und den Abkühlbedingungen in Zusammenhang stehen.