Multicrystalline Silicon for Solar Cells: Process Development by ...

Multicrystalline Silicon for Solar Cells: Process Development by ... Multicrystalline Silicon for Solar Cells: Process Development by ...

atmsp.whut.edu.cn
from atmsp.whut.edu.cn More from this publisher
13.02.2013 Views

Multicrystalline Silicon for Solar Cells: Process Development by Numerical Simulation By Christian Häûler,* Gunther Stollwerck, Wolfgang Koch, Wolfgang Krumbe, Armin Müller, Dieter Franke, and Thomas Rettelbach Continuously improving crystallization conditions and solar cell processes have lead to steadily increasing efficiencies of solar cells based on multicrystalline silicon. There is, however, still an efficiency gap between mono- and multicrystalline silicon amounting to 1±2 % (absolute) depending on the cell process used. Topographies of the local solar cell performance clearly reveal that the main contribution to this efficiency gap is due to recombination-active dislocations present in multicrystalline silicon. A further improvement of the efficiencies attainable with multicrystalline solar cells therefore is achievable by a reduction of the dislocation density. Dislocations originate from thermal stress that originates from temperature gradients inside a multicrystalline ingot during crystallization and cooling. In order to reduce this thermal stress and consequently the dislocation density we employ a numerical simulation routine, the so-called virtual crystallization furnace, for perfect control of the temperature distribution during the entire ingot fabrication process. 1. Introduction The major challenge of today's photovoltaic technology, which is still mainly based on crystalline silicon wafer technology, is the increase of solar cell efficiencies while still using cost-effective, high-throughput, and large-scale processes. Employing multicrystalline silicon (mc-Si) wafers, solar cell conversion efficiencies up to the 15 % range are presently achieved even with industrial production scenarios for crystallization and solar cell manufacturing. Solar cell efficiencies, generally, are governed by the concentration and type of impurity atoms and the density and electrical activity of extended defects such as grain boundaries and dislocations. The requirement of both the increase of process speed and material quality necessitates nearly perfectly controlled temperature ± [*] Dr. C. Häûler, Dr. G. Stollwerck, Dr. W. Koch, Dr. W. Krumbe Bayer AG Rheinuferstraûe 7±9, D-47829 Krefeld (Germany) E-mail: christian.haessler.ch@bayer-ag.de Dr. A. Müller Deutsche Solar GmbH Postcode 1711, D-09587 Freiberg (Germany) Dr. D. Franke, Dr. T. Rettelbach ACCESS e.V. Intzestraûe 5, D-52072 Aachen (Germany) profiles. For the fabrication of large mc-Si ingots this perfect control is achieved by means of a so-called virtual crystallization furnace (VCF) that represents a detailed computer model of the actual crystallization chamber and is the basis of an accurate numerical simulation of the temperature distribution during crystal growth. The unique advantage of the VCF model isÐin addition to the ability of calculating macroscopic parameters such as temperatures, process times, etc.Ðthe possibility of predicting the distribution of even microscopically small crystal defects such as dislocations. Dislocations generated by thermal stress during crystal growth have been identified to crucially determine solar cell efficiencies and, therefore, have to be effectively suppressed in high-efficient solar cells. In this work it is shown that the optimization of the crystallization process within the VCF model may result in a reduction of the dislocation density by approx. 50 % that in turn is estimated to lead to an efficiency increase of 0.5±1 % absolute. 2. Silicon-Based Solar Cells The current strong growth in the photovoltaic market is still mainly based on the crystalline silicon wafer technology. Both, monocrystalline silicon and multicrystalline (mc) Si is em- Adv. Mater. 2001, 13, No. 23, December 3 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0935-9648/01/2312-1815 $ 17.50+.50/0 1815 RESEARCH NEWS

<strong>Multicrystalline</strong> <strong>Silicon</strong> <strong>for</strong> <strong>Solar</strong> <strong>Cells</strong>:<br />

<strong>Process</strong> <strong>Development</strong> <strong>by</strong> Numerical<br />

Simulation<br />

By Christian Häûler,* Gunther Stollwerck,<br />

Wolfgang Koch, Wolfgang Krumbe, Armin Müller,<br />

Dieter Franke, and Thomas Rettelbach<br />

Continuously improving crystallization conditions and solar cell processes have lead<br />

to steadily increasing efficiencies of solar cells based on multicrystalline silicon.<br />

There is, however, still an efficiency gap between mono- and multicrystalline silicon<br />

amounting to 1±2 % (absolute) depending on the cell process used. Topographies of<br />

the local solar cell per<strong>for</strong>mance clearly reveal that the main contribution to this efficiency<br />

gap is due to recombination-active dislocations present in multicrystalline silicon. A further improvement<br />

of the efficiencies attainable with multicrystalline solar cells there<strong>for</strong>e is achievable <strong>by</strong> a reduction of the dislocation<br />

density. Dislocations originate from thermal stress that originates from temperature gradients inside a multicrystalline<br />

ingot during crystallization and cooling. In order to reduce this thermal stress and consequently the<br />

dislocation density we employ a numerical simulation routine, the so-called virtual crystallization furnace, <strong>for</strong><br />

perfect control of the temperature distribution during the entire ingot fabrication process.<br />

1. Introduction<br />

The major challenge of today's photovoltaic technology,<br />

which is still mainly based on crystalline silicon wafer technology,<br />

is the increase of solar cell efficiencies while still using<br />

cost-effective, high-throughput, and large-scale processes.<br />

Employing multicrystalline silicon (mc-Si) wafers, solar cell<br />

conversion efficiencies up to the 15 % range are presently<br />

achieved even with industrial production scenarios <strong>for</strong> crystallization<br />

and solar cell manufacturing. <strong>Solar</strong> cell efficiencies,<br />

generally, are governed <strong>by</strong> the concentration and type of impurity<br />

atoms and the density and electrical activity of extended<br />

defects such as grain boundaries and dislocations. The<br />

requirement of both the increase of process speed and material<br />

quality necessitates nearly perfectly controlled temperature<br />

±<br />

[*] Dr. C. Häûler, Dr. G. Stollwerck, Dr. W. Koch, Dr. W. Krumbe<br />

Bayer AG<br />

Rheinuferstraûe 7±9, D-47829 Krefeld (Germany)<br />

E-mail: christian.haessler.ch@bayer-ag.de<br />

Dr. A. Müller<br />

Deutsche <strong>Solar</strong> GmbH<br />

Postcode 1711, D-09587 Freiberg (Germany)<br />

Dr. D. Franke, Dr. T. Rettelbach<br />

ACCESS e.V.<br />

Intzestraûe 5, D-52072 Aachen (Germany)<br />

profiles. For the fabrication of large mc-Si ingots this perfect<br />

control is achieved <strong>by</strong> means of a so-called virtual crystallization<br />

furnace (VCF) that represents a detailed computer model<br />

of the actual crystallization chamber and is the basis of an accurate<br />

numerical simulation of the temperature distribution<br />

during crystal growth. The unique advantage of the VCF<br />

model isÐin addition to the ability of calculating macroscopic<br />

parameters such as temperatures, process times, etc.Ðthe possibility<br />

of predicting the distribution of even microscopically<br />

small crystal defects such as dislocations. Dislocations generated<br />

<strong>by</strong> thermal stress during crystal growth have been identified<br />

to crucially determine solar cell efficiencies and, there<strong>for</strong>e,<br />

have to be effectively suppressed in high-efficient solar<br />

cells. In this work it is shown that the optimization of the crystallization<br />

process within the VCF model may result in a<br />

reduction of the dislocation density <strong>by</strong> approx. 50 % that in<br />

turn is estimated to lead to an efficiency increase of 0.5±1 %<br />

absolute.<br />

2. <strong>Silicon</strong>-Based <strong>Solar</strong> <strong>Cells</strong><br />

The current strong growth in the photovoltaic market is still<br />

mainly based on the crystalline silicon wafer technology. Both,<br />

monocrystalline silicon and multicrystalline (mc) Si is em-<br />

Adv. Mater. 2001, 13, No. 23, December 3 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0935-9648/01/2312-1815 $ 17.50+.50/0 1815<br />

RESEARCH NEWS


RESEARCH NEWS<br />

ployed <strong>for</strong> the fabrication of solar cells. [1] Specifically mc-Si,<br />

however, is able to meet the requirements of actual photovoltaics,<br />

which are a cost-effective large-scale production technique<br />

and high solar cell efficiencies. Consequently, Deutsche<br />

<strong>Solar</strong> GmbH in Freiberg, Saxony, established a high-throughput<br />

block-casting production facility <strong>for</strong> Baysix mc-Si. [2,3] The<br />

capacity of this plant amounts to an annual production of<br />

approx. 16 million wafers (in the year 2000) and will be<br />

tripled <strong>by</strong> 2003. State of the art are the multicrystalline ingots<br />

solidified in a Si3N4-coated quartz crucible with a height of<br />

approx. 30 cm and weights of more than 200 kg. After crystallization<br />

these ingots are divided into several columns, which<br />

subsequently are sliced into mc-Si wafers with areas between<br />

10 ” 10 cm 2 and 15 ” 15 cm 2 , and a typical thickness of<br />

300 lm.<br />

Besides continuously rising production capacities, improving<br />

crystallization conditions and solar cell processing sequences<br />

have led to steadily increasing photovoltaic conversion<br />

efficiencies of mc-Si solar cells. Based on Baysix, mc-Si<br />

efficiencies of 13±15 % and 16±17 % have been attained <strong>for</strong><br />

industrial-type fabrication and <strong>for</strong> solar cell processes on laboratory<br />

scale, respectively. [4±7]<br />

Generally, from a technology point of view, conversion efficiencies<br />

of solar cells based on mc-Si are depending on the<br />

crystallization process, the solar cell processing sequence, and<br />

the mutual adjustment of both process scenarios.<br />

3. Improving <strong>Solar</strong> Cell Efficiencies<br />

The implementation of optional defect passivation steps,<br />

e.g., hydrogen plasma treatments, and the temperature profiles<br />

of individual process steps, such as phosphorus diffusion<br />

<strong>for</strong> the fabrication of pn-junctions, play an important role in<br />

solar cell processing.<br />

Deutsche <strong>Solar</strong> GmbH, as a supplier of silicon wafers,<br />

focuses its research and development ef<strong>for</strong>ts on the crystallization<br />

conditions. The crystal defect scenario within the mc-Si<br />

ingots <strong>for</strong>ms a basis <strong>for</strong> the solar cell per<strong>for</strong>mance level that is<br />

achievable in subsequent solar cell fabrication. From the<br />

viewpoint of defect physics the following parameters are of<br />

specific importance:<br />

l the concentration and type of impurity atoms<br />

l the concentration and electrical activity of extended defects,<br />

e.g., dislocations and grain boundaries<br />

l the interaction of impurity atoms either with other impurity<br />

atoms (complex and cluster <strong>for</strong>mation) or with extended<br />

defects.<br />

The dependence of the solar cell efficiencies on the impurity<br />

concentration is trivial, but it nevertheless puts stringent<br />

requirements on the purity of feedstock materials and on the<br />

temperature stability of construction materials. Transition<br />

metals are considered as specifically detrimental to cell efficiencies:<br />

concentrations in the 10 12 cm ±3 range or even lower<br />

should be maintained <strong>for</strong> high quality solar-grade material in<br />

this case. A key role concerning metal impurity concentra-<br />

C. Häûler et al./<strong>Process</strong> <strong>Development</strong> of <strong>Solar</strong> <strong>Cells</strong> <strong>by</strong> Numerical Simulation<br />

tions is played <strong>by</strong> the segregation process. Elements with low<br />

segregation coefficients (e.g., transition metals) are effectively<br />

segregating in the silicon melt during crystallization. After<br />

completion of the ingot solidification (the crystallization front<br />

is moving from the bottom to the top) these impurities are<br />

efficiently accumulated in the ingot top region and can, there<strong>for</strong>e,<br />

be removed easily. It is estimated that the purification<br />

effect <strong>by</strong> segregation, i.e., the impurity concentration ratio<br />

between raw material and the final solar-grade silicon wafers<br />

is equal to 100±1000 or even above <strong>for</strong> elements with low segregation<br />

coefficients. This purification <strong>by</strong> solidification, however,<br />

is affected <strong>by</strong> the crystallization velocity and generally<br />

becomes less effective <strong>for</strong> higher crystallization speeds.<br />

Extended defects, i.e., grain boundaries and dislocations,<br />

may carry electrical charge and consequently act as highly efficient<br />

recombination centers <strong>for</strong> minority charge carriers.<br />

The electrical activity of these defect centers is strongly influenced<br />

<strong>by</strong> their interaction with impurity atoms and usually<br />

tends to increase with increasing impurity concentration.<br />

Nevertheless, it was experimentally verified that the electrical<br />

activity of grain boundaries is depending on the crystallization<br />

conditions, too, and can be considerably reduced if a strictly<br />

planar solidification front is preserved during crystallization. [8]<br />

There<strong>for</strong>e, <strong>for</strong> actual mc-Si with a grain size of several millimeters<br />

up to the centimeter region and a low grain boundary<br />

activity, grain boundaries do not play a significant role concerning<br />

attainable solar cell efficiencies.<br />

In contrast to grain boundaries, crystal dislocations have<br />

been identified to be the most efficiency-relevant defect centers<br />

in mc-Si <strong>for</strong> photovoltaics. [9] Dislocations are generated<br />

during ingot growth and cooling <strong>by</strong> mechanical stress that is<br />

caused <strong>by</strong> temperature gradients inside the solidified ingot.<br />

An example of the importance of crystal dislocation is given<br />

in Figure 1, which depicts a typical topography of the short<br />

circuit current of a mc-Si solar cell. Microscopic optical inves-<br />

Fig. 1. Typical LBIC (light beam induced current) topography of the short<br />

circuit current of a solar cell based on mc-Si. The sample size is 5 ” 5 cm 2 , redcolored<br />

and dark regions indicate high and low current output, respectively.<br />

Reduced current output generally is caused <strong>by</strong> recombination-active defect<br />

centers. Microscopic investigations reveal the line-shaped structures to be due<br />

to grain boundaries whereas the more extended defect regions originate from<br />

crystal dislocations.<br />

1816 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0935-9648/01/2312-1816 $ 17.50+.50/0 Adv. Mater. 2001, 13, No. 23, December 3


C. Häûler et al./<strong>Process</strong> <strong>Development</strong> of <strong>Solar</strong> <strong>Cells</strong> <strong>by</strong> Numerical Simulation<br />

tigations reveal that the dark regions in the topography, which<br />

correspond to a reduced current output, are perfectly correlated<br />

to high dislocation density regions.<br />

From the above considerations it can be concluded that a<br />

perfect control of the temperature distribution inside the crystallization<br />

furnace is essential <strong>for</strong> high-throughput production<br />

of mc-Si <strong>for</strong> high-efficiency solar cells. The most important parameters<br />

that have to be thoroughly adjusted are<br />

l the planarity of the solidification front <strong>for</strong> a low activity of<br />

extended defects<br />

l the crystallization speed affecting impurity segregation<br />

and process times<br />

l the temperature gradients and thermal stress leading to<br />

dislocation generation.<br />

4. Simulating the <strong>Process</strong><br />

In order to optimize temperature control inside our crystallization<br />

chambers, we generally established a numerical simulation<br />

[10] of the temperature distribution as a tool <strong>for</strong> process<br />

development. The so-called VCF model [11] is employed to<br />

simulate the three-dimensional, time-dependent temperature<br />

distribution within our crystallization chambers. Basically, the<br />

VCF model consists of three major parts, which are a geometry<br />

and thermophysical data set, the software package, and<br />

the user interface.<br />

The geometry data set contains the crystallization furnace<br />

(including insulation materials, heating and cooling systems,<br />

the quartz crucible, and the silicon material) designed into a<br />

400 000 finite elements system (see Fig. 2). The thermophysical<br />

data of more than 30 different materials given as a function<br />

of temperature is employed <strong>for</strong> the simulation.<br />

Fig. 2. Numerical simulation results <strong>for</strong> the time-dependent temperature distribution<br />

during crystallization of a >200 kg mc-Si ingot (cut along the middle<br />

plane). The size of the ingot is approx. 55 ” 55 ” 30 cm 3 . The crystallization furnace<br />

chamber including heaters, insulation materials, the quartz crucible and<br />

the silicon material is designed into a 400 000 finite elements system and the<br />

thermophysical data of approx. 30 different materials are taken into account in<br />

order to simulate the time-dependent temperature profile of the entire silicon<br />

block from basic process parameters (heating power, cooling water temperature).<br />

The light yellow line represents the solidification front, i.e., the interface<br />

between liquid (above light yellow line) and solidified (below) silicon.<br />

The self-written software package CASTS (computer aided<br />

solidification technologies) is specifically designed <strong>for</strong> the simulation<br />

of technical processes. The starting point is the simulation<br />

of the temperature distribution taking into account heating<br />

power, heat flows, and the release of the latent thermal heat.<br />

The thermal gradients and the dislocation density distribution<br />

inside the silicon material are calculated from the temperature<br />

distribution. The only input parameters of the simulation<br />

are the heating power and the cooling water temperature of<br />

the crystallization chambers of our block casting facility.<br />

The accurate numerical simulation of the planarity and the<br />

speed of the solidification front (see points 1 and 2 above) are<br />

challenging problems, but nevertheless can be solved <strong>by</strong> adjusting<br />

the parameters of the numerical simulation routines in<br />

accordance with experimental data accessible <strong>by</strong>, e.g., temperature<br />

monitoring. The prediction of dislocation densities, however,<br />

is still a considerably more complicated case, because<br />

the temperature distribution only represents the starting point<br />

here, whereas the final goal is to calculate the distribution of<br />

microscopic defects in large >200 kg silicon crystals.<br />

Crystal dislocations are taken into account <strong>by</strong> considering<br />

thermal gradients and the generation of thermal stress, which<br />

is known to be the driving <strong>for</strong>ce <strong>for</strong> dislocation motion. Due<br />

to this motion, dislocation multiplication takes place during<br />

crystallization and cooling of the ingot. The multiplication<br />

process leads to a final, locally varying dislocation density at<br />

room temperature that essentially depends on the thermal<br />

history of the mc-Si material. In order to calculate the dislocation<br />

density generation <strong>by</strong> thermal stress we employ one of<br />

the basic models of dislocation multiplication that was introduced<br />

<strong>by</strong> Alexander and Haasen in 1968. [12] The basic<br />

assumptions of this model translate into the following two<br />

equations:<br />

NÇ = KNvs eff<br />

seff = sth ± spl ± A N<br />

p<br />

where, N and N Ç denote the dislocation density and dislocation<br />

multiplication rate, respectively. The parameter v denotes the<br />

velocity of the dislocations moving through the silicon crystal<br />

and s eff stands <strong>for</strong> the effective stress given as a function of<br />

the thermally induced stress sth, the plastic de<strong>for</strong>mation spl,<br />

and the dislocation density N according to Equation 2<br />

(further parameters K and A are material constants).<br />

In other words, the basic idea of our approach (see Eq. 1)<br />

considers a dislocation multiplication rate N Ç that is proportional<br />

to the dislocation density N and the respective dislocation<br />

velocity v and that is driven <strong>by</strong> an effective stress s eff. This<br />

effective stress s eff in turn is reduced <strong>by</strong> the plastic de<strong>for</strong>mation<br />

of the crystal and the dislocation generation itself and,<br />

there<strong>for</strong>e, is lower than the pure thermal stress sth arising<br />

from the temperature gradients (see Eq. 2).<br />

Generally, each parameter given in Equations 1 and 2 is<br />

dependent on temperature, time, and the position inside the<br />

Adv. Mater. 2001, 13, No. 23, December 3 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0935-9648/01/2312-1817 $ 17.50+.50/0 1817<br />

(1)<br />

(2)<br />

RESEARCH NEWS


RESEARCH NEWS<br />

mc-Si ingots. Thus, finally, a system of several combined differential<br />

equations is numerically solved <strong>for</strong> calculating the<br />

dislocation density distribution.<br />

Our self-developed numerical simulation routines were verified<br />

<strong>by</strong> comparing their results to experimental data. Structural<br />

crystal defects such as dislocations are experimentally<br />

accessible <strong>by</strong> counting micrometer small etch pits with the<br />

help of a microscope after appropriate polishing and chemical<br />

etching steps. A unique measurement system, capable of fast<br />

etch pit density (EPD) mappings even over large areas (up to<br />

10 ” 10 cm 2 wafers), was, there<strong>for</strong>e, developed <strong>for</strong> the experimental<br />

verification of the simulation results. It turned out that<br />

the simulated dislocation density distribution even in large<br />

200 kg multicrystalline ingots is in a sufficiently good agreement<br />

with the experimental results (see Fig. 3) in order to be<br />

employed <strong>for</strong> further process optimization.<br />

Fig. 3. Numerical simulation results <strong>for</strong> the dislocation density as a function of<br />

the block height in the center position of a mc-Si ingot with a weight of more<br />

than 200 kg. The simulation starts from basic parameters of the crystallization<br />

process (heating power, cooling water temperature) simulating first the timedependent<br />

temperature distribution within the crystallization furnace. By employing<br />

a physical model describing dislocation multiplication <strong>by</strong> thermal stress<br />

within the solidified silicon the final dislocation density is calculated. The experimental<br />

data was obtained <strong>by</strong> counting micrometer small etch pits originating<br />

from crystal dislocations with an optical microscope. Due to the inhomogeneous<br />

distribution of the dislocations, the measurements were averaged over<br />

an area of approx. 25 cm 2 using an image analysis system.<br />

5. Conclusions<br />

The elaboration of the dislocation density distribution<br />

within multicrystalline ingots led to the conclusion that the<br />

generation of dislocations during the fabrication process of<br />

mc-Si <strong>for</strong> photovoltaics can be divided into three important<br />

phases [13] (see Fig. 4). First, a relatively strong increase of the<br />

dislocation density is obtained in the vicinity of the solidification<br />

front immediately after the phase change from liquid to<br />

solid. The amount of dislocations generated in this first phase<br />

depends on the solidification velocity because increased solidification<br />

velocities necessitate high thermal gradients within<br />

the solidified silicon. In the second phase of dislocation generation,<br />

dislocation multiplication <strong>by</strong> thermal gradients is prevailing,<br />

whereas the third lower temperature phase


C. Häûler et al./<strong>Process</strong> <strong>Development</strong> of <strong>Solar</strong> <strong>Cells</strong> <strong>by</strong> Numerical Simulation<br />

mc-Si <strong>for</strong> solar cells. It is expected that the numerical simulation<br />

tool will be specifically valuable <strong>for</strong> future process developments<br />

addressing even larger ingots and process speeds<br />

aiming at still enhanced solar cell efficiencies.<br />

±<br />

[1] See, e.g., M. A. Green, <strong>Solar</strong> <strong>Cells</strong>, Prentice-Hall, Inc., Englewood Cliffs,<br />

NJ 1982.<br />

[2] W. Koch, C. Häûler, H.-U. Höfs, A. Müller, I. A. Schwirtlich, Solid State<br />

Phenom. 1997, 57±58, 401.<br />

[3] C. Häûler, W. Koch, W. Krumbe, S. Thurm, A. Müller, I. A. Schwirtlich,<br />

in Proc. of the 2nd World Conf. and Exhibition on Photovoltaic <strong>Solar</strong><br />

Energy Conversion Vienna (Eds: J. Schmid, H. A. Ossenbrink, P. Helm,<br />

H. Ehmann, E. D. Dunlop), European Commission 1998, p. 1886.<br />

[4] H. Lautenschlager, F. Lutz, E. Schäffer, C. Schetter, U. Schubert, R.<br />

Schindler, in Proc. of the 14th European Photovoltaic <strong>Solar</strong> Energy Conf.<br />

Barcelona (Eds: H. A. Ossenbrink, P. Helm, H. Ehmann), H. S. Stephens<br />

& Associates 1997, p. 1358.<br />

[5] A. El Moussaoui, A. Moehlecke, C. del Caæizo, A. Luque, in Proc. of the<br />

2nd World Conf. and Exhibition on Photovoltaic <strong>Solar</strong> Energy Conversion<br />

Vienna (Eds: J. Schmid, H. A. Ossenbrink, P. Helm, H. Ehmann, E. D.<br />

Dunlop), European Commission 1998, p. 747.<br />

______________________<br />

[6] R. Lüdemann, A. Hauser, R. Schindler, in Solid State Phenom. Proc. of<br />

the 2nd World Conf. and Exhibition on Photovoltaic <strong>Solar</strong> Energy Conversion<br />

Vienna (Eds: J. Schmid, H. A. Ossenbrink, P. Helm, H. Ehmann, E. D.<br />

Dunlop), European Commission 1998, p. 1638.<br />

[7] F. Duerinckx, J. Szlufcik, J. Nijs, R. Mertens, C. Gerhards, C. Marckmann,<br />

P. Fath, G. Willeke, in Proc. of the 2nd World Conf. and Exhibition on<br />

Photovoltaic <strong>Solar</strong> Energy Conversion Vienna (Eds: J. Schmid, H. A. Ossenbrink,<br />

P. Helm, H. Ehmann, E. D. Dunlop), European Commission<br />

1998, p. 1248.<br />

[8] W. Koch, W. Krumbe, I. A. Schwirtlich, in Proc. of the 11th European<br />

Photovoltaic <strong>Solar</strong> Energy Conf. (Eds: L. Guimaraes, W. Palz, C. de Reyff,<br />

H. Kiess, P. Helm), Harwood Academic Publishers 1992, p.518.<br />

[9] See, e.g., H. El Ghitani, M. Pasquinelli, S. Martinuzzi, J. Phys. III 1993, 3,<br />

1941.<br />

[10] G. Ehlen, A. Ludwig, P. R. Sahm, Giesserei<strong>for</strong>schung 1999, 51, 56.<br />

[11] I. Steinbach, D. Franke, W. Krumbe, J. Liebermann, in Proc. of the 1st<br />

World Conf. on Photovoltaic <strong>Solar</strong> Energy Conversion Hawaii, IEEE<br />

Electron Devices Society 1994, p. 1270.<br />

[12] H. Alexander, P. Haasen, Solid State Phys. 1968, 22, 27.<br />

[13] D. Franke, C. Häûler, W. Koch, J. Liebermann, in Proc. of the 2nd World<br />

Conf. and Exhibition on Photovoltaic <strong>Solar</strong> Energy Conversion Vienna<br />

(Eds: J. Schmid, H. A. Ossenbrink, P. Helm, H. Ehmann, E. D. Dunlop),<br />

European Commission 1998, p. 108.<br />

Adv. Mater. 2001, 13, No. 23, December 3 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0935-9648/01/2312-1819 $ 17.50+.50/0 1819<br />

RESEARCH NEWS

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!