narrower efficiency distribution for multicrystalline silicon solar ... - ECN

20 downloads 0 Views 111KB Size Report
May 1, 2002 - optimised for surface passivation. Using the lifetime measuring method, the Fe ..... consulting, Boulder, CO. ACKNOWLEDGEMENTS. This work ...
ECN-RX--02-025

NARROWER EFFICIENCY DISTRIBUTION FOR MULTICRYSTALLINE SILICON SOLAR CELLS BY DOUBLE-SIDE EMITTER DIFFUSION M.J.A.A. Goris A.W. Weeber J.H. Bultman

This paper has been presented at the 29th IEEE Photovoltaic Specialists Conference, 20 - 24 May 2002, New Orleans, USA

MAY 2002

Paper No IP3.6 Page 1 of 4

Presented at 29th IEEE, New Orleans, Louisiana, 2002

Narrower Efficiency Distribution for Multicrystalline Silicon Solar Cells by Double-side Emitter Diffusion M.J.A.A. Goris, A.W. Weeber, J.H. Bultman ECN Solar Energy, PO Box 1, 1755 ZG Petten, The Netherlands Phone: +31224564505, Fax: +31224568214, E-mail: [email protected]

SUMMARY Multicrystalline silicon solar cells produced with double side emitter diffusion are compared with single side emitter diffused cells. Double side emitter diffusion yields in improved current and voltage for all materials of four multicrystalline silicon wafer suppliers. Poor quality material improves approximately 3% more than good quality material. This results in a narrower efficiency distribution over all material. No relation between substitutional carbon concentration, interstitial oxygen concentration and FeB concentration and improved current due to double side diffusion was found. INTRODUCTION A major drawback of multicrystalline silicon (mcSi) solar cell processing is that it results in a very wide efficiency distribution compared to mono-crystalline silicon solar cells. This is caused by a wide distribution in material quality. Solar cell manufacturers handle this issue by demanding from wafer suppliers a certain distribution of material quality measured by effective lifetime, carbon or oxygen content or doping concentration. The wafer supplier then cuts off a certain amount of low quality material, which is not used or sold as a lower classified material. Also, cell manufacturers sell products with a wide range in power output. To avoid these negative aspects, researchers have been looking for improved processing conditions to obtain a narrower efficiency distribution. During the course of earlier work [1], we discovered such an improvement by using double side instead of single side emitter diffusion. This process improves the low quality material substantially. We do not have a proven theory to explain this improvement. The objective of this study is to find an explanation of why low quality material is improved. Also, we want to show that this improvement occurs for all kinds of material, independent of supplier. Therefore, we have collected data for mcSi wafers from four wafer suppliers and we have related these data to the solar cell characteristics after single or double side emitter diffusion. EXPERIMENTAL METHODS 2

For this experiment 100 cm mcSi wafers collected from four wafer suppliers are used. The wafers are selected from different positions in the

ingot, namely from two columns (one from the middle and one from the edge) and five heights (1%, 25%, 50%, 75% and 99% of the maximum height) For all experiments the phosphorus source is either applied on one or on both sides of the wafer by spin-coating. Cell processing was performed according to the so-called firing-through scenario (table 1). The I-V-characteristics are measured and the data are analysed with the analysis of variance method for mcSi solar cells as described by Weeber and Sinke [2, 3]. Also, spectral response is measured of selected cells. Table 1. Scenario for firing through process Firing through scenario Saw damage etching Emitter processing PECVD- SixNy:H Metallisation (front/back) Firing Furthermore, a selection of neighbouring wafers of the cell experiment has been extensively characterised by lifetime and FTIR measurements. Concentrations of substitutional carbon and interstitial oxygen in wafers are determined by Fourier Transform Infrared Spectroscopy (FTIR). Before measuring, no thermal treatment is applied and the wafers are polished by chemical etching. The calibration constants used for conversion of absorbency to concentration were the ASTM-1995 -1 standards: 1.64 ppma-cm for carbon (at ~605 cm ), -1 and 6.28 ppma-cm for oxygen (at ~1107 cm ). The effective thickness of the floatzone reference wafer and the test wafers were matched using the Si phonon bands [4]. Carbon concentrations are systematically underestimated by approximately 1030% due to the relatively small thickness of the wafer that does not allow a high-resolution FTIR. Effective lifetime of the minority carriers is measured with the Sinton QSS-equipment [7,8]. The measurements were performed after saw damage etch and applying a Remote PECVD SixNy layer optimised for surface passivation. Using the lifetime measuring method, the Fe content in the wafer can be estimated by determining the change in lifetime due to FeB pair dissociation after light soaking [5,6]. The change in inverse lifetime is proportional to the number of dissociated FeB pairs (i.e., the interstitial Fe concentration: by far most of the interstitial Fe is paired at room temperature in thermal equilibrium).

Paper No VC1.36 Page 2 of 4

RESULTS

2

supplier-A supplier-B supplier-C supplier-D

1.5

delta Voc (%)

Solar cell processing The average values of the I-V-characteristics of the solar cells with single side emitter are shown in figure 1 (Isc) and 2 (Voc). In figure 3 and 4 the gains in current and voltage for double and single side emitter diffusion (determined as double side minus single side divided by single side times 100%) are presented.

1

0.5 0 0

3.1

20

40 60 wafer position(%)

80

100

Isc (A)

3

Fig. 4. gains in Voc for double side diffused emitters compared to single side diffused emitters

2.9

supplier-A supplier-B supplier-C supplier-D

2.8 2.7 2.6 0

20

40

60

80

100

waferposition(%)

Fig. 1. short current of solar cells with single sided emitters

0.615

Voc (V)

0.605 0.595

supplier-A supplier-B supplier-C supplier-D

0.585 0.575 0

20

40

60

80

100

wafer position(%)

Fig. 2. open circuit voltage of solar cells with single sided emitters 4

supplier-A supplier-B supplier-C supplier-D

delta Isc (%)

3

2

1

0 0

20

40 60 wafer position (%)

80

100

Fig. 3. gains in Isc for double side diffused emitters compared to single side diffused emitters

The plotted values in the figures are averaged over four samples (two neighbouring wafers from two different columns). All the data-points have a certain variation that is not plotted in these figures. The current and voltage of wafers from almost all positions is for double side emitter diffusion higher than for single side emitter diffusion. Material from the bottom improves for all suppliers about 3% in current and for supplier A and D about 1.5% in voltage. Material from the top improves for supplier B, C and D about 2.5% in current and about 1% in voltage. Material from the top of supplier A is of the same quality as inside the ingots, and shows hardly any improvement. The combined effect of gains due to double side diffusion is that the efficiency distribution becomes narrower because low quality cells improve more than good quality cells. For instance, for material B, the efficiency of cells varies by 14% with single side diffusion and by 10% for double side diffusion, which is a reduction in variation by 40%. All combined experimental data is analysed using Stattgraphics [3]. As indicated by Least Significant Difference (LSD) analysis, double side emitter diffusion improves short circuit current by 2.5% compared to single side diffusion with a probability of at least 95%. It improves open circuit voltage by 1% compared to single side diffusion with a probability of at least 90%. Spectral responses are measured on solar cells originating from bottom and top of a column. Internal quantum efficiencies have been determined from spectral responses and reflection curves. In figure 5, the gain (determined as double side minus single side times 100% and divided by single side) in quantum efficiency due to double side diffusion compared to single side diffusion is shown for bottom materials (for material B, position 25% is used). In figure 6, the gain in quantum efficiency is shown for top material. The gains in quantum efficiency show that the material quality in the bottom of a column improves for double side emitter diffusion compared to single side emitter diffusion for all four suppliers. This is in agreement with the gains in short circuit current. For material in the top of a column, only material A shows less gain in red response due to double side diffusion.

Paper No VC1.36 Page 3 of 4

40

supplier A effective lifetime (µs)

30 delta IQE (%)

120

supplier B supplier C supplier D

20 10 0 400 -10

supplier A supplier B supplier C supplier D

90 60 30 0

600

800

1000

0

20

40

wavelength (nm)

20

120 80

10 0 400 -10

40

delta IQE (%)

delta IQE (%)

160

supplier A supplier C supplier D supplier B

30

0 600

800

80

100

Fig. 7. effective lifetime for wafers from several positions and suppliers

FeB content (*E+11cm-3)

Fig. 5. gains in IQE for solar cells extracted from the bottom of column (except material B is from 25%)

40

60

wafer position (%)

1000

100

supplier A supplier B supplier C supplier D

80 60 40 20 0 0

-40

20

40

60

80

100

wafer position (%)

wavelength (nm) Fig. 6. gains in IQE for solar cells extracted from the top of column

Fig. 8. estimate of FeB content 4e13*delta(1/tau) for several material

assuming

Effective lifetime and FeB concentration The effective lifetime of the minority carriers was measured on wafers from one column received from all suppliers (figure 7). The measured lifetimes show some correspondence with the quality of the cells, low lifetime material from bottom leads to low quality cells. This is not the case for the top of the ingot, where very low lifetimes do not necessarily lead to poor quality cells. According to the method explained before, a rough measure of the FeB concentration [ref5] can be calculated from the difference in lifetime before and after light soaking. The estimated FeB concentration for several materials is plotted in figure 8. All ingots show increased contamination and reduced lifetime at top and bottom. The increased contamination is due to solid state diffusion of impurities from crucible and/or coating (bottom), or back diffusion of the segregated impurities in the top of the ingot. The contamination in the central region of the ingots is due to the contamination present in the liquid silicon during crystallisation, in combination with the limited segregation of Fe. Of course, in addition to the FeB pairs, precipitated Fe and other metallic impurities may be present in the wafers. This will vary from ingot to ingot, depending on aspects

such as purity of crucible, purity and effectiveness of crucible coating, and purity of feedstock. substitutional carbon and interstitial oxygen The results from FTIR measurements are presented in figure 9 for the substitutional carbon content and figure 10 for the interstitial oxygen content measurement.

sub. carbon (ppma)

16 12

supplier A

supplier B

supplier C

supplier D

8 4 0 0

20

40 60 wafer position (%)

80

100

Fig. 9. substitutional carbon content in wafers from several positions in a column and several suppliers.

Paper No VC1.36 Page 4 of 4

int. oxygen (ppma)

25

supplier A supplier B supplier C supplier D

20 15 10 5 0 0

20

40 60 wafer position (%)

80

100

Fig. 10. interstitial oxygen content in wafers from several positions in a column and several suppliers. For materials from supplier A and B, the carbon content is strongly dependent on the position in the ingot, with the highest carbon content at the top of the ingot. For materials from supplier C and D, the carbon content is independent on the position in the ingot, and in general higher than for materials from suppliers A and B. Also in the FTIR-data from material B for wafers of position higher than 50% signals are observed that probably correspond to carbon-oxygen precipitates. Interstitial oxygen in material of supplier C is independent of position in the ingot, unlike the material of other suppliers. Also, the oxygen content for material from supplier C is much lower than all the other materials. For materials from suppliers A, B, and D, the oxygen content is dependent on the position in the ingot, with the highest values in the bottom. Wafers of supplier A contain a very high interstitial oxygen content in the bottom At the top of the column, oxygen contents are lowest and are the same for all suppliers. DISCUSSION The most significant result of our study is that a gain in short circuit current and open circuit voltage occurs for all wafer suppliers. For all suppliers the low quality material in bottom and top of a column improves after double side emitter diffusion. For suppliers B and C the improvement is most significant in the top, for A and D in the bottom. The quality of material from supplier B was exceptionally poor in the bottom. From the IQE results it is obvious that the gain in current and voltage is mainly explained by improvement of the material quality. The fact that the main improvement occurs in the top and bottom of all columns, suggests that it is related to gettering of impurities. However, the improvement is not clearly related to the measured FeB, Oi and Cs concentrations. The improvement can not be explained by the interstitial oxygen concentration because the concentration is low for material from the top section. Oxygen may play some role, e.g. in the formation and stability of precipitates, but it is not clear how important this role is. The carbon content, which is often relatively flat, is probably also not related to the improvement. For the FeB content it is shown that in the middle section the concentration is low. Nevertheless the material from this section improves after applying a double side emitter diffusion. Based on the collected wafer

characteristics it is therefore not possible to explain the exact reason why double side emitter diffusion improves the current and voltage. Probably additional contamination in the wafers is present, depending on the feedstock and crucible materials used for the crystal growth process. These can be in the form of precipitates, which dissolve during high-temperature processing. Measuring wafer characteristics after solar cell processing, like effective lifetime, substitutional carbon, interstitial oxygen and FeBcontent during processing may provide more information on this. Measuring impurity concentration with DLTS, before and after thermal processes, may provide additional information. CONCLUSION Double side diffusion narrows the cell efficiency distribution considerable compared to single side diffusion. This is due to the fact that cell efficiency improves more for low quality material than for high quality material due to the change from single side emitter diffusion to double side emitter diffusion. Double side emitter diffusion improves the current from low quality material about 3% in the bottom and 2% in the top. This improvement can not be explained by FeB concentration, substitutional C concentration, or interstitial O concentration. REFERENCES [1] M.J.A.A. Goris, A.W. Weeber, W. Jooss F. Huster; Comparison of emitters diffused Using an IR belt furnace or a POCl3 system, presented th at 17 EPVSEC, Munich, 2001 th [2] A.W. Weeber, W.C. Sinke, Proc. 25 IEEE Photovoltaic Specialist Conference, 1996, pp. 565 [3] http://www.statgraphics.com [4] A. Kempf, P. Bloechl, A. Huber, "Influence of metal contamination on minority carrier recombination lifetime in silicon", Electrochemical Society Proceedings 13 1989, pp. 221-229. [5] D. Macdonald, A. Cuevas, and J. Wong-Leung, Capture cross-sections of the acceptor level of iron-boron pairs in p-type silicon by injection-level dependent lifetime measurements", J. Appl. Phys. 89, 2001, pp. 7932. [6] for more extensive description, use, and interpretation of this technique for material and processing analysis, see L.J. Geerligs et al., to be published [7] A. Cuevas, D. Macdonald, M. Kerr, C. Samundett th and A. Sloan; Proceedings 28 IEEE PV specialists conference. Anchorage, 2000 [8] Sinton; WCT-100 photoconductance tool, Sinton consulting, Boulder, CO. ACKNOWLEDGEMENTS This work has been financially supported by the NOZPV programme from the Netherlands Agency for Energy and Environment (Novem) and by the Ministry of Economics Affairs. Much thanks to L.J. Geerligs and A. Ouchchahd for their assistance.