- Email: [email protected]
Recombination at Metal-Emitter Interfaces of Front Contact Technologies for Highly Efficient Silicon Solar Cells
Available online at www.sciencedirect.com
Energy Procedia 8 (2011) 115–121
SiliconPV: 17-20 April 2011, Freiburg, Germany
Recombination at Metal-Emitter Interfaces of Front Contact Technologies for Highly Efficient Silicon Solar Cells T. Fellmeth*, A. Born, A. Kimmerle, F. Clement, D. Biro, R. Preu Fraunhofer Institute for Solar Energy Systems ISE, Heidenhofstr. 2, 79110 Freiburg, Germany
Abstract We present an experimental approach to extract the dark saturation current density j0e-met at the emitter-metal interface of the front contact. For this purpose, 2x2 cm² sized silicon solar cells have been realized featuring different metallization fractions FM. By simply applying the one-diode-model, the dark current density j01 is determined from the open circuit voltage Voc. From the slope of the j01 over FM plot, j0e-met is extracted. However, this is only valid if the dominant recombination mechanism at Voc features a diode character that is close to unity. Hence, the local ideality factor m is determined from the suns-Voc-curve indicating the required value close to one. Three main effects are observed. First, the metallization methods which are compared show different influences on j0e-met on the same emitter configuration. Second, an emitter drive-in due to an additional short thermal oxidation lowers j0e-met. Also, the field-effect passivation of the highly n-doped selective emitter decreases j0e-met effectively. By combining the field effect passivation with a short drive-in step the very low value of j0e-met = 549 fA/cm² is reported.
© responsibility of of SiliconPV 2011. © 2011 2010 Published Publishedby byElsevier ElsevierLtd. Ltd.Selection Selectionand/or and/orpeer-review peer-reviewunder under responsibility SiliconPV. Keywords: silicon solar cell; emitter; metallization
1. Introduction Properties of the front metallization of solar cells are not only determined by conductivity, contact resistance and light shading. The metal-silicon interface of a phosphorus doped emitter provides a highly recombination active surface for minority charge carriers [1]. Hence, the impact of this loss mechanism needs to be taken into account by evaluation of front metallization methods. Unfortunately, emitter-metal recombination can not be measured directly like conductivity or contact resistance. Furthermore, the
* Corresponding author, Tel.: +49(0) 7 61/ 45 88-5652 E-mail: [email protected]
1876–6102 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of SiliconPV 2011. doi:10.1016/j.egypro.2011.06.111
116
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
degree of surface recombination always depends on the emitter profile itself [2]. The terms “opaque” and “transparent” emitter have been introduced [2] to indicate where minority charge carrier recombination in the emitter is dominant. In an “opaque” n-type emitter, holes recombine mostly in the volume however, in “transparent” emitters holes likely reach the surface and recombine there. The latter emitters are very sensitive to surface recombination properties, hence to the front metallization. In order to quantify the recombination at metal-emitter interfaces, the emitter dark saturation current density j0e-met is introduced analogously to the emitter dark saturation current density j0en+ of the passivated emitter area. Both share the same diode characteristic of n = 1. Hence, emitter recombination can be described by just adding them, weighted by area. In order to reduce recombination at the metal-emitter interface, there are three approaches in principle. The first is to simply reduce metallization fraction FM, secondly, by utilizing a metallization method that induces a low j0e-met and thirdly by utilizing an opaque emitter under the metal contact area. The latter can be done by using the selective emitter approach. The highly n-doped area under the front contact shields the minority charge carriers from reaching the surface more effectively mainly due to a higher field-effect passivation [3]. 2. Experimental approach The goal of this experiment is the extraction of the emitter dark saturation current density j0e-met at the metal contact area. Therefore, 2x2 cm² sized Si-FZ, p-type 1 :cm solar cells have been processed featuring different emitter profiles and front metallization. The screen-printing metallization is presented featuring two different pastes: x SP1: screen-printing x SP2: screen-printed seed layer plus silver light induced plating (Ag-LIP) [4] In Figure 1, a scan of a wafer is presented where eight cells are located. The front metallization fraction has been varied, resulting in varying finger pitch from 2 to 0.4 mm.
Figure 1: On the left a scan of a wafer (12.5x12.5cm²) is displayed showing eight solar cells exhibiting different finger pitches starting from 2 mm equating to C1 to 0.4 mm, that equating to C8. On the right side two cross-sections of a device featuring the SP2metallzation is displayed exhibiting in this case a selective (n++-doped) and oxidized (SiO2) emitter. Also the different emitter regions are highlighted that exhibit different dark saturation current densities.
117
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
3. Results 3.1. j0e and SIMS analysis Different emitters have also been incorporated into the experiment which is presented in Table 1. j0e-n+ is extracted in low level injection from symmetrical emitter samples using the QSSPC technique [5]. By applying the one-diode-model, the Voc limit Voc-lim of the emitters is calculated assuming a short circuit current density jsc of 35 mA/cm². For passivation of the emitter, PECVD silicon nitride (SiNx) is used. Additionally, there are samples where 10 nm of SiO2 is grown during a dry thermal oxidation at 840 °C before the deposition of the SiNx-layer is done. This process is applied on the emitter with a sheet resistance of 110 :/sq, leading to a reduction of the sheet resistance to 80 :/sq and a reduction of j0en+ from 264 to 160 fA/cm² [6]. In Figure 2 SIMS profiles of the emitters introduced in Table 1 are displayed. The phosphorus silicate glass (PSG) laser doped emitter exhibits a low sheet resistance due to the high degree of phosphorus within the first 150 nm. Also, the effect of the thermal oxidation is observable at the profile of the emitter with a sheet resistance of 80 :/sq. Compared to the 110 :/sq emitter the front phosphorus level is reduced and the tail of the profile is extended.
Rsh
j0e-n+ (fA/cm²)
Voc-lim (mV)
Passivation
(ȍ/sq) 72
347
651
SiNx
78
299
655
SiNx
110
264
658
SiNx
110 ¼ 31*
j0e-n++ = 575
637
SiNx
80
160
671
SiO2 + SiNx
110 ¼ -*
j0e-n++ = 397
645
SiO2 + SiNx
*n++-Emitter due to PSG laser doping.
1E22
1E22 PSG laser doped 31 :/sq 80 :/sq oxidized 78 :/sq 72 :/sq 110 :/sq
1E21 -3
Phosphor atoms (cm )
Table 1. Sheet resistances and emitter dark saturation current densities of the symmetrical samples. A photo generated current density of 35 mA/cm² and a ni in the base of 9.14*109 cm-3 are assumed. The evaluation is carried out at an injection level of 'n = 1015 cm-3.
1E20
1E21 1E20
1E19
1E19
1E18
1E18
1E17
1E17
1E16
1E16
1E15 0.0
0.1
0.2
0.3 0.4 depth (μm)
0.5
1E15 0.6
Figure 2: SIMS profiles of the emitters listed in Table 1. The “PSG laser doped” profile is generated by laser induced diffusion from the 110 :/sq emitter with PSG-layer acting as a phosphorus source. The PSG laser doped profile after thermal oxidation is not displayed.
3.2. Determination of the local ideality factor According to the one-diode-model, where VT denotes the thermal voltage at 25 °C and jph the photo generated current density that is set equal to jsc, Voc is directly correlated with j01.
Voc
§ j · VT Ln ¨¨ ph ¸¸ © j01 ¹
(1)
This relation is only valid, if losses related to the second diode with an ideality factor of n2 = 2 are negligible at Voc. Therefore, this assertion needs to be proven. Hence, suns-Voc measurements have been carried out [7], from which the local ideality factor m [8] at V = Voc is calculated.
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 10
3.0
450
Suns-Voc-curve
Suns
0.1
2.0
0.01 m = n1 = 1
= I-V Voc
1E-3 I-V Vmpp pseudo Vmpp
1.5
430 j01 (fA/cm²)
local ideality factor m
2.5 slope @ 1 Sun
0.50
0.65
652 650
410
648
400 646 390 644
380
1.0
0.55 0.60 Voc (V)
654
420
370
1E-4 0.45
80 :/sq SE oxidized Linear fit Voc
440
local ideality factor m
1
Voc (mV)
118
0.70
j0e-met= 850 fA/cm²
0
642 10
5 7.5 2.5 Metallization fraction FM (%)
Figure 3: On the left the suns-Voc analysis is presented. The slope at one sun of the logarithmic plotted Suns-Voc-curve yields the local ideality factor m times VT. It shows that m = n1 = 1 is valid, hence it can be stated that in this case recombination is dominated by the first diode with a global ideality factor of n1 = 1. Now, j0e-met can be extracted as it is demonstrated on the right graph. Voc declines with increasing FM and is inversely proportional to j01.
Such an analysis is displayed in Figure 3 on the left side. At decreasing voltage, m increases so that m > n1 at the pseudo-Vmpp.This indicates an influence of the second diode n2 = 2 on the I-V-Vmpp which is even below the pseudo-Vmpp. This leads to an increasing j02 influence that primarily reduces the fill factor. However on the Voc at one sun which is the illuminated Voc (I-V Voc), m indicates no influence of the second diode, so equation (1) is applicable as demonstrated on the right hand side of Figure 3. Figure 4 shows the analysis of the local ideality factor for SP1 and SP2. Increasing cell number indicates increasing FM. All results show a weak correlation between m and FM. m remains close to unity which leads to the assumption that recombination related to the first diode is dominant. Except the “110 :/sq-plot” on the left exhibits an increased m. In this case equation (1) needs to be taken with care. 1.5 72 :/sq 78 :/sq 110 :/sq 110 :/sq SE 80 :/sq SE oxidized
1.4 1.3
Screen-printing
1.4 1.3
1.2
1.2
1.1
1.1
1.0
1.0
0.9
0.9
1
2
3
4 5 Cell number
1.5
1.5
Metallization SP1:
6
7
8
78 :/sq 110 :/sq 110 :/sq SE 80 :/sq oxidized 80 :/sq SE oxidized
1.4 local ideality factor m @ Voc
local ideality factor m @ Voc
1.5
1.3
Metallization SP2: SP seed + Ag-LIP
1.4 1.3
1.2
1.2
1.1
1.1
1.0
1.0
0.9
1
2
3
4 5 Cell number
6
7
8
0.9
Figure 4: According to Figure 3 the local ideality factor m at Voc is calculated and plotted over increasing cell number which corresponds to an increasing FM.
3.3. Extraction of j0e-met As described before, j01 is calculated and shown in Figure 5. With increasing FM, the recombination at the metal-silicon interface starts to increase. By fitting of equation (2) to the j01-FM-graphs j0e-met can be
119
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
extracted.
j01
j0b j0 en ( 1 FM ) j0 emet FM
(2)
If a selected emitter is used, equation (2) needs to be extended by taking into account the area weighted dark saturation current density j0e-n++ in the highly doped n++ region. As can be stated generally, recombination on the metal-silicon interface is suppressed efficiently by utilizing a selective emitter structure. This can be seen by comparing the “110 :/sq” with the “110 :/sq SE” plot in the SP1-graph in Figure 2, where j0e-met drops by a factor of 3 from 2711 to 873 fA/cm². Thus, the more effective shielding of minority charge carriers of a highly doped area under a metal contact is directly observed. According to the results from SP2 by comparing the “110 :/sq- with the “80 :/sqoxidized-graph”, the utilized oxidation process also decreases j0e-met. This leads to the assumption that this emitter shifts from being rather “transparent” to a more “opaque” behavior. This means that total recombination shifts toward the bulk leading to a lower recombination current density in total. Also a different impact on j0e-met between SP1 and SP2 is visible by comparing results from the same emitter. Particular for the “110 :/sq-emitter” j0e-met drops tremendously from 2711 to 917 fA/cm² just by using a different paste. 72:/sq 78 : /sq j0e-met= 2711 fA/cm² 110 : /sq 110 : /sq SE 80 : /sq SE oxidized
1200 1100 j01 (fA/cm²)
1000
j0e-met= 1482 fA/cm²
900 800
1200
j0e-met= 1054 fA/cm²
600 j0e-met= 873 fA/cm²
500
j0e-met= 676 fA/cm²
400 5
7.5
12.5 10 15 17.5 Metallization fraction FM (%)
650
700 600
SP2 650
600
1000 900
700
78:/sq 110 :/sq 80 :/sq oxidized 80 :/sq SE oxidized
1100
800
700
700
1300 SP1
j01 (fA/cm²)
1300
600 j0e-met= 917 fA/cm²
550 j0e-met= 859 fA/cm²
500
550 500
j0e-met= 664 fA/cm²
450
450
500 400 20
j0e-met= 549 fA/cm²
400 5
7.5
12.5 10 15 17.5 Metallization fraction FM (%)
400 20
Figure 5: j01 is displayed obtained according to equation (1). The error bars in y-direction give the absolute error and are obtained by middle over 3 to 6 cells. The bars in x-direction are related to fluctuations of the metallization fraction and measurement inaccuracies. Note that in respect to this analysis, the non-contacted selective emitter area was determined and subtracted from j01 area weighted by considering the recombination current density j0e-n++ displayed in Table 1.
Figure 6 shows two emitters metalized by the SP2 approach. The experimental value (exp.) accords to the measured I-V Voc. The simulated curve is obtained by taking into account the corresponding dark saturation current densities determined from symmetrical samples shown in Table 1, and the experimental obtained value of j0e-met from Figure 5. By using (2) and subsequently (1) yields the Voc of the whole device. The extraction of the dark saturation current density of the rear and the bulk j0b is published elsewhere [9] and is determined to be j0b = 240 fA/cm². Obviously, measured and simulated open circuit voltage fit very nicely.
120
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 655
655
Voc (mV)
SP2 650
80 :/sq SE oxidized exp. 80 :/sq SE oxidized sim. 650 78 :/sq exp. 78 :/sq sim.
645
645
640
640
635
635
630
j0b= 240 fA/cm²
0
4 8 12 16 20 Front side metallization fraction FM (%)
24
630
Figure 6: Shows plots of the Voc for the SP2 approach. Measured Voc and the experimentally (exp.) obtained dark saturation current densities, which give the simulated (sim.) open-circuit-voltage, agree within a minimal deviation.
4. Conclusions The goal of this work was the determination of the dark saturation current density j0e-met at the metalemitter interface. Therefore, 2x2 cm² sized solar cells with varying front metallization fractions and emitter profiles have been processed. From symmetrical samples the dark saturation current density j0e-n+ of the emitters incorporated within this work has been extracted reaching a best value of j0e-n+ = 160 fA/cm². By simply fitting the j01-FM-curve from the corresponding open-circuit-voltage Voc (FM) yields j0e-met. However, this can only be done if recombination at Voc features a diode character that is close to unity. Therefore, the local ideality factor m from the suns-Voc-curve has been determined indicating the required value close to one. Three main effects are observed. First, different screen-printing pastes on the same emitter configuration showed different influence on j0e-met. Second, the short drive-in step of the oxidation that was used reduces “transparency” of the emitter to a more “opaque” behaviour, hence reduces j0e-met. Also the field-effect passivation of the highly n-doped selective emitter decreases j0e-met effectively. By combination the field effect passivation and the short oxidation process a very low value of j0e-met = 549 fA/cm² is reported which is very close to the passivated case of j0e-n++ = 397 fA/cm². By taking into account the experimentally obtained dark saturation current densities and the corresponding j0e-met-value, the measured open-circuit-voltage is reproduced within a minimal deviation.
Acknowledgements The author thanks S. Mack, A. Wolf, M. Retzlaff, U. Jäger, D. Stüwe, the whole PV-TEC team and Prof. H. Clement from the University of Tübingen as well as Prof. G. Willeke for fruitful discussions.
References [1] A. Cuevas et al., “Surface recombination velocity of highly doped n-type silicon”, Journal of Applied Physics, 1996;80(6):3370-5 [2] R.R. King et al., “Studies of diffued phosphorous emitters”, IEEE Transaction on Electron Devices, 1990;37(2):365-71 [3] U. Jäger et al., “Selective emitter by laser dopung from phosphosikicate glass”, 2009, Proceedings of the 24th PVSEC, Germany
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 [4] J. Bartsch et al., “Electrochemical methods to analyse the light-induced plating process”, Journal of Applied Electrochemistry, 2010;40(4):757-65. [5] C. Reichel et al., “Comparison of emitter saturation current densities determined by quasi-steady-state photoconductande measurement of effective carrier lifetime at high and low injections”, 2008, Proceedings of the 23th PVSEC, Spain. [6] S. Mack. et al., “Simultaneous front emitter and rear surface passivation by thermal oxidation – an industrially feasible approach to 19% efficient PERC device”, 2010, Proceedings of the 25th EU PVSEC, Spain [7] R.A. Sinton and A. Cuevas, “A quasi-steady-state open-circuit voltage methode for solar cell characterization”, 2000, Proceedings of the 16th PVSEC, UK. [8] K.R. McIntosh et al., “Lumps, humps and bumps: Three detrimental effects in the current-voltage curve of silicon solar cells”, 2001, Dissertation, University of New South Wales, Australia. [9] T. Fellmeth et al., “20.1% efficient silicon solar cell with aluminum back surface field”,2011, submitted to IEEE Electron Device Letters
121
Energy Procedia 8 (2011) 115–121
SiliconPV: 17-20 April 2011, Freiburg, Germany
Recombination at Metal-Emitter Interfaces of Front Contact Technologies for Highly Efficient Silicon Solar Cells T. Fellmeth*, A. Born, A. Kimmerle, F. Clement, D. Biro, R. Preu Fraunhofer Institute for Solar Energy Systems ISE, Heidenhofstr. 2, 79110 Freiburg, Germany
Abstract We present an experimental approach to extract the dark saturation current density j0e-met at the emitter-metal interface of the front contact. For this purpose, 2x2 cm² sized silicon solar cells have been realized featuring different metallization fractions FM. By simply applying the one-diode-model, the dark current density j01 is determined from the open circuit voltage Voc. From the slope of the j01 over FM plot, j0e-met is extracted. However, this is only valid if the dominant recombination mechanism at Voc features a diode character that is close to unity. Hence, the local ideality factor m is determined from the suns-Voc-curve indicating the required value close to one. Three main effects are observed. First, the metallization methods which are compared show different influences on j0e-met on the same emitter configuration. Second, an emitter drive-in due to an additional short thermal oxidation lowers j0e-met. Also, the field-effect passivation of the highly n-doped selective emitter decreases j0e-met effectively. By combining the field effect passivation with a short drive-in step the very low value of j0e-met = 549 fA/cm² is reported.
© responsibility of of SiliconPV 2011. © 2011 2010 Published Publishedby byElsevier ElsevierLtd. Ltd.Selection Selectionand/or and/orpeer-review peer-reviewunder under responsibility SiliconPV. Keywords: silicon solar cell; emitter; metallization
1. Introduction Properties of the front metallization of solar cells are not only determined by conductivity, contact resistance and light shading. The metal-silicon interface of a phosphorus doped emitter provides a highly recombination active surface for minority charge carriers [1]. Hence, the impact of this loss mechanism needs to be taken into account by evaluation of front metallization methods. Unfortunately, emitter-metal recombination can not be measured directly like conductivity or contact resistance. Furthermore, the
* Corresponding author, Tel.: +49(0) 7 61/ 45 88-5652 E-mail: [email protected]
1876–6102 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of SiliconPV 2011. doi:10.1016/j.egypro.2011.06.111
116
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
degree of surface recombination always depends on the emitter profile itself [2]. The terms “opaque” and “transparent” emitter have been introduced [2] to indicate where minority charge carrier recombination in the emitter is dominant. In an “opaque” n-type emitter, holes recombine mostly in the volume however, in “transparent” emitters holes likely reach the surface and recombine there. The latter emitters are very sensitive to surface recombination properties, hence to the front metallization. In order to quantify the recombination at metal-emitter interfaces, the emitter dark saturation current density j0e-met is introduced analogously to the emitter dark saturation current density j0en+ of the passivated emitter area. Both share the same diode characteristic of n = 1. Hence, emitter recombination can be described by just adding them, weighted by area. In order to reduce recombination at the metal-emitter interface, there are three approaches in principle. The first is to simply reduce metallization fraction FM, secondly, by utilizing a metallization method that induces a low j0e-met and thirdly by utilizing an opaque emitter under the metal contact area. The latter can be done by using the selective emitter approach. The highly n-doped area under the front contact shields the minority charge carriers from reaching the surface more effectively mainly due to a higher field-effect passivation [3]. 2. Experimental approach The goal of this experiment is the extraction of the emitter dark saturation current density j0e-met at the metal contact area. Therefore, 2x2 cm² sized Si-FZ, p-type 1 :cm solar cells have been processed featuring different emitter profiles and front metallization. The screen-printing metallization is presented featuring two different pastes: x SP1: screen-printing x SP2: screen-printed seed layer plus silver light induced plating (Ag-LIP) [4] In Figure 1, a scan of a wafer is presented where eight cells are located. The front metallization fraction has been varied, resulting in varying finger pitch from 2 to 0.4 mm.
Figure 1: On the left a scan of a wafer (12.5x12.5cm²) is displayed showing eight solar cells exhibiting different finger pitches starting from 2 mm equating to C1 to 0.4 mm, that equating to C8. On the right side two cross-sections of a device featuring the SP2metallzation is displayed exhibiting in this case a selective (n++-doped) and oxidized (SiO2) emitter. Also the different emitter regions are highlighted that exhibit different dark saturation current densities.
117
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
3. Results 3.1. j0e and SIMS analysis Different emitters have also been incorporated into the experiment which is presented in Table 1. j0e-n+ is extracted in low level injection from symmetrical emitter samples using the QSSPC technique [5]. By applying the one-diode-model, the Voc limit Voc-lim of the emitters is calculated assuming a short circuit current density jsc of 35 mA/cm². For passivation of the emitter, PECVD silicon nitride (SiNx) is used. Additionally, there are samples where 10 nm of SiO2 is grown during a dry thermal oxidation at 840 °C before the deposition of the SiNx-layer is done. This process is applied on the emitter with a sheet resistance of 110 :/sq, leading to a reduction of the sheet resistance to 80 :/sq and a reduction of j0en+ from 264 to 160 fA/cm² [6]. In Figure 2 SIMS profiles of the emitters introduced in Table 1 are displayed. The phosphorus silicate glass (PSG) laser doped emitter exhibits a low sheet resistance due to the high degree of phosphorus within the first 150 nm. Also, the effect of the thermal oxidation is observable at the profile of the emitter with a sheet resistance of 80 :/sq. Compared to the 110 :/sq emitter the front phosphorus level is reduced and the tail of the profile is extended.
Rsh
j0e-n+ (fA/cm²)
Voc-lim (mV)
Passivation
(ȍ/sq) 72
347
651
SiNx
78
299
655
SiNx
110
264
658
SiNx
110 ¼ 31*
j0e-n++ = 575
637
SiNx
80
160
671
SiO2 + SiNx
110 ¼ -*
j0e-n++ = 397
645
SiO2 + SiNx
*n++-Emitter due to PSG laser doping.
1E22
1E22 PSG laser doped 31 :/sq 80 :/sq oxidized 78 :/sq 72 :/sq 110 :/sq
1E21 -3
Phosphor atoms (cm )
Table 1. Sheet resistances and emitter dark saturation current densities of the symmetrical samples. A photo generated current density of 35 mA/cm² and a ni in the base of 9.14*109 cm-3 are assumed. The evaluation is carried out at an injection level of 'n = 1015 cm-3.
1E20
1E21 1E20
1E19
1E19
1E18
1E18
1E17
1E17
1E16
1E16
1E15 0.0
0.1
0.2
0.3 0.4 depth (μm)
0.5
1E15 0.6
Figure 2: SIMS profiles of the emitters listed in Table 1. The “PSG laser doped” profile is generated by laser induced diffusion from the 110 :/sq emitter with PSG-layer acting as a phosphorus source. The PSG laser doped profile after thermal oxidation is not displayed.
3.2. Determination of the local ideality factor According to the one-diode-model, where VT denotes the thermal voltage at 25 °C and jph the photo generated current density that is set equal to jsc, Voc is directly correlated with j01.
Voc
§ j · VT Ln ¨¨ ph ¸¸ © j01 ¹
(1)
This relation is only valid, if losses related to the second diode with an ideality factor of n2 = 2 are negligible at Voc. Therefore, this assertion needs to be proven. Hence, suns-Voc measurements have been carried out [7], from which the local ideality factor m [8] at V = Voc is calculated.
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 10
3.0
450
Suns-Voc-curve
Suns
0.1
2.0
0.01 m = n1 = 1
= I-V Voc
1E-3 I-V Vmpp pseudo Vmpp
1.5
430 j01 (fA/cm²)
local ideality factor m
2.5 slope @ 1 Sun
0.50
0.65
652 650
410
648
400 646 390 644
380
1.0
0.55 0.60 Voc (V)
654
420
370
1E-4 0.45
80 :/sq SE oxidized Linear fit Voc
440
local ideality factor m
1
Voc (mV)
118
0.70
j0e-met= 850 fA/cm²
0
642 10
5 7.5 2.5 Metallization fraction FM (%)
Figure 3: On the left the suns-Voc analysis is presented. The slope at one sun of the logarithmic plotted Suns-Voc-curve yields the local ideality factor m times VT. It shows that m = n1 = 1 is valid, hence it can be stated that in this case recombination is dominated by the first diode with a global ideality factor of n1 = 1. Now, j0e-met can be extracted as it is demonstrated on the right graph. Voc declines with increasing FM and is inversely proportional to j01.
Such an analysis is displayed in Figure 3 on the left side. At decreasing voltage, m increases so that m > n1 at the pseudo-Vmpp.This indicates an influence of the second diode n2 = 2 on the I-V-Vmpp which is even below the pseudo-Vmpp. This leads to an increasing j02 influence that primarily reduces the fill factor. However on the Voc at one sun which is the illuminated Voc (I-V Voc), m indicates no influence of the second diode, so equation (1) is applicable as demonstrated on the right hand side of Figure 3. Figure 4 shows the analysis of the local ideality factor for SP1 and SP2. Increasing cell number indicates increasing FM. All results show a weak correlation between m and FM. m remains close to unity which leads to the assumption that recombination related to the first diode is dominant. Except the “110 :/sq-plot” on the left exhibits an increased m. In this case equation (1) needs to be taken with care. 1.5 72 :/sq 78 :/sq 110 :/sq 110 :/sq SE 80 :/sq SE oxidized
1.4 1.3
Screen-printing
1.4 1.3
1.2
1.2
1.1
1.1
1.0
1.0
0.9
0.9
1
2
3
4 5 Cell number
1.5
1.5
Metallization SP1:
6
7
8
78 :/sq 110 :/sq 110 :/sq SE 80 :/sq oxidized 80 :/sq SE oxidized
1.4 local ideality factor m @ Voc
local ideality factor m @ Voc
1.5
1.3
Metallization SP2: SP seed + Ag-LIP
1.4 1.3
1.2
1.2
1.1
1.1
1.0
1.0
0.9
1
2
3
4 5 Cell number
6
7
8
0.9
Figure 4: According to Figure 3 the local ideality factor m at Voc is calculated and plotted over increasing cell number which corresponds to an increasing FM.
3.3. Extraction of j0e-met As described before, j01 is calculated and shown in Figure 5. With increasing FM, the recombination at the metal-silicon interface starts to increase. By fitting of equation (2) to the j01-FM-graphs j0e-met can be
119
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121
extracted.
j01
j0b j0 en ( 1 FM ) j0 emet FM
(2)
If a selected emitter is used, equation (2) needs to be extended by taking into account the area weighted dark saturation current density j0e-n++ in the highly doped n++ region. As can be stated generally, recombination on the metal-silicon interface is suppressed efficiently by utilizing a selective emitter structure. This can be seen by comparing the “110 :/sq” with the “110 :/sq SE” plot in the SP1-graph in Figure 2, where j0e-met drops by a factor of 3 from 2711 to 873 fA/cm². Thus, the more effective shielding of minority charge carriers of a highly doped area under a metal contact is directly observed. According to the results from SP2 by comparing the “110 :/sq- with the “80 :/sqoxidized-graph”, the utilized oxidation process also decreases j0e-met. This leads to the assumption that this emitter shifts from being rather “transparent” to a more “opaque” behavior. This means that total recombination shifts toward the bulk leading to a lower recombination current density in total. Also a different impact on j0e-met between SP1 and SP2 is visible by comparing results from the same emitter. Particular for the “110 :/sq-emitter” j0e-met drops tremendously from 2711 to 917 fA/cm² just by using a different paste. 72:/sq 78 : /sq j0e-met= 2711 fA/cm² 110 : /sq 110 : /sq SE 80 : /sq SE oxidized
1200 1100 j01 (fA/cm²)
1000
j0e-met= 1482 fA/cm²
900 800
1200
j0e-met= 1054 fA/cm²
600 j0e-met= 873 fA/cm²
500
j0e-met= 676 fA/cm²
400 5
7.5
12.5 10 15 17.5 Metallization fraction FM (%)
650
700 600
SP2 650
600
1000 900
700
78:/sq 110 :/sq 80 :/sq oxidized 80 :/sq SE oxidized
1100
800
700
700
1300 SP1
j01 (fA/cm²)
1300
600 j0e-met= 917 fA/cm²
550 j0e-met= 859 fA/cm²
500
550 500
j0e-met= 664 fA/cm²
450
450
500 400 20
j0e-met= 549 fA/cm²
400 5
7.5
12.5 10 15 17.5 Metallization fraction FM (%)
400 20
Figure 5: j01 is displayed obtained according to equation (1). The error bars in y-direction give the absolute error and are obtained by middle over 3 to 6 cells. The bars in x-direction are related to fluctuations of the metallization fraction and measurement inaccuracies. Note that in respect to this analysis, the non-contacted selective emitter area was determined and subtracted from j01 area weighted by considering the recombination current density j0e-n++ displayed in Table 1.
Figure 6 shows two emitters metalized by the SP2 approach. The experimental value (exp.) accords to the measured I-V Voc. The simulated curve is obtained by taking into account the corresponding dark saturation current densities determined from symmetrical samples shown in Table 1, and the experimental obtained value of j0e-met from Figure 5. By using (2) and subsequently (1) yields the Voc of the whole device. The extraction of the dark saturation current density of the rear and the bulk j0b is published elsewhere [9] and is determined to be j0b = 240 fA/cm². Obviously, measured and simulated open circuit voltage fit very nicely.
120
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 655
655
Voc (mV)
SP2 650
80 :/sq SE oxidized exp. 80 :/sq SE oxidized sim. 650 78 :/sq exp. 78 :/sq sim.
645
645
640
640
635
635
630
j0b= 240 fA/cm²
0
4 8 12 16 20 Front side metallization fraction FM (%)
24
630
Figure 6: Shows plots of the Voc for the SP2 approach. Measured Voc and the experimentally (exp.) obtained dark saturation current densities, which give the simulated (sim.) open-circuit-voltage, agree within a minimal deviation.
4. Conclusions The goal of this work was the determination of the dark saturation current density j0e-met at the metalemitter interface. Therefore, 2x2 cm² sized solar cells with varying front metallization fractions and emitter profiles have been processed. From symmetrical samples the dark saturation current density j0e-n+ of the emitters incorporated within this work has been extracted reaching a best value of j0e-n+ = 160 fA/cm². By simply fitting the j01-FM-curve from the corresponding open-circuit-voltage Voc (FM) yields j0e-met. However, this can only be done if recombination at Voc features a diode character that is close to unity. Therefore, the local ideality factor m from the suns-Voc-curve has been determined indicating the required value close to one. Three main effects are observed. First, different screen-printing pastes on the same emitter configuration showed different influence on j0e-met. Second, the short drive-in step of the oxidation that was used reduces “transparency” of the emitter to a more “opaque” behaviour, hence reduces j0e-met. Also the field-effect passivation of the highly n-doped selective emitter decreases j0e-met effectively. By combination the field effect passivation and the short oxidation process a very low value of j0e-met = 549 fA/cm² is reported which is very close to the passivated case of j0e-n++ = 397 fA/cm². By taking into account the experimentally obtained dark saturation current densities and the corresponding j0e-met-value, the measured open-circuit-voltage is reproduced within a minimal deviation.
Acknowledgements The author thanks S. Mack, A. Wolf, M. Retzlaff, U. Jäger, D. Stüwe, the whole PV-TEC team and Prof. H. Clement from the University of Tübingen as well as Prof. G. Willeke for fruitful discussions.
References [1] A. Cuevas et al., “Surface recombination velocity of highly doped n-type silicon”, Journal of Applied Physics, 1996;80(6):3370-5 [2] R.R. King et al., “Studies of diffued phosphorous emitters”, IEEE Transaction on Electron Devices, 1990;37(2):365-71 [3] U. Jäger et al., “Selective emitter by laser dopung from phosphosikicate glass”, 2009, Proceedings of the 24th PVSEC, Germany
T. Fellmeth et al. / Energy Procedia 8 (2011) 115–121 [4] J. Bartsch et al., “Electrochemical methods to analyse the light-induced plating process”, Journal of Applied Electrochemistry, 2010;40(4):757-65. [5] C. Reichel et al., “Comparison of emitter saturation current densities determined by quasi-steady-state photoconductande measurement of effective carrier lifetime at high and low injections”, 2008, Proceedings of the 23th PVSEC, Spain. [6] S. Mack. et al., “Simultaneous front emitter and rear surface passivation by thermal oxidation – an industrially feasible approach to 19% efficient PERC device”, 2010, Proceedings of the 25th EU PVSEC, Spain [7] R.A. Sinton and A. Cuevas, “A quasi-steady-state open-circuit voltage methode for solar cell characterization”, 2000, Proceedings of the 16th PVSEC, UK. [8] K.R. McIntosh et al., “Lumps, humps and bumps: Three detrimental effects in the current-voltage curve of silicon solar cells”, 2001, Dissertation, University of New South Wales, Australia. [9] T. Fellmeth et al., “20.1% efficient silicon solar cell with aluminum back surface field”,2011, submitted to IEEE Electron Device Letters
121