Ribbon Si solar cells with efficiencies over 18% by hydrogenation of defects

ARTICLE IN PRESS

Solar Energy Materials & Solar Cells 90 (2006) 1227–1240 www.elsevier.com/locate/solmat

Ribbon Si solar cells with efficiencies over 18% by hydrogenation of defects Dong Seop Kima, Vijay Yelundura,, Kenta Nakayashikia, Brian Rounsavillea, Vichai Meemongkolkiata, Andrew M. Gaborb, Ajeet Rohatgia a

University Center of Excellence for Photovoltaics Research and Education, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA b Evergreen Solar Inc., 259 Cedar Hill Street, Marlboro, MA 01752, USA Received 10 May 2005; accepted 20 July 2005 Available online 9 March 2006

Abstract We have fabricated 4 cm2 solar cells on String Ribbon Si wafers and edge-defined film-fed grown (EFG) Si wafers with using a combination of laboratory and industrial processes. The highest efficiency on String Ribbon Si wafer is 17.8% with an open circuit voltage (Voc) of 620 mV, a short circuit current density (Jsc) of 36.8 mA/cm2 and a fill factor (FF) of 0.78. The maximum efficiency on EFG Si is 18.2% with a Voc of 620 mV, a Jsc of 37.5 mA/cm2 and a FF of 0.78. These are the most efficient ribbon Si devices made to date, demonstrating the high quality of the processed Si ribbon and its potential for industrial cells. Co-firing of SiNx and Al by rapid thermal processing was used to boost the minority carrier lifetime of bulk Si from 3–5 ms to 70–100 ms. Photolithography-defined front contacts were used to achieve low shading losses and low contact resistance with a good blue response. The effects of firing temperature and time were studied to understand the trade-off between hydrogen retention and Al-doped back surface field (Al-BSF) formation. Excellent bulk defect hydrogenation and high-quality thick Al-BSF formation was achieved in a very short time (
1 s) at firing temperatures of

Corresponding author. Tel.: +1 404 894 9886; fax: +1 404 894 4832.

E-mail address: [email protected] (V. Yelundur). 0927-0248/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2005.07.008

ARTICLE IN PRESS 1228

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

740–750 1C. It was found that the bulk lifetime decreases at annealing temperatures above 750 1C or annealing time above 1 s due to dissociation of hydrogenated defects. r 2005 Elsevier B.V. All rights reserved. Keywords: Ribbon Si; String Ribbon; EFG; Hydrogenation; Silicon nitride; Solar cell

1. Introduction String Ribbon Si wafers are grown by high-temperature strings drawn through the crucible of molten Si that is currently in multi-megawatt production at Evergreen Solar. Edge-defined film-fed grown (EFG) Si wafers are pulled directly through a graphite die via capillary action. These two vertical sheet growth techniques produce low-cost Si wafers due to the high utilization of the Si feedstock and the absence of ingot sawing and wafer etching. The high quality of the processed String Ribbon and EFG wafers has been previously demonstrated [1,2] through the measurement of high minority carrier lifetimes following cell processing. Recently record highefficiency String Ribbon cell (17.7%) has been fabricated using thermal oxidation for front surface passivation, 60 min microwave-induced remote hydrogen plasma passivation, ZnS/MgF2 double-layer antireflection (AR) coating and aluminum gettering for 30 min [3]. Recent research on processing String Ribbon cells has focused on industrial-type processing using screen-printing for metallization and the relatively deep junctions necessary for firing the screen-printable inks and plasma-enhanced chemical vapor deposition (PECVD) SiNx for defect passivation. Recent cells made with screenprinting are now approaching the 16% level [4]. Therefore, we revisited laboratory cell production to demonstrate the even higher potential of this material by using industrial-type SiNx film for AR coating and bulk and surface passivation as well as AR coating. One of the laboratory processing schemes we have implemented is rapid thermal processing (RTP) for simultaneous Al-doped back surface field (Al-BSF) formation and passivation of bulk defects through hydrogenation from a SiNx layer. RTP is often used for Si solar cell fabrication due to several advantages over conventional tube furnace processing, including short cell processing time, low thermal budget, high heating and cooling rates and accurate temperature control [5–7]. In this paper we pushed the RTP process toward shorter dwell times and lower temperatures than in the past to enhance the cell performance and study the effects on the resulting AlBSF quality and defect passivation. Appropriate firing or co-firing of SiNx and Al has been shown to be an important step for significant enhancement in bulk lifetime [8]. In this paper the process sequence has been tuned to take advantage of this effect. More specifically, the co-firing temperature and time are optimized to achieve maximum hydrogenation of defects without sacrificing Al-BSF quality. The approach is directly applicable to the industrial-type screen-printing of solar cells if the front metal contact is replaced by screen-printed contact.

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1229

2. Experimental P-type String Ribbon and EFG Si wafers with an average thickness of 300 mm and a resistivity of 3 O-cm were pulled at the Evergreen Solar and RWE Schott Solar, respectively. P-type, 300 mm thick, 2 O-cm float zone (FZ) Si wafers were also used in this study as a reference. The String Ribbon Si wafers were cut to an optimum size for tube diffusion and cleaned/etched in cleaning solutions of 2:1:1 H2O:H2O2:H2SO4, 15:5:2 HNO3: CH3COOH:HF, 2:1:1 H2O:H2O2:HCl. The phosphorus diffusion was performed at 870 1C for 32 min using a liquid POCl3 source in a tube furnace to obtain an 85 O/&n+-emitter. The SiNx films were deposited in a commercial lowfrequency PECVD reactor operating at a frequency of 50 kHz and a temperature of 430 1C on the phosphorus-diffused emitters. Al paste (Ferro FX 53-038) was screenprinted on the back surface of the wafers. The SiNx on the front and the Al film on the rear were fired simultaneously in an RTP system to enhance hydrogen passivation. The ramp-up and cooling rates were set to 450 1C/s for all the firing processes of String Ribbon cells to achieve a uniform Al-BSF layer. The firing temperature was varied in the range of 700–800 1C for 1 s and 60 s firing times in order to study the effects of firing temperature and time on the cell performance. The range of firing temperature used in this study was determined from optimization of screen-printed ribbon solar cells [9]. The processing temperature was measured by a thermocouple in physical contact with the front side of wafer. The front metal grid was defined by a photolithography process followed by removal of the SiNx film in the grid region by etching in HF. Front contacts were formed by evaporating 60 nm of Ti, 40 nm of Pd and 60 nm of Ag followed by a lift-off procedure. Additional Ag was plated to increase grid thickness and reduce series resistance. Nine 4-cm2 cells on each wafer were fabricated and isolated using a dicing saw and then annealed in forming gas at 400 1C for 30 min. The emitter saturation current density (Joe) was measured by photoconductance decay method [10]. The optical properties of the SiNx were characterized using a spectroscopic ellipsometer to determine the optimal design of a double layer AR coating. In order to minimize reflectance, the SiNx thickness was adjusted to 67.8 nm by etching the film in HF. Magnesium fluoride (99.5 nm) was deposited on the SiNx by vacuum evaporation to form a double AR layer. Long wavelength internal quantum efficiency (IQE) measurements were performed to characterize the quality of the Al-BSF in finished solar cells. The thickness of the screen-printed Al was measured to be 23–28 mm by profilometry (Alpha-Step 200) after drying the screen-printed Al. The effective thickness of the printed Al was calculated by measuring the screen-printed Al weight after removal of organic constituents and solvents from the paste by annealing screen-printed Al paste at
500 1C. The thickness of the Al-BSF was measured by cross-sectional scanning electron microscopy (SEM) after delineating the Al-doped p+-region in 1:3:6 HF:HNO3:CH3COOH for 10 s. In order to study the quality of Al-BSF and its impact on the cell performance, photolithography cells were also fabricated on FZ wafers with a high-quality rapid thermal oxide on the emitter capped with ZnS/MgF2 double layers. Light beam-induced current (LBIC) measurements were carried out on best ribbon solar cells to map electrically active

ARTICLE IN PRESS 1230

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

grain boundaries and other defects. Model calculations were performed to support the experimental results and improve the understanding of these high-efficiency ribbon Si cells.

3. Results and discussion Fig. 1 shows the progress in the efficiency of three different types of ribbon Si solar cells with photolithography contacts to date, including the 17.8% efficient String Ribbon and 18.2% EFG cells fabricated in this study. Data for Dendritic Web and EFG is limited. The record efficiency String Ribbon and EFG cells made in this study were tested and verified by National Renewable Energy Laboratory (NREL). Fig. 2 shows the efficiencies of String Ribbon cells achieved in this study as a function of firing temperature and time. Fig. 2 shows that 60 s firing produced a maximum efficiency of 17.8% at 740 1C with rapid decrease in efficiency above 740 1C. Fig. 2 also reveals that cell efficiency for a much shorter firing time of 1 s was less sensitive to the firing temperatures, resulting in an equivalent maximum efficiency of 17.8% at 760 1C. It is noteworthy that a very short 1 s firing at 760 1C gives much higher efficiency than 60 s firing at 760 1C. The cell parameters are summarized in Table 1 including the best EFG cell with efficiency of 18.2% achieved by 1 s firing at 750 1C. As shown in Table 1, firing 1 s gives Voc over 620 mV for all the firing temperatures above 740 1C due to high bulk lifetime whereas firing 60 s reduces significantly Voc and Jsc above 740 1C due to reduction of lifetime. The highest efficiency achieved in this study is largely attributed to the improvement in bulk lifetime without the loss of Al-BSF quality at the optimum

19

String Ribbon 18.2

Dendritic Web EFG

18

This study

17.8

Cell Efficiency (%)

17.7 17

17.3

17.2 16.9

16.2

16

16.7

15. 2 15

15.0

15.4

14 .7

14.5

14.6 14 14.1 13 12.7 12 1980

1985

1990

1995

2000

2005

Year Fig. 1. Progress in the efficiency of laboratory scale ribbon solar cells with photolithography contacts.

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1231

22 1min firing 20

1sec firing

Efficiency (%)

17.8

17.8

18

17.6

17.3 16.5

16.2

15.5

16

14

12

10

700

740

760

800

Firing Temperature (°C)

Fig. 2. Maximum String Ribbon solar cell efficiency as a function of firing temperature and time for AlBSF/SiNx co-firing. Table 1 Light I–V data for PL cells (masked to 3.8 cm2) measured by the National Renewable Energy Laboratory Time (s)

Temp (1C)

Eff. (%)

Voc (mV)

Jsc (mA/cm2)

FF

Substrate

60 60 60 60 1 1 1 1

700 740 760 800 740 760 800 750

16.2 17.8 16.5 15.5 17.3 17.8 17.6 18.2

604 622 615 602 620 620 623 620

33.8 36.4 34.0 33.1 35.2 36.8 35.6 37.5

0.80 0.78 0.79 0.78 0.79 0.78 0.79 0.78

String String String String String String String EFG

Ribbon Ribbon Ribbon Ribbon Ribbon Ribbon Ribbon

firing temperature. Fig. 3 shows that the average bulk lifetime in String Ribbon increased from 5 to 80 ms with only 1 s firing while the lifetime of EFG Si increased from 4 to 95 ms. For 1 s firing, the bulk lifetime was maintained over 50 ms even after 800 1C co-firing of String Ribbon Si. In contrast, the bulk lifetime of String Ribbon and EFG Si dropped rapidly at the temperatures above 740 1C for 60 s firing. This suggests that hydrogenation of bulk defects takes place in a very short time and the final lifetime is determined by the release of hydrogen from the defects if hydrogen flux from the SiNx into the Si is not maintained during firing process. IQE measurements showed a significant enhancement in the long wavelength response, suggesting that in 1 s atomic hydrogen is able to passivate defects present as far as even near the rear surface. Van Weiringen and Warmoltz (VWW) measured and reported the hydrogen diffusivity according to the following equation in the

ARTICLE IN PRESS 1232

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240 100

73.7

Lifetime (µs)

80

58.5

60 38.1

32.7

40 12.6 20 4.5 0

(a)

As-grown P-diffusion 740C/1sec 740C/60sec 800C/1sec 800C/60sec Process Scheme

120 95

Lifetime (µs)

100

80 61 60

50

40

33

20 3 0

(b)

4

as-grown P-diffusion 750C/1sec 750C/10sec 750C/30sec 750C/60sec Process Scheme

Fig. 3. Effect of various process schemes on the average lifetime of (a) String Ribbon and (b) EFG Si.

temperature range of 1090–1200 1C [11]:
 DH ¼ 9:4  103 exp 0:48 eV=kT .

(1) 5

2

Extrapolation of Eq. (1) yields a diffusivity of 3.846  10 cm /s at 740 1C. In ptype Si, most of the atomic hydrogen diffuses by rapid interstitial motion at temperatures over 500 1C without any retardation by either acceptor trapping or molecule formation [12]. Hydrogen diffusion during the ramp-up and cooling can affect overall diffusion in case of very short annealing. By assuming diffusivity is a function of time, the effect can be treated by changing variables as in the following equations: qC q2 C ¼ DðTðtÞÞ 2 , qt qx

(2)

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

Z

1233

t

DðTðt0 ÞÞ dt0 ,

R¼

(3)

0

qC q2 C ¼ , qR qx2

(4)

where T(t) is the time-dependent temperature during ramp-up and cooling, D(T(t)) temperature-dependent diffusivity, C the concentration of hydrogen, and R is the integrated [13].Eq. (4) can be solved to calculate hydrogen peffffiffiffiffiffiffiffiffiffiffiffiffiffi  Dt product or (Dt) profile Cðx; tÞ ¼ C o erfcðx=2 ðDtÞeff by assuming constant hydrogen concentration at front surface (Co) during the annealing and infinite substrate thickness. Since the firing cycle is extremely rapid (1 s hold time), the surface concentration of hydrogen can be assumed as constant during the heat treatment. When heating and cooling rates are 100 1C/s and the annealing time is 1 s, the value of (Dt)eff is calculated to be 3.702 times higher than (Dt) which is obtained neglecting diffusion during the heating and cooling steps. Fig. 4 shows that hydrogen diffusion is significantly affected by ramp-up and cooling steps in case of short firing cycle and the C/Co value at the rear surface (x ¼ 300 mm) could reach 7.5  102 using VWW0 diffusivity [Eq. (1)] for 1 s firing at peak temperature and heating and cooling rate of 100 1C/s. Secondary ion mass spectrometry data in the literature [14,15] indicates that Co values after the SiNx deposition can be as high as 10186.0  1020 cm3, which suggests that the hydrogen concentration at the rear surface could reach about

Fractional hydrogen concentration (C/Co)

1.2

1.0

annealing only (1 sec) annealing (1 sec ) + ramp-up and down(100 C/sec)

0.8

0.6

114 um

59 um

0.4

C/Co=7.55x10-2

0.2

C/Co=6.26x10-4 0.0

0

100

200 300 Distance from the front surface ( µm )

400

500

Fig. 4. Calculated fractional hydrogen concentration after 1 s firing at 740 1C as a function of distance from the front surface. Solid circles represent hydrogen profiles calculated including the effects of ramp-up and ramp-down (100 1C/s).

ARTICLE IN PRESS 1234

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1018 cm3. There are other factors which can enhance the diffusion of hydrogen into Si. For example, Sopori et al. [16] predicted that hydrogen can diffuse through the entire wafer in 10 s after annealing of SiNx coated wafers at 800 1C in an RTP chamber due to dynamic distributions of hydrogen stored in process-induced mobile traps (PIT). The diffusivity of traps changes by an order of magnitude depending on Si wafers. Dube and Hanoka [17] showed that hydrogen diffusion can also be enhanced by the presence of defects, such as dislocations and grain boundaries. In order to study the effects of firing temperature and time on the quality of the Al-BSF, photolithography cells were fabricated on 2 O-cm FZ wafers with a rapid thermal oxide for emitter passivation. For all the firing conditions, the measured Joe was 2.0  1013 A/cm2 without the metal contacts to the emitter. This translates into an open-circuit voltage (Voc) [ ¼ kT/q  ln (Jsc/Jo)] of 670 mV for a short-circuit current density ðJ sc Þ ¼ 35 mA=cm2 . Since the measured Voc’s were only in the range of 630–640 mV for the FZ cells (Fig. 5) with high-quality photolithography contacts and very high bulk lifetimes, the observed dependence of the open circuit voltage on the firing temperature and time in Fig. 5 is primarily governed by the Al-BSF quality. Note that the Voc was appreciably low for the 700 1C firing but the improvement in the Voc saturated at firing temperatures above 740 1C. The IQE response of the FZ cells in the range of 750–1000 nm (Fig. 6) also indicates that the Al-BSF quality is not appreciably affected by the firing at temperatures over 740 1C for both 1 and 60 s firing times. The Al-BSF quality is determined by the depth of the p+ region, the doping concentration of the BSF layer, and the uniformity of the p–p+ junction. The junction depth and doping concentration can be calculated from Al–Si phase diagram using the following equation [18]: W BSF ¼

  tAl  rAl F ðTÞ F ðT 0 Þ  , 1  F ðTÞ 1  F ðT 0 Þ rSi

(5)

680 670

60 sec firing 1 secfiring

Voc ( mV )

660 650 640

639 638

637 638

740

760

641 638

629 630 620 610 600

700

800

Firing Temperature (°C ) Fig. 5. Average Voc as a function of firing temperature and time. Photolithography contact cells were fabricated on 2 O-cm FZ with rapid thermal oxide for emitter passivation and ZnS/MgF2 for antireflection coating.

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1235

100

95

IQE ( %)

90 740C, 1sec 800C, 1sec 85

700C, 60sec 740C, 60sec 800C, 60sec

80

75 700

750

800

850 Wavelength (nm)

900

950

1000

Fig. 6. IQE response of 2 O-cm FZ cells for different SiNx/Al firing processes.

where tAl, rAl and rSi represent the as-deposited Al thickness, and the densities of Al and Si, respectively, F(T) is the Si atomic weight percentage of the molten phase at the peak alloying temperature and F(To) is the Si atomic weight percentage at the eutectic temperature (
12%). The doping concentration of Al in the BSF is determined by the solid solubility of Al at each temperature as the BSF layer grows from the molten phase during the cooling cycle. Cross-sectional SEM micrographs of the Al-BSF layers were taken to characterize the uniformity and measure the thickness of Al-BSF after firing. Fig. 7 shows that uniform Al-BSF layers with no disconnections were obtained for all the firing temperatures and times due to the high ramp-up rate. It has been shown that the junction uniformity is affected by the heating rate and the temperature gradient [19,20]. A comparison of the Al-BSF thicknesses measured from SEM images and calculated from Eq. (5) is shown in Fig. 8. The Al thickness from the Eq. (5) was obtained by measuring the weight of printed Al (rather than the thickness) since the printed Al exhibits a porous structure as shown in SEM images. A printed Al thickness of 25 mm corresponds to a calculated Al thickness of 17.2 mm due to the porosity of the printed Al layer, which was measured to be 31% by comparing weight of wafers before Al printing and after printing/drying/burning. The measured Al-BSF thickness was found to be greater than the calculated thickness using the Al–Si phase diagram at firing temperatures below 800 1C for firing times in the range of 1–60 s. Even for 1 s firing, the Al-BSF is thicker than the predicted value and exceeds 8 mm at temperatures over 740 1C. Amick et al. [21] observed that the measured Al-BSF thickness exceeded the predicted value by more than 50% for an

ARTICLE IN PRESS 1236

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

Fig. 7. Cross-sectional SEM micrographs of the Al-BSF region in FZ solar cells for different firing processes.

16 Measured thickness, 60 sec firing

BSF Thickness (µm)

14

Measured thickness,1 sec firing

12

Calculated thickness

10 8 6 4 2 0 690

710

730

750

770

790

810

Firing Temperature (°C) Fig. 8. Al-BSF thicknesses measured by SEM in FZ solar cells and calculated based on the Al–Si phase diagram.

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1237

Al thickness of 8 mm, whereas the Al-BSF thicknesses were in reasonable agreement with the predicted values for Al layers thicker than 28 mm. It has also been reported that the Al-BSF thickness agrees with the calculated value at typical firing conditions for an Al thickness of 37 mm (temperature range of 800–850 1C, time430 s) [22]. It is noteworthy that in this study, Al–Si alloying and an 8 mm thick Al-doped Si regrowth were accomplished in just a few seconds (Fig. 7). The difference between the measured and calculated Al-BSF thickness at lower temperatures suggests that thermodynamic equilibrium is not achieved, and the front and back surfaces of the wafers may be at different temperatures during RTP firing with a fast ramp-up rate. The experimental data in Fig. 9 suggests that there is indeed a temperature difference between the front and back surfaces, which depends on the ramp-up rate. In this study, the temperature was measured by a thermocouple in contact with the front surface of the wafer. Fig. 9 shows that for the same front surface temperature, 567 or 600 1C, a higher ramp-up rate triggered early melting of the Al on the backside, providing a thicker Al-BSF. The high ramp-up rate seems to generate a temperature gradient across the wafer with the higher temperature at the Al-Si surface, which increases the uniformity and thickness of Al-BSF [23]. For example,

Fig. 9. Cross-sectional SEM micrographs of Al–Si interface in FZ solar cells before and after annealing. The temperatures were measured by a thermocouple in contact with the front surface of wafers.

ARTICLE IN PRESS 1238

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

Fig. 10. LBIC mappings of solar cells fabricated on String Ribbon (a, b) and EFG (c, d) with conversion efficiencies. All the cells are annealed for 1 s at 740 1C (String Ribbon) and 750 1C (EFG).

at 567 1C, which is below the Al–Si eutectic temperature of 577 1C, a ramp-up rate of 50 1C/s showed some evidence of Al melting resulting in the formation of an Al-BSF in some regions. A slower ramp-up rate of 10 1C/s showed no evidence of BSF formation. Likewise at 600 1C, a ramp-up rate of 50 1C/s produced a thicker and more uniform Al-BSF region relative to the ramp-up rate of 10 1C/s. Fig. 10 shows LBIC mappings of solar cells made on String Ribbon and EFG Si. Some grain boundaries are visible and remain electrically active after hydrogenation whereas some grain boundaries are not visible or show only slight decrease in LBIC response. This means that the recombination around grain boundaries can be reduced by defect passivation by a proper hydrogenation process. It was found that even
18% efficient cells (a–d) contain some electrically active regions, so there is room for further improvement by a more effective defect passivation. PC1D modeling in Fig. 11 shows that 419% efficient ribbon Si cells can be achieved by a combination of lifetime enhancement from 100 to 200 ms, reduction in back surface recombination velocity from 350 to 200 cm/s, and use of lower bulk resistivity (2 Ocm) Si material.

4. Conclusions We have successfully combined the industrial processing steps of SiNx AR coating and screen-printed Al-BSF with the laboratory processes for double-layer AR

ARTICLE IN PRESS D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

1239

20 19

Efficiency (%)

18 Current cell with 18.2 % Bulk lifetime : 90 us

17 16

bulk = 3 Ω-cm, Sb = 350 cm/s bulk = 2 Ω-cm, Sb = 350 cm/s

15

bulk = 2 Ω-cm, Sb = 200 cm/s

14 13 12

0

20

40

60

80

100 120 Bulk lifetime (µs)

140

160

180

200

Fig. 11. Efficiency of photolithography cells as a function of bulk lifetime, bulk resistivity and back surface recombination velocity. Efficiencies over 19.2% can be achieved by increasing the lifetime up to 200 ms and the reducing bulk resistivity to 2 O-cm and back surface recombination velocity to 100 cm/s.

coating, RTP co-firing, and photolithography contacts to achieve record-high 17.8%-efficient String Ribbon solar cells and 18.2% EFG Si solar cells. The bulk lifetime in String Ribbon and EFG Si improved significantly after phosphorus diffusion followed by co-firing of SiNx and Al in RTP. The 1 s firing at 740–750 1C increased the bulk lifetime from 3–5 ms to 70–100 ms suggesting that release of hydrogen from the defects is the limiting factor for maximum hydrogenation at the typical firing temperatures. SEM analysis shows that Al-BSF with a thickness of 8 mm was formed with 1 s firing at temperatures greater than 740 1C and a ramp-up rate of over 50 1C/s. The measured Al-BSF thickness was greater than the calculated thickness at firing temperatures p800 1C, suggesting that thermodynamic equilibrium may not be achieved during short and rapid firing and a temperature gradient is established through the wafer. It was found that recombination around grain boundaries can be reduced by defect passivation due to a proper hydrogenation process. Model calculations show that ribbon solar cells with an efficiency of 19% can be achieved by improved bulk lifetime and back surface recombination velocity. References [1] V. Yelundur, A. Rohatgi, J-W. Jeong, A.M. Gabor, J.I. Hanoka, R.L. Wallace, in: Proceedings of the 28th Photovoltaic Specialists Conference, Anchorage, AK, 2000, pp. 91–94. [2] A. Rohatgi, J-W. Jeong, Appl. Phys. Lett. 82 (2003) 224–226. [3] G. Hahn, P. Geiger, Prog. Photovolt: Res. Appl. 11 (2003) 341–346.

ARTICLE IN PRESS 1240

D.S. Kim et al. / Solar Energy Materials & Solar Cells 90 (2006) 1227–1240

[4] G. Hahn, A.M. Gabor, in: Proceedings of the Third World Conference on Photovoltaic Energy Conversion, Osaka, 2003, pp.1289–1292. [5] J-W. Jeong, A. Rohatgi, V. Yelundur, A. Ebong, M.D. Rosenblum, J.P. Kalejs, IEEE Trans. Electron. Dev. 48 (2002) 2836–2841. [6] S. Peters, C. Ballif, D. Borchert, R. Schindler, W. Warta, G. Willeke, Semicond. Sci. Technol. 17 (2002) 677–681. [7] R.R.S. Thakur, R. Singh, Appl. Phys.Lett. 64 (1994) 327–329. [8] V. Yelundur, A. Rohatgi, J.W. Jeong, J. Hanoka, IEEE Trans. Electron. Dev. 49 (2002) 1405–1409. [9] A. Rohatgi, V. Yelundur, J-W. Jeong, D.S. Kim, A.M. Gabor, in: Proceedings of the Third World Conference on Photovoltaic Energy Conversion, Osaka, 2003, pp.1352–1355. [10] R.A. Sinton, A. Cuevas, M. Stuckings, in: Proceedings of the 25th Photovoltaic Specialists Conference, Washington, DC, 1996, pp. 457–460. [11] A.V. Wieringen, N. Warmoltz, Physica 22 (1956) 849–865. [12] J. Pearton, W. Corbett, M. Stavola, Hydrogen in Crystalline Semiconductors, Springer, Heidelberg, New York, 1991. [13] J. Crank, The Mathematics of Diffusion, Oxford University Press, New York, 1975. [14] V. Yelundur, A. Rohatgi, J.I. Hanoka, R. Reedy, in: Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, pp. 951–954. [15] B.L. Sopori, X. Deng, J.P. Benner, A. Rohatgi, P. Sana, S.K. Estreicher, Y.K. Park, M.A. Roberson, Sol. Energy Mater. Sol. Cells 41/42 (1996) 159–169. [16] B.L. Sopori, Y. Zhang, R. Reedy, in: Proceedings of the 29th Photovoltaic Specialists Conference, New Orleans, LA, 2002, pp. 222–226. [17] C. Dube, J.I. Hanoka, Appl. Phy. Lett. 45 (1984) 1135–1137. [18] J.D. Alamo, J. Eguren, A. Luque, Solid-State Electron. 24 (1981) 415–420. [19] S. Narasimha, A. Rohatgi, A.W. Weeber, IEEE Trans. Electron. Dev. 46 (2002) 1363–1370. [20] F.M. Roberts, E.L.G. Wilkinson, J. Mater. Sci. 3 (1968) 110–119. [21] J.A. Amick, F.J. Bottari, J.I. Hanoka, J. Electrochem. Soc. 141 (1994) 1577–1585. [22] P. Lolgen, C. Leguijt, J.A. Eikelboom, R.A. Steeman, W.C. Sinke, L.A. Verhoef, P.F.A. Alkemade, E. Algra, in: Proceedings of the 23rd Photovoltaic Specialists Conference, Louisville, KY, 1993, pp. 236–242. [23] V. Meemongkolkiat, M. Hilali, A. Rohatgi, in: Proceedings of the Third World Conference on Photovoltaic Energy Conversion, Osaka, 2003, pp. 1467–1470.