Journal of The Electrochemical Society, 156 共6兲 H495-H499 共2009兲
H495
0013-4651/2009/156共6兲/H495/5/$25.00 © The Electrochemical Society
The Study of Silane-Free SiCxNy Film for Crystalline Silicon Solar Cells M. H. Kang,a,z D. S. Kim,a A. Ebong,a B. Rounsaville,a A. Rohatgi,a G. Okoniewska,b and J. Hongb a
University Center of Excellence for Photovoltaics Research and Education, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0250, USA SiXtron Advanced Materials Incorporated, Dorval, Quebec H9P 1J1, Canada
b
We deposited plasma-enhanced chemical vapor deposition silicon carbon nitride 共SiCxNy兲 antireflection coating and passivation layers using a silane-free process. We used a solid polymer source developed at SiXtron Advanced Materials to eliminate the storage and handling of dangerous pyrophoric silane gas. We used ammonia flow rate as a control for the chemical and optical properties in the silane-free process. As NH3 flow rate increases, the carbon content, refractive index, extinction coefficient, and surface charge density of the film decrease. At an ammonia flow rate of 3000 sccm, which is similar to the conventional SiNx, the extinction coefficients for the two films were similar. This led to an emitter dark saturation current density 共Joe兲 of 404 fA/cm2 for the two films on 45 ⍀/䊐 emitters. However, a stack passivation of SiO2 /SiCxNy on an 80 ⍀/䊐 emitter resulted in an emitter dark saturation current density of 95 fA/cm2, which is enough to provide a good surface passivation for high efficiency solar cells. An energy conversion efficiency of 17.4% was obtained for a 149 cm2 textured Czochralski screen-printed solar cell with this stack passivation. For a 156 cm2 nontextured multicrystalline silicon, with only SiCxNy and a 45 ⍀/䊐 emitter, we obtained 14.9% efficiency. © 2009 The Electrochemical Society. 关DOI: 10.1149/1.3116225兴 All rights reserved. Manuscript submitted January 12, 2009; revised manuscript received March 9, 2009. Published April 23, 2009.
Plasma-enhanced chemical vapor deposition 共PECVD兲 silicon nitride 共SiNx兲 films are widely used for commercial silicon solar cells. SiNx films provide excellent surface passivation on phosphorus-doped emitters because of their high, positive fixed charge density, which causes an inversion or accumulation layer in p- or n-type Si under SiNx.1 The optimum refractive index of the antireflection 共AR兲 coating layer for an encapsulated solar cell is about 2.3,2 which is achievable by using silicon-rich SiNx films. However, such films absorb light at short wavelengths and decrease the quantum efficiency. Recently, Glunz et al.3 and Martin et al.4 reported on PECVD silicon carbide 共SiCx兲 films for the surface passivation of crystalline silicon 共c-Si兲. A surface recombination velocity of less than 30 cm/s at the SiCx /c-Si interface has been reported. SiCx films also have a high refractive index 共greater than 2.3兲, suggesting that SiCx may be an improvement over plasma-enhanced chemical vapor deposited SiNx. The deposition of SiNx and SiCx films typically requires dangerous pyrophoric silane 共SiH4兲, which is a significant inhibiting factor in the cost reduction economies-of-scale strategies employed by silicon photovoltaic plants. To eliminate the need for storage and handling of silane, SiXtron Advanced Materials has developed a silane-free solid-source process and apparatus. The solid source contains atomic sources such as silicon, carbon, and hydrogen; therefore, a silicon carbon nitride film 共SiCxNy兲 with different compositions can be obtained by adjusting the source composition during the deposition or by changing the NH3 flow rate. The current solid source can be adapted to the existing PECVD production line without any modification. In this paper, we studied the optical and electrical properties of dielectric films deposited with the silane-free solid source. Complete solar cells are fabricated with SiCxNy film and compared with the conventional SiNx coated cells.
X-ray photoelectron spectroscopy and elastic recoil detection. Saw damage on the as-cut wafers was removed by etching in KOH solution followed by anisotropic etching in the mixture of KOH and isopropyl alcohol for texturing. The textured silicon wafers were cleaned in HCl + H2O2 + H2O 共1:1:2兲 and rinsed in deionized water, followed by phosphorus diffusion in a quartz tube to form the emitter. The conventional SiNx AR coating layer with a thickness of 75 nm and a refractive index of ⬃2.04 was deposited in a low frequency 共50 kHz兲 PECVD reactor with silane and NH3 gases. The SiCxNy films were deposited in the same chamber using ammonia and gas produced from a solid polymer source. The solid source was a safe and stable polymer that was heated inside a sealed pressure vessel. The polymer was from a well-known polysiloxane family. The polymer underwent a thermally activated exothermic decomposition at a specific temperature vs pressure profile. During this decomposition a gas was produced which pressurized the vessel. The vessel was then used as a gas reservoir, and the gas delivery was controlled by a standard gas panel. The gas evolved contained various molecular species containing silicon, carbon, and hydrogen. The
metal contact SiCxNy SiCx Ny AR
n+ emi emitter tter depleti deple tion on regi regio on
Experimental We used a p-type textured Czochralski 共CZ兲 Si with a resistivity of 2–3 ⍀ cm and a multicrystalline Si 共mc-Si兲 with a resistivity of 1–2 ⍀ cm as substrates for the screen-printed solar cells 共Fig. 1兲. The optical properties of the dielectric films were characterized by a J. A. Woollam Co. Inc. variable angle spectroscopic ellipsometer. The composition of the dielectric films was analyzed by Auger
z
E-mail:
[email protected]
p-type pe b bulk ulk p-ty p+ BSF Figure 1. 共Color online兲 Complete structure of screen printing solar cell.
Downloaded on 2016-09-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 156 共6兲 H495-H499 共2009兲
H496 100%
O
O
O
N
N
O
O
O
N
N
N
C
C
90% 80% 70% Composition (%)
N 60% 50% C
40%
C
C
C 30% 20%
Si
Si
Si
Si
Si
Si
SiCN(NH3:1500)
SiCN(NH3:2000)
SiCN(NH3:2500)
SiCN(NH3:3000)
SiCN(NH3:4500)
10% 0%
(a)
SiNx(NH3:3000)
18 15.4
16
H Concentration (%)
14 12
Figure 2. 共a兲 Chemical composition and 共b兲 hydrogen concentration of the dielectric films as a function of NH3 flow rate in sccm. The flow rates of SiH4 and the gas from the solid source were fixed.
12.7 11.8
11.3
10
8.8
9
SiCN(NH3:3000)
SiCN(NH3:4500)
8 6 4 2 0
(b)
SiNx(NH3:3000)
SiCN(NH3:1500)
SiCN(NH3:2000)
SiCN(NH3:2500)
gas was supplied to the PECVD reactor through a standard silane mass flow controller assuming the same correction factor as for silane. Neither a gas condensation nor a liquid formation problem was observed in the gas delivery system. The carrier lifetimes in the wafers and the emitter dark saturation current density 共Joe兲 were measured using Sinton’s photoconductance tool. We used a SemiTest SCA-2500 surface charge analyzer to measure the charge density in the dielectrics. This tool allowed contactless and nondestructive measurement of the flatband 共FB兲 equivalent charge density 共QFB, the total charge density at the flatband condition兲 in the dielectric of interest. Finally, the front and rear contacts were formed by screen printing commercial Ag and Al pastes, respectively, followed by firing in a belt furnace. Results and Discussion Effect of NH3 gas flow rate on SiCxNy film deposition composition.— Figure 2a and b shows the Si/C/N/O chemical composition of SiCxNy films and the H content as a function of NH3 gas flow rate during the film deposition. Other deposition parameters including the flow rate of polymer gas source, deposition temperature 共425°C兲, pressure 共2 Torr兲, and plasma power 共150 W兲 were fixed for all the depositions in this work. Figure 2a shows no carbon in the SiNx film formed with silane gas. However, for films deposited with the source, the nitrogen composition in the film increases with an increase in NH3 gas flow rate while the carbon content decreases. Figure 2b shows a higher hydrogen content in SiNx films than in the SiCxNy films deposited with the 3000 sccm ammonia flow rate. In general, the hydrogen composition decreases with increasing NH3 gas flow rate for the SiCxNy films. Although the SiNx films show a
higher hydrogen content at the same NH3 flow rate, SiCxNy films supply enough hydrogen to passivate defects in the bulk and Si/SiCxNy interface during the contact firing. The silicon fraction in the dielectric SiCxNy films is constant 共around 31%兲, irrespective of the NH3 gas flow rate. Thus, by adjusting the NH3 gas flow rate, the carbon composition changes without affecting the silicon composition. Therefore the NH3 flow rate serves as a tool to adjust the chemical composition of the SiCxNy film to approach that of SiNx formed with SiH4 gas. Optical properties of SiCxNy films.— Figure 3 suggests that the refractive index 共n兲 and the extinction coefficient 共k兲 decrease with an increase in nitrogen content of the SiCxNy films. The NH3 gas flow rate in the range of 1500–4500 sccm was used to obtain a similar Si/N composition to that of the SiNx film. This was tailored to suit our screen-printed contact formation process, which is optimized for the conventional SiNx film. By adjusting the source composition and gas flow rates, SiCxNy films with a refractive index in the range of 1.93–2.00 at a wavelength of 630 nm were obtained.5 Emitter passivation quality of SiCxNy film and SiO2 /SiCxNy stack.— Figure 4 compares the Joe values for the silane-deposited SiNx film and the silane-free SiCxNy film on 45, 60, and 80 ⍀/䊐 emitters. Both films show similar Joe values for 45 and 60 ⍀/䊐 emitters, but the conventional SiNx showed a slightly lower value of Joe than the SiCxNy for the 80 ⍀/䊐 emitter.6 This suggests the presence of carbon in the SiCxNy films, which may interfere with the hydrogen released from the film to passivate the silicon surface. However, with the thermally grown SiO2 at the interface, the Joe gap
Downloaded on 2016-09-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 156 共6兲 H495-H499 共2009兲 2.050 2.030
H497
0.060 2.036
0.052
Ref. Inx.
0.050
Refractive Index(n)
2.010
0.040
2.000 1.990 0.031 1.970
0.026
0.030 0.024
1.950
1.950
0.019
0.020
Extinction Coefficient(k)
Ext. Coef.
Figure 3. Refractive indexes 共n兲 and extinction coefficient 共k兲 of the SiNx and SiCxNy films as a function of NH3 flow rate in sccm. The n and k values were measured at the wavelengths of 630 and 300 nm, respectively.
1.944 0.010
1.930
1.925 0.000
1.910 SiNx(NH3:3000)
SiCN(NH3:1500)
SiCN(NH3:2000)
SiCN(NH3:3000)
SiCN(NH3:4500)
Figure 4. Joe values of 45, 60, and 80 ⍀/䊐 emitters at 3000 sccm NH3 gas flow rate. The samples were annealed in a rapid thermal processing 共RTP兲 chamber at 800°C for 2 s.
decreased for the 80 ⍀/䊐 emitter. This suggests that the hydrogen released from the film can decrease the interface trap density at the Si/SiO2 interface. Surface charge and interface trap densities for SiCxNy.— Figure 5 shows the surface charge density QFB in the dielectric films. The surface charge density of dielectrics plays a critical role in controlling the surface passivation and solar cell performance.7 The positive surface charge density in the SiCxNy film was in the range of 共1.58–1.77兲 ⫻ 1012 cm−2, which is high but slightly lower than that of the conventional SiNx film 共1.89 ⫻ 1012 cm−2兲 for these specific samples. However, the surface charge density depends strongly on the deposition conditions for SiCxNy such as NH3 gas flow. Figure 6 shows the interface trap density 共Dit兲 between Si and dielectric film interface. The trap density decreases as carbon content decreases, which suggests that a low carbon concentration is desirable to achieve good surface passivation. However, increasing the ammonium flow rate to reduce the carbon content increases the nitrogen content and decreases the refractive index. This resulted in an optimum NH3 flow rate of ⬃3000 sccm. Figure 5. Surface charge densities and carbon concentration in SiNx and SiCxNy films as a function of NH3 gas flow rate after annealing in an RTP chamber at 800°C for 2 s.
High efficiency solar cells with SiCxNy and SiO2 /SiCxNy AR coating.— Both mono- and multicrystalline Si solar cells with SiCxNy films deposited with the solid polymer source were fabricated. The SiCxNy films were used for AR coating and emitter sur-
Downloaded on 2016-09-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 156 共6兲 H495-H499 共2009兲
H498 1.60E+12
30 1.47E+12 24.7
Carbon Con.
1.20E+12
25
1.12E+12 20
9.50E+11
1.00E+12
15.9
8.00E+11
15
13.1
6.00E+11 10 4.00E+11
3.33E+11
Carbon Concentration (%)
Interface Trap Density (/cm2)
1.40E+12
face passivation. Tables I and II show the characteristic parameters for the solar cells with AR layers deposited using SiH4 and the solid source. The best results were obtained when the ammonium flow was 3000 sccm. A cell efficiency of 17.4% was achieved with an 80 ⍀/䊐 emitter on the solar cells with SiO2 /SiCxNy AR coatings using the solid source. At 45 and 60 ⍀/䊐 sheet emitters, both Si sources provided cells with comparable Jsc, Voc, and fill factor 共FF兲 causing only a 0.1% efficiency difference. However, for a high sheet emitter 共80 ⍀/䊐兲, the Voc of cells with the SiNx film was only 9 mV higher than those with the SiCxNy film despite the nonoptimized deposition conditions. This difference in the Voc is caused by a lower Joe 共Fig. 4兲 for the SiNx film. The comparable Jsc supports good optical properties for the SiCxNy films. Figure 7 shows the internal quantum efficiency 共IQE兲 and reflectance values measured on the two types of cells with 45, 60, and 80 ⍀/䊐 emitters. At short wavelength, the IQE of the 45 and 60 ⍀/䊐 emitters are comparable. However, the 80 ⍀/䊐 emitter
Dit
5 2.00E+11 0 0
0.00E+00 SiNx(NH3:3000)
SiCN(NH3:1500)
SiCN(NH3:3000)
SiCN(NH3:4500)
Figure 6. Interface trap densities of SiNx and SiCxNy films as a function of NH3 gas flow rate after annealing in an RTP chamber at 800°C for 2 s.
Table I. Performance of 149 cm2 CZ solar cells with industrial-type phosphorus diffused 45, 60, and 80 ⍀Õ䊐 emitters. Antireflective coating 共ARC兲 SiNx SiCxNy SiNx SiCxNy SiNx SiCxNy SiO2 /SiNx SiO2 /SiCxNy
NH3 共sccm兲
Voc 共mV兲
Jsc 共mA/cm2兲
FF
Efficiency 共%兲
Emitter
3000 3000 3000 3000 3000 3000 3000 3000
624 622 620 618 623 612 632 625
35.1 34.8 36.1 35.8 36.3 36.2 36.2 36.1
0.777 0.780 0.763 0.766 0.767 0.775 0.770 0.772
17.0 16.9 17.1 17.0 17.4 17.2 17.6 17.4
45 45 60 60 80 80 80 80
Table II. Performance of 156 cm2 mc solar cells with industrial-type phosphorus diffused 45 ⍀Õ䊐 sheet resistance emitters.
SiNx SiCxNy
NH3 共sccm兲
Area 共cm2兲
Voc 共mV兲
Jsc 共mA/cm2兲
FF
Efficiency 共%兲
3000 2750
156 156
614 612
31.95 31.38
0.773 0.775
15.2 14.9
100
100
90
90
80
SiNx80
80
70
SiCN80 NH3:3000
70
SiNx60
IQE (%)
60
60
SiCN60 NH3:3000 50
50
SiNx45
40
40
SiCN45 NH3:3000
30
30
20
20
10
10
0 300
400
500
600
700
800
900
1000
1100
Reflectance (%)
ARC
Figure 7. 共Color online兲 IQE responses of the solar cells with SiNx AR layer coated using SiH4 source and SiCxNy AR layer coated using the solid source with 45, 60, and 80 ⍀/䊐 emitters.
0 1200
Wavelength (nm)
Downloaded on 2016-09-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 156 共6兲 H495-H499 共2009兲 coated with SiNx showed a higher IQE because of a better surface passivation, which was demonstrated by the lower Joe value. This calls for a further optimization of the deposition conditions for SiCxNy to obtain similar values. However, the responses are similar at long wavelength for both types of cells and for all the emitters. The stack passivation of SiO2 /SiCxNy on 80 ⍀/䊐 improved the surface passivation quality leading to a Joe value reduced by ⬃50%. This resulted in 0.2% efficiency improvement. The counterpart film showed similar improvement. This suggests that the SiCxNy film can replace the SiNx films with a slight loss in solar cell performance. Table II shows the preliminary results for mc-Si solar cells with SiNx and SiCxNy antireflection coatings. Cells coated with SiNx films show 0.3% difference in efficiency, mainly because of the nonoptimized SiCxNy thickness that resulted in a lower short-circuit current density. Conclusions A silane-free solid polymer source was developed at SiXtron Advanced Materials to eliminate the dangerous pyrophoric silane gas in a manufacturing line. The polymer source can be FeDexed, which significantly removes handling and transport issues. The electrical and optical properties of the SiCxNy film were studied and compared with those of the conventional SiNx film formed with
H499
pyrophoric SiH4 as a silicon source. The ratio of carbon to nitrogen 共C/N兲, hydrogen content, refractive index, and extinction coefficient of the SiCxNy film was found to decrease with the increase in the NH3 flow rate. Comparable quality surface passivation of the SiCxNy film because of the high surface charge density and hydrogen concentration provided low Joe values of ⬃4 ⫻ 10−13 A/cm2 on a 45 ⍀/䊐 emitter. Screen-printed Si solar cells with SiCxNy AR coating layer formed from the solid source gave an efficiency of 17.4% on the textured 149 cm2 CZ and 14.9% on the nontextured mc-Si. Georgia Institute of Technology assisted in meeting the publication costs of this article.
References 1. R. Hezel and R. Schorner, J. Appl. Phys., 52, 3076 共1981兲. 2. M. Green, Solar Cells: Operating Principles, Technology and System Applications, p. 163, The University of New South Wales, Kensington 共1998兲. 3. S. W. Glunz, S. Janz, M. Hofmann, T. Roth, and G. Willeke, in Proceedings of the Fourth WCPEC, IEEE Electron Devices Society, p. 4 共2006兲. 4. I. Martin, M. Vetter, A. Orpella, J. Puigdollers, A. Cuevas, and R. Alcubilla, Appl. Phys. Lett., 79, 2199 共2001兲. 5. H. Mackel and R. Ludemann, J. Appl. Phys., 92, 2602 共2002兲. 6. M. A. Shibib and J. G. Fossum, J. Appl. Phys., 52, 1072 共1981兲. 7. S. De Wolf, G. Agostinelli, G. Beaucarne, and P. Vitanov, J. Appl. Phys., 97, 063303 共2005兲.
Downloaded on 2016-09-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).