Control of the oxidation process is one key issue in producing high-quality emitters for crystalline silicon solar cells. In this paper, the oxidation parameters of pre-oxidation time, oxygen concentration during pre-oxidation and pre-deposition and drive-in time were optimized by using orthogonal experiments. By analyzing experimental measurements of short-circuit current, open circuit voltage, series resistance and solar cell efficiency in solar cells with different sheet resistances which were produced by using different diffusion processes, we inferred that an emitter with a sheet resistance of approximately 70 Ω/□ performed best under the existing standard solar cell process. Further investigations were conducted on emitters with sheet resistances of approximately 70 Ω/□ that were obtained from different preparation processes. The results indicate that emitters with surface phosphorus concentrations between 4.96 × 1020 cm−3 and 7.78 × 1020 cm−3 and with junction depths between 0.46 μm and 0.55 μm possessed the best quality. With no extra processing, the final preparation of the crystalline silicon solar cell efficiency can reach 18.41%, which is an increase of 0.4%abs compared to conventional emitters with 50 Ω/□ sheet resistance.

The quality of the emitter has an important effect on the efficiency of a solar cell because the p-n junction is the core of the crystalline silicon (c-Si) solar cell. Currently, emitters in industrialized p-type c-Si solar cells are produced primarily by the method of high-temperature phosphorus (P) diffusion, and the factors that affect emitter quality include the surface P concentration and the junction depth of the emitter. An emitter obtained by the high-temperature diffusion method has a high surface P concentration that can even exceed the solubility of P in Si at the temperature of 850 °C (about 1 × 1021 cm−3) and form a “dead layer”.1,2 The P atoms in the “dead layer” is electrically inactive. Instead, they form recombination centers that increase the Auger recombination and cause the saturation current density to increase.3,4 Simultaneously, the quantum efficiency in the blue wavelength range becomes poor and the short-circuit current (Jsc) is reduced. Furthermore, a high surface P concentration can result in high composite current density of the emitter, which reduces the passivation effect and lowers the open circuit voltage (Voc).5 

Reducing the surface P concentration can reduce the surface recombination and improve the quantum efficiency in the blue wavelength range, but the lateral transfer resistance of the emitter and the contact resistances (Rs) of the front electrode and emitter both increase, therefore, the loss on photocurrent will increase and the fill factor (FF) will decrease. Minimization of the “dead layer” decreases the junction depth and increases Voc, but the operational windows for subsequent printing and sintering processes are reduced such that it becomes easy to produce defective products during production. Due to continuous improvements in positive silver pastes in recent years, most silver pastes on the market can produce an emitter with good ohmic contact whose surface P concentration is approximately 2 × 1020 cm−3. The Solamet® PV17x series silver introduced by DuPont can even produce emitters with surface concentrations of 1 × 1020 cm−3,6 which shows how current emitter preparation processes have great potential for improvement. With the restriction that we must ensure good contact with the electrode, it is important to determine the most suitable technology in terms of surface concentration and junction depth.

Therefore, this paper primarily focuses on optimization of the high sheet resistance structure of the emitter.7–10 As for oxygen plays an important role in diffusion, so the oxidation process is discussed in this study in detail. The optimized process can be applied easily in industrial production lines. The influence of oxygen concentration and oxidation diffusion time on the quality of the emitter was explored to optimize the surface concentration and junction depth of the emitter.

Boron-doped p-type Czochralski (Cz) Si (100) wafers with 2 Ω·cm resistivity and 180 μm thickness were used. After cleaning and texturization, a 156 × 156 mm2 p-type mono-crystalline Si wafer was doped through a diffusion process by using POCl3 in a tube furnace. The conventional diffusion process is shown in Table I. To optimize the emitter performance, the following parameters were manipulated in an orthogonal experiment: pre-oxidation time, oxygen concentration during pre-oxidation and pre-deposition, drive-in time (see Table II); the other diffusion parameters and steps stayed constant (see Table I). To reduce the complexity of the experiment, we did not consider the reciprocity of each parameter.

Table I.

Conventional diffusion process.

  Region 1Region 2Region 3Bubbling N2Diluting 
StepTime/sT1/°CT2/°CT3/°C(POCl3)/sccmN2/sccmO2/sccm
Input 420 850 850 850 25000 
Stable 600 863 866 870 25000 
Pre-oxidation 900 863 866 870 22000 3000 
Pre-deposition 1100 863 866 870 1800 20000 3000 
Drive-in 300 850 850 850 22000 3000 
Output 420 850 850 850 25000 
  Region 1Region 2Region 3Bubbling N2Diluting 
StepTime/sT1/°CT2/°CT3/°C(POCl3)/sccmN2/sccmO2/sccm
Input 420 850 850 850 25000 
Stable 600 863 866 870 25000 
Pre-oxidation 900 863 866 870 22000 3000 
Pre-deposition 1100 863 866 870 1800 20000 3000 
Drive-in 300 850 850 850 22000 3000 
Output 420 850 850 850 25000 
Table II.

Overview of the diffusion parameters for experimental groups utilizing oxidation parameter variations. The flux of POCl3 was a constant 1200 sccm. 1000 sccm = 1 L/min.

 Pre-oxidationO2 flux duringO2 flux duringDrive-inThe ratio of POCl3 to O2
Projecttime/spre-oxidation/sccmpre-deposition/sccmtime/sflux during pre-deposition
G1 1000 1500 600 180 2: 1 
G2 1000 3000 1200 300 1:1 
G3 1000 4500 1800 420 2:3 
G4 1000 6000 2400 600 1:2 
G5 2000 1500 1200 420 1:1 
G6 2000 3000 600 600 2:1 
G7 2000 4500 2400 180 1:2 
G8 2000 6000 1800 300 2:3 
G9 3000 1500 1800 600 2:3 
G10 3000 3000 2400 420 1:2 
G11 3000 4500 600 300 2:1 
G12 3000 6000 1200 180 1:1 
G13 4000 1500 2400 300 1:2 
G14 4000 3000 1800 180 2:3 
G15 4000 4500 1200 600 1:1 
G16 4000 6000 2400 420 2:1 
 Pre-oxidationO2 flux duringO2 flux duringDrive-inThe ratio of POCl3 to O2
Projecttime/spre-oxidation/sccmpre-deposition/sccmtime/sflux during pre-deposition
G1 1000 1500 600 180 2: 1 
G2 1000 3000 1200 300 1:1 
G3 1000 4500 1800 420 2:3 
G4 1000 6000 2400 600 1:2 
G5 2000 1500 1200 420 1:1 
G6 2000 3000 600 600 2:1 
G7 2000 4500 2400 180 1:2 
G8 2000 6000 1800 300 2:3 
G9 3000 1500 1800 600 2:3 
G10 3000 3000 2400 420 1:2 
G11 3000 4500 600 300 2:1 
G12 3000 6000 1200 180 1:1 
G13 4000 1500 2400 300 1:2 
G14 4000 3000 1800 180 2:3 
G15 4000 4500 1200 600 1:1 
G16 4000 6000 2400 420 2:1 

After the diffusion finished, the Rsheet was tested by the four-probe method, and an ECV instrument was used to analysis the junction depth. Partially wafers were selected for the active phosphorus concentration (APC) measurement. They were treated to a 30 min additional oxidation at 900 °C after removing the phosphorus silicate glass (PSG), in this process the flux of O2 and Diluting N2 were constant 3000 sccm and 22000 sccm respectively. According to the research of Bazer,11 the quality of the emitter was characterized by the APC, which can be calculated as the Rsheet value after the additional oxidation process divided by the Rsheet value after conventional diffusion. For other diffused wafers, PSG layer and the back p-n junction were removed and a SiNx antireflection coating layer was deposited on the front surface using PECVD. The front and rear metallic contacts were produced by screen printing silver and aluminum pastes. The cells were completed after sintered in a belt furnace. Measurements were conducted of the I-V curve, internal quantum efficiency (IQE) and Rs. Moreover, we applied IR to analyze cell defects after sintering.

In this paper, 16 groups of orthogonal experiments are presented. The results of sheet resistance measurements and I-V tests after diffusion are shown in Table III; the APC values are results from only part of the wafers.

Table III.

Results of sheet resistance measurements and I-V tests.

GroupRsheet(Ω/□)Jsc(mA/cm2)Voc(mV)FF(%)Rs(Ω)Rsh(Ω)Eff(%)APC(%)a
G1 58.64 36.8 625.8 77.81 0.008 6.09 17.92 51.20 
G2 63.42 37.0 626.3 77.65 0.008 12.88 17.95 68.26 
G3 73.78 37.2 629.7 77.12 0.009 35.69 18.03 96.98 
G4 92.37 37.2 629.0 60.95 0.013 0.73 14.25 113.35 
G5 60.07 37.0 628.3 78.05 0.008 25.31 18.13 74.91 
G6 53.34 36.8 627.0 78.08 0.008 10.68 17.98 62.97 
G7 110.94 37.4 624.8 43.08 0.021 0.17 10.05 121.08 
G8 84.96 37.4 621.0 74.59 0.010 5.66 17.30 97.33 
G9 68.86 37.4 624.0 79.19 0.008 16.98 18.41 103.07 
G10 116.98 37.3 625.8 47.84 0.018 0.21 11.15 117.64 
G11 61.04 37.1 628.3 78.09 0.008 11.81 18.16 63.06 
G12 78.73 37.2 631.0 76.80 0.008 11.56 18.03 81.42 
G13 63.58 37.1 627.5 78.05 0.008 52.83 18.15 66.13 
G14 102.5 37.4 629.5 66.58 0.012 2.36 15.64 95.75 
G15 69.08 37.1 630.7 78.18 0.008 115.53 18.29 93.95 
G16 66.36 37.2 629.0 78.35 0.008 110.72 18.32 76.16 
GroupRsheet(Ω/□)Jsc(mA/cm2)Voc(mV)FF(%)Rs(Ω)Rsh(Ω)Eff(%)APC(%)a
G1 58.64 36.8 625.8 77.81 0.008 6.09 17.92 51.20 
G2 63.42 37.0 626.3 77.65 0.008 12.88 17.95 68.26 
G3 73.78 37.2 629.7 77.12 0.009 35.69 18.03 96.98 
G4 92.37 37.2 629.0 60.95 0.013 0.73 14.25 113.35 
G5 60.07 37.0 628.3 78.05 0.008 25.31 18.13 74.91 
G6 53.34 36.8 627.0 78.08 0.008 10.68 17.98 62.97 
G7 110.94 37.4 624.8 43.08 0.021 0.17 10.05 121.08 
G8 84.96 37.4 621.0 74.59 0.010 5.66 17.30 97.33 
G9 68.86 37.4 624.0 79.19 0.008 16.98 18.41 103.07 
G10 116.98 37.3 625.8 47.84 0.018 0.21 11.15 117.64 
G11 61.04 37.1 628.3 78.09 0.008 11.81 18.16 63.06 
G12 78.73 37.2 631.0 76.80 0.008 11.56 18.03 81.42 
G13 63.58 37.1 627.5 78.05 0.008 52.83 18.15 66.13 
G14 102.5 37.4 629.5 66.58 0.012 2.36 15.64 95.75 
G15 69.08 37.1 630.7 78.18 0.008 115.53 18.29 93.95 
G16 66.36 37.2 629.0 78.35 0.008 110.72 18.32 76.16 
a

APC of more than 100% that appear in the table can occur due to the spread of P to the outside during the additional oxidation process.12 

From Table III, we can see that Jsc increases when Rsheet increases. The group of G2, G3, G6, G7 and G8 were selected for the IQE test; results are shown in Fig. 1. Rsheet increases with increasing APC, which results in lower Auger recombination and enhanced the quantum efficiency in the blue wavelength range. According to the method proposed by Bazer11 wherein APC can be used to characterize the quality of the emitter, we made a diagram of the change of Jsc with Rsheet and APC, respectively. Figure 2 shows the linear fitting and a comparison of the linear correlation R-values. From the results of the linear fitting, we find that the relevance of the APC to the Jsc is higher than the relevance of the Rsheet. Therefore, we can use this method to characterize emitter quality rather than using the ECV method.

FIG. 1.

IQE curve of the Si solar cell produced using different diffusion processes.

FIG. 1.

IQE curve of the Si solar cell produced using different diffusion processes.

Close modal
FIG. 2.

(a) The linear relationship between Rsheet and Jsc and (b) the linear relationship between APC and Jsc.

FIG. 2.

(a) The linear relationship between Rsheet and Jsc and (b) the linear relationship between APC and Jsc.

Close modal

Table III also shows that as Rsheet was increasing, the value of Voc initially increased and then decreased. The Voc can reach over 630 mV for the maximum value of Rsheet within the range from 70 Ω/□ to 80 Ω/□. The reason why Voc increases with increasing Rsheet is that the decrease of the front surface concentration contributes to a decrease of the Auger recombination rate. As a result, the reverse saturation current density is reduced. The Voc is inversely proportional to the reverse saturation current density, so it increases with increasing Rsheet. Voc begins to decrease when Rsheet climbs over 80 Ω/□ because the p-n junction is too shallow to be burned through easily; this fact is obvious if one observes the shunt resistance. We selected G2, G3, G4, G6, G7 and G8 for infrared thermal image analyses and electroluminescence (EL) tests; the results are shown in Fig. 3.

FIG. 3.

(a) Infrared thermal mappings of the wafers diffused using different processes and (b) EL mappings of the wafers diffused using different processes.

FIG. 3.

(a) Infrared thermal mappings of the wafers diffused using different processes and (b) EL mappings of the wafers diffused using different processes.

Close modal

From Fig. 3, we can see that the cells performed quite differently for different values of Rsheet under the same sinter process. In G2, G3 and G6, no obvious defects were found when Rsheet was less than 80 Ω/□. However, when Rsheet reached 84.96 Ω/□ as in G8, defects began to appear. Conversely, more obvious defects resulted from the sintering process for the higher Rsheet series represented by G4 and G7. Based on the large black zones in Fig. 2(b) for G4 and G7, we know that the electrodes of the cells were in poor contact with the emitter and even the p-n junctions were already burned through. These results are used to explain why the FF decreased rapidly after Rsheet exceeded 80 Ω/□.

Series resistance increased with increasing Rsheet. One explanation is the increased resistance of the diffusion layer, and another is that the decrease to a lower surface concentration worsened the ohmic contact between the electrodes and solar cells, which increased the contact resistance. We measured the contact resistances of G2, G3, G4, G6, G7 and G8 by using CoreScan. The corresponding results are displayed in Fig. 4(a) and Fig. 4(b), which show the trend of contact resistance with changes in Rsheet. We found a tendency for the contact resistance to stay consistent with the series resistances.

FIG. 4.

(a) The distribution of contact resistance and (b) correlation of CoreScan-measured contact resistance versus Rsheet.

FIG. 4.

(a) The distribution of contact resistance and (b) correlation of CoreScan-measured contact resistance versus Rsheet.

Close modal

Based on the final efficiency values, we conclude that solar cells with emitters having Rsheet values of approximately 70 Ω/□ presented the best performance compared with those of approximately 50 Ω/□ because their Jsc and Voc both increased. In addition, the FF remained at a higher level, and the peak efficiency reached 18.41%, which was an increase of 0.4%abs over conventional Rsheet.

Because the screen printing and sintering processes have not been optimized, the cell with Rsheet at 80 Ω/□ presented a low FF and low efficiency although the latter still reached 18.03%. A higher efficiency is anticipated when further progress is made. For this type of cell, the rapid decline in FF made the efficiency decline very steep for Rsheet over 80 Ω/□. To achieve a higher FF, the most important factor is better metallization performance even over optimizations in screen printing and sintering. However, we will not discuss this case in this paper.

According to the research of Bazer,11 the emitter quality can be characterized by the APC. Figure 5 shows the correlation between Rsheet and APC. We find that emitters with the same Rsheet values have quite different APCs when they are produced using different processes. Taking the upper blue area in Fig. 5 as an example, we observe the phenomenon that the APCs of emitters with Rsheet approximately 70 Ω/□ are higher than expected. This result indicates that the quality of these three emitter groups is better. These three groups of experiments are G3, G9 and G15. To allow further analysis of these three groups of experiments, we selected the G16 process for comparison. The ECV method was used to test the P concentrations of these four groups, and the results are shown in Fig. 6.

FIG. 5.

Correlation of active phosphorus concentration with sheet resistance.

FIG. 5.

Correlation of active phosphorus concentration with sheet resistance.

Close modal
FIG. 6.

Profiles of phosphorus concentration. The results were measured using the ECV method.

FIG. 6.

Profiles of phosphorus concentration. The results were measured using the ECV method.

Close modal

Compare the difference of four diffusion process. G9 has longer drive-in time than G3, which makes the surface concentration of the G9 lower and the junction depth deeper. G15 allows more time to oxidize than G9 and the oxygen concentration is kept at a higher level. As a result, the surface phosphorus concentration of G15 is reduced. Because the ratio of oxygen to POCl3 during the diffusion is 1:1 in G15 while that of G9 is only 2:3. From Fig. 6, we can see that the depth of the p-n junction is greater for G15 than for G9, which indicates that a low level of oxygen concentration during diffusion can increase the junction depth. Compared to G9, G16 allows less time for drive-in, but the oxidation time is longer, which makes the surface phosphorus concentration low. The oxygen concentration of G16 during the diffusion is lower, and the ratio of oxygen to POCl3 is 2:1. Although the drive-in time is short, the wafers still have deeper junction depths than those of G9. From the above analysis, we draw the conclusion shown in Fig. 7.

FIG. 7.

Influence of diffusion parameters on the surface phosphorus concentration and the junction depth.

FIG. 7.

Influence of diffusion parameters on the surface phosphorus concentration and the junction depth.

Close modal

Compared to G9, the surface phosphorus concentration of G3 is higher but the p-n junction is shallower. For this type of high surface concentration and shallow junction, recombination in the emitter is more serious and the quantum efficiency in the blue wavelength range is low. Consequently, the Jsc is low as seen from the IQE in Fig. 8(a). In this figure, we see that the IQE of G9 in the wavelength range from 300-350 nm is higher than that of G3. Within the wavelength range of 350–500 nm, the IQE of G3 is slightly higher relatively speaking.

FIG. 8.

(a) Contrast between G3 and G9 IQE data and (b) contrast between G9 and G15 IQE data.

FIG. 8.

(a) Contrast between G3 and G9 IQE data and (b) contrast between G9 and G15 IQE data.

Close modal

Compared to G9, the surface concentration of G15 is lower. There is fewer Auger recombination taking place on the surface, and it theoretically has a good response to short wavelengths. However, photon-generated carriers produced by high-frequency waves will have already recombined before they arrive at p-n junction as a result of the excessively deep junction. Generally speaking, although the surface concentration of the shallow junction G9 is high, it still has a good response to high-frequency waves as shown in Fig. 8(b); this is why the Jsc value of G9 is higher than that of G15. The Voc of G15 is 630.7 mV, which is higher than the 624.0 mV of G9. This difference occurs mainly because the surface phosphorus concentration of G15 is lower, which causes less Auger recombination on the surface, the Voc increases as surface recombination decreases. The FF of G9 is 79.19%, which is higher than in G15. This results mainly because the surface concentration of G9 is higher. With good ohmic contact between the electrodes and the silicon wafer, the contact resistance is lower and the series resistance is reduced such that the FF increases. This is proven by the contact resistance analyzed by using CoreScan in Fig. 9. Figure 9 shows that the contact resistance of G9 is 7.93 mΩ·cm2, which is slightly smaller than the 8.02 mΩ·cm2 of G15.

FIG. 9.

Contact resistance mappings of G9 (left) and G15 (right).

FIG. 9.

Contact resistance mappings of G9 (left) and G15 (right).

Close modal

From the efficiency point of view, higher surface concentration and shallower junction depth of G9 give a higher efficiency than that observed in G15, which shows that G9 also has better quality as an emitter.

In this paper, we analyzed the short-circuit currents, open-circuit voltages, series resistances and efficiencies of solar cells that have different sheet resistances and are produced by different diffusion processes. We concluded that an emitter with a sheet resistance of approximately 70 Ω/□ performed best in the existing standard process. Further optimizations of the oxidation processes suggest that an emitter prepared using an oxidation process with a phosphorus surface concentration between 4.96 × 1020 cm−3 and 7.78 × 1020 cm−3 and with a junction depth between 0.46 μm and 0.55 μm will have the best quality. With no extra solar cell production processes, the final preparation of the crystalline silicon solar cell efficiency can reach 18.41%, which is an increase of 0.4%abs over conventional emitters with 50 Ω/□ sheet resistance.

This work was supported by the National Natural Science Foundation (contract number 61176055) and a grant from the Science and Technology Project (2011A080804009) of Guangdong Province, China.

1.
M.
Finetti
,
P.
Negrini
,
S.
Solmi
, and
D.
Nobili
, “
Electrical properties and stability of supersaturated phosphorus-doped silicon layers
,”
Journal of the Electroche-mical Society
128
,
1313
1317
(
1981
).
2.
S.
Solmi
,
A.
Parisini
,
R.
Angelucci
, and
A.
Armigliato
, “
Dopant and carrier concentration in Si in equilibrium with monoclinic Si P precipitates
,”
Physical Review B
53
,
7836
7841
(
1996
).
3.
R. L.-Y.
Sah
,
Robert N.
Noyce
, and
William
Shockley
, “
Carrier Generation and Recombination in P-N Junctions and P-N Junction Characteristics
,”
Proceedings of the Institute of Radio Engineers
45
(
9
),
1228
1243
(
1957
).
4.
S.
Solmi
and
D.
Nobili
, “
High concentration diffusivity and clustering of arsenicand phosphorus in silicon
,”
Journal of Applied Physics
83
,
2484
2490
(
1998
).
5.
J.
Ostoja
,
F.
Gueri
,
P.
Negrini
, and
S.
Solmi
, “
The effects of phosphorus precipitation on the open circuit voltage in n /p silicon solar cells
,”
Solar Cells
11
,
1
12
(
1984
).
7.
Mohamed M.
Hilali
,
Kenta
Nakayashiki
, and
Abasifreke
Ebong
, “
Ajeet Rohatgi. High-efficiency (19%)Screen-printed Textured Cells on Low-resistivity Float-zone Silicon with High Sheet-resistance Emitters. Progress in photovoltaics: Research and Applications
,”
Appl.
14
,
135
144
(
2006
).
8.
M.
Hörteis
and
S. W.
Glunz
, “
Fine line printed silicon solar cells exceeding 20% efficiency. Progress in Photovoltaics: Research and Applications
,”
Appl.
16
,
555
560
(
2008
).
9.
V.
Yelundur
,
K.
Nakayashiki
,
M.
Hilali
, and
A.
Rohatgi
, “
Implementation of a homogeneous high-sheet-resistance emitter inmulticrystalline silicon solar cells
,” in
IEEE. Conference Record of the 31th IEEE Photovoltaic Specialists Conference
(
IEEE
,
New York
,
2005
) pp.
959
962
.
10.
A.
Ebong
,
V.
Upadhyaya
,
B.
Rounsaville
,
D. S.
Kim
,
K.
Tate
, and
A.
Rohatgi
, “
18% large area screen-printed solar cells on textured MCZ silicon with high sheet resistance emitter
,” in
IEEE. Conference Record of the 4th IEEE Photovoltaic Energy Conversion
(
IEEE
,
New York
,
2006
) pp.
1326
1329
.
11.
B.
Bazer-Bachi
,
E.
Fourmond
,
P.
Papet
,
L.
Bounaas
,
O.
Nichiporuk
,
N.
Le Quang
, and
M.
Lemiti
, “
Higher emitter quality by reducing inactive phosphorus
,”
Solar Energy Materials & Solar Cells
105
,
137
141
(
2012
).
12.
P.
Negrini
,
D.
Nobili
, and
S.
Solmi
, “
Kinetics of phosphorus predeposition in silicon using POCl3
,”
Journal of the Electrochemical Society
122
,
1254
1260
(
1975
).