2017 Volume 58 Issue 10 Pages 1439-1443
In the present work, an aluminum co-doping method was employed to improve the electrical properties of multi-crystalline silicon containing phosphorous and boron. The method is based on the compensation theory and on the different segregation coefficients of aluminum (kAl = 2.8 × 10−3), boron (kB = 0.8), and phosphorous (kP = 0.35) in silicon during solidification. The carrier type and carrier concentration throughout the height of the silicon ingot with co-doping of aluminum were measured. The results show that no polarity inversion occurs and a relatively low carrier concentration, i.e., low net doping concentration can be achieved with aluminum co-doping. Moreover, the resistivity and minority carrier lifetime throughout the height of the silicon ingots with and without co-doping of aluminum were measured. The results show that a relatively high and uniform resistivity and minority carrier lifetime can be achieved by adding a controlled amount of aluminum. In addition, the effect of the grain boundary on the electrical properties was studied.
A promising path to reduce the cost of solar energy is the use of low-cost silicon (Si) feedstock, or upgraded-metallurgical-grade Si (UMG-Si) feedstock, obtained by simplified purification processes1). One drawback of UMG-Si is that it contains both acceptor (boron) and donor (phosphorous) atoms in a relatively large amount, which can be considered an inevitable issue2).
Owing to compensation defined as the presence of donor and acceptor species in comparable amounts, it is possible to tolerate high concentrations of phosphorous (P) and boron (B) that are frequently difficult to remove completely by means of metallurgical refining methods3). Previous results4) have shown that in compensated Si of this sort, the minority carrier lifetime to be higher and thus solar cells to be more efficient at high compensation levels (Cl), i.e., low net doping, NA–ND. However, because of the lower segregation coefficient of P (0.35) compared to that of B (0.8), the net doping is not uniform along the ingot height and inversion of Si polarity, from p- to n-type, may occur during crystallization5). This reduces the material yield for industrial solar cell fabrication. Simulations6) using Scheil equation7) and simple models for calculating carrier mobility show that requiring a minimum resistivity of 0.5 Ω･cm and an ingot yield of at least 90% compels the use of Si feedstock containing less than 0.45 ppmw of B and 1.0 ppmw of P, which is much lower than that in common UMG-Si feedstock.
In this study, to overcome the limitations of B and P concentration in Si feedstock, we explore a method for controlling the net dopant concentration through the entire ingot height by the addition of a well-defined mix of B, P and Aluminum (Al). Al, same as B, can serve as acceptor too. The triple doping of Al, B, and P relies on the low segregation coefficient of Al (2.8 × 10−3), which enables the increased P concentration (relative to that of B) to be counterbalanced during crystal growth. The study focuses on the determination of the Al amount added to common UMG-Si feedstock (CB,0 = 1 × 1017 cm−3, CP,0 = 2 × 1017 cm−3) to get Si ingot with uniform net doping density and the improvement of the electrical properties, such as the minority carrier lifetime and resistivity, after Al co-doping. The effect of grain boundaries on the electrical properties is also characterized.
The common UMG-Si feedstock contains about 1 × 1017 cm−3 of B and 2 × 1017 cm−3 of P, which is n-type material. However, after directional solidification, which is indispensable when making Si ingot, the vertical distribution of dopants concentration along multi-crystalline Si (mc-Si) ingots follows the Scheil equation7):
| \[ C_{\rm I} = C_{{\rm I}, 0} \times k_{\rm I} \times (1 - f_{\rm S})^{k_{\rm I} - 1} \] | (1) |
Based on eq. (1), the distribution of the net dopant concentration along the Si ingot made from UMG-Si feedstock is as shown in Fig. 1(a). Owing to the different segregation coefficients between B and P, the net dopant concentration is not uniform throughout the ingot height and an inversion of the polarity type appears in Si ingot, which limits the material yield.
Calculated (a) net doping distribution profile of P-B-compensated Si and (b) net doping distribution profile of P-B-Al-compensated Si and distribution profiles of each dopants along the ingot height.
In order to suppress the polarity type inversion and obtain a relatively low and uniform net dopant concentration, the amount of Al added to the UMG-Si should qualify the following equation:
| \[ \begin{split} & 0 < C_{\rm B,0} \times k_{\rm B} \times (1 - f_{\rm S})^{k_{\rm B} - 1} + C_{\rm A1,0} \times k_{\rm A1} \times (1 - f_{\rm S})^{k_{{\rm A}1} - 1}\\ &\quad - C_{\rm P,0} \times k_{\rm p} \times (1 - f_{\rm S})^{k_{\rm P} - 1} < 1 \times 10^{17} \end{split} \] | (2) |
Then by adding 1 × 1019 cm−3 of Al, more than 90% of the ingot exhibit a net dopant concentration less than 1 × 1017 cm−3 and the polarity type inversion is suppressed, as is shown in Fig. 1(b). As a comparison, the distribution profiles of each dopant are also shown in Fig. 1(b).
The carrier type and carrier concentration, i.e., net doping concentration of the Si ingot after Al co-doping will be verified by Hall measurement and the improvement of the electrical properties of it will also be validated by experiments.
For this study, mc-Si ingots with a diameter of ～12 mm were grown using the directional solidification method. Commercially available high-purity solar-grade single crystal Si feedstock (SOG-Si feedstock) (>99.9999%) was used, to which known concentrations of B, P, and Al were added (Table 1). The concentrations of B and P correspond to those of UMG-Si feedstock. The added Al concentration was based on the calculation results shown in the last section. B and P were introduced through highly doped Si wafers, for which the doping impurity concentration was evaluated by means of four-point probe resistivity measurements correlated to the Irvin curves. Al was introduced in the form of high-purity Al powder (99.95%). All the dopant sources were accurately weighed, and then placed in the high-purity graphite crucible (99.999%) together with the SOG-Si feedstock before melting. All the Si ingots were grown with the same apparatus and same procedures: an average lowering rate of v = 0.2 mm/min and a temperature gradient of 1.1℃/mm under Ar gas protection. The graphite crucible keeps the oxygen (O) concentration in an extremely low range during the directional solidification. Therefore the effect of the X-O complex can be ignored. The experimental apparatus is shown in Fig. 2.
| Dopant | CB,0(cm−3) | CP,0(cm−3) | CAl,0(cm−3) |
|---|---|---|---|
| SOG | 0 | 0 | 0 |
| P | 0 | 2 × 1017 | 0 |
| Al | 0 | 0 | 1 × 1019 |
| B B + P |
1 × 1017 1 × 1017 |
0 2 × 1017 |
0 0 |
| B + P + Al | 1 × 1017 | 2 × 1017 | 1 × 1019 |
Experimental apparatus for the directional solidification.
The resulting ingots were then sliced into 4-mm-thick wafers along the ingot height. After mechanical polishing, four-point probe resistivity measurement (RT-70V/RG-7C, NAPSON) was performed on the wafers to determine the electrical activation of dopants. After surface passivation by 0.01 mol/L Quinhydrone/Methanol solution for 1 h in the dark, the minority carrier lifetime was measured using the Microwave Photo Conductivity Decay (μ-PCD) method (HF-300, NAPSON). Then, after etching using 200 g/L NaOH solution for 10 min, the average grain size was measured. At last, the graphite outer shell of the wafers was removed and Hall measurement (LakeShore 8400 series Hall effect measurement system) was conducted to determine the carrier type and carrier concentration.
All of the measured Hall voltages VH of P-B-Al-compensated Si samples along the ingot height were positive, which meant all the samples were p-type as shown in Table 2. This indicates that the polarity inversion is avoided in most part of the ingot and the material yield is increased by Al co-doping method.
| Ingot height (%) | 14.3 | 28.6 | 42.9 | 57.1 | 71.4 | 85.7 |
|---|---|---|---|---|---|---|
| VH (10−5 v) | 2.87 | 2.71 | 1.88 | 1.78 | 1.30 | 0.98 |
| Carrier type | P | P | P | P | P | P |
The calculated and the measured carrier concentration, i.e., net doping concentration in the P-B-Al-compensated Si samples along the ingot height are shown in Fig. 3. The measured carrier concentrations ranging from 1.36 × 1016 to 3.99 × 1016 cm−3 are in the same order of the calculated values ranging from 3.78 × 1016 to 6.62 × 1016 cm−3, although the former one is lower than the latter one. Moreover, the measured carrier concentration varies in a quite narrow range, which indicates Al co-doping effectively controls the net doping density low and uniform along the entire ingot height.
Calculated and measured carrier concentration of P-B-Al-compensated Si along the ingot height.
Figure 4 shows the resistivity measured at various points along the ingot height. We can see that the measured resistivity of P-B-Al-compensated Si ingot ranges from 0.9 to 2.7 Ω･cm, a proper range for the fabrication of high-performance solar cells. In addition, the resistivity of the P-B-Al-compensated Si ingot is comparable with that of SOG-Si ingot and much higher than those of other uncompensated Si ingots (grown from SOG-Si feedstock doped with B, P or Al), although they were fabricated with same apparatus and procedures. As to the P-B-compensated Si ingot (without Al co-doping), only the samples at 14.4% and 28.6% of the ingot height show good resistivity for the solar cells, 1.4 and 0.6 Ω･cm respectively, while the resistivity at higher ingot height is too low. According to the calculation in section 2, there should be an inversion of the polarity at about 26% of the ingot height. At ingot height higher than 26%, the dominant dopant is P. Therefore the low resistivity at higher ingot height is due to the rapidly increased P concentration.
Resistivity along the ingot height of different kinds of Si ingots.
It is well known that for uncompensated Si, the resistivity is given by:
| \[ \rho = 1/(\mu \times {\rm n}_0 \times {\rm e}) \] | (3) |
According to the available experimental evidence and the widely accepted model of Klaassen8), the net dopant mobility is lower in compensated Si than in uncompensated Si with the same net doping. At equal net dopant density n0, the resistivity will generally be higher in compensated Si than in uncompensated Si9). As a result, the resistivity specifications defined for uncompensated Si are not sufficient to determine the suitability of the compensated Si material. We also need to measure the most important property, which is the minority carrier lifetime of Si for solar cells.
Figure 5 shows the minority carrier lifetime measured at various points along the ingot height. For the P-B-Al-compensated ingot, the measured lifetime lies in the range 70–130 μs along the entire ingot. This value is much higher than those of uncompensated ingots (the minority carrier lifetime of P-doped Si ingot is too low to be detected) and even higher than that of most part of the SOG-Si ingot. As to the P-B-compensated Si ingot, only the sample at 14.4% of the ingot height show good minority carrier lifetime about 92 μs. After the polarity inversion, the minority carrier lifetime shows a severe decrease and is even too low to be detected at some part of the ingot.
Minority carrier life time along the ingot height of different kinds of Si ingots.
These results highlight the possibility of fabricating high-performance solar cells with Si feedstock containing high concentrations of B and P by Al co-doping method.
The high minority carrier lifetime of the P-B-Al-compensated Si samples can be explained by a recombination mechanism. There are three main recombination mechanisms in semiconductors: radiative recombination, Auger recombination, and recombination through defects. Radiative and Auger recombination are intrinsic recombination mechanisms. The recombination rate for p-type Si is given by:
| \[ \begin{split} R_{\rm intrinsic} ={} & n \times p \times (6 \times 10^{-25} \times {p_{0}}^{0.65}\\ & + 3 \times 10^{-27} \times \varDelta p^{0.8} + 9.5 \times 10^{-15}) \end{split} \] | (4) |
| \[ \tau_{\rm intrinsic} = \varDelta p/R_{\rm intrinsic} \] | (5) |
Figure 6(a) shows the trend of the minority carrier lifetime of SOG-Si ingot and P-B-Al-compensated Si ingot on liner scale. As a comparison, the trend of the grain size along the ingot is shown in Fig. 6(b).
(a) minority carrier lifetime on a linear scale and (b) grain size trend of SOG-Si and P-B-Al-compensated Si along the ingot height.
From Fig. 6, we can see that the minority carrier lifetime of SOG-Si ingot increases with the grain size. This is expected, as a larger grain size leads to a shorter grain boundary length and hence, fewer recombination centers. However, the minority carrier lifetime of the compensated Si ingot increases first and then decreases as the grain size increases continuously. This phenomenon can be explained by Al segregation at the grain boundaries. The Al segregation at the grain boundaries to a certain extent makes the grain boundaries electrically active. These kinds of grain boundaries can act as strong recombination centers. The increase of the grain size and the increase of the Al concentration at the grain boundaries have opposite effects on the minority carrier lifetime resulting in the trend inversion.
In addition, the segregation of Al at grain boundaries also has effects on the resistivity. From Fig. 4 we can see a drastic decreasing trend of the resistivity along the ingot height in the P-B-Al-compensated Si. The Al prefers to segregate at the grain boundaries. When the concentration of Al at grain boundaries increases to a given extent, the insulating grain boundaries will become electrically active, resulting in the decrease of the resistivity.
An Al co-doping method to increase the properties of mc-Si containing both P and B has been proposed, based on the compensation theory and Scheil's law. After calculation, the amount of Al added to the P-B-containing Si feedstock (UMG-Si) was determined. The carrier type and carrier concentration were measured and the electrical properties such as resistivity and minority carrier lifetime were determined to verify this method. The conclusions are as follows:
(1) The net dopant concentration of P-B-Al-compensated Si ingot remains uniform along the entire ingot height compared with that of uncompensated Si ingot and P-B-compensated Si ingot.
(2) The electrical properties of P-B-Al-compensated Si ingot are better than those of uncompensated Si ingot and P-B-compensated Si ingot, and are comparable to those of SOG-Si ingot after the same directional solidification procedure.
(3) The Al co-doping method effectively eliminates some of the adverse effects of B and P in Si. Thus, it may be used to increase the B and P concentration limits allowed in Si feedstock to produce high-quality solar cells.
(4) The grain boundary has a significant effect on the electrical properties and further studies are recommended.
