Experimental Study on the Interaction Between Al and P in Si Solution

The effect of impurity interactions on impurity removal efficiency in metallurgical grade silicon (MG-Si) was studied and analyzed to provide a new idea for removing impurities from MG-Si. This paper provides theoretical support for the removal of nonmetallic impurity P during the metallurgical purification and preparation of crystalline silicon. The interaction between Al and P in liquid silicon were investigated by directional solidification. The effects of different pull-down speeds, refining temperatures, holding times and Al contents on the distributions Al and P were investigated. The experimental results of the directional solidification of the Si-P-Al ternary system showed that from the bottom to the top of the sample, the contents of Al and P gradually increased. This phenomenon verifies the accuracy of the activity interaction coefficient between Al and P in the Si-P-Al ternary system calculated by the model. The mutual attraction between Al and P was also confirmed.


Introduction
Solar energy is abundant and widely distributed and is the most promising renewable energy source. Due to increasingly prominent global energy shortages and environmental pollution, solar photovoltaic power generation provides a clean, safe, convenient and efficient [1], method for generating energy in many countries around the world. The main raw material for making solar cells is solar-grade silicon (SoG-Si) [2]. However, impurities greatly impact siliconbased materials, especially high-purity silicon materials. The impurities in MG-Si must be maintained at the ppm level, and the minimum content of total metal impurities in the polysilicon matrix for solar cells is less than 0.2 ppmw. China MG-Si (GB/S2881-2014) has placed clear indicators for various impurities in MG-Si. To determine the best purification conditions, researchers have become interested in the purification of silicon [3], which has triggered in-depth research on the development of high-quality MG-Si. Therefore, removing impurities in silicon and processing it into photovoltaic polysilicon or even semiconductor polysilicon has become the direction of the entire industry.
MG-Si can reach the standard of solar-grade polysilicon materials only by greatly reducing the contents of Fe, Al, Ca, Ti, P, B, and other impurity elements. The contents of these typical impurities in MG-Si can be greatly reduced by refining in a furnace, including blowing refining [4], slagging refining [5], solvent refining [6,7], bubble flotation [8], electromagnetic purification [9], vacuum melting [10], and directional solidification [11]. For example, He [12] et al. purified MG-Si by injecting oxygen into the silicon lifting package and by refining it at 1600 ℃ for 2 h. Huang [13] et al. used directional solidification to remove impurities in silicon and studied the impurity removal mechanism. Tan Yi [14]  P is a difficult impurity to remove from silicon, and most phosphorus removal methods involve directional removal by furnace refining or electron beam refining [16][17][18]. Jafar [17] et al. used vacuum refining to remove phosphorus from silicon, and the results showed that upon increasing the temperature, the dephosphorization rate increases significantly, but this method required a high experimental temperature. Tan [19] et al. found that the phosphorus removal efficiency reached more than 99% after 1400 s at 21 kW power by electronbeam melting. Considering the defects of this method, such as the high experimental temperature and large silicon loss, it can not be used in large-scale production. Fang [20] et al. used slag refining to remove impurities in silicon, and the results showed that this method removed most metal impurities. However, the removal rate of the non-metallic impurity P was very low. Reducing P content in MG-Si can greatly improve the electrical performance of solar cells, therefore, it is of great significance to increase the theoretical research on phosphorus removal from MG-Si.
In the early stage of this experiment, scholars have carried out model calculation on it .Yan [21] et al. calculated that the infinite dilute activity coefficient of phosphorus in Si-P solution was ln 0 P =-2117∕T+ 0.20 based on the volume model of molecular interaction. Shimpo [22] et al. used a chemical equilibrium method that the self-activity interaction coefficient of phosphorus in Si solution at 1723 K was P P = 13.8 (±3.2) . Li [23] et al. used the molecular interaction volume model to calculate the self-interaction coefficient of phosphorus in liquid silicon within the temperature range of 1693-1873 K. The result was P P = -2.54 +13767.04∕T , which was consistent with the interaction coefficient determined by the chemical equilibrium method. He [24] et al. used the molecular interaction volume model to calculate the interaction coefficient of Al-P in liquid Si solution within the temperature range of 1693-1873 K. The activity interaction coefficients of Al to P in the Si-P-Al ternary system at 1773 and 1823 K measured by the same activity method were − 17.099 and − 16.211, respectively. The model calculation and experimental results showed that the activity interaction coefficients of Al on P were all negative, i.e., in a Si-P-Al ternary solution, the interaction between Al and P was greater than that between Si and P, and there were strong interaction between Al and P.
In this paper, the interaction between impurities Al and P during the refinement of MG-Si was studied by directional solidification experiments. The experimental results were compared to the results calculated by the molecular interaction volume model. The effect of impurity interactions on impurity removal efficiency during metallurgical silicon purification was demonstrated.

Experimental Principles
The most important factor for the removal efficiency of impurities by directional solidification is the element segregated coefficient. To describe the separation and coagulation characteristics of different impurities in silicon, the separation and coagulation coefficient K 0 was introduced [11].
where, Cs is the concentration of an impurity in solid-phase silicon; C 1 is the concentration of this impurity in liquid silicon at equilibrium with solid phase silicon.
The coagulation coefficient K 0 of the main impurities in silicon was measured and shown in Table 1. The K 0 value of P was close to 1, and segregation did not occur via directional solidification. The K 0 value of Al was 2 × 10 − 3 , which is far less than 1. The segregation of aluminum in silicon can occur via a segregation effect.
In the Si-P-Al ternary solution, compared with the interaction force between Si and P, if the interaction force relationship between member Al and P is f Al · P > f Si · P , then the presence of Al can reduce the activity coefficient of component P (Si · P + Al → Al · P + Si) and increase the solubility of P in Si. That is, in a Si-P-Al ternary solution, there are strong interactions between Al and P components, and the presence of Al is not conducive to the removal of impurity Al atoms in liquid silicon continuously segregate from the solid-liquid boundary to the Si solution, presenting a gradient distribution from the bottom to the top. During directional solidification, because the interaction force of Al on P is greater than that of Si on P, P will follow the segregation of Al. This causes a gradient distribution of P, thus verifying the consistency of the calculated results of the molecular interaction volume model and the experimental results.

Experimental Methods and Procedures
The reliability of the activity interaction coefficient of component Al to component P in the Si-P-Al ternary solution was verified by directional solidification experiments, i.e., the force between Al and P atoms was greater than that between Si and P atoms. The experimental raw materials used in this experiment were 0.055% Si-P parent alloy, high purity aluminum particles (99.99%).The purity of the argon gas 99.99%. The flow chart of this experiment is shown in Fig. 1.
The specific process of the experiment was: (a) First, the crushed Si-P base alloy and high-purity aluminum particles were mixed evenly in a certain proportion and placed in a high-purity graphite crucible (outer diameter: 60 mm, inner diameter: 55 mm, height: 120 mm). Then, it was heated and remelted in an intermediate-frequency induction furnace under an argon atmosphere to obtain a Si-P-Al alloy with a uniform composition. The Al and P contents determined were 1.09% and 0.028%, respectively; (b) After the Si-P-Al alloy was broken, it was placed in a high purity graphite crucible (outer diameter 50 mm, inner diameter 30 mm, height 100 mm). The graphite crucible was wrapped with thermal insulation cotton and then placed in the coil center of a mediumfrequency induction furnace. The furnace cover was closed, and a mechanical pump was used to pump the furnace into a vacuum state. Then argon gas was pumped after the pressure reached 0. The air outlet switch was opened so that the furnace was always under an argon atmosphere. Then, the heating switch was started. and the graphite crucible was heated at a rate of 303 K/min; (c) After the set temperature and holding time were reached, the pull-down device was opened and the crucible was used to perform the directional solidification experiment according to the preset pull-down speed. The specific experimental conditions are shown in Table 1; (d) After the graphite crucible was completely separated from the induction coil, the pull-down device was closed, the induction furnace switch was closed, and the power supply was turned off, so that the sample was cooled slowly to room temperature in the furnace. (e) After the sample was cooled to room temperature, the air inlet was closed and the sample is taken out. The sample is cut with a diamond wire cutting machine. One half of the samples was ground for content detection, and the other half was polished for micro morphology analysis.

Laboratory Equipment
A medium-frequency induction furnace is a commonly used laboratory silicon refining equipment. Its main body was composed of an induction furnace and a medium-frequency power controller. The refining process was protected by argon gas to refine MG-Si. Figure 2 shows the schematic diagram of the medium-frequency induction furnace. This experiment also used a diamond line cutting machine, polishing machine, electronic balance, scanning electron microscope, and plasma mass spectrometer.

Experimental Sample Processing and Testing
A diamond wire cutter was used to cut the samples obtained from the directional solidification experiment, as shown in Fig. 3. First, the sample was evenly cut into two parts along the vertical symmetry axis of the sample. Sample 1 was polished for EPMA micromorphology analysis. For analysis, sample 2 was evenly cut into five parts from bottom to top, which were numbered LB, LT, M, UB, and UT.
The experimental samples were cut and polished, and the phosphorus content was determined by inductively coupled plasma atomic emission spectrometry (ICP). SEM-EDS and EPMA were used to observe the microstructure and element distribution of the sample.

Effects of Different Pull-Down Speeds on Al and P Distributions
The effects of different pull-down speeds on the Al and P distributions during directional solidification were studied by using a constant refining temperature of 1823 K and a holding time of 30 min. The specific experimental conditions are shown in Table 2. After cutting the samples as shown in Fig. 3, they were analyzed, and the results are  upper surface of silicon melt during this process concentrated phosphorus and facilitated phosphorus volatilization. Thus, P diffused to the surface of silicon melt and formed a volatile gas that was removed. Therefore, the P content at UT on the top of the sample was different from that in other regions, so it was omitted in the analysis. The same was true for subsequent experiments. Figure 4 shows that, excluding sample processing error, the Al and P contents gradually increased from the bottom to the top of the sample. This shows that Al carried some P with it during the directional solidification process, i.e., there were interaction forces between Al and P. This verifies the accuracy of the activity interaction coefficient between Al and P in the Si-P-Al ternary solution calculated by the model. It can be seen from the figure that when the pull-down speed was 7 μm·s − 1 , the P content in regions LB, LT, M, and UB was lower than those at 10 μm·s − 1 and 15 μm·s − 1 .

Effects of Different Refining Temperatures on the Distribution of Al and P
The effects of different refining temperatures on the Al and P distributions during directional solidification were studied next. The pull-down speed was 7 μm·s − 1 , and the holding time was 30 min. The specific experimental conditions are shown in Table 2, and the experimental results are shown in Fig. 5. Al and P were gradually enriched from the bottom to the top during directional solidification at different refining temperatures. This shows the mutual attraction between Al and P. That is, Al carried P upward and segregated under the action of interaction forces. It can be seen from the figure that at 1823 K, the P content in each part was the lowest, indicating that more P was brought to UT at the top. Therefore, 1823 K was selected as the refining temperature for subsequent experiments.

Effects of Different Holding Times on the Distribution of Al and P
The effects of different holding times on Al and P distributions during directional solidification were studied. The pull-down speed was 7 μm·s − 1 , and the refining temperature was 1823 K. The specific experimental conditions are shown in Table 2, and the experimental results are shown in Fig. 6. From the bottom to the top of the sample, the Al and P contents still showed a gradual increase, but the difference between their contents in the same region was small, indicating that the enrichment degree of elements had no obvious relationship with the holding time. This was presumably because, in the intermediatefrequency induction furnace, the electromagnetic forming of the silicon melt mixing effect improved the diffusion dynamics of the impurity elements in the silicon melt. However, during the heat preservation phase, the electromagnetic force did not change, and upon increasing the holding time, the system dynamics did not significantly change. Therefore, the same parts of the gap between the impurity content were smaller. Considering the enrichment of impurities and energy consumption, the insulation time was set to 30 min.

Effects of Different Al Contents on the Distribution of Al and P
The effects of different Al contents on the distribution of Al and P during directional solidification were studied. The pull-down speed was 7 μm·s − 1 , the refining temperature was 1823 K, and the holding time was 30 min. The specific experimental conditions are shown in Table 2, and the experimental results are shown in Fig. 7. Al and P were gradually enriched from the bottom to the top of the sample upon changing the Al content. Upon increasing the Al content, the P content of the sample decreased gradually from the bottom to the top. This indicates that more P was carried to the top of the sample due to the segregation of Al, i.e., increasing the Al content enriched P. However, upon increasing the Al content, the gap between the P content in different parts of the same sample gradually narrowed. Even when the Al content reached 5.0%, the P contents in parts LB, LT, M, and UB were identical. This phenomenon indicates that the amount of P carried by Al in the upward segregation process was limited, that is, Al cannot improve the separation conditions of P infinitely. The results of the directional solidification experiment of the Si-P-Al ternary system showed that the Al and P contents gradually increased from the bottom to the top of the sample, which verified the accuracy of the interaction coefficient of Al activity on P in the Si-P-Al ternary system calculated by the model. This also confirmed the existence of mutual attraction between Al and P.

The Microstructure Characterization of Experimental Samples
In this study, electron probe microanalysis (EPMA) was used to characterize the micromorphology of the samples at regions LB, LT, M, UB, and UT at the pull-down speed of 7 μm·s − 1 , refining temperature of 1823 K, holding time of 30 min, and Al content of 1.1%. The results are shown in Fig. 8. From the bottom to the top of the sample, Al and P contents gradually increased, indicating that the impurity content gradually increased and was enriched.
To better prove the interactions between Al and P in the Si-P-Al ternary system, i.e., that Al carried some P to the top together during directional solidification, the enriched areas in different parts of the sample were characterized by EPMA. The characterization results are shown in Fig. 9. In the bottom LB, LT, and middle M regions of the sample, Al and P impurity phases were not enriched. Due to the extremely low content, EPMA scanning failed to provide an image. Al and P were highly enriched in regions UB and UT at the top of the sample, forming Al-P alloy phases. At the same time, Si will wrap Al-P or grow with Al and P and Si will wrap it. As can be seen from the EPMA scanning results of regions UB and UT, the enriched regions of P and Al were highly overlapped, and P and Al were always distributed together, which is completely consistent with the detection results of the above conditional experiment. The scanning results confirmed that Al carried some P to the top during directional solidification, indicating that there were strong interactions between Al and P, further verifying the activity interaction coefficient of Al to P.

Conclusion
In the Si-P-Al ternary solution, P and Al enrichment areas highly overlapped, and the distribution of P and Al always appeared together. However, this also confirmed that during directional solidification, Al carried some P and segregated to top together, showing the strong interactions between Al and P and further verifying the activity interaction coefficients between Al and P. In the Si-P-Al ternary system, the interaction forces between Al and P were greater than those between Si and P. The impurities in MG-Si were removed by directional solidification and impurity Al was also removed. On the contrary, the removal of impurity P will be easier if impurity Al in MG-Si is removed by acid leaching before the removal of impurity P by vacuum refining, plasma refining, or electron-beam refining.