- Email: [email protected]
Fast determination of impurities in metallurgical grade silicon for photovoltaics by instrumental neutron activation analysis
Applied Radiation and Isotopes 69 (2011) 1365–1368
Contents lists available at ScienceDirect
Applied Radiation and Isotopes journal homepage: www.elsevier.com/locate/apradiso
Fast determination of impurities in metallurgical grade silicon for photovoltaics by instrumental neutron activation analysis J. Hampel a,n, F.M. Boldt a, H. Gerstenberg c, G. Hampel b, J.V. Kratz b, S. Reber a, N. Wiehl b a
Fraunhofer Institute for Solar Energy Systems, Heidenhofstrasse 2, D-79110 Freiburg, Germany Institute of Nuclear Chemistry, Johannes Gutenberg-University of Mainz, Fritz-Strassmann-Weg 2, D-55128 Mainz, Germany c ZWE FRM-II der Technischen Universit¨ at M¨ unchen, D-85748 Garching, Germany b
a r t i c l e i n f o
abstract
Article history: Received 22 December 2010 Received in revised form 12 April 2011 Accepted 19 May 2011 Available online 27 May 2011
Standard wafer solar cells are made of near-semiconductor quality silicon. This high quality material makes up a significant part of the total costs of a solar module. Therefore, new concepts with less expensive so called solar grade silicon directly based on physiochemically upgraded metallurgical grade silicon are investigated. Metallurgical grade silicon contains large amounts of impurities, mainly transition metals like Fe, Cr, Mn, and Co, which degrade the minority carrier lifetime and thus the solar cell efficiency. A major reduction of the transition metal content occurs during the unidirectional crystallization due to the low segregation coefficient between the solid and liquid phase. A further reduction of the impurity level has to be done by gettering procedures applied to the silicon wafers. The efficiency of such cleaning procedures of metallurgical grade silicon is studied by instrumental neutron activation analysis (INAA). Small sized silicon wafers of approximately 200 mg with and without gettering step were analyzed. To accelerate the detection of transition metals in a crystallized silicon ingot, experiments of scanning whole vertical silicon columns with a diameter of approximately 1 cm by gamma spectroscopy were carried out. It was demonstrated that impurity profiles can be obtained in a comparably short time. Relatively constant transition metal ratios were found throughout an entire silicon ingot. This led to the conclusion that the determination of several metal profiles might be possible by the detection of only one ‘‘leading element’’. As the determination of Mn in silicon can be done quite fast compared to elements like Fe, Cr, and Co, it could be used as a rough marker for the overall metal concentration level. Thus, a fast way to determine impurities in photovoltaic silicon material is demonstrated. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Instrumental neutron activation analysis Transition metals Metallurgical grade silicon HCl gas gettering
1. Introduction In standard silicon solar modules, the costs of silicon wafers ¨ make up 50% of the total costs (Gotzberger et al., 2002). One reason is the high costs for near-semiconductor quality feedstock with purities of 99.99999% (7 N) to 99.999999% (8 N), as purification steps are time and resource consuming. Low-cost silicon, such as metallurgical grade (MG) and upgraded metallurgical grade (UMG) silicon with purities of approximately 2–3 N and 3–5 N could be used for new solar cell concepts, although there are various issues one has to face regarding this material. One of the crucial problems is the high amount of metallic impurities, for example Fe or Cr. These contaminations degrade the minority carrier lifetime and therefore the solar cell efficiency by forming
n
Corresponding author. Tel.: þ49 761 4588 5247; fax: þ 49 761 4588 9250. E-mail addresses: [email protected], [email protected] (J. Hampel). 0969-8043/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.apradiso.2011.05.024
recombination centers. Istratov et al. (2006) demonstrated that the distribution of metals in multicrystalline silicon has a crucial impact on the lifetime. The cell efficiency is less affected by metals in precipitates than by interstitial impurities even at lower densities. Nevertheless, starting with MG silicon, it is still important to reduce the overall concentration of transition metal impurities. Directional solidification and gettering are two ways of silicon purification. Crystallization of silicon feedstock is an early step in the solar cell process whereby the amount of metallic impurities is reduced by several orders of magnitude. The silicon feedstock is melted in a crucible and crystallized to an ingot by unidirectional solidification, which occurs from the bottom upwards. Thereby, the metallic impurities are accumulated at the top of the ingot due to the very small segregation coefficients of metals in silicon. For example, Fe and Cr have segregation coefficients of 7 10 6 and 3 10 6, respectively (Weber, 1983). The resulting silicon ingot is cut into wafers, which contain very different amounts of metals depending on the horizontal position in the former ingot.
1366
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
In addition, gettering steps are considered to be essential in the cell process to further reduce the metal content. In this work, a new gettering technique with hydrogen chloride (HCl) gas was applied. The gettering effect is due to the removal of metals from the surface of the wafers through the formation of volatile chloride compounds (Istratov et al., 2000). These compounds are transported away by the stream of gas, and further diffusion of the metals from the bulk to the wafer surface occurs. The diffusion depends on the diffusion coefficient of the respective metals. For example, Fe has a diffusion coefficient of 4 10 6 cm2/s at 1100 1C (Weber, 1983). Both above-mentioned procedures of silicon purification need to be evaluated by analytical techniques to learn about their effectiveness. A suitable analysis method is needed for the detection of metallic impurities in solar grade silicon. The ideal analysis technique should be able to detect detrimental transition metals, such as Fe and Cr, over a wide concentration range in solar grade silicon with very low detection limits. The technique should be a fast procedure without extensive sample preparation, ideally without the need to dissolve the silicon. This reduces the risk of contamination during sample preparation. Instrumental Neutron Activation Analysis (INAA) is well known and established in many different research and application areas. A lot of experience has been made applying INAA to transition metal impurities in silicon, especially in crystalline silicon (see e.g. Wiehl et al. (1982), Verheijke et al. (1989) and Istratov et al. (2003)). However, the determination of transition metals, such as Fe, Cr, and Co in silicon, which is performed using long-lived nuclides, takes a relatively long time. In this work, INAA was investigated as a possible analytical technique to control the above-mentioned procedures of silicon purification with the goal of a high sample throughput. Due to the segregation coefficients, the ratios of various transition metals along the crystallization front should be constant. Therefore, it was tested if the concentration profile of the transition metals could be determined by measuring only the profile of Mn as the ‘‘leading element’’. Mn has a high activation cross-section of 13.3 b and a half-life of 2.58 h. Consequently, quick scans of the impurity profile along the crystallization front should be possible. Therefore, we investigated the analysis of silicon columns cut from an ingot along the crystallization front.
Garching, respectively. After irradiation and decay of the background, which is caused by the beta emitter 31Si, the samples were measured with a high purity germanium (HPGe) semiconductor detector with a relative efficiency of 27.7% at the Institute of Nuclear Chemistry at the University of Mainz. Similar INAA measurements were also carried out with wafers, which were treated by HCl gas gettering and analyzed together with untreated neighboring wafers as reference to evaluate the efficiency of this new gettering technique. The determination of Ni was achieved by the 58Ni(n,p)58Co reaction. All other elements were determined by (n,g) reactions. For all analyses, high purity liquid standards of the elements Cr, Mn, Fe, Co, and Ni were irradiated and measured together with the samples. For data evaluation, the software Genie 2000 V2.1 (Canberra) was used. The applied irradiation protocols are listed in Table 1.
2.2. Sample preparation and analysis of Si columns Vertical columns with a size of approximately 1 1 20 cm3 were cut out of an ingot, which was crystallized out of upgraded metallurgical grade silicon feedstock. These columns were irradiated as a whole in the central irradiation position at TRIGA Mainz for 6 h at a neutron flux of 4 1012 cm 2 s 1. Standards of Mn and As were placed at the top and bottom of the column for the determination of the absolute metal content. After a decay time of 15 h, the columns were measured with a germanium detector with a relative efficiency of 67% whereby a collimator of lead with 10 cm thickness and a slit width of 1 cm was placed in front of the germanium detector (Fig. 1). For the measurements the columns were divided in two pieces of 10 cm length. The influence of non-perfect shielding of the collimator, i.e. gamma rays penetrating the lead, can be estimated from the mass attenuation factors of lead for the relevant energies. The estimations show that within all other uncertainties of the measurements the collimator can be assumed to be nearly perfect. The columns were scanned in 1 cm steps. The 31Si activity was used to control and correct deviations of the neutron flux along the column. The overall uncertainties were calculated from all contributing errors, which are listed in the following in the order
2. Experiments 2.1. Sample preparation and analysis of Si wafers After unidirectional crystallization of MG-Si feedstock with 99.3% purity, the ingot was cut into wafers. From three different parts of the ingot (top, middle, and bottom), wafers were cut into samples with a size of approximately 2.5 1.25 0.027 cm3 and a weight of approximately 200 mg. Afterwards, the samples were surface cleaned by chemical polishing (etching of the silicon surface by a mixture of hydrofluoric acid, acetic acid, and nitric acid) and irradiated at the research reactor TRIGA Mark II in Mainz (Hampel et al., 2006) and at the research reactor FRM-II in
Fig. 1. Schematic (top view) of collimator for scanning Si columns (not to scale).
Table 1 Irradiation protocols used for analyses of the wafers after crystallization. Program no.
Irradiation facility
Irradiation time (h)
Decay time
Measuring time (h)
Neutron flux (cm 2 s 1)
Investigated elements
1 2 3
TRIGA Mainz TRIGA Mainz FRM-II Garching
33 3 90
approx. 8 d approx. 3 h approx. 21 d
15 0.5 15
4.2 1012 4.2 1012 1.1 1014
Ni Mn Cr, Fe, Co
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
1367
of an increasing magnitude: the imperfection of the collimator, the weighting error of the standard and sample, the geometric error of the sample compared to the standard, and the error of the counting statistics.
3. Results 3.1. Detection limits The detection limits (in ng/g) of the INAA results are given in Table 2. Very low values, especially for Co, Cr, and Mn could be achieved. This is necessary because the acceptable value, e.g., of Cr in solar grade silicon wafers before solar cell processing is around 1 ng/g (Hofstetter et al., 2009). In the silicon feedstock, a higher impurity is allowed because of the purification capacity of the subsequent crystallization step. This higher impurity level makes the analysis much easier. The detection limit for Fe is higher because 58Fe, the nuclide which is utilized for the determination, has a low nuclide abundance (0.28%) and a relatively low neutron capture cross-section (1.3 b) (Magill et al., 2006). Nevertheless, the resulting detection limit is still acceptable since Fe has a higher diffusivity than Cr and can therefore be gettered more easily. The determination of Ni was achieved by the 58 Ni(n,p)58Co reaction, which occurs with fast neutrons. The detection was only possible at the TRIGA Mainz because the difference between the fluxes of thermal and fast neutrons is much smaller than that at the FRM-II Garching. The half-life of 56 Mn (2.6 h) is in the same range as the half-life of the activated silicon nuclides, but is produced from 55Mn, which has a nuclide abundance of 100% and a neutron capture cross-section of 13.3 b. These properties together with the low half-life allow a fast and reliable detection of Mn by INAA.
Fig. 2. Metal mass fractions (in mg/g) measured by INAA in metallurgical grade silicon wafers depending on the position in the ingot.
Table 3 Fe-metal concentration ratio in different ingot positions. Position in ingot
Fe–Mn concentration ratio
Fe–Cr concentration ratio
Fe–Co concentration ratio
Fe–Ni concentration ratio
Bottom Middle Top
28 7 2 23 7 2 21 7 2
117 1 2107 20 2007 20
7707 70 1360 7 110 1340 7 110
– 550 7 50 340 7 30
3.2. Si wafers The INAA results of the crystallized silicon ingot are presented in Fig. 2, showing transition metal impurity mass fractions as a function of the position in the ingot. The increase in the impurities from the bottom to the top is expected as the ingot was obtained by unidirectional solidification from the bottom upwards. The segregation coefficients of transition metals are much lower than 1, so the impurities are accumulated at the top of the ingot. While the metal concentration strongly depends on the position of the wafer in the ingot, the concentration ratios between different transition metals were found to be nearly the same throughout the ingot. This is particularly the case for the ratio of Fe and Mn concentration, which is quite similar in different positions of the ingot (Table 3), although absolute Fe concentrations increase by three orders of magnitude from bottom to top. The ratios of other metal concentrations show some deviation at the bottom of the ingot. Nevertheless, once the metal concentration ratios in an ingot and one metal profile are measured, rough profiles of some other metals could be calculated. However, this Table 2 Detection limits (in ng/g) of transition metals in approx. 200 mg silicon measured by INAA in Mainz and Garching. Element
Detection limit (ng/g)
Irradiation facility
Cr Mn Fe Co Ni
0.4 0.9 23 0.1 82
FRM-II Garching TRIGA Mainz FRM-II Garching FRM-II Garching TRIGA Mainz
Fig. 3. Mn mass fraction (in mg/g) measured by INAA in metallurgical grade silicon neighboring wafers with and without 30 min of HCl gas gettering.
assumption has to be investigated also for other feedstock materials and crystallization processes. Furthermore, this can only be assumed if no gettering was done before, because it was shown that transition metals are reduced by HCl gas gettering by a different magnitude (Hampel et al., 2009). At least, Mn can serve as a marker for the degree of purification effectiveness by unidirectional solidification. The INAA results for Mn in the wafers to which HCl gas gettering was applied are presented in Fig. 3. Gettering with HCl gas for 30 min reduces the Mn concentration in silicon drastically. A reduction of 96% can be achieved. This corresponds to a gettering efficiency of 27, which is defined as concentration of a specific element before gettering, divided by the respective concentration after gettering. Most other transition metals can be reduced by HCl gas gettering in the same way (Hampel et al., 2010).
1368
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
could be detected by INAA, from less than ng/g to mg/g. The sample preparation consists of only one surface cleaning step and the risk of contamination is much lower compared to other analyses because the silicon does not have to be dissolved before analysis. To accelerate the detection of transition metals, first experiments of locally scanning whole vertical silicon columns were accomplished, demonstrating that it is a feasible method. The fact that similar transition metal ratios were found throughout the entire ingot and the fact that Mn can be detected by INAA in a fast and reliable way, leads to the conclusion that Mn could serve as a model for other transition metals. This means that rough profiles of other transition metals can be calculated using the Mn profile. This assumption has still to be proven for various silicon feedstock materials and crystallization processes. Nevertheless, Mn can already be used as a marker for the degree of purification effectiveness by unidirectional solidification.
Acknowledgments Fig. 4. As profile of a silicon column, which was irradiated in one piece and scanned in approximately 1 cm steps by gamma spectroscopy.
3.3. Si columns In the first tests of scanning whole vertical silicon columns, a detection limit for Mn of 5 ng/g could be achieved. This is at least 1 order of magnitude below the critical impurity concentration. With the goal to use the Mn concentration as measure for all transition elements, this result shows that the used irradiation and measurement conditions are well-suited for these studies. The reliability of the measurements could be checked by the profile of the As concentration. The As concentration profile should follow Eq. (1), which is called Scheil equation (Scheil, 1942). CS ðxÞ ¼ kC0 ð1xÞk1
ð1Þ
where CS is the concentration in the solid, x is the relative ingot height, k is the segregation coefficient, and C0 is the starting concentration. With k¼0.3 for As in Si (Hull, 1999) and an As starting concentration of C0 ¼0.108 mg/g, the data are in good agreement with the theory (Fig. 4). If Mn could be used as a model for other transition metal profiles, the analysis of a Si ingot would be much faster. The determination of Mn by the presented Si column scanning method could be done in less than 24 h. For comparison, the determination of Fe or Co in a Si column takes 2–3 weeks.
4. Conclusion INAA is well-suited for the detection of transition metals in solar grade silicon. It was shown that the metals of interest, such as Cr, Mn, and Fe, can be determined with sufficient detection limits. A large range of transition metal concentrations in silicon
The authors would like to thank PV Crystalox for providing the Si wafers from MG-Si and the operators of the FRM-II Garching and the TRIGA Mainz for performing the irradiations. Furthermore, the help from the crystallization and wafering team at the Fraunhofer ISE is gratefully acknowledged. References ¨ Gotzberger, A., Luther, J., Willeke, G., 2002. Solar cells: past, present, future. Sol. Energy Mater. Sol. Cells 74, 1–11. Hampel, G., Eberhardt, K., Trautmann, N., 2006. Der TRIGA forschungsreaktor mainz. Atw Int. Z. Kernenergie 51 (5), 328–330. Hampel, J., Schmich, E., Boldt, F.M., Wiehl, N., Hampel, G., Kratz, J.V., Reber, S., 2009. Gettering of metallurgical grade silicon by HCl gas. In: Proceedings of the 24th European Photovoltaic Solar Energy Conference, Hamburg, Germany, pp. 2557–2559. Hampel, J., Boldt, F.M., Wiehl, N., Hampel, G., Kratz, J.V., Reber, S., 2010. Use of HCl gas gettering in the epitaxial wafer equivalent concept. In: Proceedings of the 25th European Photovoltaic Solar Energy Conference, Valencia, Spain, pp. 3591–3593. Hofstetter, J., Lelievre, J.F., del Canizo, C., Luque, A., 2009. Acceptable contamination levels in solar grade silicon: from feedstock to solar cell. Mater. Sci. Eng. B 159–160, 299–304. Hull, R., 1999. Properties of Crystalline Silicon. INSPEC, The Institution of Electrical Engineers, London, United Kingdom. Istratov, A.A., Hieslmair, H., Weber, E.R., 2000. Iron contamination in silicon technology. Appl. Phys. A: Mater. Sci. Process. 70, 489–534. Istratov, A.A., Buonassisi, T., McDonald, R.J., Smith, A.R., Schindler, R., Rand, J.A., Kalejs, J.P., Weber, E.R., 2003. Metal content of multicrystalline silicon for solar cells and its impact on minority carrier diffusion length. J. Appl. Phys. 94, 6552–6559. Istratov, A.A., Buonassisi, T., Pickett, M.D., Heuer, M., Weber, E.R., 2006. Control of metal impurities in ‘‘dirty’’ multicrystalline silicon for solar cells. Mater. Sci. Eng. B 134, 282–286. Magill, J., Pfennig, G., Galy, J., 2006. Chart of the Nuclides, seventh ed. European Commission, DG Joint Research Centre, Institute for Transuranium Elements. Scheil, E., 1942. Z. Metallkd. 34, 70. Verheijke, M.L., Jaspers, H.J.J., Hanssen, J.M.G., 1989. Neutron activation analysis of very pure silicon wafers. J. Radioanal. Chem. 131 (1), 197–214. Weber, E.R., 1983. Transition metals in silicon. Appl. Phys. A: Mater. Sci. Process. 30, 1–22. Wiehl, N., Herpers, U., Weber, E.R., 1982. J. Radioanal. Chem. 72, 69–78.
Contents lists available at ScienceDirect
Applied Radiation and Isotopes journal homepage: www.elsevier.com/locate/apradiso
Fast determination of impurities in metallurgical grade silicon for photovoltaics by instrumental neutron activation analysis J. Hampel a,n, F.M. Boldt a, H. Gerstenberg c, G. Hampel b, J.V. Kratz b, S. Reber a, N. Wiehl b a
Fraunhofer Institute for Solar Energy Systems, Heidenhofstrasse 2, D-79110 Freiburg, Germany Institute of Nuclear Chemistry, Johannes Gutenberg-University of Mainz, Fritz-Strassmann-Weg 2, D-55128 Mainz, Germany c ZWE FRM-II der Technischen Universit¨ at M¨ unchen, D-85748 Garching, Germany b
a r t i c l e i n f o
abstract
Article history: Received 22 December 2010 Received in revised form 12 April 2011 Accepted 19 May 2011 Available online 27 May 2011
Standard wafer solar cells are made of near-semiconductor quality silicon. This high quality material makes up a significant part of the total costs of a solar module. Therefore, new concepts with less expensive so called solar grade silicon directly based on physiochemically upgraded metallurgical grade silicon are investigated. Metallurgical grade silicon contains large amounts of impurities, mainly transition metals like Fe, Cr, Mn, and Co, which degrade the minority carrier lifetime and thus the solar cell efficiency. A major reduction of the transition metal content occurs during the unidirectional crystallization due to the low segregation coefficient between the solid and liquid phase. A further reduction of the impurity level has to be done by gettering procedures applied to the silicon wafers. The efficiency of such cleaning procedures of metallurgical grade silicon is studied by instrumental neutron activation analysis (INAA). Small sized silicon wafers of approximately 200 mg with and without gettering step were analyzed. To accelerate the detection of transition metals in a crystallized silicon ingot, experiments of scanning whole vertical silicon columns with a diameter of approximately 1 cm by gamma spectroscopy were carried out. It was demonstrated that impurity profiles can be obtained in a comparably short time. Relatively constant transition metal ratios were found throughout an entire silicon ingot. This led to the conclusion that the determination of several metal profiles might be possible by the detection of only one ‘‘leading element’’. As the determination of Mn in silicon can be done quite fast compared to elements like Fe, Cr, and Co, it could be used as a rough marker for the overall metal concentration level. Thus, a fast way to determine impurities in photovoltaic silicon material is demonstrated. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Instrumental neutron activation analysis Transition metals Metallurgical grade silicon HCl gas gettering
1. Introduction In standard silicon solar modules, the costs of silicon wafers ¨ make up 50% of the total costs (Gotzberger et al., 2002). One reason is the high costs for near-semiconductor quality feedstock with purities of 99.99999% (7 N) to 99.999999% (8 N), as purification steps are time and resource consuming. Low-cost silicon, such as metallurgical grade (MG) and upgraded metallurgical grade (UMG) silicon with purities of approximately 2–3 N and 3–5 N could be used for new solar cell concepts, although there are various issues one has to face regarding this material. One of the crucial problems is the high amount of metallic impurities, for example Fe or Cr. These contaminations degrade the minority carrier lifetime and therefore the solar cell efficiency by forming
n
Corresponding author. Tel.: þ49 761 4588 5247; fax: þ 49 761 4588 9250. E-mail addresses: [email protected], [email protected] (J. Hampel). 0969-8043/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.apradiso.2011.05.024
recombination centers. Istratov et al. (2006) demonstrated that the distribution of metals in multicrystalline silicon has a crucial impact on the lifetime. The cell efficiency is less affected by metals in precipitates than by interstitial impurities even at lower densities. Nevertheless, starting with MG silicon, it is still important to reduce the overall concentration of transition metal impurities. Directional solidification and gettering are two ways of silicon purification. Crystallization of silicon feedstock is an early step in the solar cell process whereby the amount of metallic impurities is reduced by several orders of magnitude. The silicon feedstock is melted in a crucible and crystallized to an ingot by unidirectional solidification, which occurs from the bottom upwards. Thereby, the metallic impurities are accumulated at the top of the ingot due to the very small segregation coefficients of metals in silicon. For example, Fe and Cr have segregation coefficients of 7 10 6 and 3 10 6, respectively (Weber, 1983). The resulting silicon ingot is cut into wafers, which contain very different amounts of metals depending on the horizontal position in the former ingot.
1366
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
In addition, gettering steps are considered to be essential in the cell process to further reduce the metal content. In this work, a new gettering technique with hydrogen chloride (HCl) gas was applied. The gettering effect is due to the removal of metals from the surface of the wafers through the formation of volatile chloride compounds (Istratov et al., 2000). These compounds are transported away by the stream of gas, and further diffusion of the metals from the bulk to the wafer surface occurs. The diffusion depends on the diffusion coefficient of the respective metals. For example, Fe has a diffusion coefficient of 4 10 6 cm2/s at 1100 1C (Weber, 1983). Both above-mentioned procedures of silicon purification need to be evaluated by analytical techniques to learn about their effectiveness. A suitable analysis method is needed for the detection of metallic impurities in solar grade silicon. The ideal analysis technique should be able to detect detrimental transition metals, such as Fe and Cr, over a wide concentration range in solar grade silicon with very low detection limits. The technique should be a fast procedure without extensive sample preparation, ideally without the need to dissolve the silicon. This reduces the risk of contamination during sample preparation. Instrumental Neutron Activation Analysis (INAA) is well known and established in many different research and application areas. A lot of experience has been made applying INAA to transition metal impurities in silicon, especially in crystalline silicon (see e.g. Wiehl et al. (1982), Verheijke et al. (1989) and Istratov et al. (2003)). However, the determination of transition metals, such as Fe, Cr, and Co in silicon, which is performed using long-lived nuclides, takes a relatively long time. In this work, INAA was investigated as a possible analytical technique to control the above-mentioned procedures of silicon purification with the goal of a high sample throughput. Due to the segregation coefficients, the ratios of various transition metals along the crystallization front should be constant. Therefore, it was tested if the concentration profile of the transition metals could be determined by measuring only the profile of Mn as the ‘‘leading element’’. Mn has a high activation cross-section of 13.3 b and a half-life of 2.58 h. Consequently, quick scans of the impurity profile along the crystallization front should be possible. Therefore, we investigated the analysis of silicon columns cut from an ingot along the crystallization front.
Garching, respectively. After irradiation and decay of the background, which is caused by the beta emitter 31Si, the samples were measured with a high purity germanium (HPGe) semiconductor detector with a relative efficiency of 27.7% at the Institute of Nuclear Chemistry at the University of Mainz. Similar INAA measurements were also carried out with wafers, which were treated by HCl gas gettering and analyzed together with untreated neighboring wafers as reference to evaluate the efficiency of this new gettering technique. The determination of Ni was achieved by the 58Ni(n,p)58Co reaction. All other elements were determined by (n,g) reactions. For all analyses, high purity liquid standards of the elements Cr, Mn, Fe, Co, and Ni were irradiated and measured together with the samples. For data evaluation, the software Genie 2000 V2.1 (Canberra) was used. The applied irradiation protocols are listed in Table 1.
2.2. Sample preparation and analysis of Si columns Vertical columns with a size of approximately 1 1 20 cm3 were cut out of an ingot, which was crystallized out of upgraded metallurgical grade silicon feedstock. These columns were irradiated as a whole in the central irradiation position at TRIGA Mainz for 6 h at a neutron flux of 4 1012 cm 2 s 1. Standards of Mn and As were placed at the top and bottom of the column for the determination of the absolute metal content. After a decay time of 15 h, the columns were measured with a germanium detector with a relative efficiency of 67% whereby a collimator of lead with 10 cm thickness and a slit width of 1 cm was placed in front of the germanium detector (Fig. 1). For the measurements the columns were divided in two pieces of 10 cm length. The influence of non-perfect shielding of the collimator, i.e. gamma rays penetrating the lead, can be estimated from the mass attenuation factors of lead for the relevant energies. The estimations show that within all other uncertainties of the measurements the collimator can be assumed to be nearly perfect. The columns were scanned in 1 cm steps. The 31Si activity was used to control and correct deviations of the neutron flux along the column. The overall uncertainties were calculated from all contributing errors, which are listed in the following in the order
2. Experiments 2.1. Sample preparation and analysis of Si wafers After unidirectional crystallization of MG-Si feedstock with 99.3% purity, the ingot was cut into wafers. From three different parts of the ingot (top, middle, and bottom), wafers were cut into samples with a size of approximately 2.5 1.25 0.027 cm3 and a weight of approximately 200 mg. Afterwards, the samples were surface cleaned by chemical polishing (etching of the silicon surface by a mixture of hydrofluoric acid, acetic acid, and nitric acid) and irradiated at the research reactor TRIGA Mark II in Mainz (Hampel et al., 2006) and at the research reactor FRM-II in
Fig. 1. Schematic (top view) of collimator for scanning Si columns (not to scale).
Table 1 Irradiation protocols used for analyses of the wafers after crystallization. Program no.
Irradiation facility
Irradiation time (h)
Decay time
Measuring time (h)
Neutron flux (cm 2 s 1)
Investigated elements
1 2 3
TRIGA Mainz TRIGA Mainz FRM-II Garching
33 3 90
approx. 8 d approx. 3 h approx. 21 d
15 0.5 15
4.2 1012 4.2 1012 1.1 1014
Ni Mn Cr, Fe, Co
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
1367
of an increasing magnitude: the imperfection of the collimator, the weighting error of the standard and sample, the geometric error of the sample compared to the standard, and the error of the counting statistics.
3. Results 3.1. Detection limits The detection limits (in ng/g) of the INAA results are given in Table 2. Very low values, especially for Co, Cr, and Mn could be achieved. This is necessary because the acceptable value, e.g., of Cr in solar grade silicon wafers before solar cell processing is around 1 ng/g (Hofstetter et al., 2009). In the silicon feedstock, a higher impurity is allowed because of the purification capacity of the subsequent crystallization step. This higher impurity level makes the analysis much easier. The detection limit for Fe is higher because 58Fe, the nuclide which is utilized for the determination, has a low nuclide abundance (0.28%) and a relatively low neutron capture cross-section (1.3 b) (Magill et al., 2006). Nevertheless, the resulting detection limit is still acceptable since Fe has a higher diffusivity than Cr and can therefore be gettered more easily. The determination of Ni was achieved by the 58 Ni(n,p)58Co reaction, which occurs with fast neutrons. The detection was only possible at the TRIGA Mainz because the difference between the fluxes of thermal and fast neutrons is much smaller than that at the FRM-II Garching. The half-life of 56 Mn (2.6 h) is in the same range as the half-life of the activated silicon nuclides, but is produced from 55Mn, which has a nuclide abundance of 100% and a neutron capture cross-section of 13.3 b. These properties together with the low half-life allow a fast and reliable detection of Mn by INAA.
Fig. 2. Metal mass fractions (in mg/g) measured by INAA in metallurgical grade silicon wafers depending on the position in the ingot.
Table 3 Fe-metal concentration ratio in different ingot positions. Position in ingot
Fe–Mn concentration ratio
Fe–Cr concentration ratio
Fe–Co concentration ratio
Fe–Ni concentration ratio
Bottom Middle Top
28 7 2 23 7 2 21 7 2
117 1 2107 20 2007 20
7707 70 1360 7 110 1340 7 110
– 550 7 50 340 7 30
3.2. Si wafers The INAA results of the crystallized silicon ingot are presented in Fig. 2, showing transition metal impurity mass fractions as a function of the position in the ingot. The increase in the impurities from the bottom to the top is expected as the ingot was obtained by unidirectional solidification from the bottom upwards. The segregation coefficients of transition metals are much lower than 1, so the impurities are accumulated at the top of the ingot. While the metal concentration strongly depends on the position of the wafer in the ingot, the concentration ratios between different transition metals were found to be nearly the same throughout the ingot. This is particularly the case for the ratio of Fe and Mn concentration, which is quite similar in different positions of the ingot (Table 3), although absolute Fe concentrations increase by three orders of magnitude from bottom to top. The ratios of other metal concentrations show some deviation at the bottom of the ingot. Nevertheless, once the metal concentration ratios in an ingot and one metal profile are measured, rough profiles of some other metals could be calculated. However, this Table 2 Detection limits (in ng/g) of transition metals in approx. 200 mg silicon measured by INAA in Mainz and Garching. Element
Detection limit (ng/g)
Irradiation facility
Cr Mn Fe Co Ni
0.4 0.9 23 0.1 82
FRM-II Garching TRIGA Mainz FRM-II Garching FRM-II Garching TRIGA Mainz
Fig. 3. Mn mass fraction (in mg/g) measured by INAA in metallurgical grade silicon neighboring wafers with and without 30 min of HCl gas gettering.
assumption has to be investigated also for other feedstock materials and crystallization processes. Furthermore, this can only be assumed if no gettering was done before, because it was shown that transition metals are reduced by HCl gas gettering by a different magnitude (Hampel et al., 2009). At least, Mn can serve as a marker for the degree of purification effectiveness by unidirectional solidification. The INAA results for Mn in the wafers to which HCl gas gettering was applied are presented in Fig. 3. Gettering with HCl gas for 30 min reduces the Mn concentration in silicon drastically. A reduction of 96% can be achieved. This corresponds to a gettering efficiency of 27, which is defined as concentration of a specific element before gettering, divided by the respective concentration after gettering. Most other transition metals can be reduced by HCl gas gettering in the same way (Hampel et al., 2010).
1368
J. Hampel et al. / Applied Radiation and Isotopes 69 (2011) 1365–1368
could be detected by INAA, from less than ng/g to mg/g. The sample preparation consists of only one surface cleaning step and the risk of contamination is much lower compared to other analyses because the silicon does not have to be dissolved before analysis. To accelerate the detection of transition metals, first experiments of locally scanning whole vertical silicon columns were accomplished, demonstrating that it is a feasible method. The fact that similar transition metal ratios were found throughout the entire ingot and the fact that Mn can be detected by INAA in a fast and reliable way, leads to the conclusion that Mn could serve as a model for other transition metals. This means that rough profiles of other transition metals can be calculated using the Mn profile. This assumption has still to be proven for various silicon feedstock materials and crystallization processes. Nevertheless, Mn can already be used as a marker for the degree of purification effectiveness by unidirectional solidification.
Acknowledgments Fig. 4. As profile of a silicon column, which was irradiated in one piece and scanned in approximately 1 cm steps by gamma spectroscopy.
3.3. Si columns In the first tests of scanning whole vertical silicon columns, a detection limit for Mn of 5 ng/g could be achieved. This is at least 1 order of magnitude below the critical impurity concentration. With the goal to use the Mn concentration as measure for all transition elements, this result shows that the used irradiation and measurement conditions are well-suited for these studies. The reliability of the measurements could be checked by the profile of the As concentration. The As concentration profile should follow Eq. (1), which is called Scheil equation (Scheil, 1942). CS ðxÞ ¼ kC0 ð1xÞk1
ð1Þ
where CS is the concentration in the solid, x is the relative ingot height, k is the segregation coefficient, and C0 is the starting concentration. With k¼0.3 for As in Si (Hull, 1999) and an As starting concentration of C0 ¼0.108 mg/g, the data are in good agreement with the theory (Fig. 4). If Mn could be used as a model for other transition metal profiles, the analysis of a Si ingot would be much faster. The determination of Mn by the presented Si column scanning method could be done in less than 24 h. For comparison, the determination of Fe or Co in a Si column takes 2–3 weeks.
4. Conclusion INAA is well-suited for the detection of transition metals in solar grade silicon. It was shown that the metals of interest, such as Cr, Mn, and Fe, can be determined with sufficient detection limits. A large range of transition metal concentrations in silicon
The authors would like to thank PV Crystalox for providing the Si wafers from MG-Si and the operators of the FRM-II Garching and the TRIGA Mainz for performing the irradiations. Furthermore, the help from the crystallization and wafering team at the Fraunhofer ISE is gratefully acknowledged. References ¨ Gotzberger, A., Luther, J., Willeke, G., 2002. Solar cells: past, present, future. Sol. Energy Mater. Sol. Cells 74, 1–11. Hampel, G., Eberhardt, K., Trautmann, N., 2006. Der TRIGA forschungsreaktor mainz. Atw Int. Z. Kernenergie 51 (5), 328–330. Hampel, J., Schmich, E., Boldt, F.M., Wiehl, N., Hampel, G., Kratz, J.V., Reber, S., 2009. Gettering of metallurgical grade silicon by HCl gas. In: Proceedings of the 24th European Photovoltaic Solar Energy Conference, Hamburg, Germany, pp. 2557–2559. Hampel, J., Boldt, F.M., Wiehl, N., Hampel, G., Kratz, J.V., Reber, S., 2010. Use of HCl gas gettering in the epitaxial wafer equivalent concept. In: Proceedings of the 25th European Photovoltaic Solar Energy Conference, Valencia, Spain, pp. 3591–3593. Hofstetter, J., Lelievre, J.F., del Canizo, C., Luque, A., 2009. Acceptable contamination levels in solar grade silicon: from feedstock to solar cell. Mater. Sci. Eng. B 159–160, 299–304. Hull, R., 1999. Properties of Crystalline Silicon. INSPEC, The Institution of Electrical Engineers, London, United Kingdom. Istratov, A.A., Hieslmair, H., Weber, E.R., 2000. Iron contamination in silicon technology. Appl. Phys. A: Mater. Sci. Process. 70, 489–534. Istratov, A.A., Buonassisi, T., McDonald, R.J., Smith, A.R., Schindler, R., Rand, J.A., Kalejs, J.P., Weber, E.R., 2003. Metal content of multicrystalline silicon for solar cells and its impact on minority carrier diffusion length. J. Appl. Phys. 94, 6552–6559. Istratov, A.A., Buonassisi, T., Pickett, M.D., Heuer, M., Weber, E.R., 2006. Control of metal impurities in ‘‘dirty’’ multicrystalline silicon for solar cells. Mater. Sci. Eng. B 134, 282–286. Magill, J., Pfennig, G., Galy, J., 2006. Chart of the Nuclides, seventh ed. European Commission, DG Joint Research Centre, Institute for Transuranium Elements. Scheil, E., 1942. Z. Metallkd. 34, 70. Verheijke, M.L., Jaspers, H.J.J., Hanssen, J.M.G., 1989. Neutron activation analysis of very pure silicon wafers. J. Radioanal. Chem. 131 (1), 197–214. Weber, E.R., 1983. Transition metals in silicon. Appl. Phys. A: Mater. Sci. Process. 30, 1–22. Wiehl, N., Herpers, U., Weber, E.R., 1982. J. Radioanal. Chem. 72, 69–78.