- Email: [email protected]
Effect of extended phosphorus diffusion gettering on chromium impurity in HEM multicrystalline silicon
Materials Science in Semiconductor Processing 15 (2012) 56–60
Contents lists available at SciVerse ScienceDirect
Materials Science in Semiconductor Processing journal homepage: www.elsevier.com/locate/mssp
Effect of extended phosphorus diffusion gettering on chromium impurity in HEM multicrystalline silicon Nabil Khelifati a,b,n, Djoudi Bouhafs a, Messaoud Boumaour a, Seddik-El-Hak Abaidia b, Baya Palahouane a a b
Silicon Technology Development Unit, Bd. 2 Frantz Fanon, les sept merveilles B.P.140, Algiers, Algeria University of M’hamed Bougara, Faculte´ des Sciences de l’Inge´nieur, LMMC, Avenue de l’Inde´pendance, 35000 Boumerde s, Algeria
a r t i c l e in f o
abstract
Available online 9 September 2011
We have investigated the extended phosphorus diffusion gettering (PDG) effect on chromium impurities (Cr) in p-type multicrystalline silicon (mc-Si) grown by Heat Exchanger Method (HEM). The study was made after phosphorous diffusion and according to different extended annealing temperatures. The secondary ion mass spectrometry (SIMS) analysis revealed a significant accumulation of 52Cr in heavily phosphorus doped (HPD) region. Using quasi-steady state photoconductance (QSSPC) technique, the apparent lifetime dependent minority carrier density curves have been obtained. The results showed an increment of the bulk minority carrier lifetime for specific annealing temperatures. Appropriate calculations based on QSSPC results allowed us to determine the lifetime curves associated to gettered impurities. Their fitting by Shockley-Read-Hall (SRH) model reveal that the origin of the lifetime increment is the reduction of interstitial chromium (Cri) density in the bulk. Furthermore, the estimation of electron to hole capture cross-section ratio (k¼ sn/sp) through the modelling of apparent lifetime curves using Hornbeck–Haynes model, confirmed the effectiveness of Cri gettering and identified the nature of dominant recombination centres after gettering process. & 2011 Elsevier Ltd. All rights reserved.
Keywords: HEM mc-Si Extended PDG Chromium impurity Minority carrier lifetime Modelling
1. Introduction Today, mc-Si wafers are widely used as precursor elements in solar cells manufacturing, and constitute more than half of the overall industrial market. This is due to their low cost of manufacturing as compared to that of monocrystalline silicon (c-Si) [1,2]. However, the contamination of mc-Si material by metallic impurities (Fe, Cr, Mn, Cu, Ni, Co, etc.) during its elaboration is considered one of its major disadvantages, and the presence of such contamination can provoke a high n Corresponding author at: Silicon Technology Development Unit, Bd. 2 Frantz Fanon, les sept merveilles, B.P.140, Algiers, Algeria. Tel.: þ 213 21 43 26 30; fax: þ 213 21 43 26 30. E-mail address: [email protected] (N. Khelifati).
1369-8001/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.mssp.2011.08.005
carrier recombination activity and then may be greatly limiting the efficiency potential of solar cells. Chromium is known as one of those contaminant impurities, which can strongly reduce the recombination lifetime in silicon [3]. Besides, relatively high concentrations of this impurity have been measured in mc-Si. Employing neutron activation analysis (NAA), Istratov et al. [4] determined the total chromium content in mc-Si grown by sheet technology to be up to 1.8 1015 cm 3. Similar NAA analysis has been performed by Macdonald et al. [5,6], where the chromium concentrations have been measured in different blockcast multicrystalline silicon wafers to be about 2 1013 cm 3. Despite its relatively high concentration and its harmful effect on carrier lifetime, the chromium state in
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
2. Experimental details The wafers investigated in this study were 1.5 O cm, p-type mc-Si grown by the Heat Exchanger Method (HEM) [8]. These wafers were selected from 2 cm thick brick taken from centre of the ingot. All experiments were performed on companion wafers cut adjacently, in which the features of the extended crystallographic defects were same. Before any experimental process, the as-cut wafers were slightly etched in a bath of NaOH:H2O (30%) to remove the saw damage. After this step, the wafers have undergone phosphorus diffusion on both the sides, at 900 1C for 20 min, in a tube furnace using a phosphorus oxychloride (POCL3) liquid source. Subsequently, wafers were subjected to two successive annealing under a nitrogen flow, the first was made at high temperature TH ¼930 1C for 30 min (standard gettering) and the second was performed at moderate temperatures TL varied between 600 1C and 800 1C for a fixed time of 120 min (extended gettering). This concept is known as extended gettering or variable temperature gettering [9–11]. Some wafers were processed without any phosphorus diffusion or annealing, they are considered references. Phosphorus diffusion and annealing steps were followed by a short dip into NaOH:H2O solution to remove the phosphorussilicate-glass (PSG) layer. The characterisation techniques used in this study are secondary ions mass spectrometry and quasi-steady state photoconductance. SIMS analyses were made in order to obtain the depth profiles of phosphorus (31P) and chromium (52Cr). The instrument used was Cameca IMS-4F. The quantisation of 31P and 52Cr profiles was achieved by analysing standard samples having well-known concentrations. QSSPC measurements were performed by WCT-120 tester. Before any lifetime measurement, a stripping of 10 mm from each side of wafers followed by a PECVD aSiNx:H layer deposition were made to remove the phosphorus diffusion region and to passivate the surface. These measurements allow the determination of the injection level dependent minority carrier lifetime as well as to evaluate the gettered Cri concentration after annealing steps. Lifetime values were taken at 1 1015 cm 3 excess carrier density to exclude the influence of trapping artefacts.
3. Results and discussion Fig. 1 shows the depth profiles of 52Cr and an example of the diffused 31P profile that is practically same for all the studied samples. Firstly, we note a significant accumulation of chromium impurity in HPD region ( o1 mm). According to annealing temperature TL, these profiles show the increase of chromium concentration by about one order of magnitude between reference and the sample that annealed at TL ¼650 1C, where its concentration reach the maximal value around 8 1015 cm 3. The increment of TL beyond 750 1C plays an inverse effect on the accumulation phenomenon. Indeed, the chromium concentration decreases continually to reach for TL ¼800 1C, a minimal value, which is comparable to that of reference sample. Similar result of 52Cr accumulation in phosphorus diffusion layer after gettering process has been observed by Bentzen et al. [12], and it has been explained by an effective PDG, through the increase of chromium solubility in the phosphorus doped region. The results obtained by QSSPC measurements and presented in Figs. 2 and 3, confirm SIMS results and give more details about the impact of extended gettering effect. Indeed, the improvement of bulk effective lifetime (see Fig. 3) observed for TL r750 1C, can be easily correlated with the increment of chromium accumulation in HPD region. This indicates that the carrier lifetime increment is due to the reduction of bulk recombination centres density created by chromium, and this is through its diffusion from the bulk to the HPD region. The presence of maximum value of carrier lifetime (tapp ¼ 47.7 ms) associated to the wafer gettered at TL ¼650 1C, can be explained by the fact that the chromium solubility in the bulk increases with temperature faster than in the phosphorus doped region, so reducing the segregation coefficient for higher temperatures, and on the other hand the diffusivity decreases for low
1020 1019 Concentration (cm-3)
multicrystalline silicon is still a subject of research and discussion [7]. Nevertheless, chromium as most metallic impurities dissolved in material bulk or segregated at grain boundaries, may be removed by ’’gettering’’ process. It refers to a thermal process step that activates the diffusion of interstitial impurities from the active regions of the device to less important regions created generally by phosphorus diffusion, aluminum–silicon alloying, amorphous silicon deposition, etc. In this present paper, we investigate the effect of extended PDG on the chromium impurity in p-type HEM mc-Si wafers. The process was carried out through two annealing plateaus. Wafers characterisation was mainly made by secondary ion mass spectrometry and quasisteady state photoconductance techniques.
57
31
Reference TL = 600°C TL = 650°C TL = 700°C TL = 750°C TL = 800°C
P
1018 1017 52
Cr
1016 1015
Detection limit of Cr 1014 0
1
2 Depth (μm)
Fig. 1. Chromium and phosphorus depth profiles obtained by SIMS analysis for the wafers treated at different annealing temperatures TL. The broken line represents the detection limit of chromium for depth profiling in silicon (2 1014 cm 3) associated to Cameca IMS-4F instrument.
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
1
timpurity
1014 1015 1016 Apparent minority carrier density Δnapp (cm-3) Fig. 2. Apparent minority carrier lifetime versus minority carrier density as measured by QSSPC for investigated wafers at different annealing temperatures TL, and fits using the Hornbeck–Haynes model (solid lines).
¼
1
tnongettered
Se g re
ffe ct fus ivi ty e
40 30
ga ti
on
eff ec t
Dif
20
tp0 ¼
10 0 Ref.
600 650 700 750 Annealing temperature TL (°C)
800
Fig. 3. Effective lifetimes measured at an excess carrier density of 1 1015 cm 3 for gettered wafers at different annealing temperatures TL.
temperatures preventing the impurities gettering. These two opposite effects cause the observed optimum temperature [13,14]. Previous experimental work performed by Kang and Schroder [15] confirmed clearly this phenomenon. The improvement of carrier lifetime obtained for TL ¼800 1C without any chromium accumulation is not clear. It may be due to the change of chemical state of chromium (neutralization) in the bulk or/and to an external phosphorus diffusion gettering of other metallic impurities such as Mn, Fe, etc. The bulk lifetime tnon-gettered associated to reference wafer that has not been gettered, can be written in terms of two components [16]: 1
tnongettered
¼
1
timpurity
þ
1
tgettered
tn0 ðNA þp1 þ DnÞ þ tp0 ðn1 þ DnÞ
1 N nth sp
where timpurity is the lifetime due to the getterable impurities, and tgettered is the lifetime after gettering, or alternatively, the lifetime caused by the non-getterable
ð2Þ
NA þ Dn
tn0 ¼
1 Nnth sn
ð3Þ
Table 1 lists the values of capture cross sections (sn,p) and localised energy levels corresponding to Cri, CrB, Fei and FeB, used to generate the SRH fit curves. These values have been taken from Mishra [21] and Istratov et al. [22]. Clearly, only the recombination centre capable of adequately explaining the calculated injection-level dependence for all gettered wafers is the Cri impurity. Consequently, this result implies that the lifetime 10-5 9x10-6 8x10-6
TL (°C)Density of Cri (x1011cm-3) 1.12 600 1.28 650 1.22 700 1.17 750 1.21 800
7x10-6
Cri
6x10-6 CrB 5x10-6
ð1 aÞ
ð1 bÞ
where NA is the acceptor density. n1 and p1 are, respectively, the electron and hole densities when the Fermi level coincides with the recombination centre energy. tn0 and tp0 are, respectively, the fundamental electron and hole lifetimes, and are related to the carrier thermal velocity nth, the recombination centre density N and the capture cross sections sn and sp of the specific centre in question:
Effective Lifetime (s)
Effective minority carrier lifetime τeff (µs)
tSRH ðDnÞ ¼
50
1
tgettered
A similar calculation was also used by LI et al. [17] and Schmidt and Cuevas [18] to investigate the recombination activity of Cu in mc-Si samples contaminated at different temperatures and to study boron–oxygen complexes in Cz–Si, respectively. Fig. 4 shows the results of calculation and those of attempting to SRH fit for interstitial Fe, Cr, and FeB and CrB pairs. The SRH fit curves have been generated employing the following equation [19,20]:
60 Measured at Δnapp = 1 . 1015 cm-3
or
1x10-5
pt
Reference TL = 600°C TL = 650°C TL = 700°C TL = 750°C TL = 800°C
centres. Of course, an adequate surface passivation is necessary to avoid surface effects, and this was achieved by the deposition of a-SiNx:H film, as indicated above. By measuring the injection-level dependent lifetimes before and after gettering, it is possible then to determine the injection-level dependence of the gettered impurities alone via
ce
1.5 Ωcm p-type mc-Si
ac
1x10-4
Fe B
Apparent minority carrier lifetime τapp (s)
58
Fei
1015 Excess carrier density
1016 (cm-3)
Fig. 4. Effective carrier lifetime versus excess carrier density due to the removed impurities, modelled with SRH curves for interstitial Fe and Cr, and FeB and CrB pairs (solid lines).
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
59
Table 1 Energy levels and capture cross sections used for calculation.
sn (cm2)
Energy level (eV)
13
Ec 0.22 Ev þ 0.27 Ev þ 0.38 Ec 0.23
Cri CrB Fei FeB acceptor
2.3 10 1.4 10 13 5 10 14 3 10 14
1000
σn/σp Ratio
Fei
100
FeB CrB
10
Cri Ref.
600
650
700
750
800
Annealing temperature TL (°C) Fig. 5. Variation of the electron to hole capture cross-sections ratio (k¼ sn/sp) as a function of annealing temperature TL.
increment observed for temperature TL varying between 600 and 800 1C could be explained by the gettering of Cri. An interesting result that is relevant to the type of gettering could be deduced from this calculation. Indeed, the lifetime improvement observed for TL ¼800 1C (caused by the reduction of Cri density in the bulk) without any Cr accumulation in the phosphorus diffusion region, lead us to conclude that there is certainly a local deactivation of Cri recombination activity centres (internal gettering). In spite of the narrow variation range of Cri concentration, the tendency of this variation explains well the carrier lifetime change reported above, and confirms the evolution of Cr profiles obtained by SIMS, too. However, the values of Cri concentration estimated by the modelling of lifetime curves are between four and five orders of magnitude below those obtained directly by the SIMS analysis. These results indicate that the vast majority of the gettered chromium is required to present in some much more electronically benign form, such as in precipitates. Using Hornbeck–Haynes model [23], a complete modelling of injection-level dependent lifetime curves (see Fig. 2) allows to estimate the ratio k, and so to conclude the origin of dominant recombination activities after gettering process. Fig. 5 illustrates the variation of k ratio as a function of the extended gettering temperature TL. The basic formula used for modelling measured QSSPC data can be found in [24]. More details about the calculation formula and the fitting procedure are reported elsewhere in [25–27]. For TL r750 1C, it is obviously shown that the shift of k ratio from the value associated to Cri (k ¼2.09), to values associated to FeB, CrB pairs and Fei is accompanied by the
sp (cm2) 13
1.1 10 1 10 14 7 10 17 2 10 15
k¼ sn/sp
Reference
2.1 14 714.3 15
Mishra [21] Mishra [21] Istratov et al. [22] Istratov et al. [22]
decrease of Cri content in the bulk (see Fig. 4), and an increment of the chromium accumulation in HPD region (see Fig. 1). So, this result confirms that the lifetime improvement in this annealing temperature range is the consequence of the chromium extraction from the bulk, and indicates clearly that the origin of dominant recombination activity after gettering process is mainly due to the presence of CrB, FeB pairs and Fei centres. For the wafer gettered at TL ¼800 1C, k ratio shows an important shift towards the value associated to Fei (k¼714.28), so indicating the quasi-dominance of Fei recombination activity.
4. Conclusion In summary, we have studied the effect of extended PDG on chromium impurity in p-type HEM mc-Si wafers. The correlation between results obtained by the SIMS analysis and QSSPC measurement indicated an effective extended gettering of chromium impurity. This effect was especially observed around 650 1C and 800 1C, where the effective minority carrier lifetime was, respectively, improved by more than 900% and 800%. The SRH fit of calculated effective lifetime curves associated to gettered impurities proves that the origin of lifetime improvement is the reduction of Cri recombination centres density in the bulk. Because of the Cr accumulation observed in HPD region for annealing temperature TL r750 1C, the lifetime improvement obtained was identified as the consequence of an effective PDG of Cri. While the improvement observed at 800 1C (that is due to the decrease of Cri density, too) without any Cr accumulation is probably due to a local neutralization of Cri recombination centres (internal gettering). The determination of electron to hole capture crosssections ratio using Hornbeck–Haynes model fitting of QSSPC curves leads us to conclude that after each effective gettering of Cri, the impact of Fei recombination activity becomes more dominant. Future work will be focused on the extended gettering effectiveness of iron impurity with new parameters process.
Acknowledgements This work was financed by The General Directorate for Scientific Research and Technological Development (Algeria). The authors wish to thank Prof. M. Lemiti at the INSA-Lyon University (France) for assistance with part of the QSSPC measurements.
60
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
References [1] A. Cuevas, in: Proceedings of the Materials 2003 Conference, The Institute of Materials Engineering Australasia, Sydney, Australia, 2003. [2] V. Avrutin, N. Izyumskaya, H. Morkoc- , Superlattices and Microstructures 49 (2011) 337–364. [3] J.R. Davis, A. Rohatgi, R.H. Hopkins, P.D. Blais, P. Rai-Choudhury, J.R. Mccormick, H.C. Mollenkopf, IEEE Transactions on Electron Devices ED-27 (1980) 677. [4] A.A. Istratov, T. Buonassisi, R.J. McDonald, A.R. Smith, R. Schindler, J.A. Schindler, J.A. Rand, J.P. Kalejs, E.R. Weber, Journal of Applied Physics 94 (2003) 6552. [5] D. Macdonald, A. Cuevas, A. Kinomura, Y. Nakano, In: Proceedings of the 29th EEE photovoltaic specialists conference, New Orleans, May 2002, p. 285–288. [6] D. Macdonald, A. Cuevas, A. Kinomura, Y. Nakano, L.J. Geerligs, Journal of Applied Physics 97 (2005) 033523. [7] T. Buonassisi, A.A. Istratov, M.D. Pickett, M. Heuer, J.P. Kalejs, G. Hahn, M.A. Marcus, B. Lai, Z. Cai, S.M. Heald, T.F. Ciszek, R.F. Clark, D.W. Cunningham, A.M. Gabor, R. Jonczyk, S. Narayanan, E. Sauar, E.R. Weber, Progress in Photovoltaics: Research and Applications 14 (2006) 513. [8] D. Ouadjaout, Y. Gritli, L. Zair, M. Boumaour, Revue des Energies Renouvelables 8 (2005) 49–54. [9] P.S. Plekhanov, R. Gafiteanu, U.M. Gosele, T.Y. Tan, Journal of Applied Physics 86 (5) (1999) 1 September. [10] P. Manshanden, L.J. Geerligs, Solar Energy Materials and Solar Cells 90 (2006) 998–1012.
[11] C. Zhang, S. Liu, Y. Wang, Y. Chen, Materials Science in Semiconductor Processing 13 (2010) 209–213. [12] A. Bentzen, A. Holt, R. Kopecek, G. Stokkan, J.S. Christensen, B.G. Svensson, Journal of Applied Physics 99 (2006) 093509. [13] L. Baldi, G.F. Cerofolini, G. Ferla, G. Frigerio, Physica Status Solidi A 48 (1978) 523–532. [14] C. Del Canizo, A. Luque, Journal of The Electrochemical Society 147 (7) (2000) 2685–2692. [15] J.S. Kang, D.K. Schroder, Journal of Applied Physics 65 (1989) 2974–2985. [16] D. Macdonald, in Recombination and Trapping in Multicrystalline Silicon Sola Cells, Ph. D. Thesis, The Australian National University, 2001. [17] L.I. Xiao-qiang, Y.A.N.G. De-ren, Y.U. Xue-gong, Q.U.E. Duan-lin, Transactions of the Nonferrous Metals Society of China 21 (2011) 691–696. [18] J. Schmidt, A. Cuevas, Journal of Applied Physics 86 (1999) 3175–3180. [19] W. Shockley, W.T. Read, Physical Review 87 (5) (1952) 835–842. [20] R.N. Hall, Physical Review 87 (2) (1952) 387–387. [21] K. Mishra, Applied Physics Letters 68 (1996) 3281–3283. [22] A.A. Istratov, H. Hieslmair, E.R. Weber, Applied Physics A: Materials Science and Processing 69 (1) (1999) 13–44. [23] J.A. Hornbeck, J.R. Haynes, Physical Review 97 (1955) 311–321. [24] A. Bentzen, in Phosphorus Diffusion and Gettering in Silicon Solar Cells, Ph.D. Thesis, University of Oslo, 2006. [25] D. Macdonald, A. Cuevas, Applied Physics Letters 74 (1999) 1710. [26] D. Macdonald, A. Cuevas, Applied Physics Letters 75 (1999) 1571. [27] K.R. McIntosh, B.B. Paudyal, D. Macdonald, Journal of Applied Physics 104 (2008) 084503.
Contents lists available at SciVerse ScienceDirect
Materials Science in Semiconductor Processing journal homepage: www.elsevier.com/locate/mssp
Effect of extended phosphorus diffusion gettering on chromium impurity in HEM multicrystalline silicon Nabil Khelifati a,b,n, Djoudi Bouhafs a, Messaoud Boumaour a, Seddik-El-Hak Abaidia b, Baya Palahouane a a b
Silicon Technology Development Unit, Bd. 2 Frantz Fanon, les sept merveilles B.P.140, Algiers, Algeria University of M’hamed Bougara, Faculte´ des Sciences de l’Inge´nieur, LMMC, Avenue de l’Inde´pendance, 35000 Boumerde s, Algeria
a r t i c l e in f o
abstract
Available online 9 September 2011
We have investigated the extended phosphorus diffusion gettering (PDG) effect on chromium impurities (Cr) in p-type multicrystalline silicon (mc-Si) grown by Heat Exchanger Method (HEM). The study was made after phosphorous diffusion and according to different extended annealing temperatures. The secondary ion mass spectrometry (SIMS) analysis revealed a significant accumulation of 52Cr in heavily phosphorus doped (HPD) region. Using quasi-steady state photoconductance (QSSPC) technique, the apparent lifetime dependent minority carrier density curves have been obtained. The results showed an increment of the bulk minority carrier lifetime for specific annealing temperatures. Appropriate calculations based on QSSPC results allowed us to determine the lifetime curves associated to gettered impurities. Their fitting by Shockley-Read-Hall (SRH) model reveal that the origin of the lifetime increment is the reduction of interstitial chromium (Cri) density in the bulk. Furthermore, the estimation of electron to hole capture cross-section ratio (k¼ sn/sp) through the modelling of apparent lifetime curves using Hornbeck–Haynes model, confirmed the effectiveness of Cri gettering and identified the nature of dominant recombination centres after gettering process. & 2011 Elsevier Ltd. All rights reserved.
Keywords: HEM mc-Si Extended PDG Chromium impurity Minority carrier lifetime Modelling
1. Introduction Today, mc-Si wafers are widely used as precursor elements in solar cells manufacturing, and constitute more than half of the overall industrial market. This is due to their low cost of manufacturing as compared to that of monocrystalline silicon (c-Si) [1,2]. However, the contamination of mc-Si material by metallic impurities (Fe, Cr, Mn, Cu, Ni, Co, etc.) during its elaboration is considered one of its major disadvantages, and the presence of such contamination can provoke a high n Corresponding author at: Silicon Technology Development Unit, Bd. 2 Frantz Fanon, les sept merveilles, B.P.140, Algiers, Algeria. Tel.: þ 213 21 43 26 30; fax: þ 213 21 43 26 30. E-mail address: [email protected] (N. Khelifati).
1369-8001/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.mssp.2011.08.005
carrier recombination activity and then may be greatly limiting the efficiency potential of solar cells. Chromium is known as one of those contaminant impurities, which can strongly reduce the recombination lifetime in silicon [3]. Besides, relatively high concentrations of this impurity have been measured in mc-Si. Employing neutron activation analysis (NAA), Istratov et al. [4] determined the total chromium content in mc-Si grown by sheet technology to be up to 1.8 1015 cm 3. Similar NAA analysis has been performed by Macdonald et al. [5,6], where the chromium concentrations have been measured in different blockcast multicrystalline silicon wafers to be about 2 1013 cm 3. Despite its relatively high concentration and its harmful effect on carrier lifetime, the chromium state in
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
2. Experimental details The wafers investigated in this study were 1.5 O cm, p-type mc-Si grown by the Heat Exchanger Method (HEM) [8]. These wafers were selected from 2 cm thick brick taken from centre of the ingot. All experiments were performed on companion wafers cut adjacently, in which the features of the extended crystallographic defects were same. Before any experimental process, the as-cut wafers were slightly etched in a bath of NaOH:H2O (30%) to remove the saw damage. After this step, the wafers have undergone phosphorus diffusion on both the sides, at 900 1C for 20 min, in a tube furnace using a phosphorus oxychloride (POCL3) liquid source. Subsequently, wafers were subjected to two successive annealing under a nitrogen flow, the first was made at high temperature TH ¼930 1C for 30 min (standard gettering) and the second was performed at moderate temperatures TL varied between 600 1C and 800 1C for a fixed time of 120 min (extended gettering). This concept is known as extended gettering or variable temperature gettering [9–11]. Some wafers were processed without any phosphorus diffusion or annealing, they are considered references. Phosphorus diffusion and annealing steps were followed by a short dip into NaOH:H2O solution to remove the phosphorussilicate-glass (PSG) layer. The characterisation techniques used in this study are secondary ions mass spectrometry and quasi-steady state photoconductance. SIMS analyses were made in order to obtain the depth profiles of phosphorus (31P) and chromium (52Cr). The instrument used was Cameca IMS-4F. The quantisation of 31P and 52Cr profiles was achieved by analysing standard samples having well-known concentrations. QSSPC measurements were performed by WCT-120 tester. Before any lifetime measurement, a stripping of 10 mm from each side of wafers followed by a PECVD aSiNx:H layer deposition were made to remove the phosphorus diffusion region and to passivate the surface. These measurements allow the determination of the injection level dependent minority carrier lifetime as well as to evaluate the gettered Cri concentration after annealing steps. Lifetime values were taken at 1 1015 cm 3 excess carrier density to exclude the influence of trapping artefacts.
3. Results and discussion Fig. 1 shows the depth profiles of 52Cr and an example of the diffused 31P profile that is practically same for all the studied samples. Firstly, we note a significant accumulation of chromium impurity in HPD region ( o1 mm). According to annealing temperature TL, these profiles show the increase of chromium concentration by about one order of magnitude between reference and the sample that annealed at TL ¼650 1C, where its concentration reach the maximal value around 8 1015 cm 3. The increment of TL beyond 750 1C plays an inverse effect on the accumulation phenomenon. Indeed, the chromium concentration decreases continually to reach for TL ¼800 1C, a minimal value, which is comparable to that of reference sample. Similar result of 52Cr accumulation in phosphorus diffusion layer after gettering process has been observed by Bentzen et al. [12], and it has been explained by an effective PDG, through the increase of chromium solubility in the phosphorus doped region. The results obtained by QSSPC measurements and presented in Figs. 2 and 3, confirm SIMS results and give more details about the impact of extended gettering effect. Indeed, the improvement of bulk effective lifetime (see Fig. 3) observed for TL r750 1C, can be easily correlated with the increment of chromium accumulation in HPD region. This indicates that the carrier lifetime increment is due to the reduction of bulk recombination centres density created by chromium, and this is through its diffusion from the bulk to the HPD region. The presence of maximum value of carrier lifetime (tapp ¼ 47.7 ms) associated to the wafer gettered at TL ¼650 1C, can be explained by the fact that the chromium solubility in the bulk increases with temperature faster than in the phosphorus doped region, so reducing the segregation coefficient for higher temperatures, and on the other hand the diffusivity decreases for low
1020 1019 Concentration (cm-3)
multicrystalline silicon is still a subject of research and discussion [7]. Nevertheless, chromium as most metallic impurities dissolved in material bulk or segregated at grain boundaries, may be removed by ’’gettering’’ process. It refers to a thermal process step that activates the diffusion of interstitial impurities from the active regions of the device to less important regions created generally by phosphorus diffusion, aluminum–silicon alloying, amorphous silicon deposition, etc. In this present paper, we investigate the effect of extended PDG on the chromium impurity in p-type HEM mc-Si wafers. The process was carried out through two annealing plateaus. Wafers characterisation was mainly made by secondary ion mass spectrometry and quasisteady state photoconductance techniques.
57
31
Reference TL = 600°C TL = 650°C TL = 700°C TL = 750°C TL = 800°C
P
1018 1017 52
Cr
1016 1015
Detection limit of Cr 1014 0
1
2 Depth (μm)
Fig. 1. Chromium and phosphorus depth profiles obtained by SIMS analysis for the wafers treated at different annealing temperatures TL. The broken line represents the detection limit of chromium for depth profiling in silicon (2 1014 cm 3) associated to Cameca IMS-4F instrument.
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
1
timpurity
1014 1015 1016 Apparent minority carrier density Δnapp (cm-3) Fig. 2. Apparent minority carrier lifetime versus minority carrier density as measured by QSSPC for investigated wafers at different annealing temperatures TL, and fits using the Hornbeck–Haynes model (solid lines).
¼
1
tnongettered
Se g re
ffe ct fus ivi ty e
40 30
ga ti
on
eff ec t
Dif
20
tp0 ¼
10 0 Ref.
600 650 700 750 Annealing temperature TL (°C)
800
Fig. 3. Effective lifetimes measured at an excess carrier density of 1 1015 cm 3 for gettered wafers at different annealing temperatures TL.
temperatures preventing the impurities gettering. These two opposite effects cause the observed optimum temperature [13,14]. Previous experimental work performed by Kang and Schroder [15] confirmed clearly this phenomenon. The improvement of carrier lifetime obtained for TL ¼800 1C without any chromium accumulation is not clear. It may be due to the change of chemical state of chromium (neutralization) in the bulk or/and to an external phosphorus diffusion gettering of other metallic impurities such as Mn, Fe, etc. The bulk lifetime tnon-gettered associated to reference wafer that has not been gettered, can be written in terms of two components [16]: 1
tnongettered
¼
1
timpurity
þ
1
tgettered
tn0 ðNA þp1 þ DnÞ þ tp0 ðn1 þ DnÞ
1 N nth sp
where timpurity is the lifetime due to the getterable impurities, and tgettered is the lifetime after gettering, or alternatively, the lifetime caused by the non-getterable
ð2Þ
NA þ Dn
tn0 ¼
1 Nnth sn
ð3Þ
Table 1 lists the values of capture cross sections (sn,p) and localised energy levels corresponding to Cri, CrB, Fei and FeB, used to generate the SRH fit curves. These values have been taken from Mishra [21] and Istratov et al. [22]. Clearly, only the recombination centre capable of adequately explaining the calculated injection-level dependence for all gettered wafers is the Cri impurity. Consequently, this result implies that the lifetime 10-5 9x10-6 8x10-6
TL (°C)Density of Cri (x1011cm-3) 1.12 600 1.28 650 1.22 700 1.17 750 1.21 800
7x10-6
Cri
6x10-6 CrB 5x10-6
ð1 aÞ
ð1 bÞ
where NA is the acceptor density. n1 and p1 are, respectively, the electron and hole densities when the Fermi level coincides with the recombination centre energy. tn0 and tp0 are, respectively, the fundamental electron and hole lifetimes, and are related to the carrier thermal velocity nth, the recombination centre density N and the capture cross sections sn and sp of the specific centre in question:
Effective Lifetime (s)
Effective minority carrier lifetime τeff (µs)
tSRH ðDnÞ ¼
50
1
tgettered
A similar calculation was also used by LI et al. [17] and Schmidt and Cuevas [18] to investigate the recombination activity of Cu in mc-Si samples contaminated at different temperatures and to study boron–oxygen complexes in Cz–Si, respectively. Fig. 4 shows the results of calculation and those of attempting to SRH fit for interstitial Fe, Cr, and FeB and CrB pairs. The SRH fit curves have been generated employing the following equation [19,20]:
60 Measured at Δnapp = 1 . 1015 cm-3
or
1x10-5
pt
Reference TL = 600°C TL = 650°C TL = 700°C TL = 750°C TL = 800°C
centres. Of course, an adequate surface passivation is necessary to avoid surface effects, and this was achieved by the deposition of a-SiNx:H film, as indicated above. By measuring the injection-level dependent lifetimes before and after gettering, it is possible then to determine the injection-level dependence of the gettered impurities alone via
ce
1.5 Ωcm p-type mc-Si
ac
1x10-4
Fe B
Apparent minority carrier lifetime τapp (s)
58
Fei
1015 Excess carrier density
1016 (cm-3)
Fig. 4. Effective carrier lifetime versus excess carrier density due to the removed impurities, modelled with SRH curves for interstitial Fe and Cr, and FeB and CrB pairs (solid lines).
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
59
Table 1 Energy levels and capture cross sections used for calculation.
sn (cm2)
Energy level (eV)
13
Ec 0.22 Ev þ 0.27 Ev þ 0.38 Ec 0.23
Cri CrB Fei FeB acceptor
2.3 10 1.4 10 13 5 10 14 3 10 14
1000
σn/σp Ratio
Fei
100
FeB CrB
10
Cri Ref.
600
650
700
750
800
Annealing temperature TL (°C) Fig. 5. Variation of the electron to hole capture cross-sections ratio (k¼ sn/sp) as a function of annealing temperature TL.
increment observed for temperature TL varying between 600 and 800 1C could be explained by the gettering of Cri. An interesting result that is relevant to the type of gettering could be deduced from this calculation. Indeed, the lifetime improvement observed for TL ¼800 1C (caused by the reduction of Cri density in the bulk) without any Cr accumulation in the phosphorus diffusion region, lead us to conclude that there is certainly a local deactivation of Cri recombination activity centres (internal gettering). In spite of the narrow variation range of Cri concentration, the tendency of this variation explains well the carrier lifetime change reported above, and confirms the evolution of Cr profiles obtained by SIMS, too. However, the values of Cri concentration estimated by the modelling of lifetime curves are between four and five orders of magnitude below those obtained directly by the SIMS analysis. These results indicate that the vast majority of the gettered chromium is required to present in some much more electronically benign form, such as in precipitates. Using Hornbeck–Haynes model [23], a complete modelling of injection-level dependent lifetime curves (see Fig. 2) allows to estimate the ratio k, and so to conclude the origin of dominant recombination activities after gettering process. Fig. 5 illustrates the variation of k ratio as a function of the extended gettering temperature TL. The basic formula used for modelling measured QSSPC data can be found in [24]. More details about the calculation formula and the fitting procedure are reported elsewhere in [25–27]. For TL r750 1C, it is obviously shown that the shift of k ratio from the value associated to Cri (k ¼2.09), to values associated to FeB, CrB pairs and Fei is accompanied by the
sp (cm2) 13
1.1 10 1 10 14 7 10 17 2 10 15
k¼ sn/sp
Reference
2.1 14 714.3 15
Mishra [21] Mishra [21] Istratov et al. [22] Istratov et al. [22]
decrease of Cri content in the bulk (see Fig. 4), and an increment of the chromium accumulation in HPD region (see Fig. 1). So, this result confirms that the lifetime improvement in this annealing temperature range is the consequence of the chromium extraction from the bulk, and indicates clearly that the origin of dominant recombination activity after gettering process is mainly due to the presence of CrB, FeB pairs and Fei centres. For the wafer gettered at TL ¼800 1C, k ratio shows an important shift towards the value associated to Fei (k¼714.28), so indicating the quasi-dominance of Fei recombination activity.
4. Conclusion In summary, we have studied the effect of extended PDG on chromium impurity in p-type HEM mc-Si wafers. The correlation between results obtained by the SIMS analysis and QSSPC measurement indicated an effective extended gettering of chromium impurity. This effect was especially observed around 650 1C and 800 1C, where the effective minority carrier lifetime was, respectively, improved by more than 900% and 800%. The SRH fit of calculated effective lifetime curves associated to gettered impurities proves that the origin of lifetime improvement is the reduction of Cri recombination centres density in the bulk. Because of the Cr accumulation observed in HPD region for annealing temperature TL r750 1C, the lifetime improvement obtained was identified as the consequence of an effective PDG of Cri. While the improvement observed at 800 1C (that is due to the decrease of Cri density, too) without any Cr accumulation is probably due to a local neutralization of Cri recombination centres (internal gettering). The determination of electron to hole capture crosssections ratio using Hornbeck–Haynes model fitting of QSSPC curves leads us to conclude that after each effective gettering of Cri, the impact of Fei recombination activity becomes more dominant. Future work will be focused on the extended gettering effectiveness of iron impurity with new parameters process.
Acknowledgements This work was financed by The General Directorate for Scientific Research and Technological Development (Algeria). The authors wish to thank Prof. M. Lemiti at the INSA-Lyon University (France) for assistance with part of the QSSPC measurements.
60
N. Khelifati et al. / Materials Science in Semiconductor Processing 15 (2012) 56–60
References [1] A. Cuevas, in: Proceedings of the Materials 2003 Conference, The Institute of Materials Engineering Australasia, Sydney, Australia, 2003. [2] V. Avrutin, N. Izyumskaya, H. Morkoc- , Superlattices and Microstructures 49 (2011) 337–364. [3] J.R. Davis, A. Rohatgi, R.H. Hopkins, P.D. Blais, P. Rai-Choudhury, J.R. Mccormick, H.C. Mollenkopf, IEEE Transactions on Electron Devices ED-27 (1980) 677. [4] A.A. Istratov, T. Buonassisi, R.J. McDonald, A.R. Smith, R. Schindler, J.A. Schindler, J.A. Rand, J.P. Kalejs, E.R. Weber, Journal of Applied Physics 94 (2003) 6552. [5] D. Macdonald, A. Cuevas, A. Kinomura, Y. Nakano, In: Proceedings of the 29th EEE photovoltaic specialists conference, New Orleans, May 2002, p. 285–288. [6] D. Macdonald, A. Cuevas, A. Kinomura, Y. Nakano, L.J. Geerligs, Journal of Applied Physics 97 (2005) 033523. [7] T. Buonassisi, A.A. Istratov, M.D. Pickett, M. Heuer, J.P. Kalejs, G. Hahn, M.A. Marcus, B. Lai, Z. Cai, S.M. Heald, T.F. Ciszek, R.F. Clark, D.W. Cunningham, A.M. Gabor, R. Jonczyk, S. Narayanan, E. Sauar, E.R. Weber, Progress in Photovoltaics: Research and Applications 14 (2006) 513. [8] D. Ouadjaout, Y. Gritli, L. Zair, M. Boumaour, Revue des Energies Renouvelables 8 (2005) 49–54. [9] P.S. Plekhanov, R. Gafiteanu, U.M. Gosele, T.Y. Tan, Journal of Applied Physics 86 (5) (1999) 1 September. [10] P. Manshanden, L.J. Geerligs, Solar Energy Materials and Solar Cells 90 (2006) 998–1012.
[11] C. Zhang, S. Liu, Y. Wang, Y. Chen, Materials Science in Semiconductor Processing 13 (2010) 209–213. [12] A. Bentzen, A. Holt, R. Kopecek, G. Stokkan, J.S. Christensen, B.G. Svensson, Journal of Applied Physics 99 (2006) 093509. [13] L. Baldi, G.F. Cerofolini, G. Ferla, G. Frigerio, Physica Status Solidi A 48 (1978) 523–532. [14] C. Del Canizo, A. Luque, Journal of The Electrochemical Society 147 (7) (2000) 2685–2692. [15] J.S. Kang, D.K. Schroder, Journal of Applied Physics 65 (1989) 2974–2985. [16] D. Macdonald, in Recombination and Trapping in Multicrystalline Silicon Sola Cells, Ph. D. Thesis, The Australian National University, 2001. [17] L.I. Xiao-qiang, Y.A.N.G. De-ren, Y.U. Xue-gong, Q.U.E. Duan-lin, Transactions of the Nonferrous Metals Society of China 21 (2011) 691–696. [18] J. Schmidt, A. Cuevas, Journal of Applied Physics 86 (1999) 3175–3180. [19] W. Shockley, W.T. Read, Physical Review 87 (5) (1952) 835–842. [20] R.N. Hall, Physical Review 87 (2) (1952) 387–387. [21] K. Mishra, Applied Physics Letters 68 (1996) 3281–3283. [22] A.A. Istratov, H. Hieslmair, E.R. Weber, Applied Physics A: Materials Science and Processing 69 (1) (1999) 13–44. [23] J.A. Hornbeck, J.R. Haynes, Physical Review 97 (1955) 311–321. [24] A. Bentzen, in Phosphorus Diffusion and Gettering in Silicon Solar Cells, Ph.D. Thesis, University of Oslo, 2006. [25] D. Macdonald, A. Cuevas, Applied Physics Letters 74 (1999) 1710. [26] D. Macdonald, A. Cuevas, Applied Physics Letters 75 (1999) 1571. [27] K.R. McIntosh, B.B. Paudyal, D. Macdonald, Journal of Applied Physics 104 (2008) 084503.