Available online at www.sciencedirect.com

ScienceDirect Energy Procedia 77 (2015) 586 – 591

5th International Conference on Silicon Photovoltaics, SiliconPV 2015

Influence of copper diffusion on lifetime degradation in n-type Czochralski silicon for solar cells Guilherme Gaspara,*, Chiara Modanesea, Hendrik Schønb, Marisa Di Sabatinoa, Lars Arnberga, Eivind J. Øvrelidc a

Norwegian University of Science and Technology, Dep. Materials Science and Engineering, Trondheim NO-7491, Norway b NorSun AS, Sommerrogaten 13-15, Oslo NO-0255, Norway c SINTEF Materials and Chemistry, Trondheim NO-7465, Norway

Abstract The impact of copper (Cu) contamination on the minority carrier lifetime degradation of the last solidified fraction of an n-type Czochralski (CZ) silicon ingot was investigated. Lifetime degradation of more than 50% was measured at the top of an as-cut block during the first 2000 h (~83 days). Depth profiles of Cu concentration were collected over a depth of 20 μm from the surface and an exponential decrease with depth was observed. The critical concentration for Cu precipitation was exceeded within the top 2 μm. The mechanism of Cu diffusion and precipitation towards the surface at room temperature is the suggested phenomenon for the lifetime degradation. The effect of surface potential on Cu gettering was investigated in neighboring samples by grinding the surface with different SiC grit size prior to storage in the dark. The measured Cu profiles show that an increased surface area enhances the Cu precipitation at the surface and leads to a lower concentration deeper in the bulk. Since Cu diffuses over large distance at room temperature in n-type CZ silicon and it is mainly segregated at the last solidified part of the ingot, we suggest that the tail should be detached from the main body of the ingot in order to remove a potential source of Cu contamination. © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license © 2015 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peerreview reviewbyby scientific conference committee of SiliconPV 2015responsibility under responsibility Peer thethe scientific conference committee of SiliconPV 2015 under of PSE AGof PSE AG. Keywords: silicon; copper; diffusion; precipitation; lifetime.

* Corresponding author. Tel.: +47 73594867. E-mail address: [email protected]

1876-6102 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer review by the scientific conference committee of SiliconPV 2015 under responsibility of PSE AG doi:10.1016/j.egypro.2015.07.084

Guilherme Gaspar et al. / Energy Procedia 77 (2015) 586 – 591

587

1. Introduction Monocrystalline silicon represents approximately 35% of the total solar cell production worldwide in 2014, where Czochralski (CZ) crystal pulling is the dominant fabrication process. CZ silicon has a lower defect density than multicrystalline silicon materials [1, 2] and consequently the corresponding solar cells reach higher energy conversion efficiencies. However, relatively high concentrations of metallic impurities might be found in the last solidified part of the ingot due to their low effective segregation coefficient [3]. Cu is generally measured in significant amounts in this region of the ingot, where wafers are still used for solar cells fabrication. Cu is a fast-diffusing element in silicon [4, 5], where an intrinsic diffusion length in the order of 1 mm is reached after 3 h at room temperature [6]. Cu tends to diffuse towards areas of higher chemical potential, such as extended defects and surfaces [7]. Due to both high diffusivity and low solubility of Cu in silicon, the solubility limit can easily be reached locally and precipitates are formed [6]. Cu becomes strongly recombination active for free carriers in the material, since Cu precipitates occupy a central position in the silicon band gap [8]. The formation of Cu precipitates could thus be detrimental to both bulk and emitter recombination, hence be a limiting factor for the solar cells efficiency [9, 10]. In this work, we have investigated the source of lifetime degradation directly measured on a block cut from the last solidified part of a phosphorus-doped Czochralski silicon ingot. For this purpose, the bulk impurity concentration was measured by glow discharge mass spectrometry (GDMS). Apart from the dopant, Cu was found to be the main impurity in the material, and the source of Cu contamination might be feedstock, crucible or industrial growth environment. Cu depth profiles were measured under different conditions, i.e. time after surface grinding and surface roughness. Cu concentration tends to decrease with increasing distance from the surface, reaching an almost constant concentration deeper in the bulk. Cu diffusion and precipitation close to the surface is the suggested mechanism responsible for the lifetime degradation found in the material under investigation. 2. Experimental A phosphorus-doped Czochralski crystal was grown along the direction with a diameter of 160 mm. A block was cut from the last solidified part of the crystal with a height of 130 mm, which includes 30 mm of body (Fig. 1a). The resistivity was measured by four point probe (FPP) at the top of the block and an average value of 2 ȍcm was observed. The effective minority carrier lifetime was measured at the top of the sample by transientphotoconductance decay (PCD) [11]. The bulk lifetime was measured at the center and close to the edge of the block (Fig. 1b) during a period of approximately 11000 h (~15 months). A slab was cut from the top of the block with a thickness of ~2 mm. Square samples with 20x20 mm2 size were taken from center and edge of the slab (Fig. 1c) and double-side mechanically polished. The interstitial oxygen (Oi) concentration was measured by Fourier transform infrared spectroscopy (FTIR) according to the SEMI MF1188 standard, where an average concentration of 16 ppma was detected for the whole slab. Bulk impurity concentration were measured by GDMS, which has a low detection limits [12] and, an initial pre-sputter that removes ~10 μm of material followed by four repeated analyses, for a total of 26 μm (more details of the analytical method can be found in Ref. [13]). The Cu concentration depth profiles were measured on samples from the central region of the slab after different periods of storage in the dark (2 and 8 h), i.e. the samples were kept in a black box at room temperature for the mentioned time intervals. The depth resolution of the GDMS analysis was 0.5 μm, for a total sputtered depth of 20 μm. In addition, two square samples were taken from the center of the slab and ground with different SiC papers (320 and 1200 grit size) after Cu bulk concentration measurements. After grinding, the surface roughness of the samples was measured with 15 μm step resolution along a horizontal line of 5 mm and surface roughness calculated according to Ref. [14]. Both samples were stored in the dark during 672 h (4 weeks) and Cu depth profiles measured similarly to the other samples. A neighboring sample was also analyzed, where a series of Cu depth profiles was measured in as-cut conditions, at time 0 and after 2 and 4 weeks of storage in the dark. The previously sputtered craters were removed by surface grinding with SiC paper (500 grit size) immediately after the GDMS analysis. Both storage conditions and depth profile measurement and surface grinding were undertaken under similar conditions.

588

Guilherme Gaspar et al. / Energy Procedia 77 (2015) 586 – 591

Fig. 1. (a) schematic of the position of the block in the CZ ingot; (b) top view of the block, where the position of the coil for lifetime measurements is included (black rectangles); (c) silicon slab was cut from the top of the block, where the positions of the square samples used for impurities analysis are shown (white rectangles).

3. Results and discussion Fig. 2 shows the bulk lifetime measured at the center and edge over a period of 11000 h, where similar lifetime trends were observed for both positions. The lifetime in the center of the sample was higher during the whole period compared to the edge. The values tend to stabilize after a period of approximately 2000 h, where a lifetime reduction of more than 50% was observed. The lifetime was measured with relatively high frequency at the beginning, until a plateau was observed. Before cutting the slices for further measurements (i.e. chemical analysis), the lifetime was measured once more at 11000 h to confirm the plateau trend observed at around 3000 h. The concentration of the main metallic and doping impurities measured on samples taken from the top of the block is shown in Fig. 3 (left). Among the impurities analyzed, Cu has the highest concentration besides the added dopant. Due to the relatively low concentration of boron and iron, the formation of detrimental pairs such as boron-oxygen and boron-iron pairs is neglected. Depth- and time-dependent Cu concentration profiles are shown in Fig. 3 (right). The samples were stored in the dark for 2 h and 8 h after grinding and prior to the Cu depth profile measurements. An exponential increase of the Cu concentration was found towards the surface. It is shown that constant concentrations are obtained deeper in the bulk (>20 μm) for both samples. Due to the tuning of the analyses, the top-most sputtered layer (first 0.5 μm of material) is missing in the results. The Cu concentration in this sputtered layer is expected to differ according to the storage time and may be higher for a longer storage time.

Guilherme Gaspar et al. / Energy Procedia 77 (2015) 586 – 591

589

Fig. 2. Minority carrier lifetime measured on the top of the block, both at the side and the center positions.

Fig. 3. Bulk impurities concentration at center and edge of the wafer (left) and Cu depth profiles for 2 h (S1.5 H2) and 8 h (S1.5 H8) storage of samples in the dark (right). The arrows show the trend of the depth profile with time.

Due to the gettering capabilities of defects, diffusion of Cu is enhanced and the solubility limit is easily reached locally. This results in the formation of Cu precipitates close to these sites. Above a critical intrinsic concentration of 200 ppbw [7], Cu predominantly precipitates in silicon and such concentration level was found within the top 2 μm. Cu-rich precipitates are active recombination centers since they occupy a position close to the middle of the bandgap [8, 15]. It is therefore believed that the observed lifetime degradation is associated with Cu out-diffusion and precipitation at the block surface. The formation of Cu precipitates at the surface was confirmed by measuring the lifetime, as described above, on part of the slab cut from the block. The sample was initially ground, where approximately 0.5 mm of thickness was removed, and the lifetime measured before and after sample storage in the dark. Similar lifetime degradation curve was observed compared to the block; however the large lifetime degradation took place within a period of approximately 8 h instead of 2000 h and lower values were measured. It could be explained by the fact that the reduced thickness of the slab leads to a higher contribution of the surface recombination to the measured lifetime as well as the Cu migration over much shorter distance until the equilibrium Cu precipitation is reached close to the surface. A strong effect of Cu on lifetime reduction in n-type silicon was observed earlier, even at low Cu contamination levels, demonstrating the importance of the recombination strength of Cu in this type of material [16, 17]. In n-type silicon, the room-temperature Fermi level is very close to the electroneutrality level of Cu precipitates and thus the

590

Guilherme Gaspar et al. / Energy Procedia 77 (2015) 586 – 591

nucleation of precipitates is enhanced. It is found to be more detrimental in n-type silicon compared to p-type at low concentrations since a lower Cu concentration is necessary to initiate Cu precipitate nucleation in n-type silicon. The diffusion coefficient of interstitial Cu in intrinsic silicon at room temperature was computed according to Ref. [4] and a diffusion length of approximately 30 mm was found for a period of 2000 h. This value is probably an overestimation as Cu diffusion is significantly reduced in phosphorous-doped material. Due to the significantly larger block height (~130 mm), it is believed that Cu migration towards the block surface was enhanced by electrostatic interaction between the ions in the solid and Cu precipitates formed close to the surface [7]. In order to study the effect of surface potential on Cu gettering, two 20x20 mm2 samples were ground with SiC papers of different grit size prior to depth profile measurements, i.e. the surface area of the samples was different. The surface roughness variations were measured along a horizontal line of 5 mm after grinding with 320 and 1200 SiC papers, where an increase of 8 times of the surface roughness was observed for the 320 SiC paper compared to the 1200 SiC paper. The Cu depth profiles were measured after 4 weeks of storage in the dark (Fig. 4 (left)). Since the Cu bulk concentration was observed to be slightly different from sample to sample and could thus have a direct influence on the Cu diffusion and precipitation close to the surface, the profiles were normalized over the initial Cu bulk concentration measured prior to grinding and storage. The measured Cu profiles show that an increased surface area has a higher potential to attract Cu, leading to a lower concentration deeper in the bulk. An increase of 8 times of the surface roughness resulted in a reduction of 50 % in the average bulk concentration. A higher surface roughness also leads to a smoother transition in Cu depth concentrations close to the surface and the bulk, and consequently higher Cu concentration is found within the first 3 μm. It was demonstrated earlier that Cu precipitates are positively charged in p-type silicon and negatively charged or neutral in n-type silicon [18]. A higher density of negatively charged precipitates formed due to the more extended surface area results in a stronger attraction between the precipitates initially formed close to the surface and the ions in the bulk, leading to an increased Cu concentration and precipitation towards the surface. In addition to the high diffusivity of Cu in silicon, the electrostatic interaction between precipitates initially formed at the surface and the Cu ions in the bulk is suggested to enhance the Cu diffusion towards the surface. Due to these results, a neighboring sample was analyzed in order to study the Cu gettering at the sample surface. The Cu depth profile was initially measured under as-cut conditions followed by surface grinding and storage. After storage, the Cu depth profile was re-measured and the sample re-ground before a new storage time of 2 week. The measured depth profiles are shown in Fig. 4 (right) for the as-cut conditions and after 2 and 4 weeks after the first surface grinding. The profiles show that the successive sample storage and surface grinding decreases the Cu bulk concentration and consequently less Cu precipitation takes place close to the sample surface. Therefore, the Cu precipitation at the surface with consecutive surface removal has revealed to be an efficient mechanism for bulk Cu concentration reduction in contaminated wafers. It could result in an improved material electrical properties as well as a reduction in the interaction between precipitates at the wafer surface and solar cell emitter.

Fig. 4. Cu depth profiles after 4 weeks of surface grinding (left). The profiles are normalized with the bulk Cu concentration under as-cut conditions. Cu profiles measured consecutively in the same sample for as-cut conditions, after 2 and 4 weeks of spot removal (right).

Guilherme Gaspar et al. / Energy Procedia 77 (2015) 586 – 591

591

4. Conclusions It was demonstrated in this work that Cu in n-type silicon diffuses over large distances, even at room temperature. Due to the high diffusivity of Cu towards the wafer surface, associated to the low solubility of Cu in silicon, precipitates are easily formed and consequently significant degradation of the electrical properties is obtained locally at the surface. A degradation of the minority carrier lifetime of 50% after 2000 hours was measured at the top of the block under investigation. The results show that the last solidified part of a CZ silicon ingot should be cut from the main body quite rapidly after crystallization since it is a potential source of Cu contamination for the rest of the ingot. A higher surface roughness has shown a higher potential to getter Cu close to the surface and thus a lower contamination is obtained deeper in the bulk of the wafer. Therefore, in contaminated wafers and under inefficient external gettering conditions, Cu might remain in the bulk and precipitates are formed close to the surface where the B-emitter layer is further deposited during solar cells fabrication. Acknowledgements The authors would like to thank to Norwegian Research Center for Solar Cell Technology (project number 193829) co-sponsored by the Norwegian Research Council and research and industry partners in Norway. References [1] W. Zulehner, Czochralski growth of silicon, Journal of Crystal Growth 65 1983: p. 189-213. [2] W. C. Dash, Growth of Silicon Crystals Free from Dislocations, Journal of Applied Physics 30 1959: p. 459-474. [3] K. Tang, E. J. Øvrelid, G. Tranell, M. Tangstad, Critical assessment of the impurity diffusivities in solid and liquid silicon, JOM 61 2009: p. 49-55. [4] A. A. Istratov, E. R. Weber, Physics of Copper in Silicon, Electrochemical Society 149 2002: p. G21-G30. [5] E. R. Weber, Impurity precipitation, dissolution, gettering and passivation in PV silicon, NREL - NREL/SR-520-31528 2002. [6] A. A. Istratov, C. Flink, H. Hieslmair, S. A. McHugo, E. R. Weber, Diffusion, solubility and gettering of copper in silicon, Materials Science and Engineering B 72 2000: p. 99-104. [7] C. Flink, H. Feick, S. A. McHugo, W. Seifert, H. Hieslmair, T. Heiser, A. A. Istratov, E. R. Weber, Out-diffusion and precipitation of copper in silicon: An electrostatic model, Physical Review Letters 85 2000: p. 4900-4903. [8] W. Shockley, W. T. Read, Statistics of the Recombinations of Holes and Electrons, Physical Review 87 1952: p. 835-842. [9] T. Buonassisi, M. A. Marcus, A. A. Istratov, M. Heuer, T. F. Ciszek, B. Lai, Z. Cai, E. R. Weber, Analysis of copper-rich precipitates in silicon: Chemical state, gettering, and impact on multicrystalline silicon solar cell material, Journal of Applied Physics 97 2005: p. 063503. [10] S. A. McHugo, A. C. Thompson, G. Lamble, C. Flink, E. R. Weber, Metal impurity precipitates in silicon: chemical state and stability, Physica B: Condensed Matter 273-274 1999: p. 371-374. [11] H. Nagel, C. Berge, A. G. Aberle, Generalized analysis of quasi-steady-state and quasi-transient measurements of carrier lifetimes in semiconductors, Journal of Applied Physics 86 1999: p. 6218-6221. [12] M. Di Sabatino, Detection limits for glow discharge mass spectrometry (GDMS) analyses of impurities in solar cell silicon, Measurement 50 2014: p. 135-140. [13] C. Modanese, G. Gaspar, L. Arnberg, M. Di Sabatino, On copper diffusion in silicon measured by glow discharge mass spectrometry, Analytical and Bioanalytical Chemistry 406 2014: p. 7455-7462. [14] E. Degarmo, J. Black, R. Kohser, Materials and Processes in Manufacturing, 9th ed. Wiley, 2003. [15] R. Sachdeva, A. A. Istratov, E. R. Weber, Recombination activity of copper in silicon, Applied Physics Letters 79 2001: p. 2937-2939. [16] D. C. Grupta, F. R. Bacher, W. M. Hughes, Recombination Lifetime Measurements in Silicon, Philadelphia: American Society for Testing Materials, 1998. [17] W. B. Henley, D. A. Ramappa, L. Jastrezbski, Detection of copper contamination in silicon by surface photovoltage diffusion length measurements, Applied Physics Letters 74 1999: p. 278-280. [18] A. A. Istratov, H. Hedemann, M. Seibt, O. F. Vyvenko, W. Schroter, T. Heiser, C. Flink, H. Hieslmair, E. R. Weber, Electrical and recombination properties of copper-silicide precipitates in silicon. Electrochemical Society 145 1998: p. 3889-3898.