During the passivating contact workshop in Eindhoven earlier this year more than 80 people enjoyed several in-depth presentations and discussions on this hot topic in crystalline silicon PV research. Since the PV conference low season / summer break has started now, this is a great time to look back at the workshop! After last week’s post that discussed the need for passivating contacts for c-Si solar cells and their working mechanism, I will now summarize the next two contributed talks on passivating contact materials based on silicon, which were given by Dr. Frank Feldmann (Fraunhofer Institute for Solar Energy Systems ISE) and Dr. Mathieu Boccard (École Polytechnique Fédérale de Lausanne).
(*) The write-up in this post is my personal reflection of the speakers’ presentations. The speakers have seen the text and given their permission to use their visual materials. Special thanks go to Dr. Ingrid Romijn and Dr. Paula Bronsveld for their notes taken during the presentations!
Dr. Frank Feldmann (Fraunhofer Institute for Solar Energy Systems ISE):
“Poly-Si based passivating contacts” (*)
The first ideas for polycrystalline silicon (poly-Si) contacts date back to the 1980s and 1990s when this material was used in transistor applications. In 1985, poly-Si contacts – or rather so-called semi-insulating polycrystalline silicon (SIPOS) structures as they were called back then – were successfully used in a c-Si solar cell. Already around that time, very low values of the recombination parameter (J0 < 10 fA/cm2) and high Voc values of 720 mV were achieved [1,2,3]. These poly-Si structures were characterized by a trade-off: although a low J0 could be reached, this resulted in relatively high contact resistivity (ρc). Although ρc values around 10 mΩcm2 are too high for transistor applications, they are not a major issue for contacts in c-Si solar cells. Nevertheless, it took another 20 years until poly-Si was “discovered” by the PV community.
In 2010, Sunpower released its Gen 3 solar cells, which arguably include poly-Si contacts with thick thermal oxides at the interface. Although the company has not disclosed their technology, their patents give a good indication that poly-Si contacts are in fact used in their solar cells.
In 2013, Fraunhofer ISE published a solar cell efficiency exceeding 23% involving the tunnel oxide passivating contact (TOPCon) [4]. In this approach, a very thin tunnel interfacial oxide is positioned between poly-Si and c-Si. In that particular case, n-type poly-Si contacts were used, resulting in both low J0 (4-10 fA/cm2) and low ρc (< 10 mΩcm2) values. Following this strategy, a solar cell conversion efficiency of 25.8% was reached in 2017 [5]. Since the day of the workshop earlier this year, the record efficiency for a solar cell involving polysilicon contacts has been further improved to 26.1% at ISFH using an interdigitated back contact (IBC) solar cell architecture [6]. Note that ISFH employed a thicker interfacial oxide approach when compared to the TOPCon approach of Fraunhofer ISE.
While Fraunhofer ISE and ISFH started to work on this topic around the same time, many other institutes have followed in their investigations of poly-Si during the last 5 years and very low J0 values and cell efficiencies exceeding 25% have been demonstrated. Passivating contacts based on poly-Si have now become “hot” in research, as is illustrated for instance by the fact that the amount of published papers on this topic has increased rapidly in recent years.
The TOPCon process, as developed at Fraunhofer ISE, consists of a number of steps. After the formation of an ultrathin SiOx layer on the c-Si substrate which can be made in a variety of chemical ways [7], an amorphous silicon layer is deposited next. By a high temperature annealing step this a-Si layer is transformed into a doped poly-Si layer. Finally, hydrogenation is needed to provide sufficient passivation.
When mastering these four process steps it is possible to arrive at very low J0 and ρc values. As the field-effect passivation can be influenced by the doping level of the poly-Si layer it is important to control this well, such that the dopant diffusion into the wafer is not too large.
The current transport mechanism through the oxide layer has been suggested to be dominated by tunnelling through the layer [8], tunnelling assisted by pinholes [9], entirely via pinholes without tunneling [10,11] or even by a combination of these factors [12]. No generic explanation has been found yet. This is due not only to the differences in the oxide properties, such as the preparation method and the thickness, but also by the difficulties associated with determining the barrier height. At Fraunhofer ISE, temperature-dependent IV / TLM measurements on poly-Si/SiOx structures annealed at different temperatures indicated evidence for tunnelling, like in conventional metal-insulator-semiconductor (MIS) structures [13]. On the other hand, at ISFH, indications were found for pinhole-assisted transport by means of etch back experiments, pinhole density estimations through SEM studies, and extensive modelling [14,15,16]. A unifying picture – if it can be made – probably includes a combination of both mechanisms.
So far, three types of solar cell architectures involving poly-Si contacts have appeared:
- Hybrid (poly-Si on one side and a diffused junction on the other side)
- Interdigitated back contact (n-type and p-type poly-Si regions on one side of the wafer)
- Top-rear (n-type and p-type poly-Si regions on opposite sides of the wafer)
Several benefits of poly-Si contacts can be identified:
- Similar J0 but lower ρc can be achieved in comparison to the a-Si:H based contacts in the classical SHJ structure
- Conventional high temperature processing is possible
- Overall compatibility with mainstream PV processing technology
However, there are also a number of steps towards industrialization that need to be taken and challenges that need to be addressed when aiming for commercialization of poly-Si contacts:
- TCOs may be needed to provide sufficient lateral conductivity (like in case of a-Si:H in the SHJ structure), but this yields an increased ρc after annealing [17].
- Contact firing using silver screen print pastes locally destroys the poly-Si. Firing provides effective hydrogenation, but still the J0 value after metallization is much higher than it is before metallization due to Ag spiking. The best efficiencies using this approach have been reported by ECN (21.5%) and ISFH (21.2%). Note that relatively thick poly-Si is needed here, which lowers the bifaciality [18,19].
- A non-firing approach involving a TCO/metal stack on top of poly-Si on the rear side and NiCuAg plated contacts on the front side should be further investigated, as this already resulted in a 23.4% efficient cell at Fraunhofer ISE.
When aiming to introduce poly-Si in the PV industry, upscaling needs to be possible as well. This means that a high throughput process is needed with little to no downtime. To this end, Fraunhofer ISE is working with a number of PV companies in this context and a qualification of tools is needed.
In summary, poly-Si holds the promise of facilitated back end processing, though there are obviously hurdles that still need to be overcome first.
Dr. Mathieu Boccard (École Polytechnique Fédérale de Lausanne):
“Passivating contacts based on thin-film silicon and alloys” (*)
Developments in thin film Si, specifically a-Si:H, date back to the 1960s and 1970s. During those pioneering years, photoconductivity was first observed in a-Si:H [20] and it was demonstrated that doping could be achieved [21], although at the expense of additional defect formation [22].
It was also established that hydrogen was needed to achieve a high quality material, the first a-Si:H solar cell was fabricated [23], and a light-induced decrease in the photoconductivity was discovered (now known as the Staebler-Wronski effect) [24]. In the following decades, thin film silicon grew into a PV research field on its own. During this time, the fundamental properties of thin film Si and its alloys were extensively investigated and gradual improvements in the efficiency of thin film silicon solar cells followed. However, the efficiency would never catch up with c-Si based cells, largely due to the relatively low carrier mobility and high defect density in thin film Si materials [25,26,27].
In 1992, Sanyo introduced the first SHJ solar cells, involving a c-Si absorber and a-Si:H based passivating contact stacks, with an efficiency of 18% [28]. Since then, many research groups also became involved in SHJ solar cell development and in 2014 a record open-circuit voltage (Voc) of 750 mV on c-Si devices (and a cell efficiency of 24.7%) was demonstrated by Panasonic/Sanyo [29]. Because high efficiencies have already been demonstrated by many groups, a consolidation of the field, in particular in the industry, is currently taking place. Nevertheless, the highest efficiency of all c-Si devices known until now was reported in 2017 by Kaneka (26.6% [30,31]) which underlines the potential of this type of solar cell.
To widen the bandgap of a-Si:H, alloying with C or O can be applied [32,33], while Ge alloying reduces the bandgap [34]. The bandgap can also be tuned by changing the hydrogen content from 5% up to 30%, while hydrogenated nanocrystalline silicon (nc-Si:H) can be formed as well [35]. Since the nanostructure of thin film Si materials is complex, this means that drawing an accurate band diagram of a SHJ solar cell is complex as well. All layers in the cell influence each other and band diagrams in equilibrium are quite different from band diagrams under illumination, illustrating a clear difference between Voc and maximum power point (Vmpp) conditions.
To make an efficient SHJ solar cell, the passivation layers need to be amorphous, so the transition from a-Si:H to epitaxial Si should be avoided, as epitaxial growth damages the passivation quality [36]. a-Si:H that is deposited such that this phase transition is just avoided – the so-called “edge” material – is beneficial for the passivation quality and is characterized by an elevated bandgap [37,38,39]. An annealing treatment at a moderate temperature of ~200 °C is also beneficial, as it yields strong improvements in the device performance [40]. It seems that defects at the a-Si:H/c-Si interface are comparable to defects in the a-Si:H bulk, implying that hydrogenation of the interface is key to achieve good passivation [41,42]. Some of these insights were gained in the thin film Si research field and have been conveniently used again in the further development of SHJ solar cells.
There are also distinct differences between material development for thin film Si solar cells and SHJ solar cells. For instance, while the Staebler-Wronski effect is clearly undesirable for the performance of thin film solar cells, beneficial effects on the SHJ cell performance have been reported due to prolonged light soaking as well as annealing after light soaking. Although these effects are linked to interface hydrogenation, the exact mechanism is not yet well understood [43].
The deposition of p-type a-Si:H on the intrinsic a-Si:H layer can degrade the passivation quality in case of certain types of intrinsic a-Si:H, while this does not depend on the activation energy of the p-type a-Si:H layer [44,45]. Lifetime decreases are often due to the intrinsic layer rather than the doped layers, which means that the process window for the latter is typically wider. However, when it comes to annealing treatments, the doped layers, and especially the p-layer, suffer from dehydrogenation at lower temperatures than the intrinsic layer [45].
Rehydrogenation can however be accomplished if the a-Si:H layers are thick enough (> 8 nm). Furthermore, when incorporating C in the intrinsic a-Si:H layer, more hydrogen can be incorporated as well and the layer becomes more temperature stable (i.e. the passivation quality does not significantly deteriorate when annealing up until 300 °C) [46,47].
Sputtering TCO on a-Si:H layers is a delicate process and the passivation quality provided by the a-Si:H layers can be damaged. However, this damage can be recovered in some cases by annealing [48,49]. A further challenge in SHJ cell processing is to keep the current losses due to absorption in the (doped) a-Si:H and TCO layers as small as possible [50]. The absorption in a-Si:H might be addressed by alloying it with oxygen to widen the bandgap [51,52]. Another way to mitigate optical absorption losses is to use high mobility TCOs [53,54,55], while still allowing for efficient carrier transport through the layer.
For the future, there is a great promise for IBC SHJ solar cells. This has become especially attractive since promising results have been obtained with local area n-type a-Si:H regions covered by a full-area p-type a-Si:H layer to create a tunnelling contact, such that the usual complex structuring that is common in IBC solar cells can be avoided [56].
References
[1] Y.H. Kwark et al., IEDM Tech. Digest 30 (1984) 742. DOI: 10.1109/IEDM.1984.190832
[2] E. Yablonovitch et al., Appl. Phys. Lett. 47 (1985) 1211. DOI: 10.1063/1.96331
[3] N.G. Tarr, IEEE Electron Device Lett. 6 (1985) 655. DOI: 10.1109/EDL.1985.26264
[4] F. Feldmann et al., Sol. Energy Mater. Sol. Cells 131 (2014) 46. DOI: 10.1016/j.solmat.2014.06.015
[5] A. Richter et al., Sol. Energy Mater. Sol. Cells 173 (2017) 96. DOI: 10.1016/j.solmat.2017.05.042
[6] F. Haase et al., Sol. Energy Mater. Sol. Cells 186 (2018) 184. DOI: 10.1016/j.solmat.2018.06.020
[7] A. Moldovan et al., Sol. Energy Mater. Sol. Cells 142 (2015) 123. DOI: 10.1016/j.solmat.2015.06.048
[8] H.C. de Graaff and J.G. de Groot, IEEE Trans. Electron Devices 26 (1979) 1771. DOI: 10.1109/T-ED.1979.19684
[9] J.Y. Gan and R.M. Swanson, IEEE Conf. Proc. 1990, pp. 245-250. DOI: 10.1109/PVSC.1990.111625
[10] R. Peibst et al., IEEE J. Photovolt. 4 (2014) 841. DOI: 10.1109/JPHOTOV.2014.2310740
[11] U. Römer et al., Sol. Energy Mater. Sol. Cells 131 (2014) 85. DOI: 10.1016/j.solmat.2014.06.003
[12] R. Peibst et al., Sol. Energy Mater. Sol. Cells 158 (2016) 60. DOI: 10.1016/j.solmat.2016.05.045
[13] F. Feldmann et al., Sol. Energy Mater. Sol. Cells 178 (2018) 15. DOI: 10.1016/j.solmat.2018.01.008
[14] D. Tetzlaff et al., Sol. Energy Mater. Sol. Cells 173 (2017) 106. DOI: 10.1016/j.solmat.2017.05.041
[15] T.F. Wietler et al., Appl. Phys. Lett. 110 (2017) 253902. DOI: 10.1063/1.4986924
[16] N. Folchert et al., Sol. Energy Mater. Sol. Cells 185 (2018) 425. DOI: 10.1016/j.solmat.2018.05.046
[17] L. Tutsch et al., Energy Procedia (2018) (presented at SiliconPV 2018)
[18] S. Mack et al., Phys. Status Solidi RRL 11 (2017) 1700334. DOI: 10.1002/pssr.201700334
[19] H.E. Çiftpinar et al., Energy Procedia 124 (2017) 851. DOI: 10.1016/j.egypro.2017.09.242
[20] R.C. Chittick et al., J. Electrochem. Soc. 116 (1969) 77. DOI: 10.1149/1.2411779
[21] W.E. Spear and P.G. LeComber, Solid State Communications 17 (1975) 1193. DOI: 10.1016/0038-1098(75)90284-7
[22] G. Krötz et al., Phil. Mag. B 63 (1991) 101. DOI: 10.1080/01418639108224433
[23] D.E. Carlson and C.R. Wronski, Appl. Phys. Lett. 28 (1976) 671. DOI: 10.1063/1.88617
[24] D.L. Staebler and C.R. Wronski, Appl. Phys. Lett. 31 (1977) 292. DOI: 10.1063/1.89674
[25] M. Stuckelberger, PhD thesis, École Polytechnique Fédérale de Lausanne, Switzerland (2014). DOI: 10.5075/epfl-thesis-6393
[26] J. Melskens, PhD thesis, Delft University of Technology, the Netherlands (2015). DOI: 10.4233/uuid:13d60d5e-3d68-4ac7-b2b6-a867cc393231
[27] M. Stuckelberger et al., Renew. Sustain. Energy Rev. 76 (2016) 1497. DOI: 10.1016/j.rser.2016.11.190
[28] M. Tanaka et al., Jpn. J. Appl. Phys. 31 (1992) 3518. DOI: 10.1143/JJAP.31.3518
[29] M. Taguchi et al., IEEE J. Photovolt. 4 (2014) 96. DOI: 10.1109/JPHOTOV.2013.2282737
[30] K. Yoshikawa et al., Nature Energy 2 (2017) 17032. DOI: 10.1038/nenergy.2017.32
[31] K. Yoshikawa et al., Sol. Energy Mater. Sol. Cells 173 (2017) 37. DOI: 10.1016/j.solmat.2017.06.024
[32] A. Morimoto et al., J. Appl. Phys. 53 (1982) 7299. DOI: 10.1063/1.329879
[33] K. Haga and H. Watanabe, Jpn. J. Appl. Phys. 29 (1990) 636. DOI: 10.1143/JJAP.29.636
[34] M. Stutzmann et al., J. Appl. Phys. 66 (1989) 569. DOI: 10.1063/1.343574
[35] J. Meier et al., Appl. Phys. Lett. 65 (1994) 860. DOI: 10.1063/1.112183
[36] S. de Wolf and M. Kondo, Appl. Phys. Lett. 90 (2007) 042111. DOI: 10.1063/1.2432297
[37] R. Bartlome et al., Appl. Phys. Lett. 94 (2009) 201501. DOI: 10.1063/1.3141520
[38] A. Descoeudres et al., Appl. Phys. Lett. 97 (2010) 183505. DOI: 10.1063/1.3511737
[39] A. Descoeudres et al., Appl. Phys. Lett. 99 (2011) 123506. DOI: 10.1063/1.3641899
[40] S. de Wolf et al., Appl. Phys. Lett. 93 (2008) 032101. DOI: 10.1063/1.2956668
[41] S. de Wolf et al., Phys. Rev. B 85 (2012) 113302. DOI: 10.1103/PhysRevB.85.113302
[42] J. Shi et al., Appl. Phys. Lett. 109 (2016) 031601. DOI: 10.1063/1.4958831
[43] E. Kobayashi et al., Appl. Phys. Lett. 109 (2016) 153503. DOI: 10.1063/1.4964835
[44] S. de Wolf and G. Beaucarne, Appl. Phys. Lett. 88 (2006) 022104. DOI: 10.1063/1.2164902
[45] S. de Wolf and M. Kondo, J. Appl. Phys. 105 (2009) 103707. DOI: 10.1063/1.3129578
[46] W. Beyer in “Thin-Film Silicon Solar Cells” (A.V. Shah ed.) (2010)
[47] M. Boccard and Z.C. Holman, J. Appl. Phys. 118, (2015) 065704. DOI: 10.1063/1.4928203
[48] B. Demaurex et al., Appl. Phys. Lett. 101 (2012) 171604. DOI: 10.1063/1.4764529
[49] A. Tomasi et al., IEEE J. Photovolt. 6 (2016) 17. DOI: 10.1109/JPHOTOV.2015.2484962
[50] Z.C. Holman et al., IEEE J. Photovolt. 2 (2012) 7. DOI: 10.1109/JPHOTOV.2011.2174967
[51] J.P. Seif et al., J. Appl. Phys. 115 (2014) 024502. DOI: 10.1063/1.4861404
[52] L. Mazzarella et al., Appl. Phys. Lett. 106 (2015) 023902. DOI: 10.1063/1.4905906
[53] T. Koida et al., Jpn. J. Appl. Phys. 46 (2007) L685. DOI: 10.1143/JJAP.46.L685
[54] L. Barraud et al., Sol. Energy Mater. Sol. Cells 115 (2013) 151. DOI: 10.1016/j.solmat.2013.03.024
[55] B. Macco et al., Phys. Status Solidi RRL 8 (2014) 987. DOI: 10.1002/pssr.201409426
[56] A. Tomasi et al., Nature Energy 2 (2017) 17062. DOI: 10.1038/nenergy.2017.62
Leave a Reply