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 passivating contact materials based on silicon, I will now summarize the last two contributed talks on novel, non-silicon based passivating contact materials, which were given by Dr. Martin Bivour (Fraunhofer Institute for Solar Energy Systems ISE) and Dr. Thomas Allen (King Abdullah University of Science and Technology).
(*) 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. Martin Bivour (Fraunhofer Institute for Solar Energy Systems ISE):
“Metal oxide based silicon heterojunctions” (*)
Alternatives to doped a-Si:H are desirable for a passivating contact, since the use of a-Si:H is associated with parasitic absorption, the doping efficiency in a-Si:H is relatively poor, and dangerous gases are used in the chemical vapor deposition process. To replace a-Si:H, candidate materials should ideally be transparent and have either a high or a low work function for the formation of hole- or electron-selective contacts, respectively. Such transparent contact layers are known from the OLED and organic PV research fields, but have only recently gained attention in c-Si PV research. To make these materials a success in c-Si solar cells, careful work function engineering is required to create a selective contact [1].
When replacing the p-type a-Si:H layer in a SHJ solar cell by molybdenum oxide (MoOx) or tungsten oxide (WOx) to make a hole-selective contact, as was tested for instance at EPFL [2,3] and Fraunhofer ISE [4], it possible to reach a similar Voc, higher Jsc, and a somewhat lower FF. Similarly, n-type a-Si:H can be replaced by titanium oxide (TiOx) to form an electron-selective contact. So far, TiOx has been most successful as electron-selective contact on the rear side of a homojunction solar cell with a diffusion on the front side, as demonstrated at ANU/KAUST. This approach involving TiOx has resulted in solar cell efficiencies exceeding 21-22% [5,6], while a cell that contained both MoOx and TiOx contacts reached up to an efficiency of 20.7% already [7]. The use of these metal oxides can be seen as a renaissance of MIS/SIS junctions that were explored already in the 1970s for transistor applications [8], but later on also in the fields of organic electronics [9], water splitting [10], metal source/drain MOSFET technology [11], and c-Si solar cells [12].
Generally, there are three possible types of contacts to a solar cell:
- The ideal case: contacts that provide good passivation and good carrier selectivity while ensuring that the majority carrier conductivity in the contact is much higher than the minority carrier conductivity [13,14]
- Good selectivity only: contacts that provide good selectivity but exhibit poor passivation
(iVoc = Voc) - Good passivation only: contacts that provide good passivation but exhibit poor carrier selectivity (Voc < iVoc and S-shaped IV curves) [15]
Simulations have indicated that the work function value (assuming operating conditions of a solar cell and perfect surface passivation) strongly influences the band bending and hence the Voc value due to a more or less adequate suppression of the minority carrier concentration in the contact. This makes it hard to effectively use TiOx and indium tin oxide (ITO) together. The Voc value is also affected by the work function of the adjacent metal. However, in practice it is harder to obtain high Voc values than what is predicted by the simulations due to imperfect surface passivation and Fermi level pinning. Also, the amount of band bending can be reduced after annealing treatments due to structural changes in the material like crystallization. On the other hand, certain chemical interactions at an interface can greatly improve the contact performance, as is for instance the case for the TiOx/Al interface or the interface of organic materials with silicon where dipoles play an important role in achieving carrier selectivity. Generally speaking, sufficient band bending is required to make a passivating contact, but it is not guaranteed to also provide sufficient carrier selectivity. For instance the presence of trap states may be needed in addition to achieve proper carrier selectivity, although the influence of the trap density and the distribution of traps has only been studied in simulations so far [16,17,18,19,20,21].
When assessing the suitability of a metal in a contact configuration, the value of the work function is important: a high value is preferred for a hole-selective contact, while a low value is better for an electron-selective contact. However, when bringing a metal directly in contact with silicon strong Fermi level pinning occurs, as has been reported already in the 1970s for a-Si:H Schottky diodes [22], which is very detrimental to the surface passivation quality and the carrier selectivity. When inserting another layer between the metal and the silicon, this problem can be circumvented. When inserting for instance an ITO interlayer with a high doping level the maximum Voc of the solar cell that could be achieved increases with increasing work function of the ITO, while this is still not sufficient to make an effective hole-selective contact [15]. More successful hole-selective contacts based on metal oxides that have a much higher work function than ITO include MoOx, WOx, and VOx [4,23,24,25,26,27]. Following the same work function analogy, an electron-contact based on Al, which has a low work function, can be improved by a TiOx interlayer [28].
Regarding MoOx, evaporation has so far resulted in the smallest difference between iVoc and Voc (~50 mV illustrates good carrier selectivity) when compared to the same material prepared by atomic layer deposition (ALD) or sputtering. Note that the amount of band bending for both MoOx and tungsten oxide (WOx) is similar to p-type a-Si:H before annealing. However, the amount of band bending alone does not guarantee a good contact quality. Only in case of ultrathin layers (~2 nm or thinner) transport can rely on direct tunneling, while for thicker layers a sufficiently high trap density in the metal oxide layer also has to be present to ensure good transport.
In all cases, an intrinsic a-Si:H interlayer is used to provide sufficient chemical passivation. The best solar cell efficiency of 22.6% (Jsc = 40.3 mA/cm2; Voc = 699 mV; FF = 80.2%) has so far been obtained using a stack of i-type a-Si:H / MoOx / ITO. Note that the interlayer could also be SiOx instead of a-Si:H to mitigate the optical losses. Since the a-Si:H layer contributes significantly to ρc, it is attractive to keep this layer as thin as possible. During an annealing treatment it appears that a more limited FF degradation is obtained when keeping the MoOx layer thin [19].
Given the high performance of SHJ solar cells with metal oxide based contacts realized thus far, it is worth identifying the industrial challenges associated with metal oxide contacted solar cells. These challenges can be summarized as follows:
- Metal oxide based contacts that have so far been demonstrated in lab scale solar cells are not compatible with the typical industrial 200 °C back end solar cell processing
- Upscaling of the deposition of metal oxides remains to be demonstrated
- The thermal and chemical stability of these materials has not yet been thoroughly investigated, as well as the long term stability
Despite these challenges it is good news that many groups are now working on this topic, which means that further knowledge is being acquired in the field. The high transparency and high FF that can be achieved with metal oxide based passivating contacts are encouraging for further material development, as well as the fact that the work function of metal oxides can be engineered. Considering these merits of metal oxide based contacts, there is a fair chance that the above challenges can be adequately addressed.
Dr. Thomas Allen (King Abdullah University of Science and Technology):
“Electron contacts to crystalline silicon” (*)
For the fabrication of contacts in c-Si solar cells it is worthwhile considering materials that have already been applied as contacts in other fields of research, such as organic solar cells. The quality of these novel contact materials, most of which appear to have electron-selective properties, can be assessed in terms of J0 and ρc [29,30,31].
In classical homojunction solar cells, a low ρc can be achieved, typically by resorting to an n+ diffusion. However, J0 tends to be on the order of pA/cm2, which is much too high to achieve high conversion efficiencies. For example, J0 < 5 fA/cm2 is required to achieve an efficiency of 29% [32]. This problem is typically addressed by making the contacts as small as possible and/or by using higher doping below the contacts to reduce the minority carrier concentration near the contact as much as possible [33,34,35].
When aiming for a carrier-selective contact, low J0 values can typically be reached, while achieving low ρc is a challenge. It turns out to be hard to make an electron-selective contact on n-type c-Si due to Fermi level pinning which lowers the amount of band bending (typical barrier height > 0.7 eV for Al and Ag contacts). Slightly lower barriers tend to form at the p-type c-Si / metal interface (< 0.6 eV) [36]. To circumvent these barrier problems, extremely low work function materials are attractive for electron-selective contacts.
An example of such a material is Ca (work function ~2.9 eV), which on undoped c-Si yields a low ρc value of 2-6 mΩcm2. When fabricating n-type based passivated emitter rear contact (PERC) solar cells with local Ca contacts on the rear side of the wafer, a conversion efficiency of 20.3% could be achieved (on 0.9 Ωcm c-Si) [37].
To improve the open-circuit voltage and conversion efficiency, a passivating interlayer is required. While gallium oxide and a-Si:H did not yield good results in this respect, a TiOx interlayer below the Ca layer did improve the Voc value. It appears that there is oxygen accumulation at the Ca/TiOx interface, making a very good contact with a low work function material possible. This approach yielded a conversion efficiency of 21.8% (Voc = 681 mV) [38].
Alternative approaches not involving Ca are for example other low work function materials, such as lithium fluoride (LiF) / Al (Voc = 676 mV; η = 20.6%) [39,40], magnesium fluoride (MgFx) / Al (Voc = 687 mV; η = 20.1%) [41], Mg (Voc = 637 mV; η = 19.0%) [42], and MgOx (Voc = 629 mV; η = 20.0%) [43]. Although these materials are not traditionally in the scope of c-Si PV research, they are still worth further consideration due to their ease of processing and limited additional processing complexity in solar cell fabrication.
Apart from the alkali/alkaline metals, metal oxides, and metal fluorides mentioned above, there are also still many more materials – for instance from the field of organic electronics – which have only been marginally explored as contact materials in c-Si solar cells [12,44]. Chemists do not naturally apply their knowledge to c-Si PV, but interesting materials may come out of their research for this application anyway. Therefore, it is worthwhile to be creative in finding new materials, including organic materials [20,45].
However, it is also important to be critical. Achieving good passivation is hard but essential in realizing passivating contacts; successful passivation is more likely to be guaranteed when resorting to interlayer materials such as a-Si:H and SiOx. Additionally, long term stability is still a largely overlooked topic that deserves more attention in the future. Finally, thinking outside the box is needed to identify new candidate materials for passivating contacts.
References
[1] M. Bivour et al., Energy Procedia 38 (2013) 658. DOI: 10.1016/j.egypro.2013.07.330
[2] C. Battaglia et al., Appl. Phys. Lett. 104 (2014) 113902. DOI: 10.1063/1.4868880
[3] J. Geissbuhler et al., Appl. Phys. Lett. 105 (2015) 081601. DOI: 10.1063/1.4928747
[4] M. Bivour et al., Sol. Energy Mater. Sol. Cells 142 (2015) 34. DOI: 10.1016/j.solmat.2015.05.031
[5] X. Yang et al., Adv. Mater. 28 (2016) 5891. DOI: 10.1002/adma.201600926
[6] X. Yang et al., Prog. Photovolt. Res. Appl. (2017). DOI: 10.1002/pip.2901
[7] J. Bullock et al., ACS Energy Lett. 3 (2018) 508. DOI: 10.1021/acsenergylett.7b01279
[8] J. Shewchun et al., IEEE Trans. Electron Devices 27 (1980) 705. DOI: 10.1109/T-ED.1980.19926
[9] J. Chen et al., J. Mater. Chem. 20 (2010) 2575. DOI: 10.1039/B925382C
[10] M.G. Walter et al., Chem. Rev. 110 (2010) 6446. DOI: 10.1021/cr1002326
[11] J.M. Larson and J.P. Snyder, IEEE Trans. Electron Devices 53 (2006) 1048. DOI: 10.1109/TED.2006.871842
[12] J. Bullock et al., IEEE PVSC Conf Proc. (2016) pp. 0210-0214. DOI: 10.1109/PVSC.2016.7749580
[13] P. Würfel, “Physics of Solar Cells – From principles to new concepts” (2005).
[14] U. Würfel et al., IEEE J. Photovolt. 5 (2015) 461. DOI: 10.1109/JPHOTOV.2014.2363550
[15] M. Bivour, PhD thesis, University of Freiburg (2015).
[16] M. Bivour et al., IEEE J. Photovolt. 4 (2014) 566. DOI: 10.1109/JPHOTOV.2013.2294757
[17] M. Bivour et al., Energy Procedia 92 (2016) 443. DOI: 10.1016/j.egypro.2016.07.125
[18] M. Bivour et al., Energy Procedia 124 (2017) 400. DOI: 10.1016/j.egypro.2017.09.259
[19] L. Neusel et al., Energy Procedia 124 (2017) 425. DOI: 10.1016/j.egypro.2017.09.268
[20] C. Reichel et al., Appl. Phys. Lett. 123 (2018) 024505. DOI: 10.1063/1.5010937
[21] C. Messmer et al., IEEE J. Photovolt. 8 (2018) 456. DOI: 10.1109/JPHOTOV.2018.2793762
[22] C.R. Wronski and D.R. Carlson, Sol. State Comm. 23 (1977) 421. DOI: 10.1016/0038-1098(77)90999-1
[23] C. Battaglia et al., Nano Lett. 14 (2014) 967. DOI: 10.1021/nl404389u
[24] K.-U. Ritzau et al., Sol. Energy Mater. Sol. Cells 131 (2014) 9. DOI: 10.1016/j.solmat.2014.06.026
[25] J. Bullock et al., Appl. Phys. Lett. 105 (2014) 232109. DOI: 10.1063/1.4903467
[26] L.G. Gerling et al., Sol. Energy Mater. Sol. Cells 145 (2016) 109. DOI: 10.1016/j.solmat.2015.08.028
[27] M. Mews et al., Sol. Energy Mater. Sol. Cells 158 (2016) 77. DOI: 10.1016/j.solmat.2016.05.042
[28] T. Matsui et al., Energy Procedia 124 (2017) 628. DOI: 10.1016/j.egypro.2017.09.093
[29] R. Brendel and R. Peibst, IEEE J. Photovolt. 6 (2016) 1413. DOI: 10.1109/JPHOTOV.2016.2598267
[30] U. Würfel et al., IEEE J. Photovolt. 5 (2015) 461. DOI: 10.1109/JPHOTOV.2014.2363550
[31] J. Melskens et al., IEEE J. Photovolt. 8 (2018) 373. DOI: 10.1109/JPHOTOV.2018.2797106
[32] R.M. Swanson, IEEE PVSC Conf Proc. (2005) pp. 889-894. DOI: 10.1109/PVSC.2005.1488274
[33] J. Zhao et al., Appl. Phys. Lett. 73 (1998) 1991. DOI: 10.1063/1.122345
[34] R.J. Schwartz and M.D. Lammert, IEDM Tech. Dig. (1975) 350. DOI: 10.1109/IEDM.1975.188896
[35] R.M. Swanson, IEDM Tech. Dig. (1978) 70. DOI: 10.1109/IEDM.1978.189354
[36] D.K. Schroder and D.L. Meier, IEEE Trans. Electron Devices 31 (1984) 637. DOI: 10.1109/T-ED.1984.21583
[37] T.G. Allen et al., Prog. Photovolt. Res. Appl. (2017). DOI: 10.1002/pip.2838
[38] T.G. Allen et al., Adv. Energy Mater. 7 (2017) 1602606. DOI: 10.1002/aenm.201602606
[39] J. Bullock et al., Adv. Energy Mater. 6 (2016) 1600241. DOI: 10.1002/aenm.201600241
[40] J. Bullock et al., Nature Energy 1 (2016) 15031. DOI: 10.1038/nenergy.2015.31
[41] Y. Wan et al., ACS Appl. Mater. Interfaces 8 (2016) 14671. DOI: 10.1021/acsami.6b03599
[42] Y. Wan et al., Appl. Phys. Lett. 109 (2016) 113901. DOI: 10.1063/1.4962960
[43] Y. Wan et al., Adv. Energy Mater. 7 (2016) 1601863. DOI: 10.1002/aenm.201601863
[44] A. Cuevas et al., Sol. Energy Mater. Sol. Cells 184 (2018) 38. DOI: 10.1016/j.solmat.2018.04.026
[45] U. Würfel et al., Adv. Energy Mater. 6 (2016) 1600594. DOI: 10.1002/aenm.201600594
