The sputter deposition of broadband transparent and highly conductive cerium and hydrogen co‐doped indium oxide and its transfer to silicon heterojunction solar cells

Indium oxide doped with tin (ITO) is the most commonly used material for lateral transport window layers in silicon heterojunction (SHJ) solar cells, as it currently offers the best combination of physical properties, industrial deposition capability, and module reliability. However, typically applied ITO layers by far do not exploit the full electro‐optical potential of the indium oxide material class, resulting in optical and electrical losses limiting the solar cell efficiency. In this work, cerium and hydrogen co‐doped indium oxide thin films are developed for their application as high‐performance transparent conductive oxide layers in SHJ devices. Amorphous In2O3:Ce,H layers are fabricated via radio frequency magnetron sputtering, before being crystallized during post‐deposition thermal treatments compatible with the SHJ temperature stability. The resulting excellent electro‐optical film properties are on par with values so far solely reported for reactive plasma deposited films. It is shown that the surface morphology of the substrate (planar or textured) has a strong impact on the film properties, and further, the critical role of the atmosphere present during post‐deposition annealing is elucidated. Finally, the large potential of an optimally processed In2O3:Ce,H window layer in SHJ cells is demonstrated, quantified by a gain in short circuit current density of 0.6 mA/cm2 without impairing the resistive losses in comparison to the usage of a baseline ITO layer.

hydrogen co-doped indium oxide thin films are developed for their application as high-performance transparent conductive oxide layers in SHJ devices. Amorphous technologies, tin-doped indium oxide (ITO) is still the predominantly used material for this purpose both in research labs and in SHJ production lines. Even though there is currently no fundamental shortage of indium, its comparatively uncertain supply and rather high cost strive for a reduction of the latter in SHJ cells. There exist several indium-free alternatives. Recently, the replacement of ITO by low-cost aluminum-doped zinc oxide (AZO) on the cell's rear side 1 or on both sides, 2,3 taking advantage of the wafers contribution to the lateral current transport in the case of a rear emitter cell design, 4 showed promising results, even though capping layers might be required to solve the module stability issue of zinc oxide-based TCOs. 2 Exploiting the superb electro-optical properties inherent for the indium oxide material class, another approach to reduce the indium consumption of an SHJ cell can be the reduction of the film's thickness. [5][6][7] Further, the potentially low film resistivity enables the application of a metal grid with comparatively large finger spacing, reducing the amount of Ag, while further improving front contact transparency. However, indium oxide-based materials with properties outperforming conventional ITO have been discovered and extensively studied in the past years. Here, the usage of other metal dopants (e.g., Zr, W, Ti, and Mo 5,[8][9][10][11][12] ) than tin and/or the incorporation of hydrogen [13][14][15] resulted in enlarged electron mobility (μ), being an important parameter for the conductivity-transparency trade-off.
In general, enabling low resistivity of TCOs via n-type doping is typically accomplished by (i) a substoichiometric, oxygen-poor deposition leading to donor-type oxygen vacancies within the films; (ii) the substitution of indium cations by extrinsic metal atoms; and/or (iii) the incorporation of hydrogen atoms into the film. While by these means, the electron density in the conduction band (N) can be increased, charged impurities remain, which limit the electron mobility with a related scattering cross section showing a square law dependence on the impurity charge state. 16 Oxygen vacancies in In 2 O 3 are often present as doubly charged impurities [16][17][18] ; hence, using solely them to achieve a substantial doping level is not favorable for maximizing the mobility. In the case of inserting extrinsic metal atoms to act as singly charged donors in In 2 O 3 , several aspects are considered important for a proper selection thereof. The energy level of the respective donor state needs to be high enough to effectively donate electrons to the In 2 O 3 conduction band, while hybridization between both should be avoided in order not to affect the band curvature, which might result in an increased electron's effective mass 11,19 and hence a lowered μ. When incorporated into the In 2 O 3 lattice structure, the effective radius of the ionized dopant should be close to the one of indium in order to avoid microstrain around the dopant site. 20 Further, a large oxygen affinity of the metal cation can help to decrease the density of unfavored oxygen vacancies within the TCO. Moreover, in recent years, several studies showed that the addition of hydrogen during the film growth can support high μ and N within the films simultaneously. This was attributed to the appearance of hydrogen atoms as singly charged donors, 21 their ability to passivate defects at the grain boundaries of polycrystalline films, as well as their beneficial role in the crystallization process of solid phase crystallized (spc) TCOs.
Regarding the latter, hydrogen can promote amorphous film growth such that during a post-deposition thermal treatment, high-quality polycrystalline TCO layers with low strain and large crystal grains can emerge, which allow for an exceptional high electron mobility. 14,21 Kobayashi et al. recently discovered the co-doping of In 2 O 3 with cerium and hydrogen to facilitate μ values of up to 145 cm 2 /Vs when grown via reactive plasma deposition (RPD). 20 Similarly, Koida et al. extensively characterized the microstructure and bulk properties of high μ RPD In 2 O 3 :Ce,H layers when deposited polycrystalline or after crystallization from the amorphous phase. 22 However, an equivalent film quality for this material deposited via magnetron sputtering has not been reported yet. 23 Hence, in this work, the electro-optical properties of sputter-deposited and spc In 2 O 3 :Ce,H are studied dependent on the cerium, oxygen, and hydrogen content within the films. Further, the activation energy/temperature budget required for crystallization is attached importance in order to be compatible with the SHJ thermal stability.
While there are various reports on the successful fabrication of high-mobility TCOs, in several cases, it was difficult for those films to outperform ITO when applied in SHJ cells. Reasons for this were, for instance, a too low electron density in the TCO, 5,24 being problematic for the contact formation to the adjacent a-Si:H layer, 25 or an excessive hydrogen concentration, which can degrade the contact to the metal finger during the sintering of the metal paste. 26 Another issue is that some reported high-mobility TCO results are achieved on planar substrates like glass or (SiO 2 -coated) Si for an unsophisticated TCO characterization. However, the deposition on textured surfaces and/or hydrogen-containing a-Si:H layers present in SHJ devices can lead to different TCO bulk properties. [27][28][29] Thus, we investigated the influence of the substrate type on the TCO properties and, further, the influence of the annealing atmosphere (with or without oxygen).
Eventually, as a proof of principle, we applied the optimized In 2 O 3 : Ce,H films on SHJ devices to observe the potential of this material contrasted to the ITO baseline layer.

| EXPERIMENTAL DETAILS
Approximately 105 nm thick In 2 O 3 :Ce,H films were deposited without intentional substrate heating (T substrate < 50 C) via radio frequency (RF) magnetron sputtering onto planar substrates (glass and SiO 2coated Si) as well as onto SHJ precursors with random upright pyramid surface texture. For the SHJ substrates, this leads to an effective More information about the SHJ cell fabrication process can be found elsewhere. 30,31 The charge carrier mobility and density of the thin films grown on different substrate types were determined via Hall effect measurements in the van der Pauw configuration. In order to additionally evaluate the temperature dependence of these parameters, measure- showed weak temperature dependence during the vacuum heat up (squares) to 160 C, indicating the degenerate doping of the films. At T > 160 C, a jump in both Hall parameters occurred, which is related to an increase in crystalline fraction within the layers. 32 The mobility F I G U R E 1 Carrier mobility μ (a) and density N (b) as a function of specimen temperature for In 2 O 3 :Ce,H films deposited on glass substrates under varying water background pressure. Samples were heated in vacuum from 40 to 250 C before being cooled down again. The step duration at each temperature point amounts to around 30 min before the Hall effect measurement is performed. The individual temperature range where solid phase crystallization occurred is indicated [Colour figure can be viewed at wileyonlinelibrary.com] increased for all films upon crystallization; however, a clear downward trend in the post-crystallization values exists for increasing P H2O > 10 −4 Pa. Concurrently, N decreased during crystallization within the film deposited at P H2O = 1 × 10 −4 Pa, which can be related to the elimination of oxygen vacancies during the film reorganization. 32 For higher P H2O , N went rapidly up just before crystallization.
Here, it was reported from thermal desorption spectroscopy on similar films that an enhanced water effusion occurs before crystallization. 35 Effusion of excess oxygen and the related densification of the amorphous structure possibly increase N. A further exposure to T > 180-200 C decreased the charge carrier density gradually in all crystallized In 2 O 3 :Ce 2wt% ,H films. This is likely due to the effusion of hydrogen 22 otherwise effectively acting as a dopant within the polycrystalline films. While during the cooling process down to room temperature (stars), N declined slightly further, the mobility increased in all films. The inverse proportionality between mobility and temperature demonstrates that the scattering on phonons became a dominant mechanism limiting the electron relaxation time in the crystallized In 2 O 3 :Ce 2wt% ,H films. More detailed investigations on the transport-limiting mechanism in such high-quality polycrystalline In 2 O 3 -based TCOs can be found elsewhere. 16 The impact of P H2O on the grain size of In 2 O 3 :H films fabricated under very similar conditions was investigated in a previous study. 21 Besides reducing μ and N, an excess introduction of water vapor moreover increased the activation energy for the crystallization pro- F I G U R E 3 Relation between carrier mobility and density determined electrically via Hall effect measurements (a) and optically (b) from the Drude absorption spectra. In 2 O 3 :Ce,H films deposited under O 2 /Ar ratios equidistantly varied between 0% and 0.5% were compared after crystallization in nitrogen or ambient air. The film deposited with 0% O 2 /Ar still showed substantial amorphous fraction after N 2 annealing, and thus, μ and N could not be extracted reliably from SE data [Colour figure can be viewed at wileyonlinelibrary.com] maxima decreased with Ce content. Here, it is important to note that with an increased percentage of CeO 2 also, the relative number of oxygen atoms in the deposited layer possibly increased due to the large oxygen affinity of Ce and the higher oxygen to metal atomic ratio in the stoichiometric oxide (CeO 2 : 2, In 2 O 3 : 1.5). Owing to the very moderate decrease in mobility and the effective doping, the addition of Ce enabled a very low sheet resistance (R sh ) < 25 Ω/□ for 105 nm thick films after solid phase crystallization. An important fact is that the addition of Ce results in N > 2 × 10 20 cm −3 , which is a critical factor for device integration, in particular for the formation of lowohmic contacts to charge carrier selective layers and to the metal. 5,24,25,39 As shown in Figure S1, the crystallization kinetics was affected by the incorporation of Ce. The onset temperature thereof significantly increased from around 155 C to 163 C when 0.5 wt% CeO 2 was added but then only slightly rose further (to 165 C) for a CeO 2 increment ≥1 wt%. The influence of the oxygen stoichiometry was separately depicted in Figure S2. Here, either a large oxygen excess or In order to study the evolution of the electrical film properties in more detail, the behavior of μ (μ*) and N (N*) is compared for annealing in ambient air and vacuum via temperature-dependent Hall effect measurements for films (2 wt% CeO 2 , 0.13% O 2 /Ar) deposited either on flat glass or on textured SHJ precursors ( Figure 6). Prior to the sputter deposition, the SHJ substrates were exposed to 280 C for 2 h to reduce the impact of hydrogen otherwise later effusing from the a-Si:H into the TCO (this effect will be elucidated afterwards). was observed already at a temperature of around 130 C, indicating the film's enhanced vulnerability for species (e.g., oxygen) penetrating from the ambient. After crystallization, this resulted in a highly resistive film with both low μ* and N*, in accordance to what was observed in Figure 5.
It was reported that the post-deposition crystallization of amorphously grown In 2 O 3 :H-based layers can lead to polycrystalline thin films with very low compressive strain and exceptionally large crystal grains. 22 While this allows for an electron mobility above 100 cm 2 /Vs, the lack of compressive strain was suggested to yield grain boundaries less sealed than in the case of their polycrystalline grown counterparts. One hypothesis is that this becomes particularly visible when deposited on rough surfaces. Similarly, degraded electrical properties of In 2 O 3 :H-based films were reported when grown on rough layers for the application within Cu (In,Ga)Se 2 solar cells. 27 Here, they observed formation of crack-shaped voids inside In 2 O 3 :H-based thin films, particularly in regions where the substrate layers underneath exhibited ridges or valleys. However, no such degradation was found when textured glass was used as substrate. This was explained by the larger feature size of the glass texture, which corresponds to a lower local slope of the height profile, in turn allowing for a sufficiently undisturbed TCO growth. Apart from this, it was shown that also the microstructure of the hydrogenated silicon thin films underneath can considerably influence the TCO film properties. 28 Here, they observed a more irregular initial ITO growth when comparing nanocrystalline (nc) with amorphous Si layers, which was related to the enhanced surface roughness of the nc-Si films.
Besides the surface texture of the non-planar substrates, also the effusion of hydrogen into the TCO during post-deposition thermal treatments was considered important, and the extent thereof can depend on the type of Si thin film underneath (e.g., a-Si:H(p) vs. nc-Si: H(n) 28 ). For our samples, the impact of hydrogen effusing from the a-Si:H layers on the electrical properties of In 2 O 3 :Ce,H films was investigated by comparing SHJ substrates, which prior to this were exposed to 280 C for 2 h, with counterparts not having received this predeposition hydrogen effusion treatment. In Figure 7 Figure S3, the TCO-dependent optical and electrical transport properties are quantified in Figure 9 by the short circuit current density J sc and the difference between pFF and actual FF.   A further next step will be to make use of the low film resistivity and F I G U R E 9 Short circuit current density (a) and gap between pseudo and actual fill factor (b) of SHJ solar cells with selected TCO window layers. The spacing between the fingers of the metal grid was varied. Cells containing In 2 O 3 :Ce,H films of different cerium and oxygen content were annealed either in air or nitrogen and compared to reference cells using ITO windows [Colour figure can be viewed at wileyonlinelibrary.com]