Transparent-Conductive-Oxide-Free Front Contacts for High E ciency Silicon Heterojunction Solar Cells

In order to compensate the insufficient conductance of heterojunction thin films, transparent conductive oxides (TCO) have been used for decades in both-sides contacted crystalline silicon heterojunction (SHJ) solar cells to provide lateral conduction for efficient carrier collection. In this work, we substitute the TCO layers by utilizing the lateral conduction of c-Si absorber, thereby enabling a TCO-free design. A series resistance of 0.32 Ωcm and a fill factor of 80.7% were measured for a TCO-free back-junction SHJ solar cell with a conventional finger pitch of 1.8 mm, thereby proving that relying on lateral conduction in the c-Si bulk is compatible with low series resistances. Achieving high efficiencies in SHJ solar cells with TCO-free front contacts requires suppressing deterioration of the passivation quality induced by direct metal-a-Si:H contacts and in-diffusion of metal into the a-Si:H layer. We show that an ozone treatment at the a-Si:H/metal interface suppresses the metal diffusion and improves the passivation without increasing the


Introduction
One of the fundamental challenges in the design of both sides contacted crystalline Si solar cells is the development of a contacting scheme that combines a highly conductive front contact with negligible parasitic absorption losses. The solution to this challenge involves the combination of highly conductive, opaque and locally applied materials (a metal grid) with slightly less conductive, fairly transparent layers that fully cover the front surface of the solar cell and facilitate lateral conduction to the nearest grid finger. For high temperature diffused junction solar cells, the conductive layer is the heavily doped emitter, which shows a sheet resistance typically ranging from 40 Ω/□ to 100 Ω/□ 1 . Optimization of the diffused emitter has dealt with the trade-off between contact resistivity with metal grids and Auger recombination (increases quadratically with doping density) 2 . This compromise leads to fill factor (FF) losses for lower doping densities and opencircuit voltage (Voc) losses for higher doping densities. The selective emitter reliefs this trade-off by reducing the local contact resistivity at a cost of minor increase in recombination 3 .
Silicon heterojunction (SHJ) solar cells completely avoid the highly doped diffused emitter regions by using chemical vapour deposited amorphous Si layers to form charge-selective junctions at both sides of the wafer. However, given that the amorphous contact layers are very thin and have comparably poor lateral conductivities, additional transparent conductive oxide layers have always been used since the development of SHJ solar cells featuring doped a-Si:H/intrinsic a-Si:H layer stacks in 1991 4 . While the TCO ensures sufficient lateral conductivity, it has a good but not perfect transmission that leads to losses in short circuit current density (Jsc).
In addition, the most commonly used TCO is indium tin oxide (ITO) which is economically problematic because of the high price and supply risk of indium [5][6][7] . Given that the reported highest open-circuit voltage is 727.6 mV for a diffused junction solar cell (with a very low doping density in the emitter resulting in a sheet resistance of 300 Ω/□) 8 , open-circuit voltages of >750 mV are only possible with silicon heterojunction solar cells 9,10 . Therefore, the future of crystalline Si photovoltaics would benefit greatly from any solution to circumvent the parasitic absorption in the TCOs and the use of indium.
With the recent development of back side emitter SHJ solar cells [11][12][13] , the opportunity opens up to omit at least the front side TCO as shown schematically in Fig. 1. In the design shown in Fig.   1a (front side emitter) 4,14 , holes would have to diffuse in the n-type wafer to the interface with the ITO from where efficient hole transport would happen to the metal finger. Removing the ITO in such a design would require the metal grid spacing to be much less than the diffusion length of holes in the n-type wafer. If, however, the junction is at the back as shown in Fig. 1b, the holes as minority carriers would diffuse to the back contact which is fully metallized. The electrons would drift to the front contact, where they would be collected also without the use of an ITO layer (Fig.   1c), because they are the majority carriers in the n-type wafer. Ideally, the scenario shown in Fig.   1c should enable efficient silicon solar cells with the potential downside of added resistive losses due to the lateral conduction of electrons in the n-type wafer as opposed to the ITO as shown in where d is the wafer thickness, q is the elementary charge, µ is the mobility and n is the carrier density of the respective carriers (ne= nd + Δn while nh ≈ Δn under illumination). The low sheet resistance of 46 Ω/□ for electrons at Nd = 4.5×10 15 cm -3 (1.09 Ωcm) suggests good lateral conductance of majority carriers, while the sheet resistance of holes is higher than 400 Ω/□ (Fig.   2c). While Fig. 2 suggests that there is a potential to make efficient SHJ solar cells without a TCO at the front side, the necessity for efficient photon reflection at the back contact still requires a back side TCO. This is due to the fact that TCO/metal combinations are substantially better mirrors than bare metal contacts.  Here, we show the feasibility of a TCO-free design at the front contact of a SHJ solar cell showing efficiencies > 22%. These high efficiencies are enabled by the excellent majority carrier conductivity in the c-Si absorber under illumination, leading to a low series resistance of 0.32 Ωcm 2 and a fill factor of 80.7%. The conduction of the above-mentioned SHJ solar cell is then investigated by electroluminescence (EL) imaging and illumination variation. As the TCO layer is removed, we investigated the direct metal contacts with a-Si:H layer stacks to avoid the passivation degradation. At last, the future potential of removing front side TCO for high efficiency SHJ solar cell is discussed. On the other hand, the short-circuit current density of both sides TCO-free is lower than that of front side TCO-free solar cell, which is due to the low back reflection of the rear metal surface.

Proof-of-concept TCO-free SHJ solar cells
More details about the optical spectra are shown in Fig. S4. The electroluminescence image of a front side TCO-free solar cell is shown in Fig. 4a. The 7 image shows homogeneous luminescence intensity between fingers with a pitch of 1.8 mm. The local diode voltage mapping is extracted from the electroluminescence images ( Fig. 4b) 15 , which shows a homogeneous diode voltage distribution with a fluctuation of about 1 mV. No obvious diode voltage change between the fingers is observed indicating a good lateral conductance without TCO [16][17][18] . is not measurable, while the experimental series resistance was extracted from dark-light J-V method 19 .
In Fig. 4c  simulation in Fig. S3 shows a similar trend as the experimental results. The finding of light intensity dependent series resistances indicates that even in a front junction solar cell, the absorber conduction takes part in the front side lateral collection. This is verified in the carrier distribution simulation shown in Fig. S2. Fig. 4d shows the fit of simulation results to the experimental results for rear junction SHJ solar cells without front TCO, which indicates that the variation of series resistance increases as the bulk resistivity increases.

Direct metal/a-Si:H contacts
When TCO is removed from SHJ solar cells, the metal grids form a direct contact with a-Si:H thin film stacks. To enable a high performance TCO-free SHJ solar cell, two issues have to be solved for the metal/a-Si:H direct contacts. The first issue is that the contacts must be Ohmic contacts, which is realized by using titanium as the contact metal with a-Si:H stacks in this work  Fig. 3. The best open-circuit voltage is 715.9 mV with 60 nm a-Si:H (p + )/metal contact in Fig.   3f, which is comparable to 718.2 mV with rear TCO/metal contact in Fig. 3a.
In order to prevent the metal diffusion, a thin SiOx layer (1.6 nm on polished silicon wafer by spectroscopic ellipsometer) was added to the metal and a-Si:H interface. To this end, ozone oxidation was performed on the surface of a-Si:H layers before the fabrication of metal stacks. The STEM images are shown in Fig. 5b and 5d for comparison. Compared to Fig. 5a, the diffusion is prominently inhibited by the ozone oxidation in the interface, as shown in Fig. 5b. A clear interface between the Ti and a-Si:H layer is shown in Fig. 5d. There is still a small fraction of diffusion as shown in Fig. 5d. However, the diffusion area is much smaller and the penetrating depth is less than 10 nm. Although the ozone oxidation provides a good barrier for the metal, the thin SiOx was not clearly seen in the Ti/a-Si:H interface, as shown in Fig. 5e. On the other hand, secondary ion mass spectrometry measurement shows extra oxygen signal at the Ti/a-Si:H interface (Fig. S7). It is reported that titanium reacts with the silicon oxide during evaporation process 23 and forms contacts with good adhesion. The contact resistivity between metal stacks and doped a-Si:H layers shows no prominent difference with and without ozone treatment (Fig. S6), which may be also due to the Ti-SiOx reaction.

FIG. 5 | STEM images of the metal and a-Si:H interfaces. a-b, Dark field (DF) image of metal/a-Si
The front side TCO-free SHJ solar cell performance with and without ozone oxidation are shown in Fig. 6a.

Potential of substituting TCO layer by lateral c-Si bulk conduction
Although TCO is removed and the lateral conduction between fingers relies only on the c-Si absorber, high-quality carrier collection is achieved with a small series resistance of 0.32 Ωcm 2 and a high fill factor of 80.7% in completely TCO-free SHJ solar cells. To compare the lateral conduction of TCO-free SHJ solar cell and conventional SHJ solar cell with TCOs, Quokka3 [24][25][26] was used to simulate the solar cell performances and the results are shown in Fig. 7a Fig. S9).
On the other hand, for the minor cost of 0.04 Ωcm 2 in series resistance, the parasitic absorption caused by the front side TCO is removed. Optical simulation by OPAL2 in Fig. 7b shows that there is a reduced parasitic absorption of 0.88 mA/cm 2 and a reduced reflection of 0.21 mA/cm 2 by replacing ITO by SiNx as antireflection coatings. This result fits well to the experimental results in Fig. 6b. It also suggests that when the parasitic absorption loss caused by ITO is removed, the remaining parasitic absorption loss is mainly caused by a-Si:H thin film. When the TCO layer is removed, the sputter damage 27 can also be avoided in TCO-free SHJ solar cells. However, the metal/a-Si:H contact introduces damage to the passivation, which can be solved by inhibiting the metal diffusion into the c-Si surface. In this work, an open-circuit voltage of 731 mV was achieved by applying ozone oxidation at the metal/a-Si:H interface without increasing the contact resistivity. Table 1 shows the comparison between SHJ solar cells with and without TCO. By substituting TCO layers with lateral bulk collection we are able to 1) solve the rely on indium in SHJ solar cell and 2) avoid the dilemma of transparency and conductance in TCO layers. In this work two tasks 12 are achieved: 1) low series resistance without TCO, and 2) good metal/a-Si:H contact by an oxidation treatment at the interface. Further development that contributes to TCO-free SHJ solar cells could be replacing rear side TCO by proper back reflection design 28,29 and developing more transparent front-side heterojunction materials such as MoOx 30 , nc-SiOx 31 and µc-SiC 32,33 .

Conclusions
In the present work, we demonstrate a SHJ solar cell design substituting the TCO layer by lateral conduction of c-Si absorber. Using only the c-Si bulk for lateral conduction, excellent lateral In Quokka3 simulation models, n-type silicon wafers with a resistivity of 1.09 Ωcm and a thickness of 170 µm were used. The pitch of fingers was set to be 1.8 mm which is the same as the experiment. To simplify the simulation, the tunneling transport in the interfaces was not considered in the device. Also, the work function mismatch in the interfaces is omitted and ohmic contact is assumed for all the contacts.
The current density-voltage measurement, electroluminescence image and external quantum efficiency are performed using "LOANA" characterization setup from pv-tools equipped with a SINUS 220 Wavelabs light source. In this work most of the series resistances are extracted by comparing J-V curves under different illumination intensities except for Fig. 4. In Fig. 4 we compare the series resistances under different illumination intensities, so it is important that the series resistances are extracted under that particular illumination intensity. Therefore, we extracted the series resistance in Fig. 4 by comparing the J-V curve in the dark and under illumination. Figure 1 Illustration of different SHJ solar cell structures and the path for charge carriers to electrodes. a, Sketch of conventional SHJ solar cell structure with a front emitter. b, Conventional SHJ solar cell structure with a rear emitter. c, Rear emitter SHJ solar cells using only the absorber for lateral conduction. SiNx layers are used in the present work as anti-re ection coatings (ARC).    Conduction properties of SHJ solar cells without TCO front contact. a, EL image of a front side TCO-free SHJ solar cell. b, Local voltage distribution extracted from EL images. c, Measured series resistance versus illumination intensity for different SHJ solar cells. d, Simulated series resistance versus illumination intensity for rear junction TCO-free SHJ solar cells with different wafer resistivities. The black dots in Fig. 4d is the same data as the black dots in Fig. 4c. The simulated series resistance was calculated from the integral of the resistive losses in the model, which is not measurable, while the experimental series resistance was extracted from dark-light J-V method19.  Characterization of front side TCO-free solar cells. a, Comparison between J-V performance with and without O3 oxidation as metal diffusion barrier. b, External quantum e ciency comparison between frontside TCO-free solar cells and standard SHJ solar cells with front TCO. It should be noted that the a-Si:H(n) layer thickness is 15 nm for g. 6a and 7.5 nm for g. 6b. An improved external quantum e ciency is shown by removing TCO layer in the front side, giving rise to an increase of 1 mA/cm2 in current density.

Figure 7
Comparison for SHJ solar cells with and without front-side TCO. a, Simulated performance of rear junction SHJ solar cell with and without front TCO. Solar cell performances with different TCO sheet resistance in the front surface are shown in dots and solid lines. The solar cell performances without front side TCO are shown by the dash line. The optical performance, surface passivation qualities and metal contact resistivities were kept the same with and without TCO to focus only on the lateral electrical transport properties. b, Simulated absorption and 1-re ection spectra using ITO and SiNx as antire ection coating, respectively. By replacing ITO with SiNx, the absorption loss is reduced by 0.88 mA/cm2 and the re ection loss is reduced by 0.21 mA/cm2.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryinformationTCOfreeSHJsolarcell.pdf