Selective contacts and fill factor limitations in heterojunction solar cells

Crystalline silicon‐based heterojunction (HJ) solar cells are becoming the best choice for manufacturing companies, because of the low temperature processes useful for very thin silicon wafers and the possibility to easily achieve cells efficiencies higher than 22% on n‐type silicon wafers. However, the maximum cell efficiency is still limited by the typical Fill Factor (FF) value of 82%. This issue is due to several factors, some of which are sometimes underestimated, like the base contact. Indeed, a potential mismatch between the work functions of the transparent conductive oxide and the base doped layer can give rise to a small barrier against electrons collection, which is not easy to recognize when the cell FF overcomes 80%. Also a low doping efficiency of the p‐type amorphous layer at the emitter side can negatively affect the FF. In this case, even if high efficiency cells are produced, their full potential is still unexploited. Thus, both selective contacts of the cell, even if apparently optimized to achieve very good results, can hide problems that limit the final cell FF and efficiency. In a previous work, an experimental method and a model to individuate hidden barriers at the base contact on n‐type crystalline silicon‐based HJs have been provided. In this paper, that model is applied to experimental data obtained from the characterization of both commercial and laboratory level HJ solar cells. Moreover, an easy method to recognize the presence of a barrier to the charge transport at the emitter side of the cell is illustrated.

European Renewable Energy market) project successfully converted the 3SUN production line from amorphous/microcrystalline silicon tandem thin film modules to HJ solar cells with a maximum efficiency of 25% 7 (22.4% on average) and a current annual capacity of 200 MWp that hopefully will be scaled up to 2 GWp.
Research from institutes and equipment producers is currently ongoing and continuously improving the cell conversion efficiency on pilot lines. 8 Each production step is still being optimized, and new solutions and architectures are being tested, such as using half-cells to enhance the achievable power by obtaining a higher module Fill Factor (FF). 9 This latter parameter, at cell level, is actually one of the main limiting factors for HJs. At present, typical reported FF values for HJs are indeed in the range 82-83.5%, 2 where the latter value is related to the actual record HJ cell with 25.1% efficiency. 10 Martin Green obtained from an empirical expression a theoretical upper limit of 89% for silicon solar cells FF, only limited by Auger recombination. 11 However, this is not the only effect limiting the FF, as will be discussed in the following paragraphs.
In Figure 1, a typical bifacial HJ solar cell is depicted (top), and its energy band structure (as deduced from numerical simulations of the record HJ cell 10 ) is sketched (bottom). The structure is based on a conventionally textured, n-type-doped c-Si (n-c-Si) wafer, excellently passivated on both surfaces by an intrinsic a-Si:H thin film and provided with selective contacts obtained by two doped a-Si:H films, namely (p) a-Si:H and (n) a-Si:H for the emitter and base contact, respectively.
As specifically described elsewhere, 12  contacts are ensured by films of a TCO and a screen printed silver grid on both sides, to obtain a bifacial solar cell. In this structure, the main FF limitations come from series resistance, arising from TCO lateral transport, by the specific contact resistivity between the TCO and the silver screen printed grid, and of course by the metal grid conductivity, which is usually one order of magnitude lower than that of the conventionally achieved in c-Si homojunctions, due to temperature sintering below 200 C of the screen printable silver pastes. 13 Besides these issues, other parameters can limit the HJ FF, and they have been well addressed one by one in a previous work, 14 showing how a realistic upper FF limit for this technology could be 85%. A key factor for the FF limitation is represented by the presence of barriers to charge transport which are not easily identified at Room Temperature (RT).
Nevertheless, their presence can be evidenced by measuring current density/voltage (J-V) characteristics at low temperature, when a deviation from the ideal junction behavior can be observed. Such barriers can be due to different mechanisms, not always fully considered during the optimization and fabrication of devices, like energy bandgap misalignment at one or both selective contacts. Indeed, concerning the base contact, in principle, the ohmic contact on n-type c-Si is easy to produce, because of a low barrier between n-c-Si and (n) a-Si:H at the conduction band edge. 12 However, the work function of the TCO deposited above this doped layer is not always carefully considered, even because at RT it is difficult to immediately recognize any barrier issue when the cells FFs range around 80%. Considering the energy band structure of Figure 1, it can be observed that the band bending at the (n) a-Si:H/TCO interface produces an edge which induces a depletion up to the edge of c-Si wafer and then the electron flow from the base contact to the electrode experiences an undesired obstacle. This kind of bending is clearly dependent on the TCO workfuntion (Φ TCO ): the higher the Φ TCO , the higher the mismatch with the (n) a-Si:H Fermi level, and the higher the barrier to electron collection.
In a previous work, 14 it has been shown how the optimal TCO work function on the base contact must range below 4.25 eV in order to avoid, at RT, any FF limitation coming from the presence of the barrier. We have also illustrated that a high doping of the a-Si:H layer can help in reducing the impact of a higher Φ TCO on the FF. We have finally provided a method to reveal the presence of a hidden barrier at the base contact, by measuring the J-V characteristics of the n-c-Si/ a-Si:H/(n) a-Si:H/TCO base contact alone as a function of temperature: if a limitation to the electron transport is present, the J-V characteristic of the base contact shifts from a linear (typical of an ohmic contact) to a nonlinear (similar to a Schottky barrier) curve with decreasing temperature. The analysis of the measured curves permits to define an activation energy (E act ) for this transition. Then we have F I G U R E 1 Sketch of a bifacial a-Si:H/c-Si heterojunction solar cell (top) and its band structure under dark, short circuit, and room temperature conditions (bottom) [Colour figure can be viewed at wileyonlinelibrary.com] extrapolated a theoretical curve, 14 reported as the solid line in Figure 3, which correlates E act to Φ TCO .
Several studies have been presented in the literature on the FF limitations due to the cell emitter side 12,15,16 but very few on the base contact. 17,18 We hereby carefully consider both selective contacts, their influence on the cell performances and a way to recognize if a limitation is introduced by none, one, or both of them. We detail the experimental procedure used to verify the barrier presence and apply the model to experimental data obtained from different HJ solar cells, both commercial and laboratory level.

| NUMERICAL SIMULATIONS
The models presented in this work have been obtained from numerical simulations of solar cells J-V characteristics at different temperatures and with different characteristics of the constituting materials, similarly to what already discussed in a previous work. 14  no variation of the free carriers imposed. 19 To simplify the description, the density of states at the interface between the a-Si:H buffer layer and c-Si is integrated along the thickness of the buffer layer and then reported as D it . More details can be found in Martini et al. 14 In this work, the same numerical simulator has been used to describe the emitter side behavior as a function of temperature for different doping concentration of the (p) a-Si:H layer. The most relevant properties of any material used in the device simulations, such as E g , μ, χ, optical absorptions and refractive indexes are deduced from experimental measurements and are listed in Table 1.

| Method to investigate the base contact
As already mentioned in the introduction and described in detail elsewhere, 14 it is possible to evidence the presence of a barrier at the base contact of a HJ by analyzing several J-V dark characteristics measured at different temperatures. Figure 2 shows the curve bundle relative to the base contact of two HJ cells. It can be seen that at RT the characteristic is linear, which is typical of an ohmic contact. As the temperature lowers, however, the characteristic turns to a nonlinear curve, and the effect becomes more and more pronounced with the decrease of temperature, revealing a temperature-activated process for which an activation energy (E act ) can be individuated. As explained, this effect is the evidence of a barrier at the electron selective contact, due to a nonideal band alignment between n-doped a-Si:H and TCO. At RT, the presence of this barrier is almost "hidden" in cells that have an overall FF above 80%.
To derive the activation energy of the transition from nonlinear to linear behavior, we have collected the current density values for each curve at a specific voltage, in the moderate forward bias range.
The chosen voltage corresponds to the normal working condition of a HJ solar cell at RT, when the current density is around 37 mA/cm 2 .
The Arrhenius plot of the data (insets in Figure 2) allows extracting the E act value from the slope of the linear fit of the high-temperature values.   Table 2. More details can be found elsewhere. 20 The total thickness of the emitter stack (intrinsic + doped amorphous silicon) is 25-30 nm, while for the base contact side the total thickness is 8-12 nm.

| Electron-selective base contact
The J-V characteristics of the back contact of the cells listed in Table 2 have been measured as a function of temperature, and the activation energy for each cell has been evaluated as explained in it is very difficult to evaluate the correct Φ value of air-exposed ITO.
It is well-known that environmental contamination can cause an alteration of the ITO work function during the UPS measurements, due to the effects of the ultraviolet (UV) radiation itself. 24 Surface cleaning before the measurement does not allow a correct estimation of Φ ITO , either. 25 Capacitance-Voltage (C-V) measurement is also an indirect evaluation of the TCO workfunction, 26 but it is sometimes too sensitive to the defect density at the interface between the TCO and the silicon substrate. Furthermore, the TCO film growth is influenced by several parameters, like the substrate nature and morphology 27,28 or the layer thickness, 29 Table 2 Sample V oc (mV) J sc (mA/cm 2 ) FF (%) Eff (%)  Figure 4. Comparing the base contacts of these two cells, we can see that their E act differs only by 11.5 meV, which corresponds to a difference in Φ TCO of 0.12 eV, with the cell A being below the 26 meV threshold reported in Figure 3.  31 Instead, in a HJ cell, the dark characteristic crosses the light one, as clearly evident from experimental measurements and also confirmed by simulations as reported in Figure 5A and 5B, respectively.
From Figure 5A, it can be seen that the crossing point takes place at different forward bias voltages for cells D and E. The simulations in Figure 5B reveal that such voltage is strongly dependent on the doping activation energy (E p-act ) in the emitter a-Si:H layer, while the corresponding FF values vary between 79% and 85.5%, in accordance with the trend already simulated and reported in reference. 14 Consequently, the lower the E p-act is, the higher the crossing voltage and the shown in Figure 6A. At low temperatures, a marked S-shape is evident, which strongly affects the cell FF, while around 240 K a transition to a single diode behavior is seen. Around RT the cell does not show any S-shape. It can be noticed that the cell V oc at 300 K is lower that the value reported in Table 3 and Figure 4, as an effect of the missing passivating contact on the back. F I G U R E 7 Normalized fill factor values at different temperatures of the J-V characteristics reported in Figure 6: the symbols are experimental values referred to cell D ( Figure 6A), while the blue line is obtained from the simulated curves of Figure 6C. The red line is a guide for the eye. Both experimental and simulated data show similar trends [Colour figure can be viewed at wileyonlinelibrary.com] a maximum around 240 K, which can be seen as the transition temperature toward the S-shape in the J-V characteristics. A structure similar to the one measured in Figure 6A can be simulated as an HJ cell having the base contact modeled as a recombining ohmic contact with InGa. Figure 6B and 6C show the simulated light J-V characteristics at different temperatures for two different E p-act values of 0.6 and 0.4 eV, respectively. The simulations reported in Figure 6B and 6C are quite similar but do not exactly fit the experimental data of Figure 6A.
However, they suggest that the J-V bundle lies in different voltage ranges depending on the E p-act , while the S-shape appearance in the J-V curve bundle is due to the valence band offset. Therefore, the numerical simulations are useful to remark that the band offset and the E p-act contribute in different ways to the cell FF issue related to the emitter side.
The blue curve in Figure 7 is derived from the evaluation of the FF of the simulated curve bundle in Figure 6C. The simulated FF trend is quite similar to the experimental one; therefore, it can be deduced that the S-shape seen in the J-V characteristics bundle is only due to the valence band offset, no matter the E p-act value. It is worth noticing that in all simulations a very low D it (10 9 cm −2 ) at the heterointerface is chosen, in order to exclude the influence of this parameter from the discussion.
As ΔEv depends on the difference between the energy gap values E g of c-Si and (p) a-Si:H, and since with decreasing temperature there is a higher enhancement of E g in the a-Si:H film with respect to the one in the c-Si wafer, 32,33 the corresponding ΔEv varies with temperature as shown in Figure 8. Consequently, below 240 K, when ΔEv increases, the electric field is not sufficient to ensure a complete thermionic emission of holes over that band offset, so that the Sshape appears and the FF is reduced. From the simulated curves in Figure 6B and 6C it is seen that a lower E p-act corresponds to the J-V characteristic bundle shifted to higher voltages, even if the S-shape does not disappear, thus confirming that the barrier is still present even when enhancing the built-in voltage of the cell. These simulations indirectly confirm the hypothesis that the electron affinities of a-Si:H and c-Si negligibly depend on the temperature, letting the valence band offset be the cell parameter which is most affected by the working temperature of the device. Finally, it is relevant to remark that the above-described low temperature effect on HJ cells FF does not limit the use of the cells at RT or in the temperature range of common use for HJ solar panels.

| CONCLUSIONS
This work, which can be considered as the extension of a previous more theoretical and detailed one, 14  Regarding the base contact, it is remarked that the work function of the TCO must be carefully optimized in order to obtain a perfect ohmic contact at RT. To this aim, a method to indirectly measure the actual Φ TCO value at the (n) a-Si:H/TCO interface is proposed, which consists of calculating the activation energy E act of the hidden barrier and placing the obtained value on the theoretical curve already shown in a previous work. 14 Even though the method requires measuring the J-V characteristics as a function of temperature, which is not an easy task for large area cells, however it is valid even in a restricted range of temperatures, that is, between 330 and 240 K, which can be easily reached at both laboratory and production line scale even without a specific large area cryogenic system.
FF limitations are also recognized on the emitter contacts, which are evidenced by two features observed on the J-V characteristics.
The first is the crossing of the light and dark curves at RT, and it is shown that the higher the voltage at which the crossing happens, the lower the barrier experienced by the carriers. The second is the appearance of a S-shape in the light curves at low temperatures, and it is shown that this effect is only due to the valence band energy offset at the emitter and determines the FF and V oc values of the HJ cell.
The effect can be mitigated by increasing the doping level in the