Silicon homo-heterojunction solar cells: A promising candidate to realize high performance more stably

Silicon homo-heterojunction solar cells: A promising candidate to realize high performance more stably Miao Tan,1 Sihua Zhong,1,a Wenjie Wang,1 and Wenzhong Shen1,2,a 1Institute of Solar Energy, and Key Laboratory of Artificial Structures and Quantum Control (Ministry of Education), School of Physics and Astronomy, Shanghai Jiao Tong University, Shanghai 200240, People’s Republic of China 2Collaborative Innovation Center of Advanced Microstructures, Nanjing 210093, People’s Republic of China


I. INTRODUCTION
Amorphous silicon/crystalline silicon (a-Si/c-Si) heterojunction solar cells have attracted considerable attention in the recent decade, mainly due to their small temperature coefficient, simple fabrication process and high open-circuit voltage (V OC ) benefitted from the large difference in Fermi energy of the two materials. [1][2][3][4][5] In order to passivate the c-Si surface, a thin intrinsic hydrogenated amorphous silicon (i-a-Si:H) layer is generally inserted between the doped a-Si layer and the c-Si substrate, forming the heterojunction with intrinsic thin layer (HIT) solar cells. According to the latest report, Kaneka Corp. has realized an efficiency as high as 25.1% on the HIT solar cell 5 and then achieved a record efficiency of 26.6% on a large-area (180 cm 2 ) HIT solar cell by being incorporated with interdigitated back contact technique. 6 Although the HIT solar cells have been demonstrated to be highly efficient, in mass production, it is hard to fabricate the HIT solar cells with excellent performances stably because their performances are strongly affected by various parameters. Lots of investigations have indicated that inappropriate work function of the transparent conductive oxide (TCO) layer will lead to a dramatical degradation of the performances of HIT solar cells, because the band bending in the emitter strongly depends on the work function of the TCO layer and the emitter. [7][8][9][10][11] The HIT solar cell performances are also quite sensitive to the interfacial density of states (DOS). [12][13][14][15][16] Hernandez et al., 16 demonstrated that the cell performances decline significantly when the interfacial DOS exceeds 1×10 11 cm -2 , while the cell efficiency can be improved to nearly 23% if the interfacial DOS is controlled below 5×10 10 cm -2 . Additionally, other parameters, such as the doping concentration of the hydrogenated amorphous silicon (a-Si:H) layer and defect states within it, greatly influence the solar cell performances since they directly determine the Fermi energy position in the a-Si:H layer and thus affect the built-in potential of the solar cell. 1,[17][18][19][20] Therefore, improving the stability of the performances of the heterojunction solar cell while keeping its high efficiency become urgent and are of high significance for mass production.
We have previously presented that the silicon homo-heterojunction (H-H) solar cell, which simultaneously contains homojunction and heterojunction, exhibits better tolerance for the interfacial DOS as compared to the HIT solar cell by AFORS-HET simulation. 21 In addition, it was found that the H-H solar cell shows a much higher fill factor (FF). With these two advantages, even if considering the fact that the interfacial DOS of the H-H solar cell is much higher, for example two magnitudes higher, the H-H solar cell can still realize the efficiency as high as that of the HIT solar cell but has a better stability. It should be noted that Hekmatshoar et al., 22 have already experimentally demonstrated that the H-H solar cell on a p-type c-Si (p-c-Si) substrate is less sensitive to the interfacial DOS and has a higher FF. Besides, it has been reported that the advantages of the H-H structure will still be retained if an i-a-Si:H layer is inserted to reduce the interfacial DOS. 23,24 Nevertheless, up to now, all the previous studies on the H-H solar cell are mainly focused on the sensitivity of the interfacial DOS and it is still unknown how other fabricating parameters affect the cell performances.
In this study, we have investigated the impacts of (i) the work function of TCO layer (WF TCO ), (ii) the total interfacial DOS (D it ) at the front a-Si:H/c-Si interface (hetero-interface), (iii) the peak dangling bond defect densities (N tr ) within the p-type a-Si:H (p-a-Si:H) layer, and (iv) the doping concentration (N a ) of the p-a-Si:H layer on the performances of the H-H solar cell with a structure of p-a-Si:H/p-c-Si/n-type c-Si (n-c-Si) substrate/i-a-Si:H/n-type a-Si:H (n-a-Si:H) by the aid of AFORS-HET software. We have excitingly found that, as compared with the HIT counterparts, the H-H solar cell is much insensitive to all the above parameters. As a result, the H-H solar cell has better performances in the whole range of each above parameter. Furthermore, we identify that the physical origins behind the advantages of the H-H solar cell are the decreased interfacial recombination and the reduced recombination within the a-Si:H region, which are fundamentally ascribed to the field-effect passivation from the homojunction.

II. STRUCTURES OF SOLAR CELLS AND SIMULATION DETAILS
Numerical simulations were carried out by utilizing the simulation software AFORS-HET, which is based on solving the one-dimensional Poisson equation and the continuity equations 7 and has been proven to be an effective and convenient way to analyze the influences of various parameters on the performances of heterojunction solar cells. 12 The simulated HIT cell structure, as depicted in Fig. 1(a), is TCO/p-a-Si:H/i-a-Si:H/n-c-Si substrate/i-a-Si:H/n-a-Si:H/TCO. Fig. 1(b) shows the structure of the H-H solar cell, namely: TCO/p-a-Si:H/p-c-Si/n-c-Si substrate/i-a-Si:H/n-a-Si:H/TCO. A thin ia-Si:H layer is further inserted between the p-a-Si:H layer and p-c-Si layer, labeled as H-I-H solar cell and presented in Fig. 1(c).
In the simulations, parasitic absorption in the 80 nm thick TCO layer was ignored. The back contact was assumed to be flat band in order to neglect the influence of band bending at the rear contact, while the band bending at the front contact formed between TCO and p-a-Si:H layer was carefully taken into account by changing the WF TCO . Fig. 1(d)-(f) shows gap state distributions of p-a-Si:H layer, i-a-Si:H buffer layer and n-a-Si:H BSF layer. The density of localized states in a-Si:H layers is mainly composed of exponential band tail defect states and Gaussian distributed dangling bond states. Recombination centers were modeled with 5×10 11 cm -3 and 1×10 9 cm -3 neutral defects at 0.56 eV above the valence band in the p-c-Si layer and n-c-Si layer, 23 respectively. At the front hetero-interface, defects were introduced using a Gaussian distributed DOS with the maximum at the midgap of c-Si. 17,21,25 Electrons and holes capture cross sections were equal to 1×10 -15 cm 2 . 21 In order to focus on the front side of the solar cells, the back hetero-interface was neglected during the simulations.
Regarding the carriers transport across the hetero-interface, thermionic emission model with tunneling model is adopted. 23 The surface recombination velocities of electrons and holes on both sides were set to 1×10 7 cm·s -1 . More details of the parameters used in the simulations are listed in Table I, which were mainly referred to the previous works. 7,21,26 We varied the value of the (i) WF TCO , (ii) D it at the front hetero-interface, (iii) N tr within the p-a-Si:H layer and (iv) N a of the p-a-Si:H layer to investigate how they affect the performances of the simulated solar cells. Only one parameter was varied at one time while other parameters kept at the initial values. The initial values of the above parameters were set to 5.3 eV, 1×10 12 cm -2 , 1.4×10 19 cm -3 /eV, 7.5×10 18 cm -3 , respectively. Simulation results presented in this study were obtained under AM1.5 solar spectrum with a power density of 100 mW/cm 2 and at 25 • C.

A. Influences of WF TCO and D it on the performances of simulated solar cells
As the first step of our study, we have analyzed the influences of both the WF TCO and D it on the V OC as well as the conversion efficiency (η) of the H-H and HIT solar cells. During the simulations, WF TCO varies in the range of 5.1-5.5 eV while D it varies from 1×10 10 cm -2 to 5×10 12 cm -2 . As depicted in Fig. 2(a)-(c), V OC decreases with increasing D it when WF TCO keeps constant and obvious improvement in V OC is observed with increasing WF TCO when D it keeps constant. From Fig. 2 it is also noted that the dependence of η on WF TCO and D it follows the same trends as V OC. The decrease in V OC with increasing D it can be explained by the enhanced recombination possibility, since the interface defect states work as recombination centers for photo-generated carriers. WF TCO has also been considered as a crucial factor in a-Si/c-Si heterojunction solar cells due to the fact that WF TCO determines the electrical TCO/p-a-Si:H Schottky contact properties and the band bending in the a-Si:H/c-Si junction region. 8,27,28 As shown in Fig. 2(g) and (h), higher WF TCO leads to stronger band bending at both the TCO/a-Si:H contact and the hetero-interface, which is beneficial for photogenerated holes to be collected by the front electrode and forming an effective potential barrier for electrons at the front hetero-interface. As a result, the carrier recombination at the hetero-interface is greatly reduced and hence the V OC increases.
When observing Fig. 2 more carefully, it can be found that both the V OC and η of the H-H solar cell present better tolerance for D it than those of the HIT solar cell, which is consistent with our previous study. 21 In addition, the V OC and η of the H-H solar cell are also less sensitive to WF TCO . Therefore, with the advantage of better tolerance for both the WF TCO and D it , it is not such necessary to adopt fairly rigorous process to optimize the WF TCO and D it in production lines, which is beneficial to produce heterojunction solar cells with more stable performances. Here, we want to clarify that for  the H-H solar cell, even if considering its higher D it than that of the HIT solar cell, it still has a higher or comparable η. For example, when the D it of the H-H and HIT solar cells are 1×10 12 cm -2 and 1×10 10 cm -2 , respectively (namely, the D it of the H-H solar cell is 100 times larger), the η of the H-H solar cell is still a little higher than that of the HIT solar cell, as shown in Fig. 2(i). This advantage is attributed to the better tolerance of D it and the higher FF caused by the absence of i-a-Si:H layer, which has been explained in detail in our previous study. 21 It should be pointed out that Hekmatshoar et al., 22 Fig. 3(a) and (b) at an external voltage of 0.6 V. Here, the values of WF TCO and D it are set as 5.2 eV and 1×10 12 cm -2 , respectively, and it should be noted that the results discussed below will not change even if the external voltage is set at other values. It is evident that the electric field of the H-H solar cell on the c-Si substrate side and at the hetero-interface is higher than that of the HIT solar cell. Hence it is beneficial for holes but detrimental for electrons to reach the hetero-interface, resulting in a higher hole density (p) and a lower electron density (n) at the hetero-interface, as presented in Fig. 3(b). Since carrier recombination simultaneously involves both types of carriers, the larger difference in the n and p will lead to a lower interfacial recombination. This is known as the field-effect passivation.
The field-effect passivation on the interface well explains why the H-H solar cell is less sensitive to D it . 21 Besides, it also adequately explains the better tolerance of WF TCO at a high D it , which is the condition that interfacial recombination dominates the V OC . When the WF TCO decreases, the band bending of the depletion region reduces and the barrier potential for electrons becomes lower. So the n at the hetero-interface of the HIT solar cell is greatly increased, while that of the H-H solar cell is less influenced and keeps a lower value due to the field-effect passivation from the homojunction. As a result, the V OC of the H-H solar cell is higher and shows a better tolerance for the WF TCO . Interestingly, even if D it becomes as low as 1×10 10 cm -2 , the H-H solar cell also performs a superior V OC when the WF TCO is lower than 5.3 eV. Note that, in such a case with a low D it , the recombination at the interface is not the crucial factor to determine the V OC of a solar cell and thus the higher V OC of the H-H solar cell cannot be attributed to the lower n at the hetero-interface as in the high D it condition. This indicates that the lower interfacial recombination is not the only reason for the H-H solar cell to realize the higher V OC with respect to the HIT solar cell.
In order to figure out other reasons, we have modeled the dependence of V OC on the WF TCO for the two types of solar cells without D it (namely D it =0) to totally exclude the influence of the interfacial recombination, as listed in Table II. When WF TCO <5.3 eV, the H-H solar cell exhibits a superior V OC than the HIT counterparts. Moreover, the V OC of the H-H solar cell is less sensitive to WF TCO when WF TCO varies in the range of 5.1-5.3 eV. Through integrating the recombination rate as a function of position, we can obtain the total recombination within the a-Si:H region (R a-Si:H ) of the both types of solar cells (listed in Table II). When the WF TCO is 5.1 eV, the electrons in the a-Si:H region are much less for the H-H solar cell, as shown in Fig. 3(d), because the electric field within the homojunction can strongly drag electrons back into the c-Si substrate. As a consequence, the R a-Si:H of the H-H solar cell (4.97×10 15 cm -2 s -1 ) is nearly an order of magnitude lower than that of the HIT solar cell (1.60×10 16 cm -2 s -1 ) and thus the H-H solar cell possesses a higher V OC in comparison with the HIT solar cell. Besides, for the H-H solar cell, the less variation of the n in the a-Si:H region  Fig. 3(d)) incurs the less variation of the R a-Si:H with WF TCO in the range of 5.1-5.3 eV, which well explains the less sensitivity of V OC to WF TCO . By the above analysis, we can draw a conclusion that the better tolerance of WF TCO for the H-H solar cell can be explained by (i) the less variation of the interfacial recombination at a high D it and (ii) the less variation of the R a-Si:H when the D it is considerably low. Both of them fundamentally benefit from the field-effect passivation within the homojunction.

B. Influences of N tr on the performances of simulated solar cells
Subsequently, the influences of N tr within the p-a-Si:H layer on the performances of the H-H solar cell and the HIT solar cell have been investigated, which is illustrated in Fig. 4. Here, N tr varies from 5×10 18 cm -3 /eV to 2.5×10 19 cm -3 /eV, which is reasonable according to the experimental report. 29 As can be seen from Fig. 4(a), although a higher N tr leads to a lower V OC for both the H-H solar cell and HIT solar cell due to the enhanced carrier recombination within the p-a-Si:H layer, the V OC of the H-H solar cell is much less sensitive to the N tr at either a high D it (D it =1×10 12 cm -2 ) or a low D it (D it =1×10 10 cm -2 ). Hence the V OC of the H-H solar cell keeps a higher value in the whole range of N tr . Similar to the V OC , Fig. 4(b) shows that the FF of the H-H solar cell also keeps a higher value in the whole range of N tr and is less sensitive to N tr when compared to that of the HIT solar cell. As a result, the η of the H-H solar cell is also less sensitive to N tr , as shown in Fig. 4(c). Hence, the η of the H-H solar cell with D it =1×10 12 cm -2 and N tr =2.5×10 19 cm -3 /eV is 1.87% absolutely higher than that of the HIT solar cell with D it =1×10 10 cm -2 and N tr =2.5×10 19 cm -3 /eV.
It is believed that the N tr in the doped a-Si:H layer is easily affected by dopant gas flow rate and/or hydrogen dilution during deposition 29 and the defects in the doped a-Si:H emitter play an important role in the performances of heterojunction devices. 17,26,30 Therefore, accurate control of the dopant gas flow rate and/or hydrogen dilution is required if the performances of the solar cell are sensitive to the N tr . With the advantage of a better tolerance for N tr , the H-H solar cell provides a new opportunity for photovoltaics to fabricate solar cells with stable and excellent performances.
In order to get an insight into the better tolerance of N tr for the H-H solar cell, electron densities (ns) under two different values of N tr have been extracted from the numerical simulation software. When N tr increases from 5×10 18 cm -3 /eV to 2.5×10 19 cm -3 /eV, the active doping concentration of p-a-Si:H layer decreases 17,31 and the Fermi level in p-a-Si:H layer shifts away from the valence band edge. As a result, the built-in potential reduces and more electrons have the ability to reach the heterointerface as well as the a-Si:H region. Consequently, the n at the hetero-interface and within the a-Si:H region of the HIT solar cell is greatly increased. However, the electric field of the homojunction can partly compensate the reduced electric field caused by the increased N tr so that the n of the H-H solar cell is less influenced by the N tr . Hence there is a smaller difference in n within the a-Si:H region and at the hetero-interface for the H-H solar cell (as presented in Fig. 4(d)), which leads to a smaller difference in carrier recombination, and thus the V OC as well as the FF. 10 It should be noted that, an increase in N tr with increasing N a 32,33 is not considered here for simplicity because the influences of N tr on the performances of the H-H and HIT solar cells have been discussed above. Both the V OC and FF are reduced with decreasing N a due to the weaker built-in electric field in the depletion region, 18,34-36 while higher short-circuit current density (J SC ) is observed at a lower N a owing to the diminished parasitic absorption in the p-a-Si:H layer. As an overall result, the η is found to be reduced when N a decreases. In Fig. 5(a)-(d), another remarkable characteristic is that all parameters of the H-H solar cell are much insensitive to the N a and keep higher values in the whole range of N a . It needs to be pointed out that the doping of boron is rather hard to be accurately controlled during the deposition of p-a-Si:H layer, which results in the unstable N a and greatly affects the performances of a-Si/c-Si heterojunction solar cells. 19 With respect to this, the better tolerance of N a for the H-H solar cell is of crucial importance to produce high-performance solar cells stably.

C. Influences of N a on the performances of simulated solar cells
In order to give a clear interpretation on how N a differently impacts the performances of the H-H and HIT solar cells, their energy band diagrams under different N a are compared, shown in Fig. 5(e). When the N a of the p-a-Si:H layer decreases from 1×10 19 cm -3 to 1×10 18 cm -3 , the band bending within the a-Si:H/c-Si junction of the HIT solar cell is greatly reduced, leading to the decreased potential barrier for electrons. Consequently, the n at the hetero-interface of the HIT solar cell is largely increased, as confirmed in Fig. 5(f). In contrast, with the benefit of the homojunction, the band bending within the a-Si:H/c-Si junction of the H-H solar cell is less affected by N a and hence the n of the H-H solar cell is less influenced by N a . As a result, the performances of the H-H solar cell are less sensitive to the N a .
Overall, compared to the HIT counterparts, the reasons why the H-H solar cell shows better tolerance for all the above physical parameters are ascribed to the field-effect passivation from the homojunction, which leads to the less variation of the n, and thus the less affected interfacial recombination and the R a-Si:H . Based on the above results and discussion, it is expected that the concept of H-H structure will also greatly benefit carrier-selective contact solar cells. [37][38][39] For example, the H-H structure can not only provide an excellent field-effect passivation for the hetero-interface but also help to reduce the sensitivity of cell performances to the work function of the carrier-selective contact materials. Hence a higher and more stable efficiency can be realized for the carrier-selective contact solar cells.

IV. CONCLUSIONS
The effects of the WF TCO , D it , N tr within the p-a-Si:H layer and the N a of the p-a-Si:H layer on the performances of the H-H and HIT solar cells were thoroughly compared using AFORS-HET software. An important finding was that both the V OC and η of the H-H solar cell present better tolerance for the D it and WF TCO due to the field-effect passivation from the homojunction. To be specific, the reason why the H-H solar cell is more insensitive to the WF TCO has been identified to the less variation of the interfacial recombination at a high D it and to the less variation of the R a-Si:H when the D it is considerably low. Both of them are owing to the fact that less electrons can reach the hetero-interface and the a-Si:H region under the strong electric field within the homojunction and thus the n of the H-H solar cell is less influenced by the WF TCO . Moreover, it was excited to find that, as compared to the HIT counterparts, the H-H solar cell is also less sensitive to both the N tr and N a of the p-a-Si:H layer. The reason is that the n of the H-H solar cell is less influenced by the N tr or N a since the electric field within the homojunction can partly compensate the reduced electric field caused by the increased N tr or the decreased N a . Furthermore, the performances of the H-H solar cell keep higher in the whole range of each parameter. With better tolerance for all the above parameters that greatly influence the HIT cell performances, we believe that the H-H solar cell is promising to be adopted in photovoltaic industry to produce high-efficiency solar cells more stably.