A Stacking Metal-Insulator-Semiconductor Solar Cell

Extending photovoltaic applications plays a prominent role in the following evolution of photovoltaics. Using solar cells to convert solar energy for hydrogen production (solar hydrogen) [184]-[186] can be considered a promising application (a perfect renewable energy source), as shown in Fig. 5.1(a). Generated hydrogen can convert into not only thermal energy but also electricity and mechanical energy, and return to water finally (a clean, renewable energy cycle), as shown in Fig. 5.2(b).

One of essential factors in solar hydrogen is that the Voc‟s of solar cells have to be larger than the dissociating voltage of water (1.23V). [184]-[186] However, the Voc‟s of conventional single junction solar cells are not compatible with the dissociating voltage. For matching the requirement, we can probably adopt following four strategies.

Fig. 5.1 (a) The illustration of converting solar energy for hydrogen production using photovoltaics. (b) The clean, renewable energy cycle of solar hydrogen.

55

(a) Increasing the n of solar cells

According to Eq. (1.2), increasing n can make V_oc larger, but it degrades the FF as well. [159]

(b) Reducing the surface defects of solar cells

Surface defects enlarge the surface recombination rates of solar cells, and thus lower V_oc. For reducing surface recombination rates, many surface passivation methods (such as SiNx passivation [155], and AlOx passivation [187]) are demonstrated to decrease surface defects, and therefore, enhance V_oc. Unfortunately, the augment is limited. Take crystalline silicon solar cells with excellent surface passivation for example. The measured V_oc‟s are unable to be larger than the built-in voltages of pn junctions (~0.7 V).

(c) Enlarging the bandgaps of the active layers of solar cells

For single junction solar cells, the upper limits of Voc‟s are the bandgaps of active layers. It suggests that we can use active materials with larger bandgaps to get higher Voc‟s. Nevertheless, using the active materials will make less incident photons be absorbed by solar cells. [188]

(d) Integrating solar cells with different bandgaps (tandem solar cells)

It is an effective way to raise V_oc, and incident photons can be efficiently absorbed by solar cells. [103]

Based on the concept of tandem solar cells, we propose a stacking MIS (crystalline silicon) solar cell, which integrates an n-type MIS (crystalline silicon) solar cell with a p-type MIS one, to effectively increase Voc, as shown in Fig. 2.7. There are two advantages using MIS (crystalline silicon) solar cells. First, the fabrication of MIS (crystalline silicon) solar cells is low-cost, low-temperature, simple processes. Second, the V_oc‟s of MIS (crystalline silicon) solar cells are comparable with those of pn junction (crystalline silicon) solar cells.

This is because the ultrathin oxide layers can block the majority carriers from transporting to

56

the metal electrodes, and are easy for minority carriers to tunnel through, so the effect of carrier recombination at the MS junctions is vastly decreased. [151], [152], [159], [190], [191]

Our experimental results show that by comparing with other MIS solar cells with the same output power, this stacking structure has potential to give high Voc and low Jsc, which can reduce the electrical loss in MIS solar cell applications. Furthermore, this stacking structure shows an effective way to enlarge V_oc and has potential to achieve a V_oc larger than 1.23 V, although the Voc‟s of conventional MIS solar cells are not high enough to match the requirement for dissociating water (1.23 V) [184]–[186]. Based on the stacking structure, we believe that MIS solar cells are very promising for photoelectrochemical (PEC) water splitting and that monolithic MIS photovoltaic-PEC devices can be realized in the future for both solar photovoltaics and solar hydrogen.

5.2 Results And Discussions

For obtaining the optimized process conditions of the top and bottom cells of the stacking MIS solar cell, we fabricated n-type and p-type MIS solar cells under different process conditions, and analyzed their characteristics.

The red square curves of Fig. 5.2(a) show the J-V curves of the n-type and p-type MIS solar cells with as-deposited ultrathin SiO₂ layers deposited at 20 mTorr under light illumination, and the measured parameters are listed in Table 5.1. The MIS solar cells exhibit poor photovoltaic characteristics probably because of not good interface properties between the ultrathin SiO2 layers and Si. To analyze the interface properties of the n-type and p-type MIS solar cells, high-frequency C-V curves were recorded, as shown by the red square curves in Fig. 5.2(b). Large capacitance leakage is observed in both C-V curves. It indicates that there are a great number of trap states at the interfaces, such as silicon dangling bonds and silanol groups. [192], [193] Charge carriers can flow via the trap states to cause current leakage, and the capacitance is reduced. To passivate these trap states, H₂ annealing was

57

introduced. The blue circle curves in Figs. 5.2(a) and (b) show the J-V and C-V curves of MIS solar cells with additional H₂ annealing, respectively. The MIS solar cells with additional H₂ annealing exhibit better photovoltaic properties, and their C-V curves display smaller drops at the accumulation regions. These attribute to the formation of Si-H bonds at the interfaces, and then the trap states are diminished and the leakage currents are decreased. Therefore, H2

Fig. 5.2 (a) The J-V curves of MIS solar cells with and without H2 annealing under light illumination. (b) Corresponding C-V curves. All capacitance values are normalized by the largest value of each curve.

58

annealing can improve the performance of MIS solar cells with ultrathin SiO2 layers deposited by RF sputtering because of the passivation of trap states.

The working pressure of depositing the ultrathin SiO2 layers plays an important role in the V_oc of the MIS solar cells with ultrathin SiO₂ layers. The J-V curves of n-type and p-type MIS solar cells with fixed ultrathin sputtering SiO2 layer thicknesses (n-type with about 2 nm and p-type with about 1 nm) deposited under different working pressures with H₂ annealing are shown in Fig. 5.3 and measured parameters are listed in Table 5.2. For the p-type MIS solar cells, 475 mV, 386 mV, and 350 mV open-circuit voltages, corresponding to working pressures at 20 mTorr, 30 mTorr, and 40 mTorr, respectively, are obtained. The Voc‟sof the n-type MIS solar cells are 422 mV, 339 mV, and 313 mV corresponding to working pressures at 20 mTorr, 30 mTorr, and 40 mTorr, respectively. The relationships between the Voc‟s and the working pressures during deposition are shown in Fig. 5.4(a). For both MIS solar cells, lower working pressures results in larger Voc‟s.

The corresponding C-V measurements are shown in Fig. 5.4(b). The C-V curves of the n-type/p-type MIS solar cells shift toward positive/negative bias with decreasing working pressure. At a lower working pressure, Ar⁺ ions get more energy to sputter a SiO₂ target due to a larger mean free path. As a result, more surface charges are introduced in the ultrathin sputtering SiO₂ layers on the Si substrates because of oxygen vacancies. Moreover, the Table 5.1 The measured parameters of MIS solar cells with and without H₂ annealing.

59

deposition rates of the ultrathin sputtering SiO2 layers increase with decreasing working pressure in our case. A higher deposition rate reduces the ion bombardment duration at the interfaces. As a result, less trap states are produced, and consequently, Voc is enlarged.

According to these two factors, the C-V curves of the p-type MIS solar cells shift toward a Table 5.2 The measured parameters of MIS solar cells with the fixed thicknesses of ultrathin sputtering SiO2 layers (n-type ~ 2 nm, p-type ~ 1 nm) deposited under different working pressures with H2 annealing under light illumination.

Fig. 5.3 The J-V curves of MIS solar cells with the fixed thicknesses of ultrathin sputtering SiO2 layers (n-type ~ 2 nm, p-type ~ 1 nm) deposited under different working pressures with H2 annealing under light illumination.

60

more negative voltage with a lower working pressure. However, the voltage shift toward a more positive voltage is not so significant in the C-V curves of the n-type MIS solar cells because of the presence of oxygen vacancies and the requirement of a thicker sputtering SiO₂ insulating layer. This requirement causes much longer ion bombardment duration than Fig. 5.4 (a) Voc and (b) C-V curves of n-type and p-type MIS solar cells with the fixed thicknesses of ultrathin sputtering SiO2 layers (n-type ~ 2 nm, p-type ~ 1 nm) deposited under different working pressures with H2 annealing. All capacitance values are normalized by the largest value of each curve.

61

that of the p-type MIS solar cells, and then, more trap states are generated to decrease Voc. From these results, we chose the best process conditions of the n-type and p-type MIS solar cells to manufacture the top and bottom cells of the stacking MIS solar cell. The fabrication details are introduced in Chapter 2.

The operation of the proposed stacking MIS solar cell is illustrated in Fig. 5.5. The photons incident into the stacking MIS solar cell are absorbed by the top and bottom cells, and electron-hole pairs are generated. The holes in the top cell and the electrons in the bottom cell can transport to and tunnel through the ultrathin SiO₂ layer of each cell and then are collected by the electrodes. In addition, the electrons in the top cell and the holes in the bottom cell can diffuse to and tunnel through the n⁺-p⁺ junction to recombine. The resulted V_oc of the stacking MIS solar cell is given by the sum of the top and bottom cells. For this reason, the proposed stacking MIS solar cell can give a higher Voc than an individual n-type or p-type MIS solar cell, and consequently has much more promising MIS solar cell applications.

The J-V curves of the proposed stacking MIS solar cell with and without

Fig. 5.5 Energy band diagram and operation of the stacking MIS solar cell.

62

light illumination are shown in Fig. 5.6. In the dark, it provides the nature of a diode. Under illumination, V_oc = 0.71 V, J_sc = 5.94 mA/cm², FF = 58%, and η = 2.47% are obtained. In addition, the J-V curves under illumination of an n-type MIS solar cell and a p-type one after optimization are shown in the inset of Fig. 5.6. The V_oc‟s are 0.42 and 0.47 V for the n-type and p-type MIS solar cells, respectively. It is obvious that the measured Voc of the stacking MIS solar cell is larger than that of either an n-type or a p-type MIS solar cell. For comparison, the Voc‟s of various MIS solar cells using different processes for insulating layers [151]-[153], [157]-[159], [160] are listed in Table 5.3. Note that the V_oc of our stacking MIS solar cell is superior to those of the n-type or p-type MIS solar cells with or without surface passivation.

The V_oc enhancement has been achieved by the proposed structure. However, the measured Voc is not as expected, and other relative parameters such as η are not comparable with those of conventional MIS solar cells. These attribute to non-optimized processes (such as the bonding process), current mismatching, and optical losses (such as metal reflection).

Fig. 5.6 The J-V curves of the stacking MIS solar cell with and without light illumination.

Inset: The J-V curves of the optimized n-type and p-type MIS solar cells under light illumination.

63

Table 5.3 Voc comparison between the proposed stacking MIS solar cell and other MIS and MIS-IL solar cells.

64

Figure 5.7 shows the IR images of the bonding interface of the stacking MIS solar cell after different fabrication processes. From the IR images, we know that the bonding interface is getting poor during the fabrication of the stacking MIS solar cell. This kind of bonding interface results in a damaged tunneling junction, and thus lowers the V_oc of the stacking MIS solar cell.

Current matching between subcells is a critical factor for tandem solar cells and stacking ones to obtain high efficiency. In our case, the thickness tuning is an easy approach to matching the photocurrents of the top and bottom cells. Considering silicon absorption, a rough calculation shows that the thickness of the top cell is about 4 μm, if 50% of photons in the range of 280-1100nm of AM1.5G spectrum reaches the bottom cell. The 400-μm-thick bottom cell with a perfect back reflector can absorb up to 92% of the remaining photons from the top cell. That is, the optimized thickness of the top cell is about 4 μm for the 400-μm-thick bottom cell.

The actual thickness of the top cell of the stacking MIS solar cell is hard to measure directly. We can estimate this thickness by the measured Jsc‟s of the stacking MIS solar cell and the p-type MIS solar cell shown in the inset of Fig. 5.6. The estimated thickness of the top cell is around 10 μm. This value is thinner than the one obtained from the reference n-type

Fig. 5.7 The IR images of the bonding interface of the stacking MIS solar cell (a) as-bonded stacking MIS solar cell. (b) RTA annealed stacking MIS solar cell. (c) Thinned stacking MIS solar cell.

65

Si substrate, mentioned in Chapter 2, probably because of the inappropriate thickness estimation method and the effect of nonuniform etching during the thinning process. However, the estimated thickness is thicker than 4 μm. That is, current mismatching between the top and bottom cells occurs, and accordingly, degrades the potency of the stacking MIS solar cell.

Optical loss is another major energy loss of the stacking MIS solar cell. The top electrode and the semitransparent metal layer can induce an inversion layer close to the Si/SiO2 interface to form a MIS junction, and can reduce Rs, but they also obstruct incident photons into the stacking MIS solar cell because of the characteristics of metal, and then cut down the amount of photoexcited carriers.

Substituting the semitransparent metal layer by a surface passivation layer is an effective method for increasing the photon fluxes into the stacking MIS solar cell. [194] The passivation layer can make much more photons inject into the stacking MIS solar cell because of superior transparency. In addition, the passivation layer, which can act as a metal electrode, can induce a similar inversion layer close to the Si/SiO₂ interface to form a MIS junction, to increase the diffusion length of carriers, and to improve the collection efficiency of carriers.

[194]

5.3 Summary

It is demonstrated that an MIS solar cell using a stacking structure without surface passivation exhibits a high V_oc. The obtained V_oc is up to 0.71 V, greater than those of other reported MIS solar cells. We successfully show that the proposed stacking structure is feasible to enhance V_oc. It is expected that the performance of the proposed stacking MIS solar cell can be improved by process optimization (such as interface quality after bonding), current matching (such as decreasing the thickness of the top cell), and surface passivation (such as AlOx for the top cell). Therefore, the proposed stacking MIS solar cell has great potential in the future development of solar cells and PEC water splitting.

66

Chapter 6

Conclusions And Future Works

6.1 Conclusions

With the dimension(s) of materials shrinking to 1 to 100 nm, especially close to their Bohr radii, they can exhibit special identities. We can employ them in our thin films or devices to accomplish specific purposes. In this dissertation, we utilize three kinds of ultrathin oxide layers to realize three green devices: an OBD with an interfacial Al-O compound layer, an OBD with a MoOx nanoclusterlike layer, and a high Voc MIS solar cell using a stacking structure.

In the beginning, an OBD with an n-Si/Alq3/Al structure is fabricated and its characteristics are also analyzed. We find that the bistability of the OBD results from the charge trapping in the Al-O compound layer, which is formed at the Alq3/Al interface. We can also tune the electrical features of the OBD, which are affected by the surface roughness of the Alq3 thin film, by controlling the deposition rate of the Alq3 thin film.

After that, we pay attention to another OBD, p⁺-Si/Alq3/nanostructured MoOx/Alq3/Al.

The MoOx nanoclusterlike layer (the nanostructured MoOx layer) behaves as trap sites in the OBD, and the resistance switching of the OBD can be observed in the I-V curves as charges occupy or leave the trap sites. After the OBD is switched into the high conductance state, the current keeps increasing with voltage, and no NDR exhibits in the I-V curves. This is because the effective polarized field generated by the MoOx nanoclusterlike layer is not large enough to repel the injected carriers.

In addition, a high Voc MIS solar cell using a stacking structure is reported. We stack an n-type MIS solar cell and a p-type one to form a stacking MIS solar cell. The stacking cell

67

provides 0.71 V Voc, greater than those of other reported MIS solar cells. It is worthy to mention that the stacking structure has great potential to carry out converting solar energy for hydrogen generation.

Finally, we deeply believe that the reported green devices will make a great influence on the future green technology.

6.2 Future Works

In the near future, we will/can focus on further promoting the performance of the three green devices.

In the case of the OBDs, the retention time of the OBDs is a key identity of memory.

However, the retention time of both OBDs has not fitted the requirements of the next generation nonvolatile memory yet. We can introduce deeper trap states in the OBDs (e.g., changing the shape of the MoO_x layer) or barriers adjacent to the original tap sites to prevent trapped charges from escaping, and thus their retention time is raised.

In addition, we find that the failure of the OBDs occurs after several write-read-erase-read cycle tests. The failure probably attributes to many factors, such as device degradation due to the measurement environment, and leakage paths which result from particles introduced during fabrication. However, the failures of organic memory have not caused much attention. If we can recognize the failure mechanisms, we can figure out how to tackle the causes of the failure and then can uplift the stability and endurance of the OBDs, and even perhaps can understand the mechanisms responsible for the resistance switching.

With regard to the stacking MIS solar cell, it is evident that the stacking MIS solar cell offers higher V_oc than other published MIS solar cells. But the stacking MIS solar cell can donate more superior performance as long as we can cut down its power losses. For example, we can modify (or change) our fabrication processes and then can optimize the processes for

68

the requirement of current matching and for obtaining the excellent bonding interface and high quality MIS junctions. After that, we can use a surface passivation layer (such as AlO_x) instead of the semitransparent metal layer to raise the amount of photons incident into the cell, and to induce an inversion layer for arguing the diffusion length and collection efficiency of carriers. Furthermore, from the point of view of energy collection, the performance of the stacking solar cell (such as conversion efficiency) can be further upgraded using two MIS solar cells with different bandgaps (such as Si/Ge, and Si/III-V).

69

References

[1] G. A. Meeh, W. M. Washington, W. D. Collins, J. M. Arblaster, A. Hu, L. E. Buja, W. G. Strand, and H. Teng, “How much more global warming and sea level rise?,”

Science, vol. 307, no. 5716, pp. 1769-1772, Mar. 2005.

[2] T. L. Root, J. T. Price, K. R. Hall, S. H. Schneider, C. Rosenzweig, and J. A.

Pounds, “Fingerprints of global warming on wild animals and plants,” Nature, vol.

421, no. 6918, pp. 57-60, Jan. 2002.

[3] B. Eitan, P. Pavan, I. Bloom, E. Aloni, A. Frommer, and D. Finzi, “NROM: A novel localized trapping, 2-bit nonvolatile memory cell,” IEEE Electron. Dev. Lett., vol.

21, no. 11, pp. 543-545, Nov. 2000.

[4] D. Zhao, Y. Zhu, R. Li, and J. Liu, “Simulation of a cobalt silicide/Si hetero-nanocrystal memory,” Solid-State Electron., vol. 49, no. 12, pp. 1974-1977, Dec. 2005.

[5] C. J. Amsinck, N. H. D. Spigna, D. P. Nackashi, and P. D. Franzon, “Scaling

[5] C. J. Amsinck, N. H. D. Spigna, D. P. Nackashi, and P. D. Franzon, “Scaling