**3.2. Results**

From **Figure 6a** and **Table 2**, the AZO only IOSC without an interfacial layer presents a very poor performance and improved *V*OC, *J*SC, FF, and PCE after 15 min AM 1.5G continuous illumination (light soaking), shown in **Figures 7a** and **b**. This can be attributed to the energy loss associated with the mismatch of the energy levels between the work function of sputtered AZO (4.5 eV) and the lowest unoccupied molecular orbital (LUMO) level of PCBM (3.7 eV) (shown in **Figure 8**). During the light soaking, the photogenerated electron‐hole pairs in AZO could increase the electron density, fill the trap sites, and thus lower the work function of AZO [15], which improves the electron selectivity from organic layer to AZO cathode and thus induces the improvement of device parameters. However, the performance of AZO only IOSC is still very poor, especially the low *V*OC of 0.38 V, which indicates that a modifying layer with lower work function must be inserted between AZO and P3HT:PCBM to align their energy levels.

of defects is decreased and better ZnO film quality is obtained. Simultaneously, for different ZnO film annealing temperatures of 50, 70, 80, and 150°C, the corresponding root‐mean‐square surface roughness are 0.673, 0.867, 1.108, and 1.145 nm, respectively. The increased surface roughness corresponds to the more sufficient ZnO crystallization at a high annealing temper‐ ature. It is thought that, when the annealing temperature is below 80°C, the ZnO morphology changes markedly, and when the annealing temperature is above 80°C, the ZnO morphology almost remains unchanged. From the morphology results, it is concluded that an annealing temperature of 80°C is sufficient for ZnO crystallization. And the high temperature only slightly improves the film quality, a result agreed to the similar photovoltaic performance of

According to above discussion, one can draw the conclusion that the 80°C is sufficient for ZnO annealing to obtain a relatively high film quality and act as an interfacial layer, and the resulted devices based glass or PET substrates show the similar photovoltaic performance when the ZnO annealing temperature is higher than 80°C. In short, the senior device performance and good stability show that the aqueous solution method is a more promising low‐temperature technique for depositing ZnO in IOSCs and it may be widely applied in flexible and printing

**3. ITO-free OSCs based on AZO cathodes and Ca interfacial layer**

In this section, ITO‐free IOSCs are fabricated on AZO cathodes and ultrathin Ca interfacial layer, the optimization of device performance and light‐soaking issue are mainly discussed.

The AZO (~980 nm, 2 wt% Al)‐coated glass substrates purchased from Zhuhai Kaivo co. were prepared by radiofrequency magnetron sputtering process and they have a sheet resistance of about 7.34 ohm/square and average transmission over 80% in the visible light region. The P3HT:PCBM based bulk heterojunction IOSCs with the AZO transparent cathode was fabricated, and the obtained devices have the structure of Glass/AZO/Ca/P3HT:PCBM/MoO3/

5, 10 nm) were thermally evaporated on AZO at a base pressure below 5.0 × 10−4 Pa. The detail fabrication process is the same as that in previous section. The J‐V characteristics were measured under AM 1.5G solar simulator spectrum before and after 15‐min light soaking. The stability of devices was investigated by measuring their J‐V characteristics once every 5 days

From **Figure 6a** and **Table 2**, the AZO only IOSC without an interfacial layer presents a very poor performance and improved *V*OC, *J*SC, FF, and PCE after 15 min AM 1.5G continuous illumination (light soaking), shown in **Figures 7a** and **b**. This can be attributed to the energy loss associated with the mismatch of the energy levels between the work function of sputtered AZO (4.5 eV) and the lowest unoccupied molecular orbital (LUMO) level of PCBM (3.7 eV)

. The Ca interfacial modifier with different thicknesses (0, 1,

IOSC with ZnO annealed from 80 to 150°C.

devices in the future.

168 Nanostructured Solar Cells

**3.1. Device fabrication**

in 1 month.

**3.2. Results**

Ag and an active area of 12.5 mm2

**Figure 6.** Measured J‐V characteristics for as prepared and 15‐min illuminated AZO/Ca (0, 1, and 5 nm) IOSCs. Repro‐ duced with permission [20]. Copyright 2014, Elsevier.


**Table 2.** Photovoltaic parameters (PCE, *V*OC, *J*SC, and FF) obtained from J‐V characteristics of AZO only IOSC, AZO/Ca IOSCs with different Ca thicknesses (1, 5, 10 nm), and the referenced ITO/Ca (1 nm) IOSC. The "–" represents a very large value. For the S‐shaped J‐V characteristics, the value of *R*S is so large that the data are out of the diode model used in the calculation. Reproduced with permission [20]. Copyright 2014, Elsevier.

Then, an electron transport layer of Ca is inserted between AZO and P3HT:PCBM to modify the work function of AZO cathode and the Ca modifier has already been used in ITO based IOSCs [6]. It is clear in **Figure 6** and **Table 2** that, the devices with Ca interfacial layer achieve significant improvement in *V*OC, *J*SC, FF, and PCE compared with that of AZO only IOSC. Particularly, an increased *V*OC (0.60 V) demonstrates that the Ca work function has been pinned to the PCBM Fermi level via the surface states. Consequently, the resulted ohmic contact between Ca and PCBM favors the electron transport and the rectifying contact between Ca and P3HT blocks the hole collection at the AZO cathode. Furthermore, the devices with an ultrathin Ca (~1 nm) show a superior performance with the highest PCE of 3.17% and lowest *R*S of 0.7 Ω cm2 , which is comparable to that of ITO‐based device. Thus, the low‐cost AZO electrode is one of the promising ITO alternatives.

**Figure 7.** Variations of photovoltaic parameters (PCE, *V*OC, *J*SC, and FF) during 15 min light soaking for (a) and (b) the AZO only IOSC, (c) and (d) the AZO/Ca (5 nm) IOSC. Reproduced with permission [20]. Copyright 2014, Elsevier*.*

**Figure 8.** (a) Schematic energy diagram of AZO/Ca IOSCs and ideal band diagram of Ca‐AZO contacts according to the ideal metal‐n‐type semiconductor contact [19]. (b) The AZO work function of 4.5 eV and AZO/Ca work function of 3.8 eV were determined by UPS experiments. Reproduced with permission [20]. Copyright 2014, Elsevier.

One the other hand, for the AZO/Ca (5nm) IOSC, the as prepared device has a low PCE of 1.74% as well as a S‐shaped J‐V curve, and the PCE increases to 2.69% with the vanish of the S‐shaped curve during light soaking, which is observed clearly in **Figures 6b** and **7c**, and **7d**. It is noted that the trend of increasing PCE is nearly identical to that of FF, a slight increase in *J*SC and nearly unchanged *V*OC are also observed, which is similar to the reported light‐soaking issue [15]. It is obvious in **Table 2** that the *R*S of AZO/Ca (5 nm) IOSC shows a remarkable decrease from a very large value to 2.3 Ω cm2 , which is in good agreement with an increase in FF from 38% to 58%. Combining the improved device parameters of AZO only IOSC during light soaking, the strikingly decreased *R*<sup>S</sup> indicates the less energy loss in the charge transport at interfaces and the more efficient charge collection at electrodes. Here, it is thought that the interface electron transport from P3HT:PCBM to AZO is related to the light‐soaking issue, which will be discussed in next part. The nearly unchanged VOC demonstrates that an ultrathin Ca layer of 1 nm can also effectively modify the work function of AZO cathode. More remark‐ ably, no light‐soaking issue is observed in this device and the detail discussion is shown in next section. Thus, considering the different photovoltaic performances of AZO/Ca (0, 1, 5, and 10 nm) IOSCs, the underlying reasons are deserved to be further investigated.

#### **3.3. Discussion**

Ca (~1 nm) show a superior performance with the highest PCE of 3.17% and lowest *R*S of 0.7

**Figure 7.** Variations of photovoltaic parameters (PCE, *V*OC, *J*SC, and FF) during 15 min light soaking for (a) and (b) the AZO only IOSC, (c) and (d) the AZO/Ca (5 nm) IOSC. Reproduced with permission [20]. Copyright 2014, Elsevier*.*

**Figure 8.** (a) Schematic energy diagram of AZO/Ca IOSCs and ideal band diagram of Ca‐AZO contacts according to the ideal metal‐n‐type semiconductor contact [19]. (b) The AZO work function of 4.5 eV and AZO/Ca work function of

3.8 eV were determined by UPS experiments. Reproduced with permission [20]. Copyright 2014, Elsevier.

, which is comparable to that of ITO‐based device. Thus, the low‐cost AZO electrode is

Ω cm2

170 Nanostructured Solar Cells

one of the promising ITO alternatives.

For a good understanding of different photovoltaic behaviors of AZO‐based IOSCs, the following discussions focus on the optical and electrical properties, surface morphology of the AZO/Ca (0, 1, 5, and 10 nm) films. From optical aspect, the transmission spectra of AZO/Ca (x nm) samples in **Figure 9(a)** shows that AZO has a very similar transmission tendency to that of AZO with 1, 5, and 10 nm Ca deposited on it. The lower transmission of AZO/Ca (10 nm) substrate means relatively larger light absorption loss in the active layer, which can be used to explain the relatively poor performance of the AZO/Ca (10 nm) IOSC.

**Figure 9.** (a) Transmission spectra for the ITO, AZO and AZO/Ca (1, 5, and 10 nm) coated glass substrates. Insert: en‐ larged pictures for AZO/Ca (0, 1, 5, and 10 nm) substrates in the range from 470 to 530 nm. (b) Photoconductivity measured before and after 15 min AM 1.5 G illumination for the AZO, AZO/Ca (1 nm), and AZO/Ca (5 nm) films on glass substrates. The insert shows the schematic diagram of glass/AZO/Ca (0, 1, and 5 nm)/Ag samples for photocon‐ ductivity measurement. Adapted with permission [20]. Copyright 2014, Elsevier.

The coverage of Ca (1, 5, and 10 nm) on AZO films and their surface morphology were characterized by AFM and FE‐SEM and no significant differences in morphology could be observed for AZO, AZO/Ca (1 nm), AZO/Ca (5 nm), and AZO/Ca (10 nm) films. This may be related to the very rough surface of AZO films (RMS = 11.59 nm). Fortunately, some useful information may be indirectly acquired from XPS depth profile since the coverage of Ca (1, 5, and 10 nm) on AZO films could be reflected by the different Zn and Ca elements content at the sample surface. The detail XPS analysis and Ca coverage study could be found in our reports [20]. According to the XPS study and the experience Volmer‐Weber growth or 3D island growth [21], the AZO film can be completely covered by a Ca layer about 10 nm, while the 1 nm Ca on AZO exists as unclosed islands and the 5 nm Ca could partly cover the AZO surface. From electrical aspect, the measured UPS (He I, 21.2 eV) spectra in **Figure 8** show that the AZO work function (4.5 eV) could be reduced to 3.8 eV by introducing an 5 nm Ca interfacial layer, which is well matched to the LUMO level (3.7 eV) of PCBM. Thus, a Ca modifier can be used to align the energy levels between P3HT:PCBM and AZO cathode.

To further understanding the electron transport in AZO, AZO/Ca (1 nm), and AZO/Ca (5 nm) films, the photoconductivity was investigated as following. It can be observed from **Figure 9b** that the conductivity of AZO and AZO/Ca (1 nm) does not change significantly with the continuous AM 1.5 G illumination; however, the conductivity of AZO/Ca (5 nm) film presents an obvious increase during the 15‐min light soaking. From the band diagrams of ideal Ca‐AZO contacts shown in **Figure 8**, the different work function between Ca and AZO would cause an electron transport barrier at AZO/Ca interface. For the 5 nm Ca on AZO, the low initial conductivity suggests the large energy loss in the electron transport across AZO/Ca (5 nm) interface, which can be attributed to the electron transport barrier from Ca to AZO as well as the oxidization of Ca layer. During the 15‐min light soaking, the photogenerated electron‐hole pairs in AZO could increase the electron density, fill the trap sites, and thus lower the work function of AZO, which also lowers the electron transport barrier and improves the electron transport efficiency at AZO/Ca interface, a situation agreed with an increased conductivity of AZO/Ca (5 nm) film and improved *J*SC in corresponding device. After about 15 min, no further improvement of photovoltaic parameters can be observed and the electron transport reaches a saturated case. For the AZO/Ca (1 nm) sample, its initial conductivity is comparable to the maximum conductivity of AZO/Ca (5 nm) (**Figure 9b**). More importantly, no significant improvement of conductivity is observed upon continuous illumination, which suggests the sufficient charge transport from AZO to Ca. This result is in line with the senior and stable photovoltaic parameters of AZO/Ca (1 nm) IOSC.

According to previous discussion, the high efficient electron transport at AZO/Ca (1 nm)/ organic interface may be related to the Ca coverage on AZO. As we know, the 1 nm Ca only exists as isolated islands on AZO surface, and it is thought that the ultrathin Ca may increase the number of active sites, and these isolated sites with low electric potential provide the fast pathway of electron from P3HT:PCBM to AZO. This process may be understood by introduc‐ ing the Liebig's law of the minimum [22] that the barrel capacity is limited by the shortest stave. Analogously, the active sites may act as the shortest stave and the electrons play the role of water, and the electron prefers to travel through the low‐electric‐potential pathway provided by the island‐like ultrathin Ca on AZO film, instead of the direct transport across the larger energy barrier at AZO/organic interface. This discussion agrees with the smaller *R*S, larger *J*SC, FF, and PCE of AZO/Ca (1 nm) IOSCs. In other words, the highly efficient electron transport in AZO/Ca (1 nm) film may be responsible for the removed light‐soaking issue of as prepared AZO/Ca (1 nm) IOSC. It also should be noted that the unexpected light soaking issue appears after the un‐encapsulated device has been stored in air for several days since the oxidation of Ca, but this does not happen for the device stored in N2. In short, the observed light‐soaking phenomenon is strongly related to the electron transport ability at the AZO/Ca/organic interface and the energy loss caused by the oxidation of Ca, and the more exact mechanism should be further investigated in the future work.

The coverage of Ca (1, 5, and 10 nm) on AZO films and their surface morphology were characterized by AFM and FE‐SEM and no significant differences in morphology could be observed for AZO, AZO/Ca (1 nm), AZO/Ca (5 nm), and AZO/Ca (10 nm) films. This may be related to the very rough surface of AZO films (RMS = 11.59 nm). Fortunately, some useful information may be indirectly acquired from XPS depth profile since the coverage of Ca (1, 5, and 10 nm) on AZO films could be reflected by the different Zn and Ca elements content at the sample surface. The detail XPS analysis and Ca coverage study could be found in our reports [20]. According to the XPS study and the experience Volmer‐Weber growth or 3D island growth [21], the AZO film can be completely covered by a Ca layer about 10 nm, while the 1 nm Ca on AZO exists as unclosed islands and the 5 nm Ca could partly cover the AZO surface. From electrical aspect, the measured UPS (He I, 21.2 eV) spectra in **Figure 8** show that the AZO work function (4.5 eV) could be reduced to 3.8 eV by introducing an 5 nm Ca interfacial layer, which is well matched to the LUMO level (3.7 eV) of PCBM. Thus, a Ca modifier can be used

To further understanding the electron transport in AZO, AZO/Ca (1 nm), and AZO/Ca (5 nm) films, the photoconductivity was investigated as following. It can be observed from **Figure 9b** that the conductivity of AZO and AZO/Ca (1 nm) does not change significantly with the continuous AM 1.5 G illumination; however, the conductivity of AZO/Ca (5 nm) film presents an obvious increase during the 15‐min light soaking. From the band diagrams of ideal Ca‐AZO contacts shown in **Figure 8**, the different work function between Ca and AZO would cause an electron transport barrier at AZO/Ca interface. For the 5 nm Ca on AZO, the low initial conductivity suggests the large energy loss in the electron transport across AZO/Ca (5 nm) interface, which can be attributed to the electron transport barrier from Ca to AZO as well as the oxidization of Ca layer. During the 15‐min light soaking, the photogenerated electron‐hole pairs in AZO could increase the electron density, fill the trap sites, and thus lower the work function of AZO, which also lowers the electron transport barrier and improves the electron transport efficiency at AZO/Ca interface, a situation agreed with an increased conductivity of AZO/Ca (5 nm) film and improved *J*SC in corresponding device. After about 15 min, no further improvement of photovoltaic parameters can be observed and the electron transport reaches a saturated case. For the AZO/Ca (1 nm) sample, its initial conductivity is comparable to the maximum conductivity of AZO/Ca (5 nm) (**Figure 9b**). More importantly, no significant improvement of conductivity is observed upon continuous illumination, which suggests the sufficient charge transport from AZO to Ca. This result is in line with the senior and stable

According to previous discussion, the high efficient electron transport at AZO/Ca (1 nm)/ organic interface may be related to the Ca coverage on AZO. As we know, the 1 nm Ca only exists as isolated islands on AZO surface, and it is thought that the ultrathin Ca may increase the number of active sites, and these isolated sites with low electric potential provide the fast pathway of electron from P3HT:PCBM to AZO. This process may be understood by introduc‐ ing the Liebig's law of the minimum [22] that the barrel capacity is limited by the shortest stave. Analogously, the active sites may act as the shortest stave and the electrons play the role of water, and the electron prefers to travel through the low‐electric‐potential pathway provided

to align the energy levels between P3HT:PCBM and AZO cathode.

172 Nanostructured Solar Cells

photovoltaic parameters of AZO/Ca (1 nm) IOSC.

**Figure 10.** Normalized PCE degradation for AZO/Ca (1 nm) IOSCs and ITO/Ca IOSCs stored in air or N2. Reproduced with permission [20]. Copyright 2014, Elsevier.

What is more, we investigate the stability of un‐encapsulated devices stored in air or N2 for one month and their normalized PCEs are shown in **Figure 10**. It is very clear that the AZO/Ca (1 nm) IOSC stored in N2 (measured in air) shows a good stability that its PCE could maintain 80% of the original values after one month. It is noted that the relatively stable *V*OC demonstrates that the Ca‐modified AZO cathode can provide suitable contact for electron collection. Also from **Figure 10**, a degradation of PCE by 50% for the air‐stored AZO/Ca (1 nm) IOSC during one month is observed, which can be attributed to the inevitable penetration of oxygen/water from air into the active layer and the oxidization of Ca‐modifying layer. Here, the effect of Ca oxidation on the stability could be confirmed from the results of AZO only IOSC. In detail, this device without Ca shows a better stability during the first two weeks in air than that of AZO/Ca (1 nm) IOSC, whereas the mismatch of energy levels at AZO/ P3HT:PCBM interface may be responsible for the larger energy loss in charge transport and more un‐durable PCE after two weeks. However, the PCE of ITO/Ca (1 nm) IOSC deteriorates by 50% only in 10 days and drops by 80% of the origin values in the next ten days, which further shows the effect of Ca oxidation on the device air‐stability. In addition, the increased contact resistance caused by the oxidation and mechanical damages of relatively thin Ag (70 nm) anode by metal clips during the J‐V test may also be a factor of device degradation. As a result, considering the comparable PCEs and better device air‐stability, the AZO cathode is a prom‐ ising alternative of ITO to fabricate the long‐lifetime IOSCs.
