Fabrication and evaluation of figures of merit of ZnO polymer-based hybrid UV photodiodes

Ultraviolet (UV) light detection is required for several military, civilian, and scientific applications such as missile warning, flame detection, environmental monitoring, image sensing, and satellite communication.^1,2 Inorganic wide bandgap semiconductors, including SiC, TiO₂, GaN, AlGaN, diamond, and ZnO, are widely used for UV photodetection.^3,4 Among them, the intrinsic n-type ZnO is the most popular because of its wide direct bandgap of 3.3 eV, high excitation binding energy of 60 meV, low toxicity and cost of production, high chemical stability, and remarkable electrical and optical properties.^5–8 When ZnO nanostructures are excited with UV light, oxygen molecules desorb from the surface, leaving behind free electrons that contribute to the generation of photocurrent in ZnO nanostructures.^9,10 Due to these surface reactions, charge trapping at various defect states occurs. Therefore, after the removal of UV irradiation, the phenomenon of persistent photoconductivity (PPC) in ZnO nanostructures becomes prominent.^10–12 In our previous study, four ZnO nanorod samples prepared with different seeding alcohols were excited with 365 nm UV light. A low responsivity of 0.1–0.2 A/W with a high recovery time of 75–300 s was obtained from such samples, owing to PPC.¹³ Singh et al. also reported a low responsivity of 0.022 A/W at 365 nm UV excitation for a Ni/Al:ZnO/Ni metal–semiconductor–metal UV sensor.¹⁴ Therefore, even though ZnO nanostructure-based devices show high photoconductivity on UV exposure, they, nevertheless, suffer from low photoresponsivity and high recovery times due to PPC, limiting their applicability in UV photodetectors.

A p–n heterojunction between n-type and p-type semiconductors enhances the photoresponse and photocurrent in photodiodes.^15,16 Under a reverse bias, a built-in potential at the interface of the p–n heterojunction allows the directional movement of photogenerated electrons and holes, leading to an increase in the photocurrent in the photodiodes. In that regard, p–n heterojunctions with n-type ZnO nanorods have been constructed using several inorganic CuO, NiO, p-Si, and p-GaN, as well as organic p-type semiconductors such as poly(3, 4 ethylenedioxythiophene):polystyrenesulfonate (PEDOT:PSS).^17–20 Among them, the organic PEDOT:PSS is highly beneficial because of its high electrical conductivity of up to 10³ S/cm, work function of 5.0–5.2 eV, and high photo- and electrical stability under ambient conditions.^17,21 In addition, PEDOT:PSS can be easily coated on both rigid and flexible substrates using cost-effective coating techniques such as spin-coating and spray-coating. However, PEDOT:PSS-ZnO hybrid p–n junctions suffer from large band offsets at the interface, which results in energy loss due to the recombination of photogenerated carriers and, in turn, a reduction of the photocurrent.¹⁸ Therefore, the band alignments need to be adjusted by introducing a third material between PEDOT:PSS and ZnO that bridges the potential at the interface of the layers. Subsequently, it reduces the recombination of photogenerated carriers, which enhances the photocurrent.¹⁸ 

In this study, a p-type poly(9,9-dioctylfluorene-alt-benzothiadiazole) (F8BT) polymer layer was introduced in between n-type ZnO nanorods and p-type PEDOT:PSS polymer. In fact, F8BT has its highest occupied molecular orbital (HOMO) level at −3.3 eV and lowest unoccupied molecular orbital (LUMO) level at −5.9 eV, close to the PEDOT:PSS HOMO level of −3.5 eV and LUMO level of −5.2 eV. Therefore, due to their closely aligned HOMO and LUMO levels, the F8BT layer improved the band alignment by reducing the band offset between ZnO nanorods and PEDOT:PSS, allowing a more efficient charge transfer and lower energy losses.

In our previous work, we highlighted the influence of seeding alcohols, i.e., methanol, ethanol, isopropanol, and aqueous ethanol (70% ethanol, 30% water), on the nanorod diameter, uniformity, and surface defects.¹³ In that work, photoluminescence spectroscopy revealed a higher number of surface defects in ZnO nanorods grown on methanol and isopropanol seed layers than those grown on ethanol and aqueous ethanol seed layers. In addition, scanning electron microscopy (SEM) images of these ZnO nanorods revealed that the most uniform surface coverage of the nanorod growth was for the methanol, followed by ethanol, isopropanol, and aqueous ethanol seed layers.¹³ 

In order to fabricate photodiodes in this study, similar ZnO nanorods were grown on a 100 nm thick indium tin oxide (ITO) substrate with a 20 × 15 mm² area and 20 Ω/square resistance. A step-by-step description of the photodiode fabrication process is shown in the supplementary material, Fig. S.1. For the nanorod growth, first seed layers prepared with the four alcohols were drop cast, as shown in Fig. S.1(b). Subsequently, the ZnO nanorods were grown by hydrothermal growth, as shown in Fig. S.1(c), as described elsewhere.¹³ Then, a solution of 20 mg F8BT (Sigma-Aldrich, 698 687) in 2 ml of toluene (Honeywell) was prepared, and 20 μl of the F8BT solution was spin-coated over the ZnO nanorods at a speed of 2000 rpm for 30 s, as shown in Fig. S.1(d). The ZnO–F8BT layers were annealed at 110 °C for 10 min. Then, 20 μl of the as-purchased PEDOT:PSS (OSSILA, F HC Solar) was spin-coated at a speed of 3000 rpm for 50 s [Fig. S.1(e)], and the substrate was again annealed at 110 °C for 10 min. For the electrical measurements, Ag (Chemtronics) front and back contacts were deposited on the top of the PEDOT:PSS layer and on ITO, as shown in Fig. S.1(f). Therefore, four photodiodes with a configuration of ITO/ZnO nanorods/F8BT/PEDOT:PSS/Ag were prepared, where the only differences were in the ZnO-nanorod seed layer alcohol solvents. The four photodiodes are abbreviated as device-M, device-E, device-I, and device-E_aq, where M stands for methanol, E for ethanol, I for isopropanol, and E_aq for aqueous ethanol, respectively. In addition, SEM on the photodiodes was carried out on an FEI QUANTA 250 environmental SEM FEG operating at 15 kV with a resolution of 0.5 nm. Two source measurement units (Agilent 4156) with a step size of 20 mV and maximum compliance of 100 mA were used for the electrical measurements. Their photoresponse evaluation was carried out using a 125 W Hg lamp with an output wavelength of 365 nm, an optical fiber diameter of 4 mm, and an irradiance power of ∼15 mW/cm². In order to obtain optimized polymer thickness and control the reproducibility of the photodiodes, several other devices were fabricated and tested by measuring their I–V in the dark. The results are provided in Figs. S.2 and S.3. Figure S.2 indicates that the optimal quantities of F8BT and PEDOT:PSS for device fabrication are 20 µl for both polymers. Figure S.3 suggests that there could be a ±10 mA difference in the output current among the two sets of devices. These arise from parameters such as the surface area of the top Ag contact as Ag contacts were deposited by hand using a thin wire, and the position of the Ag contacts. Nevertheless, the four sets of devices show similar I–V characteristics or good reproducibility.

Figure 1 is the cross-sectional view of the SEM images of the four photodiodes with different layer thicknesses labeled in Fig. 1(a) of device-M. The ZnO-nanorod layer is ∼1 μm thick, and the nanorod diameter is ∼100 nm. The nanorods of device-M, device-E, and device-E_aq show clear vertical growth on the ITO substrate, as observed in Figs. 1(a), 1(b), and 1(d), respectively, whereas nanorods of device-I present random orientations along with some vertical growth, as observed in Fig. 1(c). In addition, a combined F8BT-PEDOT:PSS-polymer layer of thickness of ∼150 nm is visible for device-M in Fig. 1(a). However, for other photodiodes, the polymer layers detached from the surface during SEM sample preparation and, therefore, were not visible. The bottom layer of ∼100 nm observed in SEM images is the ITO layer, coated on a glass substrate.

Figures 2(a) and 2(b) are the I–V and semi-log (I)–V characteristics of the photodiodes in the dark and under UV irradiations from −3 to +3 V. The dark I–V curves represented by dashed lines exhibit a diode-like rectifying behavior for the photodiodes,²² with the highest rectification ratio I_reverse/I_forward of 640 for device-M, followed by 80 for device-I, 70 for device-E, and the lowest rectification of 44 for device-E_aq, calculated at ±2 V. Here, I_reverse is the current at reverse bias, and I_forward is the current at forward bias. In addition, for every photodiode, the reverse current in dark tends to increase with an increase in the reverse bias voltage, attributable to the generation–recombination centers at the p–n heterojunction interface between the ZnO nanorod surface and F8BT.²³ These centers also play a role in the ideality factor (η) of the diodes, calculated using Cheung’s method;^24, η of ∼1.8, ∼2.0, ∼2.4, and ∼3.2 was calculated for device-I, device-M, device-E, and device-E_aq, respectively. In addition, η of ∼2.0 and ∼1.8 of device-M and device-I, respectively, indicate that the main current generation mechanisms in these two photodiodes are via diffusion and recombination.²⁵ On the other hand, a high ideality factor of η > 2 indicates extended defects or local non-linear shunts, due to which the electron–hole recombination current flows inhomogeneously.²⁶ Our previous study shows that the nanorods prepared using aqueous ethanol have the most non-uniform surface with several voids;¹³ therefore, the extended defects and local non-linear shunts are most plausible in device-E_aq, leading to a high η of ∼3.2.

The solid lines in Figs. 2(a) and 2(b) represent the I–V and semi-log (I)–V characteristics under UV irradiations, showing an increase in the overall output current compared to the dark current. The depletion region formed at the p–n interface between n-type ZnO nanorods and p-type F8BT/PEDOT:PSS polymers induces a built-in electric field that provides a driving force for the separation of photogenerated electron–hole pairs. Therefore, an enhancement in the overall output current under UV irradiation compared to dark conditions occurs.¹⁸ In Fig. 2(b), device-M and device-I demonstrate an increase in three orders of magnitude in the output current, owing to the higher number of surface defects in their bare nanorods, whereas, due to the lower amount of surface defects in the bare nanorods of device-E and device-E_aq, these two photodiodes exhibit only one order of magnitude enhancement in the output current under UV irradiations. Since all the other parameters of photodiode fabrication are identical, i.e., nanorod growth solution and thickness of the polymer layers, the differences in the output current can be directly correlated with the surface defects in the bare nanorods arising from the seeding layer alcohol solvent.

Figure 2(c) is the energy band diagram of photodiodes with the configuration ITO/ZnO nanorods/F8BT/PEDOT:PSS/Ag. The figure shows that the F8BT intermediate layer between ZnO and PEDOT:PSS creates a bridging potential that reduces the band offsets between ZnO and F8BT, promoting an efficient charge transfer between the layers. In addition, the n- and p-type semiconductor materials of the photodiodes exhibit band bending because of differences in their valance-conduction and HOMO–LUMO energy levels.¹⁸ In fact, electrons, being the majority carriers of n-type ZnO nanorods, migrate to PEDOT:PSS due to the upward band bending exhibited by ZnO, leaving behind holes at the surface of ZnO. On the other hand, PEDOT:PSS accepts these electrons on its surface, as it exhibits downward band bending. The collection of positive and negative charges at the interface results in a depletion region with a built-in electric field directed from ZnO towards PEDOT:PSS, generating rectifying I–V characteristics of the photodiodes. In addition, there is a potential barrier of ∼1.8 eV between the ZnO-nanorod valence band and the F8BT HOMO level. Under UV irradiation, such a potential difference supports the directional transfer of photogenerated holes from the ZnO nanorod valence band to the F8BT HOMO level, which can be further transferred to the PEDOT:PSS HOMO level. Therefore, the charge transfer mechanism explains the high photocurrent under UV irradiation compared to dark conditions for the photodiodes.

Figure 3 provides the I–t characteristics and photoresponse of the photodiodes at four different reverse biases of −0.1, −0.5, −1, and −2 V. These I–t measurements were performed under a series of UV on–off cycles; the UV-on cycles were fixed to 10 s and the UV-off cycles to 140 s, with a plausible error of 1 s due to manual operation. In particular, the magnitude of the photocurrent increases with an increase in the reverse voltage, owing to the increase in the width of the depletion region.^18,27 However, the stability of the photoresponse evaluated in the UV-on and UV-off cycles is different at different biases for the photodiodes. The most stable photoresponse is demonstrated by device-E and the second most by device-E_aq at three reverse biases of −0.5, −1, and −2 V. At the reverse bias of −2 V, both device-E and device-E_aq exhibit similar maximum photocurrents of ∼4.2 mA, as shown in Figs. 2(b) and 2(d). However, the photocurrent is higher for device-E_aq at other voltages, owing to slight differences in their surface defects. This suggests that the reverse bias of −2 V deactivates the defect states. Similarly, device-M and device-I show similar maximum photocurrents of ∼24.5 mA at −2 V in Figs. 3(a) and 3(c), while as for other biases, device-M presents a relatively higher maximum photocurrent than device-I. However, for reverse biases of −0.1, −0.5, and −1 V, there is an increase in the onset current at every UV-on cycle for device-I, as observed in Figs. 3(o), 3(k), and 3(g), respectively, suggesting that the trapped charges contribute to the photocurrent, along with excitons. The voltages of −0.1, −0.5, and −1 V are, therefore, not sufficient to compensate for the defect states in device-I. A similar behavior is also shown by device-M at −0.5 V in Fig. 3(i). In fact, due to the high number of surface defects in the bare nanorods of device-M and device-I, the probability of charge trapping in ZnO nanorods, leading to instabilities in their photoresponse, is rather high. In addition, there are differences in the nanorod layer morphology or the random orientation of nanorods for device-I in Fig. 1(c) compared to other devices, in which all the ZnO nanorods are vertically oriented. This reduces the surface and the quality of contact of the nanorods with the polymer overlayer, creating instabilities in the output current at lower voltages in device-I. On the other hand, at a reverse bias of −0.1 V, device-M and device-E_aq show a drop in their photocurrents during subsequent UV-on/off cycles, as shown in Figs. 3(m) and 3(p), respectively. The current decrease is, therefore, attributed to the trapping of photogenerated carriers in the ZnO-nanorod defect states.

Figure 4 compares the R and EQE of these photodiodes in terms of the reverse bias voltage. The figure reveals that device-M and device-I demonstrate similar performances and EQE at the same reverse bias. In fact, they both have an EQE of ∼2760% at a reverse bias of −2 V. Similarly, device-E and device-E_aq show similar performances, with similar EQE of ∼600% at the same reverse bias of −2 V. However, compared to the EQE of device-E and device-E_aq, the EQE of device-M and device-I are much higher. The higher EQE of device-M and device-I are attributed to the higher surface defects in their bare nanorods. In fact, these defects are active centers for charge generation. A higher number of charge carriers leading to both recombination and diffusion currents produce a lower ideality factor of ∼2 and ∼1.8, respectively, as compared to device-E and device-E_aq. Ideality factors of <2 imply more efficient diode operation.

However, the response times, i.e., T_rise and T_decay of device-E and device-E_aq, tend to be quicker than that of device-M and device-I, listed in Table I. In fact, T_rise is directly dependent on the total UV irradiation time. On the other hand, T_decay is the intrinsic property of the device and depends on various parameters, such as generation–recombination centers at the p–n interface, the probability of recombination of excitons, charge trapping in the defects of bare ZnO nanorods, and the width of the depletion region.^13,30 In addition, the higher the applied voltage, the quicker is the T_decay. Therefore, device-E and device-E_aq display a better T_decay of ∼60 s, compared to the longer T_decay of ∼108 s for device-I and device-M at −2.0 V. For other voltages as well, the T_decay of device-E and device-E_aq is quicker than that of device-M and device-I. Therefore, in terms of T_decay, device-E and device-E_aq are more efficient than that of device-M and device-I. However, when selecting the best photodiode, several different parameters need to be considered. In that regard, device-M appears to have better ZnO-nanorod surface coverage, the highest rectification ratio, relatively stable on/off-cycles, a η of 2, as well as good S, R, and EQE. Therefore, in this study, device-M appears to be the most suitable candidate for photodiode applications.

We have successfully fabricated ZnO nanorods and polymer-based hybrid UV photodiodes with a device configuration of ITO/ZnO nanorod/F8BT/PEDOT:PSS/Ag. In general, the photodiodes with ZnO-nanorods grown with methanol and isopropanol seeding layers possess the same EQE, which is also the highest among the photodiodes in this study. In addition, lower reverse biases are insufficient for the proper working of photodiodes, and better performances are obtained at −2 V, demonstrating the highest stability during UV on–off cycles. Furthermore, photodiodes fabricated with ZnO nanorods harboring higher amounts of surface defects, i.e., ZnO nanorods in the methanol and isopropanol seed layers, exhibited a higher EQE of ∼2760% but also produced a longer T_decay at −2 V. This suggests that higher amounts of surface defects in nanorods support one figure of merit at the detriment of the other. Furthermore, the best performing photodiode in this study was selected by comparing several of their figures of merit. In that regard, the photodiode with the methanol seed layer was considered the best performing photodiode, as it possesses the highest rectification ratio of ∼640, a η of 2, the most uniform ZnO-nanorod coverage, a high EQE of ∼2760%, a R of ∼8.14 A/W, and a good T_decay of ∼108 s at −2 V. To the best of our knowledge, this work is the first to demonstrate the effect of the seed-layer solvents on the functionalities of the ZnO nanorod-polymer based UV photodiodes. Therefore, this study provides a novel methodology to fabricate UV photodiodes, consolidated by a detailed analysis of the influence of these defects harbored by the ZnO nanorods on the figures of merit of the UV photodiodes.

A step-by-step procedure for the fabrication of photodiodes is provided in Fig. S.1 of the supplementary material. Figures S.2(a) and S.2(b) are the dark I–V and semi-log(I)–V characteristics of three ethanol nanorod based devices that were prepared by varying F8BT amounts, showing 20 µl F8BT and 20 µl PEDOT:PSS used in this work are the optimum amounts for these devices. Figures S.3(a) and S.3(b) are the dark I–V and semi-log(I)–V characteristics of two sets of four photodiodes. In this study, we used the best of two sets of devices in terms of dark I–V characteristics to further evaluate their I–t characteristics.

This work was supported by ETAG Grant No. PRG 2115 and the French National Research Agency in the framework of the “investissements d’avenir” program (Grant No. ANR-15-IDEX-02).

The authors have no conflicts to disclose.

K.N. performed all the experiments and wrote the paper. F.D. performed and confirmed experiments. I.G. contributed to the experimental analysis. E.R. participated in the experiments, discussion, and writing. P.R. conceptualized the study, and K.N. performed experiments, supervised, provided funding, and wrote and reviewed the work.

Keshav Nagpal: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Visualization (equal); Writing – original draft (lead). Erwan Rauwel: Funding acquisition (lead); Methodology (supporting); Project administration (lead); Resources (lead); Supervision (supporting); Validation (supporting); Writing – review & editing (supporting). Frederique Ducroquet: Formal analysis (equal); Funding acquisition (lead); Investigation (supporting); Validation (supporting). Isabelle Gélard: Data curation (supporting); Formal analysis (equal); Investigation (equal). Protima Rauwel: Conceptualization (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.