Introduction

On the increasing needs of clean and renewable solar energy, researchers are continuously exploring novel materials and fundamentally investigating photoelectric conversion mechanisms for the better performance of photovoltaic (PV) devices1,2,3,4. Generally, two critical processes determined the photovoltaic effect: First, the electrical-charge carries such as electron-hole (e-h) pairs are generated by absorbing photons in active layers of the devices, i.e. semiconductors, dyes5,6. Second, the photo-generated e-h pairs are simultaneously separated by a built-in asymmetry potential formed in p-n/Schottky junction7 or two electrodes with different work functions8,9. Usually, the performance of a photovoltaic device can be evaluated by measuring its open-circuit voltage (Voc) and short-circuit current (Isc), both are strongly depended on the light absorption ability of active layer, the strength and space of built-in field. Unfortunately, the performance of the semiconductor-based PV devices is mainly limited in two aspects: low Voc (typically below 1 V) and narrow space-charge region of a p-n junction7,10. Recent studies have revealed that the above two limitations can be broken in ferroelectric photovoltaic (FEPV) devices11,12,13.

In a normal structure of electrode/ferroelectric/electrode, the ferroelectric could not only generate the photo-excited e-h pairs but also provide a polarization-induced internal electric field (Epi, also called as depolarization electric field) to separate the photo-excited e-h pairs. Furthermore, electrodes can form two back-to-back Schottky barriers at the interfaces between the ferroelectric material and the electrode. The built-in fields (Ebi) due to Schottky barriers can deplete the ferroelectric layers. Generally, FEPV effect could create very high open-circuit voltage, sometimes over kV, but the short-circuit current density is very small even if under irradiation of ultraviolet (UV) irradiation14. To realize potential applications, such as light-electrical energy conversion, wireless energy transfer for a micro-electro-mechanical system (MEMS), the performance of FEPV devices must be improved significantly.

Therefore, many studies have been carried out to improve the FEPV properties by optimizing both ferroelectric and electrode materials. For instance, the number of e-h pairs will be increased significantly by using relative narrow band-gap BiFeO3 (Eg = 2.2 eV) instead of wide band-gap ferroelectric15,16, or incorporating narrow band-gap semiconductors such as Ag2O or Cu2O into Pb(Zr,Ti)O3 (PZT) matrix11,17. In addition, the photocurrents and the photo-generated voltages can be further improved by modifying the configuration of electrodes in contact with ferroelectric materials11,18.

Electrons could be emitted from the surface of a metal into the vacuum under the irradiation of lights, where the photon energy is large than the work function of the metal. Such external photoelectric effect is the basis of classic photoelectric diodes as shown in Fig. 1a . Here, an applied field is necessary to collect excited electrons from the cathode. We noticed two facts in a classic photoelectric diode and a conventional ferroelectric photovoltaic cell: the cathode metals or alloys in the former have generally low work function. This allows high rate of electrons emitted into vacuum. On the other hand, the electrodes in the latter case are usually noble metals (i.e. Pt and Au) with high work function.

Figure 1
figure 1

Illustrations of (a) a classic photoelectric diode and (b) a proposal ferroelectric capacitor consisting of the ferroelectric layer and the photoemission electrode.

A: anode, C: cathode.

Full size image

FEPV effects' studies are mostly based on an architecture of electrode/ferroelectric film/electrode, where the thickness of ferroelectric film is from several to hundreds of nanometers and the electrodes are noble metals (i.e. Pt and Au)19,20,21. Based on classic photoelectric effect theory, fewer electrons in Pt or Au can be excited by light irradiating due to a large barrier height existing at the interface which hinders the emission of electrons into the ferroelectrics. Furthermore, two interfaces between the ferroelectric film and top/bottom electrodes will result in a very complex physical picture of the device. For instance, the interface and the concomitant Schottky barriers can make ferroelectric layer partially or even totally depleted22,23. It is difficult to distinguish whether the Epi or Ebi is dominant for the electron transport and photovoltaic properties. The open-circuit voltage (usually less than 1 V) of a ferroelectric film is influenced by the film's thickness11,24.

In the present work, we report a novel concept of electrode/ferroelectric bulk/electrode capacitor combining classic photoelectric and ferroelectric effects, as shown in Fig. 1b . Here, the thickness of the ferroelectric pellet in hundreds of micrometers is much thicker than the thickness of interfaces. Therefore, the transport of carriers is dominant by Epi instead of the Ebi. In details, three types of cells were prepared in this work. The ferroelectric material is lanthanum-modified lead zirconate titanate (PLZT) and the top transparent electrodes for all samples are tin-doped indium oxide (ITO) facilitating light illumination through the top electrode. To clarify the role of photoelectric effect of metal electrodes on the FEPV cells, three different metals with work function of ΦPt = 5.5 eV, ΦAg = 4.26 eV and ΦMg = 3.66 eV were used as bottom electrodes, respectively.

Both short-circuit current (Isc) and open-circuit voltage (Voc) of the cells were improved remarkably under AM1.5 irradiation. Note that carriers in conventional ferroelectric cells are photo-generated e-h pairs. In our devices where one of the metal electrodes is of low work function, electron should be the majority carrier due to the contribution of emitted electrons from metal, in addition to the e-h pairs.

Results

Figure 2 shows XRD pattern of a pure (Pb0.97La0.03)(Zr0.52Ti0.48)O3 (PLZT) pellet, illustrating single phase perovskite PZT structure with a lattice constant of a = b = 0.402 nm and c = 0.422 nm. The inset in Figure 2 is a cross-sectional SEM image of PLZT, displaying the polycrystalline feature of PLZT with grain size ~8–10 μm. It should be noted that an excess of 10% PbO added to account for loss of lead during processing, together with La3+ substitution for Pb2+, would ensure the final PLZT to be a n-type semiconductor25,26. This is very important to realize effective photoelectric effect to be discussed in this paper.

Figure 2
figure 2

XRD pattern of PLZT ceramic bulk.

Insert: Cross-section SEM image of PLZT ceramic bulk.

Full size image

Figure 3 shows ferroelectric hysteresis loops (P–E) of PLZT with different bottom electrodes. The inset schematically illustrates the device architecture. The measured remnant polarization (Pr) is around 45–50 μC/cm2, close to those values reported in the literature26. The coercive field of the sample is ~20 kV/cm. Our experimental results imply that the measured ferroelectric properties of capacitors are not a strong function of metal electrodes.

Figure 3
figure 3

P–E hysteresis loops of the PLZT photodiodes with different bottom electrodes, Pt, Ag and Pt.

Insert: Illustration of the cross-section of a ferroelectric bulk capacitor.

Full size image

To measure the photovoltaic properties of the PLZT cells, the devices with different electrodes were poled at an applied 1200 V for 15 minutes. A poling voltage well above the coercive voltage (~600 V) will ensure a full polarization of PLZT. In our experiment, we define the positive poling as the positive voltages applied to the bottom electrodes, i.e. Pt, Ag and Mg. The negative poling is defined as the positive voltages applied to the top electrodes, i.e. ITO. With the light irradiated from either the top ITO or the bottom of metal electrodes, the open-circuit voltage and short-circuit current vs. measuring time were recorded by a self-made software through the Keithley 6517A consequently.

With irradiating light from the top (ITO), the time-dependent short-circuit current density (Jsc) and Voc were shown in Figure 4 . The observations can be summarized as: (1) both Jsc and Voc strongly depend on the bottom metal electrodes. With decreasing of work function of metals (ΦPt > ΦAg > ΦMg), both Isc and Voc increase dramatically (Jsc-Pt < Jsc-Ag < Jsc-Mg and Voc-Pt < Voc-Ag < Voc-Mg); (2) The variation of Jsc and Voc on dependence of time follows the similar trend regardless of the bottom metal electrodes. Both Jsc and Voc increase rapidly in the first several seconds. They then saturate at a certain value. Those transient responses by voltage and current may be due to the duration of the electrons transporting from top to bottom electrode, or caused by the pyroelectic effect to the current27,28; (3) Either in positive or in negative polarization, both Jsc and Voc are almost symmetric. As mentioned early, we believe that with much thicker ferroelectric material, the carrier transport is controlled dominantly by the bulk ferroelectric effect rather than that of the localized Schottky barriers at interfaces.

Figure 4
figure 4

(a) Short-circuit current density and (b) open-circuit voltage vs. measuring time (Light irradiating from ITO).

Full size image

We would like to point out that the short-circuit current and open-circuit voltage of Mg/PLZT/ITO are about 150 and 2 times respectively larger than those of Pt/PLZT/ITO. Although the mechanism of ferroelectric photovoltaic effect has not been fully understood, it has been speculated that ferroelectric photovoltaic effect is induced by the depolarization electric field which separates the photogenerated charge carriers22,29. Clarifying the relationship between Jsc and Voc can help to understand above results. In a photovoltaic device, Jsc arises from the movement of carries in the PLZT, but Voc comes from different chemical potential between the positive charges and negative charges piled on the top and bottom electrodes, respectively30. Similar to a p-n junction solar cell, the relationship between Jsc and Voccan be expressed as31:

Where n is the ideality factor of the diode, k is Boltzmann constant, T is temperature in Kelvin, q is the electron charge and J0 is reversal saturation current density of p-n junction. The much enhanced Jsc will result in larger Voc if the J0 is constant.

Usually, PLZT will exhibit photovoltaic effects under near-ultraviolet illumination due to its wide band-gap (λc = 365 nm). We have noted that the fractional of light with photon energy greater than the band-gap of PLZT in our light source is about 5%, corresponding to a power density ~7.5 mW/cm2. Therefore, such a large enhancement in Jsc and Voc in Mg/PLZT/ITO cannot be simply attributed to the absorption of light by PLZT itself. The metal electrode must play a key role in determining the device performance. We expect that the electrons in metals can be excited by light irradiation and could be emitted into PLZT matrix, similar to the classic photoelectric effect. Under such circumstances, two kinds of carriers can contribute to the total photocurrent density Jsc: one is defined as Jf, generated from e-h pairs in n-type PLZT matrix and another is Jm, generated from electrons in metal. Since the work function of Pt is high (5.5 eV), it is reasonable to ignore Jm. The Jf is ~22 pA/cm2 in a structure of Pt/PLZT/ITO. It is reasonable to assume that Jf values is ~22 pA/cm2 regardless of the electrodes used. Therefore, Jm dominates the total photogenerated current in both Ag/PLZT/ITO (Jsc = 1390 pA/cm2) and Mg/PLZT/ITO (Jsc = 3445 pA/cm2) cells. The results imply that the much enhanced photovoltaic effect in the devices is from the photoelectric effect.

To further verify our concept, we measured photovoltaic properties of metal/PLZT/ITO cells by illuminating light from the metal side. As shown in Figure 5 , the characteristics are similar to those shown in figure 4 but the values are much smaller. Since the thickness of metals (~100 nm) is thicker than the “skin depth”, no light could penetrate into PLZT. The measured photocurrent and voltages should come from the electrons emitted from metal electrodes.

Figure 5
figure 5

(a) Short-circuit current density and (b) open-circuit voltage vs. measuring time (Light irradiating from metal).

Full size image

Similar experiment was carried on Mg/PLZT/ITO. In this case, we inserted a filter JB420 between the light source and the sample during the measurement. This setup will allow the light with wavelength greater than 420 nm shine on the sample only. For PLZT with a band-gap of 3.4 eV, photocurrent can be generated by UV light at wavelengths shorter than 364 nm32. However, there still has considerable photocurrent even if there is no contribution from the photogenerated carriers in PLZT as presented in Figure 6 . Our results reveal that the visible light could generate photocurrent in Mg/PLZT/ITO device, confirming that the photocurrent in Mg/PLZT/ITO can come from the emitted electrons in Mg electrode.

Figure 6
figure 6

Short-circuit current density and open-circuit voltage vs. measuring time for Mg/PLZT/ITO.

Full size image

Figure 7 shows the temporal dependence of the photocurrent density and voltage of Mg/PZT/ITO cell. It is clear that there is no decay of photocurrent and voltage during 3 cycles of on-off the illumination light. Figure 8 shows the dark and illuminated J-V curves for Mg/PZT/ITO cell. Under standard AM 1.5 illumination condition, Voc and Jsc are 8.34 V and 3.25 nA/cm2 for lighting from ITO, 3.58 V and 1.08 nA/cm2 for lighting from Mg, respectively. Above results indicate a significant photovoltaic effect in Mg/PZT/ITO cell33.

Figure 7
figure 7

Short-circuit current and open-circuit voltage vs. measuring time for Mg/PLZT/ITO.

Full size image
Figure 8
figure 8

J–V characteristics for Mg/PZT/ITO capacitor in dark and illumination conditions.

Full size image

Discussion

To explain our experimental results, we construct a simple model. Figure 9 illustrates the energy band diagram of Metal/PLZT/ITO. The internal electric field in ferroelectric material can be classified as two independent components and written as: E = Ebi + Epi, where Ebi is the net built-in field of two back-to-back Schottky contacts at the top and bottom interfaces. The Ebi is the internal net potential (ΔΦ = ΦITO − ΦM) in the device. If we take the work function of ITO as ~4.5 eV34, the theoretical ΔΦ is about −1, 0.2 and 0.6 eV for Pt, Ag and Mg, respectively. Considering the depletion layer thickness of Schottky contact is less than 200 nm35 and the measured Voc is much larger than 1 V, the influence of localized Ebi on the photovoltaic properties is less important than Epi which is across the whole region of PLZT.

Figure 9
figure 9

Illustration of the energy band graph of Metal/PLZT/ITO, where E0 is vacuum energy band, Ec is conduct band, Ev is valance band and Ef is Fermi level: (a) Before polarization, (b) and (c) after negative and positive polarization (the e-h pairs photo-generated in PLZT weren't indicated in graph).

Full size image

The most important issue of this work is to clarify how the photoelectric effect of metal can improve the photovoltaic properties of a ferroelectric material. Table I list the most important parameters discussed in the article. Here the work function of PLZT is considered as 3.5 eV11,36. As the barrier height can be expressed as ΦB = ΦM − ΦPLZT, it is clear that the lower the work function of the metal is, the smaller of ΦB is. Theoretically, the photon energy large than 2, 0.8 and 0.2 eV can enable to emit electrons from Pt, Ag and Mg into the PLZT, respectively. However, due to the reflection loss of light on the surface and interfaces, the existing of surface states37 and the trapping of charge carriers in the defects38, the final Jsc is small. With the decrease of work function, electrons in Ag and Mg could be excited by light irradiation and are emitted into the PLZT. Different from the e-h pairs generated in PLZT, most of charge carriers in Ag/PLZT/ITO and Mg/PLZT/ITO are electrons. Compared with a conventional p-n junction diode, the recombination of e-h pairs in such capacitors can be reduced significantly due to that the photo-emitted electrons in PLZT are dominant.

Table 1 The data concerning different bottom metals
Full size table

In summary, we have demonstrated much enhanced photovoltaic effect in metal/PLZT/ITO cells by combining classic photoelectric and photovoltaic effects. By lowering the height of Schottky barrier at metal/PLZT interface, the electrons in metals could be emitted and contribute to the ferroelectric photovoltaic effect. By using a low work function metal as the electrode, i.e. magnesium, the short-circuit current density and open-circuit voltage of Mg/PLZT/ITO cell is about 150 and 2 times larger than those of Pt/PLZT/ITO cell under AM1.5 illumination. The photovoltaic response of PLZT capacitor was expanded from ultraviolet to visible spectra and its performance could be further improved by optimizing the thickness and crystalline quality of ferroelectric materials as well as using an antireflection layer to reduce light losses.

Methods

Synthesis

The detailed processing conditions to synthesize ceramic PLZT were described elsewhere39. After hot-pressing calcinations at 1240°C under a pressure 40 MPa, both sides of the sample (or pellet) were polished. The pellet was then cut into a size of 10 × 10 × 0.3 mm3. Electrodes (ITO, Pt, Ag and Mg) with a thickness around 100 nm were deposited by radio-frequency (RF) magnetron sputtering at room temperature.

Measurements

The crystallographic structure of the PLZT samples was examined by X-ray diffraction (XRD, Rigaku D-MAX diffractometer with Ni filtered Cu Kα radiation). Ferroelectric hysteresis loops were recorded by using a ferroelectric analyzer (Radiant 609B-3, USA). Open-circuit voltage and short-circuit current were measured under illumination of a 150 W Xe bulb which is used to simulate the solar spectrum of AM 1.5(100 mW/cm2)20,40. The thickness of electrodes was determined by scanning electron microscopy (SEM).