Large enhancement of the photovoltaic effect in ferroelectric complex oxides through bandgap reduction

Abstract

Tuning the bandgap in ferroelectric complex oxides is a possible route for improving the photovoltaic activity of materials. Here, we report the realization of this effect in epitaxial thin films of the ferroelectric complex oxide Bi_3.25 La_0.75 Ti₃ O₁₂ (BLT) suitably doped by Fe and Co. Our study shows that Co (BLCT) doping and combined Fe, Co (BLFCT) doping lead to a reduction of the bandgap by more than 1 eV compared to undoped BLT, accompanied by a surprisingly more efficient visible light absorption. Both BLCT and BLFCT films can absorb visible light with a wavelength of up to 500 nm while still exhibiting ferroelectricity, whereas undoped BLT only absorbs UV light with a wavelength of less than 350 nm. Correlated with its bandgap reduction, the BLFCT film shows a photocurrent density enhanced by 25 times compared to that of BLT films. Density functional theory calculations indicate that the bandgap contraction is caused by the formation of new energy states below the conduction bands due to intermixed transition metal dopants (Fe, Co) in BLT. This mechanism of tuning the bandgap by simple doping can be applied to other wide-bandgap complex oxides, thereby enabling their use in solar energy conversion or optoelectronic applications.

Large enhancement of the complex oxides through bandgap reduction

OPEN

R A

P

Hyunji An, JunYoung Han, Bongjae Kim, Jaesun Song, SangYun Jeong, Cesare Franchini, ChungWung Bark & Sanghan Lee

Tuning the bandgap in ferroelectric complex oxides is a possible route for improving the photovoltaic ferroelectric complex oxide BiLaTiO (BLT) suitably doped by Fe and Co. Our study shows that

Co (BLCT) doping and combined Fe, Co (BLFCT) doping lead to a reduction of the bandgap by more than bandgap contraction is caused by the formation of new energy states below the conduction bandsdue to intermixed transition metal dopants (Fe, Co) in BLT. This mechanism of tuning the bandgap by simple doping can be applied to other wide-bandgap complex oxides, thereby enabling their use in solar energy conversion or optoelectronic applications.

Solar energy conversion devices based on ferroelectric photovoltaics can potentially exceed the maximum efficiency of conventional pn junction solar cells15. In ferroelectrics, oriented dipole moments eectively improve the mobility of charge carriers that are generated by light absorption6. Despite this advantage, ferroelectrics are not applicable to photovoltaic devices owing to their large energy bandgaps (Eg), which lead to insufficient light absorption and consequently limit the photocurrent7. Conventional ferroelectric materials (e.g., BaTiO3, LiNbO3, and Pb(Zr,Ti)O3 crystals) exhibit large bandgaps of more than 3 eV. Thus, they can absorb mostly UV light. However, UV light (less than 400nm) comprises only 3.5% of solar radiation intensity, whereas visible light (400700nm) comprises 40% of the solar irradiation6. Therefore, to obtain enhanced photocurrents, ferroelectric materials with narrow Eg are necessary for high light absorption including visible light.

The optical bandgap can be adjusted by modulating the composition of ferroelectrics. With the exception of Mott insulators, the wide Eg of ABO3-type perovskite ferroelectric materials are governed by the charge transfer from the O 2p state to the transition metal (B cation) d state8. Therefore, most studies on the adjustment of Eg have focused on B-site substitution by a transition metal. A Bi4Ti3O12 (BiT)/LaCoO3 (LCO) superlattice thin lm is a notable example. The Eg of ferroelectric BiT is decreased from 3.55 to 2.65eV through the site-specic substitution of the Co ion on the B site of the perovskite octahedral (BO6) between the BiT and LCO interfaces3. However, this approach may not be preferred for practical photovoltaic applications because fabrication of this material requires the precise control of superlattice periodicity and the use of a complex process with multiple targets. On the other hand, a conventional doping approach is widely used for tuning Eg because of its easy process compared to that required for the fabrication of superlattice thin lms.

Korea. Faculty of Physics requests for materials should be addressed to C.W.B. (email: [email protected]) or S.L. (email: sanghan@gist. ac.kr)

SCIENTIFIC REPORTS

1

www.nature.com/scientificreports/

Figure 1. Crystal structures of the BLT, BLCT, and BLFCT lms. (a) The crystal structure of BLT. The structure shows that La substitutes specic Bi sites of BiT. (b) The X-ray 2 diraction patterns for BLT, BLCT, and BLFCT epitaxial lms on STO (001). The full width at half maximum (FWHM) for the (008) peak in (c) the BLCT and (d) BLFCT lms on STO.

Based on this point of view, we studied the adjustment of the Eg of a ferroelectric Bi3.25La0.75Ti3O12 (BLT) film using a simple doping method based on a theoretical study. Here, we report Eg-tuned Co-doped BLT (Bi3.25La0.75Co1Ti2O12, BLCT) and Fe, Co-doped BLT (Bi3.25La0.75Fe0.25Co0.75Ti2O12, BLFCT) thin lms deposited using pulsed laser deposition (PLD). The bandgaps of the BLCT and BLFCT lms are lowered to 2.58 and 2.48eV, corresponding to a bandgap reduction of ~1eV. In particular, the photocurrent density of BLCT is improved by 6 times relative to that of BLT while maintaining ferroelectricity despite the random doping of the B site in ABO3,

which is dierent from the case of the superlattice. Furthermore, the photocurrent density of BLFCT was 25 times higher than that of BLT. The large enhancement of the photocurrent density in BLCT and BLFCT conrms the signicant reduction of the bandgaps by simple doping.

Crystal structure and ferroelectricity. Figure1(a) shows the half unit cell of BLT9. BLT crystalizes in the so-called Aurivillius phase, which consists of the bismuth oxide (Bi2O2)2+ layer and the pseudo-perovskite (Bi2Ti3O10)2 layer containing the TiO6 octahedra with La substituting for Bi in the pseudo-perovskite layers. The epitaxial crystalline quality of the grown lms was determined by X-ray diraction (XRD) using a Cu K source ( = 1.5405). Figure1(b) shows the 2 scan of the BLT, BLCT, and BLFCT thin lms grown on a (001)-oriented SrTiO3 (STO) substrate. The XRD patterns reveal that all lms were grown in the 00l orientation and that the BLT crystal structure was maintained for the BLCT and BLFCT lms. This indicates that Co and Fe doping of BLT does not cause the formation of other phases and that the dopants may be substituted into the BLT structure. Rocking curves for the (008) reection were measured to determine the out-of-plane mosaic spread and the crystalline quality. As shown in Fig.1(c,d), the full widths at half maximum (FWHM) of the (008) reection rocking curves of BLCT and BLFCT are 0.12 and 0.13, respectively. This indicates that despite the Co and Fe doping, all lms exhibit reasonable out-of-plane crystallinity.

We measured the polarization (P) as a function of the electric eld (E) of the BLT, BLCT, and BLFCT lms, as shown in Fig.2. The lms are deposited on the (110)-oriented STO substrate covered with a SrRuO3 bottom electrode for an accurate comparison of ferroelectricity because the polarization orientation of BLT is in-plane and it is difficult to examine the ferroelectricity of the c-axis oriented lm10. As shown in Fig.2(a), the BLCT and BLFCT lms exhibit ferroelectricity, showing the hysteresis polarization loops, with remanent polarization (Pr) values of 2.25 and 2.60 C/cm2 for the BLCT and BLFCT lms, respectively. In particular, the BLCT lm maintains ferroelectricity at various frequencies, as demonstrated in Fig.2(b). To summarize, although they may exhibit structural perturbations due to doping, the BLCT and BLFCT lms maintain the Aurivillius phase of BLT.

Optical properties. The optical properties of the BLT, BLCT, and BLFCT lms were measured using an ultraviolet-visible (UV-vis) spectrometer. Figure3(a) shows the transmittance as a function of the wavelength

SCIENTIFIC REPORTS

2

www.nature.com/scientificreports/

Figure 2. Ferroelectric properties of the BLT, BLCT, and BLFCT lms. (a) The polarization hysteresis loops were measured using a radiant system at a frequency of 10kHz for the BLT, BLCT, and BLFCT lms on the STO (110) substrates at room temperature. (b) The polarization hysteresis loops for the BLCT lm measured in the frequency range from 1 to 50kHz. The loops are maintained at various frequencies.

Figure 3. Optical properties of the BLT, BLCT, and BLFCT lms. (a) The transmittance data depending on the wavelength (200800nm) for BLT, BLCT, and BLFCT on the STO (001) substrates. (b) The bandgap energy was estimated by extrapolating the linear part of the (h)2 versus energy plots for BLCT and BLFCT on STO. The inset shows absorbance as a function of energy for BLCT and BLFCT on STO.

(200800 nm) for the BLT, BLCT, and BLFCT films. Inspection of Fig.3(a) shows that BLCT and BLFCT absorbed visible light in the 380560nm wavelength range, whereas BLT scarcely absorbed visible light owing to its large bandgap (3.23.8eV)11,12. The plots of (h)2versus the photo-energy measured for the BLT, BLCT, and BLFCT thin lms on STO (001) are shown in Fig.3(b). The optical bandgaps of the lms were estimated using an extrapolation method proposed by Wood and Tauc13. The estimated bandgaps of BLCT and BLFCT were 2.58 and 2.48eV, respectively. Because the transmittance of the lms cannot be measured using UV-vis spectroscopy when the Eg of the substrates are lower than those of the lms, the transmittances of the BLT lms grown on the

STO substrate (Eg of STO = 3.2 eV) cannot be obtained, as shown in Fig.3(a), demonstrating thickness fringes from 380 to 800nm only. Hence, we also measured the transmittance of the BLT lm deposited on a 001-oriented LaAlO3 (LAO) substrate because the bandgap of LAO (Eg= 5.6eV) is much larger than the BLT bandgap. Based on the measured transmittance data of the BLT lms on LAO, the Eg was estimated as 3.59eV, consistent with the reported value (see Supplementary Fig. S1)11,12. The Eg values of BLCT and BLFCT are respectively, 28% and 31% lower than the experimentally obtained Eg of BLT, as shown in Fig.3(b).

It is well known that the substitution site is determined by the following two factors: the ionic size and the ionic state. Both Fe and Co are transition metals with 3d orbital states that are identical to that of Ti. Co2+/3+ and

Fe2+/3+ are stable ionic states, and their ionic sizes are smaller than that of Bi3+ but are similar to that of Ti4+. Because of this similarity in the ionic size, Co and Fe ions may prefer to substitute at the Ti sites in the perovskite blocks rather than at the Bi sites. However, their ionic states dier from that of Ti4+. Thus, owing to the mismatch of ionic states, it is not certain that Co and Fe ions can replace Ti ions. However, Choi et al. conrmed that Co can replace Ti in a BiT/LCO superlattice lm, using both the experimental result and density functional theory (DFT) electronic calculation3. Fe could also substitute for Ti in the perovskite blocks randomly, as was veried in the case of the fabrication of a Bi4Ti3O12 lm doped with LaFeO314,15. Consistent with these previously reported results, Co

SCIENTIFIC REPORTS

3

www.nature.com/scientificreports/

Figure 4. Electronic structures of the BLT, BLCT, and BLFCT lms. (a) The DOS of each lm. For the BLCT and BLFCT cases, cases A and B are shown. The bandgap is governed by the highest occupied O-p and lowest unoccupied transition metal d levels. (b) Magnication of the DOS near the Fermi level. The lowest unoccupied states are determined mostly by doped Co- and Fe-d states. (c) The eect of doping on the overall electronic levels. The bandgap is due to the d-levels of the doped transition metal ions.

and Fe ions appear to replace Ti ions at the perovskite blocks of BLT in our experiment, as we conrmed that the BLCT and BLFCT bandgaps are narrower than that of the BLT lm.

Many studies have been performed on bandgap tuning of BiT thin lms by transition metal doping16,17.

Among these, the bandgap of a BiT-LCO superlattice lm was signicantly reduced, conrming that Co ions can be the most appropriate dopants16. In the case of the BiT-LCO superlattice, the bandgap could be reduced because of the occurrence of Co ion substitution for the Ti ions in the (outer) perovskite blocks near the Bi2O2layer owing to the interface interaction between the BiT and LCO layers3. We expected that not only site-specic substitution in the superlattice but also random intermixing of Co by doping could decrease the bandgap while maintaining the ferroelectricity. The optical bandgap of BLCT, 2.58eV, is similar to the bandgap (2.65eV) of the superlattice lm. Consequently, the bandgap of BLT can be tuned using the intermixing approach, which is much simpler than the site-specic substitution approach.

Furthermore, the bandgap of the BLFCT lm decreased more than that of BLCT. Fe and Co dopants are expected to be located randomly at the Ti sites in perovskite blocks based on their mixing ratio (1:3) due to their similar ionic sizes and stable ionic states. Therefore, additional doping of Fe may reduce the bandgap of BLT more than that of Co.

Electronic structure. To reveal the mechanism of the bandgap reduction upon doping, we analysed the electronic structure using DFT calculations. Because of the mismatch between the oxidation states of Ti (4+) and doped ions (+2/+3 for Co and Fe), oxygen vacancies occur upon doping. We found that the most probable position of an oxygen vacancy was between the Bi2Ti3O10 perovskite blocks (indicated by the blue arrow in

Fig.1(a)) by more than 140meV per formula unit compared to other possible positions, consistent with previous studies18,19.

In addition to the position of the oxygen vacancy, the position of the substitutional site is an important factor. The previous site-specic doping approach uses the superlattice structure to ensure that the dopant ion replaces a Ti ion at the outer position (case A) of a perovskite block (see Fig.1(a))3. In our simpler intermixing approach, however, doped ions can be located on the inner site (case B) of a perovskite block. When no oxygen vacancies exist, doped ions prefer case B to case A. However, because the 4+ charge states are unlikely for both Fe and Co, the gap reduction was found to be highly dependent on the dopant position. By introducing oxygen vacancies, the much more stable 3+ charge state can be obtained for both Co and Fe with enhanced robustness of the gap size against dierent congurations of the dopant positions.

In Fig.4(a,b), we show the density of states (DOS) of all three cases: undoped BLT, BLCT, and BLFCT. In the second and third cases, we included oxygen vacancies for both congurations of dopant ions (cases A and B). Inspection of the DOS of the pristine BLT shows that as expected, the overall bandgap is determined mostly by the occupied O-p level and unoccupied Ti-d level. The calculated bandgap size is 2.22eV, somewhat smaller than the experimental value due to usual underestimation within the DFT-based approach but consistent with previous calculations3. When BLT is doped by Co (BLCT), clear reductions of the bandgaps by 1.86 and 1.04 eV for cases A

SCIENTIFIC REPORTS

4

www.nature.com/scientificreports/

Figure 5. Photovoltaic responses of the BLT, BLCT, and BLFCT lms. (a) The photocurrent density (J) of the BLT, BLCT, and BLFCT lms on STO (001). The dark current of the BLT, BLCT, and BLFCT lms (grey, dark grey, and black). The light current of the BLT lm (green). The light current of the BLCT lm (blue). The light current of the BLFCT lm (red). (b) The schematic of the photocurrent measurement. The ITO electrodes were deposited on the lms, and J was measured between the two electrodes. (c) The schematic of the light wavelength range that the BLT, BLCT, and BLFCT lms can absorb, which is predicted using the bandgap data.

and B, respectively, can be seen. This is due to the energy levels of the unoccupied Co-d states. Doped Co3+ has a nonmagnetic phase with t2g6 electronic congurations where the unoccupied eg level is located lower than the Ti-d levels of undoped BLT. In cases A and B, the reductions of the gap are 16% (from 2.22 to 1.86eV) and 53% (from 2.22 to 1.04eV), respectively. We assume that the position of the dopant ion is important for the eective reduction of the bandgap. Because the experimental result shows that the reduction is approximately 28% (from 3.59 to 2.58 eV), we assume that the intermixing approach results in a mixture of dierent congurations including cases A and B. The overall schematic diagram of the doping eects is shown in Fig.4(c). By comparing the DOSs for cases A and B, we nd that the location of the unoccupied Co-d states, which determine the magnitude of the Eg of BLCT, is highly dependent on the geometric position of the dopant Co ion within a perovskite block. While case B shows a split-o of Co-d (eg) states from Ti-d states to form mid-gap-like electronic states, case A shows large hybridization of Co-d and Ti-d without the mid-gap phase, consistent with a recent optical conductivity study on the supercell structure where no split-o Co-d states are found20.

In our UV-vis data (the inset in Fig.3(b)), we can see the shoulder region near 2.6 eV, which was absent for pristine BLT (see Fig. S1). We can relate this to the Co-eg states that are located closer to the Fermi level (Fig.4(b)).

The energetics shows that case B is more stable than case A with a dierence of 116 meV/f.u.; however, due to the sample fabrication process at a high temperature, two and more phases can mix, as we can conclude from the previous analysis of the bandgap data. Although the number of oxygen vacancies is not dened, considering that the 2+ and 3+charge states can both eectively reduce Eg3, Co doping is an efficient method for tuning the optical gap of a BLT system. We also investigated the eect of additional Fe doping (BLFCT), as described before. The calculated bandgaps for cases A and B are 1.56 and 1.51eV, respectively; these values are close to each other and are consistent with the experimentally observed bandgap reduction of 31%.

From the DOSs of both congurations (the fourth and h panels in Fig.4(a,b), respectively), we can observe that the overall energy levels of the transition metal ions do not change signicantly. Fe has a 3+ charge state with high-spin congurations (t2g3eg2) where unoccupied d-states are well hybridized with Ti- and Co-d orbitals without forming the mid-gap state, as schematically shown in Fig.4(c). This is also consistent with our UV-Vis data shown in the inset of Fig.3(b).

In the case of BLCT, the curve shows a steep slope starting from 3.0 eV, which is suddenly attened near the shoulder position (2.6 eV), suggesting the presence of mid-gap states. In the case of BLFCT, no such shoulder or peak is observed, with the overall form of the curve exhibiting a more gradual character than that in the case of BLCT. In general, the t2gt2g orbital hopping integrals are larger than those for the t2geg orbitals21, providing better hybridization for BLFCT than for BLCT. The electronic structures for cases A and B are very similar for BLFCT compared to BLCT, suggesting that the BLFCT sample is more robust against the disorder of the dopant ions than BLCT. Further studies using dierent dopants with various electronic states are needed to elucidate the microscopic mechanism of the gap reduction.

Photocurrent response. The photocurrent density (J) was measured to conrm the eect of narrowing the bandgaps of the BLCT and BLFCT lms. As shown in Fig.5(a), the BLT, BLCT, and BLFCT lms exhibited nearly zero current without light irradiation (dark current), while their photocurrent responses under light irradiation were clearly enhanced. Compared to the J of BLT (5.71 nA/cm2), the J of BLCT (34.58 nA/cm2) was improved 6 times and the J of BLFCT (141.92 nA/cm2) was improved 25 times, as shown in Fig.5(a,c). We note that a transparent conducting oxide, indium tin oxide (ITO), was used as the electrode on the lm to maximize

SCIENTIFIC REPORTS

5

www.nature.com/scientificreports/

the light-irradiated region. The photocurrent was measured between the ITO electrodes on the lms at a voltage from 0 to 5V under the light eld, as shown Fig.5(b).

These results conrm that the bandgap narrowed by doping also improves the photocurrent density of BLT. Conclusively, this indicates that the BLCT and BLFCT lms have enhanced optical properties and can absorb more photons eectively. Therefore, they can be applied to various devices such as photovoltaics.

Conclusion

Here, we successfully narrowed the bandgap of BLT by more than 1eV while maintaining ferroelectricity through a simple Fe, Co-doping approach. The results of the experiments conducted using a Co-doped BLT (BLCT) lm show that the simpler intermixing approach, and not only the site-specic approach, can tune the bandgap of BLT. We also observed that additional Fe doping contributed to the decrease in the bandgap of BLT and improved the photocurrent density by 25 times relative to that of BLT. This approach can be used as a method for tuning the bandgaps of a variety of perovskite ferroelectric lms by examining their substitution mechanisms and dopant eect on the decrease in their bandgaps. These results indicate that ferroelectrics with large bandgaps can generally be more easily applied for use in photovoltaic devices.

Methods

Sample fabrication. All targets for the film depositions were synthesized using a solid-state reaction method with the following starting binary oxide powders as raw materials: Bi2O3 (99.9%, Kojundo Chemical

Co. Ltd., Japan), TiO2 (99.99%, Kojundo Chemical Co. Ltd., Japan), La2O3 (99.99%, Kojundo Chemical Co. Ltd., Japan), Co3O4 (99.99%, Kojundo Chemical Co. Ltd., Japan), and Fe2O3 (99.9%, Kojundo Chemical Co. Ltd., Japan). Based on our previous results, the doping concentrations of Co and Fe atoms in the doped BLT targets were chosen to maximize the bandgap reduction, and the doping ratios were Bi3.25La0.75Co1Ti2O12 (BLCT) and

Bi3.25La0.75Fe0.25Ti2O12 (BLFCT)22,23. The BLT, BLCT, and BLFCT powders were blended thoroughly at the stoichiometric ratio in a ball mill for 24h, dried in an oven at a temperature of 100C, and calcined at 700C for 2.5h in air. The powders were pressed for 5min at a pressure of 50MPa and then sintered at 950C for 2.5h.

The BLT, BLCT, and BLFCT epitaxial thin lms (thickness of 250nm) were fabricated using pulsed laser deposition (KrF excimer laser, = 248nm), operating at a repetition rate of 10Hz, on SrTiO3 (001) substrates under 80 sccm O2 gas ow. The operating temperature for the BLT, BLCT, and BLFCT lms was 750C, and the laser energy density for these lms was 2.0J/cm2. In particular, the lms were deposited at a high frequency of 10Hz to prevent the loss of the volatile Bi.

The structural properties were analysed by XRD using Cu K radiation (Xpert PRO, PANalytical, Netherlands). The polarization hysteresis loop was measured by a ferroelectric test system (Precision Multiferroic, Radiant Technologies, Inc., USA). For the polarization hysteresis loop measurement, the BLT, BLCT, and BLFCT lms were deposited on a SrRuO3 bottom electrode covered with STO (110) substrates.

Then, a Pt top electrode (diameter of 150m) was deposited on the lms using RF sputtering. The polarization loops of the lms were measured in the c-plane orientation between the bottom electrode (SrRuO3) and the top electrode (Pt). The optical properties were analysed using an ultraviolet-visible (UV-Vis) NIR absorption spectrometer (Cary500Scan, Varian, Australia). The optical properties of the lms were measured under light with wavelengths ranging from 200 to 800nm. We use same thickness for all three lms (pristine BLT, BLCT, BLFCT) to avoid the absorption coefficient shi due to the thickness eect24. The photocurrent density was measured using a sourcemeter (SourceMeter 2410, Keithley, United States). The measurement was implemented under light illumination using a 100 W solar simulator (K3400, McScience, Korea). ITO transparent electrodes were deposited on the surface of the lms with a separation of 100m using DC sputtering. The photocurrent densities of the lms were measured between the ITO transparent electrodes aer contacting two tips with the respective electrodes.

Electronic structure calculations were performed using the projector-augmented-wave method implemented in the Vienna ab initio simulation package (VASP)25,26. We adopted the generalized gradient approximation (GGA) functional in the PerdewBurkeErnzerhof parameterization including an on-site Coulomb parameter U= 5eV for the d-orbitals of both Co and Fe within the Dudarev approach27,28,

based on previous studies in the literature on similar systems3,29. We employed supercells containing 4 formula units: Bi16Ti12O48 for BLT, Bi16Ti8Co4O46 for BLCT and Bi16Ti8Fe2Co2O46 for BLFCT. Lattice parameters were xed to the experimental values, and all internal atomic positions were fully relaxed for all cases. A plane-wave cut-o of 400eV was used with a 44 1 MonkhorstPack k-point sampling mesh.

References

1. Grinberg, I. et al. Perovskite oxides for visible-light-absorbing ferroelectric and photovoltaic materials. Nature 503, 509512 (2013).2. Nechache, R. et al. Bandgap tuning of multiferroic oxide solar cells. Nat. Photonics 9, 6167 (2014).3. Choi, W. S. et al. Wide bandgap tunability in complex transition metal oxides by site-specic substitution. Nat. Commun. 3, 689 (2012).

4. Zhang, G. et al. New high Tc multiferroics KBiFe2O5 with narrow band gap and promising photovoltaic eect. Sci. Rep. 3, 1265 (2013).

5. Yang, S. Y. et al. Above-bandgap voltages from ferroelectric photovoltaic devices. Nat. Nanotechnol. 5, 143147 (2010).6. Yuan, Y., Xiao, Z., Yang, B. & Huang, J. Arising applications of ferroelectric materials in photovoltaic devices. J. Mater. Chem. A 2, 60276041 (2014).

7. Nelson, J. The physics of solar cells. Vol. 1 (World Scientic, 2003).8. Qi, T., Grinberg, I. & Rappe, A. M. Band-gap engineering via local environment in complex oxides. Phys. Rev. B 83, 224108 (2011).9. Park, B. et al. Lanthanum-substituted bismuth titanate for use in non-volatile memories. Nature 401, 682684 (1999).10. Lee, H. N. & Hesse, D. Anisotropic ferroelectric properties of epitaxially twinned Bi3.25La0.75Ti3O12 thin lms grown with three

dierent orientations. Appl. Phys. Lett. 80, 1040 (2002).

SCIENTIFIC REPORTS

6

www.nature.com/scientificreports/

11. Wang, G. et al. Structural and optical properties of Bi3. 25La0. 75Ti3O12 ferroelectric thin lms prepared by chemical solution methods. Appl. Phys. A 76, 8386 (2003).

12. Hu, Z., Li, Y., Yue, F., Zhu, Z. & Chu, J. Temperature dependence of optical band gap in ferroelectric Bi3.25La0.75Ti3O12 films

determined by ultraviolet transmittance measurements. Appl. Phys. Lett. 91, 221903 (2007).13. Wood, D. & Tauc, J. Weak absorption tails in amorphous semiconductors. Phys. Rev. B 5, 3144 (1972).14. Wu, F. X. et al. Signicant ferrimagnetism observed in Aurivillius Bi4Ti3O12 doped by antiferromagnetic LaFeO3. Appl. Phys. Lett. 98,

212501 (2011).

15. Krzhizhanovskaya, M. et al. Aurivillius phases in the Bi4Ti3O12/BiFeO3 system: Thermal behaviour and crystal structure. Z. Anorg. Allg. Chem. 631, 16031608 (2005).

16. Choi, W. S. & Lee, H. N. Band gap tuning in ferroelectric Bi4Ti3O12 by alloying with LaTMO3 (TM= Ti, V, Cr, Mn, Co, Ni, and Al). Appl. Phys. Lett. 100, 132903 (2012).

17. Jia, C., Chen, Y. & Zhang, W. F. Optical properties of aluminum-, gallium-, and indium-doped Bi4Ti3O12 thin lms. J. Appl. Phys. 105, 113108 (2009).

18. Hashimoto, T. & Moriwake, H. Oxygen vacancy formation energy and its eect on spontaneous polarization in Bi4Ti3O12: a rst-principles theoretical study. Phys. Rev. B 78, 092106 (2008).

19. Noguchi, Y., Soga, M., Takahashi, M. & Miyayama, M. Oxygen stability and leakage current mechanism in ferroelectric La-substituted Bi4Ti3O12 single crystals. Jpn. J. Appl. Phys. 44, 6998 (2005).

20. Choi, W. S. & Lee, H. N. Strain tuning of electronic structure in Bi4Ti3O12 LaCoO3 epitaxial thin lms. Phys. Rev. B 91, 174101 (2015).

21. Khomskii, D. In Spin Electronics, Lecture Notes in Physics (eds Ziese, M. & Thornton, M. J.) 89116 (Springer, 2001).22. Han, J. Y. & Bark, C. W. Inuence of transition metal doping (X= Co, Fe) on structural, optical properties of Ferroelectric Bi3. 25La0. 75

X1Ti2O12. Nano Convergence 2, 15 (2015).23. Song, M. G., Han, J. Y. & Bark, C. W. Eects of Doping Ratio of Cobalt and Iron on the Structure and Optical Properties of Bi3. 25La0. 75

FexCo1xTi2O12 (x= 0, 0.25, 0.5, 0.75, 1). J. Nanosci. Nanotechnol. 15, 78417844 (2015).24. Singh, A., Khan, Z. R., Vilarinho, P., Gupta, V. & Katiyar, R. Inuence of thickness on optical and structural properties of BiFeO3 thin lms: PLD grown. Mater. Res. Bull. 49, 531536 (2014).

25. Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B 47, 558 (1993).26. Kresse, G. & Furthmller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169 (1996).

27. Dudarev, S. L., Botton, G. A., Savrasov, S. Y., Humphreys, C. J. & Sutton, A. P. Electron-energy-loss spectra and the structural stability of nickel oxide: an LSDA+U study. Phys. Rev. B 57, 15051509 (1998).

28. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865 (1996).29. Hong, J., Stroppa, A., iguez, J., Picozzi, S. & Vanderbilt, D. Spin-phonon coupling eects in transition-metal perovskites: A DFT+ U and hybrid-functional study. Phys. Rev. B 85, 054417 (2012).

Acknowledgements

This research was supported by the Basic Science Research Program and Global Research Network program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education and the Ministry of Science, ICT & Future Planning (NRF-2014R1A1A2053552, No. 2013005417, NRF-2014S1A2A2028361) and by the Austrian Science Fund (Project #1490-N19).

Author Contributions

H.A., J.S. and S.Y.J. fabricated the epitaxial thin lms. H.A. characterized the structural and optical properties of the grown lms. J.Y.H. fabricated the ceramic targets and performed the photocurrent measurement. B.K. performed the DFT calculation under the guidance of C.F. C.W.B. and S.L. designed the experiment and supervised the work. H.A., B.K., C.F., C.W.B. and S.L. co-wrote and commented on the manuscript.

Additional Information

Supplementary information accompanies this paper at http://www.nature.com/srep

Competing nancial interests: The authors declare no competing nancial interests.

How to cite this article: An, H. et al. Large enhancement of the photovoltaic eect in ferroelectric complex oxides through bandgap reduction. Sci. Rep. 6, 28313; doi: 10.1038/srep28313 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

SCIENTIFIC REPORTS

7