The electrical enhancement and reversible manipulation of near-infrared luminescence in Nd doped ferroelectric nanocomposites for optical switches

Abstract

Lanthanide ion doped ferroelectric oxides have unique ferroelectric and luminescence properties, which show great promise for future optoelectronic applications. Moreover, it is of great interest that the luminescence of ferroelectric materials can be effectively modulated by an external electric field. In this work, Nd³⁺ doped ferroelectric nanocomposites with near-infrared emission have been developed to study their optoelectronic properties and reversible tuning. At room temperature, the near-infrared emission has been enhanced over 4.6 times and the intensity can be controllably modulated by tuning the electric field. The mechanism behind the enhancement and reversible modulation has been investigated for designing luminescence memory or optoelectronic sensors. This work suggests a way for not only studying the luminescence properties of optical materials via polarization engineering, but also for potentially developing multifunctional materials for optoelectronic applications.

---

Introduction

The spectroscopic modification and reversible tuning of optical materials are great importance both for studying the fundamental physical processes and for a wide range of potential applications, such as non-contact ratiometric thermometry, fluorescent probes, multimode storage by optoelectronic materials, lighting phosphors, photocatalytic oxidation and bio-imaging and bio-separation.^1–8 Up to now, a lot of studies have focused widely on modulating the emission spectra by chemical methods, such as changing the active doping ions and the composition of the host materials. However, the technique of chemical tuning is usually an irreversible and ex situ process. Normally, the emission tuning of luminescent ions is strongly related to the crystal field around the active doping ions.^9,10 It is very difficult to segregate the single crystal field effect from the extrinsic effects that exist in diverse samples, including various defects and chemical inhomogeneities. Hence, it is of great interest to explore the modification of luminescence by some physical approaches, such as an electric field, a magnetic field, hydrostatic pressure, etc.^11–13 These physical methods remind us that the interaction between the dopant and the dopant host can be controlled at the micro level through external stimulation, and then some interesting optical phenomena can occur. These approaches can help researchers not only to learn about the kinetic process of photon emission, but also to develop some novel optoelectronic applications, such as luminescent sensors and memory devices.

Ferroelectric materials have excellent self-performance. When combined with other technologies, the newly fabricated ferroelectric materials have more extensive applications.^14,15 It is very interesting to explore the symmetry changes of the ferroelectric crystal structure driven by various methods, and then to study the luminescence modification of the dopant.^16,17 The ferroelectric materials can combine an electric field with the symmetry of the crystal structure, as a result of ferroelectric polarization. While the luminescence of lanthanide ions is sensitive to the symmetry of the crystal structure. These properties provide an opportunity for us to investigate multifunctional materials and controllable luminescence.^18–20 Therefore, it is possible to combine the properties of a ferroelectric material with luminescent ions. For example, Hao's group realized effective upconversion luminescence enhancement in Yb³⁺/Er³⁺ doped ferroelectric barium titanate thin film via an electric field.²¹ The optical properties of the material can be greatly improved by embedding ferroelectric nanocrystals into a glass matrix, namely, ferroelectric nanocomposites. Physical properties, such as the electrical and optical properties of the ferroelectric nanocomposites, can be optimized by modulating the forms and states of the dispersed ferroelectric phase. Here nanocrystalline LiNbO₃ embedded in transparent glass was chosen as the ferroelectric host. LiNbO₃ is characterized by a high Curie temperature, high dielectric constant, low dielectric loss, and high resistivity at room temperature. It also shows a high efficiency of nonlinear optics, especially in the generation of second harmonic light, which is widely used as a waveguide modulator and optical switch.

On the other hand, materials doped with lanthanide ions have a wide range of applications in optical telecommunication, the lighting industry, solid lasers, medical imaging and biological analysis.^22,23 A commonly employed strategy to study luminescent reactions and mechanisms is to introduce luminescent centers into different host substrates or ligand environments. In particular, the crystal symmetry of the main nanocomposites can effectively modulate the luminescence of the lanthanide ions. Here, lanthanide Nd³⁺ was used as a spectral probe to investigate crystal field effects because of its highly hypersensitive transitions. Moreover, Nd³⁺ ions can produce near-infrared (NIR) emission around 900, 1064 and 1326 nm. The Nd³⁺ doped materials possess vast prospects in biomedicine, high-power lasers and optical communication and so on.^24,25 Therefore, a kind of transparent Nd³⁺-doped ferroelectric nanocomposite has been developed in this work. The optoelectronic properties and tuning of the luminescence by an electric field have been tried to study the luminescence–electronic coupling. A methodical study was employed to investigate the potential mechanisms for the NIR luminescence response of an Nd³⁺-doped LiNbO₃ composite with an external electric field. Furthermore, it is intended to study the potential physical processes of the luminescence modulation in an external electric field. The reversibility of the ligand field around the Nd³⁺ ions under the electric field was also explored at room temperature.

Experimental

A transparent multicomponent glass matrix was selected and prepared based on a standard phase diagram of the Li₂O–Nb₂O₅–SiO₂ system. The optimized composition of the doping matrix was Li₂O–Nb₂O₅–SiO₂–Al₂O₃ = 35/25/(40 − x)/x (in mol%). Nd₂O₃ powder was used for Nd³⁺-doping at the level of 1 mol%. In a typical synthesis, ∼40 g of the raw oxide materials (99.99% purity) was uniformly mixed and melted in a corundum crucible at 1500 °C for 45 min in air. In order to avoid glass crystallization from the melting-quenching method, the melt was poured onto a cold steel plate and pressed with another brass plate. This operation allows the glass to cool quickly and the thickness of the as-produced sample was pressed to ∼1 mm thickness. After that, the samples were isothermally annealed at 450 °C, for 180 min, to release the residual stress in order to prevent breakage in subsequent experiments. The as-quenched samples went through two steps of heat treatment, with preliminary nucleation at a temperature of 580 °C for 2 h and then crystallization at 680 °C for 0, 60, 120, 180 and 240 min, respectively. The samples were labeled LNG-0, LNG-60, LNG-120, LNG-180 and LNG-240, respectively. After the heat treatment, the sample LNG-180 was divided into the required number of small pieces. And then the pieces were coated with silver paste for the ferroelectric and optical measurements.

The density of the prepared sample was measured by the Archimedes’ principle. The characteristic temperatures of glass-transition and crystallization temperatures were obtained by a differential scanning calorimeter using a NETZSCH DTA 404 PC instrument with a heating rate of 10 K min⁻¹. To identify the precipitated crystals, the fabricated samples were crushed into powders and investigated by X-ray diffraction (XRD, D2 PHASER, BRUKER) equipped with Cu/Kα radiation. The transmission and absorption spectra were collected using a PerkinElmer Lambda 900 UV-VIS-NIR spectrophotometer in the range of 200–1900 nm. The luminescence spectra were obtained using a Triax 320 type spectrometer equipped with an 808 nm laser diode (LD). The luminescence decay curves were recorded with an 808 nm pulse laser. Polarization–electric field (P–E) curves were measured under a frequency of 10 Hz by using a ferroelectric test system (Radiant Precision Premier II Technology) in a silicon oil bath to avoid electrical discharges. The microstructures in both the glass and the glass–ceramic composite were characterized by transmission electron microscopy (TEM, FEI Tecnai G2 F20 S-WTINE). All the measurements were carried out at room temperature.

Results and discussion

The precursor sample was transparent, and no resulting surface crystallites were observed after annealing. According to the DTA results, the precursor samples were heat treated at 680 °C for 0, 60, 120, 180 and 240 min, respectively. Fig. 1(a) shows the XRD patterns of the prepared samples. The amorphous matrix of the precursor sample is clearly shown with only a broad band. The broad-band curve indicates the nonexistence of grains in the precursor glass. After the heat treatment, the broad band is replaced by a well-defined peak, which proves the existence of crystal phase precipitation in the matrix. It is obvious that all the diffraction peaks ranging from 10 to 80 degrees have good counterparts in the standard pattern of JCPDS file card: 20-0631. All the diffraction peaks can be assigned to LiNbO₃ nanocrystals in the glass matrix. Compared with the standard data, the peak position of the prepared samples shifts to a smaller angle. As the Nd³⁺ (0.0983 nm) ionic radius is larger than the Li⁺ (0.076 nm)/Nb⁵⁺ (0.064 nm) ionic radius, cell distortion occurs when Nd³⁺ ions are substituted for Li⁺/Nb⁵⁺ ions in the crystal lattice. The peak shift indicates that Nd³⁺ ions have entered the LiNbO₃ phase. The peak intensity increases with the heat treatment time, which suggests that the diameter of the crystal grains grows. The average grain size can be estimated by the Scherrer formula. The crystal grain diameter in the LNG-180 sample was calculated to be about 6 nm by the Scherrer formula, and the diameter of the crystal grains enlarged with heat treatment time. The samples after heat treatment still showed excellent diaphaneity. The transmission spectra of the fabricated samples are shown in Fig. 2(b). It can be concluded that the samples still maintain good transparency after heat treatment, and the NIR transmittance can remain above 65%. There are five distinct downward peaks at 524, 585, 745, 808 and 876 nm, which can be assigned to the absorption transitions of Nd³⁺: ⁴I_9/2 → ²K_13/2 + ⁴G_7/2 + ⁴G_9/2, ⁴I_9/2 → ⁴G_5/2 + ²G_7/2, ⁴I_9/2 → ⁴F_7/2 + ⁴S_3/2, ⁴I_9/2 → ⁴F_5/2 + ²H_9/2 and ⁴I_9/2 → ⁴F_3/2, respectively. The nanocrystal dispersion, Fresnel scattering and absorption cause the main loss of the sample. It is deduced that the transmittance curve decreases slightly due to the increase in LiNbO₃ grains. In addition, the non-uniformity of the glass sample may be responsible for the lower transmittance of the LNG-120 sample at 800 nm. The morphology and structure in the nanocomposite samples were investigated by TEM and high resolution TEM image analysis. As can be seen in the TEM photograph shown in Fig. 1(c), the distribution of nanocrystal grains in the glass matrix is almost uniform. The size of the nanocrystal grains is in line with the XRD result. Fig. 1(d) shows a high resolution TEM image and d-spacing structure. The measured d-spacing value is 3.75 Å, which is consistent with the plane of the hexagonal LiNbO₃ crystal, d(012) = 3.75 Å.

Fig. 1 (a) The XRD patterns of the samples. (b) The transmission spectra of the samples. (c) The TEM image of LNG-180. (d) The high resolution TEM image of the lattice fringe.

Fig. 2 (a) The emission spectra of the samples by 808 nm pumping. (b) The NIR decay curves of the fabricated samples. (c) Energy transfer mechanism of Nd³⁺ ions. (d) The crystal structure of Nd doped LiNbO₃.

The NIR emissions of the Nd³⁺ in optical materials have been paid a lot of attention because of their wide applications in the field of laser technology. Fig. 2(a) shows the NIR fluorescence spectra of the Nd doped nanocomposites with no observable Stark splitting by 808 nm pumping. The three NIR emission bands and the related transitions are labeled. With an increase in the heat treatment time, the emission intensity around 1068 nm from Nd³⁺: ⁴F_3/2 → ⁴I_11/2 is slightly weakened. The decrease in luminescence intensity may be due to the clustering of Nd³⁺ ions, which are responsible for concentration quenching. In the present case, the Nd³⁺–Nd³⁺ ion separation (R_i) in the sample without heat treatment was found to be approximately 11.5 Å. In theory, the relaxation rate caused by concentration quenching varies with R_(i). After heat treatment, the LiNbO₃ crystal phase has been formed and Nd³⁺ has split into the residual glassy phase by reducing the inter-ionic separation to less than the 11.5 Å of the precursor glasses. This fact leads to a slight decrease in the intensity of luminescence. As the heat treatment process of the samples proceeds, the emission peaks take shape as in the microcrystalline phase and become sharper. Considering the transmission decreases with increasing the grain size in Fig. 1(b), optical losses such as Fresnel scattering may also cause the slight decrease in NIR emission. The decay curve at 1069 nm of the samples with excitation at 808 nm LD is shown in Fig. 2(b) at room temperature. The measured curve shows an exponential decay, and the excited state lifetime can be estimated from these decay curves. The lifetimes calculated from the fitted decay curves are 0.51, 0.5, 0.48, 0.47 ms, respectively. It can be seen that the excited state (⁴F_3/2 → ⁴I_11/2) lifetime decreases slightly with an increase in duration of heat treatment. This result indicates that Nd³⁺ ions are clustered in the process of forming the nanocrystalline phase. The volume of the crystals of the samples becomes larger after an extended heat treatment time, which causes concentration quenching in the 4f level electronic transition. Further investigations are underway to inspect the concentration effects and the limiting concentrations formed by clusters in this host. As revealed in Fig. 2(c), the infrared fluorescence spectra (λ_ex = 808 nm) in Fig. 2(a) of the samples demonstrated emissions peaks at 897, 1068 and 1336 nm based on ⁴F_3/2 → ⁴I_9/2, ⁴F_3/2 → ⁴I_11/2 and ⁴F_3/2 → ⁴I_13/2 electronic transitions. In Fig. 2(c), GSA, CR, NR, and R represent absorption, cross-relaxation, energy migration, non-radiative transition, and radiation transition, respectively. In perovskite crystals, Li⁺ and Nb⁵⁺ occupy a symmetric octahedral position at point C₃ or nearly C_3v. According to previous reports, rare earth ions can enter the sites of both Li⁺ and Nb⁵⁺ in the lattices of LiNbO₃,²⁶ and some rare earth ions are more likely to replace the position of Li⁺ as it enters the crystal.²⁷ However, there is an obvious difference in ionic radius and charge that will cause remarkable distortion of the lattice structure. The 3D unit cell model of the crystal is characterized in Fig. 2(d) for a deeper understanding of the structure of LiNbO₃. Each crystal lattice is represented by a different color in the figure. The blue atoms represent the Nd³⁺ ions mixed into the LiNbO₃ crystal lattice. Considering the study of the Nd³⁺ doped LiNbO₃ crystal,²⁶Fig. 2(d) clearly displays that the Nd³⁺ ions can replace Li⁺ and Nb⁵⁺. From the point of view of its ferroelectric structure, below the Curie temperature, the position of each layer shows a slight displacement, from the position of the oxygen layer (Li⁺) or the oxygen layer (Nb⁵⁺), caused by internal spontaneous polarization.²⁸

The above results confirmed that Nd doped ferroelectric LiNbO₃ nanocrystals formed in the glass matrix. The doping concentration and grain size seem to have little effect on the NIR emission. Considering the ferroelectric properties of the prepared composites, the NIR luminescence of Nd³⁺ is studied with an electric field. The fabricated samples are loaded with an electric field and the results are presented in Fig. 3. P-i (i = 0, 40, 80, 120, 160 and 200) designate that the sample is loaded with electric fields of 0, 40, 80, 120, 160 and 200 kV cm⁻¹, respectively. The P–E loops in Fig. 3(a) clearly show different polarization performances in samples loaded with elevated electric fields. After polarization with different electric fields, the P-i samples possess dissimilar remanent polarization. The breakdown electric field of the sample in this article is 232 kV cm⁻¹. Fig. 3(b) shows the effects of an electric field on the luminescence of LNG samples under excitation at 808 nm LD. It is remarkable that the luminescence intensity is enhanced with an increment in the electric field, which is related to the remnant polarization shown in Fig. 3(a). The luminescence peak at 1068 nm is significantly enhanced, with a maximum increase of 4.6 times compared to the sample without electric polarization. Nd can be employed as a spectral probe, suggesting that its ligand field is modified by an electric field. Fig. 3(c) clearly illustrates the positive relationship between the remnant polarization and luminescent intensity at 1068 nm, which was induced by the electric field. This good positive correlation provides reliable evidence for this study. There is an almost linear correlation between electric field and NIR luminescence. Thus, the change in luminescence can reflect the strength of the electric field. Fig. 3(d) shows the absorption spectra of P-0 and P-200 samples ranging from 470 to 900 nm at room temperature. It can be found that the Nd³⁺ ions have multiple energy levels, and the absorption spectra are in accordance with the electric dipole transitions from the Nd³⁺: ⁴I_9/2 ground state to diverse excited states. The absorption spectra have a strong absorption peak at 808 nm, demonstrating that the excitation by the 808 nm laser is reasonable. The absorption coefficients of P-0 and P-200 are 1.937 and 3.203 cm⁻¹, respectively, at 808 nm, which have increased by 62.6%. This phenomenon implies that the electrically polarized samples can absorb the excitation photons more effectively, and then generate NIR photons more effectively.

Fig. 3 (a) The polarization–electric field (P–E) loops of P-i (i = 40, 80, 120, 160, 200) samples. (b) The enhanced luminescence spectra with different electric fields. (c) The variation in remnant polarization (RP) and luminescent intensity (LI) at 1068 nm, which were induced by an electric field. (d) Absorption spectra of P-0 and P-200 samples.

To understand the mechanism of electrical enhancement in NIR emission, the crystal structures of the polarized samples have been recorded. The diffraction peaks around 24° of the crystal plane (012) in the P-i samples are shown in Fig. 4(a), indicating LiNbO₃ ferroelectric nanocrystals with varying degrees of remnant polarization. The 2θ of the plane (012) diffraction peaks in P-0 and P-200 were 23.7 and 23.84, respectively. According to the Bragg equation, the interplanar (012) spacing of LiNbO₃ is reduced by 0.022 Å after loading with 200 kV cm⁻¹. This result confirms that the crystal structure changes. It is known that the ferroelectric crystal lattice will change due to polarization under an electric field. After applying an electric field, the remnant polarization induces changes in the crystal structure around Nd³⁺, and thus the NIR emission is modulated. For the purpose of understanding the influence of remnant polarization on luminescence emissions, a schematic diagram in Fig. 4(b) shows the alteration of the ligand field around the Nd³⁺. After applying the electric field, the ions are polarized, leaving their original position relative to the oxygen layer.

Fig. 4 (a) XRD pattern of the (012) crystal plane in the P-i sample. (b) The positions of the atoms corresponding to the oxygen layer after polarization.

Consequently, different remnant polarizations are retained in the crystal lattice. Referring to Fig. 3(a), the formation of ordered ferroelectric domains induces a reduction in lattice symmetry which can be confirmed by the XRD pattern of Fig. 4(a). The formation of ferroelectric domains causes the lattice to be distorted under a strong electric field and these distortions induce a distortion of structural symmetry. Generally, the domains of the ferroelectric materials exhibit orientation polarization under or after loading with an electric field. Orientation polarization also affects the physical properties, including the luminescence. The ligand field of Nd³⁺ ions is modified by the distortion of the crystal lattice and orientation polarization. Then the uneven crystal field can make a mixture of opposite-parity states in the 4f configuration level, which would induce the intensification of the 4f–4f transitions from the ⁴I_9/2 level to the excited state. The slight migration of the Nd³⁺ ion position in the lattice by polarization causes the electric dipole transition probabilities to undergo a gradual increase, and thus the luminescence intensity can be effectively modulated.

Based on the external electric field induced modulation of luminescence, the sample was repeatedly loaded with electric fields of 150 and −50 kV cm⁻¹, respectively. The results of polarized NIR fluorescence are recorded for designing memory devices. Fig. 5(a) shows the modification of the luminescence intensity in accordance with the alteration caused by imposing voltage pulses with E = −50 kV cm⁻¹ for the ‘OFF’ state and E = 150 kV cm⁻¹ for the ‘ON’ state, respectively. When an electric field of 150 kV cm⁻¹ is applied, the ferroelectric nanocrystal lattice distortion causes enhanced luminescence of the dopant Nd. When the reverse voltage of −50 kV cm⁻¹ is loaded, the lattice distortion is reduced to cause decreased luminescence. After the voltage was loaded multiple times, the sample maintained this characteristic and the luminescence modulation was stable. As shown in Fig. 5(b), the samples maintained good nonvolatility and reversibility in the five tests performed after each loading of the electric field. This nonvolatile and reversible feature opens up significant avenues for reliable optical switches and information storage materials.²⁹ Such an electric control of luminescence in a reversible way can be used in optical switches. Moreover, compared with the complicated detection in MRAM and the destructive reading in FeRAM, lanthanide Nd doped ferroelectric nanocomposites have the advantages of fast electric field writing and convenient optical reading. And rare earth ions have rich luminescent levels (e.g., Nd³⁺: ⁴F_3/2 → ⁴I_9/2, ⁴F_3/2 → ⁴I_11/2 and ⁴F_3/2 → ⁴I_13/2), which can achieve multiple detection to ensure the reliability of data. As we have reported above, it can be concluded that luminescence emissions of Nd³⁺ doped LNG can be effectively modulated by an electric field, and the luminescence emission is modulated reproducibly. Ferroelectric oxide doped luminescent ions possess unique luminescence and inherent ferroelectric properties. Nonvolatile and reversible modulation in optical properties can be achieved without changing the composition. Thereby the developed ferroelectric oxides with luminescent ions have great potential in luminescence memory, optical switches and optoelectronic devices.

Fig. 5 (a) Reversible luminescence modulation of the sample when regularly loaded with electric fields. (b) Nonvolatile luminescence tuning after repeated external pulsed electric field testing.

Conclusions

In conclusion, we explored the electric-induced enhancement and reversible modulation of NIR luminescence using well-fabricated Nd doped ferroelectric nanocomposites at room temperature. The developed sample simultaneously possesses NIR luminescence and ferroelectric properties. Taking advantage of electric polarization, the NIR emission is effectively enhanced by 4.6 times due to the modulated crystallographic symmetry around Nd³⁺. More importantly, unlike chemical methods, this modulation is reversible and nonvolatile in a solid sample, which is promising for luminescence memory. XRD results and absorption and photoluminescence spectra have verified this electric-luminescence coupling phenomenon. And the results are not only applicable to the Nd³⁺ doped ferroelectric samples explored in this experiment. This will also stimulate the study of novel optoelectronic materials and devices, such as luminescent memory, photoelectric modulators, and optical switch devices.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the Chinese National Natural Science Foundation (No. 61705214, 61605192 and 61775205); and the Zhejiang Provincial Natural Science Foundation of China (No. LD18F050001).

Notes and references

1. Y. Wang, K. Zheng, S. Song, D. Fan, H. Zhang and X. Liu, Chem. Soc. Rev., 2018, 47, 6473–6485.
2. M. Chen, H. Hu, Y. Tan, N. Yao, Q. Zhong, B. Sun, M. Cao, Q. Zhang and Y. Yin, Nano Energy, 2018, 53, 559–566.
3. Q. Zhang and Y. Yin, Sci. Bull., 2018, 63, 525–526.
4. Y. Zhang, W. Jie, P. Chen, W. Liu and J. Hao, Adv. Mater., 2018, 30, 1707007.
5. H. Zhang, D. Peng, W. Wang, L. Dong and C. Pan, J. Phys. Chem. C, 2015, 119, 28136–28142.
6. Z. Xia and Q. Liu, Prog. Mater. Sci., 2016, 84, 59–117.
7. S. Lv, B. Shanmugavelu, Y. Wang, Q. Mao, Y. Zhao, Y. Yu, J. Hao, Q. Zhang, J. Qiu and S. Zhou, Adv. Opt. Mater., 2018, 1800881.
8. D. Peng, Q. Ju, X. Chen, R. Ma, B. Chen, G. Bai, J. Hao, X. Qiao, X. Fan and F. Wang, Chem. Mater., 2015, 27, 3115–3120.
9. S. Shi, L. D. Sun, Y. X. Xue, H. Dong, K. Wu, S. C. Guo, B. T. Wu and C. H. Yan, Nano Lett., 2018, 18, 2964–2969.
10. G. Bai, M.-K. Tsang and J. Hao, Adv. Opt. Mater., 2015, 3, 431–462.
11. Y. Liu, D. Wang, J. Shi, Q. Peng and Y. Li, Angew. Chem., Int. Ed., 2013, 52, 4366–4369.
12. X. Wang, H. Zhang, R. Yu, L. Dong, D. Peng, A. Zhang, Y. Zhang, H. Liu, C. Pan and Z. L. Wang, Adv. Mater., 2015, 27, 2324–2331.
13. H. Sun, X. Wu, D. F. Peng and K. W. Kwok, ACS Appl. Mater. Interfaces, 2017, 9, 34042–34049.
14. J. Wu, N. Qin and D. Bao, Nano Energy, 2018, 45, 44–51.
15. J. Wu, Q. Xu, E. Lin, B. Yuan, N. Qin, S. K. Thatikonda and D. Bao, ACS Appl. Mater. Interfaces, 2018, 10, 17842–17849.
16. Y. Yu, Z. Fang, C. Ma, H. Inoue, G. Yang, S. Zheng, D. Chen, Z. Yang, A. Masuno, J. Orava, S. Zhou and J. Qiu, NPG Asia Mater., 2016, 8, e318–e318.
17. R. Zhang, Z. Song, L. He, Z. Xia and Q. Liu, J. Alloys Compd., 2017, 729, 663–670.
18. D. K. Khatua, A. Kalaskar and R. Ranjan, Phys. Rev. Lett., 2016, 116, 117601.
19. Q. Yao, F. Wang, F. Xu, C. M. Leung, T. Wang, Y. Tang, X. Ye, Y. Xie, D. Sun and W. Shi, ACS Appl. Mater. Interfaces, 2015, 7, 5066–5075.
20. G. Bai, M.-K. Tsang and J. Hao, Adv. Funct. Mater., 2016, 26, 6330–6350.
21. J. Hao, Y. Zhang and X. Wei, Angew. Chem., Int. Ed., 2011, 123, 7008–7012.
22. W. Lu, Z. Gao, X. Liu, X. Tian, Q. Wu, C. Li, Y. Sun, Y. Liu and X. Tao, J. Am. Chem. Soc., 2018, 140, 13089–13096.
23. H. Vigouroux, E. Fargin, S. Gomez, B. Le Garrec, G. Mountrichas, E. Kamitsos, F. Adamietz, M. Dussauze and V. Rodriguez, Adv. Funct. Mater., 2012, 22, 3985–3993.
24. Z. Wu, G. Bai, Q. Hu, D. Guo, C. Sun, L. Ji, M. Lei, L. Li, P. Li, J. Hao and W. Tang, Appl. Phys. Lett., 2015, 106, 171910.
25. G. Gao, D. Busko, S. Kauffmannweiss, A. Turshatov, I. A. Howard and B. S. Richards, J. Mater. Chem. C, 2018, 6, 4163–4170.
26. L. Rebouta, J. C. Soares, M. F. D. Silva, J. A. Sanz-García, E. Dieguez and F. Agulló-López, J. Mater. Res., 1992, 7, 130–135.
27. D. Tu, C.-N. Xu, A. Yoshida, M. Fujihala, J. Hirotsu and X.-G. Zheng, Adv. Mater., 2017, 29, 1606914.
28. M. P. Fernandes Graça and M. A. Valente, MRS Bull., 2017, 42, 213–219.
29. M. Zheng, H. Sun, M. K. Chan and K. W. Kwok, Nano Energy, 2019, 55, 22–28.

---