Introduction

Ferroelectric materials possess spontaneous polarizations that usually originate from the collective displacement of ions in the crystals. Applying an electric field leads to polarization reversal that is coupled with a strain response known as the piezoelectric effect. Because of their multifunctionality, ferroelectric materials are widely used in nonvolatile memory, ultrasonic imaging systems and many other devices.1, 2, 3

Ferroelectricity was originally discovered in Rochelle salts4 and later in several other molecular systems.5 However, the rapid development of the field took place only after the discovery of ferroelectricity in perovskite oxides, for example, BaTiO3 and Pb(Zr,Ti)O3.6, 7 Although these oxide ferroelectrics are more attractive for applications because of their larger polarization and piezoelectric coefficients, they usually contain heavy metals and, therefore, are not environmental friendly. They require a high processing temperature, making them incompatible with current microelectronic systems. They are usually brittle and cannot be used in flexible devices. Organic materials, however, are much more flexible, lightweight, easy to process and are usually transparent.8 They are ideal for applications in portable/flexible electronics. During the past several years, molecular ferroelectrics have attained significant advancements, and their properties are now comparable with those of BaTiO3.9, 10, 11, 12 For example, the trigonal imidazolium perchlorate (C3N2H5ClO4) crystal has a polarization of ~8 μC cm−2 at room temperature, a high Tc of 100 °C and a low coercive field.12, 13

Despite recent intensive studies on molecular ferroelectrics, very few reports have addressed the issue of thin film growth, even though it is critical for practical applications. For inorganic ferroelectric materials, high-quality films are usually obtained by their epitaxial growth on lattice matching substrates.2, 3 However, such substrates are not available for organic molecular ferroelectrics. Furthermore, molecular materials may show growth behaviors that are drastically different from the inorganic ones.14, 15, 16, 17, 18 Thus, it is important to conduct a systematic investigation of this topic. Here, we report the room-temperature growth of high quality, flexible and transparent19 C3N2H5ClO4 films on various substrates, including polyethylene terephthalate (PET).

Experimental Procedure

Pure C3N2H5ClO4 powders were dissolved in purified water to form a 60% saturated solution at 25 °C. Pt-coated Si (Pt/Si), quartz and indium tin oxide coated PET (ITO/PET) substrates were used for the film growth. A seed layer of C3N2H5ClO4 was prepared on the substrates by spin coating the solution at 3000 r.p.m., followed by annealing at 120 °C for 1 h. The substrates were kept in the saturated solution for 0.5, 1 and 3 days, while the water evaporated at ~0.2% per day at ~25 °C. After growth, the films were dried at 120 °C for 1 h before further measurements. For comparison, the Pb(Zr0.2Ti0.8)O3 and BaTiO3 films were grown on a (0 0 1) SrTiO3 substrate with a 50-nm SrRuO3 buffer layer by using pulsed laser deposition at 680 and 710 °C, respectively. The oxygen partial pressure during the growth was 13 Pa, and the laser energy was 150 mJ per pulse.

The crystal structure of the films was characterized using an X-ray diffractometer (Brucker D8 Advance, Karlsruhe, Germany). Transmittance spectra were measured by using a UV–visible spectrometer (Shimadzu, Kyoto, Japan). The morphology and the piezoelectric force microscopy (PFM) studies were conducted on an atomic force microscope (Brucker Multimode 8, Camarillo, CA, USA). A 2-V AC voltage at 41 kHz was applied to the Co/Cr- coated conductive tip (Bruker MESP-RC) to obtain the domain images. The frequency and the DC bias-dependent piezoelectric responses were measured with a 1-V AC voltage and a conductive tip. The 100-μm diameter Pt electrodes were deposited on the films, and the polarization was measured by using a commercial ferroelectric tester (Radiant Multiferroic, Albuquerque, NM, USA). The dielectric responses were measured using a precision impedance analyzer (Agilent 4294A, Santa Clara, CA, USA). The Young’s modulus was measured by using a nanoindenter (Agilent G200).

Results and Discussion

Film growth and structural analysis

When the substrates are soaked in the saturated solution and as the water gradually evaporates, the C3N2H5ClO4 dendritic crystals appear and grow uniformly on each substrate. As an example, the film on the PET is shown in Figure 1a. Several dendritic crystals grow independently in regions A, B, C and D, and the length of the crystals in region A is over 0.5 mm. Dendritic crystal growth has been previously reported for oxides, polymers and molecular materials on amorphous substrates.14, 15, 16, 17, 18, 20 To obtain a large area high-quality film, it is preferred that one nucleus dominates the growth. Various parameters, such as temperature, solubility and the substrate–film interaction, affect the dendritic crystal growth.14, 15, 16, 17, 18 In general, the crystal growth in a liquid environment is controlled by three processes: (1) the diffusion of molecules through the liquid phase to the growth front; (2) the reorganization of molecules into the film; and (3) the heat dissipation at the growth front.18 In our experiments, molecular diffusion should not be the limiting factor because of the highly concentrated solution. Because of the extremely low growth rate, heat diffusion is also unlikely to be an issue. Thus, we suggest that molecular reorganization at the growth front is the controlling process in this case. By slowly evaporating water to control the degree of over saturation, we can prolong the time interval between the appearances of multiple nuclei. As a result, we demonstrate millimeter-scale dendritic film growth from a single nucleus on the ITO/PET (Figure 1b), quartz and Pt/Si substrates (Supplementary Figure S1). Note that AB and A′B′ are two branches originating from the same stem, which can be regarded as the stem for other dendritic crystals.

Figure 1
figure 1

Crystal growth of C3N2H5ClO4 film on ITO/polyethylene terephthalate. The optical images of C3N2H5ClO4 films grown from multiple nuclei (a) and a single nucleus (b), respectively. The surface morphology of a film after (c) 12 h, (d) 1 day and (e) 3 days in the same saturated solution at 25±0.1 °C.

The well-ordered dendritic crystal branches can merge into a seamless film as water continues to slowly evaporate. There are two prominent growth patterns: one with an ~60° angle between the stem and the branches and the other with an approximately 90° angle. These patterns appear because of the fast-growth orientations that depend on the crystal structure and growth conditions.14, 15, 16, 17, 18 In Figures 1c–e, we show films that were grown in the same solution with water evaporating at ~0.2% per day for 12 h, 1 and 3 days, respectively. The dendritic crystals grow on the substrate similar to a vine climbing on a wall during its early stage of growth (Figure 1c). After 1 day, the dendritic crystals are ~3 μm wide and 500 nm thick (Figure 1d). After 3 days, they merge completely into a continuous film, although there are still some traces of dendritic crystals (Figure 1e). Generally, a strong interaction between the film and the substrate is favorable for smooth film growth.16, 17 This method may be used to grow films of other molecular ferroelectrics from saturated solutions.

Figure 2a shows the X-ray diffraction patterns of the C3N2H5ClO4 powder, the ITO/PET substrate and the two 1 × 3-mm2 films on the ITO/PET substrate. The C3N2H5ClO4 ferroelectric phase belongs to the rhombohedral R3 m point group, the same as that of LiNbO3 (Supplementary Figure S2). In addition to the peaks from the ITO/PET substrate, the first film has only one diffraction peak at ~22° that arises from the (1 0 −1) planes. Similarly, the second film has a single diffraction peak at ~24° that arises from the (0 −1 −1) planes. The (1 0 −1) and (0 −1 −1) planes are determined by the fast-growth orientations, which are very sensitive to the growth conditions and the crystal structure.14, 15, 16, 17, 18 Similar results have been achieved on Pt/Si (Supplementary Figure S3) and quarts substrates; thus, the film growth does not depend on the crystal structure of the substrate.

Figure 2
figure 2

Structure and optical property. (a) X-ray diffraction patterns of C3N2H5ClO4 powder, (0 −1 −1) film and (1 0 −1) film on ITO/polyethylene terephthalate. (b) Optical images demonstrating the flexibility of the films. Transmittances of C3N2H5ClO4 films on quartz (c) and ITO/polyethylene terephthalate (d) substrates, showing outstanding transparency of the films.

Optical and electrical properties

After clarifying the film growth process, we then studied the properties of the films. First, by growing the C3N2H5ClO4 films on ITO/PET and quartz, we demonstrated that it is flexible and transparent in the visible range (Figure 2b). The transmittance of the film on quartz is nearly zero at a wavelength (λ) of 230 nm, which increases abruptly to >70% at λ=240 nm (Figure 2c). This suggests that the C3N2H5ClO4 film has a band gap of ~5.4 eV. Although the ITO/PET substrate absorbs most light with λ<320 nm, the flexible C3N2H5ClO4/ITO/PET sample shows a high transmittance of >70% at λ=390–800 nm, exhibiting excellent transparency in the visible region (Figure 2d).19

The ferroelectric properties of the films on the ITO/PET are characterized using both PFM and a conventional ferroelectric tester.11, 15 Although most of the film is in a single domain state, domain structures are still observed in various regions. In a 50 × 50-μm2 area of the (0 −1 −1) region, yellow/black domains that indicate upward/downward polarization are observed in the vertical PFM phase image (Figure 3a), and the contrast is barely visible in the lateral phase image (Figure 3b). The corresponding vertical and lateral PFM amplitude images are shown in Figures 3c and d. The near zero signal in Figure 3d is consistent with the polarization being perpendicular to the surface. Additionally, the (0 −1 −1) C3N2H5ClO4 region usually shows hexagonal domains, which is common in LiNbO3 with a R3 m point group when the unique polar axis is perpendicular to the surface.21, 22 However, in a 50 × 50-μm2 area of the (1 0 −1) region, domains are clear in both the vertical (Figures 3e and g) and lateral (Figures 3f and h) PFM images, which is consistent with the fact that there is an angle of ~60° between the (1 0 −1) and (0 −1 −1) planes. Similar domain patterns are also observed in films grown on Pt/Si substrates (Supplementary Figure S4). Furthermore, the temperature-dependent study reveals that the ferroelectric domains are stable up to 80 °C (Supplementary Figure S5). However, they change at 90 °C and disappear at 102 °C because of the ferroelectric-to-paraelectric phase transition at ~100 °C.13, 20 This suggests that C3N2H5ClO4 films can be used safely below 80 °C.

Figure 3
figure 3

Ferroelectric domain structures of the C3N2H5ClO4 film. (a) Vertical and (b) lateral piezoelectric force microscopy phase images and (c) vertical and (d) lateral amplitude images of the (0 −1 −1) film, and (e–h) those of the (1 0 −1) film, where the insets in (a, e) show the corresponding surface morphologies.

Typical well-saturated polarization-electric field loops are observed for the (0 −1 −1) film with the 100-μm diameter Pt electrodes (Figure 4a). The saturation polarization is ~6.4 μC− cm−2 at 100 Hz, similar to that of the single crystal.13 The coercive voltage (VC) is lower than 1.7 V for the ~2 μm thick film (that is, <8.5 kV cm−1), which is much smaller than that of perovskite ferroelectrics.2, 3 This makes it an ideal candidate for applications in low-energy consumption portable electronics. The relative dielectric constant and loss show maximum values (ɛ′~25 and tanθ~0.2) near VC (Figure 4b).

Figure 4
figure 4

Ferroelectric and piezoelectric properties of the (0 −1 −1) C3N2H5ClO4 film on ITO/polyethylene terephthalate. (a) Polarization versus voltage loops measured at 100 Hz. (b) DC field dependence of relative dielectric constant (ɛ′) and its loss (tanθ). (c) Piezoelectric response versus DC voltage and (d) piezoelectric response versus AC frequency curves for the C3N2H5ClO4 film, (0 0 1) Pb(Zr0.2Ti0.8)O3 and BaTiO3 epitaxial films, respectively.

To quantify the piezoelectric response of the (0 −1 −1) C3N2H5ClO4 film and compare it with the perovskite ferroelectrics, we record the PFM amplitudes of the (0 −1 −1) C3N2H5ClO4 film, the (0 0 1) Pb(Zr0.2Ti0.8)O3 and the BaTiO3 epitaxial films using a voltage of 1 V AC. Figure 4c shows the DC bias-dependent PFM amplitude, and Figure 4d displays the frequency dependence. The typical butterfly shape curves are observed, and the PFM amplitude of the C3N2H5ClO4 film is comparable with that of Pb(Zr0.2Ti0.8)O3 and BaTiO3; however, the driving voltage is much smaller. Furthermore, the near resonance piezoelectric response of the (0 −1 −1) C3N2H5ClO4 film is much larger than that of the Pb(Zr0.2Ti0.8)O3 and the BaTiO3 epitaxial films (Figure 4d), even though the piezoelectric d33 coefficient of the C3N2H5ClO4 single crystal (~40 pC N−1) is much smaller than that of the Pb(Zr0.2Ti0.8)O3 (>400 pC N−1) or the BaTiO3 bulk (>150 pC/N−1).20, 23, 24 This is likely because of the smaller elastic constant of C3N2H5ClO4, which makes the film less susceptible to substrate restriction (note that the Young’s modulus of C3N2H5ClO4 (~0.8 GPa) is much smaller than that of Pb(Zr0.2Ti0.8)O3 or BaTiO3 (>70 GPa). Furthermore, the near resonance piezoelectric response of the C3N2H5ClO4 film is at least eight times greater than that of the flexible polyvinylidene fluoride-trifluoroethylene copolymer (PVDF-TrFE) film.20 This behavior offers tremendous advantages for applications in portable electronics in which piezoelectric films are more desirable than bulk crystals.

Flexibility of the C3N2H5ClO4 film

To demonstrate the flexibility of the film,19 we investigate the ferroelectric domain structure of the C3N2H5ClO4 on the ITO/PET under both tensile and compressive strains using PFM by bending the sample positively and negatively, respectively. Compared with the original domains (Figures 5a and b), no significant changes are observed in the domain structure as the bending radius decreases to 7.9 mm (Figure 5c) and 3.1 mm (Figure 5d). Furthermore, in another sample (Figures 5e and f) that is negatively bent to introduce compressive strain, no obvious changes are observed as the bending radius decreases to 6.6 mm (Figure 5g) and 4.5 mm (Figure 5h). The piezoelectric responses of both positively and negatively bent films do not weaken according to the amplitude images of those bent samples measured at 41 kHz (Supplementary Figure S6). Therefore, the C3N2H5ClO4 film can be applied in flexible devices requiring a high piezoelectric response.

Figure 5
figure 5

Domain evolution of the bent (0 −1 −1) C3N2H5ClO4 on ITO/polyethylene terephthalate. The original surfaces and vertical piezoelectric force microscopy images of the first (a, b) and second (e, f) films, vertical piezoelectric force microscopy images of the first positively bent film with a radius of 7.9 mm (c) and 3.1 mm (d), and those of the second negatively bent film with a radius of 6.6 mm (g) and 4.5 mm (h).

Conclusions

C3N2H5ClO4 films with a preferred (0 −1 −1) or (1 0 −1) orientation were grown on various substrates, including transparent and flexible ITO/PET, at room temperature. The growth starts with dendritic crystals that eventually merge together to form a continuous film. Because of the smaller elastic constant of C3N2H5ClO4 compared with inorganic ferroelectric materials, the piezoelectric response of the bulk is preserved in the films. The films are also transparent and flexible. The superior properties, the low temperature required for growth and the minimal requirements on the substrates make organic ferroelectric films ideal candidates for applications in next-generation flexible electronics.