Optical, conductive, and ferroelectric properties of the first layer of dip-coated BiFeO3 films from methoxyethanol and acetic acid-based chemical dissolvents

The overall performance of the multilayer resulting in a sol-gel bismuth ferrite (BiFeO3) film will be primarily determined by the properties of the first layer, but this has yet to receive much attention, even though chemical and morphological defects of this layer can accumulate as the number of layers increases. Here, we perform an optical, conductive, and ferroelectric study of first layer (L 1) dip-coating sol-gel BiFeO3 films using two routes that vary only in the dissolvent; the first one is based on 2-methoxyethanol (MOE), and the second one on acetic acid (AA) with some MOE (AA-MOE). Tauc plots reveal a band gap of 2.43 eV and 2.75 eV for MOE (30 ± 5 nm thick) and AA-MOE (35 ± 5 nm thick) films, respectively. MOE films showed a dielectric function with features at ∼2.5 eV, ∼3.1 eV, and ∼3.9 eV, which were associated with charge-transfer transitions, but such features are absent in AA-MOE films. Advanced atomic force microscopy techniques were used to identify the fine features or defects of the BiFeO3 films: The conductive maps show that the charge transport pathways in both film routes are controlled by nanometer defects rather than grain or grain boundary defects. Current-voltage curves reveal high conductive pathway at a lower voltage for the MOE films than for AA-MOE films. The piezoelectric coefficient for MOE films was ∼20% higher than AA-MOE films. Both deposition methods yield ferroelectric films with an electromechanical strain controlled by the piezoelectric effect and minimal contribution from electrostriction. An optimization for the AA-MOE-based route in the withdrawal speed results in a significant reduction of morphological defects and a more than twofold increase in the piezoelectric coefficient. Our results broaden the understanding of optical and ferroelectric BiFeO3 films based on a chemical solution by dip-coating.


Introduction
Bismuth ferrite (BiFeO 3 ) is a lead-free and, therefore, environmentally friendly ferroelectric material that has the potential to compete with the lead zirconate titanate ferroelectric [1].BiFeO 3 can exhibit multifunctional properties such as giant polarization [1,2], high piezoelectric constant [3,4], magnetic properties [5], and resistive switching [6].Among the methods for producing BiFeO 3 film, those using chemical solution deposition (CSD) have been shown to compete, in terms of film quality and physical properties, with the wellestablished physical vapor deposition methods [7].CSD methods have proven to be an efficient route to obtain BiFeO 3 films with attractive macroscopic ferroelectric properties [2,8].Polycrystalline [2] and epitaxial structure [6] pure-phase BiFeO 3 films presenting rhombohedral structure of the space group R3c or less distorted R3m have been reported using the sol-gel route, which show remnant polarization that ranges from 0.03 µC cm −2 to >100 µC cm −2 depending on the thickness, deposition route and film quality [2,7,9].
Bismuth nitrate and iron nitrate are usually the dissolved precursors using either 2-methoxyethanol (MOE) [2] or acetic acid with MOE (AA-MOE) [10][11][12] to form the solution.Several layers of this solution are deposited on a substrate to reach the desired thickness.Between each layer, a drying process (gelation) at a relatively low temperature (100 • C-200 • C) is performed to create an amorphous film, followed by a sintering treatment (450 • C-650 • C) to reach the desired crystalline phase [7].The resulting BiFeO 3 films usually reach the wanted ferroelectric phase, but the ferroelectric properties do not always are the expected.It has been pointed out that the absence of ferroelectric features, as a ferroelectric loop, for example, is a consequence of chemical defects in the material caused by factors that govern the chemical process during heating (gelation) [7], resulting in leakage currents [10], as well as crack formation [2].An annealing treatment in oxygen atmosphere of the BiFeO 3 films has demonstrated an important reduction in the leakage current density since it reduces the oxygen vacancies with respect to those annealed in air [2,6].Besides, an increase in the bismuth content of the solution (excess bismuth) has also been shown to prevent the occurrence of oxygen vacancies, resulting in a two orders of magnitude reduction in leakage current density with respect to samples with bismuth vacancies [10].As more layers are stacked to increase the film thickness, more chemical, morphological, and even interfacial defects are created, resulting in the loss of ferroelectric properties.Therefore, understanding what happens during the formation of the first layer that will form the BiFeO 3 film in terms of defects becomes a crucial issue since subsequent layers are duplicates of this first layer.
Among variants for using the CSD method, dip-coating stands out as it allows not only reproducibility during the coating of complex geometries useful in the integration into microelectronic circuits with industrial scalability, but also offers the possibility to manipulate the morphology and thickness of the resulting film [13].On the other hand, it has been shown that for very thin BiFeO 3 films, the spin coating method causes loss of stoichiometry due to bismuth deficiency induced by the rapid volatilization of Bi 3+ , which leads to a high concentration of oxygen vacancies with increased conductivity, in addition to promoting the formation of iron-rich phases [14].In contrast, the dip coating method allows control of solvent evaporation at the surface of the liquid film through the speed of substrate extraction, also allowing control over the drying and aging times of the reagents to tune the atomic ordering [15].
In this work, we focused on studying the optical, conductive, and ferroelectric response of the first layer (L 1 ) of solgel BiFeO 3 films deposited by dip-coating.CSD methods that only vary in dissolvent were compared, one using MOE [16] and the other using the combination of AA-MOE [17].To reveal the fine features of the first BiFeO 3 layer, our study was conducted at the nanoscale employing a range of atomic force microscopy (AFM) techniques, namely, conductive atomic force microscopy (CAFM), piezoresponse force microscopy (PFM), higher order harmonic PFM, and switching PFM.To study the films in their pristine form, no annealing treatments were performed in an oxygen atmosphere; thus, the genuine defects of the resulting films were exposed.

Solution and film preparation
Precursors bismuth nitrate [Bi(NO 3 )   C in open atmosphere to prepare the BiFeO 3 solution for film deposition with a molar ratio Bi/Fe of 1.1 in 20 ml.The MOE was used as the solvent because it has good solubility with the starting reagents and provides good viscosity to the final solution [7].The films deposited with this solution were labeled as MOE-L n BFO, where n alludes to the number of layers.The solution was aged for 24 h before deposition of the films.
A second set of BiFeO 3 films was prepared, changing the dissolvent of the precursor solution.In this second set, the bismuth nitrate and iron nitrate (same molar ratio as MOE-based films) were dissolved in MOE and AA (1:3 volume) at constant stirring for 1 h.Then, citric acid (0.25 M) was added as a chelation agent and stirred for 30 min at the same temperature.AA has the function of dehydrating the water of crystallization of the metal salts [7].This set of BiFeO 3 samples was labeled as AA-MOE-L n BFO.
Both solutions were deposited on glass and FTO/glass substrates by dip-coating using a withdrawal speed of 2 cm min −1 .The films were dried at 200 • C for 1 h and then sintered at 550 • C for 1 h to reach the crystalline phase.Sintering temperatures above 550 • C were excluded as the glass slide has a glass transition temperature of 564 • C, although the FTO can be heated up to 700 • C according to the manufacturer.Besides, relatively low sintering temperatures are required to enable the incorporation of BiFeO 3 compound into devices like thin film solar cells or flexible electronic [18].
There, high temperatures can promote the emergence of secondary phases [14] in the BiFeO 3 film or decomposition of another layer constituting the device.

Structural and optical characterization
X-ray diffractograms (XRD) were acquired using CuK α radiation (λ = 1.5405Å) in a PANalytical Empyrean diffractometer with a 2θ range from 20 • to 80 • .A spectrophotometer UV-Vis-NIR Cary 5000 was used to obtain transmittance (T) and reflectance (R) spectra from which the band gap energy was calculated.The absorption coefficient α was calculated from the transmittance spectra and film thickness (d) using the formula, The direct band gap energy (E g ) for a single layer (L 1 ) BiFeO 3 films was calculated using the onset of the absorption region from the (αhν) 2 vs. energy plot, here h is the Plank constant and ν is the frequency of incident light.A fit at the linear part of the curves and extrapolation to zero (αhν) 2 gives the E g for the L 1 BiFeO 3 films, considering the proportionality αhν ∼ (hν−E g ) 1/2 .
An ellipsometer Horiba UVISEL operating from 1.5 eV to 4.7 eV was used to obtain the complex dielectric function of the L 1 BiFeO 3 films.Ellipsometry is based on the measurement of the change in the polarization state of a light beam caused by reflection on the surface of a material.From the changes in polarization state that are represented by the amplitude ratio tan Ψ and the phase difference ∆ it is possible to access relevant information such as film thickness and optical constants of the material using a suitable dispersion model.In ellipsometry, the incident beam of light is linearly polarized so that the resulting vector after reflection on the sample traces an ellipse.The interaction of light with the material obeys the laws of plane wave transmission and reflection, Fresnel's coefficients, and Snell's law.The processes that take place in the material due to the interaction with polarized light depend on the photon energy and can be electronic transitions, molecular or crystal lattice vibrations, and absorption of free carriers.In the present study, a model air/film/substrate was considered during the fit of the ellipsometry data.The glass substrate on which the BiFeO 3 films were deposited was modeled with the Cauchy dispersion relation, and the calculated parameters were kept fixed during the extraction of the dielectric response of the BiFeO 3 films from the ellipsometry data.The model for the BiFeO 3 films was Tauc-Lorentz; the imaginary part of Tauc's dielectric function describes inter-band transitions above the band edges [19].The fit was performed to the ellipsometry parameters Is and Ic, which are functions of ψ and ∆ according to Is = sin2ψ.sin∆, Ic = sin 2ψ cos ∆, and Ic ′ = cos 2ψ.The quality of the fit to the experimental data was determined from parameter χ 2, as appreciated in figure S1.

Atomic force microscopy techniques
AFM experiments were performed using an atomic force microscope Bruker dimension edge in-house upgraded with PFM, higher order harmonic PFM [20], and switching PFM techniques.These techniques are assisted by a lock-in amplifier HF2LI or Fourier analysis.The probe used for PFM, higher order harmonic PFM, and switching PFM was a BudgetSensors Tap150E-G and SCM-PIC-V2 for CAFM.Details about sample preparation for AFM measurements are presented in figure S2.
To evaluate the conductivity of the L 1 BiFeO 3 samples at the nanoscale, the CAFM technique was employed.A conductivity map of 3 µm × 3 µm at 1 V was obtained to determine the defect density on the films by counting the pixel number of the map with current lower than −40 pA divided by the total pixel of the map.To determine the pathway that follows the charge carriers, the area of the current map was reduced to 400 nm × 400 nm, and the voltage applied to the bottom electrode of the sample was increased.Curves current vs. voltage (I-V) were measured on the L 1 BiFeO 3 samples in separated loadings using linear ramps from −5 V to +5 V and from +5 V to −5 V to close the cycle.To rule out bias in the measurement, the I-V curves were obtained on map configuration of 6 × 6 in an area of 10 µm × 10 µm.The I-V curves obtained close to morphological defects or conductive defects were discarded.
To evaluate the local switching of the samples, PFM switching measurements were performed at the nanoscale.Excitation to achieve switching was performed with a train of voltagemodulated pulses containing on-field and off-field conditions.In each on-and-off condition, a sine wave with a frequency sweep spanning the contact resonance of the sample-tip system was superimposed on the pulse train [21].A lock-in amplifier using the sinusoidal signal as a reference allows demodulating the raw signal obtained from the AFM photodetector and presenting the material response as amplitude and phase signals to which a simple harmonic oscillator model is fitted [22], thus obtaining relevant parameters such as the resonance where the maximum amplitude occurs, and the quality factor of the resonance peak.The pulse train and the fit to experimental data can be appreciated in figure S3.The amplitude contains information on the electromechanical response of the material; if the amplitude loop is hysteretic with a butterfly shape and the phase also exhibits hysteresis with 180 • change, then the two signals may indicate a switching of the region being excited by the conductive probe tip.Since effects such as electrostriction, electrostatic forces, and chemical reactions can also result in the formation of a hysteretic loop [23], it is preferable to study the off-field condition of the loops in switching PFM because such effects are minimized [21].
To obtain information about the pristine polarization of the sample and domain size, piezoresponse (PR) images of BiFeO 3 samples were obtained.The out-of-plane PR was obtained from the X signal of the lock-in amplifier by reducing the experimental background through the Y-contrast reduction, as described in [24,25], see figure S4.
The higher-order harmonics technique [20] was used to separate the piezoelectric and electrostriction effects from the electromechanical signal.To evaluate the magnitude of the PR, i.e. how much mechanical strain is generated per unit voltage, a periodically poled lithium niobate test sample (ARPPLN) from Asylum Research of known piezoelectric coefficient (∼8.5 pm V −1 [26]) was used to compare with the sol-gel BiFeO 3 samples.It is important to recognize that the permittivity and thickness of the reference sample must be taken into account for proper calibration of the piezoelectric coefficient [27], but the use of ARPPLN allows us to have a rough idea of the behavior of a piezoelectric sample.Materials that present an electromechanical signal dominated by the piezoelectric effect will have a first harmonic that has a linear dependence on the applied voltage and of magnitude greater than the response of the second harmonic (which will have a quadratic dependence on the applied voltage).Materials that present an electromechanical signal dominated by the electrostriction effect will have a second harmonic signal of greater magnitude than the first harmonic.In this type of experiment, first, excite around the resonance peak and detect around that resonance, then excite at ω 0 /2 and detect at both ω 0 and ω 0 /2, the latter being the second harmonic.Sample MOE-L 1 -BFO and AA-MOE-L 1 -BFO were measured under this configuration using the same experimental conditions as the reference sample.Measurements were made in a 3 × 3 mesh within a 3 µm × 3 µm area for samples MOE-L 1 -BFO and AA-MOE-L 1 -BFO, while for the reference 5 × 5 mesh was obtained within a 25 µm × 25 µm area.The dispersion between measurements for each sample was obtained using the standard deviation, which is indicated as error bars in the graphs.To verify the integrity of the conductive probe after measuring the BiFeO 3 samples, the reference sample was measured again, and an attenuation of 3% at 2 V excitation was registered, indicating that the sample showed barely appreciable wear.The wear is because the probe scans in contact mode 256 × 256 pixels for each sample and stops at each position configured to perform the experiment.In this way, the probe always remains on the surface.
From figure 1(b), an E g of 2.43 eV and 2.75 eV for MOE-L 1 -BFO and AA-MOE-L 1 -BFO films, respectively, was determined.Note that the (αhν) 2 curve for the film MOE-L 1 -BFO presents a feature at ∼2.5 eV, possibly a sub-gap that shifts the onset of optical absorption to lower energies.The origin of this sub-gap could be a defect state in the electronic structure or evidence of an excitonic character, as discussed in [28].The feature at ∼2.5 eV can be better appreciated from the absorption coefficient plot [inset in figure 1(b)].From this inset, two other features at ∼3.1 eV and ∼3.9 eV can be seen for the MOE-L 1 -BFO film (see vertical dot-dash lines), which have been assigned as charge transfer excitations in BiFeO 3 films [28] rather than interband-transitions in semiconductors [29].These features do not appear in the AA-MOE-L 1 -BFO film.Charge transfer transitions occur from the occupied O 2p-Fe 3d states of the valence band to the unoccupied Fe 3d states of the conduction band [29,30].The occurrence of fluctuations in the valence band of Fe ions (Fe 2+ and Fe 3+ ) creates oxygen vacancies for charge compensation, resulting in low electrical resistivity and a small band gap [30].From the results in figure 1(b) it was determined that the band gap energy of the MOE route (2.43 eV) is lower than the AA-MOE route (2.75 eV), indicating that transitions in the film resulting from the MOE route are more likely, which are seen in the complex dielectric function and optical absorption.
Figure 1(c) shows the complex dielectric function calculated for MOE and AA-MOE-based films.The model parameters using the Tauc-Lorentz formula are displayed in table 1.The real part in the absorption-free region (<2 eV) of the MOE film is slightly smaller than AA-MOE film, indicating a higher absorption of the MOE sample, which is corroborated by the absorption plot (see inset in figure 1(b)).The imaginary part presents the two features at ∼3.1 eV and ∼3.9 eV for the MOE sample attributed to charge transfer transitions [28,29,31] as well the feature at ∼2.5 eV, corroborating the observed in (αhν) 2 data (figure 1(b)).The AA-MOE sample shows a smooth dielectric function without the characteristics observed in the MOE sample, indicating that the AA-MOE sample possibly has a defect-free electronic structure.The occurrence of the features at ∼2.5 eV, ∼3.1 eV, and ∼3.9 eV could also be associated with a strained structure due to the annealing process, changes in the optical properties due to strain has been observed in BiFeO 3 for other authors [32], however, as observed from the diffraction patterns, a shift in the 2θ axis is absent, indicating that the samples are not strained.

Conductivity maps and I-V curves
Figure 2 shows topography and conductivity maps acquired simultaneously on the L 1 -BiFeO 3 films for a scan size of 400 nm × 400 nm.From topography images, a root mean square roughness of 7.8 nm and 8.6 nm was measured for MOE and AA-MOE-based film.As can be appreciated from    Figures 3(a) and (b) shows typical amplitude and phase loops for the off-field conditions obtained on the MOE and AA-MOE samples; these experiments were performed at 16 different sample locations to verify reproducibility.All 16 experiments showed hysteresis for amplitude and phase; the amplitude with a butterfly shape and the phase with a switch of ∼180 • , indicating that ferroelectric switching has been achieved.Figure 3(c) summarizes the statistics for the coercive electric field E c of the PFM switching experiments.According to the data density for the E c , sample MOE presents loops with E c slightly shifted toward the negative excitation axis.In contrast, the AA-MOE sample exhibits a loop with E c shifted toward the positive excitation axis.

PFM
Figure 4 presents the topography acquired simultaneously with the PR for the two samples.It is observed that there are domains with a size of 100-200 nm (compare the diameter of the black and white contrasts with the scale bar).The black-white contrast in the PR images indicates that the domain polarization is pointing out-of-plane, while an intermediate contrast may indicate an in-plane polarization or zero PR.As expected from the polycrystalline nature of L 1 -BiFeO 3 films, in-plane polarization was also present (figure S6).

Higher order harmonic measurements
Although both MOE-L 1 -BFO, and AA-MOE-L 1 -BFO samples reveal a clear domain contrast, it is necessary to determine whether the piezoelectric effect is the dominant one and whether it is responsible for the mechanical deformation detected in the PR images and amplitude/phase loops.As mentioned in the Experimental section, another effect that may contribute to the electromechanical response is electrostriction.From reference sample ARPPLN in figure 5(a), the electromechanical or PR signal is dominated by the first harmonic, indicating that the piezoelectric effect is dominant.One of the acquired spectra leading to the formation of the left graph for the ARPPLN is presented on the right side.From those spectra, the response of the first harmonic is of higher amplitude than that of the second harmonic.Figures 6(a  (30 nm thick) presents better PR than the AA-MOE sample (36 nm thick), which is explained by analyzing the distribution of ferroelectric domains and their boundaries from figure 4. The domain boundaries are considered as another degree of freedom to tune the electromechanical response of ferroelectric films [4].From figure 4, a clear distribution of domains with well-defined boundaries is observed, while the AA-MOE sample contains barely noticeable boundaries.The presence of well-defined domain boundaries allows for more efficient local polarization switching since the clamping effect is prevented as the grains and domains are less constrained [4].
As a means of determining the relationship of the first harmonic signal to the second harmonic in figure 5, the ratio of areas between these two signals (ω 0 /[ω 0 /2]) was obtained (figure S7(b)), i.e. the area enclosed by the PR data of the first harmonic is obtained and divided by the area of the PR data of the second harmonic, provided that the PR ⩾ 0. The results indicate an area ratio ω 0 /[ω 0 /2] = 14.49, 14.51, and 13.14 for ARPPLN, MOE-L1-BFO, and AA-MOE-L1-BFO, respectively.The area ratio ω 0 /[ω 0 /2] for the two samples MOE-L 1 -BFO, AA-MOE-L 1 -BFO is very close to the reference, and consequently, the electromechanical response detected is dominated by the piezoelectric effect.

Optimization of the withdrawal speed
Since the MOE-based solution results in a lower thickness than the AA-MOE-based solution, the solution viscosity and solvent evaporation can play an important role.During the withdrawal stage in the dip-coating process, the liquid film undergoes evaporation of the solvent from its surface.Hence, a greater thickness of the liquid film will further limit the evaporation of the solvent from its interior.Consequently, the drying time will be longer for thicker liquid films resulting in a longer aging time for the reagents, which in turn increases the time for atomic ordering.Therefore, it is plausible to expect a solid film (after the annealing process) with reduced chemical defects for relatively thick single-layer films.
Such a statement was tested by considering that for each route, the solution viscosity is constant η and that the liquid film thickness h obtained in the dip-coating process retains a relationship with the withdrawal speed (U 0 ) of the form h ∼ k (η U 0 ) n as described in [33], where n = {1/2, 2/3} depending on the η and U 0 values, and k relies on the gravity force and surface tension.If the statement in the previous paragraph is true, increasing U 0 in either route will result in an improvement of the physical properties.As parameters that indicate an improved film microstructure, we use the defect density and the piezoelectric response.We limit the study to AA-MOE films as we know them to have fewer density defects.
Figures 6 (a) and (b) show the conductivity maps for U 0 of 2 cm min −1 and 10 cm min −1 , where it is possible to appreciate a significant reduction of defects as the withdrawal speed increases.Figure 6(c) shows the piezoelectric coefficients obtained for U 0 of 2 cm min −1 of 36 nm thick, 5 cm min −1 of 39 nm thick, and 10 cm min −1 of 52 nm thick plotted against applied voltage.It is observed that increasing U 0 from 2 cm min −1 to 5 cm min −1 does not cause significant variations in the piezoelectric response, only slightly reducing the scatter of the data as seen from the error bars.However, an increase of U 0 to 10 cm min −1 does result in an increase of the piezoelectric coefficient of d From d eff results, we rule out the effect of depolarization fields, as these are most noticeable in ferroelectric films with thicknesses below 10 nm where a stable paraelectric phase coexists with a metastable ferroelectric one, resulting in pinched hysteresis (resembling a double loop), as reported in [34,35].There, the injection and trapping of charges into pre-existing interfacial defects during field cycling (wake-up) screens the depolarization field, stabilizing the ferroelectricity [34,35].However, an attenuation of the PR and piezoelectric coefficient d eff in BiFeO 3 films has been reported as the thickness decreases from 150 nm to <10 nm [36].Thus, a thickness effect in BiFeO 3 film on the piezoelectric coefficient is plausible, particularly in films that were deposited under different withdrawal speeds using the AA-MOE route.Here, it is important to note that the increase in d eff for the withdrawal speed of 10 cm min −1 may be due to the thickness effect, but also highlights the reduction of defects from the conductivity map in figure 6 since the evaporation of the solvent from inside the liquid film is reduced for this withdrawal speed.This led to longer aging times for the reagents, allowing to increase the time for atomic ordering, which in turn reduced defect density, resulting in an improvement of the PR.

Discussion
BiFeO 3 films of a single layer deposited by dip-coating solgel showed a structural phase related to ferroelectric material.However, conductivity pathways were observed for both routes based on MOE and AA-MOE dissolvents, which gave place to a defect density of 0.1511% and 0.0473% for the films based on MOE and AA-MOE, respectively.The I-V curves show a more abrupt high conductive pathway for the MOE films than the AA-MOE films.The observed conductivity pathway may be due to morphological defects with nanometric size or small regions where the ferroelectric phase was not reached, either because they are in an amorphous condition or because of the presence of a non-crystalline oxide.The origin of the observed defects is the solvent evaporation in the liquid film during the withdrawal stage.As was shown, the increase of the thickness of the liquid film, controlled by the withdrawal speed, reduced the solvent evaporation from its interior, allowing a longer time for the atomic ordering, which in turn reduced defects, both chemical and morphological.
The shift of the voltage-axis of the I-V curves may be related to the pristine polarization state of the samples.It is known that the resistivity of BiFeO 3 films can be tuned by performing changes in the polarization orientation (polarizationmediated resistive switching, RS phenomenon) [6].A preferential polarization state may be a consequence of an electric field built internally due to a charge imbalance in the material, which can be caused by chemical defects.This behavior is analogous to the imprint phenomenon observed in strain loops obtained at the nanoscale, where a displacement of the loop along the excitation axis is observed [37].As can be observed from the ferroelectric characterization section, both routes give rise to well-defined hysteresis loops at the nanoscale, but they display an imprint-like behavior, which agrees with the observed in the I-V curves, indicating that the samples have a preferential orientation state and that chemical defects play an important rule on the ferroelectric behavior of a single layer BiFeO 3 films.Again, the shift of the ferroelectric loops was suppressed by reducing the solvent evaporation from its interior in the withdrawal stage (compare figure 6(d) with figure 3).In addition, a twofold increase in the PR was achieved by controlling the solvent evaporation.
On the other hand, under the same deposition condition, the BiFeO 3 films based on the MOE route presented a higher electromechanical response than the AA-MOE route but also a higher defect density.In both routes, a clear piezoelectric domain distribution was observed with an electromechanical response dominated by the piezoelectric effect.Table 2 summarizes the main results of the comparison between MOE-L 1 -BFO, and AA-MOE-L 1 -BFO films.As observed from table 2, the bandgap of one-layer BiFeO 3 films agrees with previous experimental reports [11,28,30,31].The dielectric function for the MOE route displays some features, which have been previously assigned as charge transfer transitions through analysis of the electronic structure [28][29][30][31].The MOE-based films showed lower bandgap energy than those based on AA-MOE, indicating that this route may be useful in ferroelectric-based solar as it allows light absorption closer to the visible wavelength range.Regarding CAFM maps, it has been observed in [38] that the presence of morphological defect has an adverse effect on the incorporation of BiFeO 3 films in thin film solar cells devices as they act as charge transport paths.The charge transport paths prevent effective polarization of ferroelectric BiFeO 3 to configure the internal electric field useful in the separation of photoexcited carriers [39].The optimized AA-MOE route (withdrawal speed of 10 cm min −1 ) showed almost no conductive pathways ((see figure 6(b)), indicating that material is a good semiconductor that may be useful in thin film solar cells devices based on inorganic perovskites.However, MOE and AA-MOE routes at withdrawal speed of 2 cm min −1 showed an important density of conductive pathways, indicating that defects can be avoided with appropriate deposition parameters.The effective piezoelectric coefficient [4] describes the volume change of BiFeO 3 films under the AFM tip when an electric field is produced between the conductive AFM tip and the bottom FTO electrode.The values obtained for the optimized AA-MOE route using dip coating agree with the obtained by other routes [38] using spin coating.

Conclusions
Thin films of BiFeO 3 deposited by sol-gel dip-coating technique, using MOE and AA-MOE as precursor solutions, were compared to understand what happens during the formation of the first layer.The optical, conductivity, and ferroelectric response of the first layer (L 1 ) of the films were presented in this work.A dielectric function with features at ∼2.5 eV, ∼3.1 eV, and ∼3.9 eV was found for MOE-L 1 -BFO films.The same features were observed in the absorption coefficient plot, which was associated with charge transfer transitions in BiFeO 3 films.On the other hand, for AA-MOE-L 1 -BFO films, the absorption coefficient plot, and dielectric function shows a smooth behavior.The energy band gap for MOE and AA-MOE films were 2.43 eV and 2.75 eV, respectively.From the I-V curves, an abrupt high conductive pathway at −3 < V < + 3 occurs for sample MOE-L 1 -BFO, while for sample AA-MOE-L 1 -BFO, it can be seen, a soft high conductive pathway at −2 < V < +4.The coercive electric field E c of the PFM switching experiments shows sample MOE presents loops with E c slightly shifted towards the negative excitation axis, whereas AA-MOE sample exhibits loops with E c shifted toward the positive excitation axis.This agrees with the I-V curves, indicating that the samples have a preferential orientation state due to an internal electric field resulting from a charge imbalance in the material.The PR shows domains of 100-200 nm and in-plane polarization, and the effective piezoelectric coefficient turns out to be higher for MOE-L 1 -BFO films.However, by reducing the solvent evaporation during the withdrawal stage, the defect density was reduced, and the piezoelectric response was improved in the single layer BiFeO 3 films due to the reduction of morphological and chemical defects.Our BiFeO 3 films with thicknesses lower than 100 nm using methods industrially scalable as sol-gel are desired in ferroelectric memory applications.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Figure 1 .
Figure 1.(a) x-ray diffraction patterns of BiFeO 3 films with PDF:72-2112 (dash-dot green lines).Note that more layers (n > 1) were deposited to appreciate the diffraction peaks.(b) (αhν) 2 plots as a function of the photon energy.The inset displays the optical absorption, the black solid line corresponds to MOE-L 1 sample, and the red dash-dot line corresponds to AA-MOE-L 1 sample.The vertical dash-dot lines in the inset were included to guide the eyes about the occurrence of charge transfer transitions.(c) Real (ε 1 ) and imaginary (ε 2 ) parts of the complex dielectric function for the L 1 -BiFeO 3 films.

Figure 3 .
Figure 3. Amplitude R and phase θ loops for the off-field condition obtained on the BiFeO 3 films based on (a) MOE and (b) AA-MOE chemical dissolvents.(c) Box plot distribution to explain the degree of clustering of the coercive electric field Ec data obtained from 16 (R,θ) hysteresis loops; for a normal distribution, the box ends indicate ±0.67 sigma, the vertical line inside the box indicates the mean, while the two lines outside the box are observations extending up to ±2.69 sigma, where sigma is the standard deviation.
) and (b) correspond to the experiments performed on L 1 -BiFeO 3 based on MOE and AA-MOE dissolvents.Since the higher order harmonic measurement conditions are the same for samples MOE-L 1 -BFO and AA-MOE-L 1 -BFO with respect to the reference, it is possible to roughly compare the ω 0 response to have an idea about the effective piezoelectric coefficient (d eff ) of the samples.A linear regression was performed for samples MOE-L 1 -BFO and AA-MOE-L 1 -BFO against the applied voltage and it was determined that d MOE −L1−BFO eff = 1.2 × d ARPPLN eff and d AA−MOE −L1−BFO eff = 0.9 × d ARPPLN eff (See figure S7 (a)).As can be seen from these d eff results, under the same withdrawal speed conditions (2 cm min −1 ), the MOE sample
AA−MOE −L1−BFO−10cm/min eff = 2.2 × d ARPPLN eff .Figure 6(d) shows the amplitude and phase loops for the AA-MOE film with U 0 of 10 cm min −1 , which also shows a significant reduction in the voltage required to change the polarization direction of the material (compare figure 3(b) with figure 6(d)). 3 •5H 2 O, 99.99%] and iron nitrate [Fe(NO 3 ) 3 •9H 2 O, 99.99%] from Sigma-Aldrich were dissolved in MOE under stirring for 3 h at 80

Table 1 .
Fitting parameters of the BiFeO 3 complex dielectric function.

Table 2 .
Main results of the optical, electrical, and piezoelectrical properties of L 1 -BiFeO 3 films based on MOE and AA-MOE dissolvents.