Thin-film lithium niobate（TFLN）has emerged as a promising material for integrated photonics because of its excellent properties inherited from bulk single crystals as well as its relatively high index contrast and subwavelength light confinement^［1-3］. The TFLN platform can host diverse high-performance photonic devices such as electro-optic modulators^［4-7］，acousto-optic modulators^［8-9］，frequency combs^［10-12］，and classical and quantum nonlinear frequency converters^［13-18］. Electrically pumped lasers and photodetectors have also been introduced via heterogeneous integration^［19-21］. For second-order nonlinear photonic devices^［22］，efficient frequency conversions can be obtained based on ferroelectric domain engineering in x- or z-cut TFLNs^［23-24］，and the ferroelectric domains usually possess a 180° antiparallel configuration. In addition，another type of ferroelectric domain structure exists，which is typically in either a head-to-head or tail-to-tail configuration.

The fabrication of head-to-head domain structures can be traced back to the 1980s^［25］. Periodically poled LiNbO₃ with a head-to-head domain structure was fabricated using the crystal growth striation method^［26］ and was used to demonstrate the optical properties of an ionic-type photonics crystal. Atomic force microscopy（AFM）tip poling is commonly used to fabricate domain structures in bulk and thin-film ferroelectric crystals for applications in domain-wall nanoelectronics^［27］. However，AFM poling is time consuming and cannot be used for large-area fabrication. The electric-field poling technique is based on semiconductor planar lithography and has the advantages of arbitrary domain patterning and mass production. To date，only a few studies on the fabrication of head-to-head ferroelectric domain structures in TFLN have been conducted^［28-29］. However，a comprehensive and insightful fabrication technique for conductive domain walls based on electrical-field poling is essential.

In this study，we investigated in detail the fabrication of head-to-head/tail-to-tail domain structures in x-cut congruent TFLN using the electric-field poling technique，and we characterized the domain structure using confocal second-harmonic microscopy and piezoresponse force microscopy（PFM）. The relationship between different poling electric field intensities and domain inversions was studied，and the numerical simulation results are found to be in good agreement with the domain growth trend. Based on the experimental results，a new scheme for monitoring the domain inversion process in real time was proposed，which significantly improves the time efficiency of the fabrication of large-area and high-performance nonlinear photonic devices.

In our experiments，the thickness of x-cut congruent TFLN as produced by NanoLN（Jinan Jingzheng Electronics Co.，Ltd.）was 600 nm，with a silicon substrate and a 3 μm silica buffer layer. A periodic interdigital electrode was selected to fabricate a large-area head-to-head domain structure in x-cut congruent TFLN. An asymmetric staggered electrode distribution was adopted to achieve a specific duty cycle of the domain structure，a schematic of which is shown in Fig. 1（a）. The width of the domain inversion area was W_dom. Each period consisted of two regions with lengths L_inv and L_un，respectively. Arrows indicate the direction of spontaneous polarization. Domain inversion was expected to occur in the area between two finger electrodes when a high voltage was applied. The duty cycle of the domain inversion structure could be manipulated by adjusting the ratio of L_inv and L_un. A spatial electric-field distribution near the interdigital electrode was simulated using COMSOL，as shown in Fig. 1（b）. As an example，at a voltage of 500 V，the electric field near the electrode edge on the surface of the TFLN could be higher than 50 kV/mm，which is approximately 2.5 times that of the coercive field for a congruent LN（E_sc=21 kV/mm）^［30］. This high electric-field strength ensures domain inversion. We note that the electric field increased as it approached the tip of the electrode. A rapid increase was observed at the electrode tip，where the electric field reached 100 kV/mm. The simulation results indicated that domain inversion near the tip was of higher priority during the electric poling process.

The fabrication procedures for the head-to-head and tail-to-tail domain structures are illustrated in Fig. 1（c）. A nickel electrode with a thickness of 100 nm was deposited on the x-cut congruent TFLN，and high-voltage pulses were applied to the inverted domain. Confocal second-harmonic microscopy and PFM were used to characterize the inverted ferroelectric domain structure. Both results confirmed the existence of head-to-head and tail-to-tail domain structures.

A schematic of the electric field poling process is shown in Fig. 2（a）. Electric pulses were first generated using an arbitrary wave generator（AWG）（Agilent）and then amplified 2000 times using a high-voltage amplifier（Trek）. A pair of high-voltage probes was applied to the interdigital electrode to transmit electric pulses. A series-sampling resistor was used to measure the current passing through the domain walls. The waveforms of both initial electric pulses and the domain wall current were characterized using a digital oscilloscope. A typical waveform of a poling electric pulse is shown in Fig. 2（b）. The voltage rises in 0.5 ms and holds for 17.5 ms，and then ramps down in 2 ms. The poling voltage U_pvaried from 360 V to 500 V in our experiment.

Several studies have reported that the coercive field in TFLN is larger than that in bulk LN^{［23，28，30］}. Here，we studied in detail the relationship between the poling electric field strength and domain inversion. The interdigital electrode in our experiments had a period of 30 μm，L_inv and L_un were 21 μm and 9 μm，respectively，and W_dom was set to 100 μm. For the bulk LN domain inversion，the required poling voltage was estimated to be proportional to the coercive field and to the length in the spontaneous polarization direction. In this conventional perspective，the domain inversion area at L_inv=21 μm corresponded to poling voltage U_p≈460 V. In the poling procedure，we applied electric pulse waveforms at U_p=360，420，440，460，and 500 V. Domain inversion states of the five TFLN samples were then characterized using confocal second-harmonic microscopy，as shown in Fig. 3. Figure 3（a）shows that U_pwas set to 360 V. Normally，the electric field strength is not sufficiently high for domain inversion，particularly considering that the coercive field of TFLN is much higher than that of bulk LN. However，domain inversion was observed when electric pulses of U_p= 360 V were applied，as shown in Fig. 3（a）. A small number of domains were localized and inverted at the tip of the electrode. The inverted domain was shaped like a needle and grew several micrometers in the direction of the crystal axis. This anomaly was attributed to a dramatic increase in the electric field at the tip of the electrode. In fact，the electric field at the end of the electrode could be much higher than 21 kV/mm at U_p=360 V in our simulation. When electric pulses at a higher U_p（420 V）were applied，domain inversion around the electrode tip became more evident，as shown in Fig. 3（b）. When electric pulses with U_p=440 V were applied，domain inversion occurred over the entire overlapping interdigital electrode area，as shown in Fig. 3（c）. The triangular domain structure indicated that the velocity of the domain growth in the longitudinal direction was greater than that in the transverse direction. At U_p = 460 V，the transverse direction of the domain structure was fully overgrown，whereas some areas remained in the longitudinal direction that were not completely reversed，as shown in Fig. 3（d）. A thorough domain inversion was achieved using electric pulses of U_p = 500 V. In this sample，the head-to-head and tail-to-tail domain structures were successfully fabricated，as shown in Fig. 3（e）. We concluded that the poling voltage played a significant role in the domain inversion process. In addition，for samples with different interdigital electrode periods under the same poling voltage，different numbers of high-voltage pulses（higher than the coercive field of the TFLN）should be applied to reach the same duty cycle.

Figures 4（a）and（b）present phase images of the inverted domain structure under periods of 30 μm and 20 μm，respectively. The phase difference between the inverted（orange area）and original（purple area）domains was measured as 180° using PFM. Figure 4（c）shows the phase profile of the domain structure corresponding to the white dashed lines marked in Figs. 4（a）and（b）. As Fig. 4（c）shows，the duty cycles of the inversion domains were approximately 46% and 20%，respectively，in accordance with the designed interdigital electrodes.

Although confocal second-harmonic microscopy and PFM both provide clear characterization of the domain structure，they require a considerable amount of time，thus reducing the time efficiency of domain structure fabrication. Herein，we provide a new scheme for the real-time monitoring of the domain inversion process. Domain walls in ferroelectric materials have been studied for a considerable time，and the unique conducting properties of domain walls greatly extend the function of ferroelectric crystals. Head-to-head and tail-to-tail domain walls in x-cut lithium niobate crystals exhibit conductivity ^［29］. However，the conductivities of x-cut pure congruent TFLN domain walls have yet to be reported. We preliminarily studied the conduction characteristics of the domain walls during the poling process and observed a correlation between the x-cut TFLN domain structure and its conductance，which can be further utilized to monitor the domain inversion process in real time.

Using the setup shown in Fig. 2（a），we measured the current passing through the TFLN samples. A typical result is shown in Fig. 5（a），where the red curve is the waveform of the poling electric pulse and the blue curve is the current through the sampling resistor，which also represents the current through the TFLN samples，as they are connected in series. Fig. 5（a）shows that a lower dip exists at 0‒2.5 ms. The current may have derived from the rapid change in the applied voltage and the subsequent charge-discharge process. The applied voltage corresponding to this current did not reach the coercive field and did not affect the domain inversion process. A sampling current higher than 60 μA was recorded when the 10^th pulse was applied，with a poling voltage of U_p= 500 V. To determine the correlation between the domain structure and its conductance，we measured currents in four TFLN samples. These values were converted into conductance values，as shown in Fig. 5（b）. These samples correspond to four stages in the domain inversion process，as shown in Fig. 5（c）‒（f）. We found that perceptible conductance occurred when the domain tips reached the opposite electrode，as in the case of sample 1. As more of the inverted domains growing along the TFLN z axis touched the opposite electrode，the measured conductance increased significantly（samples 2 and 3）. In sample 4，the head-to-head and tail-to-tail domain structures were completely fabricated，and the conductance tended to stop increasing. Conductance appeared to be an appropriate criterion for evaluating the domain inversion state. We attributed this phenomenon to the high conductance of the head-to-head domain boundaries^［27］. The conductance of head-to-head domain walls was verified to be several orders of magnitude greater than that of the tail-to-tail domain walls^［29］. In Fig. 5（c），the head-to-head domain walls began to form near the zigzag boundaries；at this stage，and the conductance began to increase，corresponding to the initial phase of the domain inversion. When more zigzag boundaries were formed，as shown in Fig. 5（d），the conductance increased. The increasing conductance implied that the head-to-head domain walls grew rapidly during this stage. As Fig. 5（e）shows，the domains merged in the transverse direction，and the formation of the head-to-head domain structure was nearly complete. At this stage，the conductance reaches a maximum value. When the domain inversion process was complete，as shown in Fig. 5（f），the conductance stopped increasing because no additional head-to-head domain walls were formed. The relationship between the head-to-head domain walls and conductance of the sample provides a scheme for monitoring the head-to-head domain inversion process，and the stage can be estimated in real time.

We fabricated large-area head-to-head and tail-to-tail domain structures in an x-cut TFLN platform using an electric poling technique. Interdigital electrodes were specifically designed to adjust the shape and duty cycle of the inverted domain area. First，we investigated the effects of the poling voltage on the domain inversion process. The poling voltage was confirmed as a key factor in the domain inversion process. Although the lower-voltage domain did invert，the inversion process became significantly more drastic when a higher voltage was applied. Then，under an appropriate poling voltage and pulse number during fabrication，the domain inversion process was successfully artificially manipulated. Results show that the fabricated domain walls exhibit distinct conductivities，and the conductance of the domain walls is highly dependent on the domain structure. Once the head-to-head domain structures are formed，the conductance of the domain walls increases rapidly. The conductance continues to increase as the head-to-head domain is continuously produced. When the domain inversion is nearly complete，the conductance tends to reach its maximum. Based on this phenomenon，a new scheme for monitoring the domain inversion process in real time was proposed，ensuring that the electric poling technique for congruent x-cut TFLN is more efficient. The mechanism investigated in this study was the irradiation of functional nanoelectronic devices based on domain walls. The fabricated large-area periodically poled head-to-head/tail-to-tail domain structure provides a foundation for integrated second-order nonlinear photonic and acousto-optic devices.