Photonic integrated circuits (PICs) have become a powerful paradigm in photonic circuits owing to their high integration and multi-functional advantages in communication. Power splitters, which are the basic components of PICs, have a wide range of applications in signal detection, feedback circuits, and power balance circuits^[1]. Common power splitters include directional couplers, splitters based on adiabatic conical waveguides, multimode interferometric couplers, and Y-junction photonic power splitting structures, which typically use algorithms or numerical simulations to achieve a single splitting ratio^[2]. This increases not only the time required to fabricate the device but also the area occupied in the PICs.

Recently, MMI structures that use phase change materials to control power splitting have been developed. The attractiveness of optical phase change materials (O-PCMs) lies in their ability to enable the design of tunable photonic devices^[3,4]. In general, O-PCMs exist in two states: crystalline and amorphous. The refractive index of O-PCMs changes significantly when the phase changes between the amorphous and crystalline states, thus enabling the introduction of large modulation within the confines of a compact device. This feature has been used in the designs of PIC optical switches^[5–9], reconfigurable element optics^[10,11], and neural computing^[12–14]. In addition, experiments have shown that the O-PCM’s state can be changed repeatedly from amorphous to crystalline using laser radiation or electrical heating^[15]. The transition of the two states of the O-PCM should undergo a multistage phase transition process such that the performance is nonvolatile. These characteristics allow the O-PCM to maintain a state even without a power input, as described in Refs. [3,16,17]. In Ref. [18], Huan proposed an MMI power splitter using the GSST phase-change material, which can achieve four proportions of output power using only the conversion of crystalline and amorphous states of 4pixel×16pixel. However, this structure contains significant energy loss, and because 64 pixels are needed to be manipulated to change the ratio, manipulating real photonic circuits is difficult. In Ref. [19], it was proposed that using In2Se3, three proportions of the output power could be achieved by simply changing the two states of In2Se3 without changing the number and position of the round holes. References [18,19] used the MMI structure and inverse design for optimization, but the control of the two structures remained troublesome. Second, various operational factors can cause the ratio of the power splitter to drift and deviate from the preset 1:1 ratio, which may severely affect the accuracy of some experimental results. For example, any slight deviation in the size, shape, or position of the waveguide during the manufacturing process may cause the actual power splitting ratio to drift away from the design value. Even minor errors in the manufacturing process can significantly impact the power splitting ratio. Additionally, changes in temperature cause thermal expansion or contraction of the material, which may alter the geometry or refractive index of the waveguide. This, in turn, affects the power splitting ratio as the light transmission characteristics within the waveguide are modified by these changes. The single-point tunable and nonvolatile Y-junction photonic power splitter proposed in this paper can fully solve these problems and easily correct drifts in the power splitting ratio, which is important for a variety of experiments. Improving the utilization rate of power splitters in PICs and meeting the various needs of different PICs requires a tunable, easily operated, compact, and low-loss power splitter.

In this study, we propose a single-point-controlled Y-junction photonic power splitter that operates by controlling the change in the refractive index at only a single point, thereby realizing the tunability of the power ratio between the upper and lower output ports of the Y-junction photonic power splitter. Our Y-junction photonic power splitter realizes a tunable splitting ratio in the wavelength bandwidth from 1300 to 1650 nm using a single-point control, and its loss is about 0.7 dB. In the theoretical simulations, we gradually changed the refractive index in a circular hole region from 1 to 4.5, and the ratios of power splitting gradually change from approximately 2.2 to 0.8. Finally, we took Sb2S3^[20,21] at the wavelength of 1550 nm, adjusted the intensity of optical pulse or electricity to promote the formation of the phase change material intermediate state, and realized the refractive index value between crystalline and amorphous refractive indices by controlling the proportion of the phase change material. We therewith realized 6-level tunable power splitting ratios of 1.86, 1.70, 1.50, 1.34, 1.21, and 1.14 under single-point control with a 0.4 dB insertion loss. The proposed single-point Y-junction photonic power splitter is useful for improving the integration of PICs.

Figure 1(a) illustrates the comprehensive structure of a single-point tunable Y-junction photonic power splitter on an SOI platform. The device comprises an input waveguide, two output waveguides, and a coupling region situated on the SOI platform. The dimensions of the input and output waveguides as well as the coupling region are shown in Fig. 1(a). The height (h) of the input and output waveguides and the coupling region was set to 220 nm, and the width (Wg) of the input and output waveguides was 500 nm. The width (Wc) of the coupling region was 0.5 µm at the input side, and that at the output side was 1.05 µm. The length (L) of the coupling region was 2.02 µm. The shape of the coupling region was calculated by the shape optimization algorithm^[2]. An area of the coupling region was subsequently filled with a 220 nm thick layer of phase-change material after the silicon was etched away. The power splitting ratios of the upper and lower output waveguides of the device were initially equal (1:1) in the absence of any phase change material. However, upon introducing the phase change material, the splitting ratios between the upper and lower outputs could be tuned by controlling the phase change in the added material. Figure 1(b) illustrates the power ratio, and Fig. 1(c) shows the insertion loss (IL) of the upper and lower output ports, which resulted from the modulation of the refractive index change in the phase change material region within the 1300–1650 nm band. As shown in Fig. 1(b), a larger difference between the refractive indices of the phase change material region and those of silicon corresponds to a higher power ratio. Specifically, when the refractive index of the phase change material region is smaller than that of silicon, the power ratio exceeds 1; conversely, when the refractive index of the phase change material region is larger, the power ratio is less than 1. Figure 1(c) illustrates the IL of the upper and lower output ports resulting from the modulation of the refractive index change in the phase change material region within the 1300–1650 nm band. The maximum IL was observed at 0.7 dB when the refractive index of the phase change material region was 1.

One possible explanation for the observed phenomenon is that, in a conventional Y-splitter with a symmetric structure, light undergoes a 1:1 transmission along the device to the output port. This symmetry is inherent because the device material is silicon, and the structure is symmetrical both upward and downward. However, the refractive index changes at specific locations within the device, achieved through the incorporation of phase change materials, introduce perturbations^[22]. These perturbations disrupt the initial symmetric structure, alter the original optical field, and lead to distinct power ratios between the upper and lower output ports.

The flowchart in Fig. 2 illustrates the algorithm used for determining the location of a circular hole using the direct binary search (DBS) algorithm^[23–25]. Given that variations in the perturbation locations result in different device losses, the initial application of the DBS algorithm aims to identify the optimal position for the phase change material to minimize the overall device loss. We initially set the radius of the circular hole to 100 nm using a non-silicon material. The transmittances of the upper and lower output waveguides are denoted as Tup and Tdown, respectively. The figure of merit (FOM) is defined as FOM=Tup+Tdown. The coupling region was segmented into multiple circular regions with a radius of 100 nm. By employing the DBS algorithm, we sequentially adjusted the refractive indices of these circular regions to distinguish them from silicon. The corresponding FOM values were calculated and recorded along with the position number where the refractive index was altered. Subsequently, the circular regions were restored to their original state. Each circular region was systematically evaluated, and any enhancement in the FOM resulted in an update of the stored FOM value and position number. Upon completing the calculations for all regions, the optimal location of the circular hole that maximizes the FOM and minimizes the device loss can be determined. Subsequently, exploring the impact of the refractive index and radius variations in the region on the power splitting ratio becomes feasible.

A specific amount of energy is lost once light traverses the Y-splitter. An index denoted as the IL was introduced to quantify the energy loss: IL=−10×lgTup+TdownTin,(1)where Tin represents the power of the light source at the input waveguide.

Figure 3(a) shows the optical field diagrams corresponding to various phase change material positions and refractive indices throughout the optimization process. The phase change material positions are documented as n−m and represent the rank values within the designated coupling region. The coupling region of the device is divided into rows and columns of 12pixel×20pixel, with a row spacing of 0.1 µm and a column spacing of 0.1 µm. The center of each pixel corresponds to the center of the circular hole. The DBS algorithm searches for a pixel point that minimizes the loss. For example, the center of the circular hole numbered 9–17 is located at (0.9 µm, 1.7 µm) if the lower-left corner of the coupling region in the top view of Fig. 1(a) is taken as the origin. If the entire circular hole is outside the coupling region, it is discarded and not included in the comparison of the performance metrics owing to the irregular shape of the coupling region. The optical field diagram illustrates that, irrespective of the placement of the circular hole, a notable variation in the optical power between the upper and lower output ports occurs when the refractive index of the hole differs from that of silicon. Furthermore, the power output from the upper output port increases with a greater disparity between the refractive index of the phase change material and that of silicon. Figure 3(b) shows the power ratios at different phase change material positions shown in Fig. 3(a). Figure 3(c) shows the IL at the circular hole position shown in Fig. 3(a). The plot reveals that the IL was minimized at locations numbered 9–17. Consequently, we identified the optimal locations for the circular hole as positions numbered 9–17. In the follow-up studies, unless otherwise noted, our studies are based on the optimal positions numbered 9–17.

Figure 4(a) depicts the optical field diagrams at wavelengths of 1310 and 1550 nm for the refractive indices at circular holes of 1, 1.5, 2.5, and 3.5. As shown in Fig. 4(a), when the refractive index at the circular hole is equal to 1, most of the light is directed to the upper port. The optical power of the upper and lower ports gradually equalizes as the contrast diminishes between the refractive indices at the circular hole and silicon. Figure 4(b) shows the power splitting ratios between the upper and lower output ports when the light source in the 1300–1650 nm band was used as the input as the refractive index at the circular hole changed from 1 to 4.5, when the radius of the circular hole was set to 100 nm after determining the location of the circular hole. Figure 4(b) also shows that a higher contrast in the refractive index of the circular hole with silicon leads to an increased power splitting ratio. When the refractive index at the circular hole surpasses that of silicon, the output power ratio of the upper port to the lower port falls below 1. Figure 4(c) shows the IL of the upper and lower ports after the Y-junction photonic power splitter for different wavelengths of light with a continuously varying refractive index. Figure 4(c) shows that a larger refractive index contrast results in a higher IL, whereas a decreasing contrast leads to a reduced IL. Across the 1300–1650 nm range, the maximum loss is merely 0.7 dB, indicating that the additional IL introduced by this round hole result is very small.

We also explored the effect of variations in the radius of the circular hole on the power splitting ratio and IL. Figure 5(a) illustrates the optical field diagrams with a fixed refractive index of 2.7 at the circular hole and varying hole radii of 50, 100, 200, and 300 nm. As shown in Fig. 5(a), the light transitions from an almost equal crossover state to a predominantly output state from the upper port with an increase in the radius of the hole, indicating a gradual increase in the power distribution ratio between the upper and lower ports. In Fig. 5(b), the fluctuations in the power splitting ratio within the 1300–1650 nm band are depicted as a function of the radius of the circular hole for a fixed refractive index of 2.7. Figure 5(b) clearly illustrates that the power splitting ratios between the upper and lower output ports progressively increase with the radius. This behavior arises because a larger circular hole radius introduces a more significant perturbation, causing the ratios to deviate further from the ratio of 1 that is observed in the absence of the circular hole. We also observed that when the radius of the hole was greater than 150 nm, a large wavelength of light passing through the device was associated with a greater power ratio between the upper and lower output ports. Figure 5(c) shows the IL as a function of the radius of the phase for a fixed refractive index of 2.7. When the refractive index is fixed, except for the loss in Fig. 4(c), the loss is nearly independent of the radius of the hole; the loss introduced by variations in the radius is very small. The maximum loss was only 0.35 dB at a radius of 0.3 µm and a wavelength of 1450 nm.

A designated phase change material was incorporated into the circular hole region to change the power splitting ratio through adjustments to the phase change of the material. Sb2S3 is a good phase change material because it exhibits almost zero absorption loss in the 1300–1650 nm band [the imaginary part of the refractive index (k) is 0]^[26], which makes it very suitable for mode matching with silicon waveguides. It has a crystalline refractive index of about 3.3 and an amorphous refractive index of about 2.7 in the 1300–1650 nm band (Δn≈0.6). Because the refractive index change is only 0.6, we set the radius of the circular hole to 150 nm^[27] to get a larger tuning range of the power splitting ratio. Given that the refractive index change of Sb2S3 is limited to the conversion between the crystalline and amorphous states, achieving multiple power splitting ratios through modulation is unattainable. Therefore, we propose realizing various power splitting ratios by controlling the depth of the phase transformation in 220 nm thick Sb2S3. Amorphization is achieved via rapid quenching by heating the point area to a melting point of 550°C at a heating rate of more than 109Ks−1 using a laser or electrical pulse. Crystallization is performed by heating the material to above the glass transition point at 275°C and below the melting point at 550°C, then quenching it slowly^[28]. In Ref. [28], Gao et al. utilized different pulsed lasers (550 nm wavelength, 50 fs pulse width, and 1 kHz repetition frequency) to achieve varying degrees of phase transformation of 70 nm thick Sb2S3 films, such as 150 nJ single pulse or 140 nJ double pulse to achieve 20% amorphization, 250 nJ single pulse or 170 nJ double pulse to achieve 60% amorphization, and 270 nJ single pulse or 190 nJ double pulse to achieve 90% amorphization.

The progressive phase transition of 220 nm thick Sb2S3 is illustrated in Fig. 6(a), where the red region represents amorphous Sb2S3, the green region represents crystalline Sb2S3, and the blue region corresponds to the silicon material. By modulating the heat source to achieve an incremental phase transition of 44 nm, the 220 nm thick amorphous Sb2S3 underwent gradual phase conversion into crystalline Sb2S3. This process resulted in six refractive index values encompassing both the amorphous and crystalline states, which can be controlled for modulation. Figure 6(b) illustrates the power splitting ratios and IL corresponding to the upper and lower output ports at different phase transformation depths at a wavelength of 1550 nm during the gradual phase transition of Sb2S3. As shown in Fig. 6(b), the power splitting ratios between the upper and lower output ports during the gradual phase transition of Sb2S3 were observed to be 1.86, 1.70, 1.50, 1.34, 1.21, and 1.14. In addition, the maximum IL of Sb2S3 in the amorphous state was measured at 0.34 dB. This approach, which employs Sb2S3 as the material for the circular hole and utilizes gradual phase transition, enables the realization of multiple controlled power splitting ratios using a straightforward single-point control mechanism with minimal IL. Figure 6(c) shows the optical field plots for different degrees of phase change shown in Fig. 6(a). As shown in Fig. 6(c), the power splitting ratio between the output of the upper and lower ports shown in Fig. 6(b) gradually decreases as the phase change degree deepens. Figure 6(d) presents a cross-sectional view of the optical fields at the PCM at different degrees of phase change, showing how the PCM with different degrees of phase change influences the output power of the upper and lower ports. As shown in Fig. 6(d), because the refractive index of the phase change material differs from that of silicon, the light appears to be divided into two beams of different sizes, and a greater refractive index contrast is associated with a greater degree of separation. Figures 6(c) and 6(d) illustrate the power splitting processes of the single-point tunable Y-junction photonic power splitter.

Figure 7(a) shows the results of the Y-junction splitter array, which is a cascade of a series of Y-junction splitters. By controlling the phase change of the PCM, we realized that the powers of the cascaded output ports are different and the use of Euler-bent waveguides in the bent section reduces the bending losses. Figure 7(b) shows the optical field diagram of the Y-splitter cascade. The figure shows that, when the two are divided, the powers of the upper and lower output ports are obviously different, and after the cascade is divided into four and eight ports, we can control the power of each port through the phase change of the phase change material and adjust the weight. Compared with the results in Ref. [18], in which the power splitting ratios are turned with many PCM points and correspond to complex control electronic circuits, our approach presents a minimalist control method.

Based on the preceding investigation, we established the network depicted in Fig. 7(c) by linking the single-point tunable Y-junction photonic power splitter in a cascaded manner and receiving a 16× Y-junction splitter array. Because we can easily regulate the power splitting ratios between the upper and lower output ports using an application-specific integrated circuit (ASIC), the tunable weights among the individual nodes within this network become straightforward. The Y-junction splitter array consists of 16 splitters that were formed after cascading through five stages, resulting in 32 outputs. The 6 ratios of each device can realize multiple power values at each output port after 5-stage cascading. If all splitters have a power splitting ratio of 1.8:1, the maximum and minimum power ratios at the output ports can be 18.9:1. Different power splitters can be configured at each stage to achieve higher ratios. Consequently, the single-point tunable Y-junction photonic power splitter introduced in this study has substantial potential for future applications in photonic neural networks.

The single-point tunable Y-junction photonic power splitter introduced in this study offers notable advantages, including a broad wavelength bandwidth (1300–1650 nm), low loss (0.7 dB), and a continuously tunable power splitting ratio. The device, comprising a single-phase change material with a 100 nm radius, allows for the control of the power splitting ratio between the upper and lower output ports of the Y-splitter. This is achieved through manipulation of the material’s phase change, without requiring alterations to the device structure. Furthermore, the incorporation of phase change materials, which exhibit substantial refractive index differences between the crystalline and amorphous states, coupled with a gradual phase-change approach, enables the realization of multiple power splitting ratios for the Y-splitter. This device proves instrumental in the tunable weighing of the photonic neural networks. In summary, the proposed Y-splitter regulation outlined in this paper has significant implications for advancing the integration of PICs and simplifying their manipulation, thereby contributing to the realization of optoelectronic fusion chip solutions.