**Preface to "Sonic and Photonic Crystals"**

This Special Issue on "Sonic and Photonic Crystals" is focused on broad applications of the results involving characterizations of the sonic and photonic crystal properties.

Various applications of photonic crystals are presented. The gradient cavity, waveguide, switch, and spatial beam filtering with autocloned photonic crystals are studied and discussed [Huang et al., Wang et al., Jao et al., Francis et al., Azizpour et al., Ren et al.]. Studies of the metamaterial of crystal structures can be found in [Luan, Nguyen et al., Li et al., Chang et al.], in which the negative effects of star-shaped structures were investigated and also discussed with respect to application.

Elastic wave propagations in phononic crystals is also an interesting topic. Elastic wave propagations in metamaterials, elastic piezoelectric phononic crystals, and energy harvesting phononic devices are also covered in [Le et al., Lv et al., Liang et al., Deng et al.].

Additional studies examining various photonic crystal applications also form part of this collection. The deterministic insertion of KTP nanoparticles into polymeric structures, flexible photonic nanojets with cylindrical graded-index lens, and photonic crystal fiber investigations are presented [Nguyen and Lai, Liu, Zhang et al., Yang et al.]. The design of a polarization splitter and converter based on square lattice photonic crystal fiber is investigated in [17, 18] and includes a discussion of the polarization characteristics of photonic crystals. Finally, a review paper of the recent advances in colloidal photonic crystal-based anticounterfeiting materials is included [Ren et al.].

This Special Issue presents and discusses the work of scientists studying a wide range of sonic/photonic crystal applications toward advancing this field.

> **Lien-Wen Chen, Jia-Yi Yeh** *Editors*

## *Article* **Trapping and Optomechanical Sensing of Particles with a Nanobeam Photonic Crystal Cavity**

#### **Lin Ren \*, Yunpeng Li, Na Li and Chao Chen**

School of Aviation Operations and Services, Aviation University of air force, Changchun 130022, China; ew\_radar@163.com (Y.L.); lina0629@163.com (N.L.); chenchao.19830823@163.com (C.C.)

**\*** Correspondence: renlin\_ok@163.com; Tel.: +86-13504329261

Received: 5 December 2018; Accepted: 17 January 2019; Published: 22 January 2019

**Abstract:** Particle trapping and sensing serve as important tools for non-invasive studies of individual molecule or cell in bio-photonics. For such applications, it is required that the optical power to trap and detect particles is as low as possible, since large optical power would have side effects on biological particles. In this work, we proposed to deploy a nanobeam photonic crystal cavity for particle trapping and opto-mechanical sensing. For particles captured at 300 K, the input optical power was predicted to be as low as 48.8 μW by calculating the optical force and potential of a polystyrene particle with a radius of 150 nm when the trapping cavity was set in an aqueous environment. Moreover, both the optical and mechanical frequency shifts for particles with different sizes were calculated, which can be detected and distinguished by the optomechanical coupling between the particle and the designed cavity. The relative variation of the mechanical frequency achieved approximately 400%, which indicated better particle sensing compared with the variation of the optical frequency (±0.06%). Therefore, our proposed cavity shows promising potential as functional components in future particle trapping and manipulating applications in lab-on-chip.

**Keywords:** optical force; photonic crystal cavity; particle trapping; optomechanical sensing

#### **1. Introduction**

Optical force is regarded as an ideal tool for trapping and manipulating vulnerable particles due to its contactless and nondestructive properties [1]. A photonic crystal cavity, formed by introducing a defect into a periodic arrangement of different materials, is able to build an optical potential around the defects and traps of the manipulated particles with the advantage of the cavity-enhanced optical force [2–4]. Therefore, the enlarged optical force via a high-quality factor photonic crystal cavity is promising for trapping and manipulating various vulnerable particles. Compared with two-dimensional photonic crystal structures [5–8], one-dimensional nanobeam photonic crystal cavities have attracted considerable attention due to their compact nanobeam structures [9], ease of integration of coupling waveguides [10], high quality factors [11] and small mode volumes [12]. The relationship of the optical gradient force between the quality factor, mode volume and transmission has been systematically analyzed in the nanobeam photonic crystal cavities [13]. Among these works on cavity enhanced optical forces, the primary target is to trap and manipulate particles with optical power as low as possible, since the enhanced optical absorption by the high-quality-factor cavities would have side effects on biological particles such as bacteria, proteins, and viruses due to a severe increase in temperature [14]. X. Serey et al. have compared several photonic crystal cavities and showed that polystyrene particles with different diameters could be trapped with the input optical power of 10 mW [15]. N. Descharmes et al. have presented 500-nm dielectric particles with low intracavity powers of only 120 μW for long-time optical trapping [16]. Despite these advances, optical trapping power still needs to be further lowered to reduce the influence on biomolecules. Furthermore, the optomechanical coupling between the cavity and the particle due to the self-induced back-action optical trapping is expected to isolate, analyze, or sort different particles depending on their mechanical performances [1,16–18].

In this work, we proposed to utilize a larger-center-hole nanobeam photonic crystal cavity [19,20] in silicon for trapping and opto-mechanical sensing particles. The radius of air holes increases in the center of the defect region in the designed cavity so that large optical force can be obtained. For the calculation of optical force, a polystyrene particle was introduced and located at the center above the cavity, while the trapping cavity was set in an aqueous environment. The optical force and potential on the introduced particle with radius of 150 nm were calculated at a temperature of 300 K. For particles captured at this condition, the input optical power is required to be as low as 48.8 μW to generate an optical potential of 4.14 × <sup>10</sup>−<sup>20</sup> J (10 kBT). We additionally calculate the optical and mechanical frequency shift for the particles with radius ranging from 100 nm to 500 nm. Accordingly, the change of the mechanical frequency shift can be detected and analyzed based on the optomechanical coupling between the particle and the designed cavity.

#### **2. Results**

#### *2.1. Design of a Photonic Crystal Cavity*

We designed a larger-center-hole nanobeam photonic crystal cavity fabricated on silicon-on-insulator for particle trapping and sensing. The structure of the cavity is shown in Figure 1a. The radius of air holes, *r*n, linearly increases from the mirror region to the center of the cavity defect region, where *r*<sup>n</sup> = *r*<sup>0</sup> + 4 × *n* (nm). Here, *r*<sup>0</sup> is fixed at a radius of 100 nm in the whole mirror region, and the largest hole in the center of the cavity defect region is set with a radius of *r*<sup>7</sup> = 128 nm. For the entire structure, the lattice constant *a*, the distance between two adjacent holes, is kept as 365 nm. The width and the thickness of the nanobeam are 480 nm and 220 nm, respectively.

**Figure 1.** (**a**) Top-view and (**b**) Side-view of the structure of the larger-center-hole nanobeam photonic crystal cavity, above which a polystyrene particle is introduced. The position of the polystyrene particle deviates from the center of the largest hole of the cavity in the *x* direction is denoted as Δ*x*. The distance between the center of the polystyrene particle and the top surface of the nanobeam cavity in the *z* direction is denoted as Δ*z*.

For particle trapping and sensing, the cavity was simulated in an aqueous environment with a polystyrene particle located above the center of the cavity, as shown in Figure 1b. The position of the polystyrene particle deviated from the center of the largest hole of the cavity in the *x* direction is denoted as Δ*x*. The distance between the center of the polystyrene particle and the top surface of the nanobeam cavity in the *z* direction is denoted as Δ*z*. The optical trapping force and the optical potential on the manipulated polystyrene particle were analyzed by varying its position based on the designed larger-center-hole nanobeam cavity. In the simulation, the material of the designed cavity and the particle are silicon and polystyrene, respectively, with the corresponding refractive index of 3.42 and 1.59. The whole cavity is set in aqueous environment, whose refractive index is 1.33. The Young's modulus of the polystyrene is 3300 MPa.

The optical frequency of the defect mode in the photonic band structure of the designed cavity is presented in Figure 2a. The photonic band structure is calculated by plane wave expansion (PWE) method. It can be seen that the optical resonant mode is confined in the photonic band gap of the mirror structure. And the optical wavelength of the resonant cavity mode is calculated to be 1.5477 μm without the polystyrene particle. The optical transmission spectra of the designed cavity are simulated by finite-difference time-domain (FDTD) method and presented in Figure 2b. The optical quality factor (*Q*) of the cavity can be calculated from the transmission spectra and the *Q* factor for the designed cavity is 3968.

**Figure 2.** (**a**) The photonic band structure of the larger-center-hole nanobeam photonic crystal cavity. The optical frequency of the defect mode of the designed cavity is also presented in the photonic band gap with dash line. (**b**) The optical transmission spectra of the designed cavity calculated by finite-difference time-domain (FDTD) simulation with resonant wavelength of 1.5477 μm and *Q* factor of 3968.

To study the optical trapping force and the optical potential on the polystyrene particle, the optical fields in *x* and *z* direction of the designed cavity with the polystyrene particle located at different positions were simulated and given in Figure 3. Here, the location of the polystyrene particle is set as (a) Δ*x* = 0 nm, Δ*z* = 5 nm and (b) Δ*x* = 190 nm, Δ*z* = 5 nm, respectively. It can be seen that the particle is located at the position of the largest optical field with Δ*x* = 190 nm, Δ*z* = 5 nm in Figure 3b. Accordingly a strong optical trapping can be achieved for the polystyrene particle at this position. This result can also be obtained by the calculation for the optical force and potential being applied on the polystyrene particle in the following, where a strong optical trapping potential for the polystyrene particle is located at Δ*x* = 190 nm, Δ*z* = 5.

**Figure 3.** The optical fields in *x* and *z* direction of the designed cavity with polystyrene particle located at different position (**a**) Δ*x* = 0 nm, Δ*z* = 5 nm, and (**b**) Δ*x* = 190 nm, Δ*z* = 5 nm.

#### *2.2. Optical Force and Potential on a Particle*

The optical force on the particle was calculated by the distribution of the electric field and magnetic field with Maxwell Stress Tensor [21]. In our calculation, the optical field of the designed cavity with the introduced polystyrene particle was first simulated by finite element analysis. Then, the optical force (*f*) on the particle was calculated by integrating the Maxwell stress tensor over the surfaces of the particle:

$$f = \int\_{s} T \cdot d\vec{n} \,\tag{1}$$

Where *S* is the surface of the particle, d - *n* is the surface normal and *T* denotes the Maxwell stress tensor given as:

$$\mathbf{T}\_{\mathrm{i\bar{j}}} \equiv \varepsilon \left( \mathbf{E}\_{\mathrm{i}} \mathbf{E}\_{\mathrm{j}} - \frac{1}{2} \delta\_{\mathrm{i\bar{j}}} \mathbf{E}^{2} \right) + \frac{1}{\mu\_{0}} (\mathbf{B}\_{\mathrm{i}} \mathbf{B}\_{\mathrm{j}} - \frac{1}{2} \delta\_{\mathrm{i\bar{j}}} \mathbf{B}^{2}), \tag{2}$$

Where *ε* is the electric constant and *μ*<sup>0</sup> is the magnetic constant, *E* is the electric field, *B* is the magnetic field, and *δij* is Kronecker's delta function.

It should be noted that the potential is affected by the presence of the particle, since the refractive index of the polystyrene particle is larger than that of water. Thus, the optical force and potential were calculated with the particle located above the cavity. In the simulation, the light is input from one port of the nanobeam waveguide. Its frequency is set as the optical resonant frequency of the optical cavity with the particle located 5 nm above the cavity for the x direction (Δ*x* is changed with Δ*z* = 5). The calculated results are presented in Figure 4a. The calculation of optical force and potential on the particle in the z direction, in which the particle is located at a fixed Δ*x* = 190 nm and a varied Δ*z*, is shown in Figure 4b. Here, the negative and positive optical force refer to the direction of the force on the particle. For example, in the x direction, the negative or positive optical force refers to the force direction opposite or along the x axis, respectively. For the z direction, the negative optical forces indicate that the direction of the force is opposite the z axis, so that the particle will be trapped just above the cavity.

**Figure 4.** The optical force and potential on the particle (**a**) in the x direction and (**b**) in the z direction with the input light power of 100 μW.

Since dielectric particles are attracted in the region of strongest electric field, a strong optical trapping can be realized in an optical cavity. When the light is input into the cavity, an optical potential in the *x* direction and the *z* direction around the cavity defect can be built for trapping the polystyrene particle with radius of 150 nm. For particle capturing, the minimum potential of 4.14 × <sup>10</sup>−<sup>20</sup> J (10 kBT) is required at temperature of 300 K. For x direction, the potential well depth is 8.48 × <sup>10</sup>−<sup>20</sup> J when the input power is 100 μW, as shown in Figure 4a. Since that the potential well depth increases linearly with the input light power, the input power is required to be 48.8 μW when the well depth reaches 4.14 × <sup>10</sup>−<sup>20</sup> J (10 kBT). In addition, we also calculated the optical force and potential in the z

direction while the particle is located at the strongest trapping of the potential well with Δ*x* = 190 nm. As shown in Figure 4b, when the input power is 100 <sup>μ</sup>W, the potential well depth is 9.17 × <sup>10</sup>−<sup>20</sup> J. Thus, when the well depth reaches 10 kBT, the input power is required to be 45.1 μW. Taking the required input power for both *x* and *z* direction into consideration, it can be concluded that the input optical power is predicted as low as 48.8 μW for particles captured at 300 K. It should be mentioned that the coupling loss between the input light and the trapping cavity has been considered in our simulation. Thus, the input power for particle trapping using this proposed cavity is much lower compared with other reported work (120 μW–10 mW) [15,16], which benefits to mitigate the side effect on biological particles. In addition, it is possible to attract more than one particle from the water dispersion. It can be seen from Figure 4a that there is more than one potential well in the x direction. If the input optical power goes up, more potentials can be built for trapping multiple particles by the nanobeam photonic crystal cavity.

#### *2.3. Optomechanical Sensing*

Optomechanical coupling between cavity and particle inherently exists due to the mechanism of self-induced back-action optical trapping [16]. Based on optomechanical coupling, when we distinguish different particles, both the optical and mechanical frequency shift can be read out by the light output from the cavity. Here, the optical and mechanical frequency shift for the polystyrene particles with different radiuses ranging from 100 nm to 500 nm are calculated and shown in Figure 5a and 5b, respectively. The inset of Figure 5a shows the configuration for the detected nanobeam cavity with the particle located above the cavity.

When the particles with different radiuses, ranging from 100 nm to 500 nm, appear near the cavity, the surrounding refractive indices of the designed cavity change with a small variation, accordingly. Therefore, the optical resonant frequency of the cavity will shift dependent of the particles with different sizes. The changing percent of the optical resonant frequency of the cavity, Δ*f*/*f* o, was also calculated and presented in Figure 5a. Here, the polystyrene particle is located at Δ*x* = 190 nm, Δ*z* = 5 nm. It can be seen that there is only a relative variation of ±0.06% approximatively for the optical frequency when the particle enlarged with radius from 100 nm to 500 nm.

**Figure 5.** (**a**) The optical frequency shift for the particle with different radius, ranging from 100 nm to 500 nm. The relative changing percent of the optical resonant frequency of the cavity, Δ*f*/*f* o, was also calculated. Here, the polystyrene particle is located at Δx = 190 nm, Δz = 5 nm. The inset shows the configuration for the detected nanobeam cavity with the particle above the cavity. (**b**) The mechanical frequency shift and the relative changing percent of the mechanical frequency of the cavity, Δ*f*/*f* <sup>m</sup> for the particle with a different radius. The inset shows the strain for the vibrated particle with a radius of 200 nm.

On the other hand, the mechanical frequency of the particle decreases from 4.6 GHz to 0.9 GHz with the increasing size of the particles from 100 nm to 500 nm in radius. Here, the mechanical mode of the particle is the lowest order mechanical mode with the symmetric strain in the x, y, and z axes, which can be detected by the optomechanical coupling between the cavity and the particle due to the self-induced back-action optical trapping. The corresponding changing percent of the mechanical frequency of the cavity, Δ*f*/*f* m, achieves approximately 400%. This is because the particle size significantly affects its mechanical frequency. Here, the strain for the vibrated particle with a radius of 200 nm is also given in the inset of Figure 5b. Therefore, compared with the optical frequency shift of ±0.06%, the mechanical frequency shift shows a better performance for particle sensing, which can be used to distinguish or analyze different particles with great potential.

#### **3. Discussion**

In this work, we calculated optical force and potential on an introduced polystyrene particle with radius of 150 nm generated by a larger-center-hole nanobeam photonic crystal cavity. The introduced polystyrene particle located over the cavity was simulated in an aqueous environment. For particles captured at a temperature of 300 K, the input optical power is required to be as low as 48.8 μW in the *x* direction and 45.1 μW in the *z* direction within the designed cavity. We also studied the sensing performance of the polystyrene particle dependent on different size by the resonant frequency shift of the optical cavity mode and mechanical mode. These results show a great potential of the designed nanobeam cavity for future lab-on-chip trapping and sensing applications.

**Author Contributions:** Conceptualization, L.R.; Data curation, L.R.; Formal analysis, L.R. and N.L.; Funding acquisition, Y.L.; Investigation, L.R., Y.L., N.L. and C.C.; Methodology, L.R. and Y.L.; Project administration, N.L.; Resources, C.C.; Software, L.R., Y.L. and N.L.; Supervision, N.L.; Validation, L.R., Y.L., N.L. and C.C.; Visualization, L.R.; Writing—original draft, L.R.; Writing—review & editing, L.R., Y.L., N.L. and C.C.

**Funding:** This research was funded by the National Natural Science Foundation of China (No. 61571462).

**Acknowledgments:** The authors would like to thank Chengzhi Yang for his invaluable discussion.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Polarization Converter Based on Square Lattice Photonic Crystal Fiber with Double-Hole Units**

#### **Zejun Zhang 1,\*,†, Yasuhide Tsuji 2, Masashi Eguchi <sup>3</sup> and Chun-ping Chen <sup>1</sup>**


Received: 17 December 2018; Accepted: 21 January 2019; Published: 22 January 2019

**Abstract:** In this study, a novel polarization converter (PC) based on square lattice photonic crystal fiber (PCF) is proposed and analyzed. For each square unit in the cladding, two identical circular air holes are arranged symmetrically along the *y* = *x* axis. With the simple configuration structure, numerical simulations using the FDTD analysis demonstrate that the PC has a strong polarization conversion efficiency (PCE) of 99.4% with a device length of 53 μm, and the extinction ratio is −21.8 dB. Considering the current PCF fabrication technology, the structural tolerances of circular hole size and hole position have been discussed in detail. Moreover, it is expected that over the 1.2∼1.7 μm wavelength range, the PCE can be designed to be better than 99% and the corresponding extinction ratio is better than −20 dB.

**Keywords:** polarization converter; photonic crystal fiber; square lattice

#### **1. Introduction**

With the rapid development of Internet, high frequency communication systems have been widely studied to realize the next-generation advanced communication system. Among several means of communication, the optical communication system attracts a lot of attention since it can achieve a high-speed and large-capacity data transmission. In recent years, studies on special optical fibers and high-performance optical devices promote the performance of optical communication system. Photonic crystal fiber (PCF) [1,2] is a newly emerging optical fiber with a periodic arrangement of microscopic air holes running along the fiber axis in the cladding region. For this kind of fiber, the refractive index of cladding can be easily controlled by adjusting the geometry and distribution of the holes. In comparison to conventional optical fiber, PCF shows basic properties like birefringence, nonlinearity and single-polarization transmission that can be tailored to achieve extraordinary outputs [3–5].

Polarization converters (PCs) [6–20] plays an important role in the modern communication systems and photonic integrated circuits, such as polarization division multiplexing systems [21], polarization diversity systems [22], and polarization switches [23]. Until now, several approches have been used to design PCs, which can be classified into two types. The first type is using the mode interference. The waveguide has a high birefringent core by using asymmetrical cross section, liquid crystal or with hybrid plasmonic. Two orthogonal guided modes are excited and beat together rhythmically with the light propagation. Polarization conversion can be achieved when the two guided modes accumulate a *π*-phase shift after undergoing each beat length. The second type of approach is on the basis of mode coupling. In this approach, at least two waveguides are used. Utilzing the

birefringence of the input waveguide, the specified polarization mode is coupled and converted into the adjacent output waveguide by satisfying the phase matching condition. In contrast, another polarization mode in the input waveguide will keep propagating along because of the significant effective index difference between the two waveguides, resulting in very weak coupling. According to these two design mechanisms, plenty of PCs have been proposed based on waveguides or fibers. So far, several kinds of PCF based fiber type PCs have been designed by using a liquid crystal core [11–14], asymmetric core [15,16], plasmonic core [18] and elliptical hole core [19], etc. M. Hameed et al. proposed two kinds of PCs using a liquid crystal core [12–14] and an asymmetric core [15,16], respectively. Although these PCs can achieve a high polarization conversion efficiency (PCE) and a low extinction ratio (ER), the temperature dependence of liquid crystals and the non-Gaussian field distribution caused by the asymmetric structures limit the application of these PCs. L. Chen et al. designed a PC based on hybrid plasmonic PCF in 2014. The PC offers a 93% PCE with a device length of 163 μm. However, the PC suffers from a quite high insertion loss due to the use of metal copper. After then, in 2016, Z. Zhang et al. proposed a cross-talk free PC with a symmetric distribution using several tilted elliptical air holes in the core region [19]. Almost 100% PCE is achieved with a compact device length of only 31.7 μm. Whereas, it suffering a low structural tolerance due to the difficulty of producing elliptical holes in the PCF. Consequently, the use of different air hole shapes or multiple materials increases the manufacturing difficulties and leads to a low structural tolerance. Therefore, in designing a fiber type PC, in addition to achieve a high PCE and a low ER, a Gaussian-like field distribution and a large structural tolerance are also essential.

In this study, a novel PC element based on a square lattice PCF with a simple configuration structure is reported and analyzed. Since only one type of circular air hole is adopted to consist the PC, it can be easily fabricated with the method of stack and draw [24]. The setting of geometric parameters and the light propagation behavior are illustrated in the next section. After that, considering the current PCF fabrication technology, the structural tolerances of the circular air hole size and air hole position have been discussed in detail. Moreover, the wavelength dependence of the proposed PC has also given in Section 3. In this study, the full-vectorial finite-element method (FV-FEM) and finite-difference time-domain (FDTD) method have been used to estimate the modal effective index and the light propagation, respectively.

#### **2. Square Lattice PC with Double-Hole Unit**

#### *2.1. Design of Square Lattice PC*

In this section, a PC with double-hole unit cells in the cladding is illustrated. Figure 1 shows the cross section of our proposed PC based on a square lattice PCF. The light blue region and white circular area represent the background material SiO2 and air holes, respectively. The square lattice pitch refers to Λ. For each unit cell, which is represented with red wire frame, two air holes (with the diameter *dc*) are arranged symmetrically along the diagonal. The distance between the center of two air holes is represented by Δ*D*. The angle *θ* is 45 degrees. In the periodic lattice component, one defect is introduced to confine the light as the core region. Due to the symmetrical distribution along *y* = *x*, the polarization conversion can be achieved using two excited hybrid modes that are polarized in the ±45◦ directions with respect to the *x* axis.

According to the phase matching condition, the conversion length *Lπ* is defined as

$$L\_{\pi} = \frac{0.5\lambda}{n\_{\rm eff,1} - n\_{\rm eff,2}},\tag{1}$$

where *λ* is the operation wavelength and *n*eff,1 and *n*eff,2 are the effective indices of fundamental and 1st-higher-order modes, respectively. As an incident light travels to *L* along the *z* direction, the PCE of a PC, which is an important performance indicator, is defined as follows [10]:

$$\text{PCE} = \sin^2(2\varphi)\sin^2\left(\frac{\pi L}{2L\_\pi}\right) \times 100\% \,\tag{2}$$

where the rotation angle *ϕ* is defined by the fundamental mode field:

$$\tan \varphi = R = \frac{\iint n^2(\mathbf{x}, y) H\_x^2(\mathbf{x}, y) dx dy}{\iint n^2(\mathbf{x}, y) H\_y^2(\mathbf{x}, y) dx dy}. \tag{3}$$

Here, *n*(*x*, *y*) is the refractive-index distribution, and *Hx*(*x*, *y*) and *Hy*(*x*, *y*) are the non-domain and domain magnetic fields of the fundamental mode, respectively. The modal hybrid, which is represented by *R* in Equation (3), is a quantity between 0 and 1. Therefore, *R* = 1 results in a PCE of 100% at *L* = *Lπ*.

**Figure 1.** Cross-section view of square lattice PC with double-hole unit in the cladding.

The variation of conversion length with different geometric parameters is investigated at a wavelength of 1.55 μm, as illustrated in Figure 2. Here, the refractive indices of SiO2 and air are 1.45 and 1, respectively, the lattice pitch is set to Λ = 1 μm. Simulation results using FV-FEM demonstrate that under these conditions, only two modes (the fundamental mode and the 1st-higher-order mode) are excited to dominate the polarization behavior. The parameter Δ*S* represents the horizontal distance between the center of air hole and the lattice for each unit cell. It is revealed that the PC has a shorter conversion length with a larger cladding hole size. Additionally, the variation of parameter Δ*S* has a small effect on the PC conversion length with a large hole size. In particularly, for *dc*/Λ = 0.6, the conversion length is almost unaffected by the Δ*S*. In this case, considering that adjacent air holes on the same diagonal direction cannot overlap with each other, therefore, the parameter Δ*S* between 0.22 to 0.28Λ has been investigated to make the PC manufacturable. In this paper, the Δ*S* is fixed to 0.25Λ, i.e., the hole gap <sup>Δ</sup>*<sup>D</sup>* <sup>=</sup> <sup>√</sup>2/2Λ. Moreover, the modal hybrids of fundamental and 1st-higher-order modes with different hole sizes have also been discussed. From Figure 3, as increasing the *dc*/Λ from 0.4 to 0.6, the conversion length decreases from 384 μm to 53 μm, while the fundamental and 1st-higher-order modes have almost the same modal hybrid which are increasing from 0.914 to almost 1.000. It is evident from this figure that the values of *R* are better than 0.999 for *dc*/Λ > 0.48.

Figure 4a–d show the magnetic field distributions of *Hx* and *Hy* components for fundamental and 1st-higher-order modes with *dc*/Λ = 0.6 at a wavelength of 1.55 μm. It is apparent that each mode has almost the same magnetic field distributions, and the numerical results reveal that the modal hybrid is 0.999995 for the fundamental mode and 0.999996 for the 1st-higher-order mode. The light propagation behavior through the PC with *L<sup>π</sup>* = 53 μm is shown in Figure 5. The 3D FDTD method with a grid size of Δ*x* = Δ*y* = 0.015 μm and Δ*z* = 0.05 μm has been adopted in this investigation. The obtained normalized power against the propagation length is given in Figure 6. It is obvious that with a TM mode launched into the PC, it can be completely converted into the TE mode at *z* = 53 μm. In addition, the calculated PCE is better than 99.4% and the ER is better than −21.1 dB.

**Figure 2.** Conversion length as a function of Δ*S* with different cladding hole diameters at *λ* = 1.55 μm.

**Figure 3.** Conversion length as a function of *dc* and the corresponding modal hybrids of fundamental and 1st-higher-order modes at *λ* = 1.55 μm.

**Figure 4.** Magnetic field distributions of (**a**) *Hx* and (**b**) *Hy* components for the fundamental mode, and (**c**) *Hx* and (**d**) *Hy* components for the 1st-higher-order mode with *dc*/<sup>Λ</sup> <sup>=</sup> 0.6 and <sup>Δ</sup>*D*/<sup>Λ</sup> <sup>=</sup> <sup>√</sup>2/2.

**Figure 5.** Propagation behavior of a TM mode incident light in the PC along *z* direction at *λ* = 1.55 μm. (**a**) *Hx* (**b**) *Hy* components.

**Figure 6.** Normalized power variation of the TM mode incident light against the propagation distance.

#### *2.2. Gaussian-Like Field Distribution of the PC*

Considering the connection between a PC element and a conventional single-mode fiber (SMF), a Gaussian-like field distribution of a PC is necessary to suppress the insertion loss. Unlike previous studies using the asymmetric core distributions, our proposed PC can easily offer a Gaussian-like electromagnetic field distribution. In this study, we measure the mode matching ratio (MMR) between the fundamental mode of the PC and a Gaussian field distribution using the following overlap integral:

$$\text{MMR} = \frac{\left| \int \phi(\mathbf{x}, y) \mathbf{g}(\mathbf{x}, y) dx dy \right|^2}{\int \left| \phi(\mathbf{x}, y) \right|^2 dx dy \ \int \left| \mathbf{g}(\mathbf{x}, y) \right|^2 dx dy},\tag{4}$$

where *φ*(*x*, *y*) is the field distribution of fundamental mode. The Gaussian field distribution is represented by the *g*(*x*, *y*) as follows,

$$\log(\mathbf{x}, y) = \exp(-(\mathbf{x}^2 + y^2)/\delta^2),\tag{5}$$

where *δ* is the standard deviation of *g*(*x*, *y*). Moreover, the spot size of a Gaussian field distribution, which is illustrated by *w*, is the diameter at which the light intensity |*g*| <sup>2</sup> drops to 1/*e* of its maximum value. The spot size is calculated by *w* = 2 <sup>√</sup>2*δ*. The maximum value of MMR between an excited fundamental mode of the PC and an appropriate Gaussian field distribution can be obtained by adjusting the value of *δ*. Table 1 illustrates the maximum MMR of each fundamental mode with an appropriate Gaussian field distribution (the corresponding spot size *w*/Λ is illustrated under each MMR). According to the calculated results, the excited modes of the proposed PC show quite a great

agreement with Gaussian field distributions. Therefore, we believe our proposed PC has a great potential to be used with a low insertion loss.

**Table 1.** MMR of Each Fundamental Mode with an Appropriate Gaussian Field Distribution (*w* represents spot size for each Gaussian distribution).


#### **3. Structural Tolerance and Wavelength Dependence**

So far, the propagation property of the PC with double-hole unit has been investigated. Compared to the previously proposed PCF based PCs [11–20], the optical device in this study has almost the same level of a high PCE and a low ER. However, the PC element with only one kind of circular air holes is the most prominent feature of this study. The simple structure distribution reduces the manufacturing difficulty and enlarges the structural tolerance simultaneously. Considering the current PCF fabrication technology, the structural tolerance should be discussed. In this study, we discuss the tolerances of hole size and hole position by investigating the variation of PCE, respectively. Additionally, the wavelength dependence of the PC has also been discussed. According to the calculation formula of the PCE in Equation (2), for a designed PC element, the modal hybrid (*R*) and the difference between the fixed device length (*L*) and the conversion length (*Lπ*) are two primery factors affecting the final PCE. Therefore, the variation of parameters *R* and *L* − *L<sup>π</sup>* with different conditions (the deviation of hole size, change of hole position or variation of operation wavelength) should be taken into account.

#### *3.1. Tolerance of Cladding Hole Size*

Firstly, the structural tolerance of hole size in the PC has been investigated. The variations of parameters *L* − *L<sup>π</sup>* and *R* against the hole diameter with different deviation levels have been discussed and illustrated in Figure 7. Here, we claim that the fixed device length for each case is the completely conversion length of the PC with designed air hole sizes, i.e., *L* = 199 μm for *dc*/Λ = 0.45, *L* = 119 μm for *dc*/Λ = 0.5, *L* = 77 μm for *dc*/Λ = 0.55, and *L* = 53 μm for *dc*/Λ = 0.6. It is revealed from Figure 7 that with the deviation of *dc* from −1.5% to 1.5%, the modal hybrids for *dc*/Λ =0.5, 0.55, and 0.6 remain almost 1.000. Therefore, the final PCEs are mainly dependent on the value of *L* − *Lπ*. On the contrary, for the *dc*/Λ = 0.45, the parameter *R* increases from 0.991 to 0.996, and the variation of difference between the fixed device length and the conversion length is also relatively large. Consequently, the PCE is effected by the modal hybrid and the difference between *L* and *Lπ* simultaneously.

**Figure 7.** The variation of difference between a fixed device length and the corresponding conversion length (*L* − *Lπ*) and the modal hybrid (*R*) against the hole diameter with different deviation levels at *λ* = 1.55 μm.

Figure 8 is the contour map of the PCE as a function of the deviation of hole diameter with different hole sizes from 0.4Λ to 0.6Λ. The contour line with a PCE of 99% is represented by a dash-dotted line. Simulation results reveal that within the deviation range of ±1.5%, the PC with a larger cladding hole size has a wider tolerance range to achieve the PCE better than 99%. In particularly, for *dc*/Λ =0.6, the corresponding tolerance range is ±1.5% (±9.0 nm), which is 7 times as larger as that of the PC in [19]. Therefore, we believe that a simple structure PC with a relative large tolerance is achieved.

**Figure 8.** The variation of the PCE as a function of the deviation of hole diameter with different hole sizes.

#### *3.2. Tolerance of Cladding Hole Position*

Then, the tolerance of circular hole position in each unit cell is investigated. The square lattice PC with a symmetric structure along *y* = *x* axis makes the angle between the modal optical axis and the horizontal axis to be 45 degrees. In this study, two circular air holes are arranged symmetrically along the diagonal for each unit cell to achieve the highest PCE. However, the positions of the circular holes will be slightly varied during the actual production process. The slight asymmetrical distribution of cladding holes results in a decrease of the modal hybrid. Here, we investigated the variation of PCE against the parameter *<sup>θ</sup>* with a constant hole gap of <sup>Δ</sup>*<sup>D</sup>* <sup>=</sup> <sup>√</sup>2/2Λ. The variations of the two major factors have been investigated with different rotation angles from 40 to 50 degrees, as illustrated in Figure 9. The fixed device length for each hole size is the same as previously. For the variation of *L* − *Lπ*, the differences increase as the angle *θ* deviates from 45 degrees. Additionally, for the variation of modal hybrid, the parameter *R* reaches the maximum value at *θ* = 45◦ for *dc*/Λ =0.5, 0.55 and 0.6, and the value gradually decreases as the angle varied. It is worth noting that the PC with a larger hole size has a stronger angle dependence. This is because the light is mainly confined in the core, and a larger cladding hole size leads to a stronger light confinement. Therefore, the light confinement region in the core has a strong dependence on the positions of first layer air holes in the cladding. A slight angle variation of the first layer holes may lead to a change in modal hybrid of excited modes. On the other hand, for a PC with *dc*/Λ =0.45, the light confinement is relatively weak. Moreover, as the result shown in Figure 3, the corresponding modal hybrid is lower than 1.00 at *θ* = 45◦. While for the simulation result in Figure 9, it is noted that the maximum value of modal hybrid exists at *θ* = 46◦. Therefore, for PC with a small hole sizes, the maximum value of the PCE will shift in the direction of *θ* larger than 45 degrees.

Considering the two major factors, the variation of PCE against the parameter *θ* has been discussed and illustrated in Figure 10. The contour line with a PCE of 99% is represented by a dash-dotted line. It is evident that the PCE of a PC with a cladding hole size around 0.5Λ is better than 99% for *θ* from 40 to 49 degrees. For the PC hole size larger than 0.5Λ, the corresponding angle range decreases, and

the maximum value of PCE remains at 45 degrees. While for the *dc*/Λ < 0.5, the maximum value of PCE gradually shifts to a large rotation angle, which is in agreement with the aforementioned analysis.

**Figure 9.** The variation of difference between a fixed device length and the corresponding conversion length (*L* − *Lπ*) and the modal hybrid (*R*) against the angle *θ* with different values at *λ* = 1.55 μm.

**Figure 10.** The variation of the PCE as a function of the deviation of angle *θ* with different hole sizes.

#### *3.3. Wavelength Dependence*

In order to confirm the wide-band transmission, the wavelength dependence of our proposed PC has also been discussed in detail. It is worth to note that more than two modes are excited for the PC with a short operation wavelength. Although other higher-order modes appear at a short wavelength (such as a wavelength range of 1.2∼1.45 μm for the PC with *dc* = 0.6Λ), the incident light of the FDTD simulation is only using the fundamental and the 1st-higher-order modes. Simulation results show that the other higher-order modes are not excited during the light propagation. Therefore, the effect of other modes on the polarization behavior is negligible. The variations of the difference between device length and conversion length, and the modal hybrid with different operation wavelengths are respectively illustrated in Figure 11. According to the simulation results, since the conversion length of the PC increases with wavelengths, the value of *L* − *L<sup>π</sup>* varies from positive to negative as the wavelength changes from 1.2 to 1.8 μm. Moreover, the modal hybrid of each case decreases as the wavelength increases. Consequently, the variation of PCE against the wavelength for different hole sizes is shown in Figure 12. It can be seen that the PCE is better than 99% for *dc*/Λ = 0.5 within a

wavelength range of 1.45 to 1.62 μm (about 170 nm). For *dc*/Λ = 0.6, the corresponding wavelength bandwidth reaches 500 nm (from 1.2 to 1.7 μm), which is covering the O-band to the U-band. This is a great advantage for the wide application in the future.

**Figure 11.** The variation of difference between a fixed device length and the corresponding conversion length (*L* − *Lπ*) and the modal hybrid against different operation wavelengths.

**Figure 12.** The wavelength dependence of the PC with different hole sizes.

#### **4. Conclusions**

In this paper, we focused on designing a novel square lattice PCF based PC element which has a simple construction and a large structural tolerance. Considering the fabrication technology of the PCF, only one type of circular air hole is adopted to consist the PC. In the periodic square lattice, one unit cell is defected to form the core, which allows the PC to achieve a Gaussian-like field distribution. FDTD simulation results reveal that a high PCE of 99.4% is achieved with a short device length of 53 μm, and the corresponding ER is lower than −21.8 dB. Large structural tolerances and low wavelength dependence have been demonstrated through our proposed PC. Therefore, we believe that this kind of PC has a great practical potential for optical communication systems in the future.

**Author Contributions:** Z.Z. conceived and designed the polarization converter; Y.T. gave the guideline for the paper; M.E. and C.-p.C. gave the suggestions and modifications for the paper.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Design of Polarization Splitter via Liquid and Ti Infiltrated Photonic Crystal Fiber**

**Qiang Xu 1,2,\*,†, Wanli Luo 1,†, Kang Li 2,\*, Nigel Copner <sup>2</sup> and Shebao Lin <sup>1</sup>**


Received: 31 January 2019; Accepted: 15 February 2019; Published: 18 February 2019

**Abstract:** We propose a new polarization splitter (PS) based on Ti and liquid infiltrated photonic crystal fiber (PCF) with high birefringence. Impacts of parameters such as shape and size of the air holes in the cladding and filling material are investigated by using a vector beam propagation method. The results indicate that the PS offers an ultra-short length of 83.9 μm, a high extinction ratio of −44.05 dB, and a coupling loss of 0.0068 dB and at 1.55 μm. Moreover, an extinction ratio higher than −10 dB is achieved a bandwidth of 32.1 nm.

**Keywords:** extinction ratio; polarization splitter; dual-core photonic crystal fiber; coupling characteristics

#### **1. Introduction**

Photonic crystal fiber (PCF) consists of a solid core and holes arranged in the cladding region non-periodically or periodically along the axis [1]. According to the mechanism of light transmission, PCF can be divided into refractive index guided PCF and photonic bandgap PCF. The refractive index guided PCF is similar to total internal reflection in the mechanism of light transmission. At present, refractive index guided PCF is the most mature and widely used optical fiber. PCF technology has made great progress in pharmaceutical drug testing, astronomy, communication, and biomedical engineering and sensing [2–8]. In recent years, the PCFs filled with materials have attracted great interests, because PCFs could provide excellent properties by filling different functional materials into the air holes [9–14]. Metal wire was filled into the air holes of PCFs for polarization splitting (PS) by Sun et al. and Fan et al. [10,11]. PCFs present high-quality channels that can be controllably filled with ultra-small volumes of analytes (femtoliter to subnanoliter), such as water [12], alcohol [13], and nematic liquid crystal [14].

The dual-core PCF is constructed by introducing two defect states in the periodic arrangement of air holes. When a polarized light beam is projected into the dual-core PCF with high birefringence, the coupling strength of two vertical polarization modes is weakened by the high birefringence [15]. Therefore, high birefringence could increase the difference in coupling length of the *x*-polarized mode and *y*-polarized mode of PCF, which is also beneficial to the miniaturization of PS. In general, high-birefringence fiber can be gained by breaking the symmetry implementing asymmetric defect structures, such as dissimilar air holes and elliptical holes along the two orthogonal axes, and asymmetric core design [16–18]. Another kind of high-birefringence fiber can also be manipulated by filling liquid into the air holes or hollow core [19]. At present, the dual-core PCF has a very mature application in polarization beam splitters.

The polarization handling devices based on PCFs filled with material, such as polarization splitters (PS) [20–22] and polarization rotators (PR) [23,24], have important applications in optical fiber sensing [25,26]. The PS could be divided into two fundamental modes (HE<sup>x</sup> <sup>11</sup> and HE<sup>y</sup> <sup>11</sup>) and propagate them in different directions. Its characteristics allow for significant applications in optical sensing, storage systems, communication systems, and integrated circuit systems [23,27]. In recent years, various PS structures based on PCFs have been reported in the literature [28–35]. These PS structures show good performance (see Table 1), such as an ultra-short length [31–33], a high extinction ratio [32], a low coupling loss [30,34], and an ultra-broad bandwidth [20,24,34], but these PS structures do not present these excellent properties at the same time. It is critical to design high-performance PS with ultra-short length, high extinction ratio, and low coupling loss [23]. In order to obtain a high-performance PS, we decided to use PCF filled with functional materials. It is well known that Ti demonstrates outstanding physical and chemical properties, such as light weight, anti-corrosion, biocompatibility, high melting point, durability, and high strength in extreme environments [36–38].



In this paper, a novel ultra-short PS with low coupling loss and high birefringence is proposed based on the idea of material filled PCFs by using a vector beam propagation method (BMP) [39–41]. The numerical results present a 0.0839 mm-long PS with a coupling loss of 0.0068 dB and a high extinction ratio of −44.05 dB at the wavelength of 1.55 μm. Moreover, an extinction ratio higher than −10 dB is achieved at a bandwidth of 32.1 nm.

#### **2. Physical Modeling**

The physical modeling of the proposed PS is shown in Figure 1. The BPM-based commercially state-of-the-art software RSoft (Synopsys Inc., Mountain View, CA, USA) can be used to design and analyze optical telecommunication devices, optical systems and networks, optical components used in semiconductor manufacturing, and nano-scale optical structures. Figure 1 shows the structure of PS, where *d*, *d*1, *d*2, and *d*<sup>3</sup> represent the diameters of various air holes, respectively; *a* and *b* are the major and minor axes of the elliptical air hole; Λ represents distance of hole and hole (period); the ellipticity is expressed as η = *b*/*a*; and the air-filling ratio is *d/*Λ. The refractive index of background material is set as 1.45. Ti is filled into the two yellow air holes, and liquid (ethanol) is filled into the six blue holes. For ethanol (C2H5OH), variation of refractive index as a function of wavelength at a temperature of 20 ◦C is given by Reference [42].

$$m^2 - 1 = \frac{0.0165\lambda^2}{\lambda^2 - 9.08} + \frac{0.8268\lambda^2}{\lambda^2 - 0.01039} \tag{1}$$

where λ represents wavelength of the propagating light. The refractive index of ethanol is set as 1.35 at 1.55 μm. Figure 2 shows the refractive index function of Ti versus wavelength [43].

**Figure 1.** The cross-section of the proposed dual-core photonic crystal fiber (PCF).

**Figure 2.** Refractive index of Ti as a function of wavelength [43].

The vector wave equation, which is the basis of BPM [39–41], can be expressed by

$$
\nabla^2 \mathbf{E} + k^2 \mathbf{E} = 0 \tag{2}
$$

$$
\nabla^2 \mathbf{H} + k^2 \mathbf{H} = 0\tag{3}
$$

where *<sup>k</sup>* <sup>≡</sup> *<sup>ω</sup>*√*με*. These two equations are known as the Helmholtz equations.

The electric field *E*(*x, y, z*) can be separated into two parts: the fast change term of exp(-*ikn*0*z*) and the envelope term of *φ*(*x, y, z*) of slow change in the axial direction; *n*<sup>0</sup> is a refractive index in the cladding. Then, *E*(*x, y, z*) is stated as

$$E(x, y, z) = \phi(x, y, z) \exp(ik u\_0 z) \tag{4}$$

Substituting Equation (4) in Equation (1) results in

$$k\nabla^2\phi - 2ikn\_0\frac{\partial\phi}{\partial z} + k^2(n^2 - n\_0^2) = 0\tag{5}$$

*Crystals* **2019**, *9*, 103

Assuming the weakly guiding condition, we can approximate *<sup>n</sup>*<sup>2</sup> − *<sup>n</sup>*<sup>2</sup> <sup>0</sup> ∼= 2*n*0(*n* − *n*0). Then Equation (5) can be rewritten as

$$\frac{\partial \phi}{\partial z} = -i \frac{1}{2kn\_0} \nabla^2 \phi + jk(n - n\_0)\phi \tag{6}$$

A similar expression can be written for **H**. We find that *n* = *n*<sup>0</sup> if the fields vary in the transverse direction to propagation. Light propagation in various kinds of waveguides can be analyzed by the above method.

There are four modes of dual-core PCF on the basis of the principle of coupling mode, namely, even-mode of *x*-polarization, odd-mode of *x*-polarization, even-mode of *y*-polarization, and odd-mode of *y*-polarization. The coupling length has been defined by Reference [44] as

$$L\_c^{x,y} = \frac{\lambda}{2\left|n\_{\text{even},\lambda}^{x,y} - n\_{\text{odd},\lambda}^{x,y}\right|}\tag{7}$$

where *nx*,*<sup>y</sup> even*, *<sup>n</sup>x*,*<sup>y</sup> odd* denote the effective indexes of even-mode of *x*-polarization, odd-mode of *x*-polarization, even-mode of *y*-polarization, and odd-mode of *y*-polarization, respectively. When the coupling length of dual-core PCF satisfies *L* = *m L<sup>x</sup> <sup>c</sup>* <sup>=</sup> *n L<sup>y</sup> <sup>c</sup>* , the *x*-polarization and *y*-polarization launched into core A or B can be divided [33]. Hence, the coupling ratio (CR) can be defined as

$$CR = \frac{L\_c^{\chi}}{L\_c^{\mathcal{Y}}} = \frac{n}{m} \tag{8}$$

Assuming that the incident power is coupled into a certain core, the output power of *x*- or *y*-polarized light in the core can be expressed by the following equation [45]:

$$P\_{out}^{x,y} = P\_{in}^{x,y} \cos^2(\frac{\pi}{2} \cdot \frac{z}{L\_c^{x,y}}) \tag{9}$$

where the transmission distance is denoted by *z*.

The extinction ratio is an important index to evaluate the performance of polarization splitter, which is expressed as

$$ER = 10\log\_{10}\left(\frac{P\_{out}^{x}}{P\_{out}^{y}}\right) \tag{10}$$

where *P<sup>x</sup> out*, *<sup>P</sup><sup>y</sup> out* represent the output energy of *x*-polarization and *y*-polarization, respectively [16,46]. The coupling loss of the PS can be described by

$$Loss = -10\log\_{10}(\frac{P\_{out}}{P\_{in}}) \tag{11}$$

where *Pin* is the fundamental mode power at the input core [30].

Birefringence can be expressed as

$$B = \begin{vmatrix} n\_x - n\_y \end{vmatrix} \tag{12}$$

where *nx* and *ny* are the effective refractive index of *x*-polarized and *y*-polarized fundamental modes [29].

#### **3. Results and Discussion**

First, the *Lc* and CR are examined for different period Λ, where *d*<sup>1</sup> = 0.8 μm, *d*/Λ = 0.7, *d*<sup>2</sup> = 0.7 μm, η = 0.8, and *d*<sup>3</sup> = 0.6 μm. The results are shown in Figure 3a, in which it is observed that the *Lc* is decreased when wavelength is increased for a constant period Λ. We also noticed that the *Lc* decreases with decreasing period Λ. Moreover, the coupling length of *y*-polarization is longer than the coupling length of *x*-polarization. As the period increases, the coupling between the cores becomes

difficult. Hence, the *Lc* increases with the increase of the period. Interestingly, from Figure 3b, we noticed that when Λ ≤ 1.1 μm, the size of the CR is higher for higher period Λ; when Λ ≥ 1.1 μm, the size of the CR is higher for lower period Λ. According to Equation (8), when the CR is 3/4, the effective separation of the two orthogonal polarized lights can be achieved, so we choose the period value of 0.9 μm.

**Figure 3.** Coupling length (**a**) and coupling length ratio (**b**) as a function of wavelength for different period Λ.

Next, we analyze the *Lc* and CR as a function of wavelength for different air-filling ratio *d*/Λ, when η = 0.8, *d*<sup>2</sup> = 0.7 μm, Λ = 0.9, *d*<sup>3</sup> = 0.6 μm, and *d*<sup>1</sup> = 0.8 μm. From Figure 4a, it is observed that *Lc* is decreased when wavelength is decreased for the same value of air-filling ratio *d*/Λ. Moreover, we can find that the *Lc* decreases as the value of air-filling ratio increases, when air-filling ratio *d*/Λ ≤ 0.6. This is owing to the restriction of the outer cladding to the light wave being enhanced as air-filling ratio increases. However, when *d*/Λ ≥ 0.6, the result is opposite to the above. Meanwhile, the coupling length of *y*-polarization is longer than the coupling length of *x*-polarization. According to Figure 4b, it is found that when air-filling ratio *d*/Λ ≤ 0.6, the size of the CR is higher for higher air-filling ratio *d*/Λ*;* when air-filling ratio *d*/Λ ≥ 0.6, the size of the CR is higher for lower air-filling ratio *d*/Λ. When we choose an air-filling ratio *d*/Λ of 0.7, the CR is approximately 3/4 at 1.55 μm.

**Figure 4.** Coupling length (**a**) and coupling length ratio (**b**) as a function of wavelength for different air-filling ratio *d*/Λ.

Additionally, from Figure 5a, *Lc* is shown as a function of *d*1, in which it is observed that the coupling length is increased if *d*<sup>1</sup> is increased. This phenomenon can be interpreted as the following: as the value of *d*<sup>1</sup> increases, the cores of PS can be compressed in the vertical direction, and fundamental modes in the horizontal direction will expand. Figure 5a also indicates that *x*-polarized coupling length is lower than *y*-polarized coupling length. As seen in Figure 5b, the size of the CR increases with increasing wavelength. According to the above discussed results, we determine that *d*<sup>1</sup> is 0.8 μm.

**Figure 5.** Coupling length (**a**) and coupling length ratio (**b**) as a function of wavelength for different *d*1.

Figure 6 shows *d*<sup>2</sup> dependence on the *Lc* (Figure 6a) and CR (Figure 6b). From Figure 6a, it is evident that the *L<sup>x</sup> <sup>c</sup>* decreases with increasing the value of *d*2. This result can be attributed to the following process: as the value of *d*<sup>2</sup> increases, the cores of PS can be compressed in the vertical direction, resulting in the increase of coupling length of *y*-direction. The *L<sup>y</sup> <sup>c</sup>* is shown as a function of *d*<sup>2</sup> in Figure 6a, in which it is observed that for *d*<sup>2</sup> ≤ 0.5 μm, the value of CR is higher for higher *d*2; for *d*<sup>2</sup> ≥ 0.5 μm, the value of the CR is higher for lower *d*2. This phenomenon is probably related to the ratio of compression. According to Figure 6b, we can clearly see that the size of the CR increases with a decrease of *d*2. When the CR is 3/4, the effective separation of the two orthogonal polarized lights can be achieved, so we choose the *d*<sup>2</sup> value of 0.7 μm.

**Figure 6.** Coupling length (**a**) and coupling length ratio (**b**) as a function of wavelength for different *d*2.

The *Lc* and CR with the variation of wavelength are demonstrated in Figure 7, when η = 0.65, 0.7, 0.75, and 0.8. It can be found that the *Lc* reduces with respect to wavelength. Figure 7a indicates that *L<sup>y</sup> <sup>c</sup>* is higher than *L<sup>x</sup> <sup>c</sup>* . It can also clearly be seen that the four *y*-polarized curves are extremely close to each other. This result can be explained that as the ellipticity η increases, the cores of PS can be compressed vertically, and the fundamental mode in the horizontal direction will expand, which makes the coupling of two cores easier. According to Figure 7b, the size of the CR increases with decreasing ellipticity η. According to the above discussed results, we determine that η is 0.8.

**Figure 7.** Coupling length (**a**) and coupling length ratio (**b**) as a function of wavelength for different η.

Finally, we found that the *Lc* and CR can both be impacted by *d*1, *d*2, *d*3, *d/*Λ, Λ, and η. Hence, there exist optimized structural parameters, namely, *d*1, *d*2, *d*3, Λ, *d/*Λ, and η is 0.8 μm, 0.7 μm, 0.6 μm, 0.9 μm, 0.7 and 0.8, respectively. Although the PS has many parameters, it can be easily influenced by the bulk polymerization process of polymers [47]. The optimized coupling length of *x*- and *y*-polarized direction are *Lx* = 20.91 μm and *L<sup>y</sup>* = 27.96 μm at 1.55 μm, respectively. Figure 8 shows coupling characteristics of the PS. We observed that the separation of *x*- and *y*-polarized mode is achieved at the distance of 83.9 μm at 1.55 μm. Figure 9 shows the relationship between the birefringence and filling material for the optimized structural parameters. It is observed that birefringence of PS filled with liquid and Ti is higher than birefringence of PS filled with Ti. Meanwhile, the birefringence of PS filled with liquid and Ti can attain the order 10−<sup>2</sup> at the wavelength of 1.55 μm, and the value of birefringence is about two orders of magnitude higher than that in References [29,48].

**Figure 8.** Normalized power as a function of propagation distance at optimized structural parameters.

Figure 10 shows the variation of coupling length with filling material for the optimized structural parameters. It can be seen that the coupling length of the PS with filled Ti is higher the coupling length of the PS with filled liquid and Ti. For the PS with filled Ti, coupling length of *x*- and *y*-polarized direction are *Lx* = 29.25 μm and *Ly* = 48.28 μm at 1.55 μm, respectively. According to Equation (8), the length of the PS with filled Ti is about 234 μm, which is much longer than the PS with filled liquid and Ti. Therefore, the PS filled with liquid and Ti has a shorter length and higher birefringence than the PS filled with Ti.

**Figure 9.** Birefringence as a function of wavelength for different filling materials.

**Figure 10.** Coupling length as a function of wavelength for different filling materials.

Figure 11 shows the extinction ratio of PS with respect to wavelength at optimized structural parameters. The extinction ratio is measured in dB according to Equation (10). The PS has an extinction ratio of −44.05 dB at 1.55 μm, an extinction ratio better than −10 dB, and a bandwidth of 32.1 nm from 1567 nm to 1535 nm. Figure 12 shows the coupling loss of PS as function of wavelength. We observe that the PS has a coupling loss of 0.0068 dB at 1.55 μm. Performances in factors such as the length, coupling loss, and bandwidth are better than or at the same order of magnitude as those of the early works mentioned above (see Table 1).

**Figure 11.** ER of PS as a function of wavelength at optimized structural parameters.

**Figure 12.** Coupling loss of PS as a function of wavelength at optimized structural parameters.

Figure 13 shows the mode field distribution of odd- and even-mode in *x*- and *y*-polarization direction. When a PS is incident upon core A or core B, both the odd- and even-mode of that polarization can be generated [49].

**Figure 13.** (**a**) Even-mode of *x*-direction, (**b**) odd-mode of *x*-direction, (**c**) even-mode of *y*-direction, and (**d**) odd-mode of *y*-direction for PS.

#### **4. Conclusions**

In conclusion, a novel ultra-short PS based on Ti and liquid infiltrated PCF with high birefringence have been demonstrated by using a vector beam propagation method. The designed PS shows an ultra-short length of 83.9 μm, a coupling loss of 0.0068 dB, a high extinction ratio of −44.05 dB, and a bandwidth of 32.1 nm at a wavelength of 1.55 μm. In addition, the birefringence of PS can attain the order 10−<sup>2</sup> at the wavelength of 1.55 μm. The ultra-short PS with highly birefringent and low coupling loss properties is suitable for optical sensing, communication systems, storage systems, and integrated circuit systems.

**Author Contributions:** Analysis and writing, Q.X.; data curation, W.L.; project administration, S.L.; methodology, K.L. and N.C.

**Funding:** This work is supported by the National Natural Science Foundation of China (Grant No. 11647008) and the Scientific Research Program funded by Shaanxi Provincial Education Department (Grant No. 18JK0042).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
