1. Introduction

Polarization being one of the essential properties of light, its manipulation is important for numerous applications such as those in optical communication, industrial inspection, and imaging [1]. In developments of polarization manipulation with compact optical devices, synthetic structured materials, such as metamaterials [2,3], photonic crystals (PhCs) [4], metasurfaces [5,6], and PhC slabs [7,8], have drawn great attention due to the freedom of geometry design in controlling the light-matter interaction. Importantly, synthetic devices can achieve maneuverable and pinpoint polarization state manipulation, and can replace or even outperform conventional optical approaches such as bulky wave plates. Thereby, they are highly integrated candidates to perform polarization manipulation for classical and quantum light. Typical wave plates or polarizers with compact size and/or much more superior functionalities have been theoretically proposed and experimentally demonstrated in various frequency ranges [9–12]. Recently, the physical concept of bound states in the continuum (BIC) has been introduced to manipulate the momentum dependent polarization states [13–19], and to generate the so-called vectorial laser beams [20,21]. The polarization state manipulation driven by BICs has some advantages, one of which is the high quality (Q) factor of the guided resonance in close to BICs that can provide a high sensitive polarization manipulation. The other advantage is that the topological nature of BICs provides a momentum dependent polarization state manipulation [22]. Although the polarization state can be momentum dependent, the output polarization state is determined at a fixed momentum position once the device is designed. In addition, when employing a resonance in a mirror-symmetry system, not just limited to the resonance driven by BICs, the polarization conversion cannot be perfect and always has an upper limitation due to the non-zero residual energy in the original excited polarization channel. It leads to a lot of energy waste [23–26]. More importantly, non-zero residual light in the excited polarization channel leads to the shrink in the scope of polarization manipulation, and a perfect polarization conversion cannot occur. Therefore, to constrict residual light energy in the original excited polarization channel would not only enhance the polarization conversion efficiency toward a perfect polarization conversion, but also enlarge the scope of polarization manipulation.

In this work we argue that above mentioned drawbacks can be overcome with the aid of the concept of coherent perfect absorption/absorber (CPA) [23] initially proposed in non-Hermitian optics with managed gain or loss. A CPA refers to a resonant device that can fully absorb all incident light without outward scattering, which usually required a critical coupling between the Ohmic- and the scattering loss rates [27–29]. Physically, a CPA can be viewed as a time-reversal enantiomer of a laser, and is enabled due to the interference among multiple excitations [30]. Besides the null scattering, the interference mechanism in CPA also provides a possible means of dynamically modulating the system response in real time when we change the relative phase of the multiple excitations. CPAs have been theoretically proposed and experimentally demonstrated in various synthetic systems [31–33] such as PhCs [27,28] and metasurfaces [29,34,35].

Albeit the critical coupling is crucial to achieve CPA, to fulfil the requirements of other applications and purposes we can replace the Ohmic losses by the analogues ones (pseudo Ohmic-losses), for examples, other loss degrees of freedoms in the energy conversion process with respect to the original exciting channel such as fluorescence [36], polarization conversion [37,38], diffraction [39], and the excitation of polaritons [40]. Here we would show that such a kind of pseudo Ohmic-losses approach can be applied to achieve a coherent polarization conversion. We find that the critical coupling to achieve CPA coinciding with the perfect polarization conversion can be released. The only requirement is that the working frequency is the resonant one, as long as we set the excited polarization to a circular polarization state instead of a linear one. It greatly diminishes the difficulties to achieve CPA. To consolidate our theoretical discussion, a temporal coupled-mode theory (TCMT) [41] model with a single guided resonance is provided. When resorting to our theoretical conclusion from TCMT to a realistic design, we adopt a guided mode resonance driven from symmetry protected BIC in a dielectric PhC slab, which not only can possess a momentum dependent radiative coupling rate, but also provide a resonance with high Q factor. With the help of this particular resonance characters, we prove that the pseudo CPA can be indeed achieved under a circular polarization excitation without the critical coupling requirement. A ring of pseudo CPAs possessing scalar singularity is shown to be achievable in the momentum space near BIC. We then show that the scope of the output polarization state can be efficiently enlarged, and a perfect polarization conversion can be indeed achieved. The output polarizations can also be dynamically controlled by the phase delay between two incident ports with a fixed polarization. Under a coherent two-ports incidence, the output polarization can be tuned from circular to linear then to the opposite circular polarizations. It can span the entire Poincare sphere by combining the momentum dependent radiative coupling rate character driven by BIC and the phase delay in this coherent approach. We believe such a feature can deliver important benefits for a variety of applications, including improving the sensitivity of polarization measurements and modification of the degree of polarization entanglement for quantum light [42].

2. Theoretical description

Generally, the scattering response due to an inherent resonance at a fixed parallel wave vector ${{\textbf k}_\parallel } = ({{k_x},{k_y}} )$ can be analytically described using TCMT [41]. For the sake of simplicity, let us consider a PhC slab consisting of only one resonant mode with a complex amplitude q in the frequency range of interest and having mirror symmetry in the normal ($z$) direction and C₄ symmetry around the $z$ axis, schematically shown in Fig. 1(a). Being the periodicity confined in the x and y directions and with a period smaller than the wavelength in free space, the resonant mode can only couple to the zeroth-order diffraction channels of either transverse-electric (TE) or transverse-magnetic (TM) polarized plane waves. The equation for the excitation and re-radiation of this resonant mode is

 

Fig. 1. Schematic view of coherent approach (a) and the output Stokes parameters (b-e) under coherent left-handed circular polarization incidence as a function of scattering coupling rate ratio ${{0.5\arctan [\beta } / \alpha }]$ and phase delay $\Phi $ worked at resonance. These fixed parameters are $\sigma = 1$, $\chi = 0.1188$.

Download Full Size | PDF

Here $\zeta$ is a global phase factor, and ${\chi ^{\textrm{TE,TM}}}$ is used to indicate the direct transmission/reflection coefficient.

Due to the mirror symmetry in the normal ($z$) direction, the resonant mode should to be either even or odd and can be denoted by $\sigma \,\textrm{ = } \pm 1$, respectively. The radiative coupling ${\textbf K}$ between the input light and the resonant mode is generally in the form of ${\textbf K} = \left( {\begin{array}{{cccc}} {{d_{\textrm{TE}}}}&{\sigma {d_{\textrm{TE}}}}&{{d_{\textrm{TM}}}}&{\sigma {d_{\textrm{TM}}}} \end{array}} \right)$, where ${d_{\textrm{TE}({\textrm{TM}} )}}$ illustrates the radiative coupling coefficient for TE (TM) polarized lights. ${\textbf D}$ indicates the radiative coupling between the output light and the resonant mode. In the presence of energy conservation and time-reversal symmetry, there are some constraints as ${\textbf D}{{\textbf D}^\dagger } = 2\gamma$, ${\textbf C}{{\textbf D}^\dagger } + {{\textbf D}^{\rm T}} = 0$, and ${\textbf K} = {\textbf D}$[41]. The scattering matrix ${\textbf S}$ of the system can be defined by ${\textbf b} = {\textbf {Sa}}$

We should emphasize that the phase difference between ${d_{\textrm{TE}}}$ and ${d_{\textrm{TM}}}$ is critical to determine the radiative polarization state, and is constrained by the direct coupling pathway without resonance [8]. If ${\textbf C}$ is given by that one of an isotropic homogeneous background at normal incidence, we can assume ${\chi ^{\textrm{TE}}} = {\chi ^{\textrm{TM}}} = \chi$,which, in turn, tells that the phase difference between ${d_{\textrm{TE}}}$ and ${d_{\textrm{TM}}}$ can only be $N\pi$, where $N$ is a integer number. This particular phase difference of $N\pi$ between ${d_{\textrm{TE}}}$ and ${d_{\textrm{TM}}}$ indicates that the radiative state is linearly polarized [8,43]. Once the incident angle is nonzero finite smaller, the radiative state is a general elliptical polarization unless ${d_{\textrm{TE}}} = 0$ or ${d_{\textrm{TM}}} = 0$. But since the incident angle is very small (in close to the $\Gamma $ point), the phase difference between ${d_{\textrm{TE}}}$ and ${d_{\textrm{TM}}}$ is approximately close to $N\pi$, the radiative polarization state has a very small degree of circular polarization, so the radiative state can be considered being approximately linearly polarizated [8,14]. Note that to enhance the degree of circular polarization, radiative circularly polarized state in the far field can be achieved by symmetry breaking of the structure [17,43] or making the incident angle around the boundary of Brillouin zone [44]. Here, we utilize a resonant mode with a high Q factor providing by a symmetry-protected BIC at $\Gamma $ point, for this reason, the incident angle should be very small, the difference between ${\chi ^{\textrm{TE}}}$ and ${\chi ^{\textrm{TM}}}$ is correspondingly small for an isotropic homogeneous background, i.e. ${\chi ^{\textrm{TE}}} \approx {\chi ^{\textrm{TM}}}$. For simplicity, in below discussion we would assume ${\chi ^{\textrm{TE}}} = {\chi ^{\textrm{TM}}} = \chi$ and $N = 0$ to approximately indicate the output polarization state under a circular polarization incidence with a small incident angle. Under one port circular polarization excitation, the transmission matrix ${\textbf t}$ in the circular polarization base is given by

If the incident light is a left-handed circular polarization with ${\textbf a} = \frac{1}{{\sqrt 2 }}{\left( {\begin{array}{{cccc}} i&0&1&0 \end{array}} \right)^\textrm{T}}$, the transmission amplitude in the crossed polarization channel is $|{\sigma {e^{i\sigma \chi }}{{({\alpha + i\beta } )}^2}} / [2(if - i{f_0} + {\alpha^2} + {\beta^2} ) ] |$, where $\alpha$ ($\beta$) is the amplitude of ${d_{\textrm{TE}}}$ (${d_{\textrm{TM}}}$). We can find the polarization conversion cannot be perfect and the maximum amplitude of the transmission coefficient is only 50% at resonance [23–26]. So the polarization conversion has an upper limitation. Besides this limitation, the polarization of transmitted state is also fixed and cannot be modified. For example, at resonant frequency, the transmitted polarization state under the left-handed circular polarization incidence is $\left( {\begin{array}{{ccc}} {{{{S_1}} / {{S_0}}}}&{{{{S_2}} / {{S_0}}}}&{{{{S_3}} / {{S_0}}}} \end{array}} \right) = \left( {\begin{array}{{ccc}} { - \cos({2\varsigma + 2\sigma \chi } )}&{ - \sin({2\varsigma + 2\sigma \chi } )}&0 \end{array}} \right)$, where ${S_{0,1,2,3}}$ are Stokes parameters. We can find the transmitted polarization state is determined by the background response represented by $\chi$ and the radiative coupling rates described by $\varsigma = \arg ({\alpha - i\beta } )$. Once again, we should emphasize that the obtained theoretical results are based on the above mentioned assumption, which are approximate but having clear physics when the angle of incidence is nonzero while infinite small. To obtain a detailed transmitted polarization state with a higher accuracy at finite incident angle, the difference between ${\chi ^{\textrm{TE}}}$ and ${\chi ^{\textrm{TM}}}$ should be taken into account.

To overcome these limitations, we can utilize the physics of coherent control by manipulating the interference among multiple input channels. Coherent control has been widely explored after the proposal of CPA [23]. The null output in the original polarization channel can be viewed as a pseudo CPA, where the Ohmic loss degree of freedom is analogously replaced by the polarization conversion effect.

Now, under coherent two-port excitation, the output in each side of the structure can be manipulated by both the amplitude difference and the phase delay between the two ports. It does provide an additional opportunity to dynamically control the output polarization. For example, the output coefficient in the lower side of the structure ${\textbf o}$ under two ports excitation with a phase delay $\Phi $ in the circular polarization base ${\textbf a} = \frac{1}{{\sqrt 2 }}{\left( {\begin{array}{{cccc}} { \pm i}&{ \pm i{e^{i\Phi }}}&1&{{e^{i\Phi }}} \end{array}} \right)^\textrm{T}}$ can be found from

The first (second) subscript of o indicates the output(input) circular polarization, where ${\pm}$ denotes left(right)-handed circular polarization. The output in the original polarization channel is zero, i.e. $|{{o_{ +{+} }}} |= |{{o_{ -{-} }}} |={=} 0$, under the condition of

In other words, the output in the original polarization channel can be null under a coherent symmetric (anti-symmetric) circular polarization incidence at resonance, as given by Eq. (6). It is similar to traditional CPA that supports a null output, so we nominate it as a pseudo CPA. But unlike traditional CPA, where the null output is due to a full absorption of light, here the null output in the excited polarization channel is due to a perfect polarization conversion, and the polarization conversion plays a similar role of the Ohmic loss in traditional CPA. What more, unlike the traditional CPA which requires not only a coherent linear polarization incidence but also the satisfaction of critical coupling [23,24,26,45], this pseudo CPA only askes for a circular polarization incidence at the resonant frequency of $f = {f_0}$, and does not require additional critical coupling condition. These features can produce many obvious advantages. The output efficiency in the crossed polarization channel can reach 100% at the pseudo CPA condition of Eq. (6), which definitely breaks the upper limitation under one port excitation.

Besides the perfect polarization conversion, the output polarization can also be obviously tuned via the phase delay $\Phi $. For example, under a coherent left-handed incidence, ${\textbf a} = \frac{1}{{\sqrt 2 }}{\left( {\begin{array}{{cccc}} i&{i{e^{i\Phi }}}&1&{{e^{i\Phi }}} \end{array}} \right)^\textrm{T}}$, the output polarization state $\left( {\begin{array}{{ccc}} {{{{S_1}} / {{S_0}}}}&{{{{S_2}} / {{S_0}}}}&{{{{S_3}} / {{S_0}}}} \end{array}} \right)$ in the lower semi-space is

We can see the phase delay $\Phi $ between the two excitation ports provides an additional degree of freedoms to manipulate the output polarization, which does not occur in one port excitation. The dependence of normalized Stokes parameters ${{{S_{1,2,3}}} / {{S_0}}}$ on the phase delay and radiative coupling rate ratio under a left-handed circular incidence is shown in Fig. 1(b) to (d). We can see under the coherent two ports incidence, the output polarization can be changed from circular to linear ones, and even to the opposite circular polarization. By including radiative coupling rate design, especially the momentum varied radiative coupling ratio, the output polarization state can span the entire Poincare sphere, as illustrated in Fig. 1(e). Based on the TCMT conclusion, we can find that the coherent control provides a great feasibility to dynamically control the output polarization state by manipulating the interference among multiple input channels with a tunable phase delay. And a perfect polarization conversion can occur under the condition of a pseudo CPA.

3. Dielectric PhC slab design and numerical verification

To consolidate our theoretical discussion, we adopt a guided mode resonance driven from a symmetry protected BIC, which not only provides a resonance with high Q factor, but also a momentum varied radiative coupling due to its topological nature [14]. With this particular resonance driven by BIC, a ring of pseudo CPA possessing scalar singularity can be achieved in the momentum space near this BIC singularity, which extends CPA occurring only at a discrete parameter position. In detail, we consider a dielectric (silicon) PhC slab with C_4v symmetry, which possesses a class of symmetry protected BIC states at Γ point [46]. The non-radiative loss due to Ohmic loss can be fully ignored here. The unit cell of the concrete PhC slab is a square lattice with periodicity of $d = 450 \mu$ m. The thickness of this slab is h. An air hole with radius ${r_c}$ is in the center of each unit cell and passes through the slab. This PhC slab with C_4v symmetry holds symmetry protected BIC states at Γ point. Note that our theoretical conclusion based on TCMT is not limited to the C_4v-symmetric PhC slabs discussed here, it can be applied to other resonant PhC or metasurface systems with different geometries and symmetries. The additional consideration is that the symmetry constraint on radiative couplings should be included. Here, we fixed the thickness of PhC slab to $h = 200 \mu$ m, and optimize the radius of air hole to tune BIC at Γ point. For a fixed radius such as ${r_c} = 144 \mu$ m, the guided resonance band [47] is illustrated in Fig. 2. One TE-like (the 4^th band) and one TM-like (the 5^th band) BIC modes are supported in the frequency region between two degenerated radiative bands, and have opposite band curvature (dispersion slope). The obtained Q factor diverges at Γ point and the corresponding field distribution, which are shown in Figs. 2(b) to (d), clearly indicate the nature of TE- and TM-type BICs.

 

Fig. 2. (a) Numerical band structure along the two symmetric directions in designed PhC slab with C_4v geometry. (b) The Q factors for the 4^th and 5^th bands both along Γ-X and Γ-M directions. The parallel wave vector along Γ-M (Γ-X) direction ${k_{/{/}}} = ({k,k} )$ (${k_{/{/}}} = ({k,0} )$) is in unit of ${\pi / d}$. (c)-(d) The z component of magnetic (electric) field distribution for the 4^th (c) and 5^th (d) bands at Γ point. The inset schematically shows the unit cell of designed PhC slab.

Download Full Size | PDF

We also explore the transmitted polarization state around resonance under a circular polarization incidence with a small incident angle. When the incident polar angle is fixed, such as $\theta = {2^\circ }$, the numerical transmitted polarization state for the TE-like resonance around 0.303 THz under a left-handed circular polarization incidence is displayed in Fig. 3, which is in a good agreement with the analytical prediction. We can see that the transmitted polarization state is fixed once the incident angle is determined. Although our theoretical prediction under the simple assumption is approximate for a small finite incident angle, our theoretical prediction could grasp the main features of the physics mechanism emphasized in this work. The maximum deviation in ${{{\textrm{S}_3}} / {{\textrm{S}_0}}}$ is only about 2%, and is mainly due to the small fluctuation of the resonance around the azimuthal angle and the small difference between ${\chi ^{\textrm{TE}}}$ and ${\chi ^{\textrm{TM}}}$.

 

Fig. 3. The analytical (Solid line) and numerical (Scatter) transmitted Stokes parameters for symmetric resonant mode under the left-handed circular polarization incidence as a function of varied incident azimuthal angle $\varphi$ with fixed incident polar angle $\theta = {2^\circ }$. These fitting parameters are $\sigma = 1$, ${\beta / \alpha } = \tan [{2\varphi } ]$, $\chi = 0.1188$.

Download Full Size | PDF

Due to the existence of mirror symmetry along the z direction, the resonance driven by the TE-like BIC is corresponding to an even mode ($\sigma = 1$), while the resonance driven by the TM-like BIC is corresponding to an odd mode ($\sigma ={-} 1$). So we can use coherent symmetric and anti-symmetric incidence to selectively excite resonance driven by either the TE-like or the TM-like BIC. We numerically display the output intensity in the lower semi-space under either symmetric or anti-symmetric left-handed circular polarization incidence in Fig. 4 to clearly indicate the appearance of TE- and TM-like BICs at Γ point. Around 0.3027 THz, as shown in Figs. 4(a) and (b), the resonance is divergent with a null full width at half maximum (FWHM), i.e. the Q factor approaches infinity and a TE-like BIC appears at Γ point. Under the condition of the pseudo CPA given by Eq. (6) for a symmetric incident circular polarization, such as the left-handed circular polarization, a perfect cross-polarization convention to the right-handed circular polarization can be obtained, when the resonance is deviated from Γ point in the leaky region. Along the two symmetric directions, e.g. the Γ-X and Γ-M directions, the pseudo CPA with almost zero output in the original excited circular polarization channel, ${|{o_{ +{+} }^\textrm{s}} |^2} \buildrel\textstyle.\over= 0$, can continuously occur on resonance with varied incident angle except at Γ point, as shown in Fig. 4(a). Correspondingly, the perfect polarization conversion with ${|{o_{ -{+} }^\textrm{s}} |^2} \buildrel\textstyle.\over= 1$ can be found at resonance coinciding with ${|{o_{ +{+} }^\textrm{s}} |^2} \buildrel\textstyle.\over= 0$ except at Γ point. In the symmetric direction, radiative couplings take $\alpha \ne 0$ ($\beta \ne 0$) and $\beta = 0$ ($\alpha = 0$) along the Γ-X (Γ-M) direction, where the critical coupling condition is not met. But pseudo CPA coinciding with perfect polarization conversion can be found for circular polarization incidence, as shown in Figs. 4(a) and(b), which clearly show that this effect does not require a critical coupling. Similar conclusion can be also applied to TM-like resonance, as shown in Figs. 4(c) and (d), where the pseudo CPA follows the TM-like branches shown in Fig. 2(a).

 

Fig. 4. Output intensity in the lower semi-space under symmetric (a, b) and anti-symmetric (c, d) left-handed circular polarization incidence along two symmetric directions (Γ-M and Γ-X). The superscript s (as) indicates coherent symmetric(anti-symmetric) incidence with phase delay $\Phi = 0(\pi )$.

Download Full Size | PDF

At resonance, as given by Eq. (6), a zero output in residual polarization channel and a simultaneously perfect polarization conversion in the crossed polarization channel can be always found. This effect holds true even when the resonant frequency is continuous varied or sustained constant with a periodic varied parameter, and transformed into a ring-like of pseudo CPA for either symmetric or anti-symmetric circular polarization incidence. Paying attention to the PhC slab structure discussed above, in the momentum space near Γ point, the resonant frequency fluctuates very weak, and can be approximately considered as a constant when the angle of incidence is sufficiently small. Now we can find a ring of CPA for circular polarization at this fixed resonant frequency around the Γ point. As displayed in Fig. 5, around 0.303 THz, under symmetric incidence, a ring of almost null output for the initial polarization state can be found around ${{{k_\parallel }} / {{k_0}}} \approx 0.035$, where ${k_0}$ is the wave vector in vacuum, and, at the same time, a maximum ${\sim} 100\%$ output of the crossed polarization can be found. For the anti-symmetric circular polarization incidence, similar phenomenon exists around 0.3045 THz at ${{{k_\parallel }} / {{k_0}}} \approx 0.035$. Unlike CPA occurring at discrete parameter position, the ring of CPA for circular polarization can be found in momentum space without additional critical requirement of critical coupling.

 

Fig. 5. Output intensity in the lower semi-space under symmetric (a, b) and anti-symmetric (c, d) left-handed circular polarization incidence around each resonance in momentum space.

Download Full Size | PDF

Besides pseudo CPA for circular polarization, coherent approach can also provide a feasibility to manipulate the output polarization by only changing the phase delay of the two inputs. Combing the radiative coupling manipulation in momentum space using BIC and the coherent approach indicated by the phase delay, the output polarization can span the full Poincare sphere in an already designed photonic system. As shown in Fig. 6, we can find that the output polarization can indeed span the full Poincare sphere using resonant mode driven by TE-like BIC in above investigated PhC slab. Analytical prediction from Eq. (7) finds a good agreement with the numerical simulation at a selected momentum position possessing determined radiative coupling rates, as shown in Fig. 7. In contrast with previous studies on polarization manipulation based on BIC [14–17], our approach discussed here can dynamically control the output polarization state at a fixed momentum position, which expands the scope of polarization control. With the help of coherent approach, in a singlet resonance system with mirror symmetry, a perfect polarization conversion can be achieved in the proposed coherent approach, which breaks the limitation for the polarization conversion [25]. And the critical coupling requirement is not necessary in our proposed coherent approach, making our investigation more easily realizable in realistic resonant systems. Note that unlike former approaches where active materials are introduced into synthetic devices in order to dynamically manipulate the output light [48,49], the dynamical control in our proposed coherent approach is due to the interference mechanism. Consequently, without making any change on the designed/fabricated functional device, the output polarization can be flexibly and precisely manipulated on the entire Poincare sphere by combing coherent approach and momentum dependent radiative coupling coefficient provided by BICs.

 

Fig. 6. The output Stokes parameters under coherent left-handed circular polarization incidence as a function of varied incident azimuthal angle $\varphi$ with fixed incident polar angle $\theta = {2^\circ }$, and varied phase delay $\Phi $ worked at symmetric resonance.

Download Full Size | PDF

 

Fig. 7. The analytical (solid line) and numerical (scatter) output Stokes parameters for symmetric resonant mode under coherent left-handed circular polarization incidence as a function of phase delay $\Phi $ when incident angle is selected at $\theta = {2^\circ }$, $\varphi = {30^\circ }$. These fitting parameters are $\sigma = 1$, ${\beta / \alpha } = \tan [{{\pi / 3}} ]$, $\chi = 0.1188$.

Download Full Size | PDF

4. Conclusion

In conclusion, here we show that CPA coinciding with a perfect polarization conversion can be indeed achieved under circular polarization excitation without critical coupling requirement, when transferring Ohmic losses freedom to polarization conversion. In a PhC slab possessing symmetry protected BIC consolidating our theoretical prediction, we also find a ring of pseudo CPA, which usually occurs at discrete parameter position. Most importantly, the scope of output polarization state can be efficiently enlarged, especially including a perfect polarization conversion, and can span the entire Poincare sphere by combining momentum dependent radiative coupling rate driven by BIC and the phase delay in coherent approach. Note that in our theoretical exploration using a dielectric PhC slab, the dissipation loss is very small and can be ignored. When applied our theoretical results to realistic device, the influence of the dissipation loss on the polarization state should be considered. In the present work only one resonance is adopted in the theoretical discussion, multiple resonances can be explored in future works. We believe the mechanism and effects discussed here can deliver important benefits for a variety of applications in light control and management, including improving sensitivity of polarization measurements and modification of degree of polarization entanglement for quantum light.