3.1. Transmission Spectra and Field
We analyze the the transmission spectra under different fractal factor difference
while keeping
constant during simulation. The results are shown in
Figure 2. Firstly, for
(symmetric structure), there is only one Fano resonance peak at 17.21
m. As
increases, the in-plane symmetry is broken, the BIC energy can be radiated outside, leading to transition between BIC mode and the continuum free space, which produce additional Fano resonance. Thus, two extra resonances appear in the transmission spectra. Among the three resonances, we refer the one with the longest wavelength as mode3, the one with the shortest wavelength as mode1, and the rest as mode2. It can be observed that mode3 has ultra-narrow resonance, indicating ultrahigh
Q factor. Mode1 and mode2 have wider resonance linewidths, suggesting that the symmetry breaking increases the radiative channels at the resonance, resulting in more energy leakage and corresponding decrease in the
Q factor. Additionally, as
increases, the asymmetry of metasurface increases, leading to increasing linewidths of the three resonances, accompanied by a blue shift in the resonance wavelength. Besides, the distance between mode1 and mode2 also becomes significantly larger. For mode3, the asymmetry caused by the difference of fractal scale factors at diagonal positions compensates for the spatial symmetry mismatch between the BIC mode and the incident excitation. A radiative channel is formed between the metasurface and free space, allowing the BIC energy to radiate outside, thus transforming the BIC mode into a quasi-BIC mode with high
Q factor and finite linewidth. The smaller the symmetry breaking is, the narrower the radiative channel is, resulting in higher
Q factor. However, when
becomes too small, the linewidth of resonance is extremely narrow, but its modulation depth is also low, resulting in difficulty in sensing applications under the case that there is external interference. On the other hand, when
is large, the linewidth of resonance increases rapidly resulting in lower
Q factor. To make a trade-off, we choose
as the operating parameter for this sensor.
According to the simulation results, all three resonances are of Fano resonance profile. To calculate
Q factor of the fano resonances, we fit transmission spectra with Fano model by using the classical Fano formulation (
1) [
48,
49].
where
represents the resonant angular frequency,
is the resonance linewidth,
refers to transmission offset,
means the continuum-discrete coupling constant, and
q is the Fano parameter determining the asymmetry of the resonance profile.
Figure 3 shows the fitted results under
. It can be observed that the fitted results match the simulated curve closely. The fitted parameters of mode 3 are
,
,
, and
. Then the resonant
Q factor of mode3 is calculated as 40793 by
. Similarly, fitted results of mode1 and 2 show the linewidths of 35.48 and 58.50,
Q factors of 2054.2 and 3112.9 respectively. Therefore, it can be concluded that the resonance linewidth of mode3 is significantly smaller than the others, resulting in the highest
Q factor. Thus, mode3 is chosen as the main operating mode for the sensor proposed.
We further investigate the relationship between the
Q factor of mode3 and the asymmetric parameter
, which is defined here as the ratio of area difference (
) between original second-order fractal square and its symmetrically broken square to area (
S) of original second-order fractal square.
where
and
are shown in
Figure 1.
Based on previous studies [
27], the
Q factor of the symmetry-protected quasi-BIC exhibits an inverse quadratic law with the asymmetric parameter of metasurface (
).
In similar way, we calculate
Q factors for a serious different asymmetric parameters
=0.0416, 0.0654, 0.091, 0.12, 0.151, 0.185, 0.223, 0.265, 0.312 and 0.422. The results is shown in
Figure 4. As
approaches zero, the resonant
Q factor tends to infinity. The relationship between
Q factor of mode3 and the asymmetric parameter follows the inverse quadratic law. It indicates that the resonance of mode3 is excited by the symmetry-protected quasi-BIC resulting from the symmetry breaking.
To further investigate the physical mechanism of all resonances under symmetric and symmetric broken cases, the electric and magnetic fields of the metasurface at the resonance wavelength are calculated and shown in
Figure 5(A) to
Figure 5(L).
Figure 5(A),
Figure 5(E) and
Figure 5(I) are the electric and magnetic fields under symmetric case. As shown in the figures, the electric field vector forms annular distribution surrounding the axis of the dielectric nanodisk along y, which generates a magnetic moment in y direction. MD mode is excited. Under asymmetric cases, the fields at the three resonances are shown in the
Figure 5(B),
Figure 5(F),
Figure 5(J),
Figure 5(C),
Figure 5(G),
Figure 5(K) and
Figure 5(D),
Figure 5(H),
Figure 5(L). The electric field vectors of mode1 (
Figure 5(F)) and mode2 (
Figure 5(G)) still exhibit annular distribution, indicating MD mode resonances. However, unlike that in symmetric case under which the axis of annular electric field is along y. The axis of annular electric field for mode1 (
Figure 5(B)) is rotated 45 degrees clockwise from y axis (
Figure 5(A)). Similarly, axis of annular electric field of mode2 is approximately orthogonal to mode1 (
Figure 5(C)), rotating 45 degrees counterclockwise from y axis. Besides, as is shown in
Figure 5(A),
Figure 5(B) and
Figure 5(C), the electric field intensity for three resonances above is confined in the left and right gaps of the unit cell in the horizontal direction. While
Figure 5(E),
Figure 5(F) and
Figure 5(G) show the electric field intensity concentrates at the upper and lower ends of the nanodisks in z direction. Such concentrated field distribution provides higher sensitivity for the sensor proposed. According to their annular distribution feature of electric field vector, all the three resonances above are MD mode resonance. Therefore, when symmetry is broken, the original single resonance under symmetric case is transformed into two resonances as shown in
Figure 2.
However,
Figure 5(D),
Figure 5(H) and
Figure 5(L) shows the electric and magnetic field of mode3. In this case, the electric field vectors form two adjacent loops which rotates in opposite directions in y-z plane (
Figure 5(H)), which develop two magnetic moments along x axis, but in opposite direction. Consequently, they provide an annular magnetic field in x-z plane (
Figure 5(L)). Such magnetic field can be effectively represented as magnetic dipoles arranged end-to-end along the circumference of the loop, resulting in a TD moment in y direction. Therefore, mode3 resonance is different from MD resonances discussed above since it is excited by TD. Besides, the magnetic field enhancement is much lager than that of other modes. Thus, it can suppress radiation loss effectively, resulting in a large
Q factor. Besides,
Figure 5(D) shows the electric field intensity of mode3 is confined in the up and down gaps of the unit cell in the horizontal direction. While electric field intensity concentrates at the middle of the nanodisks in z direction as shown in
Figure 5(H). Such concentrated field distribution provides higher sensitivity for the sensor proposed.
3.2. Multipole Decomposition
In order to further understand the properties of each resonance, we use multipole decomposition in the Cartesian coordinate system for quantitative analysis. Firstly, we extract the displacement current density based on the electric field
. Here,
represents the relative permittivity at the position vector r. By integrating the current density within the dielectric, we can calculate the multipole moments. Then, we calculate the far-field scattering power spectra of each multipole moment. By summing up scattering powers of all multipole moments, we obtain the total scattering power spectra. For this metasurface, higher order multipole moments can be ignored, so we calculate eight types of multipole moments as follows:
,
,
,
,
,
,
and
, which represent electric dipole, magnetic dipole, toroidal dipole, electric quadrupole, magnetic quadrupole, toroidal quadrupole, electric octupole, and magnetic octupole. The formulas for calculating each multipole moment are as follows[
50,
51]
Here, r represents the position vector,
c is the speed of light,
refers to the angular frequency, and
represent the x, y, z directions respectively. For the quadrupole and octupole moments, a simplified notation is used. For example,
indicates that the second term (right side of the arrow) is obtained by interchanging the subscripts
and
in the first term (left side of the arrow) while keeping
unchanged.The far-field scattered power of the multipole moment can be calculated as[
51]
The results of the multipole decomposition are shown in
Figure 6. It can be observed that the dominant multipole mode contributing to mode3 is the TD, with a subsequent contribution from the magnetic quadrupole. On the other hand, the dominant multipole mode contributing to mode1, mode2 and symmetric mode is MD, with a subsequent contribution from the electric quadrupole. These results are consistent with the electric and magnetic field distributions presented in
Figure 5.
3.3. Polarization and Dimension Influence
To analyze the performance of the designed metasurface under different liner polarization, the transmitted spectra for x and y polarizations are shown in
Figure 7. The dashed line is the case of x-polarized wave incidence, while the solid line represents the case of y-polarized wave incidence. From the results, it can be observed that the two MD resonances and one TD resonance can be excited under both polarization. Besides, there is almost no difference in their modulation depth and resonance wavelengths. It indicates that the obtained resonances are insensitive to x and y polarization.
To analyze the influence of geometric parameters on transmission spectra, we calculate the transmission spectra for various thickness of substrate
m=0.4
m, 0.6
m, 0.8
m, various height of nanodisk
h=4
m, 4.25
m, 4.5
m and various period
P=9.75
m, 10
m, 10.25
m. The corresponding results are shown in
Figure 8(A),
Figure 8(B) and
Figure 8(C) respectively. We set fractal factor difference
=0.04, and other geometric parameters the same as the parameters used in
Figure 1 except the variable geometric parameter.
It can be seen that as the height of dielectric nanodisk h and the thickness of substrate m increase, all three resonances have a redshift, while the resonance linewidth remains relatively unchanged. Among them, mode 3 exhibits a larger variation in resonance wavelength when changing m, while modes 1 and 2 show a larger variation when changing h. Generally, h has a more significant effect on the resonance wavelength. When period P increases, all three resonances have a blueshift. Furthermore, the resonance linewidth of mode 3 varies less than mode1 and mode2 versus P. Linewidth is the least at period P=10m. In conclusion, compared with linewidth, resonant wavelength of these three resonances is more sensitive to geometric parameters. Besides, the linewidth of mode 3 changes less than mode1 and mode2. The modulation depth of these three resonances keeps stable. Since only wavelength varies significantly with geometric parameters for resonance of mode 3, it provides possibility for adjusting resonance wavelength by manipulating geometric parameters.
3.4. Refractive Index Sensing
The all-dielectric metasurface sensor proposed based on second-order fractal can excite high Q Fano resonances. Thus, it is well suited to detecting minimal wavelength shift which can be used to measure refractive index of liquid.
Firstly, we studied the relationship between the sensor’s wavelength shift caused by microliquid sample and its height. We calculate wavelength shift by assuming that the gaps among silicon nanodisk are filled with water (refractive index
). The variation of wavelength shift versus the height of water sample is achieved in
Figure 9. As the height of water sample increases from 0
m (unloaded) to 4.25
m, redshift of mode3 occurs. Wavelength shift increases slowly when the height of sample is less than 1
m and increases quickly between sample height of 1
m and 3.5
m. Then, tends to a constant when the sample height is larger than 3.5
m, which is close to the height of silicon nanodisk. Such characteristic can be attributed to the distribution of electric field intensity of mode 3 in z direction (
Figure 5(H)). Wavelength shift increases quickest at middle range of nanodisk height where electric field has greatest concentration. Similarly, when the height of sample is near to the top of nanodisk (4.25
m), the wavelength shift becomes much smaller. Thus, the height of sample should be close to top range of nanodisk so that influence of sample height on refractive index measurement can be ignored.
To investigate sensing performance of the sensor proposed, we choose five different liquid samples with refractive indices ranging from 1.333 to 1.489 and develop simulation model to retrieve the refractive index. According to discussion above, the height of liquid samples are set to the same as height of nanodisk of 4.25
m to avoid influence of sample height on refractive index results. The liquid samples chosen have refractive indices
n as follows: water sample with
, ethanol sample (
) with
, pentanol sample (
) with
, carbon tetrachloride sample (
) with
, and benzene sample (
) with
. We compare sensitivity for the three resonance modes. Transmission spectra loaded with different liquid samples are shown in
Figure 10(A) (for mode3) and
Figure 10(B) (for mode1 and mode2). The results indicate that the larger the refractive index of the liquid sample is, the larger the resonance wavelength shift is for all the three resonances. Here, the refractive index sensitivity (
) is defined as the resonance wavelength shift caused by a unit change in refractive index, i.e
. Besides, to quantitatively describe the sensing performance of the sensor, we define FOM=
·
/2
c=
, which concerns sensitivity, Q factor and resonance wavelength. where
is the resonant wavelength, linewidth
is obtained by fitting the transmission spectra loaded with liquid sample using equation (
1).For each resonance mode, average linewidths at five refractive indices are used to calculate FOM.
Figure 10(C) shows linear fitting results on resonance wavelength shift versus corresponding refractive index. Based on that, refractive index sensitivity and FOM results for the three modes are listed in
Table 1. It shows that mode3 has the highest FOM among the three modes.
Ultimately, We compare the performance of presented metasurface sensor with some other sensors in
Table 2. It shows that our sensor has a great sensing performance.