Wavelength-Flattened Directional Coupler Based Mid-Infrared Chemical Sensor Using Bragg Wavelength in Subwavelength Grating Structure

In this paper, we report a compact wavelength-flattened directional coupler (WFDC) based chemical sensor featuring an incorporated subwavelength grating (SWG) structure for the mid-infrared (MIR). By incorporating a SWG structure into directional coupler (DC), the dispersion in DC can be engineered to allow broadband operation which is advantageous to extract spectroscopic information for MIR sensing analysis. Meanwhile, the Bragg reflection introduced by the SWG structure produces a sharp trough at the Bragg wavelength. This sharp trough is sensitive to the surrounding refractive index (RI) change caused by the existence of analytes. Therefore, high sensitivity can be achieved in a small footprint. Around fivefold enhancement in the operation bandwidth compared to conventional DC is achieved for 100% coupling efficiency in a 40 µm long WFDC experimentally. Detection of dichloromethane (CH2Cl2) in ethanol (C2H5OH) is investigated in a SWG-based WFDC sensor 136.8 µm long. Sensing performance is studied by 3D finite-difference time domain (FDTD) simulation while sensitivity is derived by computation. Both RI sensing and absorption sensing are examined. RI sensing reveals a sensitivity of −0.47% self-normalized transmitted power change per percentage of CH2Cl2 concentration while 0.12% change in the normalized total integrated output power is realized in the absorption sensing. As the first demonstration of the DC based sensor in the MIR, our device has the potential for tertiary mixture sensing by utilizing both changes in the real and imaginary part of RI. It can also be used as a broadband building block for MIR application such as spectroscopic sensing system.


Introduction
To implement industrial process control, security and surveillance, environmental analysis, and clinical/biomedical monitoring, numerous sensors with small footprints, high stability, low cost, and low power consumption are demanded. Nanophotonics sensors are promising to fulfill these  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range. The schematic of our MIR SWG-based WFDC is illustrated in Figure 1b. The coupling region is formed by inserting SWG into the conventional DC structure. The RI of the equivalent homogeneous material is determined by period Λ and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is the chosen material platform since its fabrication is mature and stable.

Concept and Design Optimization
In order to illustrate how SWG structure helps to increase the operation bandwidth of DC, we compare the dispersion of waveguide with and without SWG structure. Figure 1c shows the theoretical dispersion of the fundamental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, n SiO 2 = 1.4 and infinitely thick SiO2 cladding. The dispersion is simulated by finite difference analysis using Lumerical Mode Solution. Figure 1d presents the dispersion of the floquet mode in an SWG with Λ = 0.86 µm and duty cycle = 0.25 while the rest of the parameters are the same as slab waveguide. Floquet mode's dispersion is calculated numerically by the effective medium theory which will be explained in details in the next paragraph. Unlike the linear dispersion presented in 1 and an odd mode Nanomaterials 2018, 8, x FOR PEER REVIEW Figure 1a shows the schematic of conventional DC which consists of two slightl waveguides. Owing to evanescent wave coupling, an even mode ɸ1

Concept and Design Optimization
and an odd mode ɸ2 ex coupled structure according to the coupled mode theory (CMT). The input EM wave excite and ɸ2.
The coupling between these two modes allows the EM wave to transfer between t waveguides. The required coupling length for 100% coupling efficiency is analytically calcu L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can achieved in a limited wavelength range. The schematic of our MIR SWG-based WFDC is illustrated in Figure 1b. The coupling formed by inserting SWG into the conventional DC structure. The RI of the equivalent homo material is determined by period Λ and duty cycle = a/Λ. In our study, silicon-on-insulato the chosen material platform since its fabrication is mature and stable.
In order to illustrate how SWG structure helps to increase the operation bandwidth o compare the dispersion of waveguide with and without SWG structure. Figure 1c sh theoretical dispersion of the fundamental mode in a slab waveguide with h = 0.4 µm, n SiO 2 = 1.4 and infinitely thick SiO2 cladding. The dispersion is simulated by finite difference using Lumerical Mode Solution. Figure 1d presents the dispersion of the floquet mode in with Λ = 0.86 µm and duty cycle = 0.25 while the rest of the parameters are the sam waveguide. Floquet mode's dispersion is calculated numerically by the effective medium which will be explained in details in the next paragraph. Unlike the linear dispersion pre 2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both 3 of 13 nventional DC which consists of two slightly spaced pling, an even mode ɸ1 and an odd mode ɸ2 exist in the mode theory (CMT). The input EM wave excites both ɸ1 des allows the EM wave to transfer between these two for 100% coupling efficiency is analytically calculated by elength and neff1 and neff2 are the effective RI of modes ɸ1 WFDC is illustrated in Figure 1b. The coupling region is nal DC structure. The RI of the equivalent homogeneous ty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is ation is mature and stable. re helps to increase the operation bandwidth of DC, we th and without SWG structure. Figure 1c shows the mode in a slab waveguide with h = 0.4 µm, n si = 3.4, . The dispersion is simulated by finite difference analysis presents the dispersion of the floquet mode in an SWG hile the rest of the parameters are the same as slab alculated numerically by the effective medium theory xt paragraph. Unlike the linear dispersion presented in 1 and 3 of 13 n c of conventional DC which consists of two slightly spaced ave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the upled mode theory (CMT). The input EM wave excites both ɸ1 two modes allows the EM wave to transfer between these two length for 100% coupling efficiency is analytically calculated by he wavelength and neff1 and neff2 are the effective RI of modes ɸ1 ases, the stronger modal confinement tends to equate neff1 and in neff1 − neff2. Hence, the desired coupling efficiency can only be ge.
nventional directional coupler (DC) and (b) the SWG-based oupler (WFDC) in SOI platform. T is the transmitted power the evanescently coupled power to the adjacent waveguide. w, width, waveguide height, coupling gap and coupling length, G period, silicon width and silicon dioxide width respectively. ɸ1, ode presented in the DC. ɸ3 is a weakly coupled even mode. (c) e in a slab waveguide as shown in the inset. The slab waveguide nd infinitely thick silicon dioxide cladding. (d) Dispersion of the in the inset. The black glowing line in the insets of (c,d) show the ates in the waveguide.
-based WFDC is illustrated in Figure 1b. The coupling region is nventional DC structure. The RI of the equivalent homogeneous and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is s fabrication is mature and stable. structure helps to increase the operation bandwidth of DC, we ide with and without SWG structure. Figure 1c shows the mental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, adding. The dispersion is simulated by finite difference analysis ure 1d presents the dispersion of the floquet mode in an SWG 0.25 while the rest of the parameters are the same as slab ion is calculated numerically by the effective medium theory the next paragraph. Unlike the linear dispersion presented in 2 . The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n e f f 1 − n e f f 2 , where λ is the wavelength and n eff1 and n eff2 are the effective RI of modes 3 of 13 conventional DC which consists of two slightly spaced oupling, an even mode ɸ1 and an odd mode ɸ2 exist in the d mode theory (CMT). The input EM wave excites both ɸ1 modes allows the EM wave to transfer between these two h for 100% coupling efficiency is analytically calculated by avelength and neff1 and neff2 are the effective RI of modes ɸ1 the stronger modal confinement tends to equate neff1 and ff1 − neff2. Hence, the desired coupling efficiency can only be tional directional coupler (DC) and (b) the SWG-based r (WFDC) in SOI platform. T is the transmitted power vanescently coupled power to the adjacent waveguide. w, h, waveguide height, coupling gap and coupling length, iod, silicon width and silicon dioxide width respectively. ɸ1, presented in the DC. ɸ3 is a weakly coupled even mode. (c) slab waveguide as shown in the inset. The slab waveguide finitely thick silicon dioxide cladding. (d) Dispersion of the e inset. The black glowing line in the insets of (c,d) show the n the waveguide. d WFDC is illustrated in Figure 1b. The coupling region is tional DC structure. The RI of the equivalent homogeneous duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is rication is mature and stable. ture helps to increase the operation bandwidth of DC, we with and without SWG structure. Figure 1c shows the al mode in a slab waveguide with h = 0.4 µm, n si = 3.4, g. The dispersion is simulated by finite difference analysis d presents the dispersion of the floquet mode in an SWG while the rest of the parameters are the same as slab ion atic of conventional DC which consists of two slightly spaced wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled mode theory (CMT). The input EM wave excites both ɸ1 se two modes allows the EM wave to transfer between these two g length for 100% coupling efficiency is analytically calculated by s the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 creases, the stronger modal confinement tends to equate neff1 and p in neff1 − neff2. Hence, the desired coupling efficiency can only be range.
conventional directional coupler (DC) and (b) the SWG-based l coupler (WFDC) in SOI platform. T is the transmitted power is the evanescently coupled power to the adjacent waveguide. w, e width, waveguide height, coupling gap and coupling length, WG period, silicon width and silicon dioxide width respectively. ɸ1, d mode presented in the DC. ɸ3 is a weakly coupled even mode. G-based WFDC is illustrated in Figure 1b. The coupling region is conventional DC structure. The RI of the equivalent homogeneous Λ and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is its fabrication is mature and stable. G structure helps to increase the operation bandwidth of DC, we guide with and without SWG structure. Figure 1c shows the damental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, cladding. The dispersion is simulated by finite difference analysis igure 1d presents the dispersion of the floquet mode in an SWG = 0.25 while the rest of the parameters are the same as slab 2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate n eff1 and n eff2 , resulting in the significant drop in n eff1 − n eff2 . Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range. The schematic of our MIR SWG-based WFDC is illustrated in Figure 1b. The coupling region is formed by inserting SWG into the conventional DC structure. The RI of the equivalent homogeneous material is determined by period Λ and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is the chosen material platform since its fabrication is mature and stable.

Concept and Design Optimization
In order to illustrate how SWG structure helps to increase the operation bandwidth of DC, we compare the dispersion of waveguide with and without SWG structure. Figure 1c shows the theoretical dispersion of the fundamental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, n SiO 2 = 1.4 and infinitely thick SiO2 cladding. The dispersion is simulated by finite difference analysis using Lumerical Mode Solution. Figure 1d presents the dispersion of the floquet mode in an SWG with Λ = 0.86 µm and duty cycle = 0.25 while the rest of the parameters are the same as slab waveguide. Floquet mode's dispersion is calculated numerically by the effective medium theory which will be explained in details in the next paragraph. Unlike the linear dispersion presented in  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1

Concept and Design Optimization
and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2.
The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1

Concept and Design Optimization
and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2.
The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range. and an odd mode ɸ2 exist in the pled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 ɸ2.
The coupling between these two modes allows the EM wave to transfer between these two veguides. The required coupling length for 100% coupling efficiency is analytically calculated by = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and , resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be ieved in a limited wavelength range.
The schematic of our MIR SWG-based WFDC is illustrated in Figure 1b. The coupling region is formed by inserting SWG into the conventional DC structure. The RI of the equivalent homogeneous material is determined by period Λ and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is the chosen material platform since its fabrication is mature and stable.
In order to illustrate how SWG structure helps to increase the operation bandwidth of DC, we compare the dispersion of waveguide with and without SWG structure. Figure 1c shows the theoretical dispersion of the fundamental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, n SiO 2 = 1.4 and infinitely thick SiO 2 cladding. The dispersion is simulated by finite difference analysis using Lumerical Mode Solution. Figure  dispersion is calculated numerically by the effective medium theory which will be explained in details in the next paragraph. Unlike the linear dispersion presented in the slab mode, the effective RI of the floquet mode in SWG rises drastically as λ approaches the Bragg wavelength λ B . Such tremendous effective RI boost has a different influence on n eff1 and n eff2 . n eff1 is appreciably elevated while n eff2 is less affected by the index perturbation due to  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range. The schematic of our MIR SWG-based WFDC is illustrated in Figure 1b. The coupling region is formed by inserting SWG into the conventional DC structure. The RI of the equivalent homogeneous material is determined by period Λ and duty cycle = a/Λ. In our study, silicon-on-insulator (SOI) is the chosen material platform since its fabrication is mature and stable.

Concept and Design Optimization
In order to illustrate how SWG structure helps to increase the operation bandwidth of DC, we compare the dispersion of waveguide with and without SWG structure. Figure 1c shows the theoretical dispersion of the fundamental mode in a slab waveguide with h = 0.4 µm, n si = 3.4, n SiO 2 = 1.4 and infinitely thick SiO2 cladding. The dispersion is simulated by finite difference analysis using Lumerical Mode Solution. Figure 1d presents the dispersion of the floquet mode in an SWG with Λ = 0.86 µm and duty cycle = 0.25 while the rest of the parameters are the same as slab waveguide. Floquet mode's dispersion is calculated numerically by the effective medium theory which will be explained in details in the next paragraph. Unlike the linear dispersion presented in 2 's anti-symmetry [61]. The resultant increase in n eff1 − n eff2 compensates for its reduction as λ decreases. Consequently, n eff1 − n eff2 is preserved and WFDC can be realized. Figure 2a demonstrates our method to obtain the RI of the equivalent homogenous material of 3D SWG on the SOI platform. The study is conducted for the wavelength of 3.62 µm, assuming Λ = 0.86 µm and duty cycle = 0.25 without loss of generality. The 3D SWG structure (left) is firstly compressed into an equivalent 2D SWG by reducing the z dimension using the effective index method. Here, we use the commercial simulation tool Lumerical Mode Solution to derive the effective RI. After this step, the 3D SWG can be regarded as a 2D SWG in the xy plane (middle). The red strips possess an effective RI of 2.6 determined by the effective RI of the fundamental mode of 0.4 µm Si slab covered by SiO 2 cladding. The grey strips have an effective RI of 1.4 since it is compressed from a structure consisting solely of SiO 2 . Then, according to Amnon Yariv and Pochi Yeh [62], the effective RI n eff of the equivalent homogenous material of the compressed 2D SWG can be analytically solved by: where c = 3 × 10 8 m/s is the speed of light in vacuum, K is the Bloch wave number, ω is the angular frequency of the EM wave determined by the wavelength, Λ is SWG's period while a and b equal to [Λ × duty cycle] and [Λ × (1 − duty cycle)], respectively, n 1 and n 2 are the effective RI of the two material layers in the 2D structure (in our case n 1 = 2.6 and n 2 = 1.4), k 1x and k 2x are the wave vector along the propagation direction, and β is the projection of the wave vector along the boundary plane which equals to 0 since normal incidence is assumed in our study. After numerical calculations using Equations (1)-(5), the 2D SWG in the middle is finally simplified to an equivalent homogeneous material with RI = 2.0 (right).
Since the solution is determined by Λ = a + b and duty cycle = a/Λ in the SWG structure, we optimize these two parameters accordingly. The targeted wavelength range for flattening is 3.66-3.895 µm which is available in our laser setup. We fix the duty cycle at 0.25 first in order to optimize Λ. Initially, we aim to locate the Bragg wavelength λ B only slightly below 3.66 µm whereby n eff increases most significantly to compensate for the drop in n eff1 − n eff2 as λ decreases. Nonetheless, this scheme is risky. In the case when Λ of the fabricated device is larger than the design due to fabrication imperfection, the Bragg wavelength will red shift and become λ B > 3.66 µm. Subsequently, a small wavelength range of (λ B − 3.66 µm will undergo Bragg diffraction. To minimize this risk, we position the Bragg wavelength at 3.62 µm instead. Figure 2b presents the Λ optimization result. n eff boosts at Λ = 0.75 µm after a gradual rise from Λ = 0.1 µm to Λ = 0.75 µm. A maximum n eff = 1.984 is reached at Λ = 0.86 µm, after which, mathematically, Equation (2) is not solvable since the right hand side is larger than unity and the Bragg diffraction happens physically. Thus, 0.86 µm is chosen as the optimized Λ. Duty cycle mainly affects the excitation of ɸ3 in the SWG-based DC (see ɸ3 in Figure 1b). ɸ3 is interpreted as the supermode caused by the coupling of the second order modes of individual waveguides. Once excited, ɸ3 could interfere with ɸ1 and ɸ2 to cause spurious power transfer. Duty cycle around 0.2 is chosen since it could effectively suppress the excitation of ɸ3 as suggested by Halir et al. [61]. Figure 2c plots the dependence of neff on the duty cycle. A positive quasi-linear relation is observed. This positive relation is reasonable since larger duty cycle grants a higher Si ratio in the SWG to elevate the effective RI. Although a smaller duty cycle could more effectively suppress ɸ3, it requires more stringent fabrication. Subsequently, duty cycle = 0.25 is selected such that the critical dimension of 215 nm is 20% larger than our current fabrication limit of 180 nm linewidth using 248 nm DUV lithography.

Device Fabrication and Characterization
The waveguide dimension of h × w = 0.4 µm × 1.2 µm is chosen to achieve low loss single mode waveguide [29]. The gap g for the conventional DC and SWG-based WFDC are 0.5 µm and 1 µm respectively. The length of the grating in SWG is 10 µm. Λ = 0.86 µm and duty cycle = 0.25 are selected as the center parameters while some variations are considered to investigate the influence of Λ and duty cycle on coupling efficiency. Conventional DC and SWG-based WFDC with varying Lcs are fabricated in order to achieve different coupling efficiencies.
The fabrication starts from a commercially available 8-inch SOI wafer with a 220 nm Si device layer and 3 µm SiO2 BOX. A 180 nm silicon blanket is epitaxially grown to top up the device layer to 400 nm. The devices are patterned by deep ultra-violet (DUV) photolithography followed by silicon reactive ion etching (RIE). Cladding oxide of 3 µm is then deposited by plasma enhanced chemical vapour deposition (PECVD). Finally, a deep trench with more than 100 µm in depth is etched for butt fiber coupling. The experimental setup for optical testing is presented in Figure 3. The dashed lines show the equipment connection while the glowing lines exhibit the light path. Light is emitted from the MIR laser (Daylight Solution) and passes through a half-wave plate (Thorlab) for polarization control. Transverse-electric (TE) mode is used in the experiment. Next, the light is modulated by a . Concept and Design Optimization Figure 1a shows the schematic of conventional DC which consists of two slightly spaced aveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the oupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 nd ɸ2. The coupling between these two modes allows the EM wave to transfer between these two aveguides. The required coupling length for 100% coupling efficiency is analytically calculated by π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 nd ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and eff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be chieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1

Concept and Design Optimization
and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2.
The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slig waveguides. Owing to evanescent wave coupling, an even mode ɸ1

Concept and Design Optimization
and an odd mode ɸ coupled structure according to the coupled mode theory (CMT). The input EM wave ex and ɸ2.
The coupling between these two modes allows the EM wave to transfer betwee waveguides. The required coupling length for 100% coupling efficiency is analytically c L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equ neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency achieved in a limited wavelength range. n Optimization s the schematic of conventional DC which consists of two slightly spaced o evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the ording to the coupled mode theory (CMT). The input EM wave excites both ɸ1 between these two modes allows the EM wave to transfer between these two uired coupling length for 100% coupling efficiency is analytically calculated by 2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 0]. As λ decreases, the stronger modal confinement tends to equate neff1 and ignificant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two slightly spaced waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode ɸ2 exist in the coupled structure according to the coupled mode theory (CMT). The input EM wave excites both ɸ1 and ɸ2. The coupling between these two modes allows the EM wave to transfer between these two waveguides. The required coupling length for 100% coupling efficiency is analytically calculated by L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective RI of modes ɸ1 and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to equate neff1 and neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficiency can only be achieved in a limited wavelength range.  Figure 1a shows the schematic of conventional DC which consists of two sl waveguides. Owing to evanescent wave coupling, an even mode ɸ1 and an odd mode coupled structure according to the coupled mode theory (CMT). The input EM wave and ɸ2. The coupling between these two modes allows the EM wave to transfer betw waveguides. The required coupling length for 100% coupling efficiency is analytically L π = (λ/2)/(n eff1 − n eff2 ), where λ is the wavelength and neff1 and neff2 are the effective R and ɸ2 respectively [60]. As λ decreases, the stronger modal confinement tends to e neff2, resulting in the significant drop in neff1 − neff2. Hence, the desired coupling efficien achieved in a limited wavelength range. 3 , it requires more stringent fabrication. Subsequently, duty cycle = 0.25 is selected such that the critical dimension of 215 nm is 20% larger than our current fabrication limit of 180 nm linewidth using 248 nm DUV lithography.

Device Fabrication and Characterization
The waveguide dimension of h × w = 0.4 µm × 1.2 µm is chosen to achieve low loss single mode waveguide [29]. The gap g for the conventional DC and SWG-based WFDC are 0.5 µm and 1 µm respectively. The length of the grating in SWG is 10 µm. Λ = 0.86 µm and duty cycle = 0.25 are selected as the center parameters while some variations are considered to investigate the influence of Λ and duty cycle on coupling efficiency. Conventional DC and SWG-based WFDC with varying L c s are fabricated in order to achieve different coupling efficiencies.
The fabrication starts from a commercially available 8-inch SOI wafer with a 220 nm Si device layer and 3 µm SiO 2 BOX. A 180 nm silicon blanket is epitaxially grown to top up the device layer to 400 nm. The devices are patterned by deep ultra-violet (DUV) photolithography followed by silicon reactive ion etching (RIE). Cladding oxide of 3 µm is then deposited by plasma enhanced chemical vapour deposition (PECVD). Finally, a deep trench with more than 100 µm in depth is etched for butt fiber coupling. The experimental setup for optical testing is presented in Figure 3. The dashed lines show the equipment connection while the glowing lines exhibit the light path. Light is emitted from the MIR laser (Daylight Solution) and passes through a half-wave plate (Thorlab) for polarization control. Transverse-electric (TE) mode is used in the experiment. Next, the light is modulated by a chopper which serves as an external reference signal to the lock-in amplifier (Stanford Research System) to reduce MIR detector noise. The light is then launched to the ZrF 4 MIR fiber (Thorlab) and coupled to the device sitting on the sample stage (Kohzu). Fine fiber alignment is achieved by the 6-axis stage. Finally, the output light is captured by another MIR fiber and routed to the MIR detector (Horiba).   To ensure that the SWG-based WFDC exhibits good DC performance, we measured the self-normalized coupled power X/I at several Lc, where X is the power coupled evanescently through the DC and I = X + T is the total power measured at the DC output (See Figure 1a). According to the theoretical DC model analyzed by the CMT, X/I should satisfy the sine squared function: The experimental result of X/I vs. Lc at 3.7 µm is shown in Figure 4b. The data is fitted well by the sine squared function with adj. R-square of 0.997, demonstrating the good DC performance of our SWG-based WFDC.
We study the influence of Λ on the coupling efficiency of our SWG-based WFDC. Λ is varied from 0.81 µm to 0.85 µm in steps of 0.01 µm while the duty cycle and the number of SWG periods are fixed at 0.25 and 30 respectively. The result is presented in Figure 4c. Devices with Λ = 0.81 µm and Λ = 0.82 µm show gradual increase of coupling efficiency throughout the wavelength range of 3.66-3.895 µm. In contrast, a local maximum of coupling efficiency can be observed in devices with Λ = 0.83 µm, 0.84 µm and 0.85 µm. Meanwhile, the local maximum shifts to a longer wavelength with increasing Λ as indicated by the blue dashed arrow in Figure 4c. These local maximums are caused by the rapid increase of neff as the wavelength approaches the Bragg wavelength (see Figure  2b). The high neff enhances the coupling between the waveguides, leading to higher coupling efficiency. The local maximum shifts due to the fact that larger Λ corresponds to larger Bragg wavelength according to λ B = 2n eff Λ. Apart from the local maximum, another observation is that higher Λ provides stronger coupling at individual wavelength as indicated by the orange dashed arrow. This could be attributed to the stronger coupling offered by higher neff since the larger Λ is closer to the Bragg diffraction zone as illustrated in Figure 2b.  To ensure that the SWG-based WFDC exhibits good DC performance, we measured the self-normalized coupled power X/I at several L c , where X is the power coupled evanescently through the DC and I = X + T is the total power measured at the DC output (See Figure 1a). According to the theoretical DC model analyzed by the CMT, X/I should satisfy the sine squared function: The experimental result of X/I vs. L c at 3.7 µm is shown in Figure 4b. The data is fitted well by the sine squared function with adj. R-square of 0.997, demonstrating the good DC performance of our SWG-based WFDC.
We study the influence of Λ on the coupling efficiency of our SWG-based WFDC. Λ is varied from 0.81 µm to 0.85 µm in steps of 0.01 µm while the duty cycle and the number of SWG periods are fixed at 0.25 and 30 respectively. The result is presented in Figure 4c. Devices with Λ = 0.81 µm and Λ = 0.82 µm show gradual increase of coupling efficiency throughout the wavelength range of 3.66-3.895 µm. In contrast, a local maximum of coupling efficiency can be observed in devices with Λ = 0.83 µm, 0.84 µm and 0.85 µm. Meanwhile, the local maximum shifts to a longer wavelength with increasing Λ as indicated by the blue dashed arrow in Figure 4c. These local maximums are caused by the rapid increase of n eff as the wavelength approaches the Bragg wavelength (see Figure 2b). The high n eff enhances the coupling between the waveguides, leading to higher coupling efficiency. The local maximum shifts due to the fact that larger Λ corresponds to larger Bragg wavelength according to λ B = 2n eff Λ. Apart from the local maximum, another observation is that higher Λ provides stronger coupling at individual wavelength as indicated by the orange dashed arrow. This could be attributed to the stronger coupling offered by higher n eff since the larger Λ is closer to the Bragg diffraction zone as illustrated in Figure 2b. The dependence of coupling efficiency on the duty cycle is shown in Figure 4d. Λ and the number of SWG periods are fixed at 0.81 µm and 30 respectively. As the duty cycle varies from 0.23 to 0.29 in steps of 0.02, the increasing trend of coupling efficiency throughout 3.66-3.895 µm maintains, revealing that ɸ3 is successfully suppressed by the small duty cycle to achieve a stable coupling. The coupling efficiency is positively related to duty cycle at each individual wavelength as indicated by the orange dashed arrow. This is a result of higher duty cycle offering higher neff (see Figure 2c) so that coupling is strengthened. Additionally, this suggests that SWG could be adopted in DC to reduce the device footprint as well due to its capability of offering stronger coupling.
In the following discussions, the devices are all designed with the same Λ = 0.83 µm and duty cycle = 0.25. From Figure 4b, we could identify Lπ which provides 100% coupling efficiency for 3.7 µm EM wave. Similarly, we extract Lπ at individual wavelength for both conventional DC and SWGbased WFDC. The result is shown in Figure 5a. It is clear that Lπ drops almost linearly in conventional DC and yet is more stable in SWG-based WFDC. We define the percentage change in Lπ as: In conventional DC, the percentage change in Lπ is 41.4% while our SWG-based WFDC could achieve 16.7%. A more stable Lπ suggests the SWG-based WFDC is more resistant to wavelength change compared to conventional DC. Figure 5b presents the comparison of simulated mode profiles of conventional DC and SWG-based WFDC under different conditions. The simulation is performed using Lumerical FDTD Solutions [63]. For conventional DC, while Lc = 60 µm guarantees 100% coupling efficiency at 3.77 µm, a substantial amount of power is transmitted through the original waveguide as the wavelength rises to 3.89 µm. Thus, 100% coupling efficiency is compromised. However, in SWG-based WFDC with Lc = 43.16 µm, 100% power coupling ratio could be maintained even if the wavelength changes from 3.77 µm to 3.89 µm. The dependence of coupling efficiency on the duty cycle is shown in Figure 4d. Λ and the number of SWG periods are fixed at 0.81 µm and 30 respectively. As the duty cycle varies from 0.23 to 0.29 in steps of 0.02, the increasing trend of coupling efficiency throughout 3.66-3.895 µm maintains, revealing that 3 of 13 nventional DC which consists of two slightly spaced pling, an even mode ɸ1 and an odd mode ɸ2 exist in the mode theory (CMT). The input EM wave excites both ɸ1 des allows the EM wave to transfer between these two or 100% coupling efficiency is analytically calculated by elength and neff1 and neff2 are the effective RI of modes ɸ1 e stronger modal confinement tends to equate neff1 and neff2. Hence, the desired coupling efficiency can only be 3 is successfully suppressed by the small duty cycle to achieve a stable coupling. The coupling efficiency is positively related to duty cycle at each individual wavelength as indicated by the orange dashed arrow. This is a result of higher duty cycle offering higher n eff (see Figure 2c) so that coupling is strengthened. Additionally, this suggests that SWG could be adopted in DC to reduce the device footprint as well due to its capability of offering stronger coupling.
In the following discussions, the devices are all designed with the same Λ = 0.83 µm and duty cycle = 0.25. From Figure 4b, we could identify L π which provides 100% coupling efficiency for 3.7 µm EM wave. Similarly, we extract L π at individual wavelength for both conventional DC and SWG-based WFDC. The result is shown in Figure 5a. It is clear that L π drops almost linearly in conventional DC and yet is more stable in SWG-based WFDC. We define the percentage change in L π as: In conventional DC, the percentage change in L π is 41.4% while our SWG-based WFDC could achieve 16.7%. A more stable L π suggests the SWG-based WFDC is more resistant to wavelength change compared to conventional DC. Figure 5b Figure 5c shows the wavelength-flattened performance by comparing the operation bandwidth of SWG-based WFDC with that of conventional DC at 100% coupling efficiency respectively. The acceptance range is defined as 98-100%. Similar to our previous work, the lower limit and upper limit of the 98-100% range is defined as the first wavelength that stays in this range and the wavelength with the highest coupling ratio respectively [64]. As shown in Figure 5c where Lc = 43.16 µm in the SWG-based WFDC, the coupling efficiency is maintained between 98% and 100% over the wavelength range of 3.67-3.845 µm (175 nm span). Nevertheless, in conventional DC with Lc = 49 µm, the corresponding wavelength range is only 3.765-3.8 µm (35 nm span). Fivefold enhancement is realized for 100% coupling efficiency. The drastic drop of coupling efficiency (or the trough) at 3.67 µm is observed which could be explained by the SWG reflection as wavelength approaches the Bragg wavelength. This trough is utilized for RI sensing in the following context. The simulation results of coupling efficiency derived by 3D finite-difference time domain (FDTD) simulation are also presented in Figure 5c, which is consistent with the experimental data.

Investigation of Sensing Performance
Dichloromethane (CH2Cl2) is a germinal organic liquid with important applications in industry as a solvent. The detection of CH2Cl2 is critical since it is hazardous while being colorless and volatile. Here we investigate the sensing capability of our device for CH2Cl2 detection in ethanol (C2H5OH) by simulation. Figure 6a shows the complex RI of both CH2Cl2 and C2H5OH. The imaginary part of RI of CH2Cl2 is much lower than that of C2H5OH, indicating mixture with higher CH2Cl2 concentration will cause weaker light attenuation. The difference in the real part of RI between CH2Cl2 and C2H5OH exceeds 0.03 across 3.65 µm to 3.9 µm. Such a difference is able to induce a significant shift of Bragg wavelength when CH2Cl2 concentration changes in the mixture. The complex RI of CH2Cl2 and C2H5OH are adopted from [65]. The complex RI of the mixture is calculated using Arago-Biot equations which states both the real and imaginary part of RI of the mixture are the linear combination of the two ingredients with their concentration as the linear coefficients respectively [66].
We investigate RI sensing and absorption sensing enabled by the change of the real part and imaginary part of RI, respectively, by simulation performed in a device 136.8 µm long using Lumerical FDTD Solutions. For RI sensing, the normalized transmitted power (T/I) spectrum of mixtures with different CH2Cl2 concentration is presented in Figure 6b. The trough blue shifts due to the rising surrounding effective RI caused by the drop of CH2Cl2 concentration. Figure 6c presents the zoom-in of Figure 5b to the low CH2Cl2 concentration region for better visualization. The first derivative of 0% CH2Cl2 concentration curve in Figure 6b is derived and plotted in Figure 6d. The magnitude of this first derivative indicates the sensitivity of T/I to wavelength change, and thus  Figure 5c shows the wavelength-flattened performance by comparing the operation bandwidth of SWG-based WFDC with that of conventional DC at 100% coupling efficiency respectively. The acceptance range is defined as 98-100%. Similar to our previous work, the lower limit and upper limit of the 98-100% range is defined as the first wavelength that stays in this range and the wavelength with the highest coupling ratio respectively [64]. As shown in Figure 5c where L c = 43.16 µm in the SWG-based WFDC, the coupling efficiency is maintained between 98% and 100% over the wavelength range of 3.67-3.845 µm (175 nm span). Nevertheless, in conventional DC with L c = 49 µm, the corresponding wavelength range is only 3.765-3.8 µm (35 nm span). Fivefold enhancement is realized for 100% coupling efficiency. The drastic drop of coupling efficiency (or the trough) at 3.67 µm is observed which could be explained by the SWG reflection as wavelength approaches the Bragg wavelength. This trough is utilized for RI sensing in the following context. The simulation results of coupling efficiency derived by 3D finite-difference time domain (FDTD) simulation are also presented in Figure 5c, which is consistent with the experimental data.

Investigation of Sensing Performance
Dichloromethane (CH 2 Cl 2 ) is a germinal organic liquid with important applications in industry as a solvent. The detection of CH 2 Cl 2 is critical since it is hazardous while being colorless and volatile. Here we investigate the sensing capability of our device for CH 2 Cl 2 detection in ethanol (C 2 H 5 OH) by simulation. Figure 6a shows the complex RI of both CH 2 Cl 2 and C 2 H 5 OH. The imaginary part of RI of CH 2 Cl 2 is much lower than that of C 2 H 5 OH, indicating mixture with higher CH 2 Cl 2 concentration will cause weaker light attenuation. The difference in the real part of RI between CH 2 Cl 2 and C 2 H 5 OH exceeds 0.03 across 3.65 µm to 3.9 µm. Such a difference is able to induce a significant shift of Bragg wavelength when CH 2 Cl 2 concentration changes in the mixture. The complex RI of CH 2 Cl 2 and C 2 H 5 OH are adopted from [65]. The complex RI of the mixture is calculated using Arago-Biot equations which states both the real and imaginary part of RI of the mixture are the linear combination of the two ingredients with their concentration as the linear coefficients respectively [66].
We investigate RI sensing and absorption sensing enabled by the change of the real part and imaginary part of RI, respectively, by simulation performed in a device 136.8 µm long using Lumerical FDTD Solutions. For RI sensing, the normalized transmitted power (T/I) spectrum of mixtures with different CH 2 Cl 2 concentration is presented in Figure 6b. The trough blue shifts due to the rising surrounding effective RI caused by the drop of CH 2 Cl 2 concentration. Figure 6c presents the zoom-in of Figure 5b to the low CH 2 Cl 2 concentration region for better visualization. The first derivative of 0% CH 2 Cl 2 concentration curve in Figure 6b is derived and plotted in Figure 6d. The magnitude of this first derivative indicates the sensitivity of T/I to wavelength change, and thus RI sensitivity. Four wavelengths namely 3.668 µm and 3.676 µm with high first derivative, 3.685 µm with a medium first derivative, and 3.727 µm with a near-zero derivative are studied. T/I is plotted against different CH 2 Cl 2 concentrations in Figure 6e. The slopes of the fitted curves represent the sensitivity at each wavelength. Sensitivity of −0.47%, −0.45%, −0.17%, and 0% T/I change per percentage of CH 2 Cl 2 concentration is demonstrated at 3.668 µm, 3.676 µm, 3.685 µm, and 3.727 µm respectively. This result shows that the first derivative of the T/I spectrum serves as a good reference for the selection of sensing wavelength as its high value returns high sensitivity while its near-zero value returns near-zero sensitivity. µm with a medium first derivative, and 3.727 µm with a near-zero derivative are studied. T/I is plotted against different CH2Cl2 concentrations in Figure 6e. The slopes of the fitted curves represent the sensitivity at each wavelength. Sensitivity of −0.47%, −0.45%, −0.17%, and 0% T/I change per percentage of CH2Cl2 concentration is demonstrated at 3.668 µm, 3.676 µm, 3.685 µm, and 3.727 µm respectively. This result shows that the first derivative of the T/I spectrum serves as a good reference for the selection of sensing wavelength as its high value returns high sensitivity while its near-zero value returns near-zero sensitivity. The capability of absorption sensing is also examined. This sensing mechanism is especially enabled by operating in the MIR region. Since CH2Cl2 and C2H5OH have distinct imaginary parts of RI, their mixture shows different absorption strength in changing CH2Cl2 concentrations. As shown in the inset of Figure 6f, [area B (green)] shows the integration of total output power (X + T) over the spectrum when the mixture is free of CH2Cl2 while [area A (grey) + area B (green)] presents the integration when the mixture is free of C2H5OH. Pure CH2Cl2 allows stronger light transmission since its low imaginary part of RI causes less absorption. We plot the integration of X + T (or power integration) of the mixture with different CH2Cl2 concentrations. The slope of the fitted linear curve represents a sensitivity of 0.12% change in the normalized total integrated output per percentage of CH2Cl2 concentration in absorption sensing.

Conclusions
In summary, we design, fabricate, and characterize a compact wavelength-flattened directional coupler based chemical sensor for the MIR. Broadband performance is achieved by incorporating a subwavelength structure to the directional coupler for dispersion engineering. Meanwhile, the 10% 5% 2% 1% 0% Figure 6. Simulated sensing results of CH 2 Cl 2 detection in C 2 H 5 OH using 3D FDTD simulation. (a) Wavelength dependent RI of CH 2 Cl 2 and C 2 H 5 OH adopted from [65]. The upper and lower panel show the imaginary part and real part respectively. (b) Self-normalized transmitted power (T/I) spectrum in different CH 2 Cl 2 concentration. (c) Zoom-in self-normalized transmitted power (T/I) for low CH 2 Cl 2 concentration sensing. (d) The first derivative derived from 0% CH 2 Cl 2 curve in (b). (e) The self-normalized transmitted power (T/I) versus CH 2 Cl 2 concentration at different wavelengths. The sensitivities can be extracted from the slope of the fitted linear curves. (f) Normalized total integrated output power (X + T) versus concentration. The slope shows the sensitivity. Inset: Spectrum of X + T. Area B is the power integration for pure C 2 H 5 OH while Area A + B is the power integration for pure CH 2 Cl 2 .
The capability of absorption sensing is also examined. This sensing mechanism is especially enabled by operating in the MIR region. Since CH 2 Cl 2 and C 2 H 5 OH have distinct imaginary parts of RI, their mixture shows different absorption strength in changing CH 2 Cl 2 concentrations. As shown in the inset of Figure 6f, [area B (green)] shows the integration of total output power (X + T) over the spectrum when the mixture is free of CH 2 Cl 2 while [area A (grey) + area B (green)] presents the integration when the mixture is free of C 2 H 5 OH. Pure CH 2 Cl 2 allows stronger light transmission since its low imaginary part of RI causes less absorption. We plot the integration of X + T (or power integration) of the mixture with different CH 2 Cl 2 concentrations. The slope of the fitted linear curve represents a sensitivity of 0.12% change in the normalized total integrated output per percentage of CH 2 Cl 2 concentration in absorption sensing.

Conclusions
In summary, we design, fabricate, and characterize a compact wavelength-flattened directional coupler based chemical sensor for the MIR. Broadband performance is achieved by incorporating a subwavelength structure to the directional coupler for dispersion engineering. Meanwhile, the sensitive trough at the Bragg wavelength introduced by the subwavelength grating structure allows a compact sensor with high sensitivity to RI change. Around fivefold enhancement in the operation bandwidth compared to the conventional directional coupler is demonstrated experimentally for 100% coupling efficiency in the device with a small length of~40 µm. Dichloromethane (CH 2 Cl 2 ) detection in ethanol (C 2 H 5 OH) is investigated by 3D FDTD simulation to examine sensing performance and obtain sensitivity. The sensing capability of a device with 136.8 µm length reveals −0.47% change in the self-normalized transmitted power per percentage of CH 2 Cl 2 concentration in RI sensing, while 0.12% change in total integrated output power is realized in absorption sensing. Our device can potentially work for sensing of tertiary mixture as well as for MIR applications that require broadband operation such as spectroscopic sensing systems.