1. Introduction

Narrowband perfect absorption and emission in the infrared (IR) region are critical prerequisites for modern spectroscopic applications such as multi-wavelength pyrometers [1–4], nondispersive infrared (NDIR) gas sensors[5–9], IR radiative heaters[10–12], and thermophotovoltaics [13–15].

Within the past decade, artificial photonic structures, such as plasmonic metamaterials [16–22] and photonic crystals [23–27], have been used to realize spectrally selective absorbers and emitters. Unfortunately, these subwavelength patterned structures often entail complex and high-cost lithography processes, which severely complicate the fabrication on large-scale surfaces. Moreover, the inherent losses of conductive elements in the IR region arguably dampen surface electromagnetic (EM) waves on nanopatterned metallic structures and thus impair the spectral resolution of plasmonic absorbers and emitters [28].

Planar photonic structures have the great advantage of scalability without the burden of intricate patterning. For example, lossy dielectric films have been used to tailor the selective absorption by modulating the overall phases of reflections for destructive interference effects [29–31]. However, the absorption bandwidths of these structures are often relatively broad for some demanding spectroscopic applications. In another approach, Tamm plasmon structures in which plasmon polaritons are supported by one-dimensional multilayers have been demonstrated for narrowband thermal emitters [32–35]. In recent years, Tamm plasmon structures have shown rapid progress toward the enhancement of quality factor (Q-factor), for example, by coupling the resonances of a cavity and Tamm plasmon structure to achieve sharper emission peak [34]. Recently, Zhu et al. proposed a tunable narrowband thermal emitter with an additional bilayer cavity and phase-changing material [36]. Optical cavities, which have been used as fundamental building blocks for lasers, amplifiers, optical switches, and interferometers, are a powerful configuration to achieve a high spectral resolution due to their strict resonance conditions. Therefore, efforts have been made to utilize the interference effect of resonant cavities to achieve narrowband thermal absorbers and emitters, or to design color filters [37–43].

In this study, we propose and realize ultra-narrowband perfect absorbers and emitters based on the multiple beam interference in a Gires-Tournois (GT) etalon with the presence of low metallic loss. A GT etalon (also named as asymmetric Fabry-Perot etalon) is a standing-wave optical cavity, which reflects light of all wavelengths but exhibits a strong dispersion of the phase delay. The phase delay is maximal at the resonance wavelength, allowing high energy confinement within the cavity. With the presence of a small amount of inherent metallic loss in a GT etalon, ultra-narrowband near-unity absorption/emission can be achieved.

We aim to explore how the multiple parameters of GT cavity in various configurations impact its spectral characteristics and performance, thus providing an applicable framework to optimize spectral resolution and directionality for IR spectroscopic applications. Using rigorous coupled-wave analysis (RCWA), we systematically examined GT resonator-based perfect absorbers in three distinct configurations: lossless dielectric cavity on metal (DM), metal-dielectric-metal (MDM), and distributed Bragg reflector (DBR)-dielectric-metal (DDM). Following the designs rationally determined during the optimization, we realized both MDM and DDM resonator-based emitters targeting the molecular absorption bands of carbon monoxide (at about 2.3 μm) [44]. We also report, to the best of our knowledge, the first on-chip DDM resonator-based IR detector targeting the molecular absorption band of methane (at 3.3 μm) [44]. The DDM resonator-based detector exhibits excellent spectral resolution.

 

Fig. 1 (a) Schematic illustration of a GT resonator-based perfect absorber. Three distinct configurations of the GT resonator-based perfect absorbers examined in this work: (b) lossless dielectric cavity on metal (DM), (c) metal–dielectric–metal (MDM), and (d) distributed Bragg reflector–dielectric–metal (DDM).

Download Full Size | PDF

2. Results and discussion

2.1. Absorption mechanism

In its original form, a GT resonator consists of a semi-transparent mirror in the front and a highly reflective mirror in the back, which are separated by a lossless dielectric spacer (see Fig. 1(a)). The front semi-transparent mirror is the Fresnel reflection either from a dielectric/air interface (see Fig. 1(b)), ultra-thin metallic mirror (see Fig. 1(c)), or pairs of DBRs (see Fig. 1(d)). A highly reflective metallic layer in the IR region is used as the back mirror of the cavity to provide high reflection and relatively low metallic absorption.

The reflection coefficient r_GT of a GT etalon in the presence of metallic loss can be derived from the superposition of multiple reflections [45]:

 

Fig. 2 Rigorous coupled-wave analysis of the EM responses of GT resonator-based perfect absorbers. Schematic diagrams and intensity plots of the electric field, magnetic field, total absorption and reflectivity, absorptivity, transmissivity characteristics of the (a) SiC (165 nm)–Au (200 nm) resonator-based absorber, (b) Au (15 nm)–SiO₂ (740 nm)–Au (200 nm) resonator-based absorber, and (c) 3[SiO₂ (415 nm)–Si (175 nm)]–SiO₂ (820 nm)–Au (200 nm) resonator-based absorber at the resonance wavelength, λ_res = 2.37 μm.

Download Full Size | PDF

To demonstrate the performance of GT resonator-based absorbers, we numerically simulated the EM responses of the absorbers in three configurations; DM, MDM, and DDM. The absorptivities, electric field, magnetic field, and total power absorption distribution of the GT resonator-based perfect absorbers were calculated using the commercial RCWA package (DiffractMOD;, Synopsys’s Rsoft) [46]. In the simulation setup, the excited EM field propagates along the z-axis and the electric field oscillates along the x-axis. The electric and magnetic fields are normalized to the input fields. The transmittivities (T) and reflectivities (R) were calculated across the simulation domain. The absorptivity A of the absorbers was calculated with the following equation: $A=1-\left(T+R\right)$. The dielectric functions of Au, Al, Si, SiO${ }_2$, and Al${ }_2$O${ }_3$ were taken from the literature [47]. Fig. 2 shows the schematic diagrams, simulated electric and magnetic field distributions, total absorption and reflectivity, absorptivity, and transmissivity characteristics for the three configurations. The DM absorber is composed of a 165 nm thick SiC film, which is coated on a 200 nm thick Au back reflector and serves as a resonant cavity. The MDM absorber consists of a 740 nm thick SiO${ }_2$ cavity sandwiched between 15 and 200 nm thick Au layers. The DDM absorber comprises an 820 nm thick SiO${ }_2$ cavity sandwiched between a DBR with three SiO${ }_2$ (415 nm)–Si (175 nm) pairs and a 200 nm thick Au back reflector.

The simulation results for all the three configurations reveal that the absorbed EM waves centered at 2.37 μm resonate within the optical cavities, yet with different degrees of confinement. Firstly, the single SiC cavity on the Au back reflector exhibits a low field amplitude and confinement that can be attributed to the low reflectance at the SiC/air boundary (Fig. 2(a)). As described in Equation (3), the low reflectance of the front reflector undermines the phase accumulation of the resonant waves, thus resulting in an inevitably low absorptivity (A $\approx$ 0.2). Due to the poor performance of DM-type absorbers, we used this configuration only for the comparison in numerical simulation. Fig. 2(b) shows that the simulated electric and magnetic fields of the Au–SiO${ }_2$–Au absorber are highly localized in the cavity, resulting in a near-unity absorptivity (A $\approx$ 0.97). The reason for this is that the 15 nm thick Au partial reflector exhibits a higher reflectivity than the SiC/air interface and thus further promotes the multi-reflection in the cavity. A closer look at the total absorption distribution of the MDM configuration reveals that the absorption intensity is much higher at the front reflector than at the back reflector because a substantial fraction of light encounters the inherent losses while penetrating the front metallic layer before being confined within the cavity for further multi-beam interference. Finally, the DBR with various pairs of alternating low- and high-index dielectric films was adopted because it exhibits extremely low absorption losses in the IR region compared with metallic reflectors. The reflectivity of a DBR at its photonic stopband can be extremely high if the index contrast between the two alternating dielectric materials of the Bragg pairs is sufficiently high and the number of layers is large. Therefore, the thin metallic front layer can be replaced by a DBR, which is expected to provide a higher reflectivity while avoiding unwanted metal losses. In the proposed configuration, the thickness of the alternating DBR layers was designed to match its photonic stop band with the resonance mode of the cavity. As expected, the simulated spectra of the 3(SiO${ }_2$–Si)–SiO${ }_2$–Au absorber in Fig. 2(c) show unity absorption and a remarkably high spectral resolution compared with that of the Au–SiO${ }_2$–Au absorber. Specifically, the full widths at half maximums (FWHMs) of the Au–SiO${ }_2$–Au and 3(SiO${ }_2$–Si)–SiO${ }_2$–Au absorbers are approximately 120 and 20 nm, respectively. The DDM design is efficient enough to be directly integrated in a thermal sensing platform because the overall absorption occurs at the metallic back reflector and eventually turns into heat. This heat can be readily harvested and effectively converted into electrical signals if the DDM absorber is loaded on a pyroelectric or thermoelectric transducer. The achieved ultra-narrowband resolution leads to a great advantage of this DDM design over currently available photoconductive IR sensors with a typical spectral resolution of about 1 μm.

 

Fig. 3 (a) Simulated and (b) experimental absorptivity of the MDM-type Au–SiO₂–Au absorber with varying thickness of the top metallic layer (t_f = 15 nm, 30 nm, and 50 nm). (c) Simulated and (d) experimental absorptivity of the DDM-type n_pair(SiO₂–Si)–SiO₂–Au absorber with varying number of DBR pairs (n_pair = 1–5).

Download Full Size | PDF

2.2. Optimization of the spectral resolution and absorptivity

The MDM-type Au–SiO${ }_2$–Au absorbers with various front reflector thicknesses and DDM-type $n_{pair}$(SiO${ }_2$–Si)–SiO${ }_2$–Au absorbers with different numbers of $n_{pair}$ of DBR pairs were simulated and experimentally verified. The front Au reflector thickness of the Au–SiO${ }_2$–Au absorbers was varied from 15 to 30 to 50 nm. The thicknesses of the SiO${ }_2$ cavity and Au back reflector were 600 and 200 nm, respectively. Figs. 3(a) and 3(b) show the simulated and experimental absorptivities. As expected, the spectral absorptivity curves sharpen when the thickness t_f increases; however, the absorptivity gradually decreases and vanishes. As mentioned above, there is always a tradeoff between the absorptivity and spectral resolution due to the absorption losses at the metallic front reflector. The absorption peaks shift towards the longer wavelengths as the metallic front layer becomes thinner, which is due to the fact that the phase shift of the thinner film upon reflection at the top thin metallic film decreases.

The number of DBR pairs of the DDM-type $n_{pair}$(SiO${ }_2$–Si)–SiO${ }_2$–Au absorbers was varied from one to five. The SiO${ }_2$ cavity was 820 nm thick. The alternating SiO${ }_2$ and Si layers of the DBR pairs were 415 and 175 nm thick, respectively. The simulated and experimental absorptivities are presented in Figs. 3(c) and 3(d). The unchanged resonance position reflects the fact that the phase of the light waves remains constant during transmission through the DBR stack. The absorptivity increases proportionally to the number of pairs and reaches the highest value at $n_{pair}$ = 3 before deteriorating at $n_{pair}$ = 4 and 5. This is because the increase of the number of pairs in the DBR enhances the reflectivity of the front reflector but reduces the amount of light penetrating the cavity. These results provide a comprehensive insight into the optimization of the GT resonator-based absorber for intended applications.

 

Fig. 4 Tuning the resonance wavelength of GT resonator-based absorbers. (a–d) Cross-sectional scanning electron microscope (SEM) images and (e–h) absorptivity. The blue and red lines are the simulated and measured spectra, respectively. (a), (e) Au(15 nm)–Si(220 nm)–Au(200 nm); (b), (f) Au(15 nm)–Si(245 nm)–Au(200 nm); (c), (g) 3[SiO₂(415 nm)–Si(175 nm)]–SiO₂(820 nm)–Au(150 nm); and (d), (h) 3[SiO₂(570 nm)–Si(240 nm)]–SiO₂(1170 nm)–Au(150 nm).

Download Full Size | PDF

2.3. Tuning the resonant wavelength

Eq. (2) indicates that the resonance peak can be continuously tuned across the mid-IR wavelength range by adjusting the cavity thickness. The condition for constructive resonance in the cavity is $2\phi =2m\pi$, where m is an integer. In the case of the DM structure, the optical thickness of the SiC cavity is designed to be close to quarter-wavelength because the phase shifts of the reflecting waves at the front and back reflectors are 0 and π, respectively. On the other hand, the optical thicknesses of the Si and the SiO${ }_2$ cavities in the MDM and DDM structures are designed to be approximately half-wavelength because light reflected from metal surfaces and boundaries, which is incident from lower- to higher-index materials, acquires a phase shift of π. Note that the compensation for a phase shift less than π at the front metallic reflector should be considered to precisely tailor the absorption peak during the optimization.

We manifest the ability to tune the resonance wavelength of the MDM-type Au–Si–Au absorber with a resonance wavelength targeted at the absorption spectra of two hazardous gases: carbon dioxide (λ = 2.1 μm) and carbon monoxide (λ = 2.3 μm) [44]. The Si cavity thicknesses were set to 220 and 245 nm, while those of the front and back Au reflector were 15 and 200 nm, respectively. The cross-sectional scanning electron microscope (SEM) images of the fabricated structures are presented in Figs. 4(a) and 4(b). One fabricated device exhibits a prominent peak at the center wavelength of 2.1 μm, with an absorptivity of 0.73 and a FWHM of 400 nm (Fig. 4(e)). The other one exhibits an absorption peak at the center wavelength of 2.33 μm, with an absorptivity of 0.7 and a FWHM of 410 nm (Fig. 4(f)).

The resonance wavelength of the DDM-type 3(SiO${ }_2$–Si)–SiO${ }_2$–Au absorber is aimed at 2.37 μm and 3.31 μm. At ${\lambda }_{res}$ = 2.37 μm, the optimum parameters are $t_{cavity }$= 820 nm, t_Si = 175 nm, and $t_{Si{O}_2}$ = 415 nm. At ${\lambda }_{res}$ = 3.31 μm, the optimum parameters are $t_{cavity}$ = 1170 nm, t_Si = 240 nm, and $t_{Si{O}_2}$ = 570 nm. A thin film of titanium (5 nm) was deposited as an adhesion layer between SiO${ }_2$ and Au. The thicknesses of the Au back reflectors in both cases were 150 nm. The cross-sectional SEM images of the structures are presented in Figs. 4(c) and 4(d). The experimental absorptivity reaches 0.88 with a FWHM of 28 nm at 2.37 μm (Fig. 4(g)) and 0.79 with a FWHM of 30 nm at 3.31 μm (Fig. 4(h)). The simulated and measured results are in excellent agreement. The slightly broadened bandwidths and lower absorptivities of the experimental spectra compared with the simulated data in both configurations can be attributed to the influence of the lossy adhesion layers (2 nm thick Ti) and the inherently granular and nonuniform nature of the deposited films.

 

Fig. 5 Ray-optics representation of the propagating light in the cavities with (a) high and (d) low refractive indices depending on the incident angles. (b), (e) Simulated and (c), (f) experimental characteristics of the angular-dependent absorptivityof the Au–Si–Au and Au–Al₂O₃–Au absorbers, respectively.

Download Full Size | PDF

2.4. Customization of directionality

The directionality of GT-based absorbers can be customized to meet specific application requirements. A closer look at Equation (2) indicates that the apparent optical thickness $nd\cos \theta$ varies with the propagation angles of the light in the cavity such that the cavity thickness seems to become thinner at an oblique incidence. Hence, the resonance peak is shifted to shorter wavelengths. Snell’s law of refraction indicates that the refracted rays in the high-refractive index cavity (n_H) are less affected than those in the low-refractive index cavity (n_L) for a given change in the incident angle $\mathrm{\Delta }{\theta }_i$: $\mathrm{\Delta }{\theta }_H$ < $\mathrm{\Delta }{\theta }_L$ (see Figs. 5(a) and 5(d)), where $\mathrm{\Delta }{\theta }_H$ and $\mathrm{\Delta }{\theta }_L$ are the changes in the angle of refraction in high- and low-refractive index cavities, respectively. In other words, if the refractive index of the cavity is high, it can effectively suppress the angular sensitivity of a GT cavity. In contrast, to achieve high angular selectivity, one must choose the cavity with low refractive index. To demonstrate the above-mentioned mechanism, we numerically simulated and fabricated two MDM-type absorbers with low ($n_{A{l}_2{O}_3}\approx$ 1.6) and high $(n_{Si}\approx$ 3.4) refractive indices, that is, Au (15 nm)–Al${ }_2$O${ }_3$ (600 nm)–Au (200 nm) and Au (15 nm)–Si (245 nm)–Au (200 nm). The simulated results are shown in Figs. 5(b) and 5(e); the experimental results are shown in Figs. 5(c) and (f). As expected, the absorptivity of the Au–Al${ }_2$O${ }_3$–Au structure is significantly shifted to shorter wavelengths at incident angles above 20${ }^o$, while that of the Au–Si–Au structure retains high values and a constant resonance wavelength at incident angles up to 70${ }^o$. The angular insensitivity of high-index cavity absorbers is useful for IR stealth applications, which require perfect omnidirectional absorption. Apparently, the directionality of MDM-type absorbers can be customized by choosing cavity material with an appropriate refractive index.

 

Fig. 6 (a) Simulated and (b) experimental angular-dependent absorptivity of 3(SiO₂–Si)–SiO₂–Au resonator-based perfect absorbers.

Download Full Size | PDF

The angular-dependent characteristics of the DDM-type absorber were also investigated. More specifically, the spectral directionality of the 3(SiO${ }_2$–Si)–SiO${ }_2$–Au absorber was numerically simulated (see Fig. 6(a)) and experimentally verified (see Fig. 6(b)). The absorptivities of the structure are significantly shifted towards shorter wavelengths at incident angles above 10${ }^o$. This can be attributed to the inherently dispersive characteristics of the Bragg reflectors. The angular sensitivity of the DDM absorber is suitable for angularly selective applications such as IR detectors for production lines, automobiles (which utilize small fields of view and need to exclude environmental noise) [48], and high-efficiency solar energy conversion [49].

 

Fig. 7 Narrowband thermal emitters based on the MDM-type and DDM-type GT resonators. (a) measurement setup; (b) measured reflectivity and spectral emissivity characteristics of a MDM-type emitter, inset: photo of the fabricated thermal emitter; (c) measured reflectivity and spectral emissivity characteristics of a DDM-type emitter.

Download Full Size | PDF

2.5. MDM and DDM resonator-based thermal emitters

In this section, we present both MDM and DDM resonator-based emitters for the application in low-cost, compact NDIR gas sensors. According to Kirchhoff’s law of thermal radiation, the emissivity is equal to the absorptivity in thermodynamic equilibrium. This means that a perfect absorber is also a perfect emitter. Au(15 nm)–Al${ }_2$O${ }_3$(600 nm)–Au(200 nm) structure and 3[SiO${ }_2$(415 nm)–Si(175 nm)]–SiO${ }_2$(820 nm)–Au(150 nm) structure were deposited on 28×5 mm${ }^2$ silicon substrates. Two ends of the Au back reflector were used as electrodes of the emitter. To aim for practical NDIR gas sensing applications, the resonant emission was targeted at around the molecule-specific absorption band of carbon monoxide, ${\lambda }_{res}$ = 2.3 μm. Fig. 7(a) illustrates the measurement setup to characterize the emission spectra of the emitters. The sample was set inside an ultra-high vacuum (UHV) chamber with a base pressure of 10${ }^{-8}$–10${ }^{-9}$ Torr. The emitter was gradually heated by increasing the electrical current on the Au back reflector. The IR emission was guided through a zinc selenide window to an external port of a Fourier-transform infrared (FTIR) spectrometer. The measured reflectivity, spectral emissivity of the MDM emitter and DDM emitter are shown in Figs. 7(b) and 7(c), respectively. The MDM emitter exhibits a Q-factor of 21 and an emissivity of 0.8. The DDM emitter exhibits a Q-factor of 85 and an emissivity of 0.82. The Q-factor of the resonator is defined as the ratio of the resonance wavelength ${\lambda }_{res}$ to the FWHM of the emission peak, that is, Q = ${\lambda }_{res}$/FWHM. The slight discrepancy between the reflectivity and emission spectra is attributed to the influence of the thermal expansion of the constituent materials.

 

Fig. 8 GT resonator-based narrowband pyroelectric IR detector: (a) schematic illustration, (b) actual implementation, (c) measured reflectivity and spectral responsivity, and (d) schematics of the measurement setup.

Download Full Size | PDF

2.6. DDM resonator-based thermal detector

In a conventional detector, a narrowband optical filter is placed in front of the sensing element for spectrally selective detection, resulting in a bulky design and limited wavelength selection. In our design, we directly integrated a 3(SiO${ }_2$–Si)–SiO${ }_2$–Al absorber in a pyroelectric LiTaO${ }_3$ single-crystal wafer (Fig. 8(a)). We intentionally chose Al as the material for the back reflector for three reasons: first, it exhibits low-loss plasmonic properties similar to Au in the near-to-midIR region [20]; second, no adhesion layer is required to fix the oxide layer on Al; and third, Al is widely used as one of the industry-compatible base metals. The parameters of the GT detector were optimized for greenhouse gas sensing applications to match the absorption band of methane, ${\lambda }_{res}$ = 3.3 μm. The optimal parameters are the following: DBR pairs: $t_{Si{O}_2}$ = 570 nm, t_Si = 240 nm, SiO${ }_2$ cavity thickness: 1170 nm. The 10 mm (height) × 10 mm (width) × 0.02 mm (thickness) single-crystal LiTaO${ }_3$ slab was sandwiched between two 200 nm thick Al electrodes. Note that the upper Al electrode also served as the back reflector of the DDM-resonator-based absorber. This way, IR radiation at the resonance wavelength was selectively absorbed and transferred to the single-crystal LiTaO${ }_3$ slab in the form of heat. The rise in the temperature modifies the polarization of the LiTaO${ }_3$ crystal, resulting in a vertical voltage difference at the upper and lower Al electrodes. To characterize the performance of the device, we used an optical parametric oscillator (OPO; TOPAS Prime, Spectra-Physics) combined with a Ti:Sapphire regenerative amplifier (Solstice, Spectra-Physics) as tunable IR source to measure the spectral response in the range from 2.8 to 4 μm. The laser pulse width and repetition rate were 104 fs and 1 kHz, respectively. The thermal electromotive voltage was amplified with a preamplifier (RS560, Stanford Research Systems) and demodulated with a lock-in amplifier (LI5640, NF Corporation). The IR laser beam was guided onto the sensing area of the detector. The output electrical signal was amplified with a preamplifier and demodulated with a lock-in amplifier (Figs. 8(b) and 8(d)). In Fig. 8(c), the measured responsivity of the detector is plotted against the reflectivity spectrum. Note that the broadened spectrum of the responsivity curve originates from the convolution of the IR laser linewidth and the spectral response of the detector. The spectral responsivity curve exhibits an absorption peak at 3.3 μm, which is highly consistent with the measured reflectivity dip. The FWHM of the reflectivity spectra is as narrow as 22 nm, corresponding to a Q-factor of 151, which is two to four times higher than that of recently reported multilayered absorbers [32, 34]. To the best of our knowledge, this is the first time that an on-chip GT resonator-based perfect absorber approaches such a high spectral resolution. The absorption band is narrow enough to distinguish and label specific gases in a complex mixture. The demonstration of our detector devices presented here paves the way for a new class of complementary metal-oxide-semiconductor (CMOS)-compatible, custom-made gas analyzers as well as other spectroscopic devices operating in the IR region.

3. Conclusion

In summary, we report high-performance narrowband IR emitters and detectors by combining GT etalon-based perfect absorbers with CMOS-compatible platforms. The absorption mechanism and spectral characteristics of GT resonator-based perfect absorbers in various configurations were both numerically and experimentally investigated to establish the guiding principle for the optimization of the spectral resolution and customization of the directionality. The Au–Al${ }_2$O${ }_3$–Au resonator-based emitter exhibits a resonantly enhanced emissivity of 0.8 with a quality factor of 21. The 3(SiO${ }_2$–Si)–SiO${ }_2$–Au emitter exhibits a Q-factor of 85 and an emissivity of 0.82. The 3(SiO${ }_2$–Si)–SiO${ }_2$–Al resonator-based pyroelectric detector exhibits an extremely sharp resonant absorption at the absorption peak of methane (3.3 μm) with an excellent quality factor of 151. The GT resonator-based absorber proposed in this study is lithography-free and can be easily fabricated on large scale surfaces. All these features make the GT resonator-based perfect absorber an outstanding candidate for many IR spectroscopic applications ranging from NDIR gas sensors and multi-wavelength pyrometers to radiative heaters and thermophotovoltaics.