1. Introduction

Recent progress regarding the size reduction and wafer level vacuum packaging (WLVP) for the uncooled microbolometer type infrared (IR) detector technology paved the way for IR technology to be utilized in civilian market by enabling low-cost fabrication of focal plane arrays (FPAs), thus offered a wide range of applications [1–2]. One of the major trends in the market is the pixel size reduction in order to achieve lower cost per FPA and to enhance the image resolution without compromising from the responsivity performance of the sensor [3–4]. However, the reduced pixel size also reveals the issue of reduced absorber area per pixel, resulting in a degradation in responsivity performance of the individual pixels. The issue gets more acute with the typical bolometer topology, where the IR absorber layer and the thermal isolation legs share the same area on the vertical space since a trade-off emerges between the reserved spaces for the thermal isolation legs and IR absorber layer, Fig. 1(a). In literature, this problem is typically solved by employing high fill-factor structures, where the thermal isolation and the IR absorption performances are decoupled to some extent by physically separating the thermal isolation legs apart from the absorber layer on the vertical space, Figs. 1(b), 1(c), and 1(d), [5–10]. There are two prominent solutions for the decoupling of these performances. First solution suggests the thermal isolation legs to take up almost entire pixel area underneath another layer consisting of both the temperature sensitive part and the absorber layer, which is also called as the double-layer (two-level) approach, Fig. 1(b) [5–6]. The other solution suggests only the absorber layer to be spaced apart from the rest of the pixel body, which lacks the interconnect routing reaching up to the uppermost layer of the microbolometer. Thus, the underneath space is mostly used for elongating thermal isolation legs as well as for the inclusion of the temperature sensitive part, which is also known as the umbrella solution, Figs. 1(c) and 1(d) [7–10]. Although both double-layer and the umbrella solution approaches allow independent handling of the absorption and the thermal isolation performances, the former approach requires more complex fabrication since the interconnect routing has to reach up to the temperature sensitive element placed at the uppermost layer in the double layer design. It is also reported that the thermal sensitivity can be further increased by modifying the umbrella body layer aside from employing high fill-factor structures, such as by introducing subwavelength size holes cut through the absorber layer; i.e., perforation, Fig. 1(d). Various studies report such performance enhancement [9–11], which is either attributed to an artificial increase in the sheet resistance (R_s) of the absorber metal, or attributed to the lowering of the thermal time constant due to the reduced thermal mass via perforation.

 

Fig. 1. 3D illustrations of various microbolometer topologies. A typical one-level topology involves the absorber layer, the temperature-sensitive part, interconnect routing, and the thermal isolation legs on the same vertical space, with which either the filling factor is too small or the thermal isolation is too low (a). Isolation legs are vertically spaced apart from the temperature-sensitive part and the absorber layer in order to make more room for longer isolation legs with a typical two-level topology (b). Umbrella solution leaves the interconnect routing in the first level and helps increasing the filling factor in the second level (c), and perforations to the umbrella layer is introduced to further enhance the overall detector performance (d).

Download Full Size | PDF

It is also suggested that the perforation can offer multi-spectral operation capability, such as Visible-LWIR or SWIR-LWIR, without increasing the detector space already reserved for the single-band operation [12]. In another study, it is reported that the vacuum requirement for the packaging is relaxed since the molecular heat transfer ratio is decreased through lowered number of collisions with the reduced interaction volume due to perforation of the umbrella layer [13]. It is also observed that this technique can be used for easing the final release of the microbolometer pixels since the interaction volume of releasing process with the sacrificial layer is increased.

In this paper, subwavelength (SW) size perforated absorbing umbrella structure comprised of a dielectric and a thin absorbing metal layer is examined for almost unity absorption and maximum thermal-mass reduction within the LWIR regime. Unlike the previous studies, this report explains the underlying absorption phenomena of the perforated metal-dielectric stack and how it changes with perforations of different dimensions through exhaustive testing.

2. Design and optimization

Typical absorber topology employed with the microbolometers is based on the quarter-wave impedance matching (QWIM) technique, which exploits the air gap as the quarter wavelength transmission line, which is also utilized for ensuring the thermal isolation. The topology requires a dielectric slab with a specific optical thickness of quarter-wavelength, which is backed with a reflector on one side, terminated with a thin resistive metal layer on the other side, whose sheet resistance (R_s) is matching the free space wave impedance, i.e., 377Ω/□. The thickness of the dielectric layer determining the optical path between the reflector layer and the absorber layer is adjusted to be quarter-wave (QW) in length, ensuring a resonance [14–17]. Within the scope, the QW based absorber is realized with a canonical topology of a suspended perforated umbrella body layer composed of a silicon nitride (SiN_x) layer terminated with a ni-chrome (NiCr) based absorbing metal layer, which is separated from a fully reflective gold (Au) layer through air as the dielectric spacer. Thus, such topology, decoupled from an actual pixel topology, can still represent a typical microbolometer pixel employing an umbrella layer without loss of generality [18–20]. The topology is proposed to examine the conditions for maximizing the absorption while ensuring both the high-fill-factor condition and maximum thermal-mass reduction. Figure 2(a) shows the considered design parameters such as the air gap thickness (t_Air), the body nitride thickness (t_SiN), the sheet resistance (R_s) of the absorbing metal layer, the perforation ratio (w/Λ), the post diameter (d), and the pixel pitch size (p) for two possible selections of unit-cell (UC) for the simulations, namely the perforation UC and the pixel UC. While the perforation UC is only used for examining the effect of the perforation, thus lowering the simulation cost; the pixel UC involves the effects of both the post diameter and the finite pixel pitch size, thus demanding high simulation cost in return for more realistic predictions. Lumerical 3D electromagnetic (EM) solver is used for the numerical simulation of the unit cells defined in Fig. 2(a) with suitable boundary conditions (BCs) applied. Since the metal layer is modeled in a non-dispersive manner, by only utilizing sheet resistance value, in order to ease the computational load, the absorbing metal is modeled with a 2D layer having a certain sheet resistance in the FDTD software. The same layer can also be modeled by generating a resistive layer with a finite thickness and a bulk conductivity corresponding to the desired (modeled) sheet resistance. Periodic boundary conditions (PBCs) are used to lower the simulation volume since the perforations reside in a 2D array format. Incident plane wave is employed for the excitation with the perfect matched layer (PML) BC on one side while the perfect electric conductor (PEC) BC is used to model the reflector on the other side realized by reflective Au layer. Finally, the absorption is calculated by subtracting the reflection from unity incident radiation while ensuring that neither diffraction nor scattering mechanisms are invoked, which is achieved by making period size of the perforations smaller than the operation wavelength (Λ < λ_op); i.e., perforations should be subwavelength (SW) in size, and a suitable diameter is determined for the posts holding the umbrella layer.

 

Fig. 2. 3D illustration of perforated umbrella structure (a) and numerical simulations regarding the average LWIR absorption performances with changing perforation ratios for various sheet resistances (b), for various period sizes (c), also the normalized average LWIR absorption with changing post size (d/p) determining the upper bound for maximum absorption (d).

Download Full Size | PDF

Previously, it is observed that t_Air = 2.3 µm, t_SiN = 200 nm, R_s = 400Ω/□ are the optimum values for achieving a measured average absorption of 96% in the LWIR regime (λ=8-12 µm) without any perforation applied to the umbrella structure [21]; thus, the optimization of the perforation is performed centering around these previously optimized values. Figure 2(b) shows that the perforation neither improves nor degrades the average LWIR absorption up to a cut-off value of w/Λ=0.5 considering the umbrella layer employing a metal layer with the sheet resistance of R_s=400Ω/□, although the R_s is effectively increased, which can be explained through robustness of the topology centering around the optimized values [21]. It is also observed that if an R_s value lower than the optimum value of 400Ω/□ is employed for instance, 100Ω/□, then the cut-off point turns into an optimum point and shifts to a higher w/Λ ratio of around 0.8. This clearly shows that for metal-dielectric stack that is optimized for the region of interest, introducing perforation cannot further improve the absorption performance. Thus, if the plain metal-dielectric stack is, for some reason, not optimized to match the free space impedance, a “pseudo” enhancement is observed with the introduction of perforation of certain sizes since the perforation will effectively make the metal have a higher R_s. In summary, if optimum parameters of the non-perforated case are employed in the first place, then the perforation will not be able to further improve the absorption performance.

Moreover, Fig. 2(c) shows that practical period sizes such as Λ=2.0 µm, 3.0 µm, and 4.0 µm do not change much the characteristic response for the average LWIR absorption versus perforation ratio (w/Λ). The slight discrepancy between responses is due to the difference between the deep SW (Λ << λ_op) and SW (Λ < λ_op) limits since the SW assumption is challenged more as the period size approaches to the operation wavelength. Figure 2(d) shows the practical maximum diameter of the post for a pitch size of p for not disturbing the SW assumption within LWIR regime, which is predicted to be around d/p<0.24 where the normalized LWIR absorption starts to degrade for various perforation sizes such as w/Λ=0, 0.5, and 0.8.

Simulation results suggest that the perforation ratio can be increased by depositing a thicker absorbing metal; i.e., by lowering the sheet resistance, for maximizing the thermal-mass reduction while also maintaining a maximum average LWIR absorption. It is also concluded that the diffraction is not an issue within LWIR regime with the aforementioned period sizes (Fig. S3 and Fig. S4) confirming the SW nature of the perforation. Thus, a physical insight can be obtained through effective medium approaches in order to explain for the effect of the SW size perforation on the absorption performance. In other words, a cascaded transmission line (CTL) approach can be developed such that the perforation based discontinuous surfaces are considered as continuous ones with effective parameters such as effective optical constants and effective sheet resistances as shown in Figs. S1 and S2, whose values are mainly determined by the perforation ratio [22–23]. Thus, it becomes clear that the QWIM technique with effective parameters can still be utilized for predicting the absorption performance of the SW size perforated structures as an analytical alternative for the simulation, which gives more accurate predictions within deep SW regime. In other words, the absorption performance of the SW perforated umbrella structure is still limited with the highest performance of that QWIM can achieve. Thus, a lithography mask is designed with pitch sizes of p=18 µm and 25 µm; periods of Λ=3.0 µm and 4.0 µm; and with various perforation ratios such as w/Λ=0.38, 0.50, 0.63 and 0.75 to examine the predictions mentioned above.

3. Fabrication

Aforementioned topological variations are realized in single batch process of the same fabrication flow using a three-mask process. Firstly, 100 nm gold layer is deposited on Si wafer to ensure full reflection within the LWIR regime. Next, polyimide (PI 2610) based sacrificial layer is spin-coated over the mirror layer to form the air gap of around 2.3 µm. The patterning of the posts holding the umbrella layer is performed with the reactive ion etching (RIE) of the sacrificial layer with the first mask using a positive tone photoresist. Then the umbrella body layer is formed with the deposition of SiN_x via plasma enhanced chemical vapor deposition (PECVD) employing a stress-free deposition recipe. Next, the NiCr absorbing metal layer is deposited using sputtering with thickness values corresponding to the related sheet resistances. NiCr thickness is measured to be around 19 ± 2 nm for R_s=100Ω/□, using a profilometer at several points on the device (Fig. S5). The thickness of the metal for the R_s=400Ω/□ is estimated to be around 5 nm using both sheet resistance relation to thickness and extrapolating the data of deposition time against thickness at constant deposition rate.

The patterning of the NiCr layer is performed through lift-off technique, and it is followed by deposition of a thin SiN_x based passivation layer of around 30 nm thickness to protect the thin metal layer from oxygen plasma employed for the release. The third mask is used to define the perforation and individual umbrella pixels in array format mimicking an FPA format.

Figure 3 shows some of the successfully released, suspended, perforated umbrella type absorber structures with various pitch sizes of p=18 µm and 25 µm; and perforation ratios of w/Λ=0.91, 0.55, 0.38, and 0.25, which are different than the designed ones to some extent due to the fabrication related imperfections.

 

Fig. 3. Scanning electron microscope (SEM) images of the successfully released perforated umbrella type absorbers with various perforation ratios. Top view images of perforated cases with a pitch size of p=25 µm, period size of Λ=3.0 µm, and perforation ratios of w/Λ=0.91 (a), w/Λ=0.55 (b), w/Λ=0.38 (c); lastly the cross-sectional view for a perforated absorber body with w/Λ=0.25 and a pitch size of 18 µm.

Download Full Size | PDF

4. Discussion

Spectral IR absorption is obtained through FTIR based reflection measurements by subtracting the normalized reflected radiation from unity within the range of (2.5 µm-20 µm), where the diffraction and scattering is eliminated by design for the LWIR sub-range. Figure 4 shows spectral IR measurement results in comparison with various simulation techniques and topologies for an air gap of around t_Air=2.3 µm, a body silicon nitride of t_SiN=200 nm, a period size of Λ=3.0 µm, sheet resistances of both R_s=400Ω/□ and R_s=100Ω/□, a post diameter of d=2.6 µm, and perforation ratios of w/Λ=0.40, 0.77, and 0.90, which are partly different than the designed ones due to fabrication imperfections.

 

Fig. 4. A comparison of both numerical (Perforation UC, Pixel UC) and analytical (CTL) simulations with the FTIR measurement results for spectral absorption performances of various umbrella type absorbers changing in sheet resistances, and perforation ratios. The pitch size is p=25 µm while the period size is Λ=3.0 µm for all cases.

Download Full Size | PDF

The feasibility of the cascaded transmission line (CTL) technique is also examined as an analytical alternative for the simulation, which is incorporated with the effective medium approach (EMA) to explain for the effect of the SW discontinuities caused by perforations. It is shown that CTL not only predicts the field intensity successfully but it can be also used for providing a physical insight for the obtained results. The discrepancy between the simulations using entire pixel and the perforation UC is mainly due to the existence of the post holding the umbrella structure and the finite pitch size. The difference between the analytical CTL and the numerical solutions of pixel UC and perforation UC is mainly due to the fact that CTL is a one-dimensional analytical method that does not take geometry into account, and assumes perfect continuous surfaces. Moreover, due to its no-geometry nature CTL assumes deep SW regime, which is not the actual case for the entire range of (2.5 µm-20 µm). The discrepancy between various simulation techniques and the FTIR measurement result concentrates on the lower wavelength regime since the diffraction effect is more pronounced within this regime and the FTIR tool has a finite aperture; on the other hand, the discrepancy for higher wavelength regime is mostly due to fabrication imperfections. Although the percentage root-mean-square (RMS) error between the measurements and the simulations can be as high as 6.9%, it should still be considered within acceptable range since the qualitative predictions regarding the variation of the spectral absorption performance for various perforation ratios and for different sheet resistances are accurate.

Figure 5 summarizes other key findings of this study, which compares measured average LWIR absorption results with the simulation based predictions with changing perforation ratios for both sheet resistances of R_s=400Ω/□ and Rs=100Ω/□, for different pitch sizes of p=18 µm and p=25 µm; and for different period sizes of Λ=3.0 and Λ=4.0 µm. It is concluded that for an optimized absorber topology, satisfying the QWIM conditions already for non-perforated case, the SW perforation is not able to further enhance the absorption performance, but it can be kept around 96% up to a cut-off perforation ratio of 50% (w/Λ=0.5) for the umbrella configuration of t_Air=2.3 µm, t_SiN=200 nm, and R_s=400Ω/□. In other words, the thermal-mass reduction is still possible without degrading the absorption performance for the already optimized non-perforated configuration, which demands a sheet resistance of around R_s=377Ω/□. Nevertheless, the thermal-mass reduction can be further increased above 50% by lowering the sheet resistance of the absorber metal; i.e., by depositing a thicker absorbing metal layer. Figure 2(b) shows the case where an average LWIR absorption of 93% is achieved with a perforation ratio of 77% by decreasing the sheet resistance down to 100Ω/□; i.e., by depositing a thicker NiCr layer. Deposition of a thicker layer degrades the maximum available performance for a non-perforated case that can be achieved by the QWIM technique, which is measured to be around 65% since the QWIM condition for the termination impedance is disturbed. Nevertheless, perforating the absorber layer with subwavelength features restores back the maximum performance since the sheet resistance of the thicker layer is effectively increased and restored back to the optimum impedance value of around 377Ω/□. As the simulation and measurement trends, summarized in Fig. 5, are in accord; the deviations are due to fabrication imperfections that could not be included in the analytical or numerical modeling, especially for higher perforation ratios, where the edge rounding is more prominent and NiCr layer is more prone to oxygen plasma damage due to higher surface area exposed.

 

Fig. 5. A comparison of numerical simulations with the measurements examining the effect of the perforation ratio on the average LWIR absorption performance for various pitch sizes of p=18 µm and p=25 µm; period sizes of Λ=3.0 µm and Λ=4.0 µm; and sheet resistances of R_s=400Ω/□ (a) and R_s=100Ω/□ (b).

Download Full Size | PDF

5. Conclusion

A detailed analysis is performed regarding the effect of the subwavelength perforation of the umbrella type absorbers, employing QWIM technique, for achieving high absorption performance within LWIR regime. Perforations of changing pitch sizes, perforation ratios, and period sizes are considered in the analysis to convey the trend of a generic impedance matched metal-dielectric slab in free space, which can be integrated into any optical device in the wavelength region of interest

Furthermore, it is demonstrated that thermal-mass reduction can be maximized with larger perforation ratios through deposition of a lower sheet resistance (R_s < 377Ω/□) absorbing layer without sacrificing from the optical performance. Both numerical and simplified analytical (CTL) modeling are presented that agree with the measured results. Finally, it is reported that, a grid-like absorber structure is formed for high performance absorption, which can be also utilized for multi-wavelength operation such that highly perforated structure becomes invisible for lower operation wavelengths. From microfabrication point of view, grid-like structures also ease the release of the sacrificial layer deposited underneath; thus, also having the potential to speed up the batch process line.