There is rich phenomenology in the terahertz (300 GHz–3 THz), longwave (8–12 μm), and midwave (3–5 μm) infrared bands, which enables the detection and imaging of chemical, biological, and environmental phenomena [1–3]. Some of these applications include remote sensing of the atmosphere and pollutants, detection and classification of biological materials, and diagnosis of medical conditions. Most practical commercial detectors in these bands lack spectral selectivity, being responsive over broad bands of the electromagnetic spectrum. Those that function at room temperature are thermal, and thus have poor spectral resolving power ($R=\lambda /\mathrm{\Delta }\lambda$), typically less than 3 [4]. When applications require spectral selectivity, components must be added before the detector—such as fixed filters or interferometers—increasing the cost and complicating the device.

The ideal hyperspectral imager or spectral detector is uncooled while having high spatial resolution and resolving power. It should have no moving parts, allowing for high-speed and compact operation. No current technology can meet all these goals; for example, filter wheel cameras have low operational speed and are limited to a small number of preset spectral bands. Imagers utilizing dispersive elements, such as gratings, sacrifice one dimension of the detector array for spectral sensing, requiring that they be mechanically push broom scanned over the target, while interferometers require a highly precise mechanical scanner coupled with wide detector readout bandwidth [5,6].

We propose and implement an infrared thermal detector where the detecting element itself (a pyroelectric material) is an integral part of a metamaterial. Pyroelectric detectors are a promising technology because they are thermal, operate over a huge range of the electromagnetic spectrum, and are uncooled [7]. Unlike microbolometers [8], the classic pyroelectric detector utilizes a bulk unpatterned planar element, which significantly reduces fabrication costs, and the need for hermetic packaging is eliminated. However, traditional broadband pyroelectric detectors utilize carbon or metal nanoparticles to absorb incident radiation, which is then conducted as heat to the pyroelectric capacitor [9]. Instead, our device utilizes a novel configuration where the pyroelectric element serves as the dielectric layer of a metamaterial detector (MMD). The detector exploits the multifunctional ability of metamaterial absorbers [10–12], that is, the metallic portions serve as the electrodes for the pyroelectric effect while ensuring that absorbed energy is dissipated preferentially in the dielectric spacer element and the composite structure achieves near unity absorption of a narrow band of electromagnetic radiation.

We have designed, simulated, and fabricated a series of fast, narrowband, metamaterial absorber-based detectors spanning a portion of the thermal infrared band of 8–11 μm. A typical design has a center wavelength of 10.73 μm and a full width at half maximum (FWHM) of 557 nm. An SEM image of the aforementioned MMD is shown in Fig. 1(a), and a schematic of the unit cell is shown in Fig. 1(b), which consists of a gold structure patterned directly on top of a 575 nm thick layer of lithium niobate [13] followed by a 100 nm continuous gold ground plane and a further 500 μm ${\mathrm{LiNbO}}_3$ substrate. The MMD unit cell is a split cross resonator, with symmetry that provides a polarization independent response. The detector active area—shown in Fig. 1(c)—is square with a side length of 150 μm and consists of 780 metamaterial unit cells.

 

Fig. 1. (a) Electron micrograph of metamaterial top surface, (b) schematic of unit cell for 10.73 μm detector, and (c) micrograph of detector pixel consisting of $28\times 28$ unit cells.

Download Full Size | PDF

The spectral response of the fabricated devices was characterized by measuring the optical reflectivity with a Fourier transform infrared (FTIR) spectrometer (Bruker Vertex 80v) and an infrared microscope (Hyperion 2000) with a $36\times$ 0.5 NA objective lens. To find absorbance, the transmission was assumed to be negligible due to the gold ground plane. To contrast the performance enabled by the metamaterial architecture, we also fabricated and measured a reference lithium niobate detector, which consisted of the same grid without the 1.9 μm arms. In Fig. 2(a), we show the simulated (black curve) and experimental (red curve) absorbance of the MMD as a function of wavelength. Good correspondence between simulated and experimental absorbance data is evident. Also shown are four other metamaterial detectors, Fig. 2(b), demonstrating that our design may be scaled across much of the thermal infrared through modification of the top metal geometry (see Supplement 1), while using the same pyroelectric thickness throughout. We find peak absorbance values of 85.1% at 8.30 μm, 85.8% at 8.72 μm, 90.8% at 9.82 μm, and 91.0% at 10.13 μm, with corresponding FWHM values of 820 nm, 787 nm, 558 nm, and 554 nm, respectively.

 

Fig. 2. (a) Simulated (black curve) and measured (red curve) optical absorbance spectra of a typical metamaterial detector structure. (b) Measured optical absorption spectra of several MMDs. (c) Detector response (blue curve) of MMD shows good correspondence with the measured optical absorbance (red curve). Also shown is the detector response of a non-resonant grid (green curve), showing that metamaterial resonance is necessary to absorb energy into the pyroelectric element.

Download Full Size | PDF

Detector performance was characterized with the same spectrometer, where the signal output of the MMD was connected to a high-impedance voltage pre-amplifier ($Z=100\mathrm{M}\mathrm{\Omega }$, SRS model SR560) and then fed into the spectrometer analog-to-digital input. The MMD spectral voltage response is shown in Fig. 2(c) (blue curve), and it realizes a good match with the optical absorption (red curve) replotted from Fig. 2(a). Also shown in Fig. 2(c) is the response for the reference detector (green curve). As can be observed, without the metamaterial pattern, there is no appreciable detector response in this range of wavelengths, that is, the green curve is equal to the noise floor of our setup.

The 10.73 μm MMD was further characterized by measuring the responsivity and noise-equivalent power (NEP) as a function of modulation frequency. The interferometer mirror speed was varied, so that the modulation frequency of the response peak at 10.73 μm varied from 94 Hz to 2.34 kHz, and, by using a thermal power meter, the absolute power was calibrated (see Supplement 1). We determined detector noise statistics by acquiring ten two-minute scans and averaging. Assuming Gaussian noise, NEP is equal to the variance between scans.

The responsivity (dark gray curve) and noise equivalent power (blue curve) versus modulation speed at the peak absorption wavelength (10.73 μm) are shown in Fig. 3(a) and demonstrate peak performance comparable to other commercial small detectors. It is important to note that we calculate the MMD responsivity using the voltage at the detector element, before any amplification is applied. We observe that as modulation speed increases, NEP drops, as is typical for pyroelectric detectors. There are a variety of frequency-dependent noise sources, such as $1/f$ noise, which can originate from the high-impedance amplifier, spectrometer electronics, and detector materials and contacts.

 

Fig. 3. (a) Measured responsivity and NEP versus modulation frequency. Voltage is referenced at the element (before amplifier). Bars on responsivity show standard deviation. (b) Specific detectivity of the metamaterial detector.

Download Full Size | PDF

Thermal simulations of the MMD, using the loss pattern from electromagnetic simulations (both assuming an infinite array and energy at normal incidence), yield a thermal time constant of 34.6 Hz (${\tau }_t=28.9\mathrm{ms}$) determined centered underneath a metal arm. Using a precision LCR meter, we determined the RC time constant of the pixel was 7.56 kHz (${\tau }_{\mathrm{RC}}=132.3\mathrm{\mu s}$) including the capacitance of the bond pads and wiring, which were of a substantially larger area than the detector itself (Fig. 1). Using the difference in metal area between this device and a larger detector ($\text{750}\times 750\mathrm{\mu m}$), we extracted the wiring parasitic capacitance to find the detector-only capacitance, corresponding to an RC time constant of 97.0 kHz (10.3 μs). Over the measured modulation range, from 100 Hz to 2 kHz, there is an approximate 3 dB/decade decrease in responsivity. This correlates well with the nearly flat response expected in the region between the thermal and electrical (RC) poles [14]. While both responsivity and NEP decrease with modulation frequency, specific detectivity ($D^{*}$), a time- and area-normalized measure of signal-to-noise ratio, shows an increase in performance through the measured range [Fig. 3(b)], reaching at least $9.59\times {10}^6\mathrm{cm}\sqrt{\mathrm{Hz}}/\mathrm{W}$, which should peak just outside the measured range.

Although the metamaterial detector presented here is a thermal device, the fast response suggests that it is usable for a scanned imaging or detection system. The resonant performance of the MMD exceeds commercial detectors. In particular, the OPHIR Pyrocam IV [15], a pyroelectric imaging array, has 75 μm square pixels and is specified for a $13\mathrm{nW}/\sqrt{\mathrm{Hz}}$ NEP; this is an order of magnitude worse than our demonstrated device. There are tradeoffs, such as thermal conductivity and amplifier impedance, which decrease the performance of conventional discrete pyroelectric detectors as speeds increase. The Gentec-EO QS-IF series exhibits a comparable detectivity around ${10}^6\mathrm{cm}\sqrt{\mathrm{Hz}}/\mathrm{W}$ and a thermal frequency of 3.5 Hz [16]. The performance of our MMD can potentially be improved further by directly wire-bonding the detector to a transimpedance amplifier, as is found in commercial pyroelectric detectors.

Our MMD design has a fundamental advantage over other thermal detectors because optical energy is directly converted to heat inside the pyroelectric material, that is, the detection element is part of the metamaterial. Without the MMD architecture, it would be highly inefficient due to the index mismatch between lithium niobate and free space, combined with the relatively low absorption of the material. However, the metamaterial perfect absorber design results in a resonance with unity theoretical absorption by the structure. It is designed such that $\epsilon$ and $\mu$ at resonance has a high effective loss (${\epsilon }^{\prime \prime }$), while maintaining a perfect match to free space ($\sqrt{{\mu }_r{\epsilon }_r}=1$). Dimension errors in fabrication and differences in assumed material properties reduce this absorbance, but with iterative adjustments to the structure, absorption in excess of 99% can be achieved [12].

The multifunctional property of metamaterials exploits the same structure for both optical and electrical response. We have optimized our MMD such that the DC field, which forms the pyroelectric capacitor, significantly overlaps the optical absorption and subsequent thermal response. Figure 4 depicts simulations of the optical power absorption (left column), the resultant temperature distribution (middle column), and the DC field between the top metal layer and the ground plane at resonance (right column). Since the structure of our MMD is sub-wavelength, there is no direct correlation between metal coverage and absorption efficiency; a large amount of optical to thermal conversion occurs underneath the metal. In contrast, in a conventional pyroelectric detector design, metal contacts reflect incident optical energy, and, therefore, the absorber layer must be thick enough to convert optical energy to heat before it is reflected back into free space. The heat then must be conducted into the pyroelectric capacitor, which results in a slow device thermal time constant, slowed proportionally to the amount of added thermal mass.

 

Fig. 4. Simulated optical power absorption density, temperature, and DC electric field through the MMD structure at resonance. One-half watt of vertically polarized IR energy is applied to the unit cell. The multifunctional nature of the metamaterial enables the DC field to directly overlap much of the thermal excitation.

Download Full Size | PDF

At the resonant wavelength, simulations indicated that 82.5% of total incident energy was absorbed in the lithium niobate, while the remaining 9% and 8% were dissipated in the top metal and bottom ground plane, respectively. Remarkably, only 0.5% of incident energy is reflected. This translates into an effective absorption coefficient of ${\alpha }_{\mathrm{eff}}=15,156{\mathrm{cm}}^{-1}$ in the dielectric, a factor of 8.2 times greater than the $1,840{\mathrm{cm}}^{-1}$ we extracted for the bulk lithium niobate (comparable to the literature value of $1,000{\mathrm{cm}}^{-1}$ [17]).

At typical detector modulation frequencies, noise in the pyroelectric element is primarily caused by temperature fluctuations and loss tangent noise [18]. Temperature fluctuations, also called phonon noise, are due to the quantized radiation of heat exchange between the detector and the background, and they limit the NEP and detectivity performance of a pyroelectric detector. It can be shown (see Supplement 1) that the fundamental limiting NEP of a pyroelectric detector is

While the analysis is straightforward for traditional bulk infrared pyroelectric detectors, there is a subtle, yet significant advantage afforded with the metamaterial absorbers used in our design. One difference is that in the MMD, the effective optical area, which matches the physical size of the metamaterial array, is different from the detector electrical area (the area encompassed by the capacitor fields), which also varies from the effective thermal area (the area with significant optical power absorption). Intense field concentration and thermal conversion in the active layer is not necessarily present in all metamaterial and plasmonic structures, such as hole arrays [19], and must be carefully designed. Absorbers are a good class of metamaterials since they may be designed to absorb all the incident energy in a subwavelength structure and largely in the dielectric spacer layer itself. The metamaterial absorber design creates field concentration within the pyroelectric material, and thus, the effective thermal radiative area ($A_{\mathrm{eff}}$) is smaller than the physical area ($A_{\mathrm{phys}}$), which by Eq. (1) leads to a decrease in limiting NEP by a factor of $\sqrt{A_{\mathrm{eff}}/A_{\mathrm{phys}}}$. This factor can be significant—simulations show that 90% of the absorbed optical power in the pyroelectric element occurs in only 2.4% of the volume. There is thermal conduction away from these absorption points, but the volume that is heated to at least 50% of the peak temperature is still concentrated to 14.0% of the total. Assuming a thin device where temperature distribution is invariant in the vertical axis, this means that the NEP contribution from thermal fluctuations can be 2.67 times better than a conventional planar absorber.

The electrically continuous, yet physically small, nature of the metamaterial reduces detector capacitance, which reduces both the RC time constant and loss tangent noise (Johnson noise caused by the equivalent series resistance). This noise is significant at frequencies above 100 Hz and scales proportionally to $\sqrt{C}/f[\mathrm{A}/\sqrt{\mathrm{Hz}}]$ [20]. For the 10.73 μm MMD previously presented, the metal coverage is only 40.8%, and neglecting fringing fields, the capacitance is reduced by the same fraction compared to a continuous metal electrode.

We next highlight another major benefit of the metamaterial detector design paradigm. Spectral response is largely dependent on the metal structure, so the resonant wavelength can be shifted by modifying the dimensions of the top pattern. This is not possible in interference-based absorbers, including Woltersdorff metal films, which necessitate changes in the thickness of the substrates [21] and thus, if multiple resonant frequencies are required within one device, complex fabrication procedures are required. We demonstrated this tunability by fabricating multiple detectors that resonated from 8.3 to 10.7 μm wavelengths on a single thickness substrate. The unit cell size varied from 3.5 to 7.5 μm and the width of the metal cross was either 650 or 750 nm, all within the capability of standard lithography techniques. Each detector was the same size, $150\times 150\mathrm{\mu m}$. Figure 2(b) shows that the high optical absorbance and narrowband response is maintained over the entire spectral range. This supports the assertion that the split cross metamaterial absorber structure only shows a minor dependence on substrate thickness, which also enables robustness against process variations. The ability to span the entire thermal IR band simply by changing the metal pattern provides the core of a low-cost high-performance hyperspectral detector.

The multifunctional capabilities of MMDs permit integration of optical, thermal, and electrical functionality into a single structure. We found that there are many advantages to this approach when compared to traditional detectors. First, the detector itself is spectrally selective but can be tuned by simply changing the dimensions of the top metal structure, eliminating external filter wheels or interferometers in a system and allowing for simple fabrication in an array. The metamaterial exhibits field concentration: the absorbed optical energy is converted to heat in a concentrated region, which is also effectively overlapped by the DC field between the electrodes. Not only is the thermal time constant improved while preserving responsivity, but we show that this reduction in effective thermal area improves the limiting phonon noise of the detector. Finally, the metamaterial structure only covers a fraction of the area of the pyroelectric material, reducing detector capacitance. This improves not only the electrical time constant, but the reduction in capacitance and the higher detector operating frequency reduces detector loss tangent noise and $1/f$ noise from readout electronics. With these benefits combined, the improved responsivity, detector speed, and reduced noise exceed the capabilities of conventional pyroelectric detectors.

We have designed, fabricated, and characterized a narrowband metamaterial detector based on a resonant perfect absorber, with theoretical unity absorption. The multifunctional metastructure uses a metal layer and ground plane to efficiently absorb optical energy from free space directly into the pyroelectric element while simultaneously serving as electrodes to effectively capture the generated electric field. Because the multifunctional metamaterial design does not require heat transport from a separate absorber, the detector is extremely fast—with a detectivity peak greater than 2 kHz—and shows excellent responsivity and NEP, without the use of a membrane or insulating bridge. An important highlight to this performance is that the metamaterial absorber creates an intense optical field and thermal concentration, reducing the effective thermal radiation area, and improving the thermal fluctuation limiting NEP. The electrodes are not continuous planes and thus capacitance is reduced, improving the RC time constant, and decreasing loss tangent noise. The latter noise source dominates at modulation speeds greater than 100 Hz, where our device can operate.

The thermal speed of these detectors can be improved with substrate thinning and membrane techniques, as we found that most of the heat flow was through the relatively thick substrate. Additionally, a pre-amplifier mounted directly beside the detector will reduce electrical noise. We have shown that the split cross resonator metamaterial enables broad tunability while maintaining the spectral resolution and absorption efficiency by simply changing the design of the top metal layer. Arrays of multicolor pixels can be created, forming a hyperspectral imager.