1. Introduction

Optical gas sensing (OGS) has a key role in several disciplines and applications [1–6]. To name a few, in the gas and oil industries, OGS is used in monitoring hazardous gas leaks, and controlling production processes to improve their efficiency; in environmental science and engineering – for watching and tracking air pollution and greenhouse gases; in earth science – to measure climate changes and atmospheric composition; in the engine manufacturing industry – to analyze combustion gases; in military and homeland security – for remote chemical-imaging; and in medicine – for breath analysis. Optical gas imaging (OGI) devices, together with image processing algorithms, enable gas visualization, and supply spatial characteristics of the gas plume. OGI’s pixel analysis is generally based on detection of specific absorption or emission bands in the infrared (IR) spectral region that characterize gas molecule vibrational–rotational modes. OGI methods can be divided into dispersive, such as grating-based hyperspectral imaging [7], Fourier transform infrared spectroscopy (FTIR) [8, 9], and non-dispersive [10], which utilize one or several predefined band-pass (BP) filters to identify the gas by its IR signature.

Dispersive methods may achieve high spectral resolution and accuracy, but require scanning, which thwarts the possibility of acquiring fast phenomena. This tradeoff between acquisition rate and resolution typically leads to low spatial resolution relative to broadband imagers. Furthermore, the costs of these imaging spectral-radiometers or hyperspectral cameras are very high, making them unsuitable for many applications. Non-dispersive methods can be divided into mono-, bi-, and multispectral band-based acquisition. In monospectral band acquisition, one BP filter is selected to match the most significant absorption or emission band of the gas, named the in-band filter. Such a method may lead to ambiguity between the target gas and the surrounding objects in the field of view, which display a flat spectral response in the band. Such objects are called clutter. The ambiguity mandates human interpretation of the image, which limits its practicality. Common bispectral sensors have one additional BP filter that matches a transparent band of the gas, and is called the out-band filter. Capturing the same scene with each filter individually eliminates the ambiguity almost completely and allows automatic interpretation by comparing the two bands. The bispectral filter-changing mechanism, however, consumes either time—thus limiting the possibility of acquiring fast phenomena—or the detection area, meaning that it decreases the spatial resolution or field of view. Furthermore, multispectral sensors, though increasing gas detection and identification confidence [11], are still suffering from the significant acquisition-time or spatial-resolution limitations mentioned above. In this paper, we demonstrate an experimental setup of a fast and electrically switchable bispectral band-stop, i.e., a notch filter (NF) based on a liquid crystal (LC) layer inside an IR transparent cell.

2. Simulation and experimental results

In the following simulation, we analyze the signals detected by two pixels of an OGI system. We show the monospectral ambiguity, and then illustrate how the switchable NF discriminates between gas and clutter signals. Figure 1(a1) depicts an OGI setup with a single BP filter. Figure 1(b) illustrates the spectral transmittance of the BP filter (green dashed line), ${\tau }_{BP}\left(\lambda \right)$, which in this simulation has a Gaussian shape with a central wavelength (CWL) at 8µm and a full-width-half-max (FWHM) of 250nm. The purple dotted line in Fig. 1(b) shows the gas plume in-band transmission,${\tau }_{GAS}\left(\lambda \right)$ with its significant absorption line, while the cyan solid line shows the clutter in-band spectrally constant transmittance, ${\tau }_{CLUTTER}$. The incoming radiation from a clutter object is given by$P_T\left(\lambda \right){\tau }_{CLUTTER}$, where$P_T\left(\lambda \right)$is the background’s radiation at temperature$T$, and ${\tau }_{CLUTTER}$ represents the open path attenuation together with the background's intensity variation. $P_T\left(\lambda \right){\tau }_{GAS}$ is the incoming radiation from the target gas including the open path attenuation. We assume that the gas reaches thermal equilibrium with the ambient, and there is a sufficient temperature difference between the gas (and the ambient) and background. Without loss of generality, we analyze the case where the background is hotter than gas.

 

Fig. 1 Simulation of two gas detection setups: (a1) Using a single BP filter, and (a2) adding a switchable NF. (b) Example of gas, clutter, and BP transmissions separately using the a1 setup. (c) Transmissions of the NF (a2 setup) at ‘on’ and ‘off’ states aligned to gas resonance. (d) Total transmission in ‘on’ and ‘off’ states for the NF a2 setup.

Download Full Size | PDF

Looking at two different, $S_{CLUTTER}$, and$S_{GAS}$, representing the signals of a gas pixel and a clutter pixel acquired simultaneously from the same scene. The signals are formulated by:

 

Fig. 2 A setup consisting of a switchable NF and BP acquire gas and clutter signals (a). Spectral transmissions of the objects in the clutters and the system components (b). The 'on' to 'off' signal ratio $R_{CLUTTER}$is constant for any spectrally flat radiation, and differs for the gas,$R_{GAS}$. This allows us to distinguish between gas and clutter pixels.

Download Full Size | PDF

The LC-based NF utilizes two fundamental properties of the LC: i. The absorption strength, which depends on the optical-axis orientation of the LC molecule relative to the polarization of the incident light: the ordinary state for orientation parallel to the light polarization vs. the extraordinary state for the perpendicular one; ii. The fact that its molecular absorption bands in the thermal IR spectral region overlap those of most hydrocarbon gases (and others). When voltage is applied to the LC cell, the orientation of the molecules switches between ordinary and extraordinary, which significantly changes the transmittance at the absorption lines. Two perpendicular cells modulate the transmittance of unpolarized light without the losses associated with the traditional use of cross-polarizers. Applying alternating voltage may result in a high temporal-frequency transmittance modulation, in the order of 100Hz, which enables the detection of dynamic scenes such as gas plumes. This concept may be useful not only for OGS, but also for detection of liquid and solid objects containing hydrocarbons, for example, in the fields of hazardous liquid detection, and geosciences.

Before describing the experiment, it is important to note that LC substrates in the nematic phase are wavelength-dependent birefringent [12,13], i.e., their complex refractive index,$\tilde{n}\left(\lambda \right)= n\left(\lambda \right)+i\kappa \left(\lambda \right),$ where $n$ is the real part of the refractive index, κ is the extinction coefficient, and $\lambda$ is the wavelength, changes its value from $\widetilde{n_o}\left(\lambda \right) ,$ for an ordinary state, to $\widetilde{n_e}\left(\lambda \right)$, for an extraordinary state. The real part represents the birefringence, $\text{Δn}\left(\lambda \right)= n_o\left(\lambda \right)-n_e\left(\lambda \right)$, which is often used in LC displays and has either a positive or negative value. In addition, LC molecules exhibit several absorption bands in the IR spectral region, due to carbon–carbon (C-C, C = C), carbon–hydrogen (C-H), and carbon–nitrogen (C≡N) stretching and deformation resonances. $\kappa \left(\lambda \right)$ is small outside these bands, but within them, is significant [12,13]. For elongated molecules, such as LCs, these bonds induce anisotropic absorption, with respect to the molecule orientation, where the ordinary extinction coefficient, $k_o\left(\lambda \right)$, differs from the extraordinary one, $k_e\left(\lambda \right)$, by $\text{Δk}\left(\lambda \right)$, which as the real part, may have a positive or negative value and can be wavelength dependent.

As depicted in Fig. 3(a), the experimental setup includes a 1cm X 1cm Germanium LC-cell (Instec) with a 50µ cell-gap, filled with E7, a multicomponent LC mixture (Merck) with 5CB, depicted in Fig. 3(c), as its major component (51%). Optical properties of E7 and 5CB are described in [13], and the composition of the E7 mixture and its components' chemical structures are described in [14]. The inner sides of the Ge plates were coated with polyamide (PL) to align the E7 molecules such that their principal symmetry axis direction is parallel to the cell sides, the Y-axis in Fig. 3(a). A ZnSe wire grid polarizer (Edmund) was set in front of a cavity blackbody (CI-Systems) at 1200°C such that the radiation propagated in the X direction, and was polarized in the Y direction. To rotate the LC molecules from an ordinary state (Y-axis) to an extraordinary one (X-axis), the cell was connected to a switchable 20V DC voltage source (Keithley 2400). The graphs in Fig. 3(b) demonstrate the measured spectral transmission of the E7-filled Ge NF in the ‘on’ (dot-dashed red) and ‘off’ (dashed black) states, and the transmittance of the cell without the E7 (empty cell, solid blue line). The table indicates the LC’s absorption bands, A–F, associated with the vibrational stretching and deformation resonances in the mid- and long-wave infrared range. Figure 3(d) and its inset show the measured rise time from the ‘off’ state to the ‘on’ state to be ~20 milliseconds while the discharging time is ~10 seconds. As can be seen in the transmission plots, the A and F absorption bands exhibit negative $\text{Δk}$behavior, while the B–E bands exhibit a positive$\text{Δk}$, consistent with [13]. Looking at vapor phase infrared spectrum library [15, 16], we found that the A–F bands significantly overlap with the absorption and emission bands of many materials with similar bonds and resonances, especially hydrocarbons. For example: methane has bands at A and D; propanol has bands at A, and E; freon-123A has bands at A, D, E, and F.

 

Fig. 3 The experimental setup and results. (a) The setup for cell transmittance measurements. (b) The uniform spectral-response of an empty cell (solid blue line) compared to the E7LC mixture in the ‘on’ state (red dot-dashed line) and ‘off’ state (dashed black line). The E7 absorption lines are associated with the vibration modes of the LC [13] (see table in the figure). (c) The chemical structure of LC 5CB – the major component of the E7 mixture, adapted from [14]. Time response measurements of the E7-filled cell (d) show a fast rise time of 20milliseconds after applying the voltage (arrow-ON), and slow fall time of 10 seconds after switching off the voltage (arrow-OFF).

Download Full Size | PDF

In the experiment, we used an E7 LC to detect 1,1,1-Trifluoroethane (R143a), a refrigerant gas, which has a high global warming potential and thus needs leak-monitoring and tracking [17,18]. We used a 10cm-long gas-cell (Pike Technologies) with infrared transparent ZnSe windows, and a polarizer. The Ge LC-cell was placed inside the radiometer between the fore optics and the detector. We measured the radiation intensities collected by the spectroradiometer (SR5000N, CI-Systems, spectral resolution < 2% of the wavelength) when the gas-cell contained air only, and when it was filled with 2000 ppm-m R134a in the ‘off’ state,$S_{OFF}\left(\lambda \right)$, and after applying 20V, in the ‘on’ state, $S_{ON}\left(\lambda \right)$. Figure 4(a) shows the measured transmittances of the E7 when in the ‘on’ (dot-dashed red line) and ‘off’ (dashed black line) states, compared to 2000 ppm-m for R134a gas (pink dotted line). The overlap of the E7 and R134a absorption bands is optimal in the spectral region centered at a central wavelength (CWL) of 8.5μ (solid green vertical line), where a positive $\text{Δk}$ is significant. For this reason, the entire spectral band of interest is between 8.2μ–8.8μ. To optimize the detectability, it is important to match the FWHM of the LC resonance line, which is about 150nm, to the BP filter. In addition the CWL of the BP-filter must be optimized for maximal clutter to gas contrast, defined as:

 

Fig. 4 (a) Transmittance measurements of R134a gas (dotted pink) compared to E7 transmission in the ‘on’ (red dot-dashed) and ‘off’ states (dashed black). (b) Detection of R134a gas by comparing $R_{GAS }$ (dashed black) to air’s $R_{CLUTTER}$ (red dotted), and the contrast (blue solid) for several CWL of BP filters with FWHM = 170nm.

Download Full Size | PDF

To do this, a circular variable filter (CVF) with a series of BP filters, each with an FWHM of ~150nm around different CWLs, is added in front of the detector. The plots in Fig. 4(b) show the ‘on’ to ‘off’ state ratios: $R_{GAS}$ for R134a (dashed black) and $R_C$ for lab’s air (dotted red), respectively. At 8.5µm $R_{GAS}=1.18$ and$R_C=1.6$. Also, the$CONTRAST$is plotted as a function of the BP CWL (solid blue). As can be seen, the optimal CWL is at 8.5µm where$CONTRAST>0.3$.

3. Discussion

This technique can be further optimized by the following: (1) Polarizer elimination – To handle unpolarized radiation with minimal transmission losses due to the polarizer, the setup should be modified to have two identical adjacent cells in a series such that the LC-molecules’ orientation in each cell is anchored perpendicular both to the light direction, and one to the other. (2) Anti-reflective (AR) coating – Ge’s refractive index in the IR is about 4 [19]; thus, the transmittance of an empty cell, having four interfaces, is low, and cannot exceed the theoretical threshold ~0.17. An AR coating should improve the transmittance dramatically and reduce the temperature difference between the background and the gas needed for detection. (3) Switching-time – To minimize the switching time, especially the decay time, a quasi-static and high voltage DC pulse [20,21] should be applied, and also using a scattering-free polymer LC network would be effective [22]. (4) LC and hydrocarbon–gaseous absorption-band overlap – Though hydrocarbons and LCs have quite similar resonances, for several of these bands, the overlap is suboptimal due to the influence of the intermolecular bonds that exist in liquid only, and result in a shift in the CWL of the liquid relative to gas. Tuning the resonance of the LC could be achieved by adding solvents with different polarizability to the LC.