1.

Introduction

Carbon capture and storage have been identified as effective means to control global warming.¹ ${\mathrm{CO}}_2$ sequestration is newly developed carbon storage approach, through which ${\mathrm{CO}}_2$ is pumped into the ground for long-term storage. Real-time large-range ${\mathrm{CO}}_2$ monitoring on ground and underground helps to estimate the ${\mathrm{CO}}_2$ movement and leakage, and is essential for ${\mathrm{CO}}_2$ sequestration.²

Traditional electrochemical or semiconductors ${\mathrm{CO}}_2$ sensors are cross sensitive to other chemical species, and cannot work under high temperature and high humidity, hence are not appropriate for ${\mathrm{CO}}_2$ sequestration applications.^3,4 Laser spectroscopic sensors measure the gas-specific molecular vibration energy level and are thus crosstalk free. If the laser is delivered via optical fiber, the totally passive configuration of the sensing probe ensures that it can work under critical conductions. ${\mathrm{CO}}_2$ has a strong absorption at middle infrared range, however, the attenuation of middle infrared light in optical fiber is very high and only be delivered to hundred of meters away.⁵ In our system we selected to use 1.57 μm light, at which ${\mathrm{CO}}_2$ has a weak unique absorption, and the attenuation of such light in silica optical fiber is $0.2\mathrm{dB}/\mathrm{km}$ only. The absorption strength around 1.57 μm is so weak that the interaction length of the light and ${\mathrm{CO}}_2$ is usually of dozens of meters or even longer, which is not practical for ${\mathrm{CO}}_2$ sequestration applications. In this letter we use correlation and optimization approaches for data processing, and realize 1% ${\mathrm{CO}}_2$ accuracy for gas cell as short as 20 cm.

2.

Redundant Linear Equations

As shown in Fig. 1, light from a DFB laser passes through a long silica optical single mode fiber (SMF-28), gets modulated by the concentration of ${\mathrm{CO}}_2$ in the gas cell, passes silica fiber again and is then detected by a photo detector. Under the control of a computer, the wavelength $\lambda$ of the laser is scanned across an absorption line of ${\mathrm{CO}}_2$ around 1572 nm by tuning the working temperature of the laser, and according to the Beer’s law, the light power received by the photo detector can be expressed as

Fig. 1

Laser spectroscopy setup for gas sensing.

In our application, the gas pressure is approximately 1 ATM, thus $\sigma (\lambda )$ can be regarded as a fixed Lorentzian-like curve. We pre-measured the spectral ratio $R_{100}(\lambda )$ for 100% ${\mathrm{CO}}_2$, and use it as a reference to solve the optimal $\rho$ solution of a redundant linear equations:

There are only two unknown variables in Eq. (3), but $N$ is usually over 1000. This redundant condition helps to improve the detection accuracy of ${\mathrm{CO}}_2$ concentration.

3.

Spectral Shift Correction

Figure 2(a) shows four typical absorption spectral ratio curves we detected under the same condition. They should be the same in theory, however, due to the hysteresis phenomenon of the temperature control, a spectral error up to 0.1 nm was observed among different scans. As Eq. (3) depends not only on the absorption amplitude but also on the spectral information, such wavelength instability will degrade the demodulation accuracy of ${\mathrm{CO}}_2$concentration significantly. A spectrum correlation approach⁶ is used to solve the problem:

Figure 2(b) shows the correction results for spectra given in Fig. 2(a).

Fig. 2

Compensate the instability of thermal tuning. (a) original collected spectra for pure ${\mathrm{CO}}_2$ whose absorption lines shift randomly; (b) the shift was compensated by correlation algorithm.

4.

Experimental Results

By changing the operation temperature, the wavelength of the DFB laser was scanned from 1571.0 nm to 1572.5 nm, covering one of the ${\mathrm{CO}}_2$’s absorption lines at 1572.23 nm. ${\mathrm{CO}}_2$ samples whose concentration ranged from 0% to 90%, were tested, and the data were analyzed with correlation and optimization approaches as described in Secs. 2 and 3. Measurement is repeated 10 times to estimate the test stability. Figure 3 shows typical absorption spectra obtained for eight different ${\mathrm{CO}}_2$ concentrations, and the estimated concentration values are given in Fig. 4, where error bars are used to show the repeatability of the results under same test conditions. The error bar demonstrates that the standard deviation is $\sim 1\%$. We hence conclude that the system detection accuracy is 1% and the resolution is higher than 1%.

Fig. 3

Typical measured spectra at different ${\mathrm{CO}}_2$ concentrations.

Fig. 4

Estimated concentrations with error bar.

5.

Conclusion

For ${\mathrm{CO}}_2$ sequestration application, the background (concentration of ${\mathrm{CO}}_2$ in normal air) is around 0.03%, while the signal level (leakage concentration) usually ranges from 1% to 100%. The 1% detection accuracy we have obtained is sufficient for ${\mathrm{CO}}_2$ sequestration monitoring. Moreover, the system is based on near-infrared, the low attenuation and reliable performance of silica fiber ensures the long distance and large range remote sensing.