A Direct Bicarbonate Detection Method Based on a Near-Concentric Cavity-Enhanced Raman Spectroscopy System

Abstract

Raman spectroscopy has great potential as a tool in a variety of hydrothermal science applications. However, its low sensitivity has limited its use in common sea areas. In this paper, we develop a near-concentric cavity-enhanced Raman spectroscopy system to directly detect bicarbonate in seawater for the first time. With the aid of this near-concentric cavity-enhanced Raman spectroscopy system, a significant enhancement in HCO₃⁻ detection has been achieved. The obtained limit of detection (LOD) is determined to be 0.37 mmol/L—much lower than the typical concentration of HCO₃⁻ in seawater. By introducing a specially developed data processing scheme, the weak HCO₃⁻ signal is extracted from the strong sulfate signal background, hence a quantitative analysis with R² of 0.951 is made possible. Based on the spectra taken from deep sea seawater sampling, the concentration of HCO₃⁻ has been determined to be 1.91 mmol/L, with a relative error of 2.1% from the reported value (1.95 mmol/L) of seawater in the ocean. It is expected that the near-concentric cavity-enhanced Raman spectroscopy system could be developed and used for in-situ ocean observation in the near future.

1. Introduction

The carbon cycle in the ocean is a dynamic component of the global carbon budget, but the diverse sources and carbon sinks as well as their complex interactions in the ocean remain poorly understood [1,2]. The dissolved inorganic carbon (DIC) in the ocean is the sum of HCO₃⁻, CO₃²⁻, H₂CO₃ and CO₂ [3]. The concentration of DIC is about 2.30 mmol/L, with ${\mathrm{HCO}}_3^{–}$ (about 1.95 mmol/L) making up the majority of it [4]. Therefore, bicarbonate is a key factor in carbon cycle research. The modern methods for oceanic carbonate system measurement are mainly involved in total carbon and CO₂ partial pressure (pCO₂) investigations [5,6]. However, the direct detection of HCO₃⁻ and CO₃²⁻ in seawater it is still a huge challenge due to their low concentrations. Raman spectroscopy is potentially capable of overcoming this challenge on the condition of a significant ensitivity enhancement [7].

In 2004, scientists reported the first attempt to detect HCO₃⁻ in water solutions directly by using Raman spectroscopy [8], with a limit of detection (LOD) of 7.5 mmol/L after an integration time of 20 min. They had also tried to enhance the HCO₃⁻ signal with an introduced liquid core waveguide [9], eventually obtaining an amplification ratio of 7.8× in the peak intensity. All those efforts still do not make the direct Raman detection of HCO₃⁻ in seawater practical due to the fact its concentration is as low as 2 mmol/L. The strong background signal resulting from SO₄²⁻, which has a concentration of ~28 mmol/L in seawater [10], makes direct detection of HCO₃⁻ even more difficult. With a similar waveguide-based enhancement mechanism, the multi-pass cavity concept has been presented as an effective approach in gas Raman detection with desirable enhancement [11,12,13,14]. Taking such an approach in direct seawater Raman detection is the motivation for the work presented in this article. A near-concentric cavity-enhanced Raman spectroscopy system was specially built for liquid sample detection. With the aid of this system, direct HCO₃⁻ Raman detection in seawater becomes readily feasible as expected.

2. Experiment Setup

As shown in Figure 1, the near-concentric cavity-enhanced Raman spectroscopy system is adapted on the basis of a formerly reported design [14]. In order to detect liquid samples, a sample cell of fused silica with 10 mm × 10 mm × 40 mm inner size and 1 mm wall thickness is used and placed at the center of the cavity. In this system, a diode-pumped, frequency-doubled Nd:YAG CW laser (532 nm) with a power of 300 mW is used as the Raman excitation laser source. The polarization of the laser beam is rotated by 90° through passing a half wave plate (HWP), and then the laser beam is compressed and focused into the chamber by a couple of lens (a planoconvex lens with a focal length of 100 mm and a planoconcave lens with a focal length of −75 mm). The near concentric cavity is composed of two identical spherical mirrors M1 with 25.4 mm diameter and 25 mm focal length. The reflectivity of these two mirrors is over 99% from 500 to 700 nm. These two mirrors are spaced face to face at a distance of 104.2 mm, and the mirror M1-2 is clockwise rotated about 0.04° to create the near concentric cavity reflection mode. Because of the strong absorption of water, the laser beam number in the cavity is 18, which is less than the number in the cavity of a gas sample cell. The scattering signal from the center of cavity is collected by an achromatic doublet lens L1 (f = 30 mm) with a diameter of 25.4 mm. A mirror M2 (f = 12.5 mm), placed on the opposite side beyond the sample chamber, is used to enhance the collection efficiency of the signal. Another lens L2 (f = 30 mm) is used to couple the Raman signal into the delivery optical fiber bundle (19 μm × 100 μm, NA = 0.22), with a long pass filter placed on its front to get rid of Rayleigh scattering from the collected signal. From the other end of the fiber, the signal is then coupled into a spectrometer (Acton SP2500i, Princeton Instruments, Trenton, NJ, USA), with 1200 g/mm grating and 10 μm entrance slit width. The Raman spectra are recorded by a CCD detector (SPEC-10: 400B, Princeton Instruments, Trenton, NJ, USA) with a 1340 × 400 imaging array and 20 μm × 20 μm pixel size operating at −40 °C. The resulted spectral range is 800~2000 cm⁻¹ with a resolution of 2 cm⁻¹. To get a relative desirable Raman signal and prevent the CCD from overexposure, each spectrum is accumulated five times with an integration time of 10 s, so the total acquisition time is 50 s for one sample. Five measurements are taken for each sample to evaluate the repeatability.

3. Results

The reported LOD for SO₄²⁻ by using a liquid core optical fiber Raman system is about 1.5 mmol/L [15]. As a comparison, we prepared 15 Na₂SO₄ solutions (0.10 mmol/L–2.00 mmol/L) to evaluate the LOD of the NC-CERS system. The LOD is defined as the ratio of three times the noise intensity (σ) to the slope of calculation curve. The spectra of four solutions are shown in Figure 2a, and the Raman signal of SO₄²⁻ can be observed clearly in the spectrum of the 0.10 mmol/L solution. Given this, the NC-CERS system has a LOD less than 0.10 mmol/L for SO₄²⁻ which is 15 times lower than that reported before. The linear relationships between concentrations and peak intensity of the SO₄²⁻ are shown in Figure 2b and the response of the NC-CERS system for liquid detection has good linearity with R² = 0.996 over the whole range.

To evaluate its performance for HCO₃⁻, we prepared nine NaHCO₃ solutions with concentrations ranging from 0.40 mmol/L to 4.00 mmol/L. The corresponding concentrations of HCO₃⁻ are from 0.37 mmol/L to 3.79 mmol/L calculated by using the carbon balance mode. All of the concentrations of HCO₃⁻ are corrected by using the equilibrium mode of carbon components in water. Figure 3a shows the spectra of five of these concentrations. The peak intensity of the HCO₃⁻ Raman signal increases as a function of the concentration, and the Raman signal of HCO₃⁻ can be distinguished in the solution with 0.37 mmol/L NaHCO₃. It is proved that the LOD of the NC-CERS system for HCO₃⁻ is less than 0.37 mmol/L, which is much lower than its concentration in seawater. The linear relationship between concentrations and peak intensities of the HCO₃⁻ is shown in Figure 3b and it shows good linearity with R² = 0.978.

Furthermore, we prepared simulated seawater with fixed a NaSO₄ concentration (28.00 mmol/L) and different NaHCO₃ concentrations (0.00, 2.00, 4.00 mmol/L) to evaluate its ability to quantify HCO₃⁻ in seawater. The corresponding concentrations of HCO₃⁻ are 0.00, 1.93 and 3.79 mmol/L. The corresponding spectra are shown in Figure 3c, and the detailed spectra are shown in Figure 3d. Compared with the intensity of SO₄²⁻, the intensity of the HCO₃⁻ signal is much weaker. Even so, the Raman signal of HCO₃⁻ can be detected in the spectrum of the mixed solution with 1.93 mmol/L NaHCO₃. This shows the ability of the NC-CERS system for direct HCO₃⁻ detection in seawater, but the strong and adjacent SO₄²⁻ signal of seawater has an impact on the quantitative analysis of HCO₃⁻.

In order to reduce this impact, we developed an adaptive signal extraction method for HCO₃⁻ and the process is shown in Figure 4a–c. The spectrum is obtained from a mixed solution with 28.00 mmol/L Na₂SO₄ and 1.93 mmol/L NaHCO₃. In this spectrum, the Raman peaks of SO₄²⁻ (981 cm⁻¹) and H₂O (1640 cm⁻¹) can be clearly recognized, while the peak of HCO₃⁻ (1021 cm⁻¹) is almost submerged in the background. Given this, the original spectrum is normalized according to the peak intensity of H₂O to reduce the impact of laser power fluctuation as shown in Figure 4a. The next step is baseline correction. We have compared four baseline fitting results with four methods (Voigt, Gaussian, Lorentzian and bi-exponential fitting methods), and the polynomial function shows the best performance among them. The baseline corrected result is shown as a blue line in Figure 4b. The last step is the weak HCO₃⁻ signal extraction from the neighboring strong SO₄²⁻ signal. The right wing of the SO₄²⁻ signal is regarded as the baseline of the HCO₃⁻ signal and is fitted by using a bi-exponential function in the region of 1000–1050 cm⁻¹. We again tried four baseline fitting methods (Voigt, Gaussian, Lorentzian and bi-exponential fitting methods), and the polynomial function showed the best performance among them. The fitted baseline is shown as the green dotted line, and the corrected spectrum is shown as the black circle dots in Figure 4c. The Raman peak of HCO₃⁻ is extracted and fitted by using a Gaussian function which is shown as the red line in Figure 4c.

Based on this data processing method, the Raman signals of HCO₃⁻ in simulated seawater solutions are extracted and shown in Figure 4d. The signal of a deep seawater sample is also shown as the cyan line in Figure 4d. The linear relationship between concentrations and peak intensities of fitted HCO₃⁻ is shown in Figure 4e and it shows high linearity with R² = 0.951 over the range. We chose a mixed solution with 28.00 mmol/L NaSO₄ and 1.50 mmol/L NaHCO₃ as a blind sample. The test result is 1.35 mmol/L with 5.0% relative error, which is shown as the pink point in Figure 4e. Furthermore, we tested deep seawater, knowing the concentration of HCO₃⁻ is 1.91 mmol/L which is shown as a blue point in Figure 4e. Compared with the reported HCO₃⁻ concentration value of 1.95 mmol/L in seawater [4], the relative error was about 2.1%.

4. Conclusions

In this paper, we developed a near-concentric cavity-enhanced Raman spectroscopy system for liquid sample detection based on a near-concentric cavity and directly detected bicarbonate in seawater for the first time. Systematic experiments have been carried out for system evaluation. As shown by the NC-CERS system tests, the LOD for SO₄²⁻ is less than 0.10 mmol/L, which is 15 times lower than the result reported before. The LOD for ${\mathrm{HCO}}_3^{–}$ is less than 0.37 mmol/L, which is much lower than its concentration in seawater (1.95 mmol/L). The detection performance of the NC-CERS system for HCO₃⁻ in seawater is evaluated by analyzing mixed solutions of fixed 28.00 mmol/L NaSO₄ and different NaHCO₃ concentrations. The Raman signal of 1.93 mmol/L HCO₃⁻ in mixed solution can be clearly detected, which shows the ability of direct HCO₃⁻ detection in seawater. To further realize the quantitative analysis for HCO₃⁻, a specially developed data processing scheme is introduced. With this data processing method, the weak HCO₃⁻ signal is extracted from the neighboring strong SO₄²⁻ signal, hence a quantitative analysis with R² = 0.951 is made possible. Based on this calibration curve, the HCO₃⁻ concentrations of a blind sample and a seawater sample are calculated. The calculated values are 1.43 mmol/L and 1.91 mmol/L with 4.9% and 2.1% relative error, respectively. It is hoped that this NC-CERS system could be fully developed for field ocean observations in the near future. CO₃²⁻ with a concentration down to 1/10 of the HCO₃⁻ in seawater will be the next detection object.