SHINING A LIGHT ON TDLAS Airat Amerov and Nicole Rose, Ametek, USA, evaluate how a tunable diode laser absorption spectroscopy (TDLAS) gas analyser can effectively measure carbon monoxide and methane in hydrogen gas. H ydrogen is heavily consumed in refineries, by the cryogenics industry, and in the production of ammonia and methanol. Petroleum refineries are both consumers of hydrogen and large producers of hydrogen. Refinery hydrogen is often used for hydrotreating naphtha, in hydrodesulfurisation, and in cycle oil hydrogenation. Traditionally, catalytic reforming has been used to produce hydrogen as a byproduct of the production of high octane aromatic compounds used in gasoline, with steam reforming as the dominant production method. The hydrocarbon feed is mixed with steam at high temperatures and then passed through catalyst filled tubes to produce synthesis gas, which is an equilibrium mixture of hydrogen, carbon monoxide and carbon dioxide. The synthesis gas that exits the reformer is cooled before entering a shift converter, where more hydrogen is produced as carbon monoxide and steam are converted to form carbon dioxide and hydrogen. Hydrogen gas purity is largely dependent on the purification method, as residual methane in the hydrogen gas reduces its purity. To HYDROCARBON ENGINEERING Reprinted from May 2016 Table 1. Composition of product hydrogen Gas component Wet scrubbing PSA Hydrogen, % 95 - 97 99 - 99.99 Methane, % 2-4 Down to 100 ppm CO and CO2, ppm 10 - 50 10 - 50 Nitrogen, % 0-2 0.1 - 1.0 Figure 1. A calculated spectra of carbon monoxide and methane, at a temperature of 50˚C and atmospheric pressure in an optical path of 1 m. reduce the amount of impurities, wet scrubbing hydrogen plants commonly use shift converters, in combination with carbon dioxide removal and methanation. Depending on the temperature of the shift converter and the reformer steam to carbon ratio, the amounts of carbon monoxide and methane at the outlet of a shift converter can vary from 0.5 - 4% and from 2 - 8%, respectively. In newer plants, wet scrubbing has been replaced by pressure swing adsorption (PSA), in which adsorbent beds are used to remove impurities. The composition of the hydrogen product depends on the technology used. Typical compositions are shown in Table 1. Hydrogen management is a critical concern for refiners because various processes require different levels of hydrogen purity and pressure. Additionally, the control of the impurities provides opportunities for optimising the hydrogen production process, which may result in both time and energy savings, as well as improvements in the quality of the final product. Near-infrared tunable diode laser absorption spectroscopy (TDLAS) has gained much attention for industrial applications. This technology has three key attributes: specificity for the analyte, high sensitivity and fast response time. The specificity is the result of the extremely high spectral resolution achievable. Emission bandwidths for tunable diode lasers are on the order of 10-4 - 10-5 cm-1, which results in the ability to isolate a single ro-vibrational transition line of an analyte species. The second attribute of TDLAS is the ability to rapidly tune the lasers, so techniques such as wavelength modulation spectroscopy (WMS), which typically yield two orders of sensitivity enhancements over a direct absorption approach, are easily implemented. Because TDLAS is an optical technique, it offers a very fast response time. The high specificity, sensitivity and response time of TDLAS make it very suitable for a variety of process measurements. Reprinted from May 2016 HYDROCARBON ENGINEERING The specificity of TDLAS for an analyte is dependent on the sample matrix. For some applications it is difficult to find and use an absorption line for the analyte species that is free of interference from all other species in the sample matrix. By combining TDLAS with chemometric strategies that have been used in other fields of spectroscopy, new application possibilities emerge. The measurement of impurities – carbon dioxide, carbon monoxide and methane – in product hydrogen gas is an attractive application area for TDLAS. The measurement of carbon dioxide is straightforward and has previously been demonstrated by one of the authors in a similar sample matrix. However, making reliable measurements of carbon monoxide in this sample matrix is complicated by the spectral overlapping with methane in the vicinity of 2334 nm. The use of multivariate analysis enables resolution of overlapping responses. The objective here is to characterise a new TDLAS-based analyser. The analyser was set up to perform an all digital implementation of WMS. A distributed feedback (DFB) laser, with an emission centred at 2334 nm, was used to measure the concentration of carbon monoxide in product hydrogen gas. A key feature of this instrument was the use of a sealed reference cell, which contained known amounts of the analytes, for referencing the emission wavelength of the laser. The use of a reference cell enabled both the ability to line lock the laser and continuously validate the analyser performance. The instrument The TDLAS analyser evaluated in this work was an Ametek 5100 HD, which was modified to operate with a 2334 nm DFB laser. Reference spectra of carbon monoxide and hydrocarbons in the vicinity of 2334 nm, calculated from Pacific Northwest National Laboratory (PNNL) data, is shown in Figure 1. Of the three hydrocarbons shown in the figure, only the methane had any substantial peak structure overlapping with the carbon monoxide line, which would interfere with measurements of low concentrations of carbon monoxide. Figure 1 also demonstrates that, with increased molecular weight of the hydrocarbon molecule, there is a significant reduction in the absorbance spectral profile. Corresponding 2F spectra of heavier hydrocarbons results in close to zero levels and could be neglected. However, the carbon monoxide line close to 2334 nm has minimal overlap with methane, as compared to other lines in this region of the spectrum. This absorption line also has more than two orders of magnitude higher intensity than carbon monoxide absorption lines in the vicinity of 1570 nm. Additionally, the line at 2334 nm has negligible spectral interference from carbon dioxide, which is another component of the product hydrogen gas. Scanning the laser over a range of wavelengths containing peaks for both carbon monoxide and methane compensates for background interference. A schematic representation of the analyser used for the measurement of carbon monoxide concentrations is shown in Figure 2. The laser produced an optical power of approximately 3 milliwatts (mW). A Herriott cell provided an optical path length of 16 m. In addition, a fibre-optic coupled DFB laser and a gradient refractive index (GRIN) lens based collimator were used in this configuration. In contrast to the analyser configurations published previously by this group, the reference cell is in line with the sample cell. The reason for this configuration was the absence of available fibre-optic beam splitters for use at this wavelength. The sample cell and the reference cell were assembled into one optically aligned module of the analyser. This module is shown in Figure 3. At the output of the sample cell, a plate beam splitter, designed for 45˚ orientation, was used to divide the optical power in a 50:50 ratio. One beam was directed through the reference cell, which was filled with carbon monoxide. The detection of the 2334 nm energy required the use of extended InGaAs photodiode detectors. The sample channel photodiode was located close to the beam splitter. The reference cell signal was used to lock the laser wavelength on the selected absorption line for carbon monoxide. The reference cell contained carbon monoxide in an optical path of 1 cm. Two nested levels of temperature control were employed to maintain the operation of the DFB lasers at the proper wavelength. The first level is a simple PID control loop, which maintained the laser at a target temperature. In the second level, the outer control loop, the spectra of the analyte samples in the reference cells were monitored. Minor shifts in the observed peak positions were used as a feedback signal for the temperature set point of the inner control loop. The outer control loop provided a fine adjustment for the inner control loop. While this work describes a TDLAS-based analyser performing WMS, all aspects of the experiment were performed with a digital implementation of the technology. The WMS experiment was implemented by using a digitally sampled sine function, summed with a staircase, and the resulting signal was used to drive the tunable DFB lasers. Specifically, the sine function was sampled at 21 discrete points, evenly spaced across each period. This digitally synthesised signal provided a very close approximation to an analogue WMS experiment. Further, signals produced by the detectors were digitised, prior to applying signal processing (e.g., phase sensitive detection, smoothing, etc). The duration for the discrete steps in the waveform was stored as a setup parameter in the analyser, so the modulation/demodulation frequencies were software selectable. Additionally, in contrast with the common practice of using second harmonic detection (2F), the detection/demodulation in this instrument was performed at the laser modulation frequency (i.e. ‘1F’ detection). Using the 1F detection scheme enabled the analyser to normalise the spectra, without the need for a separate measurement of the laser power. Specifically, the magnitude of the power envelope of the laser output was contained in the spectra produced by 1F demodulation. After the 1F spectra were normalised, ‘2F’ spectra were calculated as derivative of the 1F spectrum. Methods Sample gases for this study were prepared using a programmable gas mixer to blend commercial carbon monoxide, methane and other background gases to simulate hydrogen recycle gas of varying compositions. The concentration ranges for carbon monoxide and methane were set in agreement with real orders provided by customers. Figure 2. A schematic diagram of the TDLAS analyser for carbon monoxide measurements at low concentration levels. Figure 3. Sample and reference cells assembly. Figure 4. Carbon monoxide spectra with a methane background. A calibration was developed for the TDLAS instrument covering a carbon monoxide concentration range of 0 - 100 ppm and a methane concentration range of 0 - 1000 ppm balanced hydrogen. A multivariate regression was used to establish the calibration models for carbon monoxide. For the low concentration range of carbon monoxide, a calibration was developed for the range of 0 - 100 ppm, with 1000 ppm methane and a balance of hydrogen. Following the calibration, instrument performance was evaluated. The overall measurement accuracy for carbon monoxide was established, and corresponding limit of detection (LOD) was determined. Results and discussion Solving the multivariate problem was achieved with an inverse least squares regression (P-matrix calibration). The HYDROCARBON ENGINEERING Reprinted from May 2016 Figure 5. Carbon monoxide readings for a series of concentration challenges in the range of 0 - 100 ppm. Figure 6. Analyser zero drift test. Table 2. Tested analyser performance parameters Analyser LOD, ppm Accuracy, ppm Repeatability, ppm Drift, ppm 1 2.8 1.0 0.2 1.1 2 4.5 2.1 0.4 1.0 3 2 0.5 0.2 1.0 response variables used in the regression were the integrated values observed over three spectral bands in the 2F spectra. Specifically, a band centred at the peak in the carbon monoxide spectrum and two additional bands were used. All concentrations are cited by volume (i.e. ppm and % [V/V]), and the estimates were calculated as: 3 nn Cj = Σ αj,i Ri i=l Where: Cj = concentration estimates for each component; αj,i = regression coefficients; and Ri = integrated band intensities. Example spectra collected with the 2334 nm laser are shown in Figure 4. Each of the spectra in the figure covers a range of 2333.4 - 2334.1 nm, and is the average of 100 separate scans. In these data, the peak amplitude and area of the 2F carbon monoxide and methane signals were proportional to the concentration of carbon monoxide and methane in the cell. With increasing concentrations of Reprinted from May 2016 HYDROCARBON ENGINEERING carbon monoxide, a peak at 2333.72 nm became more notable. Spectra recorded for two separate concentrations of carbon monoxide (50 and 100 ppm) and methane spectra for a concentration of 1000 ppm are shown in Figure 4. The methane spectral features are clearly observed on the long wavelength side of the spectra. This background spectrum was significant in comparison with the absorbance of the carbon monoxide at these concentration levels. Hence, the spectra of the sample matrix components needed to be built into the calibration model. The regression vector that was calculated as a result of the inverse least squares regression is also shown in Figure 4. The calculated regression vector is analyte specific and corresponds to regions where the analyte spectra are changing with concentration. The calibration procedure for the TDLAS instrument was conducted, and a multivariate calibration model for the carbon monoxide measurement was developed. All spectral data were recorded for gas samples under atmospheric pressure and a temperature of 60˚C. Spectra were recorded for a series of different carbon monoxide concentrations, ranging from 0 - 100 ppm, with 1000 ppm methane and a balance of hydrogen. Analysis of this application was based on the performance of three separate TDLAS analysers, which have subsequently been placed in the field. According to stream composition data provided for these installation sites, the background gas contained approximately 1000 ppm of methane in a balance of hydrogen and corresponded to purified hydrogen gas at the outlet of a purification unit. Figure 5 shows analyser responses to a series of carbon monoxide challenges over the concentration range for one of the selected analysers. The duration for each challenge was approximately 10 minutes, and the data were acquired at a rate of 2 seconds per measurement. A return to the background gas baseline was also tested before and after the validation measurements. A methane-hydrogen gas mixture was used as the background gas. The response time (T90) was measured to be 48 seconds, which was limited by the larger volume of the sample cell and the propagation of the gas in the sampling system with a flow rate of 2 l/min. The results of the drift test for one of the analysers are shown in Figure 6. Background gas was running through analyser sample cell during the entire test. The drift was evaluated as the difference between maximum and minimum carbon monoxide readings recorded during the test. The results of the validation and drift testing for all three tested carbon monoxide analysers are shown in Table 2. Conclusion A new TDLAS-based analyser was evaluated for the measurement of carbon monoxide in product hydrogen gas. A DFB laser was used in the analyser. Scanning the laser over a range of wavelengths sufficient to capture the overlapping absorption lines for carbon monoxide and methane enabled the simultaneous measurement of carbon monoxide. The digital signal processing methods that were employed in this system were successful in implementing a multivariate calibration that enabled the analyser to accurately measure carbon with accuracy of 2 ppm over a concentration range of 0 - 100 ppm range in the gas stream of 1000 ppmv of methane balanced hydrogen.
© Copyright 2026 Paperzz