Raman spectroscopy for natural gas process applications

Hele tekst

(1)Raman spectroscopy for natural gas process applications An instrumental and operational survey of theory and practice Author: Christiaan Mul Bsc Supervisors: Dr. Freek Ariese (VU), Dr. Jan-Hein Hooijschuur (ASaP) Master thesis Chemistry track analytical sciences, Date: 12-Dec-2017.

(2) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 1 A BSTRACT This study focusses on online spectrometric applications of the Raman effect for process control equipment in natural gas utilizing plants. This subject springs from the upcoming variable calorific value in the Dutch natural gas distribution grid which makes innovation necessary to maintain safe and competitive processes. Further insight in the industrial landscape and the innovative drive is given in the first chapter. Available techniques are discussed and the strengths and weaknesses that come with the introduction of Raman technology are considered. To gather insight into the technology a full chapter is focused on Raman theory and natural gas. The fundamental principles upon which the technology is based are discussed as well as the concept of spectroscopy. Natural gas is a complex mixture of components and impurities. The varying concentrations of the components in the natural gas influence key parameters. For a detailed look on natural gas both literature and computer experiment are deployed to report composition ranges, and predicted spectra. One of the main subjects is the experimental setup that can be used to measure Raman scattering. Following the route of the light, the components are each discussed and their function in the whole described. A completely new process interface is designed for this application and subjected to robustness and efficiency simulations. Other simulations and experiments are done to attempt optimization of the interface between the collection fiber and spectrometer entrance slit. The physical instrumentation only has a detector signal whereas the composition and key parameters are the valuable results. The processing of a detector signal to a proper result is stepwise discussed with examples and flow charts. Calibration of the detector, cosmic ray detection, and dark current and background correction are shown. Additionally some ideas are shared about the implementation and restrictions of multivariate modelling. Finally at the discussion it is discussed what worked well, and where the instrumentation may be optimized with suggested alterations. Both the instrumentation and the operational effectivity of the application are discussed whereby the results are taken into account. It was found that only few goals were met, spectra can be recorded from the main components although they differ little over their tested concentration range. Future research should be focused on increasing the sensitivity of the measurement, which is found to be the main weakness, and on developing advanced algorithms for the determination of key parameters of natural gas. The study reveals that to maintain optimal cost and process control, fast and accurate analysis methods need to be developed that can measure the anticipated compositional changes1. The main question is ” is Raman spectroscopy a viable technology for the compositional analysis of natural gas mixtures?”. Page 2 of 116.

(3) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 2 T ABLE OF C ONTENTS. 1. ABSTRACT ............................................................................................................................................ 2. 2. TABLE OF CONTENTS ......................................................................................................................... 3. 3. RAMAN SPECTROSCOPY FOR NATURAL GAS ASSESSMENT ...................................................... 5 3.1 THE RISE OF LIQUEFIED NATURAL GAS (LNG) AND ITS INTRODUCTION INTO THE DUTCH MARKET..................... 5 3.2 RAMAN TECHNOLOGY................................................................................................................................................... 7 3.2.1 Diffraction spectrometer, robustness from a static design .................................................................... 7 3.2.2 Fast analysis time can improve process control ....................................................................................... 8 3.2.3 In situ analysis can reduce error and emissions .....................................................................................10 3.2.4 Sensitivity and selectivity ................................................................................................................................11 3.3 ALTERNATIVE MEASUREMENTS TECHNIQUES ............................................................................................................11. 4. IN THEORY ......................................................................................................................................... 13 4.1 THE RAMAN-EFFECT ....................................................................................................................................................13 4.2 NATURAL GAS IN DETAIL .............................................................................................................................................16 4.2.1 Literature spectra ...............................................................................................................................................17 4.2.2 Theoretical prediction of spectra ..................................................................................................................18. 5. INSTRUMENTATION OF THE EXPERIMENTAL SETUP ................................................................. 22 5.1 LIGHT SOURCE, LASER .................................................................................................................................................23 5.2 EXCITATION FIBER ........................................................................................................................................................25 5.3 OPTICAL PROBE ...........................................................................................................................................................26 5.3.1 Collimation lenses .............................................................................................................................................28 5.3.2 Filters and mirrors .............................................................................................................................................28 5.4 EXTENSION TUBE, IMMERSION PROBE .......................................................................................................................30 5.5 LIGHT PATH GEOMETRY ...............................................................................................................................................32 5.6 LENSES AND WINDOWS ..............................................................................................................................................37 5.7 COLLECTION FIBER .......................................................................................................................................................39 5.8 SPECTROGRAPHS .........................................................................................................................................................44 5.8.1 Entrance slit .........................................................................................................................................................45 5.8.2 Observation times..............................................................................................................................................47 5.8.3 Mirrors and grating ...........................................................................................................................................47 5.8.4 Detector .................................................................................................................................................................48 5.9 MEASUREMENT CELL ...................................................................................................................................................52. 6. SIGNAL TO SPECTRUM .................................................................................................................... 54 6.1 SIGNAL CALIBRATION ..................................................................................................................................................54 6.1.1 Wavelength ..........................................................................................................................................................54 6.1.2 Raman shift ..........................................................................................................................................................56 6.1.3 Intensity .................................................................................................................................................................57 6.2 SIGNAL PREPARATION .................................................................................................................................................57 6.2.1 Number of data points.....................................................................................................................................57 6.2.2 Cosmic ray detection ........................................................................................................................................60 6.2.3 Dark current correction ...................................................................................................................................61. Page 3 of 116.

(4) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS 6.2.4 Background correction .....................................................................................................................................63 6.2.5 Noise filters ..........................................................................................................................................................66 6.3 RESULTS AFTER SIGNAL PREPARATION.......................................................................................................................68 6.4 SIGNAL ANALYSIS BY MODELLING ..............................................................................................................................72 6.4.1 Univariate model ...............................................................................................................................................72 6.4.2 Multivariate models ..........................................................................................................................................73 7. DISCUSSION AND CONCLUSION ................................................................................................... 75. 8. REFERENCES ...................................................................................................................................... 78. 9. LISTS OF FIGURES, EQUATIONS, AND TABLES ............................................................................ 83 9.1 9.2 9.3. FIGURES ........................................................................................................................................................................83 EQUATIONS ..................................................................................................................................................................86 TABLES ..........................................................................................................................................................................86. 10. GLOSSARY OF TERMS ...................................................................................................................... 87. 11. APPENDIX A ...................................................................................................................................... 88. 11.1 11.2 11.3 12. APPENDIX B ..................................................................................................................................... 110. 12.1 12.2 12.3 12.4 13. COMPOSITION OF GAS STANDARD ‘HIGH CALORIFIC NATURAL GAS’ ................................................................. 110 KEY PARAMETERS OF GAS STANDARD ‘HIGH CALORIFIC NATURAL GAS’ ............................................................ 111 COMPOSITION OF GAS STANDARD ‘LOW CALORIFIC NATURAL GAS’ .................................................................. 112 KEY PARAMETERS OF GAS STANDARD ‘LOW CALORIFIC NATURAL GAS’ ............................................................. 113. APPENDIX C ..................................................................................................................................... 114. 13.1 14. THEORETICALLY PREDICTED SPECTRA CALCULATED OF PURE COMPOUNDS IN NATURAL GAS.............................88 COMBINED PREDICTED SPECTRA ............................................................................................................................. 102 COMBINED PREDICTED SPECTRA FROM TYPICAL COMPOSITIONS ........................................................................ 108. FORMULAS AND CONSTANTS USED TO CALCULATE ISOTOPIC INFLUENCE .......................................................... 114. APPENDIX D .................................................................................................................................... 115. 14.1 14.2. SECOND MANUFACTURER COMPARISON OF SPECTROGRAPH ............................................................................. 115 SPECTROMETER DETECTOR SPECIFICATIONS .......................................................................................................... 116. Page 4 of 116.

(5) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 3 R AMAN SPECTROSCOPY FOR NATURAL GAS ASSESSMENT Quality measurements of natural gas are essential for a multitude of processes. The quality is characterized by certain key parameters following from the composition. This study focusses on the direct assessment of the natural gas quality with the use of Raman spectroscopy. The development’s innovation cannot be found merely in the application of Raman spectroscopy on natural gas, but also in the probe design which facilitates insitu analysis. The setup and use of available components is studied, in theory and where possible in practice. The thesis concludes upon weighing the results of the experiments and evaluates their practical use.. 3.1 T HE. RISE OF LIQUEFIED NATURAL GAS. (LNG). AND ITS INTRODUCTION. INTO THE D UTCH MARKET “The interest in liquefied natural gas (LNG) has recently intensified due to the development of significant global gas reserves and more advanced techniques for their recovery. An increased awareness of the human carbon footprint has led to advanced carbon accounting. When natural gas is used for electricity generation it can diminish carbon dioxide emission, compared to coal, by 10%2.The abundance of natural gas combined with the worldwide demand for energy leads to many questions about the efficient production and transportation of LNG3. Which method is optimal for the transportation of natural gas from the production site to the customer site is a complex discussion. A few options to choose from include: pipelines, compressed natural gas, gas to liquids, and gas to solids. To use available resources effectively, the distance covered per energy unit should be maximised4. The energy density per volume in LNG is approximately a factor 600 higher than that of natural gas. This property makes it profitable to transport volumes of LNG in containers over longer distances than would be feasible with pipelines. Energy transactions are accounted for using energy content per volume per currency5. The energy content of natural gas is readily measured with conventional techniques such as gas chromatography or with the faster micro gas chromatography methods. The developments in analysis are aimed at the development of faster and more precise machines so that a more accurate calculation of the energy content can be made.” (Citation from Mul, 2015)6 The primary objective in every company is to maximize profit while saving resources. When focused on the (petro-)chemical or energy industry this can be done by the use of process analysis. In this study special attention is given to natural gas compositional analysis due to the anticipated challenges formed by the shift of supply and quality of the gas7. Figure 1 shows the caloric value distribution throughout the Netherlands. There are two main problems that arise from the new Liquefied Natural Gas (LNG) supply. First, the change in heat index, the so-called H-gas contains less nitrogen and has a higher caloric value compared to the conventional G-gas commonly used in the Netherlands. Second, there are plans to add hydrogen to the gas in order to reduce carbon dioxide emissions. This would create gas with less carbon emissions, but simultaneously an even broader array of possible caloric values available from the natural gas distribution network.. Page 5 of 116.

(6) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 1 H-gas map of the Netherlands, note the LNG-import dock labelled 'LNG'. Source: gasunietransportservices.nl accessed 16-11-17. Gas turbines can be a major source for relatively clean electricity, certainly when methane is burned only carbon dioxide and water is formed. Still the gas does not come free of cost, and the turbine needs to control the air and fuel flows. Currently these processes are primarily controlled by Gas Chromatographs (GCs) or Wobbe index analyzers. These classic technologies each have their own positive and negative properties. For instance, GCs rely on the physical separation of the analyte components, and are therefore limited in the achievable speed. In addition, Wobbe index analyzers cannot adjust for the different hydrogen oxidation stoichiometry compared to carbon containing fuels8. In the future it is therefore expected for these gas turbines to run less efficiently than theoretically possible, a waste of money and fuel. In short, the current generation of analysis equipment is not sufficiently capable of dealing with fast changing gas compositions. Innovation in fast and accurate compositional analysis would increase the possibility for control of cost of the used natural gas, and the control of the fuel to air mixture with all its advantages. During the course of this study a survey will be made of both the instrumentation and operation of Raman technology to ensure the highest efficiency. The extent of some experiments is mainly theoretical whereas also a complete measurement rig is built to bring theory into practice for further evaluation. The purpose of the survey is aimed at finding an ideal setup to measure the composition and quality of natural gas with Raman spectroscopy.. Page 6 of 116.

(7) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 3.2 R AMAN. TECHNOLOGY With the image of mammoth tankers and gas turbines it might be hard to imagine laboratory Raman spectroscopy setup amidst the hard hatted coveralls. But nothing is further from the truth, Raman spectroscopy is a developing field in process analytical technology9,10,11,12 with a growing number of applications. This development is expected to proceed as optical components improve in quality, ruggedness, and affordability.. 3.2.1 Diffraction spectrometer, robustness from a static design An advantage of the Raman spectroscopic method is that it can be fully functional without the use of any moving parts. All major components consist and can operate based on fixed components. Both Figure 2 and Figure 3 show a diffraction spectrometer, one schematic and an Avantes ULS with an opened casing. Not only are fixed parts easier to control, the lack of bearings and motors also reduces maintenance. The application locations are often vibrating due to pumps, turbulent flows, or other heavy equipment. Ruggedness of the spectroscopy equipment is therefore considered an advantage.. Figure 2 Schematic view of diffraction spectrometer including lightpaths. Figure 3 Picture of opened spectrometer, Manufacturer: Avantes13. Both a diffraction spectrometer and a filtered band spectrometer can be used for measurements on a range or selected bandwidths. Whereas the diffraction spectrometer is static, without moving parts, the filtered spectrometer needs a driver motor and bearing for the filter wheel, as can be seen on Figure 4. A common practice is to replace the bearing, driver axis/snare, and motor after a number of operation hours when they become prone to faults. The lifetime of these moving parts is significantly less, 1 to 5 years, compared to the static components that can last an analyzer’s lifetime. Each method of light separation and detection has interesting characteristics, that will be discussed in chapter 5.8. The increased stability and improved maintenance interval make the diffraction spectrometer the preferred option in a process environment.. Page 7 of 116.

(8) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 4 Schematic view of filtered band spectrometer. 3.2.2 Fast analysis time can improve process control Processes can be regulated in multiple ways, the most used concepts are feed-back and feedforward control. As an example we will examine the Claus process where H2S is partially combusted and catalytically reformed to elementary sulfur and water, see Figure 5 for a schematic view of the process. The possibility to analyze the sample within seconds makes it possible to rethink control mechanisms used in industrial processes. In traditional feed-back control, the Claus process is equipped with an analyzer to sample the final gaseous output, so-called ‘tail gas’, to determine the ratio H2S to SO2. This information shows, among other things, whether the combustion ratio was correct. If the ratio is out of bounds more or less air is added to the combustion to keep the ideal ratio for the catalytic conversion. The time needed for the feed gas to combust and flow through multiple condenser and converter steps is, depending on the specific installation, around 30 minutes. This means if the composition in the feed gas changes, the process can only be adjusted 30 minutes later, resulting in excessive SO2 output or insufficient sulfur recovery. In feed forward control the feed gas is analyzed before it reaches the combustion furnace. With the use of computers the needed amount of air is then calculated while the process can still be controlled. The correct adjustment of the process should be known in time, so adjustments to the trim air can be made in time, from sample to adjustment this would be approx. 10 seconds. A tail gas analyzer is still needed to observe the catalyst activity, but it will definitely convert more sulfur if the ratio is precisely controlled. The planned regulation of the process will prevent unnecessary loss of efficiency or damage to the environment. The example above shows the possibility of improving multiple qualities dependent on efficient process control: environmental damage, loss of raw materials, and loss of heat. The development of the feed forward control was only possible with the development of a fast spectroscopic analysis method combined with adequate sample handling. Raman spectroscopy has possible applications to instantly know exactly what is fed into a furnace, turbine, or reactor so that the process itself can be improved. Page 8 of 116.

(9) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 5 Schematic view of the Claus process for sulfur recovery, indication of feedforward and feedback loop. Page 9 of 116.

(10) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS 3.2.3 In situ analysis can reduce erro r and emissions In situ, or in English “in the original place”, comprehends a whole different approach to analysis. In extractive methods the sample is taken from the pipeline, transported, adjusted for analysis, and only then analyzed. With in situ methods the sample is analyzed directly inside the process. Interest in this type of measurements is fueled by the possibility to eliminate a loss of accuracy resulting from the sampling and sample handling method. Another positive effect resulting from in situ measurements is the reduction of undesirable emissions. Whereas normally the sample is extracted, analyzed, and discarded, the sample can now be fully used in the continued process. A good example is the measurement of residual oxygen in combustion furnaces for the optimization of the burner efficiency. This type of measurement can be done in situ or extractive with the use of a Zirconium Oxide-sensor14, also known as a lambda-sensor. The sensor is flow-sensitive, therefore the in situ probe is designed to rely on a diffusion principle where the direct flow over the sensor is always stable. When the sensor is used extractive, for instance when also a combustion sensor is needed or for easier maintenance access, a positive flow over the system is needed. A maladjustment of the flow would lead to a biased signal on the extractive oxygen measurement compared to the in situ measurement, solely due to the sampling handling. 𝛿(𝑖𝑛 𝑠𝑖𝑡𝑢 𝑍𝑟𝑂2 𝑠𝑒𝑛𝑠𝑜𝑟) 𝑉𝑆. 𝛿(𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑎𝑛𝑑 𝑠𝑎𝑚𝑝𝑙𝑒 ℎ𝑎𝑛𝑑𝑙𝑖𝑛𝑔) + 𝛿(𝑒𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑣𝑒 𝑍𝑟𝑂2 𝑠𝑒𝑛𝑠𝑜𝑟) To most effectively reduce the total error from an extractive method the possibility to measure in situ should be considered. Some of the residual oxygen measurement methods that can be applied extractive but are hard to use in situ are paramagnetic sensors and IR-absorption spectrometry14. To equally compare methods not only the sensor accuracy should be compared, but the total system accuracy. 𝛿(𝑖𝑛 𝑠𝑖𝑡𝑢 𝑍𝑟𝑂2 𝑠𝑒𝑛𝑠𝑜𝑟) 𝑉𝑆. 𝛿(𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑎𝑛𝑑 𝑠𝑎𝑚𝑝𝑙𝑒 ℎ𝑎𝑛𝑑𝑙𝑖𝑛𝑔) + 𝛿(𝑒𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑣𝑒 𝐼𝑅 𝑠𝑝𝑒𝑐𝑡𝑟𝑜𝑚𝑒𝑡𝑟𝑦 ) The same discussion can be applied to natural gas measurement applications. International standards prescribe the use of gas chromatography to determine the composition of natural gas. The precision of this equipment can be determined in the laboratory by performing a set of experiments involving standard gas mixtures. To compare Raman-spectroscopy applications to the current techniques the following equation including the sample handling should be used. 𝛿(𝑖𝑛 𝑠𝑖𝑡𝑢 𝑅𝑎𝑚𝑎𝑛 𝑠𝑝𝑒𝑐𝑡𝑟𝑜𝑠𝑐𝑜𝑝𝑦) 𝑉𝑆. 𝛿(𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑎𝑛𝑑 𝑠𝑎𝑚𝑝𝑙𝑒 ℎ𝑎𝑛𝑑𝑙𝑖𝑛𝑔) + 𝛿(𝑒𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑣𝑒 𝐺𝐶 𝑎𝑛𝑎𝑙𝑦𝑠𝑖𝑠) Naturally, in situ analysis also has negative points. Turbulent streams in the process pipeline require a proper probe design to avoid intensity fluctuations and beam steering15. The latter phenomena is caused by different densities in the gas that causes the whole spectrum to shift. Next to that, maintenance on the probe might be harder to execute, because the process is hard to access. Block and bleed valves need to be installed for safe access. In situ analysis compared with extractive techniques is a trade-off where a good analytical and low-maintenance design has the possibility to add value.. Page 10 of 116.

(11) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS 3.2.4 Sensitivity and s electivity The main challenges in Raman spectroscopy for natural gas applications are the sensitivity and selectivity for the analytes. Sensitivity to the sample depends on the total available signal, which is impaired by 2 main factors described further on. The selectivity of the method to distinguish between different analytes is essentially the result of the chemical qualities of the analytes themselves. The signal is expected to be less intense for applications on gasses compared to applications on liquids or solids due to the reduced sample density, fewer molecules will be present to interact with the light and produce a signal. Naturally the density will increase with pressure, so not only is a high pressure good for the signal intensity, also the method is inevitably pressure dependent. Some authors even report a change in the spectra at different pressure16, indicating not only the sensitivity, but also the selectivity might be affected by the process pressure. The efficiency of the Raman-effect is low when compared to fluorescence spectroscopy. The reported quantum efficiencies for Raman range from 10-6-10-8 whereas for fluorescence up to 0.8 has been reported17,18, the expected Raman signal is thus much weaker. Due to the physical nature of these limitations solutions should be found with the instrumentation, the signal will need to be carefully collected and analyzed, and not with the sample. Natural gas consists mainly of hydrocarbons which are relatively similar to each other, the main difference is their chain length and molecular weight. Vibrations in the molecules are similar for all hydrocarbons, because they consist of equal types of atoms, hydrogen and carbon. Raman analysis measures these vibrations and the selectivity is expected to be a challenge. The method should render sufficient resolution to separate the vibrations and generate adequate data. In case the selectivity proves to be overly challenging the subsequent calculations are aimed at deconvolution of the peaks, or the method will focus on the key parameters. Raman analysis may detect the complete composition with a single measurement, therefore also a lack of one component could indicate an increase in another. Composition is the ultimate goal for this analyzer, though with the calculation of the Caloric value and Wobbe index from the spectrum the analyzer should be able to work in a process control loop.. 3.3 A LTERNATIVE. MEASUREME NTS TECHNIQUES When considering the application, measurement of composition and key properties of natural gas, it also makes sense to determine the alternatives on the market. These measurements are based on various techniques, each with their own strengths and weaknesses. In the columns of Table 1 techniques are shown that are currently used or have the potential to be used for metering and regulation analysis. The rows list a number of properties following the previous paragraphs completed with key differences. Due to the virtually infinite number of possible varieties of techniques this is by no means a complete comparison. Gas Chromatography19 and Wobbe-index measurement20 can be considered wide-spread techniques. IR-spectroscopy is a recent addition to the market and the ‘Tunable Filter Spectroscopy’-application can be readily used for process regulation21. An interesting alternative to the spectroscopic methods is the combined sensor-modelling technique22, multiple signals are combined to compute key characteristics of the natural gas. Naturally also Raman-spectroscopy is listed to complete the comparison. Page 11 of 116.

(12) Christiaan Mul. SensorComputational. + + + + + +. Wobbe-index measurement. + + + + + + +. Gas Chromatography (TCD-detector). Measurement speed Sensitivity to O2, N2, H2 (Compositional analysis) Possible with complex mixtures Possible in situ measurement Linearity of signal Signal intensity Cost of ownership Static design possibilities. IR spectroscopy. Raman spectroscopy. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. + + + + -. + + + -. + + + +. Table 1: A comparison of the advantages and disadvantages of competitive techniques.. As was described in the paragraph ‘Fast analysis time’ it can make quite a difference to the process if the measurement is quick enough to regulate feed-forward. Although certain GCs can have update times of 60 seconds the average time is 3 to 5 minutes for optimal separation and analysis. All other methods have update times of less than 10 seconds, significantly faster, resulting in more data-acquisition and better process regulation. Compositional analysis can only be done with measurement methods selective to the different analytes. Only the Raman method and gas chromatography are theoretically capable of complete compositional analysis of the components in natural gas, respectively by deconvolution of individual signals and separation of components prior to detection. IR cannot measure homonuclear diatomic molecules, since there will be no shift in the dipole-moment during the vibration. An estimation can be made for nitrogen 100 % - measured % = nitrogen %, though this is obviously biased and easily influenced if hydrogen or oxygen is present. Wobbe index measurement is generally done by measurement of the residual oxygen after combustion, an indirect measurement which does not correlate to compositional differences. In a similar matter sensor-computational methods are based upon indirect measurement and are unable to provide a complete composition. The direct measurement of valuable components is a prerequisite for a full composition result. A minor concern is the possibility to measure atomic gasses, such as argon. Over the past years gas chromatography has become more expensive due to the need of inert carrier gasses. Often the solution is to exchange helium for argon, hydrogen, or nitrogen, since these gasses are more cheap even though similar resolution can be obtained. An unfortunate result is that the Thermal Conductivity Detector cannot measure the difference between the carrier gas and the analyte which is the carrier gas, for instance: with carrier gas hydrogen, the analyte hydrogen cannot be measured properly. Whereas Raman technology cannot measure atomic gasses, since there are no molecular vibrations, the possibility is expected to become gradually more expensive with chromatography. Another distinct difference of Raman technology and gas chromatography is closely linked to the ability to do a complete compositional analysis. The possibility to analyze trace components, such as dihydrogen sulfide or carbonyl sulfide, is a valuable addition. Opportunities arise when complementary analyses can be done with a single unit. Page 12 of 116.

(13) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 4 I N THEORY 4.1 T HE R AMAN - EFFECT As was described in 1928 by C. V. Raman and K. S. Krishnan23 there are 2 types of scattering events, unmodified and modified, also called elastic and inelastic respectively. Elastic scattering of light, known as Rayleigh scattering, does not influence photon energy of the incident light whereas inelastic scattering shows the changes in the molecular vibrational and rotational energy during the scattering event23. In the Jablonski diagram, see Figure 6, these scattering events are schematically drawn for clarity. The use of the words excitation and relaxation are common for scattering events even though the molecule is not truly excited to an electronic or vibrational state. A disputed so called virtual energy state is sometimes referred to, but could also be described as a short lived electron cloud polarization. Therefore, some authors prefer to use pro- and demoted.. Figure 6 Jablonski diagram showing energy states of different scattering events.. The depicted energy levels in Figure 6 are different for each molecular compound, whereby the Sstates signify various electron configurations, and the V-states multiple vibrations thereof. Figure 7 shows a –CH2– group where the hydrogen atoms vibrate around the carbon. One should note the different energy levels correspond to different modes of vibration. The depicted vibrations do not account for the remainder of the molecular vibrations, recoil of the carbon atom, or the vibrations on the rest groups for instance. These interactions make that the vibrations are slightly different depending on the weight of the atoms themselves, and the rest groups. Both the Jablonski diagram and the Molecular vibration diagram are thus merely clarifying schematic approaches of the underlying theory. Page 13 of 116.

(14) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS The measured parameter is the difference between the energy of the incident light and the energy of the scattered light. For Stokes Raman scattering this gives a positive (red-)shift, and for the antiStokes Raman scattering an equal but opposite energy (blue-)shift, hence the colored arrows in Figure 6. Both Raman shifts contain the same molecular information24. Because of the more populated ground (S0) state25, compared to the vibrational states the intensity of the Stokes Raman scattering is higher compared to the anti-Stokes Raman scattering. Therefore, often only the Stokes Raman scattered light is used for spectrum interpretation.. Figure 7 Molecular vibrations in a -CH2- group, LRTB: Symmetrical stretch, Asymmetrical stretch, Scissoring, Rocking, Wagging, and Twisting. Whereas the arrows display the initial direction of the vibrations on the plane of the paper, the + and – show the movement perpendicular to this plane.. For all analytical purposes a high S/N ratio is desirable, where the noise is mainly determined by the instrumentation. The signal is also fundamentally dependent on the wavelength of the scattered radiation. It was found by Lord Rayleigh in 1871, that the relative intensity of the signal is approximately proportional to λ-4. The correlation26 of the incident light frequency to the signal intensity is also expressed in Equation 1 where 𝜐̃0 is an intensity variable. A photon with more energy would render a higher signal, though if the incident light is too energetic fluorescence or other interfering effects might occur. Contrary to an absorbance spectrum, the shape of Raman spectra are independent of the used wavelength of excitation. The Raman shift, expressed in reciprocal centimeters (cm-1), would not change although the emitted wavelengths would27. This property of the technique makes it possible to use different excitation energies such as to prevent photo degradation, absorbance, or fluorescence from the sample. Another positive effect is that all Raman spectra taken with different light sources can be directly compared. Page 14 of 116.

(15) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS The intensity of Raman scattering is approximately a fraction of 10-6 to 10-8 compared to Rayleigh scattering28,29. Thus only a very small fraction of the light shows interaction with the sample and creates a viable signal. Overall the intensity of the scattering can be summarized in the formula given in Equation 1, see Table 2 for the applicable units and a description of the symbols.. Equation 1: Intensity of Raman scattering, equation reproduced from30. Symbol IR 𝜂 I0 n (d σ)/(d Ω) Ω Le 𝜐̃𝑅 𝜐̃0. Description Intensity of Raman effect The experimental factor intensity of the incident light particle density differential Raman cross section collection optic angle effective length of the sample cell Wavenumber Raman scattered light Wavenumber incident light. Unit J/s No unit J/s cm-3 cm2/sr sr cm cm-1 cm-1. Table 2 Definition of symbols and units in Equation 1. The intensity of the Raman scattering is a direct function of the energy of the incident and scattered light, as described above. Other parameters are: 𝜂, the experimental factor for the yield of the experimental setup, I0, the intensity of the incoming light (laser power), n, the particle density, (dσ)/(dΩ), differential Raman cross section of the analyte31, Ω, collection optic angle, and Le, the effective length of the sample cell30. Some of these parameters are physical constants, other can be influenced by the experimental method and setup.. Page 15 of 116.

(16) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 4.2 N ATURAL. GAS IN DETAIL Natural gas is a complex mixture and can be found with several impurities. The majority of the gas consists of methane, and depending on the source or means of transport the other components can be hydrocarbons or inert components. A limited overview of the expected components in natural gas from Groningen and Qatar can be found in Table 3. Bio-gas usually has a higher sulfur impurity content due to its biological background. These compounds do not only hinder the comfortable consumer usage, but their accompanied acidity may result in corrosion of wetted parts32. Most notably bio-gas contains close to 100% methane, more than any other type of natural gas because it is made in methanogenesis33. Nitrogen is another component present in natural gas, gas fields in Groningen are reported to contain on average more than 14 %34. In contrast, Liquefied Natural Gas (LNG) contains due to its physical nature practically no nitrogen. The gas is liquefied to approx. -163 °C, at which nitrogen remains gaseous and can be separated. Because nitrogen does not burn the relative quantity has consequences for the fuel quality. Common names: Methane Ethane Propane iso - Butane normal - Butane. C1 C2 C3 C4 C4. Molecular formula: C1H4 C2H6 C3H8 C4H10 C4H10. iso - Pentane C5 normal - Pentane C5. C5H12 C5H12. neo - Pentane Hexane. C5 C6. C5H12 C6H14. Nitrogen Carbon Dioxide Hydrogen Oxygen. N2 CO2 H2 O2. N2 CO2 H2 O2. Groningen 81.30% 2.85% 0.37%. Qatar 88.2% 6.1% 2.3%. 0.14% C4 lumped. 1.0% C4 lumped. 0.04% C5 lumped. 0.0% C5 lumped. 0.05% C6+ lumped 14.35% 0.89%. 2.5% 0.0%. 0.01%. Table 3 main components in sales gas, with composition in volume % for NG from Groningen34 (Wobbe index of 43,7 MJ/m3) and LNG from Qatar (mixed to a Wobbe index of 54 MJ/m3 for the Dutch market). Fuel quality can be expressed in a number of ways depending on the operation. Most commonly the superior calorific value35, expressed as MJ/m3, is used for transport and cost calculation. Other expressions are the methane number36, expressed in the methane/hydrogen mixture knocking equivalents37, and the Wobbe index35,38, expressed in MJ/m3. These units are valuable regulation parameters in gas fueled engines and furnaces. For the composition of the mixture to be analyzed it is important that either the components are separated (as is done in a GC) or the signal can be separated. Several applications are known to analyze mixtures up to C3 and separating the Raman signals11,12,39,40. Other analysis methods, discussed in 3.3 Alternative measurements techniques, determine these key factors directly without use of the composition. Page 16 of 116.

(17) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS 4.2.1 Literature spectra Table 4 displays a list consisting of the analytes, their measured Raman shift, and the difference between consecutive peaks. The list immediately shows the difficulty of the sample, many of the components are alike. The CH-stretch vibrations, around 2900 cm-1, are similar for all carbohydrates.. Components unordered Raman shift (cm-1) Methane Methane Ethane Ethane Propane Propane iso - Butane iso - Butane normal - Butane normal - Butane Nitrogen Hydrogen Hydrogen Oxygen Carbon Dioxide Carbon Dioxide. 2917 1535 2914 993 2908 870 2880 794 2890 827 2331 4156 587 1555 1388 1285. Components ordered after Raman shift Raman shift Difference -1 (cm ) between peaks (Δcm-1) Hydrogen 587 iso - Butane 794 207 normal - Butane 827 33 Propane 870 43 Ethane 993 123 Carbon Dioxide 1285 292 Carbon Dioxide 1388 103 Methane 1535 147 Oxygen 1555 20 Nitrogen 2331 776 iso - Butane 2880 549 normal - Butane 2890 10 Propane 2908 18 Ethane 2914 6 Methane 2917 3 Hydrogen 4156 1239. Table 4 Raman shift (cm-1) of most common components, data reproduced from Kiefer et al. (2008)11. The literature values in Table 4 are likely to only relate to the main isotopes, 1Hydrogen and 12 Carbon. Other isotopes would change the vibrations since the weight would alter the oscillation properties. In Appendix C some calculations are noted for the approximation of isotopic influence on the vibrational frequency from a classical mechanical point of view. The results shown in Table 5 are for the C-H stretch vibration from methane and show two important trends. Firstly the change of 12Carbon to 13Carbon shows a small decrease in vibrational frequency. Hereby it should be noted that the calculation does not adjust for any other bonds of the carbon atom. For methane more influence would be expected than for propane, since the carbon in propane has heavier side groups. From this data the isotope effect is expected to be only a minor effect and, depending on the instrumental resolution, might be seen as minor broadening of the signal. Secondly the results show that in case deuterium is present, the vibration would be less energetic. Such a shift would show the peak on a different place in the spectrum, though it would be hard to measure these vibrations. 2H has a natural occurrence of 0.015 %, which would mean for the hydrogen rich molecule hexane 0.2% of the signal would be changed in such manner. Shortly, the deuterium shift is not expected to be a major signal interference or cause of a lost signal.. Page 17 of 116.

(18) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Bond components Hydrogen Carbon12 Hydrogen Carbon13 Deuterium Carbon 12 Deuterium Carbon 13. Wavenumber (cm-1) 2917 2908 2142 2130. from literature Calculated Calculated Calculated. Table 5 Results of isotope vibrations approximation.. A difficulty for this application is the gaseous phase of the sample. Raman spectroscopy on gasses is not new41,42, though it is a developing science. Certain effects are to be expected with increasing pressure. Firstly, a higher signal is expected. The signal intensity equation, Equation 1, showed that the particle density of the sample is proportional to the signal30. For the same reason scientists have been building high pressure sample cells43,44. Secondly, not only the spectrum intensity changes, also the peaks can shift with pressure45. Molecules are more or less dispersed in the sample, when the density changes they have more or less interaction with each other, thus influencing the shape of the spectrum43,44. For methane-ethane mixtures this can be a good thing, for instance a built in pressure measurement46.. 4.2.2 Theo retical predictio n o f sp ectra As a part of the research the theoretically predicted vibrations were calculated with the use of Amsterdam Density Functional, ADF, software. The program is based on Density Functional Theory and can be controlled with a Graphical User Interface, GUI, where the user can build the applicable molecule(s) and select the correct calculation mode. There was no in-depth analysis of the used algorithms, a description of the main computation is given in Van Gisbergen et al. (1990)47, only the setup and results were operated to analyze natural gas. The first step in using the ADF-software is building the molecules, atom by atom, in separate job files. The molecules built to calculate their individual spectra are shown in Table 3. Consecutively the geometry of the molecules is optimized with the built in algorithm, this is important due to the symmetry in the molecules. When the built molecule does not have an optimized geometry symmetric vibrations are not properly calculated. The last step in the calculation setup is the calculation mode selection and the input of laser frequency. Calculation preset ‘frequencies’ was selected and modified to include the full Raman prediction and a photon energy of 2.3305 eV calculated from a laser wavelength of 532 nm. Eventually all vibrations were calculated, both IR and Raman active, a typical Raman spectra output is shown in Figure 8.. Page 18 of 116.

(19) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 8 ADF-Calculation output for methane. In case the symmetry is entered correctly into the software, apart from the frequencies also the vibrational symmetry, degeneracy, Raman intensity, and linear depolarization ratio is calculated. Linear and non-linear polarization47 is not applicable to this project, and is further ignored. For this study only the Raman active vibrations were collected and used. Peak data is extracted from the ADF software and loaded into MatLab™ for further calculations. From the peak frequency and intensity a normal distribution is made in which the mean is the peak frequency, sigma = 10, and the total area is the intensity. These distributions are added up to plot the spectra in Appendix A, where the separate component spectra and peak data are collected. For illustration Figure 9 is shown, all separate peaks are summed and displayed as a theoretically expected spectrum. It was attempted to calculate such parameters for a mixture of hydrocarbons. These jobs would have similar settings, although with multiple methane molecules built with the GUI. Unfortunately these simulations proved to be too computationally intensive for the used computers and had to be abandoned.. Page 19 of 116.

(20) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 9 Theoretically predicted spectrum of methane, data from ADF plotted in MatLab. A comparison of the literature data with the predicted data shows certain discrepancies. Firstly the carbon dioxide peaks from the literature on 1285 and 1388 cm-1 were not expected. The main peak from the calculations is predicted as a symmetrical stretch vibration with wavenumber 1182 cm-1. The wavenumber shift and doublet formation is likely a result from Fermi resonance48,49. Secondly the comparison of the hydrogen literature and predicted values show an interesting difference. Where the theory only predicts a single vibrational mode ( 3N-5 for linear molecules), two peaks are reported50, at 587 and 4155 cm-1. The first a rotational Stokes shift, and the latter a vibrational shift. In Figure 10 the predicted spectra are combined in an overlay view wherein the component spectra are multiplied by their mole-fraction. The largest contribution to the signal originates from methane, since it has the highest typical concentration in Groningen gas. More detailed overlay spectra, not-concentration corrected, or corrected for typical Qatar gas can be found in paragraphs 11.2 and 11.3. Lastly a comparison can made from the concentration corrected spectra from Groningen and Qatar, the two main gas sources in the Netherlands from own production and LNG import. Figure 11 shows the absolute theoretical difference when the one is subtracted from the other. It can be seen that there is a difference, approximately 10 % change of the signal forms the full change in concentration, and thus with the correct sensitivity and resolution it can be measured.. Page 20 of 116.

(21) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 10 Overlay of Theoretically predicted spectra from the components in Groningen-gas multiplied by their typical concentrations.. Figure 11 Theoretical absolute difference of spectra of natural gas from Groningen and Qatar, one subtracted from the other.. Page 21 of 116.

(22) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 5 I NSTRUMENTATION OF THE EXPERIMENTAL SETUP When it comes to consideration how to apply Raman spectroscopy we have to ask the question: What do we consider to be essential feats for Raman spectroscopy for natural gas process applications? Safety first, the use of high power lasers and flammable gasses is a combination to consider. Secondly the signal is expected to be low, so all needs to be done to gather as much light as possible. In this chapter the experimental setup is discussed in the same order the light travels.. Figure 12 Schematic view of the experimental setup with the three main components; laser, optical probe and spectrometer.. Figure 12 shows the used optical setup used for the practical experiments. There are three main components selected for this application. The laser, a Cobolt 04-01 series51,52 Samba™, can be set to up to 156mW output on 532nm wavelength. The Optical probe, an InPhotonics Ramanprobe™53 incl. Reaction Ramanprobe™54. Multiple spectrometers from different manufacturers were used, to compare which one would perform the best. These might be the main components, but their components will be discussed in detail to attempt optimization. Page 22 of 116.

(23) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 5.1 L IGHT. SOURCE , L ASER Light sources can be characterized upon many properties including; spectral purity, emitted wavelength, and the power stability. For the Raman application several characteristics are important but foremost the light source should provide a single well defined wavelength, so called monochromatic light. Monochromatic light can be made with the use of a broadband light source and various filters41, as Raman did in his original experiment23. In the last few decades, lasers, Light Amplification by Stimulated Emission of Radiation, LASER, have become widely available and are commonly used as a light source in Raman spectrometers due to their inherent properties, and single wavelength emission. A narrow emission bandwidth of the laser is directly relatable to the resolution of the spectrum. In a previous paragraph it was explained that the wavelength of Raman radiation is relative to the initial excitation wavelength. A broad excitation band would result in a broad banded spectrum, with low resolution. Generally the emitted peak is characterized as the spectral linewidth in FWHM. The selected laser, a Cobolt 04-01 series Samba™, has a spectral linewidth (FWHM) of < 1 MHz52. From a calculation over the coherence length it follows the linewidth (FWHM) is <1 * 10-6 nm, much exceeding the separation of the expected peaks. The spectral purity of the laser is defined as > 60dB. This measure defines the intensity of the emitted wavelength relative to the non-lasing emission, the noise. Although this comprises an important property, this unit is not always given in the datasheets. Nd:YAG, a type of Diode Pumped Solid State Laser, can be frequency doubled to change the emitted wavelength from 1064 nm to 532 nm. Generally speaking, these resonant non-linear coupled lasers are known to have high spectral purity55–57 and can be utilized for spectroscopic applications. The selected wavelength has to be determined per application. A shorter laser wavelength would increase the Raman intensity, and some studies try to utilize a deep-UV laser58,59,60. Such UV-range lasers often lead to high fluorescence or photo decay of the analytes61. Additionally UV lasers have a safety drawback since the light is harmful to but cannot be seen by the naked eye. Multiple applications are based upon visible lasers with a wavelength (𝛌) of 532 nm62,11,40. Longer wavelength lasers would generate Raman signals with even lower intensities and a longer wavelength63,64, but less fluorescence is expected. The excitation wavelengths are part of a trade-off for intensity, photo degradation, and fluorescence. As was discussed in paragraph 4.1 not only the signal intensity, but also the signal wavelength is dependent on the excitation wavelength. With the literature values of the Raman shift then an estimate can be made of the expected spectrum from a wavelength point of view, which is important when considering the necessary resolution of the spectrometer. When a shorter wavelength is chosen the resulting signals will become more compressed. The aimed for application, in process measurement of natural gas, does not suffer from photo degradation. In the process the sample can be continuously refreshed by new sample, thus when it would occur it is expected to have a small effect on the spectrum. Fluorescence could provide some issues, sales gas is cleaned from impurities, whereas raw gas from the source may contain aromatic compounds or metal-complexes. Based upon the results found in literature made with a 532 nm laser, and the expected resolution, fluorescence, and safety considerations a 532 nm emission wavelength is selected for the experiments.. Page 23 of 116.

(24) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS Additionally the emitted wavelength of the laser should be as stable as possible. Many diode lasers are temperature dependent for their exact emission wavelength, a minor temperature shift will result in a change of the spectrum, the peaks will shift equally to the laser wavelength. For stable operation it is therefore necessary the emission wavelength does not drift, or that the drift is known to establish a proper calibration interval. The selected laser has a wavelength stability of 2 pm over ± 2 °C and 8 hours52, showing the importance of temperature control. Another characteristic that determines if a laser can be used for an application is the laser power. As was described in Equation 1, the Raman intensity is directly proportional to the light intensity of the excitation beam. Not only should the light intensity be powerful enough to render a significant signal, also it should be stable in time to prevent the intensity of the signal to vary. In literature applications with pulsed lasers can be found58,63,65, these applications rely interpretation of a single spectrum. The aimed for application should be stable over longer periods of time, where preferably all spectra are of equal intensity. For this reason a continuous wave laser was selected for these experiments. The use of lasers in an environment with flammable gasses is strictly regulated66. Limits are set for focusing beams, and power per irradiated area. Therefore the laser power should be well controlled and as low as possible, while maintaining a proper spectrum, though not less than one-third of its maximum power for stability. A well-defined irradiated area also adds to the safety of the setup. The 150 mW laser was selected and used on max power, 156 mW, for the experiments, though if possible to be changed for a less powerful source in the future. For this study the factory calibration of the laser internal power meter was controlled with a Coherent FieldMaxII-TO power meter67. Figure 13 shows the streaming data during the experiment setup, hereby it was attempted to mount the optical probe in front of the power sensor with the highest efficiency. Later by adjusting the laser power from the laser software, see Figure 14 for a typical display, the internal laser power limiter could be compared to the external power measurement. This measurement showed lower power on the external measurement, set 156 mW, measured 151.7 mW with RSD of 0.19%. The difference in intensity is likely the result of the laser to fiber and fiber to power meter interfaces. The verification of the internal power meter showed the set power and internal power measurement can be considered true.. Figure 13 Display of Coherent FieldMaxII-TO power meter software during setup of Laser clamp. 1 measurement per second.. Page 24 of 116.

(25) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 14 Typical display of Cobolt Samba laser control software.. 5.2 E XCITATION. FIBER Both the laser and the optical probe were selected to be pigtailed, equipped with permanently mounted optical fibers. Hereby the chance to erroneously connect the fibers or spend too much time on optimizing the interfaces is decreased. Although a single fiber connection sounds simple enough, special attention was given to the type of the connector and the size of the fiber core. Optical fiber, sometimes called glass fiber, can be made in many different types and sizes. it is important to note these fibers guide the light through the core based on total internal reflection. This principle, based upon the refractive indices of the inner core and the cladding around it, traps the light in the core allowing it to advance over great distances. Optical fibers can be optimized for the type of light propagating though the core, there are fibers for single mode light, polarized light, and multimode light. Also there are fibers optimized for a specific light wavelength by doping the glass with rare earth elements or specialized coating. Page 25 of 116.

(26) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS The size of the fiber should be such that the collimating of the laser light can be done without significant loss of intensity. Also the smaller the fiber, the harder it becomes to properly align the core for collimating the light into a beam. Practically the lowest core diameter for an excitation fiber is around 50 µm, smaller cores have a too small numerical aperture for effective use. In this study a 100 µm core multimode fiber was used for the excitation fiber, both on the laser and the optical probe. For scientific purposes the main fiber optic connectors are SMA, ST, and FC/PC, which are only a fraction of the commercially available choices as can be seen in Figure 15. The remainder of the connectors is mainly used for telecommunications, hence also some duplex connectors. The FC/PC, Ferrule Connector / Physical Contact, connectors are spring loaded and appropriate to be used in high vibration surroundings because they will maintain a stable pressure on the contact surface. To prevent light reflecting back from the optical probe into the laser cavity the connector can be polished at an 8 degree angle, a so called APC, Angled Physical Contact, connector. That is not possible with an ST connector since they are not key-aligned. The used connector for the coupling of the laser to the optical probe is a type of, an APC connector which has an 8 degree Angled Physical Contact.. Figure 15 Various types of connectors, edited from source68.. 5.3 O PTICAL P ROBE Optical probes are fundamentally different from sample probes. The purpose of a classic probe is to sample representatively and maintain the integrity of the sample for transport to the analyzer. Optical probes are not made to transfer the sample, but the signals from and to the analyzer69. Sample probes are not part of this research, from this point on all referenced probes are optical probes. Although optical process probes become more commercially available, robust optical probe for the use in field applications remain relatively scarce. Optical probes can be roughly categorized in two groups; Fiber Optical Probes (FOPs) and Lens Collection Array (LCAs). Note these descriptions and abbreviations are self-coined, by lack of alternatives. Immersion probe70 is a coined concept, an optical probe that can be submerged in a liquid or gaseous sample. Though this only means the optical components are protected with a (mostly sapphire) window71,72, it does not give any information about the optics itself. Furthermore also fiber optical probes can be submerged in compatible samples73. Therefore: FOPs and LCAs for definition see below.. Page 26 of 116.

(27) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS Fiber Optical Probes (FOPs) are highly modifiable74–76, as Figure 16 illustrates. The excitation fiber can be modified in width, as can also the collection fibers. The number of collection fibers and their distance to the excitation fiber can be altered. Also the tip can be beveled, which allows for a higher collection angle and a better overlap of the excitation and collection cones, thus a more effective collection of scattered photons. A typical challenge associated with FOPs is cross-talking, when the excitation and collection fibers show interference77,73.. Figure 16 Various Fiber Optical Probe Configurations. Left to right: Single-fiber with dichroic mirror, two fiber flat tipped with separated excitation and collection fiber, six around one flat tipped, single fiber with dichroic mirror and lens, two fiber beveled with separated excitation and collection fiber, and six around one beveled tip. After Cooney et al. (1996)78.. Lens Collection Arrays (LCAs) are another way to collect scattered light for further processing41,42. The LCA can be mounted on a breadboard and because of its adjustability it is often used in nowadays research setups65,79. Combinations of optical fibers with LCA-probes can also be found80 and provide an excellent structure for integration of a Raman instrument into a process environment. Due to special interest the extension tube with the lens is discussed separately. For the experiments a LCA from a commercial party, InPhotonics RamanProbeTM,53 in combination with the InPhotonics Reaction RamanProbeTM,54 was used based on their selection of associated components, see Figure 17.. Figure 17 Component overview of the factory standard optical probe, copied from80.. Page 27 of 116.

(28) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS 5.3.1 Collimation lenses When the light arrives with the excitation fiber the first step would be to modify the light into a parallel beam, to collimate. The benefit of this would be that consecutive treatment of the light becomes a more standard beam setup with efficient light coupling. The last step in the probe would be to converge the Raman signal from a parallel beam into the collection fiber. For the excitation fiber a 100 µm core fiber was used, and for the collection fiber a 600 µm fiber. These increased core diameters also increase the numerical aperture and influence the collimation properties. The excitation light has to be collimated, measurement of the beam showed a beam diameter of approx. 2,5 to 3 mm. This was done by pointing the beam toward a measurement paper without a lens. The beam diameter is influenced by the NA of the fiber, and the distance to the collimating lens. For the excitation fiber a small NA is preferred to diminish reflections into the fiber. The ideal beam diameter is based upon the laser power and the LIDT, Laser Induced Damage Threshold, of the optical components. The signal collection fiber is used to transport the Raman radiation to the spectrometer. To optimize the signal strength the convergence is done into a fiber with a large diameter core and high NA. Different to the excitation light collimator the convergence lens has multiple wavelengths from each signal to diffract. Diffraction of light is wavelength dependent and the efficiency of the collimation may differ per wavelength. For some applications a reflective collimator, or GRIN lens may be used to prevent wavelength dependent deviations. A large core diameter may help to accommodate the focal length shift from a chromatic collimating lens. Further increment in the core diameter would lead to issues on the other side of the fiber where it has to be coupled to the spectrometer.. 5.3.2 Filters and mirro rs A few filters are used in the experimental setup; a bandpass filter, a long pass filter, and a dichroic mirror. The selection of these filters was done by the supplier of the probe, the below description and knowledge was found in an attempt to self-build an optical probe, see Figure 18. The development of the alternative probe did not follow through due to the foreshadowed miniaturization and production issues. The theoretical evaluations of components provides valuable insight into the operation of commercially available models.. Figure 18 3D rendering of an alternative optical probe with collimating lens, optical filters, long pass dichroic mirror, parabolic collimating mirror, and collection fiber.. Page 28 of 116.

(29) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS General considerations apply when working with optical filters that may be evident. Filters are fragile, for laser applications it is important to know the LIDT to prevent the laser burning the top layer. Filters have an optimal orientation with respect to the light beam, firstly to diminish reflections, secondly, some types have a recommended light direction as well. Dielectric filters change their transmission maximum with temperature due to the expansion of the dielectric layers, a temperature log may be recommended. Filters are evaluated on their transmission, which cannot be 100%, too many filters even in the transmitted range will smother the light intensity. With their remaining properties optical filters make a powerful method to clean up both the excitation source and the signal to gather a proper spectrum. A bandpass filter can have multiple functions within an experimental setup, in definition the filter only has transmittance for a specific wavelength range. The main function of the filter in this setup is to clean the laser excitation light from interfering signals, fiber Raman signals, hence the alternative name ‘laser clean-up filter’. The interference signals are considered to come from impurities in the core of the fiber itself, since the spectral purity from the laser should be sufficient. The quality of a bandpass filter should be defined with at least these parameters: Central wavelength of transmission, FWHM of the highest peak, the transmission of this peak, and the blocked wavelengths. By installing this filter the quality of the excitation light can be ensured. To clean up the Raman signal from interfering light multiple filter types may be used. a notch filter could be used to filter the laser light out to prevent detector saturation and reduce background signals63, one would then be able to look at both the Stokes and the anti-Stokes shift. Alternatively, if the application is known and only a small band of the spectrum is of interest this band could be passed on while the rest is filtered out by the use of a bandpass filter81. This principle is sometimes also used in specific setups where the spectrograph is exchanged by a filter wheel and an intensity measurement. High and low pass filters are easily available and commonly used to clean up the signal. These filters can eliminate Stokes or anti-Stokes Raman scattering together with Rayleigh scattering from the laser82,83. In this experiment a long pass filter was used for cleanup of the signal.. Figure 19 Short and long pass dichroic mirror, the colors of the arrows indicate the relative wavelength of the light beams. Figure 17 shows the use of a short pass dichroic mirror in the optical probe to separate the laser light from the Raman signal. Dichroic mirrors are characterized by their cut off wavelength and their type, short or long pass, shown in Figure 19. Depending on the quality of the mirror the cut of wavelength can be sharper, or the transmission or reflection higher or lower. The use of a short pass dichroic mirror has the advantage that the laser light can only reach the signal pathway by accidental reflection. Page 29 of 116.

(30) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 5.4 E XTENSION. TUBE , IMMERSION PROBE The concepts extension tube and immersion probe are similar and generally only differentiated based on the sample phase. Figure 20 shows a collection of immersion probes with and without further optical components. The primary function is to separate the optics, such as filters and mirrors, from the sample and provide a surface for a leak tight fit. The second function of these tubes is to extend the reach in narrow places, or submerge the probe entirely. During these experiments various tubes have been tested whereby certain opportunities have become evident, an alternative is proposed and evaluated.. Figure 20 Various commercially available (optical) probes and immersion tubes. Left top Immersion probe made by Solvias84, Right top AirHead™ Gas-phase Raman Probe made by Kaiser Optical Systems inc85, Left bottom Bioprocess in-line Raman Analyzer (probe only) made by Resolution spectra Systems86, Right bottom Fiber Optic Raman Probes made by Wasatch Photonics87.. To reach the optimal in situ place for process measurement some sort of extension tube will be necessary. Figure 21 shows a cross section of a process pipeline, A, a typical flanged probe tie in point is mounted upon a nozzle. From, C, the flow through the pipeline that flows with a laminar flow profile, it follows that the probe tip has to be closer to the middle than the side. A rule of thumb is the sample should be from the middle one-third of the process line. B, the ideal probe length is thus the sum of the tie in point nozzle length and one third of D, the pipe diameter. Unfortunately, the length of the extension tubes is limited because the signal worsens with increasing length, a simple extension to the ideal length is therefore not possible. Page 30 of 116.

(31) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. Figure 21 Schematic cross section of a process pipeline showing: A Flanged tie in point, B Ideal probe length, C Laminar flow profile, D pipe diameter.. The probe will have to be tough enough to withstand the stress and strain of the process conditions. Wake frequency calculations are aimed to simulate both the static and the dynamic stress on the proposed structure to determine if this is within permissible limits88. These calculations have not been supplied or made as a part of this study for any of the extension probes. It is recommended to assess the final structures before process implementation. To effectively address the application’s need, an alternative extension tube is proposed, as can be seen in Figure 22. The new design minimizes the length of the extension tube and mounts the optical probe at the tip. First the alternative was modeled in 3D modelling software, Solidworks 2017, and then theoretically further evaluated with Zemax OpticStudio 16.5.. For information about this part of the document, please contact Analytical Solutions and Products bv.. Figure 22 Impression renderings of designed alternative probe.. Page 31 of 116.

(32) Christiaan Mul. RAMAN SPECTROSCOPY FOR NATURAL GAS PROCESS APPLICATIONS. 5.5 L IGHT. PATH GEOMETRY All probes from Figure 20 use a backscatter geometry and are equipped with a lens at the probe tip. The function of the lens is to focus as much light as possible on a single point making this the most probable location for the Raman-effect to occur. The lens is also used to collimate the Raman signal back into the extension tube in direction of the optics. For an achromatic lens both focus points would exactly overlap, since they are corrected for wavelength dependent diffraction. Regular lenses would show chromatic aberration, see Figure 23, where the signal is diffracted differently than the excitation light. Other studies reported besides backscatter77 also 90 degree11, or a free collection angle63.. Figure 23 Lens at the tip of the extension tube, light path illustrates chromatic aberration over the length of the extension tube.. Optical simulations in Zemax Optical Studio 16.5 were performed to evaluate the backscatter geometry. Figure 24 shows the simulated experimental setup, certain aspects are altered for the experiments to test their influence. Main components are detector 1 (1 mm behind lens) and detector 2 (30 mm behind lens), together with the light guides they simulate the extension length. Unless otherwise specified, a light source a 532 nm unidirectional point source with one million rays was simulated, and scattering rays are set off. The location of the point source is in the focal point of the lens at 532 nm calculated in the sequential mode. To simulate different scenarios, selected components were changed or altered to investigate its influence. Chromatic aberration, together with optical alignment, is one of the main reasons the extension length of the probes is limited. By reducing the extension length in the alternative design as much as possible an attempt is done to counteract the effects. Another applied solution is the use of bigger lenses with different curvatures. Most extension tubes use 4 to 7 mm lenses, for the alternative design a 12.7 mm diameter lens was selected. From Snell’s law we know chromatic aberration is most severe where the diffraction angle is highest, so in case mainly the middle of the lens is used, where the angles of refraction are smallest, the effects are minimized. A simulation was done to estimate the degree of chromatic aberration. To do so first the focal point of the lens was calculated in the sequential mode when irradiated with a 532 nm light source, such as when the laser is focused on a point in the sample. In the Non sequential mode a point source with a wavelength of 630 nm (methane Raman signal) was simulated on this distance from the lens. point source from the 532 nm. It was shown that with a lens of 12.7 mm diameter the chromatic aberration was negligible. Page 32 of 116.

No results found