Spectroscopic monitoring of gas in liquid

Abstract

There is described a spectrometer for analysis of a gas species within a liquid, comprising: one or more optical sources configured to generate first probe light at first wavelengths and second probe light at second wavelengths, the first wavelengths including a wavelength of an absorption feature of the gas species in the liquid and the second wavelengths offset from the absorption feature; a controller arranged to sequentially trigger the probe light; a sample interface for interfacing the probe light with the liquid comprising the gas species; sample beam path optics arranged to guide the probe light through the sample interface to a first photodetector; reference beam path optics arranged to guide the probe light along a reference path to a second photodetector without passing through the sample interface; an analyser arranged to provide a quantitative determination relating to the gas species in the liquid based on signals received from the photodetectors at the first wavelengths and the second wavelengths.

Description

[0001] Spectroscopic monitoring of gas in liquid

[0002] The present invention relates to a spectroscopy apparatus and method of quantitative analysis of the amount of gas within a liquid. The spectroscopy method uses a dual band technique for accurate determination of a gas held within a liquid. Embodiments are particularly applicable to measuring carbon dioxide in liquid water.

[0003] Introduction

[0004] Methods of absorption spectroscopy are well-known and commonly use the absorption of electromagnetic radiation in the infrared and ultraviolet bands to determine the presence of, and amount of, a particular substance in a sample. Such techniques are commonly applied to liquids, solutions, gases etc. However, measurement of the amount of a gas in a liquid is more difficult.

[0005] Climate change has resulted in a number of novel approaches to be tried in order to reduce the amount of greenhouse gas such as carbon dioxide being emitted in to the atmosphere. Capturing the carbon dioxide and storing has also become desirable. The carbon capture and sequestration industry requires trustable and accurate tools for determining the amount of greenhouses gases being emitted by an activity, a facility or a process, or how much is being removed from the atmosphere to be sequestrated.

[0006] For some techniques greenhouse gases are held within a liquid matrix, which means that traditional gas phase measurement methods are no longer valid. Detecting, identifying and quantifying a gas with a liquid matrix is challenging. Industrial applications also require a capability of fast measurements taking only a few seconds to provide an output, as part of monitoring live (online) running processes. Accuracy, selectivity and insensitivity to interfering chemicals are also important requirements to enable trust and transparency of analysis, even when complex liquid mixtures are used.

[0007] Summary of the invention

[0008] The present invention provides apparatus and methods for quantitative analysis of gas within a liquid such as in solution. The apparatus and methods exploit the signature of gas vibrational bands with a liquid, for example CCkor CH4in water, by measuring on- resonance and off-resonance absorption using two offset wavelength ranges, such as in the mid-infrared. Other measurements may be made by the system to eliminate environmental and instrument factors.

[0009] 17257738.MDE.MDE In particular, the present invention provides a spectrometer for analysis of a gas species within a liquid, the apparatus comprising: one or more optical sources, the one or more optical sources configured to generate a first probe light at first wavelengths and second probe light at second wavelengths, the first and second probe light may be an output or beam, the first wavelengths including a wavelength of an absorption feature of the gas species in the liquid and the second wavelengths offset from the absorption feature; a controller arranged to sequentially trigger the first probe light then the second probe light, or vice Versa; a sample interface for interfacing the first and second probe light with the liquid comprising the gas species; sample beam path optics arranged to guide the first and second probe light through the sample interface to a first photodetector; reference beam path optics arranged to guide the first and second probe light along a reference path to a second photodetector without passing through the sample interface; an analyser arranged to provide a quantitative determination of a gas species in the liquid sample or provide an absorption spectrum of the gas species within the liquid based on signals received from the first and second photodetectors at the first wavelengths and the second wavelengths. The absorption feature of the gas species in the liquid is preferably a vibration absorption band such as arising from the gas species being in the liquid. That is, the usual well-resolved gas-phase vibration and rotation bands are not observed in solution because the gas species are solvated or surrounded by the molecules of the solution, strongly perturbing the vibrational and rotational modes. This broadens the individual lines to the extent that they are no longer resolved.

[0010] The measurements by the second photodetector may be used to measure and account for the presence of the gas species in air in the instrument which provides some absorption included in the measurements as a result of the probe light passing through air to the sample. The off-resonance measurements may be used to measure and account for absorption due to the liquid such as water. Off-resonance measurements by the first photodetector also take account of the effects and variations of the sample cell, such as any path length variation. The use of two lasers allows changes in laser characteristics, such as thermal drift, to be monitored and taken account of. Once measured, these effects may be removed from the sample measurement. By taking account of these effects, the requirement for calibration is reduced or eliminated.

[0011] The first wavelengths may comprise a wavelength of peak absorption of the gas species in the liquid, and the second wavelengths are spaced apart or offset in wavelength from the peak absorption of the gas in the liquid, such as by at least 5cm'1or more preferably by at least 10cm'1, at least 20cm'1, at least 30cm'1or at least 40cm'1.

[0012] 17257738.MDE.MDE The spectrometer may further comprise a beamsplitter arranged to split the first and second probe light and direct a first portion to the sample interface and a second portion to the reference path.

[0013] The one or more one optical sources may be configured to scan or sweep the first and second probe light across the first wavelengths and second wavelengths respectively. Alternatively, the one or more optical sources may be broad band sources configured to simultaneously output all of the first wavelengths and the second wavelengths and the photodetector comprises a filter arranged to scan across the first wavelengths and second wavelengths.

[0014] The one or more optical sources may comprise a first optical source for generating the first probe light at first wavelengths and a second optical source for generating the second probe light at second wavelengths. Alternatively, the one or more optical sources may comprise a single optical source configured to generate multi-band light, that is, the single optical source may be configured to generate both the first probe light at the first wavelengths and the second probe light at the second wavelengths.

[0015] The first and second optical sources may be tuneable laser sources or broadband sources.

[0016] The first wavelengths and second wavelengths may be in the mid-infrared, such as between 2 and 20 pm. For example, for measuring CO2 the wavelengths may be at wavenumbers between 2000 and 2500 cm'1, for measuring methane the wavelengths may be around 1200cm'1such as between 1000 and 1500cm'1or for measuring ozone the wavelengths may be around 1000cm'1such as between around 750 and 1250 cm1. Other wavenumber ranges may also be used, such as for measuring other species.

[0017] The reference beam path optics may be configured to guide the first and second probe light to the second photodetector along an atmospheric pathway through the apparatus. An atmospheric pathway is a pathway that is open to atmospheric or local environmental conditions.

[0018] The reference beam path may include a reference gas cell, such as a low pressure gas cell for fine laser frequency calibration.

[0019] The sample interface may comprise a sample cell or an ATR probe or crystal.

[0020] The controller may be configured to trigger the first probe light and the apparatus is configured to detect signals at the first photodetector and second photodetector. The controller may also be configured to trigger the second probe light and the apparatus is configured to detect signals at the first photodetector and second photodetector.

[0021] 17257738.MDE.MDE The analyser may be configured to receive: a first signal based on the first probe light incident on the first photodetector, the first signal indicative of absorption at the gas species absorption feature wavelengths of the gas species in the liquid sample; a second signal based on the first probe light incident on the second photodetector, the second signal indicative of absorption at the gas species absorption feature wavelengths along the reference path; a third signal based on the second probe light incident on the first photodetector, the third signal indicative of absorption of the sample at wavelengths offset from the gas species absorption feature wavelengths; and a fourth signal based on the second probe light incident on the second photodetector, the fourth signal indicative of absorption at the offset wavelengths along the reference path. In embodiments, more than four signals may be used.

[0022] The analyser may be configured to generate transmission and / or absorption spectra based on any one or more of the first to fourth signals.

[0023] The analyser may be configured to generate the transmission and / or absorption spectra by performing a wavenumber (or wavelength) assignment or calibration based on known absorption lines of the gas species.

[0024] The analyser may comprise a processor and memory configured to perform steps of: generating a forward model based on spectroscopic data of a reference liquid sample measured at the first wavelengths and the second wavelengths and the reference path measured at the first wavelengths and the second wavelengths. The forward model may be configured to estimate spectra and / or provide a quantification of species in a sample based on an amount of the gas species in the liquid sample. The analyser may be further configured to invert the model to generate a quantification of gas species in a sample based on measured spectra. The spectroscopic data of a reference liquid may be known data such as found in literature or measured during manufacture or development of the analyser. In this way the analyser does not require calibration with samples of known concentration during operation.

[0025] The analyser may be configured to determine a quantification of gas species in the liquid sample based on using the second to fourth signals to subtract from the first signal absorption due to water in the sample and the gas species in air.

[0026] The present invention further provides a method of spectral analysis of a gas species within a liquid, the method comprising: triggering first probe light at, or scanned across, first wavelengths and splitting the first probe light to direct a first portion through a sample interface to a first photodetector and direct a second portion along a reference path to a second photodetector unit, wherein the first wavelengths comprise a wavelength of an

[0027] 17257738.MDE.MDE absorption feature of the gas species in the liquid; detecting, at a first photodetector, first probe light that has passed through the sample interface; detecting, at a second photodetector, first probe light that has propagated along a reference path; triggering second probe light at, or scanned across, second wavelengths and splitting the second probe light to direct a third portion through a sample interface to a first photodetector and direct a fourth portion along a reference path to a second photodetector unit, wherein the second wavelengths are wavelengths offset from the absorption feature; detecting, at a first photodetector, second probe light that has passed through the sample interface; detecting, at a second photodetector, second probe light that has propagated along a reference path; analysing signals from the first and second photodetector based on the detected probe light to determine a spectrum or concentration of the gas species in the liquid.

[0028] The step of analysing may comprise: generating a forward model based on spectroscopic data of a reference liquid sample measured at the first wavelengths and the second wavelengths and the reference path measured at the first wavelengths and the second wavelengths, the forward model configured to estimate spectra and / or provide a quantification of species in a sample based on an amount of the gas species in the liquid sample; and inverting the model to generate a quantification of gas species in a sample based on measured spectra.

[0029] The present invention further provides a quantitative method of analysis of gas species in a liquid sample, the method comprising: generating a forward model of a spectroscopic analysis instrument based on spectroscopic data of a reference liquid sample measured at first wavelengths and second wavelengths and a reference path measured at the first wavelengths and the second wavelengths, the forward model configured to estimate spectra or absorption based on an amount of gas species in the liquid sample; receiving data representing: a first signal indicative of absorption at gas species absorption feature wavelengths of the gas species in the liquid sample based on first probe light incident on a first photodetector; a second signal indicative of absorption at the gas species absorption feature wavelengths along a reference path based on the first probe light incident on a second photodetector; a third signal indicative of absorption of the sample at wavelengths offset from the gas species absorption feature wavelengths based on second probe light incident on the first photodetector; and a fourth signal indicative of absorption at the offset wavelengths along the reference path based on the second probe light incident on the second photodetector, and inverting the model to generate a quantification of gas species in the sample based on the received data.

[0030] 17257738.MDE.MDE The forward model may be configured to estimate the transmission or absorption along a sample beam path to the first detector based on the length of the beam through air and the length of the beam path through the liquid sample; and the forward model estimates the transmission or absorption along a reference beam path through air to the second photodetector based on the length of the beam through air.

[0031] The method may further comprise generating a vector comprising data representing spectra of the four signals; and using an inversion of the forward model on the vector to determine the amount or concentration of gas species in the liquid.

[0032] The present invention provides a computer-readable medium storing instructions which, when executed by a processor, causes the processor to carry out the method set out herein.

[0033] Brief description of the drawings

[0034] Embodiments of the invention will now be described, with reference to the drawings, of which:

[0035] Figure 1 is a schematic diagram of a spectrometer apparatus according to embodiments of the present invention for quantitatively analysing a sample comprising a gas in a liquid;

[0036] Figure 2 is a schematic absorption spectral diagram comparing the absorption coefficient of water and carbon dioxide in the 2250 to 2400 cm'1wavelength range;

[0037] Figure 3 is a detailed spectral diagram showing absorption spectra for water, gaseous CO2, aqueous CO2 and the combined absorption for all three;

[0038] Figure 4 is a flow-chart showing a method of operating the apparatus of figure 1 ;

[0039] Figure 5 is a flow-chart showing steps of initial processing of four sets of measurement data received at the two detectors from the two sources;

[0040] Figures 6a-6f are plots of measurement data showing the processing performed according to the steps of figure 5, where figures 6a, 6c and 6e respectively show for the on- resonance laser the raw data, dark current corrected data with a third order curve fit, and a baseline corrected wavenumber based curve, and where figures 6b, 6d and 6f show corresponding data for the off-resonance laser;

[0041] Figure 7 is a graph showing spectrum noise calculated based on the data of figures 6a-6f;

[0042] Figure 8 is a flow-chart describing pre-processing of measurement data before being used in a forward model;

[0043] 17257738.MDE.MDE Figure 9 is a flow-chart setting out the steps of using a forward model for quantification of the gas in the liquid;

[0044] Figure 10 is a panel of eight graphs with four showing example of data output from second level processing using data recorded for aqueous CO2 and four graphs showing corresponding residuals;

[0045] Figure 11 is a panel of four graphs showing example results obtained from second level processing from a deployment measuring CO2 dissolved in water, the graphs showing mass fraction CO2, ppm CO2, expansion of the sample cell and a reduced / 2 measure; and

[0046] Figures 12a and 12b show mass fraction and uncertainty respectively, comparing between "single point" processing, where only two fixed laser frequency measurements are used, namely one on resonance and one off resonance, versus the full spectral analysis according to the present invention.

[0047] Detailed description of embodiments

[0048] Figure 1 schematically shows apparatus 100 for quantitatively analysing a sample comprising a gas in a liquid. The sample may be provided to the apparatus in a sample cell, although as we will describe below other sample interface means may alternatively be used. The apparatus comprises one or more optical sources such as a pair of lasers 110, 120. Laser 110 provides first probe light which may be scanned across first wavelengths. Laser 110 is selected such that its wavelength range covers one or more vibration bands of the gas species or molecules when dissolved in the liquid and preferably also covers one or more vibration bands of the gas species in gaseous form. Hence, absorption of the first probe light occurs when the wavelength matches a vibration band of the gas species either in gaseous form or dissolved in the liquid.

[0049] Laser 120 provides second probe light which may be scanned across second wavelengths. Laser 120 is selected such that its wavelength range is outside or offset from the peak of the vibration band of the gas species in dissolved form. For example, it is preferable if the wavelength range of laser 120 is well-spaced from the peak absorption wavelength of the gas species in aqueous form, such as spaced by at least 10 cm-1but more preferably at least 20, 30 or 40 cm-1. The second probe light will still excite vibration bands of the gas species in gaseous form. The lasers 110, 120 may be quantum cascade lasers (QCLs) but other types of laser and optical sources may be used. The probe light may form beams from the lasers which are directed to beamsplitter 130 which directs a first portion of each beam along a sample arm beam path 145 and directs a second portion of each beam along a reference arm beam path 155. Sample beam path 145 comprises

[0050] 17257738.MDE.MDE sample beam path optics and sample interface. The respective probe light laser beams are directed through the liquid sample at the sample interface. In figure 1 the sample interface is a sample cell 140. The sample cell may be a CaFI cell. After passing through sample interface or sample cell, the probe light is directed to a first photodetector 160 which generates a voltage VA based on the intensity of the received probe light. Reference laser beam 155 is directed from beam splitter 130 to a second photodetector 170 which generates a voltage VB based on the intensity of the received probe light. Also shown in figure 1 are other optical components and features. For example, the output from each laser is collimated or focussed by an aspheric lens 111 , 121 , to produce a laser beam. The lasers themselves may be air cooled. The lasers may be arranged side-by-side. If this is the case, as shown in figure 1 , then a mirror or reflector 125 may be needed such that the two beams are directed at different parts of the beamsplitter. The beamsplitter may be a plate, cube or thin film beamsplitter. The two beams are incident on the beamsplitter orthogonally or transversely, with the beamsplitter at around 45° to each beam. The beamsplitter preferably provides a 50:50 split of the beam, directing half of each beam along the sample path arm and reference path arm. One of the lasers, such as laser 120 in the figure, may alternatively be arranged at 90° to the other laser 110 with the laser beam directed straight at the beamsplitter without requiring additional mirror or reflector 125. Other arrangements for the two lasers are also possible.

[0051] Referring now to the sample arm beam path 145, the beam may be directed at mirror or reflector 142 after first leaving the beamsplitter 130. Alternatively, no mirror or reflector may be used and the beam travels straight towards the sample cell. Again other arrangements to the sample arm beam path and the reference arm beam path are possible. Before and after passing through the sample cell may be provided an iris 147 or aperture. These are included to prevent stray light from reaching the detector. The sample cell 140 comprises a flow cell through which a sample or process liquid stream may pass through. The cell may be a tube with opposing windows that are transparent to the wavelengths of the laser beams. Having passed through the sample cell the probe light is focussed on the first photodetector 160 by lens 148 which may have a 50mm focal length. Other optical arrangements for the sample beam path and sample cell are possible.

[0052] Instead of a sample cell, the sample interface may use an attenuated total reflection (ATR) probe comprising an ATR crystal. ATR crystals reflect light at an interface with the sample in such a way that an evanescent wave extends into the sample. The beam may reflect at the interface in this way multiple times. Using a probe with an ATR crystal allows the sample to be measured without using a small size flow cell which could be at risk of

[0053] 17257738.MDE.MDE fouling depending on the liquid passing through. An ATR probe could be inserted into a wider flow pipe or from a window at the flow pipe and interface with the sample.

[0054] The reference arm beam path 155 takes the portion of light from the beamsplitter and directs it at second photodetector 170, with lens 148 focussing the light on to the photodetector. The lens 148 which may have a 50mm focal length.

[0055] The reference arm path may include an optional low pressure gas cell which is inserted in the reference path to provide a mean for fine laser frequency calibration. This could be inserted between the beam splitter 130 and photodetector 170, and preferably between beamsplitter 130 and lens 148, for example at around the location of the arrow numbered 155. The low pressure gas cell may used to check the frequency of the laser(s) I optical sources. As we will describe in the following, the gas absorption lines that are being detected may be narrow so it is desirable to be able to readily calibrate the optical frequency I wavenumber of the optical sources to maintain precision and accuracy to the measurements. The gas in the low pressure gas cell provides well known reference resonant frequencies that will appear in the signals, such as at photodetector 170. With such frequencies appearing in the experimental spectra, the relationship to features in the measured sample spectra can be determined. For example, the relationship between the laser frequency sweep index and the absolute frequency emitted by the laser sources can be determined. Calculations, such as by modelling using a polynomial function or other smoothly varying function, can be performed to calibrate the spectra recorded across the full wavelength of interest. By including the low pressure gas cell in the optical set up, a calibration is included in each spectrum measured making the measurements more robust.

[0056] A controller 180 may control the triggering of turning on and off of the optical sources such that they are turned on alternately. The controller 180 may also control scanning of the sources across first and second wavelengths. For example, the controller may control the scanning of sweeping of the optical sources by around 2 cm1in wavenumber. An analyser may receive signals, such as based on voltages from the first photodetector 160 and second photodetector 170 and analyse the signals to determine spectra and / or quantify the amount or relative amount of gas species dissolved in the liquid or solution. The analyser may comprise a microprocessor, memory and user interface. The analyser may be a computing device incorporated in the spectroscopic apparatus. Alternatively, the analysis may be performed by a separate computing device that is supplied with data collected from the photodetectors. In embodiments the controller and analyser may be combined and integrated into the spectroscopic apparatus.

[0057] 17257738.MDE.MDE The apparatus described in figure 1 allows absorption measurements to be made on resonance and off resonance as we now describe with reference to figures 2 and 3. The approach uses the unique spectral signature of gas vibrational bands within a liquid. In the following we describe the example of CO2 gas dissolved in water. For example, for carbon sequestration purposes it may be desirable to know how much carbon dioxide is dissolved in a fluid such as water. The apparatus may operate in the middle infrared, such as at any wavelength between 2 and 20 microns, where the spectral signatures are the fundamental vibration bands. These vibration bands are intense, thereby allowing for high measurement precision to be achieved.

[0058] The example of CO2 is shown in figure 2. Towards the top of the figure the absorption line of CO2 in water in shown at 4.27 pm. This is superimposed on the water absorption signal which is shown by the line showing gradually decreasing absorption coefficient with increasing wavenumber. The lower lines in figure 2 show absorption with the absorption from water subtracted to leave the peak for CO2 absorption in water. The present invention uses a dual wavelength approach to measure a small subset of wavelengths “on resonance” (meaning at the CO2 peak) and “off resonance" (meaning well away from the CO2 peak). These ON and OFF wavelengths are indicated by arrows in figure 2 at 2343.5 cm-1and 2380 cm-1respectively. Hence, first probe light may be at 2343.5 cm-1and second probe light may be at 2380 cm-1.

[0059] Figure 2 is a simplification of the absorption spectra of CO2 and water, which is shown in more detail in figure 3. Since the probe light passes through air on its way to the sample interface we also need to consider the absorption of CO2 in gaseous form in air. Figure 3 shows spectra as follows:

[0060] A - transmission spectrum of aqueous CO2 (corresponding to the absorption line for aqueous CO2 in figure 2);

[0061] B - transmission spectrum of water (corresponding to the gradually decreasing absorption line in figure 2);

[0062] C - transmission spectrum of gaseous CO2 showing various vibration absorption peaks; and

[0063] D - combined transmission spectrum of A, B and C together.

[0064] Also shown in figure 3 are two circled regions representing spectral coverage of the probe light of the two optical sources.

[0065] Hence, the spectra generated by the instrument include absorptive effects from gaseous CO2, aqueous CO2, liquid water, and also from optical surfaces in the instrument. As the optical sources may be tuned to vary the laser frequency, the measured spectra will

[0066] 17257738.MDE.MDE also include laser power variation inherent to the tuning. The spectra in figure 3 shows a calculated transmission spectrum that includes the various absorptive effects across a range from 2320 cm-1to 2390 cm-1.

[0067] It is clear from figure 3 that there is a lot of information in the spectra that a singlepoint analysis would miss, or even wrongly interpret, as for example, in the case of a change in water transmission, or by not taking into account the multiple absorption peaks of gaseous CO2.

[0068] The apparatus is directed to making a quantitative measurement of the amount of gaseous species, such as CO2, in solution. However, the beams from the optical source pass through air on the way to, and after passing through, the sample. Since air comprises CO2 it is desirable to be able to measure this so that it can be corrected for. Furthermore, the water itself has an absorption effect that needs to be corrected. Hence, the reference beam path and detector B, 170, provides a measure of the CO2 in the atmosphere, which may vary, especially in enclosed spaces. The off-resonance laser through the sample provides a measure of the contribution of the absorption of water in the sample. By analysing and combining the four absorption measurements the absorption by CO2 in aqueous form can be determined and hence a quantitative measure provided.

[0069] Figure 4 is a flow-chart describing a method of operating the apparatus. At step 410 the “on resonance” optical source 110 is triggered turning it on. The optical source may be scanned across first wavelengths such as the spectral range 2343 to 2346 cm-1circled in figure 3. The probe light from optical source 110 is split at beamsplitter 130 to direct a first portion of the probe light to the sample interface 140 and on to the first detector 160. A second portion of the probe light is directed from the beamsplitter along reference path 155 to second detector. The beamsplitter preferably splits the probe light such that the intensity of the first portion and the second portion are approximately equal. At step 420 signals are measured at first photodetector 160 for the sample interface path and at the second photodetector 170 for the reference path. Preferably, the first optical source and data is collected at various wavelengths in the spectral range. For example, 100s or preferably around 1000 measurements may taken at each photodetector at wavelengths spread across the spectral range. After these two sets of measurements are collected the “on resonance” optical source is turned off as indicated at step 430. At step 440 the “off resonance” optical source 120 is turned on. The off-resonance optical source may be scanned across second wavelengths such as the spectral range 2381 to 2385 cm-1circled in figure 3. The probe light from optical source 120 is split at beamsplitter 130 to direct a first portion of the probe light to the sample interface 140 and on to the first detector 160. A

[0070] 17257738.MDE.MDE second portion of the probe light is directed from the beamsplitter along reference path 155 to second detector. The beamsplitter preferably splits the probe light such that the intensity of the first portion and the second portion are approximately equal. At step 450 signals are measured at first photodetector 160 for the sample interface path and at the second photodetector 170 for the reference path. Preferably, the second optical source and data is collected at various wavelengths in the spectral range. For example, 100s or preferably around 1000 measurements may taken at each photodetector at wavelengths spread across the spectral range. After these two sets of measurements are collected the “off resonance” optical source is turned off as indicated at step 460. As a result of the two sets of measurements, using two optical paths and two photodetectors, four sets of data are generated, as follows: i) data based on the “on resonance” first probe light incident on the first photodetector, having passed through the sample; ii) data based on the “on resonance” first probe light incident on the second photodetector, having propagated along the reference path; iii) data based on the “off resonance" second probe light incident on the first photodetector, having passed through the sample; and iv) data based on the “off resonance” second probe light incident on the second photodetector, having propagated along the reference path.

[0071] At step 470 the four measured signals or data are processed, such as by analyser 190, to generate absorption spectra and / or a measure of the concentration or amount of the gas dissolved in the liquid sample.

[0072] Although we have described scanning the laser wavelength across a spectral range, in an alternative embodiment the optical sources could be a single broadband optical source that emits across both spectral ranges. In such a case a filter at the two photodetectors is scanned such that its pass window scans across the “on resonance” spectral range and then across the “off resonance” spectral range. Alternatively, two broadband optical sources may be used, with one for the “on resonance” spectral range and the other for the “off resonance” spectral range. Again, filters would need to scan a pass window across the detectors.

[0073] Figure 5 is a flow-chart showing steps of initial processing of the four sets of measurement data. Figures 6a-f are plots of measurement data showing the processing performed.

[0074] The measurement data may be collected from the photodetector and received at the analyser 190 or may be stored and later passed to the analyser. The measurement

[0075] 17257738.MDE.MDE data comprises the four sets of measurements as mentioned, and may also comprise information about the laser conditions for each measurement, such as temperature and current, and also a time stamp. The measurement data may be received at step 510 as one file such as a TDMS file. The TDMS file may comprise all of the information measurement. The analyser may divide up the data into four separate arrays, one for each of the measurement conditions: (i) detector A, on-resonance laser, (ii) detector A, off- resonance laser, (iii) detector B, on-resonance laser, (iv) detector B, off-resonance laser. At step 520 the sub-threshold dark signal from each spectrum is subtracted. The subthreshold dark current is measured at laser currents before the laser starts emitting as we will discuss in relation to figure 6. The data processing then comprises:

[0076] 1 . Perform baseline fit (step 530);

[0077] 2. Generate preliminary transmission spectra (step 540);

[0078] 3. Perform wavenumber assignment (step 550);

[0079] 4. Repeat the baseline fit using the wavenumber assignment (step 560); and

[0080] 5. Calculate spectrum noise.

[0081] In figure 6 the left-hand column of plots (figures 6a, 6c and 6e) show example spectrum and processing based on data recorded by detector A with the on-resonance laser, and the right-hand column of plots (figures 6b, 6d and 6f) show example spectrum and processing based on data recorded by detector B with the off-resonance laser. Figures 6a and 6b show raw spectra. The spectra comprise 1000 data points and start below the laser threshold to allow the dark signal to be calculated. The shaded rectangles below threshold indicate the parts of each spectrum that were used to calculate the dark signal. In all cases the start of each spectrum showed a small transient in the dark signal, which was likely caused by the end of the laser ramp from the previous laser scan. Hence, for the purposes of calculating the dark signal the first part of every spectrum was neglected. This was more prominent in the spectra recorded with the on-resonance laser. In the case of the off-resonance laser (figure 6b), there was a large number of sub-threshold data points (almost 250). In this example analysis data-points 75 to 175 were chosen to calculate the dark signal. In the case of the on-resonance laser (figure 6a), there were fewer subthreshold data points, and data points 35 to 50 were chosen for the dark signal measurement. Other numbers of sub-threshold data points may be used.

[0082] Towards the end of the on-resonance spectrum (figure 6a) there is a large- amplitude absorption feature from gaseous CO2 which is not fully resolved. This feature was found to perturb the baseline analysis and so the last part of the on-resonance spectra

[0083] 17257738.MDE.MDE (highlighted by the rectangle at datapoints 900-1000) was disregarded from further analysis.

[0084] Figures 6c and 6d show the baseline fitting of the dark-signal-corrected spectra, indicated in figure 5 by step 530. The baseline fit was performed using a 3rd order polynomial function. The off-resonance spectra shown in figure 6d comprise mostly data points that were off-resonance from CO2(g) absorption features. In figure 6d the off- resonance data can be seen with the dips in detector signal at around datapoints 200, 400 and 600. The baseline fitting is also shown which matches the general curve but continues without fitting to the dips. The baseline fitting is a good fit. The baseline fit for the spectra recorded by the on-resonance laser is more complex, as shown in figure 6c. The spectrum includes a large amplitude CO2(g) absorption feature. A 3rd order polynomial was again used to fit the baseline and is shown in figure 6c.

[0085] After the baseline fitting, initial transmission spectra were calculated (step 540) by dividing by the baseline function. The initial spectra were then used for the wavenumber assignment. The main peaks in the initial absorption spectra were identified (step 550), as shown by the circled dips in the spectra of figures 6c and 6d. An algorithm or software routine was used to find the main peaks in each spectrum. The peak positions were then used as a priori centroids in fits of a Gaussian function of each absorption feature. The centroids of each of these were then used as absorption line centres. For each spectrum three dominant absorption features were used in the frequency / wavelength assignment of the spectra and for each spectrum a quadratic function was fitted using the measured centroids as the X data. For the Y data known standard absorption line frequencies were used, such as HITRAN line frequencies. For example, the following HITRAN line frequencies may be used:

[0086] - On-resonance laser: 2345.53610 cm-1, 2344.36573 cm-1, 2343.39377 cm-1; and

[0087] - Off-resonance laser: 2383.35519 cm-1, 2382.49970 cm-1, 2381.61852 cm-1.

[0088] After the wavenumber assignment the baseline analysis was repeated, generating a baseline function that uses wavenumber as the argument (step 560), shown in figures 6e and 6f. For the off-resonance spectra, three absorption lines can be seen. For the on- resonance laser a relatively much greater amount of absorption can be seen, which corresponds to the absorption by aqueous CO2.

[0089] During a measurement period many such measurement spectra may be collected, such as over a period of hours. The processing described above is performed for each spectrum of a given type (Det A, on-res; Det A, off-res; Det B, on-res; Det B, off-res). The various measurements over a period of time may be stored in one file. Spectrum noise may

[0090] 17257738.MDE.MDE then be calculated, as indicated at step 570. The spectrum noise may be calculated using a covariance matrix of all the spectra in a given file to obtain the standard deviation on every spectral data point. The data in figure 7 shows increased noise around the positions of spectral features, which may indicate laser frequency noise.

[0091] After the spectrum noise has been calculated, the initial processing is complete and the data may be saved. The data may include: (i) all spectra, (ii) the wavenumber scale for each spectrum, (iii) baseline parameters for each spectrum, (iv) array of time stamps associated with the spectra, and (v) an ‘error catch array’ that records whether the baseline fitting procedure was a success for each spectrum (a ‘0’ indicates a success, a ‘1 ’ indicates a failure).

[0092] To determine the concentration or amount of gas species with the liquid, such as CO2 dissolved in water, a second level of processing is used based on a model of the instrument. The model analyses all four spectra simultaneously, by forming a concatenated vector containing all the spectra, in the order [detector A_off-resonance, detector B_off- resonance, detector A_on-resonance, detector B_on-resonance]. This simultaneous use of spectral information allows the second level of processing to efficiently deconvolve multiparameters. The processing by the model includes a reference measurement taken for a sample of pure water. More details on the processing by the model are shown in figure 9 and will be discussed in the following. However, we first discuss figure 8 which describes pre-processing of the measurement data before being used in the model. The second level pre-processing comprises the following steps:

[0093] 1 . Load (step 810) in the first level data (i.e. the outputs from figure 5), standard absorption line data (such as HITRAN data) and COa(aq) cross-section reference data;

[0094] 2. Create summed error-catching array (step 820);

[0095] 3. Generate regular wavenumber array (step 830);

[0096] 4. Interpolate all spectra, the spectrum noise and the CO2(aq) cross-section to the regular wavenumber array; and

[0097] 5. Produce concatenated arrays for the spectra, wavenumber, relative wavenumber and spectrum noise.

[0098] We now describe each step in more detail. For each measurement, a combined error-catching array (step 820) is created by summing the error-catching for the baseline fits of the individual spectra. Any set of four measurements with a non-zero number flags a measurement not to be processed, as likely having quality issues. Further error catching is then performed using the wavenumber scaling. For each laser, wavenumber boundaries

[0099] 17257738.MDE.MDE are determined as the maximum range that overlap all the valid spectra contained in the file. Regular wavenumber arrays are then generated, at step 830, with the resolution set by wavenumber range found from the overlap analysis and the number of points per spectrum. The spectra, spectrum noise and the COa(aq) cross-section are interpolated (step 840) to the new regular common wavenumber array. If a spectrum’s pre-interpolated wavenumber array contains values that are far away from the interpolation range (using a user-defined threshold), then the previously determined wavenumber assignment is considered to have failed and the measurement associated with this spectrum is rejected. Concatenated arrays for the spectra, wavenumber, relative wavenumber and spectrum noise, are generated at step 850.

[0100] As mentioned, the second level of processing uses a model of the apparatus. The model is a Forward Model (FM). The FM allows the determination of the instrument output y - namely the spectra, knowing the parameters we seek to measure x - in our case primarily, but not only, the aqueous CO2 concentration. We then have y = F(x) with F() being the FM. The second level analysis is based on mathematical inversion methods that allow to invert the function F() and solve x knowing y.

[0101] As shown in figure 9, once the forward model of the instrument has been generated, at step 910, the data is input to the model at step 920. The data will include the spectral data and may be the concatenated arrays produced at step 850. Inversion methods are used on the forward model, at step 930, to determine the parameters we are seeking to measure, namely a quantitative measure such as concentration of the gas species dissolved in the liquid 940. In the examples described herein the quantitative measure is that of CO2 dissolved in water.

[0102] We now describe further the forward model of the apparatus for the embodiment relating the carbon dioxide dissolved in water. To generate the modelled instrument output forming y, the physical model is described by equations 1 - 4, as follows:

[0103] TBoff(w) = poff(co) ■ exp [-ag(Cg; P; T; co) ■ Lair] (1 )

[0104] TB°" (CO) = pon(co) ■ exp [-ag(Cg; P;T; co) ■ Lair] (2)

[0105] Toff(co) = koff ■ pow(co) ■ exp [-ag(Cg; P ; T ; co) ■ (Lair+5Lair) - (aaq(Caq; co) + aH2o(co)) ■ (I— cell + 6L cell)] (3)

[0106] TA011( O) = kon ■pon(c ) ■ exp [— Qg(Cg; P;T; co) ■ (Lair + 6Lair) — (oaq(Caq; co) + OH2O(CO)) ■ (Lcell+5Lcell)] (4)

[0107] 17257738.MDE.MDE Equation 1 describes the transmittance, TB°”, detected at detector B (reference path) for the off-resonance laser. The transmittance can be seen to vary based on the concentration of gaseous CO2 (Cg) and the path length of the laser through air (Lair), as well as various other factors including the absorption coefficient of gaseous CO2 (ag). Equation 2 for the transmittance, TB011, is that detected at detector B for the on-resonance laser, and is described similarly to TB0” in equation 1 except for a different laser baseline function p.

[0108] Equation 3 describes the transmittance, TA0”, detected at detector A (sample interface path) for the off-resonance laser. The transmittance can be seen to vary in a similar manner to equation 1 but has an additional term to take into account the absorption of aqueous CO2 and the absorption of water. These absorptions are determined based on the concentration of aqueous CO2 (Caq), and the path length of the laser through sample interface or cell. A variation in the sample cell size Seen is taken into account based on changes in pressure and temperature. The term 5Lairtakes into account the difference in path length through air compared to the reference path.

[0109] Equations 1 - 4 also include the following functions:

[0110] • Poff / on(w) is the "baseline" functions for the off- and on-resonance lasers, which describes the power modulation of the laser and the response of the detection chain. Their units are signal voltage. For each laser the function is the same for both photodetector signals, because the laser power modulation is the same, with a k scaling parameter that accounts for differences in the transmission and detection efficiency between the two beam paths (differences originate from: transmission through different optical elements, different alignment of the beam onto the two photodetectors, and a different responsivity between the two photodetectors). The baseline function has the form shown in equation 5:

[0111] PoffZon( j) = P ff / on +p1off / on ’ ) + p2off / on ‘ J2+ P^off / on ‘UJ2(5)

[0112] • ag(Cg; P; T; w) is the absorption coefficient of CC g), which is equal to the cross section (calculated from a standard database such as the HITRAN database) multiplied by the CO2(g) concentration.

[0113] • aaq(Caq; w) is the absorption coefficient of CC>2(aq). This is the product of the mass fraction of dissolved CO2 (Caq, units of mg / kg) and the absorption cross-section (aaq, units of kg mg-1rrr1). This cross-section can be determined experimentally using FTIR measurements.

[0114] 17257738.MDE.MDE • aH2o(w) is the absorption coefficient of water. This can also be determined experimentally using FTIR measurements and may be described by the following expression: OH2O(W) = 2037cm-1- (wavenumber x0.766). OH2O (W) has units of cm-1, and so should be multiplied by a length measured in units of centimetres when calculating the transmission.

[0115] The instrument output y may be a vector, with y = [TBoff, TBon, TA0”, TAon] made up of the four concatenated spectra, such as with each made up of 1000 data points. The state vector x is made of the parameters in equations 1 - 5. We set out in detail in Table 1 below each of the parameters of the equations.

[0116] Parameter Description Reference mode Sample mode

[0117] P°off Oth order baseline coefficient, off-res O O laser (V) 4.31 ± 1.19 4.31 ± 1.19

[0118] P1off 1st order baseline coefficient, off-res O O laser (V / cnr2) -0.89 ± 2.47 -0.89 ± 2.47 2off 2nd order baseline coefficient, off-res O O laser (V / cm-2) 0.03 ± 1 .48 0.03 ± 1 .48

[0119] 3rd order baseline coefficient, off-res O O

[0120] |33off laser (V / cm-3) -0.050 ± 0.26 -0.050 ± 0.26

[0121] Oth order baseline coefficient, on-res O O laser (V) 3.50 ± 0.80 3.50 ± 0.80

[0122] 1st order baseline coefficient, on-res O O laser (V / cm-1) -1 .37 ± 1 .70 -1.37 ± 1.70

[0123] 2nd order baseline coefficient, on-res O O laser (V / cm-2) 0.41 ± 1.16 0.41 ± 1.16

[0124] 3rd order baseline coefficient, on-res O laser (V / cm-3) -0.15 ± 0.24

[0125] Baseline scaling parameter, on-res laser O ®

[0126] (dimensionless) 1 .30 ± 1 .30 1.52

[0127] Baseline scaling parameter, off-res O ® laser (dimensionless) 1 .30 ± 1 .30 1.28

[0128] CO2(g) concentration O O

[0129] (PPm) 400 ± 200 400 ± 200

[0130] CO2(aq) concentration ® O

[0131] (mg / kg) 0 5000 ± 5000

[0132] P Air pressure (torr) ® 760

[0133] 17257738.MDE.MDETAmbient temperature ® ®

[0134] (kelvin) 295 295

[0135] Lceii Set length of cell ( m) ® 12

[0136] SLceii Cell expansion (pm) O 8.0 ± 8.0

[0137] Beam path through air to det B ® ®

[0138] (cm) 18 ± 1 18 ± 1

[0139] 6Lair Beam path correction for det A ® (cm) 1.1

[0140] Table 1

[0141] Table 1 indicates whether each of the parameters was a fit parameter (indicated by an unfilled circle symbol) or a fixed parameter (circle symbol with a cross in). Under each symbol are typical values taken from example measurements. The value is either (i) the a priori value and uncertainty (for fit parameters) or (ii) the fixed value (for fixed parameters).

[0142] The instrument was operated in two modes: (i) reference mode with pure water measurements, and (ii) sample mode with aqueous solution measurements made under pressure. Accordingly, Table 1 has two columns of parameter values for these two modes. For most of the parameters the value and type (fixed or fitted) are the same in both modes. However, some do vary between operational mode:

[0143] • on / off are fit parameters in the reference mode measurements. Reference mode measurements may be recorded periodically such as at the start and end of the day. The k values may then be calculated for the start and end of the day and a linear interpolation used to establish the variation (if any) of the k values over the course of the sample measurements during the day, and these values used as fixed parameters for that day’s sample measurements. Alternatively, only one set of water measurements may be used to establish the k values for a measurement run such as a day, and these values were used in the analysis of all CO2(aq) measurements for that day. Typically, little variation was seen in the k values.

[0144] • Caq, the concentration of aqueous CO2 is fixed to zero for the reference mode measurements as pure water was used.

[0145] • SLcell is fixed at zero for the reference mode measurements as these were taken at low pressure, where laboratory measurements show that there is no cell expansion.

[0146] Lair is preferably measured experimentally prior to deployment. 5 Lair takes into account a different path length through air to reach the different detectors. For example,

[0147] 17257738.MDE.MDE analysis of an embodiment found that this is well described by a difference of 1 .1 cm. Lceii was set to 12 urn, defined by the thickness of a spacer in the optical cell, in an embodiment in which a sample cell was used.

[0148] The 6LCeii takes into account that, for aqueous solutions, the cell path length is very short, a few tens of microns, and hence prone to vary with pressure at micron levels. In one example measurement deployment, the pressure of the fluidics system varied from around 10 bar to 13.5 bar, and the effect of this on the measured spectra was very clear. A crucial advantage of the dual-band measurement method according to the present invention is that the detector A signal from the off-resonance laser is only weakly affected by CO2(aq) but is strongly affected by water absorption, and so the off-resonance measurements serve as a good method of measuring the total path length through the cell to fully compensate for this drift.

[0149] Figure 10 shows an example of data output from the second level processing using data recorded for CO2(aq). The upper panel shows experimentally measured and fitted theoretical photodetector signals. However, to the resolution of the figure the lines representing the two signals match so closely that they effectively overlie each other. The figure shows data for a single measurement of the four signals: i) Detector A, off-resonance laser; ii) Detector B, off-resonance laser; iii) Detector A, on-resonance laser; and iv) Detector B, on-resonance laser. The data from the four signals was combined into a single measurement vector. The vertical dashed lines mark the separation between the spectra. The lower panel shows corresponding residuals, that is the difference between measured values and fitted model, which are typically on the order of 1% of the measured signals. Some increased values occur around the strong CC>2(g) absorption features, where the shape of the residuals indicates a frequency offset. As can be seen in the top panels of figure 10 the curves for the off-resonance laser are similar to figures 6b and 6d. The larger absorption features for the on resonance laser for the sample path can also be seen in the fourth of the upper panels of figure 10.

[0150] Figures 11 and 12 shows example results obtained of the second level processing from a deployment measuring CO2 dissolved in water. In figure 11 the top panel shows the CO2(aq) mass fraction, with typical values in the range from 8000 mg / kg to 9500 mg / kg. The horizontal axis is time over a period of three days. The mass fraction data shows a periodic structure with a period close to 4 minutes. Investigation of the system being analysed indicates that this is a real feature that results from pressure variation in the system. The second panel from the top shows the measured concentration of CO2(g). This shows significant variations in concentration. This will be affected by the amount of

[0151] 17257738.MDE.MDE gaseous CO2 in the local environment to the instrument. This has been linked to the room / enclosure in which the apparatus was located being closed with occupants present increasing the CO2 levels periodically until doors of the enclosure are opened. It can be seen that the model well describes the CO2(g) concentration independently of the CO2(aq) mass fraction, and the latter shows no correlation with the former, indicating that potential biases due to gas phase CO2 are well rejected.

[0152] The third panel in figure 11 shows 8LCeii describing the sample cell length path variation over time. Under positive pressure, such as from sample flow, the windows of the cell could be forced outwards, increasing the path length through the sample. The spectral analysis allows this to be captured, and the associated bias corrected. The variation in the cell expansion correlates well with the expected variations in system pressure.

[0153] The bottom panel of figure 11 shows the value of the reduced x2 for the fits. These are typically in the range from 0.1 to 3. Any measurement with a %2 value greater than 10 is preferably excluded from the final dataset.

[0154] Figures 12a and 12b show comparisons between "single point" processing, where only two fixed laser frequency measurements are used, namely one on-resonance and one off-resonance, versus the full spectral analysis described herein. Both the CO2(aq) mass fraction values (figure 12a) and uncertainties (figure 12b) for the single-point analysis (indicated by “1 P”) and the full spectrum analysis (indicated by “FS”) are shown. It can be seen that the single-point (1 P) analysis suffers from significant drift in the measurements based on the high uncertainty values shown. This is likely dominated by variations in the cell path length. Conversely, the results of the full spectrum (FS) analysis are to a very large extent immune from this source of bias. Figure 12b shows that the full spectral (FS) analysis produces significantly reduced uncertainties than the single-point (1 P) analysis typically by an order of magnitude.

[0155] The person skilled in the art will readily appreciate that various modifications and alterations may be made to the above described apparatus and methods without departing from the scope of the appended claims. For example, different mirrors, detectors or beamsplitter may be used.. Adjustments to the optical arrangement may also be made without departing from the scope of the present invention, as defined by the appended claims.

[0156] 17257738.MDE.MDE

Claims

CLAIMS:1 . A spectrometer for analysis of a gas species within a liquid, the apparatus comprising: one or more optical sources, the one or more optical sources configured to generate first probe light at first wavelengths and second probe light at second wavelengths, the first wavelengths including a wavelength of an absorption feature of the gas species in the liquid and the second wavelengths offset from the absorption feature; a controller arranged to sequentially trigger the first and second probe light; a sample interface for interfacing the first and second probe light with the liquid comprising the gas species; sample beam path optics arranged to guide the first and second probe light through the sample interface to a first photodetector; reference beam path optics arranged to guide the first and second probe light along a reference path to a second photodetector without passing through the sample interface; an analyser arranged to provide a quantitative determination relating to the gas species in the liquid or generate an absorption spectrum of the gas species within the liquid based on signals received from the first and second photodetectors at the first wavelengths and the second wavelengths.

2. The spectrometer of claim 1 , wherein the first wavelengths comprise a wavelength of peak absorption of the gas species in the liquid, and the second wavelengths are offset in wavelength from the peak absorption of the gas in the liquid, such as by at least 5cm1or more preferably by at least 10cm1or at least 20cm1.

3. The spectrometer of any preceding claim, further comprising a beamsplitter arranged to split the first and second probe light and direct a first portion to the sample interface and a second portion to the reference path.

4. The spectrometer of any preceding claim, wherein the one or more one optical sources are configured to scan the first and second probe light across the first wavelengths and second wavelengths respectively.

5. The spectrometer of any of claims 1 to 3, wherein the one or more optical sources are configured to output the first wavelengths and the second wavelengths and the17257738.MDE.MDEphotodetector comprises a filter arranged to scan across the first wavelengths and second wavelengths.

6. The spectrometer of any preceding claim wherein the one or more optical sources comprises a first optical source for generating the first probe light at first wavelengths and a second optical source for generating the second probe light at second wavelengths.

7. The spectrometer of claim 6, wherein the first and second optical sources are tuneable laser sources or broadband sources.

8. The spectrometer of any preceding claim, wherein the one or more optical sources comprise a single optical source configured to generate both the first probe light at the first wavelengths and the second probe light at the second wavelengths.

9. The spectrometer of any preceding claim, wherein the first wavelengths and second wavelengths are in the mid-infrared.

10. The spectrometer of any preceding claim, wherein the reference beam path optics are configured to guide the first and second probe light to the second photodetector along an atmospheric pathway through the apparatus.11 . The spectrometer of any preceding claim, further comprising a gas cell arranged to receive first and / or second probe light such that a reference signal is generated at the first or second photodetector for calibration of the first and / or second probe light.

12. The spectrometer of any preceding claim, wherein the sample interface comprises a sample cell or an ATR probe.

13. The spectrometer of any preceding claim, wherein the controller is configured to trigger the first probe light and the apparatus is configured to detect signals at the first photodetector and second photodetector, and the controller is configured to trigger the second probe light and the apparatus is configured to detect signals at the first photodetector and second photodetector.17257738.MDE.MDE14. The spectrometer of any preceding claim, wherein the analyser is configured to receive: a first signal based on the first probe light incident on the first photodetector, the first signal indicative of absorption at the gas species absorption feature wavelengths of the gas species in the liquid sample; a second signal based on the first probe light incident on the second photodetector, the second signal indicative of absorption at the gas species absorption feature wavelengths along the reference path; a third signal based on the second probe light incident on the first photodetector, the third signal indicative of absorption of the sample at wavelengths offset from the gas species absorption feature wavelengths; and a fourth signal based on the second probe light incident on the second photodetector, the fourth signal indicative of absorption at the offset wavelengths along the reference path.

15. The spectrometer of claim 14, wherein the analyser is configured to generate transmission and / or absorption spectra based on any one or more of the first to fourth signals.

16. The spectrometer of claim 15, wherein the analyser is configured to generate the transmission and / or absorption spectra by performing a wavenumber assignment or calibration based on known absorption lines of the gas species.

17. The spectrometer of any of claims 14 to 16, wherein the analyser comprises a processor and memory configured to perform steps of: generating a forward model based on spectroscopic data of a reference liquid sample measured at the first wavelengths and the second wavelengths and the reference path measured at the first wavelengths and the second wavelengths, the forward model configured to estimate spectra based on an amount of the gas species in the liquid sample; and inverting the model to generate a quantification of gas species in a sample based on measured spectra.

18. The spectrometer of any of claims 14 to 17, wherein the analyser is configured to determine a quantification of gas species in the liquid sample based on using the second to17257738.MDE.MDEfourth signals to subtract from the first signal absorption due to water in the sample and the gas species in air.

19. A method of spectral analysis of a gas species within a liquid, the method comprising: triggering first probe light at first wavelengths and splitting the first probe light to direct a first portion through a sample interface to a first photodetector and direct a second portion along a reference path to a second photodetector unit, wherein the first wavelengths comprise a wavelength of an absorption feature of the gas species in the liquid; detecting, at a first photodetector, first probe light that has passed through the sample interface; detecting, at a second photodetector, first probe light that has propagated along a reference path; triggering second probe light at second wavelengths and splitting the second probe light to direct a third portion through a sample interface to a first photodetector and direct a fourth portion along a reference path to a second photodetector unit, wherein the second wavelengths are wavelengths offset from the absorption feature; detecting, at a first photodetector, second probe light that has passed through the sample interface; detecting, at a second photodetector, second probe light that has propagated along a reference path; and analysing signals from the first and second photodetector based on the detected probe light to determine a spectrum or quantification of the gas species in the liquid.

20. The method of claim 19, wherein analysing comprises: generating a forward model based on spectroscopic data of a reference liquid sample measured at the first wavelengths and the second wavelengths and the reference path measured at the first wavelengths and the second wavelengths, the forward model configured to estimate spectra or provide a quantification of gas species based on an amount of the gas species in the liquid sample; and inverting the model to generate a quantification of gas species in a sample based on measured spectra.17257738.MDE.MDE21 . A quantitative method of analysis of gas species in a liquid sample, the method comprising: generating a forward model of a spectroscopic analysis instrument based on spectroscopic data of a reference liquid sample measured at first wavelengths and second wavelengths and a reference path measured at the first wavelengths and the second wavelengths, the forward model configured to estimate spectra or absorption based on an amount of gas species in the liquid sample; receiving data representing: a first signal indicative of absorption at gas species absorption feature wavelengths of the gas species in the liquid sample based on first probe light incident on a first photodetector; a second signal indicative of absorption at the gas species absorption feature wavelengths along a reference path based on the first probe light incident on a second photodetector; a third signal indicative of absorption of the sample at wavelengths offset from the gas species absorption feature wavelengths based on second probe light incident on the first photodetector; and a fourth signal indicative of absorption at the offset wavelengths along the reference path based on the second probe light incident on the second photodetector, and inverting the model to generate a quantification of gas species in the sample based on the received data.

22. The method of claim 21 , wherein the forward model estimates the transmission or absorption along a sample beam path to the first detector based on the length of the beam through air and the length of the beam path through the liquid sample; and the forward model estimates the transmission or absorption along a reference beam path through air to the second photodetector based on the length of the beam through air.

23. The method of claim 21 or claim 22, comprising generating a vector comprising data representing spectra of the four signals; using an inversion of the forward model on the vector to determine the amount or concentration of gas species in the liquid.17257738.MDE.MDE24. A computer-readable medium storing instructions which, when executed by a processor, causes the processor to carry out the method of any of claims 21 -23.17257738.MDE.MDE

Images