Using laser absorption spectrometry for the measurement of stable isotopes of atmospheric

As atmospheric

Traditionally, high-precision stable isotope measurements are done using isotope ratio mass spectrometry (IRMS)

Optical (infrared) spectroscopy now offers this possibility following strong developments in recent years in Fourier-transform infrared (FTIR) spectroscopy and, especially for the laser light sources, to perform isotopologue measurements showing precision close to, or even surpassing, IRMS measurements

In this paper we present the performance, in terms of precision and accuracy, of an Aerodyne dual-laser optical spectrometer (CW-IC-TILDAS-D) in use since September 2017, for the simultaneous measurement of

The optical bench as depicted in Fig.

Scheme of the optical board of the SICAS (figure adapted from

After passing the cell, the lasers are led to a thermoelectrically cooled infrared detector, measuring the signal from the lasers in the spectral range (Fig.

Typical absorption spectrum, transmission spectrum and residual for laser 1

The gas inlet system, depicted in Fig.

Gas inlet system of the SICAS with one VICI multi-valve inlet port, connected to three high-pressure natural-air tanks and 12 free ports for samples. The includes an extra inlet port for the working gas, also a high-pressure natural-air tank.

The gas handling procedures are different for measurements of air from cylinders or flasks. For the cylinders, single-stage pressure regulators are in use (Rotarex, model SMT SI220), set at an outlet pressure of

The Allan variance as a function of the integration time in seconds for a single gas measurement plotted for both the measured isotopologue abundance

The SICAS measurement performance was evaluated by determining the Allan variance of the four measured isotopologue abundances and the three isotope ratios as a function of measurement time on a single whole-air sample in the sealed optical cell. The isotope ratios, defined here as the ratio of the rare isotopologue (636, 628 and 627) and the most abundant 626 isotopologue, are r636, r628 and r627

Note that the r628 and r627 differ strictly speaking from the isotope ratios (r18 and r17) by a factor of 2.

. This experiment was first done in September 2017 and repeated in July 2019 to see the development in time of the measurement precision (Fig.The precision became significantly worse for all species but isotopologue 627 in the time period between September 2017 and July 2019. In this same period a gradual but significant decrease (of about 50 %) in the measured laser intensity was observed. For most species this led to an increase of the optimal integration time, which is logical given the fact that the minimal precision was higher, such that the increase due to drift influences the acquired precision at a higher integration time and also at a higher variance level. Figure

The decreased laser intensity, potentially leading to a deteriorated signal-to-noise ratio, was caused by contamination of the mirrors in the optical cell, most likely due to precipitation of ultra-fine salt-based aerosols from the sample air occurring during evacuation of the cell. The majority of flask samples measured on the SICAS are from the atmospheric measurement station Lutjewad, which is located on the northern coast of the Netherlands in a rural area dominated by cropland and grassland mainly used for dairy cows. The aerosol composition at Lutjewad is therefore expected to be dominated by sea salt and ammonium nitrate from agricultural emissions. Hence, we were able to clean the mirrors and retrieve

To reduce short-term instrumental drift, all sample measurements needed to be alternated with measurements of the working gas, as then the drift-corrected signal can be expressed as

A tank was measured

Relative standard deviations for

Cross-contamination, being the dilution of a small volume of the working gas in the sample aliquot that is being measured, and vice versa, as described for a dual-inlet IRMS in

The stable isotope composition of atmospheric

Three experiments have been conducted over the last 2 years to determine the CMFD and to assess its stability over time. These experiments were conducted in December 2017 (experiment 1), in December 2018 (experiment 2) and in May 2019 (experiment 3). Experiment 1 was conducted in cooperation with the Institute for Marine and Atmospheric research Utrecht (IMAU) and served as the initial determination of the CMFD on the SICAS. Experiments 2 and 3 were meant to assess the stability of the CMFD over time. A methodology to determine the CMFD of the

For determination of the CMFD on the SICAS, this approach was used with some adjustments. The SICAS is designed for the measurement of atmospheric samples of which the relevant range of

Air samples for experiments 2 and 3 were prepared manually in our own laboratory; for the detailed procedure, see Appendix A. The dilutor gas consists of natural air scrubbed of ^{®} (sodium-hydroxide-coated silica, Sigma-Aldrich) and Sicapent^{®} (phosphoric anhydride, phosphorus(V) oxide), which results in dry,

In the next two paragraphs we will discuss the results of the above-described experiments for evaluation of the two sources of CMFD according to

The first source described by

Residuals (expressed in ‰ relative to the measured amount fraction) of the linear fit of the rare isotopologue abundances as a function of the

Measured

The second source for CMFD, described by

Isotope ratios are susceptible to instrumental drift, but delta values are drift-corrected as the uncalibrated delta value

Slopes derived from the linear fits of the three measured deltas and

Mean residuals for correction of the CMFD of the three measured deltas using three different scenarios; lin and

The results for

Although most of the variance occurring in the observed CMFD of the deltas (especially of the

Four high-pressure gas tanks (40 L Luxfer aluminum, alloy 6061, max. pressure of 200 bar) containing reference gases are used in the daily measurement procedure of the SICAS: a working gas used for drift correction and possibly for a first calibration step, a quality control tank that is being treated as a sample and two tanks containing a high mole fraction reference gas and a low mole fraction reference gas, from now defined as the high reference and the low reference, which can thus be used for CMFD corrections. The high and low reference cover a great part of the

It is known that for laser spectroscopy the composition of the sample air affects the absorption line profiles by pressure broadening effects (“matrix effects”), with non-negligible consequences

CMFD for samples of the same

The gas tanks were produced in-house from dry, compressed natural air collected on the roof of our institute using a RIX compressor (model SA-3). The high and low reference were produced as follows: the high-reference cylinder was filled up to ^{®}. After the low-reference cylinder was filled up to ^{®} filled tube was removed, and the filling was continued until the pressure of both cylinders was

The

Calibrated whole-air working standards used in daily operation of the SICAS measurements.

Aliquots of all four tanks have been analyzed at the MPI-BGC in Jena by IRMS to link the

Aliquots of the working gas and quality control gas were analyzed for their

Two different calibration strategies are discussed in this section. The calibration strategies are based on the two main approaches for calibration of isotope measurements, as also described by

The RM, being very similar to calibration strategies applied by isotope measurements using DI-IRMS

The IM has the advantage that there is no need to take the introduced CMFD into account

The measurement procedure that is used for both calibration methods is based on the alternating measurements of samples/reference gases and the working gas, so the drift-corrected measurement value can be calculated as in Eq. (1). Per sample/reference gas measurement, there are nine iterations of successive sample and working gas measurements, from now on called a measurement series, before switching to the next sample/reference gas measurement series. One measurement series lasts

In the RM, measured isotopologue mole fractions are used for the estimation of isotope ratios (Eq. 1), which are calibrated to the international VPDB-

Up to this point, the procedures are more or less identical to those for IRMS measurements (but without the ion correction and

The IM as described by

A complete overview of all calculation steps of both the RM and IM can be found in Appendix C.

Quality control gas

To capture the very small signals in time series of the isotope composition of atm-

Mean residuals and standard errors of the quality control measurements in the three different measurement periods.

From the results we learn that the difference in performance between the two methods is minimal. The precision of the quality control gas measurements shows the same results, while the accuracy shows small differences between the methods for the different periods. High-quality performances are reached in period 1 for the

A combined uncertainty consisting of measurement uncertainties and scale uncertainties is calculated for the sample measurements. Measurement uncertainties include the standard error of the sample measurement, the repeatability of all (usually four) measurements of the quality control gas throughout the measurement sequence and the residual of the mean of the quality control gas measurements from the assigned value. The measurement uncertainties will therefore vary with each measurement/measurement sequence. We observe a high repeatability in all sequences included in the analysis of Fig.

The scale uncertainties, which are fixed for all measurement sequences in which the working gas, low reference and high reference are used for the sample calibration, were simulated using the Monte Carlo method. Input values were generated by choosing random numbers of normal distribution with the assigned value and uncertainty as in Table

Besides the uncertainties introduced by the scale uncertainties of the delta values, the calibrated measurements of the IM are also affected by the scale uncertainties of the

Reducing the combined uncertainty of the

To test the accuracy of SICAS flask measurements over a wide range of

The lab intercomparison is presented in the usual way: the mean and standard deviation of the differences between our SICAS

Lab intercomparison of ICP sausage 90–94 results, only including data points within the

Results of isotope measurements of quality control gas from the tank and quality control gas air in flasks (calibrated with the RM) at different periods after the flask filling procedure. The last column shows the number of cylinder measurements or the number of flasks that were used to calculate the average and the standard deviation.

When we compare the

To check the performance of the SICAS for both the IM and RM over the wide

Results of the intercomparison of

With the direct measurement of

Quality control gas

Average of the residuals from the assigned value and mean of the standard error of the quality control gas

All results show too enriched values according to the assigned values, which is probably due to the fact that the assigned

Due to the lower seasonal variations of the

In this study we discuss the measurement performances of our Aerodyne dual-laser absorption spectrometer in static mode for stable isotope measurements of atm-

A comparison of SICAS results, for both calibration methods, with results from the MPI-BGC from the sausage ICP shows that sample results within the

The pure ^{®} (sodium hydroxide coated silica, Sigma-Aldrich) and Sicapent^{®} (phosphoric anhydride, phosphorus(V) oxide), which results in dry,

Individual isotopologues of standards of known

For a more elaborated explanation of these equations, see

Calculate ratios from the measured isotopologue abundances:

Use only the relevant interval (in our case 30–60 s) from measured ratio and [

Do a drift correction and calculate the uncalibrated delta value by

Group all

Define the variable “sborder” (sborder

Calculate for all values in the series the (absolute) deviation from the median of the series, resulting in a new series containing the distance from the median (DM).

Calculate the MAD (median absolute deviation), by

Calculate per value of the series the deviation with the following equation:

If the deviation of a value is higher than 1, the value is identified as an outlier.

Calculate the mean and standard error per measurement series, excluding the identified outliers.

Do first a one-point-calibration on all mean values using the known values of the working gas, by

Calculate the means of the

Use the two calculated residuals, together with a residual of 0 for a hypothetical working gas measurement, to do a quadratic fit (

Calculate the combined uncertainty by

Use only the relevant interval (30–60 s) from measured isotopologue abundances per measurement.

Do a drift correction by

Group all

Calculate the mean per measurement series, excluding the identified outliers.

Calculate the quadratic calibration curves (

Calculate the calibrated isotopologue mole fractions of all four isotopologues for all measurements, so not for the mean of the grouped measurements but for all drift-corrected

Calculate the calibrated delta values using the calibrated isotopologue abundances for all sample measurements.

Group all

Calculate the mean and standard error of all

Calculate the combined uncertainty of the measurement:

The setup is as follows: a high reference, filled up to ^{®}-filled cartridge that removes all ^{®}. A needle valve installed before the cartridges creates a low flow to ensure the complete removal of the

Setup for the preparation of the low reference. The cylinder on the left is the high reference, filled until

To determine whether cross-contamination has the potential to affect isotope measurements on the SICAS, a simulation was conducted in which we use the measurement procedure described in this paper. Input in the simulation is an experimentally derived value, which expresses how much a measured sample is affected by the sample that was measured in the optical cell before. The experiment was conducted as follows: the high reference was measured eight times in a row, each time letting in a new aliquot, followed by the low reference which was also measured eight times in a row, and this procedure was repeated three times. The usual flushing procedure was applied every time there was a switch between the cylinders. It can be expected that the first measurement of a series of eight of the low reference is affected the most by the preceding measurement of the high-reference gas. The last measurements of a series of eight will be affected less and will be closer to the “true” value. We quantified this effect by applying the following equation to all series of measured isotopologues:

A simulation for a measurement sequence was set up in Excel, following the measurement procedure as described in this paper, only using three sample measurements per measurement series instead of nine. Included in the simulation are measurements of the low- and high-reference gas and two hypothetical samples with

These simulated, normalized measurements of the low- and high-reference gases are used to do a linear fit as a function of the true value and in doing so calculate the calibration curves. These curves are used to calculate the calibrated sample measurements, and the measured

All data that have been used for this study which were measured at the SICAS can be found in the Supplement.

The supplement related to this article is available online at:

PMS, HAS and HAJM conceived the experiments, which were conducted by PMS and HAS, and PMS carried out the data analysis. DDN and JBM optimized the fit and contributed with technical assistance for development of the gas handling system as well as solving problems with the instrumentation. The manuscript was written by PMS, and HAS and HAJM contributed with discussions and comments throughout the writing process.

Authors Dave D. Nelson and J. Barry McManus work for Aerodyne Research Inc., which is the company that developed the instrument described in this study.

We would like to thank Heiko M. Moossen, as well as his colleagues at the MPI-BGC, for measuring our reference cylinders and providing us with the data that we required for the intercomparison. We greatly acknowledge the help of Getachew Adnew from IMAU who prepared the samples for the first CMFD experiment and helped measuring them. He also measured the

This research has been supported by the European Metrology Programme for Innovation and Research (grant no. 16ENV06SIRS).

This paper was edited by Tim Arnold and reviewed by Edgar Flores, David Griffith, and one anonymous referee.