A comparison of isotope ratio mass spectrometry and cavity ring‐down spectroscopy techniques for isotope analysis of fluid inclusion water

RATIONALE Online oxygen (δ18 O) and hydrogen (δ2 H) isotope analysis of fluid inclusion water entrapped in minerals is widely applied in paleo-fluid studies. In the state of the art of fluid inclusion isotope research, however, there is a scarcity of reported inter-technique comparisons to account for possible analytical offsets. Along with improving analytical precisions and sample size limitations, interlaboratory comparisons can lead to a more robust application of fluid inclusion isotope records. METHODS Mineral samples-including speleothem, travertine, and vein material-were analyzed on two newly setup systems for fluid inclusion isotope analysis to provide an inter-platform comparison. One setup uses a crusher unit connected online to a continuous-flow pyrolysis furnace and an isotope ratio mass spectrometry (IRMS) instrument. In the other setup, a crusher unit is lined up with a cavity ring-down spectroscopy (CRDS) system, and water samples are analyzed on a continuous standard water background to achieve precisions on water injections better than 0.1‰ for δ18 O values and 0.4‰ for δ2 H values for amounts down to 0.2 μL. RESULTS Fluid inclusion isotope analyses on the IRMS setup have an average 1σ reproducibility of 0.4‰ and 2.0‰ for δ18 O and δ2 H values, respectively. The CRDS setup has a better 1σ reproducibility (0.3‰ for δ18 O values and 1.1‰ for δ2 H values) and also a more rapid sample throughput (<30 min per sample). Fluid inclusion isotope analyses are reproducible at these uncertainties for water amounts down to 0.1 μL on both setups. Fluid inclusion isotope data show no systematic offsets between the setups. CONCLUSIONS The close match in fluid inclusion isotope results between the two setups demonstrates the high accuracy of the presented continuous-flow techniques for fluid inclusion isotope analysis. Ideally, experiments such as the one presented in this study will lead to further interlaboratory comparison efforts and the selection of suitable reference materials for fluid inclusion isotopes studies.


| INTRODUCTION
In speleothem research, fluid inclusion isotope records have been used in combination with U-Th dating to reconstruct rainfall δ 18 O values and associated climate change through time. [5][6][7][8][9][10][11][12] Fluid inclusion isotope data of halite deposits have been used to reconstruct the isotope history of ocean water. 13, 14 When applied to vein-type deposits, fluid inclusion isotope data can be used to reconstruct basin-scale fluid flow circulation in the subsurface, which mainly finds an interest in ore geology, [15][16][17][18] petroleum geology, 19,20 and petrology. 21 Isotope analysis of fluid inclusions is not straightforward, as inclusions are only microns to hundreds of microns in size. As a consequence, bulk analytical techniques have mostly been developed, in which the integral volume of a large number of fluid inclusions within a mineral sample is released and subsequently analyzed. The first techniques emerged in the 1970s and relied on water release through thermal decrepitation in an off-line preparation device and subsequent analysis on a dual-inlet mass spectrometer. [22][23][24][25] In general, only hydrogen isotope ratios are usable, due to considerable uncertainties associated with measuring oxygen isotope ratios with these techniques. [26][27][28] Thermal decrepitation techniques have been applied mainly on vein-type deposits because fluids involved in the deposition of hydrothermal (ore) minerals are typically characterized by large isotope variations. 15,16,[29][30][31] In the 2000s, fundamentally different techniques for fluid inclusion isotope analysis were developed that allow for analysis using continuous-flow isotope ratio mass spectrometry (CF-IRMS). 32,33 Within these setups, the spectrometer is connected online to a mechanical crusher unit that is maintained at a relatively low temperature (120 C-130 C). A pyrolysis reactor within the line converts released water vapor into hydrogen and carbon monoxide gas, which are subsequently separated in a gas chromatographic column before analysis by IRMS. Reliable data for both oxygen and hydrogen isotope ratios can be acquired using IRMS techniques for sub-microliter amounts of fluid inclusion water, as demonstrated in various studies on speleothems. 6,7,34 Furthermore, the technique has successfully been applied on hydrothermal vein mineralization to gain insight into subsurface fluid flow dynamics. [18][19][20]35 As a more cost-efficient alternative, fluid inclusion isotope ratios may also be measured on an IRMS instrument using an off-line extraction method. 36 Since the beginning of the 2010s, new online setups have been developed using cavity ring-down spectroscopy (CRDS), which was already an established technique for the accurate isotope analysis of water samples. 37,38 An advantage of laser absorption spectrometry is its ability to measure δ 18 O and δ 2 H values of water vapor without the need to first split the water molecules into separate oxygen and hydrogen species. Laser spectroscopy techniques developed for fluid inclusion isotope analysis use crusher units similar to IRMS setups and run on a dry nitrogen carrier gas, 10,39 on a "wet" carrier gas containing a constant standard water background, 40,41 or under vacuum. 21 For continuous-flow IRMS and CRDS setups, reported precisions for fluid inclusion δ 18  The previously cited speleothem studies demonstrate that interlaboratory differences may exist for fluid inclusion isotope analysis. For a more robust application of fluid inclusion isotope techniques in paleo-fluid studies, the field, therefore, needs more thorough interlaboratory comparison experiments, ideally based on reliable fluid inclusion-bearing standard materials available to all laboratories that use the techniques. This is particularly valid for ongoing discussion on potential diagenetic exchange of oxygen between fluid inclusion water and host calcite in speleothems. 44,45 As long as the interlaboratory reproducibility of particularly δ 18 O fi data is underdefined, it remains difficult to distinguish between diagenetic effects and possible analytical artifacts.
In this contribution, we present an extensive comparison study of two commonly used continuous-flow setups for online fluid inclusion isotope analysis, both available at the Max Planck Institute for Chemistry in Mainz, Germany. One setup uses a crusher unit and cryogenic trap connected to a pyrolysis furnace and an IRMS instrument (cf. Vonhof et al 33 ). The second technique couples a crusher unit to a CRDS instrument and runs on humidified nitrogen carrier gas (cf. Affolter et al 40 ). Several fluid inclusion-bearing mineral samples, including speleothem, travertine, and vein material, were run on both setups. In doing so, we can present a robust data comparison between the two main analytical techniques currently used in the field.

| Design of the line
The setup for fluid inclusion isotope analysis using an IRMS system is similar to that first presented by Vonhof et al. 33 It consists of (a) a preparation unit, (b) a continuous-flow pyrolysis furnace The preparation unit consists of a crusher placed in an oven at 120 C. Water released within the crusher is vaporized and transported by He carrier gas to the TC-EA. A cryogenic trap is positioned in the line between the crusher and the TC-EA to collect water that is being released. The cryogenic trap consists of a 60 cm coiled 1/16 in. stainless steel capillary, which is located outside the oven so that it can be cooled down by submersion into an ethanolliquid nitrogen sludge at −100 to −90 C. At these temperatures, water vapor is effectively trapped, whereas contaminant species in the carrier gas and common inclusion gas phases such as CO 2 or CH 4 are flushed through. The cryogenic trap can be heated to $150 C with a flash heater to generate a water pulse short enough to be isotopically analyzed. In the TC-EA pyrolysis reactor, water vapor is transformed into H 2 and CO gas by reaction with glassy carbon at 1400 C. 46 A gas chromatographic column at 65 C separates the H 2 and CO gases before entry into the mass spectrometer. The gas flow into the mass spectrometer is controlled by a ConFlo IV Universal continuous-flow interface (Thermo Scientific). In the mass spectrometer, a rapid magnet peak jump between the analysis of H 2 and CO gases allows for the acquisition of both hydrogen and oxygen isotope ratios from a single water sample ( Figure 2).
Within the preparation unit, a bypass (through port D in Figure 1

| Analytical performance
Standard waters were analyzed by injection into the septum port of the crusher to monitor the stability and precision of the setup. on fluid inclusion isotope ratios, however, is considerably higher, as the analysis of mineral samples involves a more complex analytical F I G U R E 2 Analytical output of the isotope ratio mass spectrometry (IRMS) instrument for a 0.2-μL water sample. H 2 and CO gases are chromatographically separated and independently measured through a magnet peak jump between the entry of both gases into the IRMS instrument. Isotope ratios of the sample peaks are calibrated with respect to reference gases. The analysis of a single water sample (standard water injection or sample crush) takes about 10 min [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 3 Design of the crusher used in both the isotope ratio mass spectrometry and the cavity ring-down spectroscopy setups. A, The crusher consists of a lower part where a sample can be loaded, an upper part with a piston that can be lowered, and a clamp to close the crusher. B, View of the crusher when closed. C, Cross section of the crusher. The crusher tip is convex to achieve effective crushing when the piston is lowered onto the sample. Gas-tight seals are created using fluorinated ethylene propylene gaskets. Water standards can be injected directly into the crusher via a septum port [Color figure can be viewed at wileyonlinelibrary.com] protocol to account for memory and size effects (see section 2.3).
The upper sample size limit is approximately 0.7 μL as peak separation at higher water amounts is incomplete.

| Analytical procedure and data calculation
Prior to analysis of a mineral sample, a standard water is repetitively measured to condition the analytical line. The mineral sample in the crushing cell is crushed and measured as soon as the mass spectrometer is recording stable values for the injected standard water. Direct calibration of fluid inclusion isotope ratios is performed by comparing recorded isotope ratios of the sample crush with bracketing standard water measurements that are run directly before and after the crush. Cold-trap times are kept equal for both standard water injections and sample crushes (usually 4 min). Volumes of standard water injections following the sample crush are tailored to the water yield of the sample crush. This is to minimize possible size effects in the direct comparison of the sample with the standard water.
The setup exhibits both a memory effect and linearity, which become evident when analyzing isotopically different standard waters (Table 2). Due to the sample-to-sample memory effect, about three to four consecutive measurements of the same standard water are needed to reach stable values. For the series of standards in Table 2 Table 2.
Linearity factor δ 18 Memory effect δ 18 Equations 5 and 6 are used for the final calculation of true sample crush δ 2 H fi and δ 18 O fi values, respectively. The equations include a correction for both memory effect and linearity. The subscript "crush" refers to the isotope results of the sample crush. The subscript "STD" refers to a standard water measurement of the approximately same amount as the water yield of the crush, and the subscript "STD true" to the true isotope ratios of this standard water: Calibrations presented in Table 2

| Fluid inclusion isotope results
The analytical performance of the IRMS setup was determined by running a set of mineral samples following the protocol outlined earlier. Sample material includes a fluorite vein, travertine material, and speleothems from tropical, temperate, and high-altitude settings (  water was used to produce an average background level for δ 18

| Analytical performance
The performance of the CRDS setup was monitored by measuring standard waters ( demonstrates that isotope analyses on the CRDS setup need to be corrected for linearity and sample size effects. Memory effects in this setup are absent ( Table 5). The linearity of the instrument was values. It is noteworthy that because a 0.5 -μL micro-syringe was used, the standard water measurements for the amounts above 0.5 μL represent two rapid consecutive injections. Although this leads to a slightly different peak shape, it does not affect data quality.  Figure 9. The data of standard water injections were collected over a month, without any detectable drift in the linearity and size effects.

| Fluid inclusion isotope results
For isotope analysis of fluid inclusion water, mineral samples are crushed to analyze the liberated water. Opening and closing the crusher to load a sample leads to instability in the water background, which requires a 10-15 min waiting time for restabilization ( Figure 6).
During the stabilization period, atmospheric moisture and water adsorbed onto the sample are flushed out of the system. Because the subsequent crush and isotope analysis take another 10 min, the total time to load and analyze one mineral sample is 25-30 min.
First, a size effect correction is applied to isotope ratios of sample crushes using the size-dependent logarithmic trends shown in Precise calculations behind the corrections are explained in more detail in the supporting information (S1).
Fluid inclusion isotope data were collected on the CRDS setup for the same samples as were analyzed on the IRMS setup (Table 3).  Ocean are similar to present-day rainfall in the region 50 and lie close to unpublished drip water isotope data ( Table 3). The Huagapo-7 sample formed during the Late Holocene at an elevation of 3850 m above sea level in the Andes Mountains, where present-day rainfall has comparatively low isotope ratios ( Table 3). The Holocene Scladina stalagmite sample from The Meuse valley in Belgium also closely matches present-day drip water ( Table 3). The higher isotope variability of sample HV-SA-11 from Saudi Arabia is probably related to internal heterogeneities in the fluid inclusion water, as the subsamples were not taken from a single growth layer of this stalagmite. The water content of the HV-SA-11 subsamples is also considerably more variable than that of the other samples (Table 6).
We have no drip water of the now-dry cave, but the δ 18   are strikingly similar to those that we determined for the Huagapo sample (Table 7).