Quantification of trace element contents in frozen fluid inclusions by UV-fs-LA-ICP-MS analysis

We have developed a new analytical setup for the determination of trace element concentrations in fluid inclusions by UV-fs-LA-ICP-MS. Laser ablation was performed at a low temperature of 40 C by using a modified heating–freezing stage as the ablation cell. With this method it was possible to successfully analyse 53 of 55 frozen synthetic NaCl–H2O fluid inclusions in quartz, covering a size range between 8 mm and 25 mm down to a depth of 50 mm. The high success rate could be achieved as the 194 nm UVfs-laser allows excellent control over the opening procedure of frozen fluid inclusions. Trace element analyses were performed with a fast scanning magnetic sector field ICP-MS. The lower limits of detection for fluid inclusion analysis vary from 0.1 mg g 1 (for Bi) to 10 mg g 1 (for K). The typical analytical uncertainty, depending on the element and respective concentration level, ranges between 10% and 30% (1RSD), based on the reproducibility of experimentally synthesized fluid inclusions. All elements from a stock solution, which behaved inert during the HP/HT experiments (B, K, Cd, Te, Tl, Pb and Bi), could be recovered in the synthetic inclusions at concentrations that correspond within their specific analytical uncertainties to their original concentration of 53 mg g . The method represents a highly efficient tool for the determination of accurate trace element data on low concentration levels in small fluid inclusions with a high success rate of >90%. The latter is particularly advantageous considering the commonly time consuming characterization of fluid inclusions.


Introduction
Fluid inclusion studies are commonly conducted in order to gain information about pressure-temperature conditions, and in particular about the chemical composition of deep crustal uids. 1 Concerning the latter, laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) is regarded as the most reliable technique for the determination of concentrations of solutes in uid inclusions for a large number of elements. 2 This technique provides a high dynamic range which allows detecting elemental concentrations from the mg g À1 to the wt% level. Laser ablation-ICP-MS measurements can provide trace element concentrations (e.g. Cu, Zn, Au, Sn, Sr, Rb, Cs, Mo, Pb) as well as major element concentrations (e.g. Na, K, Ca, Mg) from small sample volumes of uid inclusions from short transient signals acquired by ICP-MS. 3 Even isotope ratio determinations have been carried out on uid inclusions using this highly versatile technique. 4 The basic principles for LA-ICP-MS analysis of uid inclusions have been pioneered by Günther et al. 5 using a nanosecond UV laser ablation system (193 nm, ArF excimer) in combination with a quadrupole ICP-MS (QMS). The method was rened during the last decade by Günther et al., 6 Halter et al., 7 Heinrich et al. 3 and Allan et al., 8 but did not change in general. Besides the common determination of cation contents, it is also possible to quantify elements that exist as anions in the uid inclusions, such as Cl, Br, and S. 9,10 High control over the laser ablation process is required to achieve representative uid inclusion analysis. Problems can occur when fragments of the host mineral (most commonly quartz) quarry out during ablation and the uid is lost through cracks before the ablation process reaches the inclusion. Other problems arise from incomplete mobilization of the inclusion, or the lack in mass spectrometric acquisition speed leading to an under-estimation of the elemental concentrations. In order to achieve representative analysis the complete volume of the inclusion, which may be composed of gas, liquid, and crystals, needs to be mobilized and transported to the ICP-MS.
Opening the inclusions is described as one of the most critical steps in uid inclusion analysis. 5 Especially the analysis of shallow inclusions (<10 mm depth) is oen accompanied by uncontrolled release of the uid inclusion content due to overpressure in the inclusions and splashing of the material out of the ablation pit (Fig. 1). Allan et al. 8 estimated that the sampling process frequently controls the precision achieved for uid inclusion analysis. Depending on their extent, most elemental concentrations can be determined with a precision on the order of 20-30% RSD. Better precision values have been reported for elements with high concentration levels (e.g. up to the wt%-range) in relatively large uid inclusions. 2 In this study, we focus on smaller inclusions with an elemental load of #53 mg g À1 , which is of high signicance for the chemical characterization of natural uid inclusion but analytically challenging.
Quadrupole MS systems have been favoured over magnetic sector eld mass spectrometers as they are capable of fast peak scanning which is essential for the sequential analysis of the transient signal that is typically generated during laser ablation analyses of uid inclusions. Sweep times as short as 260 ms have been used for QMS analyses of a set of 20 isotopes. 2 However, more recently fast scanning sector eld mass spectrometers with a high transmission and a large dynamic detection range such as the Element XR™ from ThermoScientic™ have been developed. These are particularly suitable for short laser ablation analyses of low concentration levels.
Advantages of 196 nm UV-fs-LA over UV-ns-LA have been reported for the determination of isotopic ratios of heavy stable isotope systems 11-14 as well as Pb-U mineral dating, 15 but only preliminary experience has been documented for trace element analyses with single collector ICP-MS applications. 2 To the best of our knowledge, only one study has been published using femtosecond laser ablation (fs-LA) for the analysis of uid inclusions with respect to solute quantication. Borisova et al. 16 used a NIR-fs-LA-QMS system in their study. Pettke et al. 2 speculated that, particularly for uid inclusion analysis with fs-LA, problems such as cracking at the crater bottom and associated material loss or phase explosions and an uncontrolled opening of the inclusion may arise. Nevertheless, the greatest advantage of fs-LA is the minimal heat transfer from the laser spot into the sample during ablation 17 resulting in minimized elemental fractionation effects. 15 In this study, we present a new analytical setup for the determination of major, minor and trace element concentrations in uid inclusions. A 194 nm UV-fs laser is coupled with a fast scanning sector eld ICP-MS. To use the low heat transfer of the fs-laser to its full capacity, we adapted a heating-freezing stage as the ablation cell and performed the ablations at low temperatures (e.g. À40 C). This new approach was tested in this study by the analysis of different types of synthetic uid inclusions with known concentrations.

Instrumentation
Analyses have been carried out with an Element XR™ fast scanning sector eld inductively coupled plasma mass spectrometer (ThermoScientic™, Bremen, Germany) in combination with an in-house build laser ablation system which is based on a Spectra-Physics™ femtosecond (Ti:Sapphire) laser (Sol-stice™) operating in the deep UV at 194 nm. The laser system produces a pulse energy of 70-90 mJ in the fourth harmonic. This ultra short pulsed laser generates a so ablation with high control and avoids elemental fractionation at the sample site. The ablations of the standard reference materials (NIST610) were carried out with a repetition rate of 10 Hz. For uid inclusion analysis, the repetition rate was adjusted depending on the depth of the inclusions. For shallow inclusions, between near surface and 30 mm depth, a repetition rate of 2-5 Hz was used, resulting in signal intensities signicantly above the respective detection limits (Table 1). Inclusions deeper than 30 mm were analyzed with 10 Hz for faster drilling. The laser repetition rate controls the signal shape, intensity and nally the limits of detection (LOD) for the measured elements. A higher repetition rate results in shorter and higher signals and subsequently in better LODs. 2 The selected spot size on the sample surface is chosen to be bigger than the analyzed uid inclusion in order to guarantee that the whole uid inclusion is ablated and subsequently transported to the ICP-MS. An adjustable aperture in the beam path controls the spot size. It is possible to use spot sizes of up to 30 mm for the ablation of quartz. Ablations of quartz with bigger spot diameters would result in insufficient ablation rates, because of a reduced laser energy density. The spot size was held constant during the analysis which differs from the ablation procedure described by Günther et al. 5 With our technique there is no need for stepwise opening of the inclusions, because an explosion of the inclusion or splashing of the

Isotope
Background 3SD background Signal Flinc material out of the ablation pit can be excluded when the inclusions are frozen. Especially the analysis of CO 2 bearing inclusions are expected to be simplied since the pressure is stringly decreased upon phase transformation of CO 2(gas) to CO 2(solid) . We expected that this approach works especially with femtosecond laser pulses due to their low thermal energy transfer, keeping even uid inclusions with low ice melting temperatures (e.g. À65 C) in their frozen state. The heataffected zone during fs-LA has been investigated in earlier studies. Hirayama and Obara 18 showed that the layer affected by heating with a femtosecond laser is only of a few nm thickness. In contrast, nanosecond laser ablation results in signicant conductive heat transfer within a layer of several mm. 19 For analyses, the Element XR™ is operated in "speed mode" which provides an optimized data acquisition for short transient signals as produced during LA of uid inclusions. Due to a faster magnetic scanning and a faster scanning of the electrical eld in the electrostatic analyzer unit, the sweep time is much shorter than that commonly used for analyses with sector eld instruments. For 20 isotopes ranging between 9 Be and 209 Bi, one sweep takes 477 ms. A short sweep time is essential for uid inclusion measurements, because the signal peak commonly occurs subsequent to the opening of the uid inclusion (Fig. 2) and a slow data acquisition may result in a preliminary signal cut-off and articially fractionated element concentrations. Further ICPMS operational settings are shown in Table 2.
For this study, we used a modied heating-freezing cell for laser ablation in order to overcome some fundamental limitations of LA-ICP-MS analysis of uid inclusions. One of the aims was to improve the uid inclusion opening procedure and thereby avoid the explosive ejection of the material during opening. With frozen inclusions and the small ablation rate of the UV-fs laser, we aimed to enhance the control during the opening process and as a result to improve the success rate of uid inclusion analyses.
While frozen, cracking at the bottom of the uid inclusion would only result in the loss of gases, which are not frozen and cannot be detected with the instrumentation used. Furthermore, we expected that the transient signal could be stretched in time using low repetition rates and frozen solid uid inclusion. Additionally, we aimed to improve the standardization process by using frozen matrix matched calibration standards prepared from standard solution in addition to the NIST610 and NIST612 glass standards. The matrix-matched standards enabled precise quantications even at low counting rates. Independent of the external standard used, the precision of the analyses still depends on the size, shape and position of the inclusions, as well as on the laser repetition rate.
The analytical setup was realized with a modied 'HCS622V' cell from INSTEC™, Colorado, USA. The cell is vacuum tight and provides an appropriate cell window to sample distance suiting the use of an objective with a focal length of 20 mm. It provides precise temperature control over the range from À190 C to +600 C. To optimize the cell volume, the cell was modied to carry a removable 3 cm 3 cell made from Teon, which is sealed to the main cell by means of O-rings. The small cell connects to the gas inlet and outlet with Teon tubes. With this modication the washout time is reduced dramatically. Helium, a gas with much higher thermal conductivity compared to Argon, can be used with ow rates of up to 0.8 L min À1 limiting the lowest reachable temperature to À100 C. Without the small Teon cell we observed a strong signal loss between room temperature and À40 C which exceeded 70% for all measured isotopes due to the rising viscosity of helium at lower temperatures. This deciency was reduced to a signal loss of <5% when the modied Teon cell was inserted. Furthermore, the transient signal shape from raster analysis, e.g. on NIST glasses, does not differ from those obtained at ambient temperature. The cooled sample area is large enough to provide space for a standard reference material (SRM) and several sample chips. Hence it is not necessary to open the cell during the data acquisition for calibration.

Calibration
To prove that LA analysis at low temperatures results in adequate values, we measured the NIST612 glass against the NIST610 glass at À40 C. Results indicate that the preferred concentration values of most elements (taken from the GeoRem database (http://georem.mpch-mainz.gwdg.de)) can be measured with an accuracy of AE5%.
To test the performance of our method for analysis of frozen liquids, we analyzed a self-prepared standard solution against the NIST610 glass (Fig. 3). As low partition coefficients of most elements between aqueous solution and ice results in the formation of micro-nuggets on grain boundaries, it was necessary to freeze the solution with a high cooling rate to hamper the growth of ice crystals and generate small grain sizes resulting in a more homogeneous elemental distribution. Raster ablations (2 Hz, 30 mm spot size) on larger areas were carried out to contribute to homogeneous sampling. The calculated results agree within analytical uncertainties (1RSD) with a specic value of 108 mg g À1 (Fig. 3).
Because of easier handling, for uid inclusion analysis the NIST610 glass rather than frozen standard solutions was used for external calibration of element ratios in the uids. The analyses were performed using the standard-bracketing method, with a standard ablation aer every fourth sample. Final ice melting temperatures were determined by microthermometry and provided the NaCl eq. values which were used as the internal standard for the calculation of elemental concentrations 5 using the SILLS soware.

Samples/synthetic uid inclusions
To test our analytical setup we used synthetic uid inclusions in quartz, which were generated from a standard solution of known concentration. The solution was prepared as a 1 : 1 mixture from the MERCK™ VIII ICPMS-multi-element-standard and a 20% NaCl solution; its composition is given in Table 3.
Preparation for uid inclusion synthesis generally followed the workow described by Bodnar et al. 20 with minor modications. Cores of 2.5 mm diameter and ca. 2 mm length were drilled from inclusion free alpine quartz, heated to 350 C and subsequently immersed in concentrated hydrouoric acid for 10 minutes to widen the cracks.
Two experiments were performed, one in a gold (ID188-Au) and one in a platinum (ID189-Pt) capsule, respectively, with dimensions of 20 mm length, 3.2 mm outer diameter and a wall thickness of 0.2 mm. In each capsule one pre-fractured quartz cylinder was placed together with 5 mg silica gel (to accelerate crack healing) and 30 ml of the standard solution. The capsules were pressurized to 200 MPa and heated isobarically in a rapidheat/rapid-quench hydrothermal autoclave of a design described by Matthews et al., 21 using argon as pressure medium. Uncertainties of temperature and pressure measurements are considered to be AE5 C and AE5 MPa, respectively. Aer a runtime of 5 days the autoclave was slowly cooled to room temperature. The recovered capsules were weighed to check for potential leaks during the run. The quartz cylinders were cleaned, dried and embedded in Araldite to be cut and polished to chips of 300 mm thickness.
From both experiments the quartz chips contained abundant synthetic uid inclusions and 55 inclusions were selected for microthermometric and LA-ICP-MS analyses. Microthermometric measurements of ice melting temperatures T m (ice) of À6.2 AE 0.1 C correspond to a salinity of 9.47 AE 0.13 mass% NaCl eq. , calculated aer Bodnar, 22 which is in full agreement with the standard solution composition.

Ablation procedure
In this study we followed the workow for uid inclusion analysis reported by Heinrich et al. 3 Since the inclusions are analyzed in a frozen state, problems such as explosive opening and sputtering of the inclusion content during ablation could be excluded, and we were able to use the so-called 'straight ablation' for the measurements. In contrast to the stepwise opening procedure which is used especially for polyphase inclusions, the "straight ablation" procedure holds some advantages, as reported by Pettke et al. 2 These are (1) a lower amount of surface contamination, (2) higher signal/background ratios and (3) lower limits of detection (LOD).
Prior to ablation the samples and reference materials were cleaned with deionized water and acetone, and the heatingfreezing cell was wiped out carefully with dilute nitric acid. The positions of the inclusions were mapped off-line prior to the analyses using a standard petrographic microscope, which simplies the search with the video system of the LA stage.  In the rst step the uid inclusions are frozen quickly by lowering the sample temperature to À100 C. Depending on the uid inclusion chemistry, the temperature is subsequently raised to a temperature sufficiently below the solution's eutectic point, so that the inclusions remain entirely frozen. Reference materials were analyzed at the same temperatures since the tuning of the mass spectrometer is also performed under these conditions. For the synthetic uid inclusions the heatingfreezing cell was set to À40 C, since they were generated from a simple binary NaCl-H 2 O uid.
A gas blank of at least 40 seconds was recorded prior to each single analysis before the start of ablation. The scan speed for the raster pattern performed on the used SRM was set to 20 mm s À1 .

Data processing
For calculation of trace element concentrations in uid inclusions it is necessary to separate the inclusion signal from the chemically distinct signal of the host mineral. We used the data reduction soware SILLS, 23 which is able to separate the short transient signal of the inclusion by a matrix correction. The soware follows the procedures and equations from Allan et al., 8 Halter et al., 7 Heinrich et al. 3 and Longerich et al. 24 As silicon in the inclusion is negligible compared to silicon in the quartz, it was used as the matrix-only tracer. Three integration windows for background, matrix and inclusion signal were dened for each single analysis. The length of the integrated inclusion signal was adjusted to that of Na, because Na was used as the internal standard for the quantication of the elemental concentrations. Sample compositions were calculated using the mass balance approach of SILLS.
As reported by Pettke et al., 2 the best way to determine concentration values and analytical uncertainties for LA-ICP-MS analyses of uid inclusions is to calculate the mean value and the external error from a batch of analysis of individual inclusions belonging to the same assemblage. The analytical error is based on the external precision (1SD) and is dened as the relative standard deviation (RSD) in %. Since all inclusions in our samples are considered to be chemically identical, we calculated the average concentration and the RSD from all single analyses of one sample.

Accuracy and precision
We used synthetic inclusions to check the analytical precision and accuracy of our method. By dividing the measured concentration with the theoretical value of the starting solution used for the high pressure/high temperature (HP/HT) experiment, the relative accuracy of these measurements can be estimated. The experimental solution contains 24 elements from the MERCK™ VIII ICPMS-standard-solution with known concentrations (Table 3). However, only 10 of them could be recovered in the synthesized uid inclusions.
The missing elements probably either reacted with the capsule material (e.g. the transition metals with Au or Pt) or have been enriched at the surface of the quartz host phase due to diffusive processes, like Li, Mg, Al and K. Consequently the original concentration of $53 mg g À1 in the experimentally synthesized uid inclusions was found for only a few elements (Fig. 4). The two samples ID188-Au and ID189-Pt were analyzed in separate sessions. In total 53 of 55 measured inclusions could be analyzed successfully. Only two inclusions that were located directly under the sample surface could not be used for quanti-cation, as a matrix correction could not be performed due to the missing separation of the host mineral signal from the inclusion signal. Single values, which lie beyond the 2SD threshold, were considered as outliers and have not been used for the calculation of mean values and standard deviations (Tables 4 and 5). Note that 195 Au and 197 Pt are not contained in the MERCK™ VIII standard solution and were likely leached out from the capsule material. The results for 11 B, 111 Cd, 125 Te, 203 Tl, 208 Pb, and 209 Bi (Fig. 4, Tables 4 and 5) indicate that these elements have been nearly completely recovered in the synthetic uid inclusions and were measured with an accuracy relative to the starting composition of the uid of >90%. The calculated mean value for these elements is in accordance with the original uid compositions and is within analytical uncertainties (1RSD) i.e. between 13% ( 203 Tl) and 25% ( 125 Te) for sample ID188-Au and 17% ( 111 Cd) to 25% ( 11 B) for sample ID189-Pt (Fig. 5). 39 K was only measured for sample ID189-Pt and could also be detected in the host mineral (quartz). Aer matrix correction, the determined K concentration of 42 mg g À1 (RSD ¼ 37%) is in accordance with the original uid.

64
Zn was only analysed in some inclusions of sample ID189-Pt, with an average value of 37 mg g À1 (25% RSD). The concentrations determined for 9 Be, 59 Co, 69 Ga, 88 Sr and 137 Ba in the synthesized uid inclusions of both samples are signicantly below the concentration of the original standard solution. Furthermore, the reproducibility of these elements was lower than that of the other elements in most cases. Beryllium, Co, and Ba show the highest variability with RSDs between 40% and 60%. Gold and Pt have also been analysed, because they were expected to be leached out of the capsule walls during the experiments. The inclusions in sample ID188-Au which were synthesized in gold capsules contain 1.6 wt% Au (RSD ¼ 29%), which seems to be an extraordinary high value compared to Au concentrations in natural uid inclusions that do not exceed the lower mg g À1 range. 25,26 The mixture between the MERCK™ VIII solution, which is 1 M HNO 3 , and the NaCl-uid may support the leaching of the capsule material during the experiments by formation of nitro-hydrochloric acid. Likewise, the inclusions in sample ID189-Pt which were synthesized in platinum capsules contain higher concentrations of Pt (124 mg g À1 ) which is heterogeneously distributed (with 73% RSD) and 11 mg g À1 Au (RSD ¼ 26%). The inclusions can be separated into two groups with $200 mg g À1 and $50 mg g À1 Pt, respectively. Both groups have relatively homogeneous Pt concentrations with an RSD of 38% respective 34%. This may indicate formation of two generations of uid inclusions in experiment ID189-Pt, which can only be identied due to their Pt concentration since all other elements show similar values in both groups. None of the other analyzed elements, except Pt and Au, could be identied to originate from the capsule material.

Conclusions
We have tested a new analytical setup for the determination of trace element concentrations in uid inclusion with fs-LA-ICP-MS. Ablation with femtosecond laser pulses allows analyzing uid inclusions in their frozen state. With this technique, 53 of 55 synthetic uid inclusions in quartz, covering a size range between 8 mm and 25 mm and down to a depth of 50 mm, could be analyzed successfully. The 194 nm UV-fs-laser allows excellent control over the opening procedure of uid inclusions. Due to the low thermal heat transfer onto the sample surface the inclusions do not melt while the sample material is mobilized. It is even possible to measure very shallow inclusions that frequently explode, resulting in material loss during analysis under room temperature conditions with a nanosecond laser. The use of a fast scanning magnetic sector eld ICP-MS with a SEM and a Faraday detector allows the detection of elements across the concentration range from mg g À1 to wt%. The lower limits of detection for uid inclusion analysis vary between 0.1 mg g À1 (for 209 Bi) and 10 mg g À1 (for 39 K). The detection of 20 isotopes over the whole mass range takes 477 ms. The typical analytical uncertainty ranges between 15% and 30% (1RSD). This is a signicant improvement in precision compared to earlier studies for samples with such low concentration levels. Elements from the stock solution which did not react with the capsule material or host mineral during the HP/HT experiments could be fully recovered in the synthetic uid inclusions at their original concentrations. The results illustrate that our method is able to produce adequate data from natural uid inclusions.
With respect to the time consuming work for preparation and characterization of uid inclusion studies, especially microthermometry, our method offers a highly efficient tool for solute quantication in uid inclusions with a success rate of >90%. First tests with natural inclusions in quartz have shown that the success rate can be expected to be similar to our results. Given that most natural samples oen host only very few inclusions of sufficient size ($10 mm), this method provides new possibilities for uid inclusion studies. The applicability of our approach to the widespread UV-ns-LA systems needs to be tested. If successful, it may become a useful tool for the opening of uid inclusions by LA-ICP-MS analysis.