Environ. Sci. Technol. 2003, 37, 2267-2273
Accuracy of Continuous Ice-Core Trace-Element Analysis by Inductively Coupled Plasma Sector Field Mass Spectrometry STEFANIE KNU ¨ S E L , * ,†,‡ D A V E E . P I G U E T , ‡ M A R G I T S C H W I K O W S K I , * ,‡ A N D H E I N Z W . G A¨ G G E L E R † , ‡ Paul Scherrer Institute, CH-5232 Villigen PSI, CH-5232 Switzerland, and Department of Chemistry and Biochemistry, University of Berne, Freiestrasse 3, CH-3012 Berne, Switzerland
Trace elements trapped in glaciers are important indicators for the characterization of past biogeochemical cycles, the identification of numerous sources and their varying strength, and thus indirectly provide insight into past climate variations. However, this necessitates highly resolved and continuous records of trace elements in ice. To obtain records corresponding to these requirements, a continuous ice-core melting (CIM) device was coupled to an inductively coupled plasma sector field mass spectrometer (ICP-SFMS). Accuracy of this newly developed method was tested by replicate analysis of longitudinally cut icecore sections (reproducibility) and by comparing results of the continuous method with the conventional decontamination and analysis procedure. The new, fast method is suited to accurately determine concentrations of a number of elements, such as Li, Na, Mg, Ca, Mn, Co, Br, Sr, Mo, and Tl. However, for 18 elements (including Al and lanthanides) observed concentrations were underestimated when analyzed using the continuous method. Possible explanations of these low concentrations are (i) incomplete dissolution of mineral dust particles contained in the ice resulting from a delayed acidification step and/or (ii) adsorption of dissolved trace elements or mineral dust particles on the surface of the ice melting device.
Introduction Impurities trapped in ice cores give precious information on the paleo atmosphere and consequently deliver insight into past climate variations, knowledge of which is a precondition for the prediction and interpretation of future climate. Trace elements (herein referred to elements with a concentration range in ice of pg g-1 to µg g-1, although some may be major or minor constituents of the earth crust) are of increasing interest because they provide the ability to identify various sources (1, 2) and to understand biogeochemical cycles (3). In particular the study of trace elements in dust contributes to the understanding of past circulation systems and the estimation of source strengths (4-7). Additionally trace elements analysis is suited for the determination of natural background concentrations as well as anthropogenic pol* Corresponding authors phone: +41 56 310 4397; fax: +41 56 310 4435; e-mail:
[email protected] (S.K.) and margit.
[email protected] (M.S.). † University of Berne. ‡ Paul Scherrer Institute. 10.1021/es026452o CCC: $25.00 Published on Web 04/17/2003
2003 American Chemical Society
lution in ice cores from around the globe. Recent investigations of Antarctic ice cores provided evidence that quiescently degassing volcanoes are the major source of trace metals, whereas continental dust is of minor importance (8). Anthropogenic pollution was observed in both Antarctic (9) and Arctic ice cores, where for instance it has been demonstrated that lead pollution in the atmosphere in the second half of the 20th century was caused by lead additives in gasoline (10). Increased concentrations of platinum group metals in Greenland snow dated to the mid-1990s (11) and decreasing lead concentrations in Greenland snow over the past 20 years appear to be related to the use of automobile catalytic converters and consequently the diminishing use of lead additives in gasoline (12). Ice-core records of trace elements from mid- and low-latitudes have also provided insight into pollution on a more regional scale, e.g. in the Alps (13-15) or Andes (16). Conventional methods to analyze trace elements in ice include a decontamination procedure performed by chiselling away outer, possibly contaminated layers under extremely clean conditions (17, 18). The chiselling procedure also includes the risk of carrying contamination from the outer layers to the inner, “clean” portion of the core. Furthermore, this procedure is labor-intensive and timeconsuming, because the very low natural concentrations of trace elements in snow and ice can easily be influenced by contamination. Therefore the number of samples processed as well as the spatial resolution is limited. Thus most of the trace element records available in the literature are discontinuous (e.g., refs 8, 11, 15, and 16) or recorded with a coarse depth resolution of about 10-20 cm (8, 12, 19). There is a need for continuous, highly resolved records of trace elements particularly in areas with low accumulation or in glaciers with strong layer thinning. To obtain higher spatial resolution, new decontamination methods have been developed including the use of an ice holder, which provides a depth resolution of 2 -3 cm (20). Another approach was presented by Reinhardt et al. (21) who applied laser ablation inductively coupled plasma mass spectrometry (ICP-MS) to frozen ice cores and who achieved an extremely high resolution of 0.3-1 mm. However, this method might be disturbed by either the inhomogeneity of dust-containing ice layers or the inclination of layers. McConnell et al. (22) reported a promising method which included the coupling of a melting head to a quadrupole ICP-MS. The melting head decontaminated the ice by separating the meltwater from the inner part of the core from the surface meltwater, thus greatly simplifying the decontamination procedure. In addition to the benefits of a shorter procedure and the reduced risk of contamination of this modified process, a high spatial resolution of about 1 cm was obtained. In the study of McConnell et al. (22) only the reproducibility of the continuous method was tested, whereas the results were not compared with those from the conventional method for analyzing trace elements in ice cores. Additionally, neither procedural blank concentrations nor detection limits were reported. Here we present a detailed study of the analysis of trace elements in ice using continuous ice melting (CIM) ICP sector field mass spectrometry (SFMS), including a determination of procedural blanks and reproducibility as well as a comparison with the conventional analysis method. Our main findings are that this new method appears to be suitable for the analysis of one group of trace elements, while serious difficulties were observed for a number of trace elements that often occur in nature as silicates. VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2267
FIGURE 1. Stratigraphy of the Illimani ice-core section from a depth of 45.45-46.1 m. Oval air bubbles (ovals) of about 0.5 mm in width and 1-2 mm in length, round air bubbles (circles) of about 0.3-0.5 mm in diameter, and a yellow dust layer (hatched mark) were observed. Black vertical lines mark breaks.
Methods Materials. All sample tubes (30 and 125 mL bottles, highdensity polyethylene (HDPE), 30 mL bottles, fluorine ethylene polypropylene (FEP)), auto sampler vials (1.5 mL, polystyrene), syringes (10 mL, polypropylene (PP)), and filter holders used for trace element analysis were washed three times with ultrapure water (18.2 MΩ quality) and twice with 0.2 M HNO3 and were dried in a class 100 flow bench prior to use. Throughout the preparation and analysis of samples, wrist length PE gloves were worn by the analyst. Blank ice-core sections produced by freezing ultrapure water in previously cleaned PE tubes were employed for determining the procedural blanks. Sample tubes utilized for ion chromatography (IC) and ICP optical emission spectroscopy (OES) were rinsed five times with ultrapure water. Ultrapure nitric acid (Ultrex, J. T. Baker) was used throughout the analysis. Standards for ICP-SFMS were prepared every 3 weeks in HDPE bottles by diluting a stock standard solution made from single-element standards in a 30 mL FEP bottle. Stability of these standards was monitored. Ice Core Sections. Two ice-core sections were used to test reproducibility and accuracy, one from a temperate glacier at the Jungfraujoch, Switzerland (7°59′E, 46°32′N, ∼3500 asl), and the other from a cold glacier at the Illimani, Bolivia (23). Figure 1 illustrates the stratigraphic features of the Illimani ice-core section, including a clearly visible 1.2 cm yellow dust layer at a depth of 45.49 m. The ice-core sections (diameter: 7.6 cm) from the Jungfraujoch and Illimani with densities of 0.88 g cm-3 and 0.9 g cm-3, respectively, were cut longitudinally in a cold room (-20 °C) to yield to seven subcores each with a cross section of 2.2 cm × 2 cm for the continuous analysis. For this purpose a commercial band saw was modified and cleaned according to Eichler et al. (24). The cross section of these subcores is sufficiently large to cover the CIM melting head (inner diameter: 10 mm, see below). The opposing rough ends at each break within both ice-core sections were decontaminated in the cold room by removing the outermost 5 mm using a ceramic knife precleaned with ultrapure water. Without touching the ends of the cores, the ice-core sections were stacked in a Plexiglas tube (i.d. 32 mm) that was then fixed above the melting head. Blank ice-core sections (length: about 10 cm) were placed before and after the natural ice core to determine the procedural blank during an analysis. The melting head was rinsed 10 times with 0.5 mL ultrapure water prior to use. Continuous Ice Melting Device. The newly developed continuous ice melting (CIM) ICP-SFMS device (Figure 2) is comprised of a melting head coupled to a double-focusing sector field ICP-MS (ThermoFinnigan, Element 1). The melting device, set up in a freezer at -20 °C, consists of a coated aluminum melting head, which is similar to the one used by Sigg et al. (25). Major improvements of our melting head include the per fluor alkoxy (PFA) coating, a heating with three 100 W heating cartridges and the design of the 2268
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 10, 2003
FIGURE 2. Design of continuous ice melting (CIM) inductively coupled plasma sector field mass spectrometry (ICP-SFMS) device. Thick black lines with arrows mark capillaries, enabling the transport of meltwater from the melting head to degasser, HPLC pump, mixing tee, and nebulizer. The meltwater is acidified at the mixing tee by continuously adding a solution of 1 µg g-1 103Rh in 13.5 M HNO3 to the flow of ice-core meltwater. The melting head is located in a freezer at -20 °C. drains. These drains, consisting of PTFE tubes inserted into “pilot” holes bored in the aluminum melting head, are connected to the tubing so that the meltwater is never in direct contact with the aluminum. The diameters of the inner and outer circle of the melting head are 10 and 36 mm, respectively. The heating is controlled by a temperature sensor. The design of the melting head allows continuous decontamination of the ice cores while melting, by maintaining separation between the meltwater flow originating from the inner part of the ice core and from the potentially contaminated outer part. This system is suited only for ice and not for porous firn, because capillary forces in firn (26) may transport contaminated meltwater from the outer to the inner part of a firn core. The speed of ice melting and the pump rate were adjusted so that there was a flow from the inner circle to the outside. To avoid clogging of the capillaries (see Figure 2) by insoluble particles (e.g. dust) in ice, a PP filter with a pore size of 45 µm was placed on the inner circle of the melting head. The melting head was heated to 45 °C, resulting in a melting speed of 0.4-0.5 cm min-1. To achieve a constant melting speed a weight of 115 g was placed on top of the melting ice core. This weight was connected to a preset potentiometer, which recorded the progress of the melting process. The meltwater from the inner part of the ice core flowed, due to gravitational forces, through a polytetrafluorethylene (PTFE) capillary of 0.5 mm inner diameter (i.d.) through a degasser into an inert HPLC pump (Sykam, Peek (poly-etherether-ketone) equipped model with a micro pumping unit) with a flow rate of 100 µL min-1. The degasser removed air bubbles, which were contained in the ice. A 20 cm long (0.05 mm i.d.) pressure capillary connected the HPLC pump with a mixing tee (pressure: about 20 bar). A pump alert system gave an acoustic warning when the flow dropped. At the mixing tee an infusion pump (Precidor, type 5003, flow rate: 2 µL min-1) equipped with a gastight syringe (PTFE plunger tip) continuously added an internal standard solution consisting of 1 µg g-1 103Rh in 13.5 M HNO3 to the ice-core meltwater, continuously pumped by the HPLC pump. After the mixing tee, the meltwater from the ice core became an acidic solution of 0.26 M HNO3 containing 26.2 ng g-1 of 103Rh. Both pumps were able to deliver a highly precise flow, which is necessary to maintain this 103Rh concentration in the meltwater. The acidified meltwater was
TABLE 1. Nuclides Analyzed Using ICP-SFMS in Low- and Medium-Resolution Mode low resolutiona
7Li, 23Na, 24Mg, 27Al, 75As, 79Br, 85Rb, 88Sr, 95Mo,
medium resolutionc
44Ca, 51V, 55Mn, 56Fe, 59Co, 103Rhb
103Rh,b 111Cd, 121Sb, 133Cs, 138Ba, 139La, 140Ce, 141Pr, 146Nd, 147Sm, 205Tl, 208Pb, 209Bi, 232Th, 238U
a Low resolution mode: 300 (10% valley definition). standard. c Medium resolution mode: 3500.
b
Internal
pumped through a 0.17 mm i.d. PTFE capillary to the inlet system of the ICP-SFMS. The total volume between the melting head and the nebulizer is about 450 µL, which is equivalent to a retention time of 4.5 min. The spatial resolution in the ice core obtained by this continuous technique is estimated to be 1 cm, which cannot be reduced further because of mixing in the inner circle of the melting head, which tapers off, and mixing, turbulence and diffusion processes during the transport from the melting head to the nebulizer. Conventional Decontamination. Conventional decontamination was performed according to Boutron and Batifol (27). All operations were carried out under class 100 to 1000 laminar flow benches. An ice-core section with an original diameter of 7.6 cm was cut longitudinally to a quarter of its former cross section and to a maximum length of 30 cm. This subcore was fixed in a Plexiglas holder and decontaminated in a cold room at -20 °C using stainless steel scalpels precleaned with ultrapure water (28). The operator wore shoulder length gloves under wrist length PE gloves during the decontamination procedure. Consecutive outer layers of the ice core were removed in three steps. Between each step, the scalpels and sample holder were cleaned of snow shavings and gloves were replaced. The decontaminated icecore section was cut in half directly into 125 mL HDPE containers without touching the decontaminated surfaces of the ice. A depth resolution of about 6-10 cm was obtained. Prior to melting, the ice was acidified to give 0.2 M HNO3 (14). The meltwater was filtered using a filter holder equipped with a PP filter of 45 µm pore size, which was rinsed prior to the use with 10 × 10 mL 2 M HNO3, 5 × 10 mL 0.2 M HNO3, and 5 × 10 mL ultrapure water. The purity of the filter was checked by analyzing the filtered ultrapure water. Twenty microliters of an internal standard solution consisting of 1.3 µg g-1 103Rh solution in 0.24 M HNO3 was added to 980 µL of the sample before analysis. To determine the procedural blank, ice-core sections prepared of ultrapure water were decontaminated according to the procedure described above. Moreover, the efficiency of the decontamination procedure was monitored by analyzing ice removed from the outer layers of the Illimani ice-core section, resulting in a three-step depth profile from the outer to the inner sample. ICP-SFMS. The inlet system of the ICP-SFMS consisted of a PFA nebulizer (Elemental Scientific, aspiration rate of 100 µL min-1) and a water-cooled (3 °C) Scott-type spray chamber. The plasma was operated at 1350 W. Tuning parameters such as gas flows and ion optics were adjusted daily to optimize the intensity of In in the respective resolution mode using a tune-up solution (Merck) diluted to 2 ng g-1. The sensitivity achieved in the low-resolution mode (LRM, resolution: 300) was 150 000 to 200 000 cps per ng g-1 In. Mass calibration in the LRM as well as in the mediumresolution mode (MRM, resolution: 3500) was performed daily. Medium resolution had to be applied for the analyses of certain elements that were interfered with by oxides, argides, or doubly charged ions in the LRM (e.g., ref 29). Table 1 lists the nuclides analyzed and their corresponding resolution. 103Rh was used as internal standard to correct for intensity fluctuations in the inlet system.
Calibration. An external calibration was performed using five standards of adjusted concentration for each element in order to cover the concentration range of the ice samples. The intensities of the standards were fitted using linear regression, and the y-axis intercept at zero concentration, which was assumed to represent an average blank of the standards, was subtracted. The correlation coefficient of the calibration line was g 0.999. Data Analysis. Raw data from the ICP-SFMS were evaluated with a MATLAB (MathWorks Inc.) program established for this specific statistical analysis. Initially, a first outlier test was performed using a 4σ limit of the 103Rh data, which was calculated from the logarithmic values truncated on both sides by 1.5-10%. The 103Rh outliers were assumed to originate from an inconstant flow of either the HPLC- or the infusion pump and therefore raw data associated with outlying 103Rh concentrations were deleted. A second outlier test, the Dixon test (30), was then used to remove outliers attributed to spikes in the mass spectrum, which were observed to increase the peak area up to 1 order of magnitude. Such noise spikes in the mass spectrum may have been caused by the nebulization of insoluble dust particles contained in the ice, re-evaporation of droplets situated on the wall of the spray chamber, or by the fall of drops in a poorly designed drain chain (31). The outlier-corrected data set was then averaged to a depth resolution of 1 cm (corresponding to five scans). Procedural blanks were subtracted, and for concentrations smaller than the detection limit (Table 2) a value of half of the detection limit was inserted. Alternative Analysis Methods. Samples for IC analyses (Na) and ICP-OES (Mg, Ca) were decontaminated and analyzed according to Eichler et al. (24). The depth resolution obtained by this method was about 3 cm.
Results and Discussion Procedural Blanks and Decontamination Efficiency. Concentrations of all elements in ultrapure water, represented by the y-axis intercept, were below or equal to the most diluted standard containing 1 pg g-1 of each element (except Al, Br, Ca, Mg, and Na: concentration range of the most diluted standard: 10-2000 pg g-1). Thus, the use of the ultrapure water for the preparation of standards and blank ice-core sections is justified. Procedural blanks were determined from blank ice-core sections, analyzed according to the decontamination and measurement procedures described above. The detection limit was calculated using the German industry norm (Deutsche Industrie Norm, DIN) 32645 (32) utilizing four blanks, three measurements per blank, the one-sided statistical value with a degree of freedom of three and a 95% confidence level. Tables S1 and 2, containing elements selected according to the behavior in the accuracy test (see below), list the resultant detection limits and the procedural blank concentrations for the conventional and CIM procedures. CIM blanks listed in Tables S1 and 2 represent blank concentrations prior to the analysis of the Illimani ice-core section. CIM blanks for Li, Na, Mn, Co, Br, and Tl (Table 2) were less than or equal to the conventional blanks, whereas elevated blank concentrations were observed for the elements Mg, Ca, Sr, and Mo (Table 2), which were probably caused by unfiltered air in the freezer (see Figure 2). CIM blanks reported in Table 2 equate to 0.3-17% of the average concentration in the Illimani ice-core section, with the exception of Li, Br, and Mo, where the blank contributions were 70-100%. Li and Br showed high blanks in both methods, which were probably due to interferences by 14N2+ and 40Ar38Ar1H+ in the low-resolution mode. In this case a higher resolution could not be applied, as sensitivity would then have been too low due to the reduced transmission. Mo had an elevated blank in the CIM procedure, which was VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2269
TABLE 2. Detection Limits (D.L.)a and Procedural Blank Concentrations of the Conventional and the CIM Analysis (pg g-1)d detection limita blank conv.b blank CIMc
d
Li
Na
Mg
Ca
Mn
Co
Br
Sr
Mo
Tl
16 97 (8.3 137 (44
660 1714 (120 2190 (82
55 390 (17 624 (79
1031 5510 (338 9605 (1400
44
