Heng
Chen
,
Zhen
Tian
,
Brenna
Tuller-Ross
,
Randy L.
Korotev
and
Kun
Wang
*
Department of Earth and Planetary Sciences, McDonnell Center for the Space Sciences, Washington University in St. Louis, One Brookings Drive, St. Louis, MO 63130, USA. E-mail: wangkun@wustl.edu
First published on 5th December 2018
In the last a few years, several groups proposed new methods of high-precision K isotope analysis using Multiple Collector Inductively Coupled Plasma Mass Spectrometer (MC-ICP-MS) either through “collision gas” or “cold plasma” methods. Here we report a detailed method analyzing K isotopes in high-precision using Neptune Plus MC-ICP-MS in cold plasma and conduct an inter-laboratory comparison. The precision of measurements of 41K/39K ratios of a single run in this study can routinely reach ∼0.05 per mil (95% confidence interval; n = 10). The long-term (20 months) robustness and reproducibility of this method have also been evaluated (0.11 per mil; 2 standard deviation; n = 890). We also report K isotopic compositions of 20 geological reference materials and compare this method with recent methods from other laboratories. These independent measurements of the same reference materials agree well with each other within reported analytical uncertainties. Because of the newly improved methods and observed K isotopic fractionation among reference materials, the International Union of Pure and Applied Chemistry (IUPAC) recommended values for the weight of potassium and the atomic fractions of K isotopes need to be revised.
Potassium has two stable isotopes, 39K (93.2581%) and 41K (6.7302%), and one naturally occurring radioactive isotope 40K (0.0117%) with a long half-life (t1/2 = 1.277 × 109 years). The K stable isotopic composition is typically expressed in delta notation, where δ41K = ([(41K/39K)sample/(41K/39K)standard − 1] × 1000). The branched decay of 40K to 40Ca and 40Ar is an essential radioactive heating source and may have been even more important than U and Th in the early Earth. It may also be the possible energy source that kick-started the geodynamo, which produces the magnetic field of the Earth.8–10 The K–Ca–Ar system is also useful in chronology and is one of the most popular dating tools. The precise determination of K isotopic ratios (40K/39K) and the understanding of the K stable isotope fractionation among minerals would help to further improve the accuracy of K–Ca–Ar geochronology.
The natural variation of K isotopic compositions was not well known until very recent. The earliest measurements of potassium isotopic ratios were done with Dempster style mass spectrographs in 1920–30s.11–13 Brewer14 was the first to investigate the variation of K isotopic compositions among rocks, minerals and plants. This and many other studies in this era concluded that although there is measurable isotopic difference among biological samples (e.g., normal vs. cancerous tissues), the K isotopic variations of geological samples are negligible.14–24 Taylor and Urey25 successfully fractionated K isotopes by chemical exchange with zeolites; while Brewer26 didn't observe any K isotopic fractionation during evaporation experiments. After decades of relative inactivity, studies on K isotopes resumed in the 1970s.27–34 These investigations covered various types of samples, including terrestrial samples such as basalts, granites and mantle xenoliths, and extraterrestrial samples such as chondrites, eucrites, and lunar samples, especially after the return of Apollo missions. The major discovery during this period is that lunar regolith is significantly enriched in heavier K isotopes (i.e., 41K), and this observation was interpreted as volatilization and partial loss of K during the micro-meteorite impacting and solar wind sputtering on the surface of the Moon.27,31,35 Most of these studies were conducted on Thermal Ionization Mass Spectrometers (TIMS), and the error bars of these early measurements are typically larger than 1% (relative precision here and after) for 41K/39K ratios.34
Humayun and Clayton36,37 made a major contribution to the K isotope systematics of both terrestrial and extraterrestrial samples achieving the best analytical precision (0.5 per mil for 41K/39K ratio) at the time by using the Secondary Ion Mass Spectrometer (SIMS) technique. Intriguingly, there was still no detectable K isotopic difference between meteoritic, lunar, and terrestrial samples (except for lunar regoliths as had been discovered previously), even though there is a clear trend of volatile depletion as indicated by K/U ratios between the most primitive chondrites and the highly depleted Moon and Vesta. Potassium isotope studies were also carried out on tektites and microtektites using SIMS,38,39 and there were no detectable K isotopic differences between tektites on average and their possible source materials. Although experiments showed that kinetic isotope fractionation processes such as evaporation and thermal (Soret) diffusion would produce measurable K isotopic fractionation,40–42 no such fractionation has been observed in natural samples such as chondrules that exhibit loss of alkali metal volatiles including K.43,44
During the last two decades, the fast development of Multiple-Collector Inductively-Coupled-Plasma Mass-Spectrometry (MC-ICP-MS) has made it possible to improve the analytical precisions of many isotopic systems dramatically.45–48 The importance of K isotopes and the lack of high-precision K isotopic data have been realized by the geochemistry community. Morgan et al. pioneered the development of a high-precision K isotopic analytical method using MC-ICP-MS.49–52 More recent endeavors have also been made by Santiago Ramos and Higgins,53 Wang and Jacobsen,54,55 Li et al.,56,57 Morgan et al.,58 Santiago Ramos et al.,59 Hu et al.60
Here we report a method using a Thermo Scientific Neptune Plus MC-ICP-MS to further push the analytical limits on K isotope measurements and to test the long-term (20 month) robustness and reproducibility of such method. We also report a detailed procedure of ion-exchange chromatography to purify K with low blanks. We apply this protocol to 20 well-studied reference materials in various petrology and mineralogy to survey the range of K isotopic fractionation between igneous and sedimentary rocks and minerals, and also for the purpose of inter-laboratory comparison and calibration.
Sample dissolution protocol | |
---|---|
1 | Weigh 10–50 mg of each sample by electronic balance (depending on each sample's K concentration) |
2 | Dissolve samples in concentrated HF/HNO3 mixture (VHF![]() ![]() ![]() ![]() |
3 | Heat in Parr high-pressure digestion vessels at 150 °C in Fisher Scientific Isotemp 500 oven for two days |
4 | Evaporate the digested samples completely under heat lamps |
5 | Re-digest the samples by HCl/HNO3 mixture (VHCl![]() ![]() ![]() ![]() |
6 | Evaporate the samples again under heat lamps |
Big column purification | |
---|---|
1 | Big columns (ID = 1.5 cm, BIO-RAD Econo-Pac) filled with 17 mL AG50-X8 100–200 mesh cation-exchange resin |
2 | Resin cleaning: 100 mL 6 mol L−1 HCl |
3 | Resin conditioning: 50 mL 0.7 mol L−1 HNO3 |
4 | Load sample: the samples are digested in 1.1 mL 0.7 mol L−1 HNO3. After centrifuging the solution, load 1 mL supernatant onto the big columns |
5 | Matrix eluted by adding 82 mL 0.7 mol L−1 HNO3 |
6 | Pre-cut: 5 mL 0.7 mol L−1 HNO3 to monitor the recovery of K |
7 | K-cut: potassium collected in Teflon beakers by addition of 107 mL 0.7 mol L−1 HNO3 |
8 | Post-cut: 5 mL 0.7 mol L−1 HNO3 to monitor the recovery of K |
9 | Purified solutions are dried down under heat lamps overnight for small column |
Small column purification | |
---|---|
1 | Small columns (ID = 0.5 cm, BIO-RAD glass Econo-column) filled with 2.4 mL AG50-X8 100–200 mesh cation-exchange resin |
2 | Resin cleaning: 20 mL 6 mol L−1 HCl |
3 | Resin conditioning: 15 mL 0.5 mol L−1 HNO3 |
4 | Load sample: the samples are digested in 1 mL 0.5 mol L−1 HNO3, then loaded all onto the small columns |
5 | Matrix eluted by adding 14 mL 0.5 mol L−1 HNO3 |
6 | Pre-cut: 2 mL 0.5 mol L−1 HNO3 to monitor the recovery of K |
7 | K-cut: potassium collected in Teflon beakers by addition of 18 mL 0.5 mol L−1 HNO3 |
8 | Post-cut: 2 mL 0.5 mol L−1 HNO3 to monitor the recovery of K |
9 | Purified solutions are dried down under heat lamps overnight for analysis |
The 20 samples include 11 igneous rocks, 7 sedimentary rocks, and 2 potassium-rich minerals. They have been well studied and are readily available to other groups for future inter-laboratory comparison. The igneous rocks used in this study include three basalts from three tectonic settings: a continental flood basalt (BCR-1), a mid-ocean ridge basalt (BIR-1) and an ocean island basalt (BHVO-2). We also investigated one andesite (AGV-2), two granites (GS-N and G-2), one granodiorite (GSP-1), one quartz latite (QLO-1), one rhyolite (RGM-1), one tonalite (TLM-1), and one obsidian (NIST-278). Two pure K-bearing minerals were also studied: one sodium feldspar (NIST-99a) and one potassium feldspar (FK-N). The sedimentary rocks used in this study include one limestone (NIST-1c), two sediments (NIST-2704 and BSK-1), one marine mud (MAG-1), and three shales (SCo-1, SDO-1 and SGR-1). The sample names, petrological types, locations and literature data are listed in Table 2.
Sample name | Type | Location | Distributed bya | K (%) this study | K (%) literatureb | δ41K SRM3141a (‰) | 95% C.I.c | n |
---|---|---|---|---|---|---|---|---|
a USGS = U.S. Geological Survey; ANRT = Association Nationale de la Recherche Technique (Paris); NIST = U.S. National Institute of Standards and Technology. b The K concentrations are from ref. 63 except that the K concentration of TLM-1 and BSK-1 is from ref. 62 and those of AGV-1, BCR-1, BHVO-2, BIR-1, G-2, RGM-1 are from ref. 71. c 95% confidence interval. d Number of measurements. | ||||||||
Mafic | ||||||||
BCR-1 | Basalt | Portland, Oregon | USGS | 1.44 | −0.49 | ±0.05 | 8 | |
BHVO-2 | Basalt | Hawaiian Volcanic Observatory | USGS | 0.43 | −0.46 | ±0.04 | 13 | |
BIR-1 | Basalt | Reykjavík, Iceland | USGS | 0.025 | 0.024 | −0.51 | ±0.06 | 9 |
![]() |
||||||||
Intermediate | ||||||||
AGV-1 | Andesite | Lake County, Oregon | USGS | 2.44 | −0.43 | ±0.08 | 10 | |
![]() |
||||||||
Felsic | ||||||||
GS-N | Granite | Senones, Vosges, France | ANRT | 4.03 | 3.84 | −0.43 | ±0.08 | 13 |
G-2 | Granite | Bradford, Rhode Island | USGS | 3.74 | −0.46 | ±0.04 | 19 | |
GSP-1 | Granodiorite | Silver Plume, Colorado | USGS | 3.56 | 4.57 | −0.44 | ±0.05 | 10 |
QLO-1 | Quartz latite | Lake County, Oregon | USGS | 3.15 | 2.99 | −0.40 | ±0.04 | 11 |
RGM-1 | Rhyolite | Glass Mountain, Siskiyou County, California | USGS | 3.68 | 3.56 | −0.35 | ±0.04 | 16 |
TLM-1 | Tonalite | Side County, California | USGS | 1.38 | 1.31 | −0.46 | ±0.06 | 16 |
278 | Obsidian | Clear Lake, Newberry Crater, Oregon | NIST | 3.16 | 3.45 | −0.36 | ±0.07 | 12 |
![]() |
||||||||
Mineral | ||||||||
99a | Feldspar, sodium | Spruce pine pegmatite district of North Carolina | NIST | 3.98 | 4.30 | −0.08 | ±0.08 | 9 |
FK-N | Feldspar, potassium | Tamil Nadu, India | ANRT | 8.37 | 10.63 | −0.26 | ±0.09 | 8 |
![]() |
||||||||
Sedimentary | ||||||||
1c | Limestone, argillaceous | Putnam County, Indiana | NIST | 0.25 | 0.23 | −0.58 | ±0.03 | 14 |
2704 | River sediment | Buffalo River, Buffalo, New York | NIST | 1.67 | 2.00 | −0.41 | ±0.05 | 9 |
BSK-1 | Sediment | Kesterson National Wildlife Refuge, California | USGS | 2.16 | 2.30 | −0.37 | ±0.05 | 15 |
MAG-1 | Marine mud | Wilkinson Basin, Gulf of Maine | USGS | 2.17 | 2.95 | −0.44 | ±0.04 | 13 |
SCo-1 | Shale | Natrona County, Wyoming | USGS | 2.33 | 2.30 | −0.43 | ±0.04 | 16 |
SDO-1 | Shale | Morehead, Kentucky | USGS | 2.56 | 2.78 | −0.22 | ±0.08 | 9 |
SGR-1 | Oil shale | Green River formation, Wyoming | USGS | 1.02 | 1.38 | −0.26 | ±0.06 | 10 |
![]() |
||||||||
Other | ||||||||
Suprapur | 99.995% purity KNO3 | Merck KGaA | 0.00 | ±0.04 | 14 | |||
Blank K | Total-procedure blank | −1.31 | ±0.08 | 4 |
![]() | ||
Fig. 1 Big-column calibration using reference material BHVO-2 in 0.7 mol L−1 HNO3 (see Table 1 for details). |
The first column is a BIO-RAD Econo-Pac column (1.5 cm ID, polypropylene) filled with 17 mL AG50-X8 100–200 mesh cation-exchange resin. Columns were first cleaned with 100 mL 6 mol L−1 double-distilled HCl and then conditioned with 50 mL 0.7 mol L−1 double-distilled HNO3. After centrifuging, samples dissolved in 1 mL 0.7 mol L−1 HNO3 were loaded onto the first columns. Matrix elements were eluted by adding 82 mL 0.7 mol L−1 HNO3 and then pre-cuts were collected into disposable polypropylene tubes by adding 5 mL 0.7 mol L−1 HNO3. Potassium was recovered into acid-cleaned Teflon beakers by adding 107 mL 0.7 mol L−1 HNO3. Post-cuts were collected into additional disposable polypropylene tubes by adding 5 mL 0.7 mol L−1 HNO3. The recovered K was dried down and redissolved with 1 mL 0.5 mol L−1 HNO3 before loading onto the second column.
The second column is a BIO-RAD Econo column (0.5 cm ID, borosilicate glass) filled with 2.4 mL AG50-X8 100–200 mesh cation-exchange resin. Columns were first cleaned with 20 mL 6 mol L−1 double-distilled HCl and then conditioned with 15 mL 0.5 mol L−1 double-distilled HNO3. All 1 mL samples were loaded onto the second columns. Matrix elements were eluted by adding 14 mL 0.5 mol L−1 HNO3 and then pre-cuts were collected into disposable polypropylene tubes by adding 2 mL 0.5 mol L−1 HNO3. Potassium was recovered into acid-cleaned Teflon beakers by adding 18 mL 0.5 mol L−1 HNO3. Post-cuts were collected into additional disposable polypropylene tubes by adding 2 mL 0.5 mol L−1 HNO3. The recovered K was dried down and redissolved with 2% HNO3 before measuring with the mass spectrometers.
For our analysis, we need at least 10 μg K for ten measurements (1 μg K for each measurement: 1 ppm solution × 1 mL). For most samples, we aim for 100 μg K or more to be used for repeat measurements and for suppressing the blank contamination. In the minimal case of 10 μg K from the sample, the sample/blank ratio is 38. Assuming that the K isotopic composition of the blank is −1.31 per mil (in delta notation), the blank contamination will introduce a ∼0.03 per mil error (≈−1.31/38), which is well below the analytical uncertainty of our K isotope measurements. For most samples with >100 μg K, the sample/blank ratio is above 385, and thus the influences from the blank contamination are negligible. Low-blank chemistry is especially crucial for any low-K samples; fortuitously, most terrestrial crustal rocks are high in K abundance, and the blank contamination is negligible for all samples in this study.
We ensure that the yields of all samples through the column chemistry are more than 99% by monitoring the pre-cuts and post-cuts for both steps of the column chemistry. The concentrations of K in pre-cuts, post-cuts, and K-cuts of each sample were measured with a Thermo Scientific iCAP Q quadrupole ICP-MS. The yield of each sample is calculated as
Yield = (CK-cut × VK-cut)/(Cpre-cut1 × Vpre-cut1 + Cpost-cut1 × Vpost-cut1 + Cpre-cut2 × Vpre-cut2 + Cpost-cut2 × Vpost-cut2 + CK-cut × VK-cut) |
The Cpre-cut1, Cpost-cut1, Cpre-cut2, and Cpost-cut2 represent concentrations of K in the pre-cuts and post-cuts collected during the first and second steps of column chemistry. The CK-cut represents the concentrations of the final K solution collected after the second column. The VK-cut, Vpre-cut1, Vpost-cut1, Vpre-cut2, Vpost-cut2 are the final volumes of these solutions.
We also compare the K concentrations of the sample calculated from the final K-cuts and sample weights to the K concentrations reported in the literature.62,63 Considering the potential sample loss during digestion and transfer between containers and errors from weighing, pipetting and the quadrupole ICP-MS analysis, the K concentrations of the sample analyzed in this study agree well with the literature data within 10% (see Table 2). Both the calculated yields and this comparison to literature data confirm that we had collected ∼100% K from the samples and there is no possible isotopic fractionation via column chemistry.
![]() | ||
Fig. 2 Result of column fractionation tests. The shaded area is the isotopic composition of the standard by definition (0.00 ± 0.05‰; 95% C.I.). |
This “cold plasma” method relies on Neptune Plus' pseudo-high resolution (no total separation of the two peaks) to resolve the mass difference between 41K and 40Ar1H+ at a cold plasma condition. At regular ICP-MS operation state (high RF forward power of ∼1350 W), Ar+ and ArH+ are the major background peaks. At lower RF forward power (∼600 W or less), NO+ is the primary ion observed, while the Ar+ and ArH+ signals can be attenuated dramatically.64 In addition, because the 38Ar/40Ar is naturally extremely low, the interference of 38Ar1H+ on 39K can be neglected in this situation. The interference of 40Ar1H+ on 41K can be resolved at the “shoulder” of the peak (see Fig. 3).
Because the “cold plasma” mode would considerably suppress the intensities of potassium isotopes, we use an APEX Ω high sensitivity desolvation system produced by Elemental Scientific as the sample introduction system to increase the signal intensity. The typical parameters of the MC-ICP-MS and APEX Ω desolvation system are listed in Table 3. With this configuration, we can achieve ∼15 V for 39K at 1 ppm concentration, which is comparable to what we previously achieved with the “collision gas” method.54 All measurements were done with sample-standard bracketing technique, and the standard we used here is NIST SRM3141a.
MC-ICP-MS parameters | |
RF power | 600 W |
Cool gas flow rate | 15 L min−1 |
Intermediate gas flow rate | 1.0 L min−1 |
Sample gas flow rate | 1.15 L min−1 |
Focus quad | − 5.5 V |
Dispersion quad | 0.00 V |
Integration time | 8.389 s |
Number of integrations | 1 |
Number of cycles | 25 |
Number of blocks | 1 |
Cup configuration | L2 (39K), L1 (40Ar), Center (41K), H3 (44Ca) |
![]() |
|
APEX Ω parameters | |
H2 flow rate | 10 mL min−1 |
Ar (sweep gas) flow rate | 10 L min−1 |
Sample uptake rate | 100 μL min−1 |
Spray chamber temperature | 140 °C |
Peltier cooler temperature | 3 °C |
Desolvator temperature | 155 °C |
We measure each sample ∼10 times at 1 ppm concentration (±1%), and the averages of the ∼10 measurements are reported. The internal (within-run) reproducibility (95% confidence interval) of ∼10 measurements is given for each sample in Table 2, and the typical internal reproducibility is ∼0.05‰. The external reproducibility is calculated as the 2 standard deviations (2SD) of the measured aliquots of same reference materials (BHVO-2 and G-2) that passing through six columns (see Table 4). The typical external reproducibility (2SD) is ∼0.08‰. This analytical precision is also comparable to that of “collision gas” methods reported in literature.54,56
δ41K SRM3141a (‰) | 95% C.Ia | n | |
---|---|---|---|
a 95% confidence interval. b Number of measurements. | |||
BHVO-2 (basalt) | |||
Column 1 | −0.54 | ±0.10 | 10 |
Column 2 | −0.46 | ±0.07 | 9 |
Column 3 | −0.44 | ±0.10 | 10 |
Column 4 | −0.46 | ±0.05 | 10 |
Column 5 | −0.43 | ±0.06 | 9 |
Column 6 | −0.49 | ±0.08 | 10 |
Average | −0.47 | ||
2SD | 0.08 | ||
![]() |
|||
G-2 (granite) | |||
Column 1 | −0.50 | ±0.08 | 10 |
Column 2 | −0.43 | ±0.07 | 8 |
Column 3 | −0.47 | ±0.08 | 9 |
Column 4 | −0.45 | ±0.14 | 8 |
Column 5 | −0.48 | ±0.08 | 10 |
Column 6 | −0.53 | ±0.07 | 8 |
Average | −0.48 | ||
2SD | 0.07 |
In this study, we measured the K isotopic compositions of both the Merck KGaA Suprapur® 99.995% KNO3 and the NIST SRM3141a (see Table 2). We found that these two standards are indistinguishable from each other in term of K isotopes. This result is not coincidental or surprising since both standards might have been produced from the same industrial source of sylvite deposits. This identity of Suprapur KNO3 and the NIST SRM3141a is fortunate because we can directly compare the data from these four groups50–52,54–56,60 without conversion. However, for future analysis, we suggest further inter-laboratory calibrations of the standards used by different institutions.
In order to eliminate the matrix effects, a two-step column chromatography procedure has been implemented. We have checked the elemental compositions of all samples with a Thermo Scientific iCAP Q quadrupole ICP-MS after running the sample through the two-step column procedure. In rare cases, the only interference element present after two columns is chromium. According to the calibration curves of the AG50-X8 100–200 mesh cation-exchange resin in 0.7 mol L−1 HNO3 (see Fig. 1), the Cr peak has an overlap with the K peak. Other major rock-forming elements such as Ca, Fe, and Mg do not fall into the K cut (see Fig. 1).
We tested the matrix effect of Cr by doping different amounts of Cr (1, 2, 5, and 10% relative to K). The result in this study is shown in ESI Table S2† and Fig. 4, where we also plotted the matrix effect of Mg, Al, Ca, Ti, V, Cr and Rb reported in literature.56,58,60 We found that when matrix elements are <2% in solution, no resolvable matrix effects are observed in K isotopes. This matrix effect needs to be monitored and it is critical for any samples with relatively low K concentrations. If more than 2% matrix element is found in the final K solutions, additional column chemistry is needed to purify the K in the samples further.
![]() | ||
Fig. 4 Potassium isotopic compositions of standards doped with potential residual matrix elements (Mg, Al, Ca, Ti, V, Cr and Rb) measured by this study and in literature.54,58,60 Both “cold plasma” and “collision gas” methods produce no measurable matrix effects for samples containing less than 2% matrix elements. |
![]() | ||
Fig. 5 K isotopic compositions of reference material BHVO-2 measured in 90 days over 20 months from Jan. 2017 to Aug. 2018 (890 individual analyses). Each data point represents the average of 5–15 measurements of BHVO-2 during a 24 hour analytical session. Error bars represent the 95% confidence interval of repeated analyses during each 24 hour analysis session. The grand average of all 890 analyses is −0.480 ± 0.006‰ (95% C.I.) which agrees well with the Bulk Silicate Earth (BSE) value (−0.479 ± 0.027‰; 2SD) defined by basalts from three different tectonic settings.54 |
![]() | ||
Fig. 6 K isotopic compositions of all samples in this study. Error bars represent the 95% confidence interval of about 10 repeated analysis. Igneous rocks have a confined range of variation in K isotopes, while sedimentary rocks have a much larger variation. The shaded area is the Bulk Silicate Earth (BSE) value (−0.48 ± 0.03‰) defined by basalts from three different tectonic settings.54 |
The only two igneous rock samples that are different from the BSE value are also the two samples with the highest SiO2 contents (see Fig. 7): one rhyolite from Glass Mountain, Siskiyou County, California (RGM-1) and one obsidian from Clear Lake, Newberry Crater, Oregon (NIST 278). These two high-Si felsic samples have K isotopic compositions of −0.35 ± 0.04‰ and −0.36 ± 0.06‰, which are significantly heavier than the igneous average (−0.45 ± 0.07‰) reported in this study. As also shown in Fig. 6, the K isotopic compositions of felsic rocks do exhibit slightly heavier values than those of mafic rocks.
![]() | ||
Fig. 7 K isotopic compositions of all igneous rocks in this study versus their SiO2 abundances. The SiO2 data are from literature.74 The shaded area is the Bulk Silicate Earth (BSE) value (−0.48 ± 0.03‰) defined by basalts from three different tectonic settings.54 Although most igneous rocks are indistinguishable from this BSE value, two high-Si felsic igneous rocks (rhyolite and obsidian) are significantly enriched in heavy K isotopes. |
The two minerals (NIST 99a and FK-N) from the igneous rocks, sodium and potassium feldspars, have even larger K isotopic fractionations relative to the BSE value than bulk samples. The sodium feldspar from Spruce Pine pegmatite district of North Carolina (NIST 99a) has the heaviest K isotopic composition in this study. It is consistent with the observation by Morgan et al.58 that pegmatites (consisting of feldspars as the major K phases) show >1‰ K isotopic fractionation.
In summary, most igneous rocks (mafic, intermediate and felsic) have the same K isotopic composition, which agrees well with the BSE value previously defined from the average of basalts (mid-ocean ridge basalt, ocean island basalt, and continental flood basalt). However, we have also observed that two high-Si (degassing-influenced) felsic rocks (rhyolite and obsidian) and two K-rich minerals (sodium and potassium feldspars) are significantly enriched in heavy K isotopes. This variation observed between different rocks and minerals indicates that high-temperature K isotope fractionations are finally resolvable with the new high-precision K isotope analysis method.
The first issue is the different standards used by different laboratories: Merck KGaA Suprapur® 99.995% KNO3, NIST SRM 999b, and NIST SRM 3141a. As discussed above in Section 3.2, none of these standards are ideal for isotopic analysis because they are not certified isotopic standards; however, they are the only presently available choices. We measured both Suprapur and NIST 3141a standards in this study (see Table 2 and Fig. 6), and we found no resolvable difference in terms of K isotopes within the current typical analytical errors (∼±0.05‰). Regardless, more effort should be made in the future to coordinate between research laboratories so that the same isotopic standard of K is used by all groups. At this time, however, it is acceptable to use either Merck KGaA Suprapur® 99.995% KNO3 or the NIST SRM3141a, since they have no measurable difference in the K isotopic composition.
The second issue is that different laboratories use different methods to remove the Ar based interference; see Section 3.1 for a description of both the “cold plasma” and “collision gas” methods. It has not been rigorously tested whether these two methods would produce a systematic consistency in measuring 41K/39K ratios. It needs to be verified that the data produced using one method agrees well with data produced using the other when analyzing the same sample. We have measured K isotopes using both the “cold plasma” and “collision gas” methods.54 Here we analyzed five USGS standards (BCR-1, BHVO-2, AGV-1, GSP-1, and G-2) in this study, which are the only ones that have been reported by different groups.54–56,58,60 As shown in Fig. 8, our results measured with the “cold plasma” or “collision gas” method are indistinguishable from those data measured by different groups using different methods.54–56,58,60 The new data show that both methods, using either Isoprobe, Neptune Plus or Nu Plasma MC-ICP-MS, produce data that are comparable to each other.
![]() | ||
Fig. 8 K isotopic compositions of reference materials measured in this study using the “cold plasma” method and literature data using “cold plasma” and “collision gas” method.54–56,58,60 Error bars represent the 95% C.I. (Confidence Interval) for this study and literature data. The shaded area is the Bulk Silicate Earth (BSE) value (−0.48 ± 0.03‰) defined by basalts from three different tectonic settings.54 |
In summary, the high-precision K isotopic compositions measured with both “cold plasma” and “collision gas” methods relative to both Merck KGaA Suprapur® 99.995% KNO3, and NIST SRM 3141a in the literature can be compared directly. There is no systematic shift found between data from different laboratories.
We calculated the absolute abundances of K isotopes and the atomic weight of K in seawater and the Bulk Silicate Earth based on the absolute abundances of K isotopes measured in NIST-SRM985 and relative differences between NIST-SRM985 and the seawater and the Bulk Silicate Earth (see Table 5). Notably, the value of the atomic weight of K in Bulk Silicate Earth is revised from 38.0983 to 38.0982 a.m.u. (atomic mass unit). The newly calculated atomic weight of K agrees with the one proposed recently by Morgan et al.58 However, the new atomic fractions of K isotopes of the Bulk Silicate Earth are different between two groups due to (1) the slightly different (yet same within errors) estimates for the Bulk Silicate Earth value, and (2) different estimates of 40K abundance. Morgan et al.58 used the current IUPAC value for 40K abundance,70,72 while in this study we propose to use the more precise value from Naumenko et al.73 In addition, we also report the newly calculated absolute abundances of K isotopes and the atomic weight of K in the CI chondrite and the Bulk Silicate Moon (see Table 5). Therefore, we suggest IUPAC to revise their recommended values for absolute abundances of K isotopes and the atomic weight of K.
Measured 39K/41K | Measured 40K/41K | Measured 40K/39K | Measured δ41K SRM985a (‰) | 39K (atomic fraction) | 40K (atomic fraction) | 41K (atomic fraction) | Atomic weight | |
---|---|---|---|---|---|---|---|---|
a The data were converted from the original in the literature through the 0.259 per mil difference between NIST-SRM985 and seawater measured by ref. 58. | ||||||||
NIST-SRM985 | ||||||||
Garner et al.32 | 13.85662 | 0.0017343 | 0.932581073 | 0.000116722 | 0.067302205 | 39.0983 | ||
Naumenko et al.73 | 0.000125116 | |||||||
IUPAC70 | 0.932581 | 0.000117 | 0.067302 | 39.0983 | ||||
Garner et al.32 + Naumenko et al.73 | 0.932581112 | 0.000116681 | 0.067302207 | 39.0983 | ||||
Seawater58 | −0.259 | 0.932597383 | 0.000116667 | 0.067285950 | 39.0983 | |||
Bulk Silicate Earth 54 | −0.838 | 0.932633759 | 0.000116637 | 0.067249604 | 39.0982 | |||
CI chondrite55 | −0.893 | 0.932637214 | 0.000116634 | 0.067246152 | 39.0982 | |||
Bulk Silicate Moon55 | −0.397 | 0.932606052 | 0.000116660 | 0.067277288 | 39.0983 |
We applied this protocol to a wide variety of selected geological samples that have been well documented in the major elements, trace elements, and other isotopes. The results show that
(1) The igneous rocks have homogeneous K isotopic composition and the average K isotopic composition of igneous rocks in this study agrees with the Bulk Silicate Earth value previously defined;
(2) Two high-Si felsic samples such as rhyolite and obsidian and K-bearing minerals such as feldspars are enriched in heavy K isotopes, which indicate that high-temperature igneous processes could fractionate K isotopes;
(3) Sedimentary rocks have a much broader range of variation in K isotopic compositions compared to igneous rocks, which suggests significant isotopic fractionations during low-temperature geochemical processes;
(4) Although different groups use different standards and methods to analyze K isotopes in high precision, the inter-laboratory comparison in this study shows no significant nor systematic difference among data from different groups;
(5) The atomic weight of K and the atomic fractions of K isotopes for natural samples need to be updated to reflect the newly observed variation.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8ja00303c |
This journal is © The Royal Society of Chemistry 2019 |