Research ArticleGEOLOGY

Boutique neutrons advance 40Ar/39Ar geochronology

See allHide authors and affiliations

Science Advances  11 Sep 2019:
Vol. 5, no. 9, eaaw5526
DOI: 10.1126/sciadv.aaw5526

Abstract

We designed and tested a compact deuteron-deuteron fusion neutron generator for application to 40Ar/39Ar geochronology. The nearly monoenergetic neutrons produced for sample irradiation are anticipated to provide several advantages compared with conventional fission spectrum neutrons: Reduction of collateral nuclear reactions increases age accuracy and precision. Irradiation parameters within the neutron generator are more controllable compared with fission reactors. Confidence in the prediction of recoil energies is improved, and their likely reduction potentially broadens applicability of the dating method to fine-grained materials without vacuum encapsulation. Resolution of variation in the 39K(n,p)39Ar neutron capture cross section at 1.3 to 3.2 MeV and discovery of a strong resonance at ~2.4 MeV illuminate future pathways to improve the technique for 40Ar/39Ar dating.

INTRODUCTION

Precise geochronology is the basis for quantification of processes in Earth and planetary sciences and allied fields such as archeology. The 40Ar/39Ar technique is the most versatile dating technique available in terms of the applicable age range and diversity of geological environments, having been used to date materials from the formation of our Moon ca. 4 billion years ago (1) to historically recorded volcanic eruptions (2). It is widely used, e.g., for dating of volcanic tephra in sedimentary sequences recording human evolution and climate variability (3), the timing and causes of biotic evolution and mass extinctions (4), and determination of exhumation rates in mountain belts (5). The rock clock is based on the natural radioactive decay of 40K to 40Ar*. The technique requires neutron irradiation of the sample to transmute 39K via a neutron capture, proton emission (n,p) reaction to 39ArK, which is then a proxy for 40K and allows the critical daughter/parent isotope ratio to be measured by noble gas mass spectrometry. After correction for Ar isotopes arising from air contamination and collateral reactions from the irradiation, the measured 40Ar*/39ArK of a sample is converted to 40Ar*/40K through geological age standards that are co-irradiated with samples. Conventionally, 40Ar/39Ar samples are irradiated in nuclear fission reactors providing a wide spectrum of neutron energies resulting in two limitations: (i) Kinetic energy of the neutron impact and proton emission is partially transferred to the produced 39Ar, resulting in a recoil effect, its redistribution, and, potentially, 39Ar loss. “Recoil” loss leading to erroneously old ages is significant for samples with large surface per volume, i.e., fine-grained materials and materials with significant volume of nonretentive sites (6). The effect was resolved for grain dimensions smaller than ~50 μm and results in inaccuracies up to ~20% for materials with grain sizes around 5 μm (7). This places materials such as calcium-aluminum–rich inclusions in meteorites (the oldest materials in the solar system), clay minerals, or volcanic glass shards (of interest in crystal free tephra) out of limits for accurate conventional 40Ar/39Ar dating. (ii) Multiple reactions induced by the broad spectrum of the neutrons in a reactor core produce Ar isotopes from K, Ca, and Cl (8) and mass interfering 80Kr from Br (9), e.g., the reaction 42Ca(n,nα)36Ar. Here, the interfering reactions that create 36Ar are the most critical as 36Ar is used to determine the amount of atmospheric Ar in a sample and, accordingly, subtract the atmospheric 40Ar from the total 40Ar to determine the radiogenic component relevant for the age calculation. Because of an atmospheric 40Ar/36Ar ratio of 298.56 (10), uncertainty of the 36Ar measurement propagates with more than two orders of magnitude into the age. Corrections for all known interfering reactions are possible, but significantly add to the uncertainty in ages (11). Besides undesired reactions, the production of 37Ar via 40Ca(n,α)37Ar and 38Ar via 37Cl(n,γ,β)38Ar are useful tracers of Ca and Cl contents. They allow monitoring, e.g., the degassing composition of phases with Ca/K zonation or hydrous alteration through increased Cl contents. In the case of 36Ar and 37Ar produced from Ca, the intricacies of collateral reactions and recoil mingle: 37ArCa is used for the correction of 36ArCa, but their production involving alpha and proton emission, respectively, result in a different recoil loss and a flawed 36ArCa correction, typically making the ages erroneously young (12, 13).

Renne et al. (11) proposed to build a deuteron-deuteron (D-D) fusion neutron generator to produce quasi-monoenergetic neutrons of ~2.5 MeV [D(D,n)3He]. Compared with fission neutrons, they anticipated it to reduce 39ArK recoil ranges by a factor of 3 and exclude or minimize by energy threshold or small neutron cross section all undesired interference reactions while still allowing the production of 37Ar as a valuable tracer for Ca (11). The reduction of 38ArCl production as a tracer for Cl, which is predicted to be about three orders of magnitude smaller than in fission reactors, relative to 39Ar production from K (14), may be a disadvantage for some applications. Such a proposed apparatus, dubbed the high flux neutron generator (HFNG), has been constructed at the University of California (UC) Berkeley, commissioned in 2013, and has undergone continuous upgrades subsequently. Its design and technical capabilities were recently detailed in (15). Here, we report (i) the first dating and experiments to quantify (ii) the reduction of recoil, (iii) interference reactions, and (iv) the 39K(n,p)39Ar cross section at 1.3 to 3.2 MeV.

RESULTS

(i) To test our ability to date “unknowns,” we co-irradiated the geological age standard Alder Creek sanidine [ACs; 1.1848 ± 0.0006 million years (Ma); uncertainties are 1σ throughout] (16), sanidine from the Campanian Ignimbrite [MDP-1; 39.85 ± 0.14 thousand years (ka)] (17), and sanidine from the 79 CE eruption of Vesuvius documented by Pliny the Younger (1.939 ka; AM-1). (ii) For a first-order quantification of recoil effects, we produced grain size fractions of MDP-1 ranging down to <15 μm by crushing and sieving pristine 180- to 300-μm crystals. (iii) To investigate production of collateral reactions on Ca and K, we irradiated CaF2 and a glass with 11.3 weight % (wt %) K. (iv) In ancillary experiments, we determined the cross section of 39K(n,p)39Ar as a function of neutron energy. We used the capabilities of the HFNG to provide different neutron energies at different angles as a result of the two-body fusion reaction [~2.76 MeV forwardmost to ~2.2 MeV backward; (15), with respect to the incident deuteron beam]. We expanded on this energy range by using the LICORNE (Lithium Inverse Cinematic ORsay NEutron source) apparatus at the Accélérateur Linéaire et Tandem à Orsay (ALTO) facility of the Institut de Physique Nucléaire d'Orsay (IPN) that produces 0.5- to 4-MeV neutrons (18), based on the reaction p(7Li,n)7Be.

Dating, recoil, and interference reactions

We irradiated 2 to 5 mg of each sample in wells arranged in a triangular pattern over an area of 12 mm by 10 mm with 12 aliquots of a 180- to 300-μm control fraction of MDP-1, bracketing and interleaving the other samples to monitor for 39Ar production rate variation (Fig. 1). F values (40Ar*/39ArK) of the dispersed 180- to 300-μm fractions indicate little variation in 39ArK production rate in the central sample array (Fig. 1). Only the F value of the sample in marginal hole 17 differs from the other holes outside uncertainty at the 2σ level. The ratios of F values (R values) (19) of the different samples are an independent measure of relative age; they are given in table S2 and are concordant with reported literature values. Using MDP-1 as the geological age standard, we calculate inverse variance–weighted mean ages of 1.183 ± 0.046 Ma for ACs and 2.3 ± 1.1 ka for AM-1.

Fig. 1 Map of the sample holder used for the main HFNG experiment showing determined F values.

Dark gray fields are the MDP-1 control samples to monitor production rate variations. Transparent holes are not-available (NA) analyses (NA). ps indicates peak suppression; c, accidental contamination. F values are in sample-specific colors for ease of orientation. Figure S1 provides a photograph of the actual sample holder.

To test for recoil loss of 39Ar through grain surfaces, we normalized F values of the fine-grained fractions of MDP-1 with the F values of the surrounding control fraction aliquots (Figs. 1 and 2). Deviation of the normalized F value from unity is equivalent to the age bias caused by 39Ar recoil loss. A similar approach was used by Jourdan et al. (12), who analyzed different grain sizes of sanidine and a sanidine and plagioclase mixture, and by Paine et al. (7), who analyzed individual platy grains of biotite of different thicknesses. For comparability, we calculated surface area per volume (SAV) ratios of individual grains and fractions, assuming cube and cylinder shapes for sanidine and biotite, respectively. Jourdan et al. (12) found a positive age bias for sanidine fractions <40 μm (SAV larger than ~0.1 μm−1), which they interpreted as being due to 39Ar recoil loss, and a negative age bias for the sanidine and plagioclase mixed fractions <40 μm, which they interpreted to result from undercorrection of 36ArCa because the tracer 37ArCa experienced relatively more recoil loss (Fig. 2A). Paine et al. (7) found a positive age bias for grains with SAV >0.08 μm−1 interpreted as 39ArK recoil loss. While uncertainties are comparably higher in our experiment, all our grain size fractions yielded F values of unity within uncertainty, indicating irresolvable recoil effects for ~2.75-MeV neutrons in this experiment.

Fig. 2 40Ar/39Ar age bias as a function of grain geometry for D-D fusion neutron or fission neutron–irradiated minerals.

Positive age bias of high SAV biotite grains irradiated with fission neutrons is interpreted to represent recoil-related 39Ar loss [data points are partially shown as means and envelope only; (7)]. Negative age bias of a sanidine-plagioclase mixture is interpreted to reflect undercorrection of 36ArCa (12). Biotite grains (7) were irradiated separately, while sanidine fractions [this study and (12)] were irradiated with multiple grains in contact, allowing potential reimplantation and decreasing apparent recoil.

To quantify the production of Ar isotopes from Ca and K, we analyzed three samples of CaF2 and K-rich glass each. As desired, the production of the short-lived 37Ar (T1/2 = 34.95 ± 0.04 days) (20) is well quantifiable: The inverse variance–weighted mean production rate ratio of 37ArCa/39ArK per Ca/K weight ratio is 0.249 ± 0.011, compared with a production rate ratio in fission reactors around 0.5. This confirms the usefulness of 37ArCa as a tracer of Ca content using D-D fission neutrons for irradiation. From CaF2 analyses, we determined inverse variance–weighted 40Ar/36Ar, 38Ar/36Ar, and 39Ar/37Ar ratios of 299.3 ± 2.3, 0.189 ± 0.027, and 0.005 ± 0.012, which are indistinguishable from atmospheric values of 298.56 ± 0.31, 0.1885 ± 0.0003, and 0, respectively (10). This indicates no resolvable production of 40Ar, 39Ar, 38Ar, and 36Ar from Ca. Similarly, no quantifiable production of 40Ar from K via 40K(n,p)40Ar could be resolved from the K-rich glass that provides an inverse variance–weighted mean 40ArK/39ArK production rate ratio of 0.8 ± 6.4.

39K(n,p)39Ar cross section

We co-irradiated K-bearing targets (KBr- or K-rich glass) and metallic neutron fluence monitors (In or Ni foil) with neutrons of different energies. Following irradiation, the K-bearing target was analyzed for its 39Ar content and the metallic neutron fluence monitors for their γ activity. Cross sections of the reaction 39K(n,p)39Ar were calculated against well-known cross sections of neutron-absorption reactions on In and Ni.

Our new cross section data resolve structure in the energy range of 1.3 to 3.2 MeV that has not been previously described (Fig. 3 and Table 1). The cross section of the 39K(n,p)39Ar reaction shows nonstatistical behavior with variability exceeding a factor of 5 over <0.2 MeV. The experiments in the HFNG with an energy resolution around 0.02 MeV resolve this variability in detail in the range of 2.4 to 2.8 MeV. The LICORNE data with a generally coarser energy resolution suggest a nonsmooth function down to energies ~1.5 MeV: The cross section determined for ~1.46 MeV with a narrower neutron energy spectrum is higher than the cross section with a wider neutron energy spectrum averaging ~1.54 MeV (Fig. 3 and Table 1).

Fig. 3 Experimental cross section of the reaction 39K(n,p)39Ar as a function of neutron energy and their comparison with the most recent ENDF evaluation.

Based on this study, (34) and (14). Error bars in neutron energy represent 67% limits of the neutron energy spectrum; error bars in the cross section represent analytical uncertainty. The strong variability in the cross section offers the possibility to magnify 39Ar production rates by choosing the right neutron energy. ENDF, Evaluated Nuclear Data File.

Table 1 Summary of irradiations used to determine the cross section (σ) of 39K(n,p)39Ar in the energy range of 1.2 to 3.2 MeV.

View this table:

DISCUSSION

Dating

The aliquots of the control fraction of sample MDP-1 in 10 holes display little spatial variation in F value (39ArK production rate) outside uncertainty with one exception (Fig. 1). The F value of hole 17 exceeds the average by 25% and outside 2σ uncertainty. We can only speculate on the reasons: Most likely, a grain of the neighboring K-rich glass contaminated this sample unnoticed. Hole 17 aside, our experiment shows that the setup used with a multi-aperture extraction plate can create 39ArK production rates that vary in the single-digit percent range over >4 cm2. This is on the higher side of gradients commonly observed in fission reactors (21) but sufficient for high-precision 40Ar/39Ar geochronology.

The determined R values and calculated absolute ages of ACs, AM-1, and MDP-1 are concordant with literature values (table S2) and provide the proof of concept for dating samples using the HFNG. The larger age uncertainties of our one to four aliquots as compared with previous work from experiments with several dozen aliquots are a statistical effect and not representative of the ultimately higher precision and accuracy achievable with the HFNG. This ultimately higher accuracy of the HFNG is anticipated on the basis of avoiding or reducing the need for correction of interfering reactions, with the most significant being 42Ca(n,nα)36Ar for high Ca/K samples (11).

Recoil

The potential effect of recoil on 40Ar/39Ar geochronology was first recognized by Turner and Cadogan (22). Subsequent studies contributed to the understanding and quantification of Ar recoil by either theoretical estimation or experimental studies of different grain sizes of one mineral separate or vacuum encapsulation of mineral grains. Onstott et al. (23) calculated a theoretical estimate of 39Ar recoil ranges by calculating 39Ar recoil energies based on a fission flux spectrum and a smoothed 39Ar(n,p)39Ar cross section function and by simulating the transport range of 39Ar ions with the resulting energies in silicate glass. A caveat of such a theoretical estimate of recoil energies is its limited transferability because the neutron flux spectrum is an important input to the calculation, but is specific to reactor and irradiation position, varies over time, and is not a smooth function, because cross section functions of moderating materials are not necessarily smooth [cf. Figure 1 of (21) for a simulated neutron spectrum for a 40Ar/39Ar irradiation position]. While neutron moderation will generally decrease recoil energies, it is difficult to accurately predict how the complex structure of the 39K(n,p)39Ar cross section function affects the resulting average recoil energies when it is convolved with a complex neutron energy spectrum. In any case, we estimate 39Ar recoil reduction with D-D neutrons by comparing ~2.7 MeV with the average neutron energy in fission reactors performing the 39K(n,p)39Ar reaction, which is the average of the convolution of the neutron energy flux spectrum and the 39K(n,p)39Ar cross section. This is relevant because the recoil of 39Ar depends primarily on the energy of the neutron and, only to a much lesser extent, on the emission of the proton (22). We convolved the smoothed Evaluated Nuclear Data File (ENDF) cross section (14) with a Watt thermal fission flux spectrum and derived an average neutron energy of 4.1 MeV performing 39K(n,p)39Ar. This suggests a reduction of average recoil energies by a factor of 0.65 when 2.7-MeV neutrons are used. The estimate of a reduction by a factor 0.3 by Renne et al. (11) was based on the erroneous assumption that calculations for a perfectly inelastic collision provide the maximum recoil energy. Our approach cannot capture the complexity in cross section structure and flux spectrum discussed above, nor can any approach until details of these features are resolved. Notably, average recoil energies are compared, which may oversimplify the complexity of loss and redistribution when recoil energies and, thus, recoil distances range over an order of magnitude (23). These caveats highlight the advantages of a quasi-monoenergetic neutron source that allows an accurate theoretical prediction of recoil energy and allows well-controlled experiments to test it.

The experimental approach to recoil quantification chosen in this study is to assess 40Ar*/39ArK differences as a function of the grain SAV (7, 12). This approach is regarded to mostly capture 39Ar recoil loss through grain surfaces but may contain effects of loss from low-retentivity sites as a function of irradiation temperature or short circuit diffusion pathways that are grain size dependent (12). These processes are invoked by Jourdan et al. (12) to explain the difference between their sanidine data and the biotite data of Paine et al. (7) that indicate about a factor 5 more apparent recoil per SAV (Fig. 2). Another complicating factor is the reimplantation of recoiled 39Ar into neighboring grains that makes the experiments difficult to compare: The biotite grains of Paine et al. (7) were separated by millimeter-thick aluminum, while in our experiment and the experiment of Jourdan et al. (12), multiple sanidine grains were irradiated in contact.

A more direct experimental approach is the vacuum encapsulation of samples during irradiation, in which recoiled 39Ar is captured and can be analyzed separately. This method showed that clay minerals may lose ~30%; hornblende and biotite around 0.4%, but up to 1.2%; and sanidine 0.04 to 0.18% of the produced 39Ar (6, 24). Considering the grain geometries, these amounts exceed the expected recoil loss through surfaces and indicate mineral-dependent loss from nonretentive sites that, in many phases, exceeds loss from surface ejection (6, 24).

Our experiment indicates that fine multigrain separates down to <15 μm can be irradiated without currently resolvable apparent 39Ar recoil loss in the HFNG. It remains unclear how much reimplantation into neighboring grains contributes to this outcome. Our experiment, which was tailored around the current relatively low neutron flux capability of the HFNG, needs to be considered as a starting point for further investigation. Once the necessary neutron fluences can be achieved by the HFNG in acceptable irradiation time, experiments investigating Ar recoil from single encapsulated grains need to be carried out to quantify recoil loss through nonretentive sites and quantify possible reimplantation. These experiments can serve as a benchmark for simulation of recoil distances because the spectral character complicating the recoil energy calculation for fission spectrum neutrons becomes irrelevant.

Moreover, the HFNG’s anisotropic neutron flux—sourcing from a disc-shaped region next to the samples—can allow previously inconceivable experiments such as the semidirectional neutron irradiation of grains with planar inhomogeneities (e.g., exsolutions) or advancement of implantation experiments [e.g., (25)] to determine oriented recoil distances. These could, for example, quantify phase-dependent recoil distance, test the effect of 39Ar ion channeling, and improve our understanding of the degassing behavior of mixed phases.

Collateral reactions

Our analyses support the anticipated reduction of unwanted interference reactions from Ca and K while easily quantifiable amounts of 37ArCa are produced, thus retaining its use as a tracer. However, at the current production levels of the HFNG, even the production rate ratios known from fission reactors that are expected to be higher could not be resolved analytically. Further experiments with higher neutron flux are required to enable more precise quantification of the reduction with reasonable irradiation times. Rendering the correction of interfering reactions obsolete or reducing them does not only reduce uncertainty in ages but also ameliorate the differential recoil loss of 36ArCa and 37ArCa, resulting in erroneous age corrections (12). The largely advantageous results of eliminating thermal neutrons in the HFNG come at the cost of losing the sometimes useful capability to produce noble gas isotopes from halogens, e.g., via 37Cl(n,γ,β)38Ar, that allows tracing aqueous alteration or fluid and melt inclusions (26). To overcome this drawback, a potential moderation of neutrons by using a 20 cm thick polyethylene sarcophagus encasing the HFNG is currently under testing. Initial results obtained by enclosing the instrument except for its bottom indicate a fast/thermal neutron flux ratio around 170, which would be sufficient for an 39ArK/38ArCl production rate of 0.004. To be useful for tracing Cl, an 39ArK/38ArCl production rate >1 would require a fast/thermal flux ratio of 0.6; however, other advantages of the HFNG would be lost. In principle, the instrument configuration could be tailored to whether or not the desirability of measuring 38ArCl outweighs the benefits of reducing the thermal neutron flux.

39K(n,p)39Ar cross section

Our new results resolve a complex structure in the 39K(n,p)39Ar cross section at 2.4 to 2.8 MeV and indicate additional complexity around 1.5 MeV. This is similar to a recently found resolved resonance region for the reaction 35Cl(n,p)35S in the same energy range (27) and is not entirely expected, given the low-level density in nuclei near the N,Z = 20 shell closure. This discovery informs future instrumental design to include 40Ar/39Ar sample irradiation capabilities at higher angle and, thus, lower neutron energy: A minimally threefold higher cross section at ~2.4 MeV (~112° at 100-keV D acceleration) results in higher production rates even if the neutron yield is at only 55% of the forward direction.

The resolved structure also has implications for fission reactors that require further quantification: The folding of a nonsmooth cross section function with a nonsmooth neutron spectrum that varies between and within reactors may explain variations observed in production rates that are commonly considered to be a result of neutron fluence gradients alone. We conclude that further experiments would be fruitful for elaborating the 39K(n,p)39Ar cross section in further detail.

Additional benefits

Compared to conventional fission neutron irradiation, the HFNG offers several advantages beyond the merits discussed above: (i) It allows for irradiation at about room temperature compared to temperatures of, for example, ~270°C determined for using Al disks as sample holders in CLICIT (cadmium-lined in-core irradiation tube), the most utilized irradiation facility for 40Ar/39Ar geochronology at the Oregon State University research reactor. This provides an opportunity to date thermally unstable phases and those prone to Ar diffusion at low temperatures. (ii) The HFNG provides reproducible irradiation conditions. (iii) It reduces production of radioactive waste and exposure of personnel to radiological hazards. (iv) As a small-scale facility, it provides independence from a shrinking pool of research reactors worldwide. (v) The cost of irradiation with the HFNG is expected to be lower compared to a fission reactor because costly safety and environmental aspects are absent or less complex. In addition, application of D-D neutrons to 21Ne cosmic ray exposure dating offers promise for illuminating planetary surface processes (28).

CONCLUSIONS

We provide the proof of concept for using a D-D fusion neutron generator for an 40Ar/39Ar sample irradiation. We successfully reproduced the ages of three well-characterized samples ranging from 1.9 ka to 1.2 Ma. We find no grain size–dependent age bias for grain fractions down to <15 μm of a 40-ka sanidine from the Campanian Ignimbrite, supporting the anticipated reduction of 39Ar recoil. This promising result requires further experimental quantification at increased neutron flux. A first quantification of interference reactions also supports the anticipated reduction of unwanted interference reactions, while the tracer reaction 40Ca(n,α)37Ar remains useful. In ancillary experiments, we resolve previously unrecognized structure in the 39K(n,p)39Ar cross section. Our results motivate further development of D-D fusion neutron generators for 40Ar/39Ar geochronology, with emphasis on increasing the 39Ar production rate. This can be achieved by increasing the neutron flux and/or optimizing the design to exploit the newly found structure in the 39K(n,p)39Ar cross section.

MATERIALS AND METHODS

Irradiation and analysis of geological age standards, K-rich glass, and CaF2

The geological age standards (MDP-1, AM-1, and ACs), CaF2, and K-rich glass were loaded in a triangular drill pattern in an aluminum sheet. This sheet also held the samples of irradiation 3 described in the “Irradiation targets” section (Fig. 4). Figure 1 and fig. S1 show the sample array. After loading, the wells in the aluminum sheet were closed with aluminum foil and then sealed with Kapton tape. MDP-1 of the control fraction (180 to 300 μm; Fig. 1) bracketed AM-1, ACs, and MDP-1 of other grain size fractions to allow normalization of production rate variations over the sample array.

Fig. 4 Irradiation setup of neutron cross section experiments.

(A) Geometry of samples in the HFNG with respect to the neutron source area where the D+ beamlets impinge on the Ti layer hosting the interstitially implanted D. The neutron energy spectrum each target was exposed to is given in Fig. 5). (B) Setup of the LICORNE experiment and irradiation geometry. Table 1 and table S3 list the irradiation details.

We used one ion source of the HFNG using a multi-aperture extraction plate, directing 31 beamlets of D+ ions at a titanium surface 8 mm above the samples. Impinging with an energy of 100 keV, they produced neutrons through fusion with already embedded D (15). The setup resulted in a neutron source region with a diameter of 19 mm centered on the samples. The irradiation lasted 210.2 hours at an ~2.75-MeV neutron flux of ~2 × 107 n cm−2 s−1. Quantified by the dimensionless irradiation parameter J used to characterize fission reactors (29), this translates to a J/hour of 5.2 × 10−10, about six orders of magnitude below common fission neutron irradiation facilities. Following irradiation, the samples were transferred into a CO2 laser extraction system and analyzed using mass spectrometry. We calculated inverse variance–weighted mean F values (40Ar*/39ArK) for each geological sample.

After irradiation, the front of the sample holder was opened hole by hole, and the material was moved—using a wetted paintbrush—onto weighing paper and, after drying and weighing (2 to 5 mg), was loaded into wells in a copper disk. On the basis of the weight and expected Ar signals, some samples were split into multiple aliquots (table S1). During the loading procedure, two samples were accidentally contaminated and discarded (Fig. 1).

The Ar was released from the samples by incremental heating using a CO2 laser with a beam-shaping lense providing a flat energy profile of adjustable diameter. The released gas was purified for 300 s using a cold trap at −130°C to remove H2O and a GP-50 SAES getter to remove reactive gases. After purification, the gas was allowed to expand into an Nu Instruments Noblesse 5 collector sector field noble gas mass spectrometer. Time series of ion beam intensities of Ar isotopes 36 through 40 were measured over 700 s and extrapolated to determine time-zero intercepts. We corrected for analysis blanks and mass spectrometer ion source mass discrimination. Radiogenic 40Ar* was determined by subtraction of the atmospheric component by subtracting 36Ar multiplied by 298.56 (30). For the geological age standards, F values, i.e., 40Ar*/39ArK, were determined for each step. 37Ar decay (T1/2 = 34.95 ± 0.08 days) (20) in the analyzed CaF2 was corrected for, accounting for decay during irradiation, irradiation breaks, and the time lag between irradiation and analysis. Uncertainties were linearly propagated. The largest source of uncertainty in the recoil experiment was the correction for air 40Ar that amounted to ~90% of total 40Ar in MDP-1. Analysis of five samples showed initially rising and strongly curved Ar isotope ion beam intensities. This observation is commonly interpreted to be an effect of contamination of the released Ar with other gases that are either inert or exceed the capabilities of the gas purification system, resulting in the initial suppression of Ar ionization (9). No meaningful time-zero intercepts could be determined for these samples, and they were discarded.

From the results of all heating steps of a sample, a weighted mean F value was determined (table S1). The weights were determined by inverse variance. To determine the R values (19) of MDP-1, ACs, and AM-1, we calculated the inverse variance–weighted mean F values of the MDP-1 samples of the control fraction surrounding the AC and AM-1 samples in the sample holder. Table S2 lists the R values (19) of the samples ACs, MDP-1, and AM-1 calculated from literature data given in the main text and compares them with the R values determined in this study (based on Fig. 1). They are indistinguishable within uncertainty.

Determination of 39K(n,p)39Ar cross section

Overview. To determine the cross section of 39K(n,p)39Ar, we irradiated a K-bearing material (KBr- or K-rich glass) along with a metallic neutron fluence monitor (In or Ni foil) with neutrons. We used two different reactions as a neutron source for energies ranging from 1.3 to 3.2 MeV: D-D fusion and p(7Li,n)7Be. Following irradiation, the K-bearing material was analyzed for its 39Ar content, and the neutron fluence monitor was analyzed for its γ activity to calculate the neutron fluence. Cross sections of 39K(n,p)39Ar were then calculated against the well-known cross sections of neutron capture reactions on In and Ni.

Irradiation targets. We used KBr crystals and a K-rich glass (11.3 wt % K) as targets for the 39K(n,p)39Ar reaction. The merits of KBr are its comparably weak hygroscopic properties, stoichiometric and thus weighable and high K content, and commercial availability of high-purity large single crystals. The Ar retentiveness of KBr was ensured in an ancillary experiment by irradiation of KBr in a fission reactor, followed by stepwise degassing (9). For irradiation 1 (Table 1) in the HFNG, we used a cuboid target (10 mm by 10 mm by 1 mm) that was manufactured from a large single crystal, taking advantage of its perfect cleavage. The targets for LICORNE were unpolished cut disks (diameter, 12 mm; thickness, 4 mm) that were purchased through ISP Optics Corp. The KBr targets contained significant quantities of atmospheric Ar (~3 × 10−12 mol 40Ar/g). To remove this atmospheric Ar before the irradiation, all targets were degassed at 450°C in ultrahigh vacuum (UHV; 2.7 × 10−6 Pa) for 24 hours. The targets were then removed from the furnace and exposed to atmosphere. One was reinserted and analyzed for its Ar content to ensure complete degassing. Insignificant atmospheric Ar was detected. The K-rich glass was acquired through Corning glass, broken and sieved. It was degassed in UHV to remove atmospheric Ar. We ensured homogeneity of the produced glass shards by electron microprobe analysis.

Neutron sources.
High flux neutron generator

The HFNG is a prototype D-D fusion neutron generator developed at UC Berkeley’s Department of Nuclear Engineering in collaboration with the Berkeley Geochronology Center. Deuterium was ionized in the plasma source by radio frequency excitation. It was then extracted through hole(s) in an extraction plate: Irradiation 1 used a single extraction hole with a diameter of 4 mm, while irradiation 3 used 31 holes with a diameter of 1 mm, arranged in a triangular pattern [cf. (15) for details on extraction plates]. The resulting beamlets were accelerated by a high-voltage differential potential (here 100 kV) and implanted in a Ti target. D+ fuses with already implanted D to form 3He with emission of a neutron. Technical details were recently reported by Ayllon et al. (15). For this experiment, we used its capabilities to provide different neutron energies with comparably high fluxes as a function of angle. The primary irradiation target (see the “Irradiation and analysis of geological age standards, K-rich glass, and CaF2” section) was positioned 8 mm behind the fusion reaction interface and irradiated by neutrons emitted in the 0°, forward angle (laboratory frame) that have energies around 2.7 MeV (Fig. 4A). Secondary targets were positioned at various angles up to 113° (Fig. 4A) for which the neutrons range downward in energy to ~2.4 MeV.

The neutron flux spectra were determined using an analytical model for the neutron source combined with a Monte Carlo neutron transport model. This approach was developed and benchmarked by Ayllon et al. (15). The analytical model of the neutron source was based on the well-characterized kinematics of the D(D,n)3He fusion reaction, simulation of beam transport physics, and target interactions. It takes into account the effects of acceleration voltage, beam current, ion beam optics, and deuterium diffusion within the target. The neutron flux spectra of individual samples were calculated using a Monte Carlo N-particle (MCNP) transport code using a discretized form of the source function as an input to the model. Figure 5A shows the neutron energy spectra of the individual samples.

Fig. 5 Simulated neutron energy spectra of cross section experiments.

(A) Neutron energy spectra of the six HFNG irradiation targets calculated using the MCNP. The spectra are a function of the radial distribution of the neutron beam, the energy-angle correlation of the neutron source, the intensity-angle correlation, and the solid angle. Table 1 lists the irradiation details. (B) Simulated neutron energy spectra of the four LICORNE-irradiated samples calculated. The spectra are a function of hydrogen cell target geometry, 7Li energy, and solid angle covered by the sample. Table 1 and table S3 list the irradiation details.

LICORNE

The LICORNE was operated at the ALTO facility of the IPN (18). The tandem particle accelerator produces an intense 7Li beam (typically 100 to 200 nA) that is directed through a Ta foil (2.79 μm) into a 2 cm long hydrogen gas cell operated at a pressure of 0.15 MPa and a flow rate of 0.5 cm3/s (31). There, the reaction p(7Li,n)7Be creates a cone of neutrons in the forward direction toward the sample. The energy of the neutrons and the maximum opening angle of the cone depend on the incident energy of the 7Li. We used incident 7Li energies of 15.8 to 17.5 MeV (measured before the Ta foil) to create neutrons of up to ~3.2 MeV. Ancillary to the main neutron energy, a satellite peak was created at a lower intensity and a lower energy of 0.7 to 0.9 MeV (18). It arises from the neutrons emitted “backward” in the center of the mass frame that are still moving forward in the laboratory frame. The two peaks were merged at a 7Li energy of 15.8 MeV and started to diverge at increasing energies. Narrowing the solid angle covered by the sample, i.e., the distance, was the main variable to narrow the full width half maximum of the main neutron spectrum (table S3). The limitations are the resulting lower neutron flux and the necessity to produce a well-quantifiable amount of 39Ar. To optimize separation of primary and satellite peak and neutron flux, we placed the samples at varying distances of 1 to 5 cm behind the hydrogen gas cell (Fig. 4B and table S3). We tested for the influence of the satellite peak on the total 39Ar activation by folding the neutron spectra with the latest ENDF 39K(n,p)39Ar cross section (14) and found it to be in or below the per mille range.

Direct measurements of the neutron spectrum close to the LICORNE inverse kinematics source are difficult because of the complex way in which the neutron spectrum and flux vary in three-dimensional space. The exact energy spectrum at a particular position, thus, needs to be calculated via Monte Carlo simulations, which must be validated by a direct measurement of the neutron spectrum at 0° using the time-of-flight (TOF) technique. We used a code based on the Geant4 Monte Carlo package that was developed for LICORNE to calculate the neutron spectrum received by each sample as a function of 7Li energy and distance from the gas cell. The model has been validated in previous experiments (18) and verified by repeated TOF measurements using an EDEN scintillation detector module (32) at 3-m distance from the hydrogen gas cell the day before starting the reported irradiation experiments. Pulsing and bunching the 7Li beam at 400-ns intervals allowed measurement of the TOF of the neutrons relative to the gamma peak [repetition of experiment described in (18)].

Neutron fluence determination. The neutron fluence received by the irradiation targets was determined using co-irradiated In or Ni foils, making use of the reactions 115In(n,n′)115mIn and 58Ni(n,p)58Co as neutron fluence dosimeters. The foils covered the same solid angle as the target. After the end of irradiation, the 336-keV gammas from the decay of 115mIn and 811-keV gammas from the decay of 58Co in the foil were measured using an HPGe (high-purity germanium) detector at ~10-cm sample detector separation. The detector was calibrated using 152Eu and 133Ba sources certified to 1.5% accuracy. Decay during irradiation and delay between irradiation and during the measurement were corrected for.

We calculated the 115In(n,n’)115mIn and 58Ni(n,p)58Co cross sections for each sample by folding the simulated neutron spectrum (Fig. 5) with interpolated cross section functions (14). The derived cross sections are given in Table 1.

Both LICORNE and HFNG exhibit fluctuations in neutron flux over the course of a multihour experiment. In LICORNE, these resulted primarily from inhomogeneity of the 7Li source, requiring retuning of the ion source every few hours. In the HFNG, the loading of the deuterium in the Ti getter on the surface of the cathode results in a ramping up of the neutron flux over the first ~1.5 hours of the experiment, while the neutron flux can also vary throughout a run when ion source parameters are tuned to maximize efficiency. The 4.5-hour half-life of the 115mIn made monitoring of these fluctuations a necessity: The weight of each time increment of the irradiation experiment on the final activity of the monitor material is based on the half-life of the activation product and irradiation time. As a simple example, for a 4.5-hour irradiation, the contribution of the first time increment on the final 115mIn activity will be half of the contribution of the last increment because 50% of the isotopes produced during the first increment of the irradiation have already decayed by the end of the irradiation. For the LICORNE irradiation, we used an EDEN scintillation detector module (32), and for the HFNG irradiation, a Bonner sphere to monitor relative neutron flux fluctuations. Equation 1 represents a correction factor R, where A is the arithmetic mean of the measure of neutron flux (e.g., counts), λ is the decay constant of the produced isotope, tn is the irradiation time, and xi is a measure of neutron flux at any time increment.R=Ai=0neλ(tnti)i=0neλ(tnti)χi(1)

As an example, fig. S2 shows the recorded time series of the Bonners sphere readback of irradiation 1 (Table 1) in the HFNG and the weight each time increment has in the calculation of the fluence from the In foil. A few-minute-long interruption toward the end of the irradiation evens out the initial phase where the flux is ramping up, resulting in a correction factor R of 0.99. Reported fluences in Table 1 include the correction factors. No such correction was necessary for the Ni foil where the half-life of 71 days was long compared to the irradiation length.

Analysis of 39Ar contents of KBr- and K-rich glass. To release the 39Ar for mass spectrometry, the KBr was fused in a resistance furnace and the K-rich glass was fused using a CO2 laser. The released gas was purified and analyzed following procedures described in (16) for the Noblesse noble gas mass spectrometer. The sensitivity of the instrument was determined by analysis of the known radiogenic 40Ar content of weighed aliquots of standards GA1550 and HD-B1 (33) before, during, and after analysis of the unknown samples. The sensitivity did not vary over the course of the measurements, and a mean value was determined. Expansion volumes were used to reduce Ar concentrations for the standards, but not for the unknown samples with smaller Ar contents. The relative volumes of furnace, laser cell, and expansion volumes were calibrated using repeated expansion and analysis of small aliquots of atmosphere. The difference in sensitivity between KBr- and K-rich glass analyses (Table 1) is related to the addition of the furnace volume for the KBr analyses. Uncertainties related to the radiogenic 40Ar content of the standards (~0.8%), their weighing (~0.1%), volume calibration (~0.5%), and analysis (~0.1%) contribute to uncertainty of the sensitivity.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/9/eaaw5526/DC1

Fig. S1. Photograph of the sample holder after loading all samples but the <15-μm fraction of MDP-1.

Fig. S2. Impact and correction of neutron flux fluctuation biasing the determined average neutron fluence.

Table S1. Results of noble gas mass spectrometric analysis of geological age standards, K-rich glass and CaF2.

Table S2. Comparison of R values based on literature data cited in the main text and this study.

Table S3. Irradiation details of LICORNE irradiations.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We acknowledge use of the ALTO facility, technical assistance by L. Morgan, and facilities support from the Ann and Gordon Getty Foundation. We thank C. Hall, A.A.P. Koppers, and two anonymous reviewers for the constructive criticism and K. Hodges for editorial handling. Funding: This research was funded by NSF grant EAR-0960138 (to P.R.R.), Deutsche Forschungsgemeinschaft (DFG; German Research Foundation) RU2065/1-1 (to D.R.), and U.S. DOE LBNL contract no. DE-AC02-05CH11231 (to L.A.B.). Author contributions: D.R., P.R.R., and T.A.B. designed the experiments. D.R., P.R.R., K.v.B., and L.A.B. acquired the funding. D.R., J.M., L.Q., M.A., J.B., J.J., S.-A.C., W.L.H., M.W., A.M., C.J., G.W., P.A.A., and H.M. operated the neutron sources. J.M. and L.Q. simulated the neutron energy spectra. D.R., J.M., and L.Q. performed the gamma spectrometry analysis. D.R. performed the noble gas analysis and evaluated the analytical data. D.R., J.M., and L.Q. developed the methodology and software. K.v.B., J.B., J.W., and M.L. managed the neutron sources. P.R.R., K.v.B., J.W., and M.L. oversaw the experiments. D.R. and J.M. visualized the results. D.R. wrote the original draft. All authors reviewed the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article