1 Introduction

When four hydrogen nuclei (protons) fuse to form one helium nucleus in the solar core, two of them convert to neutrons, releasing two neutrinos in the process. One of the definitive tests of this so-called Standard Solar Model is to measure the flux of those neutrinos. In order to detect these elusive particles, physicists have devised a number of experiments that broadly fall into two categories.

One group of experiments (Sudbury Neutrino Experiment, Super-Kamiokande, IceCube, Borexino) measures the light that is emitted when neutrinos scatter off electrons in water or a scintillation fluid. A second group (Homestake, Gallex, Sage) measures the radiation produced by neutrino reaction products such as Ar, Ge, and B (Bahcall et al., 1996). Because neutrino interactions are so rare, most of these experiments are massive in size and cost, with only one notable exception.

In 1976, Melvin Freedman proposed that the reaction ²⁰⁵Tl (ν,e^-) ²⁰⁵Pb could form the basis of a natural neutrino detector with the following advantages over alternative experimental designs (Freedman et al., 1976):

1. Thanks to the relatively large nuclear cross section of the reaction and the 16 Myr half life of ²⁰⁵Pb, sufficient amounts of the reaction product can accumulate over geologic time to be measured by (Shottky) mass spectrometry on a sample of just a few kilograms of thallium-bearing minerals such as lorandite (TlAsS₂).
2. The resulting neutrino flux is a long term average over geologic time that may be more informative than the snapshot view of solar activity provided by artificial detectors.
3. Tl is sensitive to a wider range of neutrino energies than any other detector.

The world’s largest accumulation of Tl-bearing minerals, and the only one suitable as a neutrino detector, is found in the Allchar mine in the former Yugoslavian republic of Macedonia. This deposit contains an estimated 500 tonnes of thallium, mostly in the form of lorandite with a geologic age of 4.3 Ma (Neubauer et al., 2009). In 1983, the international LOREX (LORandite EXperiment) collaboration was set up with the aim to investigate the feasibility of Freedman’s idea (Pavićević, 1988). It was quickly realised that the Achilles heel of the proposal was the relatively shallow depth (120m) of the Allchar mine (Neumaier et al., 1991).

Besides neutrino reactions, a second production mechanism for ²⁰⁵Pb is by cosmic ray muons. Whereas ²⁰⁵Pb production by neutrinos is effectively independent of depth, the muon flux decreases exponentially with depth. But at 120m, the (fast) muon pathway still produces a significant background signal of ²⁰⁵Pb. This paper shows that the burial depth of the lorandite may have been significantly greater in the past, because 4.3 million years worth of erosion may have removed a significant amount of overburden. The erosion rate, and hence the magnitude of the muogenic ²⁰⁵Pb contribution, may be estimated by analysing other cosmogenic nuclides such as ³⁶Cl and ¹⁰Be.

In 1991, the steady-state erosion rate of the Allchar area was estimated by a single ³⁶Cl measurement in limestone (Dockhorn et al., 1991). The ³⁶Cl concentration was found to be high, leading to the conclusion that erosion had been negligible, and that the lorandite had spent most of its 4.3 Ma lifetime at or near a depth of 120m. This conclusion all but terminated the geological neutrino detector and the physics community moved on to other experiments. This paper raises several issues with the Dockhorn et al. (1991) study, suggesting that the lorandite project may have been aborted prematurely (Section 2).

Figure 1: Topographic (left, Jarvis et al., 2008) and geologic map (right) of the Allchar area with indication of the Crven Dol (North) and Centralni Deo (South) mine locations, and the sample locations (yellow circles). A – modern alluvium, B – Quaternary sediments, including (fluvio-)glacial deposits, C – Pliocene andesites, rhyolites and tuffs, D – Jurassic serpentinites, E – Triassic limestones and dolomites, F – Mesozoic sandstones and slates, G – Precambrian gneiss.

Pavićević et al. (2016) recently conducted a cosmogenic ³⁶Cl – ²¹Ne – ²⁶Al study to re-evaluate the erosion rates in the Allchar area. They estimated erosion rates to fall in the 50-100 m/Ma range, which is much higher than the values obtained by Dockhorn et al. (1991). Unfortunately, the Pavićević et al. (2016) study suffers from two methodological issues. First, it primarily focuses on the two lorandite-bearing mines (Crven Dol and Centralni Deo, Figure 1), which were considered to be the most relevant to the neutrino experiment. These sites also suffer from significant anthropogenic disturbance. This intrinsically leads to over-estimated erosion rates. Second, Pavićević et al. (2016) chose not to report any of their ¹⁰Be results because “the nominal erosion rates calculated on the basis of these ¹⁰Be AMS measurements were considerably smaller than those obtained on the basis of ²¹Ne, ²⁶Al, and ³⁶Cl concentrations”. We find this line of reasoning to be questionable, because ¹⁰Be is generally considered to be the most reliable and least problematic cosmogenic nuclide.

To improve on these previous research efforts, we here present the results of a thorough cosmogenic nuclide investigation combining ³⁶Cl and ¹⁰Be measurements in carbonates and quartz from bedrock samples and Majdanska River sediments (Sections 3-5). This two-pronged approach allows us to quantify the spatial variability of apparent erosion rates that may have affected previous erosion rate studies (Dockhorn et al., 1991; Pavićević et al., 2016), whilst simultaneously providing us with more robust catchment-wide erosion rate estimates (Section 6).

2 Previous erosion rate studies

Dockhorn et al. (1991) present the preliminary results of an uncompleted depth profile study. They report the ³⁶Cl/Cl ratio of a single sample of carbonate collected from 23m depth at an undisclosed location. No further compositional information is provided, although the total chlorine content of the sample is speculated to have been overestimated. At a depth of 23m, the poorly constrained muogenic production of ³⁶Cl far outweighs the much better constrained nucleogenic component. The lack of analytical detail and the suboptimal sampling strategy put into question the value of this erosion rate estimate. Furthermore, cosmogenic nuclide geochronology has greatly matured as a science since 1991.

A lot more is known now about the complex production systematics of ³⁶Cl in carbonates, including the effect of thermal neutron reactions on ³⁵Cl (Bierman et al., 1995; Stone et al., 1998; Alfimov and Ivy-Ochs, 2009; Schimmelpfennig et al., 2009), and the first order effect of sample preparation on the meteoric ³⁶Cl component (Merchel et al., 2008). These factors, if not accounted for, lead to an overestimation of the spallogenic ³⁶Cl content. Thus, the erosion rates of the Allchar area may have been greatly underestimated and ought to be re-evaluated using modern insights and methodologies.

In a recent study, Pavićević et al. (2016) presented a dataset of ²⁶Al (in quartz; 15 samples), ³⁶Cl (in carbonate; 3 samples), ³He (in pyroxene; 1 sample), and ²¹Ne (in quartz, sanidine or pyroxene; 8 samples). These samples were collected in bedrock directly above the two lorandite ore bodies (Crven Dol and Centralni Deo) in order to generate the most relevant erosion rate estimates for the lorandite neutrino project. Double-dating of hydrothermal vein quartz with ²¹Ne and ²⁶Al yielded discordant erosion rate estimates, with an excess of stable ²¹Ne relative to radioactive ²⁶Al. This discordance may be attributed to a complex exposure history, or simply to the presence of nucleogenic or magmatic ²¹Ne in the samples. Unfortunately, no ¹⁰Be measurements are reported that could distinguish between these two scenarios.

Pavićević et al. (2016) propose erosion rates of 50-100 m/Ma for the two lorandite localities, with individual estimates covering a huge range from 20-370 m/Ma. This range does not allow a clear-cut decision as to whether the long term erosion rate exceeds the 50 m/Ma cutoff required for the geological neutrino detector to be feasible. We address this problem with a different sampling strategy that combines bedrock samples and modern river sediments collected from the entire catchment area.

3 Sampling strategy

A two square kilometre area near the village of Majdan (41.157^∘N, 21.947^∘E) was combed out in search of suitable samples for cosmogenic nuclide analysis (Figure 1.a). The sampling strategy included four different kinds of sites:

1. The summits of the highest hills, which are covered by andesitic volcanics, and are expected to yield the lowest erosion rates.
2. Rare bedrock exposures in the canyons of tributaries to the Majdanska River that are carved out in dolomite and marble and are expected to yield the highest erosion rates.
3. Bedrock exposures on the steep slopes in between the above two settings, which consist of andesite, tuffs, marble and every conceivable reaction product of these end members.
4. Modern sand and gravel from the Majdanska River, which contain a wide range of lithologies including many quartz bearing phases (gneiss, granite). These are expected to yield intermediate erosion rates that are representative for the catchment-wide average of the area upstream of Majdan.

Unsurprisingly, we found that the vicinity of the lorandite ore bodies was severely affected by human activity. Thus, although these areas have the highest relevance to the proposed neutrino detector, they are the least well suited for cosmogenic nuclides studies as these require steady-state conditions. In contrast, the area between the Crven Dol and Centralni Deo sites has seen little anthropogenic disruption. We would argue that the apparent erosion rates from this area are more representative of the long term trends in the field area.

At each sampling location, the orientation of the sampled surface and the azimuth and elevation of the horizon were carefully measured, as these are needed to correct the ³⁶Cl and ¹⁰Be concentrations for topographic shielding (Table 1). A total of 19 samples were collected, including 10 carbonates, 5 volcanic rocks, and 4 samples of modern river sediment (gravel and sand).

4 Methods

Upon their arrival in the UK, the hand specimens were cut into thick sections and their chemical and mineralogical composition were analysed by QEMSCAN (Quantitative Evaluation of Minerals by SCANning electron microscopy, Allen et al., 2012) at UCL. This reveals that the interplay between recent volcanic activity and the carbonate basement has produced a wide diversity of lithologies in the field area. The carbonate samples exhibit the full range of compositions from nearly pure dolomite to nearly pure calcite. Meanwhile, the volcanic samples feature sufficiently large and abundant phenocrysts for cosmogenic ³⁶Cl analysis. After completion of the QEMSCAN analyses, all the samples were shipped to CEREGE for cosmogenic nuclide analysis using ³⁶Cl (18 samples) and ¹⁰Be (6 samples).

Carbonate and silicate samples were crushed and sieved to 250-500μm grain size. The magnetic fraction was removed from the silicate samples. Before any chemical treatment, whole rock sample splits were kept aside for analysis of the chemical composition by ICP-OES (major oxides) and ICP-MS (trace elements) at SARM-CRPG (Nancy, France). The samples were washed, and for the carbonates (silicates), 10 wt% (20 wt%) were etched of the grain surfaces by 2M HNO3 (a mixture of concentrated HF and 2M HNO) and discarded. In case of the silicates, a 1 g split was taken from the resulting material for analysis of the major oxides (to know the concentrations of the target elements for ³⁶Cl production Ca, K, Ti and Fe) at SARM-CRPG. The carbonates and the remaining sample material of the silicates were dissolved after adding a spike enriched in ³⁵Cl.

In case of the carbonates, a split of this solution was taken for analyses of the target elements Ca and K by ICP-OES at CEREGE. AgNO3 was added to precipitate AgCl, which was extracted and redissolved with NH3. Ba(NO3)2 was added to precipitate BaSO4 , which was filtered out and discarded. The pH was lowered and AgCl precipitated once more, which was extracted and dried for measurement at the ASTER Accelerator Mass Spectrometer (AMS) facility in CEREGE (Arnold et al., 2013).

Sediment samples were prepared similarly, but had the magnetic fraction removed in a Frantz separator first, before dissolving the carbonate fraction as described above. The undissolved silicate minerals were retained for Be measurement. Samples for ¹⁰Be analysis were crushed, sieved and washed. The magnetic fraction was removed by Frantz magnetic separator. Carbonate was removed with HCl. The grains were then leached in a mixture of HCl and H2SiF6 . Atmospheric ¹⁰Be was removed by etching 3 times ~10 wt% off the surface of the remaining quartz grains with HF.

A ⁹Be carrier was added to the residuum which was subsequently dissolved in hydrofluoric acid. HF was evaporated and the sample redissolved in HCl. Raising the pH by addition of NH3 yielded Be(OH)2 precipitate which was separated, dried, and redissolved with HCl. Fe and Mn were removed by ion exchange columns loaded with DOWEX 1×8 resin. Beryllium was recovered and Be(OH)2 precipitated with NH3 , separated, and dried again. The samples were redissolved in HCl, and loaded onto ion exchange columns of DOWEX 50W×8 resin. B was removed, and finally Be recovered, precipitated, centrifuged, redissolved in HNO3 and finally dried down in porcelain crucibles. The samples were oxidised to BeO in a furnace at 700^∘C, before preparation for AMS analysis at ASTER (Arnold et al., 2010).

Erosions rates were inferred from the ³⁶Cl and ¹⁰Be data using the spreadsheet of Schimmelpfennig et al. (2009) and CosmoCalc version 3.0 (Vermeesch, 2007). Production rates were determined using the scaling model of Stone (2000), assuming Sea Level and High Latitude (SLHL) spallation values of 42.2 at[³⁶Cl]/(g[Ca]⋅yr) for Ca (Schimmelpfennig et al., 2011), 148.1 at/(g⋅yr) for K (Schimmelpfennig et al., 2014), 13 at/(g⋅yr) for Ti (Fink et al., 2000), and 1.9 at/(g⋅yr) for Fe (Stone et al., 2005). Catchment-wide erosion rates for samples 11, 15, 17 and 20 were calculated using the average latitude and elevation obtained from a digital elevation model (von Blanckenburg, 2005), using only the area occupied by carbonates and quartz-bearing rocks for ³⁶Cl and ¹⁰Be, respectively.

5 Results

The Allchar deposit is located in the catchment of the Majdanska River, which is underlain by Triassic dolomite, and andesitic lavas and rhyolitic tuffs of Pliocene age (Figure 1.b). The dolomite has undergone various degrees of contact metamorphism and hydrothermal alteration. This is reflected in the chemical and mineralogical composition, as determined by QEMSCAN and ICP-OES/MS (Table 2). Carbonate rocks range from the original dolomite to completely recrystallised marbles made of pure calcite. ³⁵Cl concentrations follow a bimodal distribution, with the marbles containing an order of magnitude more Cl than the dolomites (2-9 vs. 39-58 ppm, Table 2). Thermal neutron-producing U and Th is only present in rhyolitic tuff samples 5 and 6.

All samples yielded measurable amounts of ³⁶Cl (in carbonates and silicates, Table 2) and ¹⁰Be (in quartz, Table 3). ³⁶Cl concentrations range from 28-393×10³at/g, except for sample 10, which contains a much higher 25×10⁶at/g with a 30% analytical uncertainty at 1σ. Because of this high uncertainty and the fact that the ³⁶Cl concentration exceeds the secular equilibrium value, sample 10 is not considered further in this paper.

Similarly, the relatively high U, Th and Cl content of samples 5 and 6 is incompatible with their low ³⁶Cl concentration. Thermal neutrons produced by U and Th are expected to be absorbed by ³⁵Cl to generate excess ³⁶Cl (Alfimov and Ivy-Ochs, 2009). However, no such excess is observed in samples 5 and 6 and so the only way to obtain a finite erosion rate is to assume a physically implausible zero crystallisation age for this rhyolitic material.

On a different note, it is useful to point out that samples 5 and 6 were collected at the same location, at depths of 40 and 110 cm below the surface respectively. Thus, these two samples form a depth profile of sorts. As expected, the ³⁶Cl concentration of the shallow sample exceeds that of the deeper sample (50 vs. 28×10³at/g, Table 2) with the difference agreeing very well with a simple exponential trend. This, again, appears to be incompatible with the thermal neutron production mechanism, which would exhibit a ‘bulge’ at shallow depths (Schimmelpfennig et al., 2009). Apart from samples 5, 6 and 10, all other samples are retained for further interpetation.

The exposure history of the Majdanska River Valley is poorly understood. Although the Balkan peninsula is known to have experienced extensive glaciation during the last Ice Age (Menkovic et al., 2004) and the Majdanska River valley bottom is reportedly covered by ‘Pleistocene glacial deposits’ (Figure 1.b), previous studies have assumed that the ice did not reach the < 1000 m elevations of the Allchar mines (Pavićević et al., 2016). The V-shaped morphology of the Majdanska River valley appears to support the latter scenario (Figure 1.a). Nevertheless, we will quantify the possible effect of glacial erosion by considering two end-member scenarios.

The first scenario assumes that the Allchar area was completely stripped clear by glacial ice, which retreated at 20 ka (= finite exposure scenario). The second scenario assumes an erosion steady-state, in which the field area was never covered by glacial ice. These two scenarios lead to minimum and maximum estimates for the long-term erosion rate, respectively (Table 2 and 3). Three measurements are incompatible with the finite exposure scenario, as they contain too much ³⁶Cl (for sample 7) or ¹⁰Be (for samples 11c and 15c). For the remaining samples, the difference in erosion rate between the two scenarios is between 2 and 48%.

6 Discussion

With the exception of sample 10, the ³⁶Cl concentrations are invariably lower than in the Dockhorn et al. (1991) sample, and exhibit an order of magnitude in spacial variability, ranging from 28 to 393×10³at[³⁶Cl]/g[sample]. This variability is not surprising for bedrock samples, as the assumption of steady-state erosion is only approximately valid at best. Cosmogenic nuclides predominantly form in the upper 1-2 m below the surface and erosion is very variable at this scale.

Three samples contain too much ³⁶Cl (sample 7) or ¹⁰Be (samples 11 and 15) to be compatible with a finite exposure history. Taken at face value, these concentrations argue against glacial erosion of the Majdanska River Valley during the last Ice Age and support the hypothesis of steady-state erosion. In order to further investigate the dispersion of the bedrock erosion rates, let us now partition the ³⁶Cl estimates into topographic and lithological categories (Figure 2).

Figure 2: ³⁶Cl-based apparent erosion rates for the Allchar area grouped according to topographic location (left) and lithology (right). Vertical bars connect the erosion rate estimates assuming steady-state (top) and a 20 ka exposure history (bottom), respectively. Similarly, the geometric mean erosion rates are shown as boxes whose top and bottom margins correspond to the steady-state and finite exposure scenarios. Samples 5 and 6 contain high concentrations of thermal neutron-producing U and Th for which an erosion rate can only be calculated under the assumption of an unrealistic zero crystallisation age. Sample 7 contains too much ³⁶Cl to be compatible with the 20 ka exposure scenario.

As expected from the sampling strategy outlined in Section 4, samples collected from ridges exhibit the lowest erosion rates, with values ranging from 25 to 112 m/Ma and geometric¹ mean values of 54 and 100 m/Ma for the 20 ka exposure and steady-state erosion histories, respectively. In contrast, the bottom of small canyons hosting minor tributaries of the Majdan River exhibit the highest erosion rates (range: 339-915 m/Ma, means: 559-611 m/Ma).

Samples collected from the slopes between the ridges and the canyon bottoms are characterised by intermediate erosion rates (range: 109-456 m/Ma, means: 208-281 m/Ma). Similarly, catchment-wide erosion rates based on ³⁶Cl in the coarse fraction of modern sediment samples 11, 15 and 20, as well as bedrock sample 19 collected at the bottom of the main river channel also yield intermediate erosion rate estimates (range: 66-648 m/Ma, means: 204-264 m/Ma).

Grouping the bedrock samples according to lithology shows that by far the lowest ³⁶Cl-based erosion rate estimates are observed in the hard andesites of samples 7 and 8 (25-89 m/Ma). At this point it is useful to recall the observation that the calcite marble contains an order of magnitude less natural Cl than the dolomite. This suggests that Cl is lost from dolomite during contact metamorphism. The low ³⁵Cl concentration of the calcite reduces the importance of the thermal neutron production pathway of ³⁶Cl. We would therefore expect the dolomite to contain more cosmogenic ³⁶Cl per gramme of Ca (spallation + thermal neutron absorption) than the calcite (spallation only). This is indeed what is observed, with geometric mean erosion rates of 282-358 m/Ma for the marble and 222-303 m/Ma for the dolomite.

The dispersion of the individual estimate around these mean values (84-864 m/Ma and 109-915 m/Ma, respectively) is admittedly too high to draw any firm conclusions. But what is clear is that the effect of thermal neutrons on the erosion rate estimates is at modest at best, because the difference between the dolomite and the calcite would be much greater if it were not.

As expected, catchment-wide erosion rates based on ¹⁰Be-in-quartz are consistently lower than the ³⁶Cl-based estimates (in either bedrock or sediment). This indicates that silicate lithologies erode more slowly than carbonates, a result that is entirely consistent with the ³⁶Cl concentrations in andesite discussed above. Additionally, it is also useful to contemplate the fact that half life of ¹⁰Be is more than four times longer than that of ³⁶Cl. This means that ¹⁰Be averages erosion rates over longer time scales than ³⁶Cl.

The lower ¹⁰Be-based erosion rates could also be taken as evidence for an acceleration of the erosion rates in the Majdanska River Valley over the past million years. Unfortunately, it is impossible to assess the likelihood of this interpretation with the current data.

The ³⁶Cl concentrations in modern river sediments range from 69 to 155×10³at/g. This dispersion should not surprise us given the small size of the area occupied by carbonate lithologies (13, 11 and 20km² for samples 11, 15 and 20, respectively), which may be insufficient to average the upstream heterogeneity in erosion rates. This is made worse by the poor resistence to mechanical abrasion of the carbonate clasts, which biases the detrital carbonate record to nearby sources. Contrastingly, the ¹⁰Be concentrations of the fine and the coarse fractions of samples 17 and 20 are remarkably consistent, with four aliquots all containing 67-68×10³at[¹⁰Be]/g[SiO2].

Samples 11 and 15 contain much more ¹⁰Be. There is no satisfactory explanation for these values, although two separate observations cast some doubt on their validity. First, the fact that only one size fraction was analysed for these samples reflects the difficulty in finding sufficient high quality quartz for cosmogenic nuclide analysis in this particular sample. Second, one cannot help but notice that that samples 11 and 15 were collected immediately below the lorandite mining area (Figure 3), which has seen the greatest anthropogenic disturbance. In any case, the striking consistency of the remaining four ¹⁰Be erosion rate estimates has led us to accept them as the most representative values for the long-term erosion rate of the Allchar area.

Figure 3: Catchment area of the Majdan River and its tributaries upstream of the four modern sediment samples, with indication of the ¹⁰Be-based catchment-wide erosion rate estimates. (c) refers to the coarse and (f) to the fine fraction.

7 Conclusions

This study has re-evaluated the erosion rate of the Majdanska River Valley in southern Macedonia using cosmogenic ³⁶Cl in carbonates from bedrock and modern river sediment, and ¹⁰Be in fluvial quartz. Bedrock samples exhibit the greatest range of erosion rate estimates, from 51 to 915 m/Ma, reflecting the small scale variability of erosion rates in both time and space.

The order of magnitude range in erosion rates obtained from our study is not surprising. It is a result of our sampling strategy, which specifically targeted the fastest (canyons) and slowest (ridges) landforms. This is a very different situation from the study of Pavićević et al. (2016). Their results exhibit a similar degree of dispersion to ours, but for unclear reasons.

In the presence of the observed levels of dispersion, it would be unwise to rely on a single sample of bedrock to determine the erosion rate for the entire Allchar area, as was done by Dockhorn et al. (1991). Samples collected on ridges, in canyons, and in the anthropogenically disturbed Crven Dol and Centralni Deo mining areas are unlikely to yield reliable erosion rates. Carbonate samples collected on slopes and in the Majdan River yield mutually consistent steady-state erosion values of 260-280 m/Ma.

¹⁰Be-derived catchment-wide erosion rate estimates for samples 17 and 20 are similarly consistent but significantly lower at 113-125  m/Ma, again assuming an erosional steady-state. These samples were collected upstream and downstream of the lorandite mining area, respectively, and constrain the erosion rate of the quartz-bearing lithologies in the Majdanska River Valley, which occupy roughly twice as much of the draining area as the carbonate lithologies.

The factor of two difference between the carbonate and silicate erosion rates is entirely consistent with the local geology and geomorphology. Closer inspection of the topographic and geologic maps (Figure 1) reveals that the quartz-bearing lithologies (sandstone, gneiss and rhyolite) occupy the higher elevations, whereas the limestones are found in the valleys.

The Pliocene volcanics are currently found on either side of the Majdanska River Valley, at altitudes of more than 850m, where they form a hard protective cap on top of the much softer Triassic carbonates. It is likely that these Pliocene deposits once filled the valley itself, until they were removed by fluvial or fluvioglacial erosion. Once the Majdanska River cut through this hard layer and reached the comparably soft carbonates, erosion would have accelerated to produce the deep valley that is currently observed. Under this scenario, the ¹⁰Be-based erosion rates represent the pre-incision values, whereas the ³⁶Cl-based erosion rates represent the current erosion rates of the carbonate areas.

Conservatively extrapolating the ¹⁰Be-derived values of 100-120 m/Ma over the 4.3 Ma lifespan of the lorandite deposit at Allchar would suggest the removal of > 450 m of overburden. Adding this amount of shielding to the present 120 m depth of the lorandite deposit would reduce the effect of the muogenic ²⁰⁵Pb contribution sufficiently to be able to see the neutrinogenic component.

Importantly, a recent acceleration of the erosion rate as implied by the ³⁶Cl measurements would further improve the prospects for the lorandite neutrino detector. This is because such an acceleration would mean that the lorandite spent most of its existence at greater depths, only to be exhumed to the surface relatively recently.

Jointly considering the entire body of our twenty-five new erosion rate estimates provides sufficient evidence to discard the single ³⁶Cl measurement of Dockhorn et al. (1991). We would therefore urge the physics community to re-evaluate the feasibility of the lorandite project. Much work remains to be done to make Melvin Freedman’s vision a reality (Pavićević et al., 2012). But given the unique advantages of the geological neutrino detector, we would argue that the geological neutrino detector certainly deserves a second chance.

Authors’ contributions

PV conceived the study, obtained the funding, carried out the field work and wrote the paper, MR carried out the fieldwork and the sample analyses, IS and LB contributed to the sample analyses and the data interpretation, the ASTER team carried out the AMS measurements.

Funding

This research was supported by Leverhulme Grant #RPG-2014-410.

Acknowledgments

PV would like to thank Prof. Günther Korschinek and an anonymous referee for feedback on the submitted manuscript.

References