Holocene sea-level rise evidenced in Hells Bells 234U/238U ratio and geochemical composition

Hells Bells are underwater secondary carbonates discovered in sinkholes (cenotes) southeast of Cancun 16 on the north-eastern Yucatán Peninsula, Mexico. These authigenic calcite precipitates, reaching a length of up to 17 4 m, most likely grow in the pelagic redoxcline. Here, we report on detailed 230Th/U-dating and in-depth geochemical 18 and stable isotope analyses of specimens from cenotes El Zapote, Maravilla and Tortugas. Hells Bells developed 19 during the end of MIS5b/c (~96–90 thousand years ago) and again since the early Holocene, with active growth 20 until present day. The temporal evolution of the geochemistry and isotope composition of Hells Bells calcites 21 appears closely linked to the mid to late Holocene sea-level rise, which reflects changing hydrological conditions 22 of the aquifer. A stabilization of sea level combined with aquifer occlusion during the past ~8 thousand years 23 probably led to a reduction in hydraulic conductivity and a desalinization of the freshwater layer as indicated by 24 decreasing Sr/Ca values. In addition, initial (234U/238U) activity ratios (δ234U0) in the Bells calcite decrease from 55 25 to 15‰ as sea level converges toward its present state. We propose that the Holocene sea-level rise drives 26 desalinization and subsequent deceleration of leaching of excess 234U from the previously unsaturated bedrock. 27


28
Impressive bell-shaped speleothems hanging from cavern ceilings and walls were recently discovered in 30-40 m 29 water depth in a small cluster of meromictic sinkholes (e.g., cenote El Zapote) on the north-eastern Yucatán 30 Peninsula (YP) in Mexico (Fig. 1) 1 . These up to 4 m long conically downward expanding calcareous structures 31 called Hells Bells were formed underwater 1 . Recently, Hells Bells formation was suggested to initially result from 32 CO2-degassing of ascending gas bubbles that accumulate at cave irregularities and further growth of these 33 structures in the carbonate-saturated freshwater layer of cenote El Zapote 2 . Alternatively, Ritter et al 3 proposed 34 that the actual growth of Hells Bells is most likely restricted to the pelagic redoxcline, a 1-2 m thick zone of steep 35 redox-gradients of electron acceptors and reduced chemical species above a sulfidic halocline. The redoxcline 36 overlaps with a distinct milky white horizontal cloud, the turbid layer, into which some Hells Bells partly reach or 37 completely hang inside. Based on hydrogeochemical profiles, a biologically induced authigenic calcite precipitation 38 within the turbid layer was hypothesized 1,3 . In any case, the growth of these speleothems is most likely not 39 continuous. Instead, the corroded lobes of dog-tooth spar crystals and microcrystalline calcite layers of Hells Bells 40 may represent alternating phases of growth and intermittent dissolution, which suggest an episodic elevation of the 41 halocline and, thus, the redoxcline and zone of calcite precipitation 1,3 . So far, the timing of Hells Bells growth 42 remained uncertain since only two small specimens have been 230 Th/U-dated yielding ages between 5,200 and 43 300 years 1,2 . To better constrain the timing and growth rate of Hells Bells, we conducted a systematic study using 44 230 Th/U-dating on several large specimens. In addition, small nodules growing on a drowned tree trunk infer active 45 growth within the redoxcline. Moreover, we have investigated the geochemical composition of these Bells and 46 discovered a small but systematic temporal trend in initial ( 234 U/ 238 U) activity ratios and the Bells Sr/Ca ratios that 47 seems closely linked to the terminal sea-level rise of the mid to late Holocene. ( 234 U/ 238 U) activity ratios are reported 48 here in delta notation (δ 234 U values), representing the deviation of ( 234 U/ 238 U) from secular equilibrium (δ 234 U = 49 ( 234 U/ 238 U)-1). The initial value of δ 234 U (δ 234 U0) can be calculated from the measured δ 234 Um by correcting it for the 50 decay of excess 234 U since the time of sample formation (t): 51 234 0 = 234 * 234 * 52 234 U and 238 U as well as Sr and Ca enter freshwater and seawater from host rock dissolution with preferentially 53 leaching of 234 U due to the processes of alpha-recoil. In general the Sr/Ca ratios and δ 234 U0 values are suspected 54 constant for seawater while in cave drip waters and speleothems, Sr/Ca ratios may depend on the amount of 55 precipitation and other processes 4-6 . The δ 234 U0 value is influenced by several processes, such as the alpha-recoil 56 process, host rock dissolution, and redox-behavior of Uranium 7-10 . In general, the δ 234 U0 value of secondary 57 carbonates from the northern YP is close to secular equilibrium [11][12][13]  level of the cenotes in the north-eastern YP is known to be roughly equal to sea level due to connective passages 68 with the Caribbean Ocean and the low hydraulic gradient of about 1-10 cm km −1 24 . Precipitation rapidly infiltrates 69 through the porous limestone into the underlying coastal aquifer consisting of a meteoric water mass, the freshwater 70 lens, above saline water intruding from the coast 25 . The thickness of the freshwater lens varies between <10-100 71 m and is generally thinner towards the coast 26 , resulting in a higher salinity of the freshwater lens close to the coast 72 than in inland areas 25 . The halocline separates the meteoric and marine water bodies and is usually characterized 73 by undersaturation with respect to CaCO3, leading to cave formation and conduit enlargement in the coastal 74

Hydrogeochemistry of cenote El Zapote 83
Uranium concentrations are nearly constant at ~12 nmol L -1 in the upper 35 m of the oxygenated freshwater layer 84 and instantly drop by an order of magnitude as soon as the water becomes anoxic ( Supplementary Fig. 1b δ 234 U0 values steadily decline in all studied Bells from three different cenotes from 55‰ to around 20‰ between 100 ~8-4 ky, followed by a minor decrease to values of ~15‰ to present (Fig. 2). and 3). The determined ages are mostly in stratigraphic order. Layers within ~3.5 cm distance from the marine 112 limestone slab, yielded ages between 15.5 and 10 ky (Fig. 3, Supplementary Fig. 2c and 3). One of these samples 113 was taken immediately adjacent to a growth interruption, which is macroscopically identified as a black rim 114 conformed by pyrite ( Supplementary Fig. 2c). This sample yielded a significantly older age (15.51 ± 0.21 ky) than 115 the two samples above, thus representing a major stratigraphic inversion. A similar age of 15.02 ± 0.13 ky was 116 obtained from a sample of the root of Big Bell, which was also sampled near a thin layer of pyrite (Supplementary 117 Fig. 2b). Since we do not know whether the pyrite layeror the processes that led to its formationmay have had 118 an influence on the geochemistry of the adjacent calcite layers and thus the 230 Th/U ages, or whether this strong 119 inversion is the result of a mixture of Hells Bells material of different ages, we exclude these older samples from 120 the following discussion and focus on the mid-to late Holocene samples. From 7.7-2.5 ky, i.e., from 3.7-53.5 cm, 121 the ages of ZPT-7 suggest more regular growth with an average growth rate of about 100 µm yr −1 (Fig. 3). At 2.5 122 ky (53.5 cm), a growth interruption is evident with a duration of about 1 ky (Fig. 3, Supplementary Fig. 3). The tip 123 of ZPT-7 at 56.7 cm is dated to an age of 1.284 ± 0.083 ky, which possibly corresponds to the timing of Bell    Table 2.

153
Hells Bells carbonates show strongly correlated Sr/Ca and Ba/Ca ratios, and the linear fit intersects with the Sr/Ca 154 and Ba/Ca ratio of the host rock (Fig. 4b). The ratios of these samples show a decreasing trend with increasing 155 distance from the apex and decreasing age, respectively (Fig. 4b). As for the stable isotopes, the Sr/Ca and Ba/Ca 156 ratio of the oldest samples (96-90 ky) show slightly different values than the general trend. 157 Dark, brown-colored layers identified on the polished half of ZPT-7 correspond to elevated Mn/Ca and Fe/Ca values 158 ( Fig. 3) and show a large scatter. The molar ratio of Mg/Ca (24-40 × 10 -3 ) appears rather constant through time 159 (Fig. 3). In contrast to the metal/calcium ratios, the multivalent non-metal Sulphur (S) reveals an opposing trend 160 with an increasing S/Ca ratio of ZPT-7 from ~0.1 to 2.7 × 10 -3 with decreasing age (Fig. 3). These geochemical 161 trends are thus evident throughout all studied Hells Bells, and even in the nearby cenotes Maravilla and Tortugas 162 (Fig. 4). the ones from Maravilla and Tortugas show lower growth rates (~4-18 µm a -1 ) than ZPT-7 (~100 µm a -1 ). One 194 explanation for that might be that those Bells may have experienced continuous changes between growth and 195 dissolution of calcite, resulting in lower net growth since these Bells were hanging in greater depths, i.e., closer to 196 the acidic water below the redoxcline. This assumption is supported by the strong lamination of these samples 197 ( Supplementary Fig. 4). The ages of 2.8 and 1.3 ky, determined for the lowermost parts of Big Bell and 198 respectively, may refer to the times when these specimens broke off the cave ceiling and fell on the cave floor, 199 where they stopped growing. Whether these break-offs were gravitationally triggered by the weight of the Bells or 200 even by a devastating event (e.g., earthquake) remains speculative. Interestingly, the age of 1. shown that heavy rainfall events (i.e., hurricanes) can cause turbulent mixing between the marine and the meteoric 262 waters, leading to an increased freshwater salinity 54-56 . Thus, the authors suggest that the decline in aquifer salinity 263 during the last ~7 ky likely reflects a change in hydrology (drying trend) associated with a decreased freshwater 264 flow in the aquifer 25 . Similarly, benthic microfossils from cenote Aktun Ha, 75 km south of the Hells Bells cenotes, 265 indicate a gradual decrease in freshwater salinity over the past ~4.3 cal ky BP 20 . They found convincing 266 relationships between changing aquifer salinity and late Holocene precipitation patterns. However, the authors also 267 point out that other factors may influence the salinity of the aquifer, such as aquifer occlusion and coastal 268 sedimentation, both long-term processes resulting from sea-level stabilization, which may gradually reduce 269 hydraulic conductivity and turbulent mixing of the aquifer, leading to a desalinization of the freshwater layer 20 . 270 Reconstructions of middle to late Holocene sea-level rise on the YP (Fig. 6a)  The pattern of δ 234 U0 values matches the progression of middle to late Holocene sea-level rise on the YP even 280 better than that of Sr/Ca ratios, presenting an average deviation of only 6% from the sea-level fit (Fig. 6). Over the 281 past ~8 ky, δ 234 U0 values decrease continuously from values around 55-60‰ to values between 15-20‰ as sea 282 level converges toward its present state, roughly equivalent to an 8-10 m sea-level increase (Fig. 6c). of uranium in the saltwater body is reduced due to anoxic conditions in which U behaves particle reactive. In 293 contrast, the overlying freshwater seems rather homogeneous and is close to secular equilibrium (~16‰, 294 Supplementary Fig. 1). 295 In a recent study, Wendt et al 15 interpreted δ 234 U0 values of subaqueous calcite from Devils Hole 2 cave as a proxy 296 for water-rock interactions in the regional aquifer. They propose that changes in the elevation of the water table are 297 responsible for changes in the amount of leached excess 234 U from the bedrock and that variations in ( 234 U/ 238 U) 298 activity ratios therefore coincide with interglacial-glacial cycles. Although the setting of the YKA is distinctly different 299 from that of the Devils Hole in southwest Nevada, they both are subject to recurrent changes in water level elevation 300 on interglacial-glacial timescales. In Nevada, water table fluctuations are driven by variations in recharge amount  301 to the local groundwater flow system 58 , whereas on the YP, they are associated with glacio-eustatic changes in federal state of Quintana Roo (Fig. 1). Cenote El Zapote (20°51′27.78″N 87°07′35.93″W) is water-filled and 334 connected to the surface by a 28 m deep vertical shaft (Fig. 1a). The freshwater lens and the saline water mass 335 are separated by a thick halocline reaching from 36.7-51.7 m water depth (Fig. 1a)  Uncertainties are reported at the 2σ-level, and do not include half-life errors. 383

Major and trace element analysis 384
About 3 mg of each powdered carbonate were digested in 2 mL of 10% HNO3 for major and trace element analyses. 385 Subsequently, concentrations of Ca, Mg, Sr, Ba, S, Fe and Mn of diluted aliquots were determined by ICP-OES at 386 Heidelberg University. Quality control of the measurement was performed using reference materials SPS-SW1 and 387 SPS-SW2 with recovery rates of ~100% for the analyzed elements. Quality control of the digestion of the carbonate 388 material was performed with analyses of parallel aliquots of the limestone reference material ECRM 752-1. The 389 recovery rate of the certified values was ~100% for the elements Ca, Mg, Sr, Fe and Mn, while a yield of ~90% 390 was achieved for the element Ba and ~80% for the element Fe. The resulting element to Ca ratios are presented 391 as molar ratios. 392

Competing interests 414
The authors declare that they have no conflict of interest