Petrographic and geochemical constraints on the formation of gravity‐defying speleothems

Cave carbonates, seemingly growing in defiance of gravity, have attracted the community's interest for more than a century. This paper focusses on ‘helictites’, contorted vermiform speleothems with central capillaries. Petrographic, crystallographic and geochemical data of calcitic and aragonitic helictites (recent to 347 ka) from three caves in Western Germany are placed in context with previous work. Aragonitic helictites from one site, the Windloch Cave, form exceptionally large and complex structures that share similarities with the celebrated helictite arrays in the Asperge Cave in France. Aragonitic and calcitic helictites differ significantly in their crystal fabrics and internal geometry. Calcitic helictites are best described as a composite crystal fabric consisting of fibrous mesocrystals. Aragonite helictites display a complex fabric of acicular to platy crystals, some of which show evidence for growth‐twinning and perhaps crystallisation via a monoclinal precursor stage. The micro‐tomographic characterisation of several orders of channels and their complex architecture raises important questions regarding fluid migration and helictite architecture. In terms of their isotope geochemistry, helictites are depleted in 13C to various degrees, isotope values that are controlled by the mixing of fluids and mineralogy‐related fractionation. Regarding their δ18O values, most helictites overlap with other calcitic and aragonitic speleothems. Previous models explaining the twisted morphology of helictites are discussed from the viewpoint of fluid migration and CO2 degassing rates, mineralogy and helictite petrography. For the complex aragonitic helicities documented here, the most likely mechanisms to explain the contorted growth forms include the internal capillary network combined with localised (sector) growth at the helictite tip. The morphologically simpler calcitic helictites are best explained by capillary and surface flow. Future work should include geomicrobiology to assess the significance of induced mineralisation and transmission electron microscopy analysis to more quantitatively assign crystallographic properties.


| INTRODUCTION
are contorted cave carbonates twisting in any direction and protruding out, up and down from the cave floor, wall, ceiling or from parent stalactites or stalagmites (Collett, 1878;Davis, 2019;Dolley, 1886;Spencer, 1903; see also review in Hill & Forti, 1997or in Reinhold, 1998. When helictites form on cave floors, they are sometimes referred to as 'heligmites', but this term is not widely used. Perhaps due to their intriguing growth forms, these features have attracted significant interest in the community (Cabrol, 1979;Cunningham et al., 1995;Graeme, 1981;Hill & Forti, 1997;Martini, 1987;Onac, 1992;Onuk et al., 2014;Reinhold, 1998;Richter & Neuser, 2007;Tisato et al., 2015; see also 'Helictite Cave' in Cantabria, Spain, or the journal 'Helictite' published by the Australian Speleological Federation). When these speleothems grow, their c-axis orientation is skewed, creating narrow, curvilinear tubes or more complex three-dimensional structures, including filament, beaded, vermiform, ribbon, antler or butterfly types, among others (Hill & Forti, 1997;Rowling, 2001). Here, the label 'helictite' (from 'helix') is applied as an umbrella term for all forms. Most helictites are short, only one to several centimetres in length, but also complex, decimetre-sized structures have been reported (Tisato et al., 2015). Both aragonitic and calcitic helictites have been described. Most of these features form in the vadose environment. Davis (1987) and Rowling (2003) described exceptional phreatic case examples.
Several studies proposed models for the formation of helictite speleothems. Hypotheses published thus far included helictite formation via hydrostatic capillary pressure, biologically mediated processes (fungi, bacteria and algae), deposition from aerosols, changes in piezoelectric pressure potential and osmosis (see discussion in Hill & Forti, 1997;Reinhold, 1998;Tisato et al., 2015). Previous studies focussing on calcitic helictites included descriptive (morphological and petrographic) approaches (Göbel, 1972), but also made use of experimental work (Huff, 1940;Slyotov, 1985), applied petrographic (Reinhold, 1998) and optical methods or geochemical tools (Onuk et al., 2014). However, despite significant advances and numerous hypotheses published, there is no conclusive, overarching model explaining the formation of these intricate features. Arguably, this might be related to the plethora of processes involved, potentially making each cave a case on its own, and the fact that the majority of studies focus on one cave and one helictite type.
Here, a state-of-the-art petrographic and geochemical data set of aragonitic helictites from the recently discovered Windloch Cave, Western Germany, yielding some of the largest and most spectacular helictite occurrences in Western Europe, is presented, discussed and placed in context with previous hypotheses explaining helictite formation. Helictite data from the Windloch Cave are complemented by such from other caves in this region (Dechen Cave and Huettenblaeser Cave; Figure 1). The purposes of this paper are: (i) to present state-of-the-art optical data (macrophotography, scanning electron imaging, electron backscattered analysis and microtomography), U-series age dating and isotope and elemental geochemistry (helictites and cave waters); (ii) integrate data shown here with previous work to discuss the petrography and geochemistry of helictites; (iii) present and evaluate models to explain helictite formation and morphology.

| Windloch Cave
The entrance to the Windloch Cave (50°59′39″N/07°27′03″E; opens at 199 m a.s.l.; Figure 1A,B) lies ca 2 km ENE of the town of Engelskirchen. The Windloch Cave was discovered in 2019 in a systematic regional cave survey by the Klutert Caving Society. By early 2022, the cave passageways reached a total known length of 8.5 km ( Figure 1B), making the Windloch Cave the ninth largest cave in Germany, but the cave is likely much larger. The cave has developed in partially dolomitised, Lower Devonian limestones of the Hohbraek Formation reaching several tens to one hundred metres in stratigraphic thickness. The host rocks are mainly built by reefal facies with breccia, arenitic and muddy intervals. Locally, argillaceous horizons are interbedded. Upsection, the carbonates grade into clastic sedimentary rocks. The extensive labyrinthine Windloch Cave system developed around a saddle structure with meteoric fluids leaching a pre-existing fracture system in a phreatic environment. Abundant cryogenic calcites are found in many cave chambers and passages and assess the significance of induced mineralisation and transmission electron microscopy analysis to more quantitatively assign crystallographic properties.

K E Y W O R D S
aragonite, calcite, crystallography, geochemistry, helictites, speleothem suggest that water in cave ponds froze during glacial periods. Among the various speleothems in several galleries, extensive gypsum deposits and uncommonly large and complex aragonitic helictite 'bushes' (including anthodites and soda straw features) are found (Figure 2A,B,C). The authors collected helictite and water samples in the Wunderland ('wonderland') chamber in fall 2021.

| Dechen Cave
The Dechen Cave (51°21′56″N/07° 8′41″E; opens at 170 m a.s.l; Figure 1A,C) is used as a touristic cave (the nearby German Cave Museum opened in 2006) and lies in the Gruener River Valley, between the towns of Iserlohn and Letmathe. The Dechen Cave forms part of the Gruener River Valley cave system (ca 60 caves) with an overall known length of cave passageways of 17 km and was discovered in 1868 by railway workers. Across the Gruener River Valley system, a total of five cave levels formed that are correlated to terraces of the nearby Lenne River (Niggemann et al., 2018). The known cave passages and chambers assigned to the Dechen Cave proper reach a length of 0.9 km ( Figure 1C). The distance between the entrance of the Dechen and that of the Huettenblaeser caves is less than 1 km. The Dechen Cave was formed in Upper Devonian limestones, predominantly with biostromal facies. The cave formed due to limestone dissolution of a fracture system in a phreatic diagenetic system during the Mid-Pleistocene. Nearly all galleries of the Dechen Cave are richly decorated by various speleothems (Figure 2D), of which the oldest date back to ca 500 000 years bp (Dreyer et al., 2010;Niggemann et al., 2000). The cave floor is covered by up to 3 m of cave loam from which several thousand bones and teeth of Pleistocene mammals, including those of the cave bear (Ursus spelaeus), were collected. The authors collected calcitic helictites and water samples in a crevasse near the Kanzel ('pulpit') Chamber in fall 2021.

| Huettenblaeser Cave
The Huettenblaeser Cave (51°22′8.1″N/07°39′16.8″E; opens at 193 ad 173 m a.s.l) is another cave of the Gruener River Valley system. The site is approximately 45 km to the NNE of the Windloch Cave ( Figure 1A,D). The Letmathe caving society first explored the cave in 1993. By the end of 2021, a total of 4.85 km (vertical distance of 46 m; Figure 1D) of cave passageways has been explored, but the Huettenblaeser Cave is likely significantly larger. The cave formed in wellbedded limestones, particularly reefal and lagoonal facies (upper Middle Devonian; Givetian stage; Krebs, 1974). Variscan fault systems locally affect the limestones, allowing localised hydrothermal dolomitisation of the host rock F I G U R E 2 Images showing helictite speleothems in the three caves studied. (A) Exceptionally large and complex, aragonitic helictite 'bush' (dubbed 'hydra', ca 1.4 m across) forming at the ceiling in a cave passage of the Windloch Cave. Cave decorations also feature soda straw and anthodite speleothems with transitional features between the former two and helictites. 230 Th/U-age dating suggests that some of these features are up to 350 ka old, but active growth continues to the present day. (B) Complex speleothem including a stalagmite at the base and an uncommonly large, aragonitic helictite structure expanding from cave ceiling downwards towards the stalagmite (dubbed 'tree of happiness'). The structure is 0.85 m tall. (C) Accumulation of stalactitic soda straws (red arrows) and helictite fragments accumulating on the cave floor. Wherever possible, fragments from the cave floor were sampled so as not to damage the active speleothems. Scale refers to the foreground. Images A, B and C (Windloch Cave) courtesy of G. Steffens, German mining museum. (D) Various calcitic stalactites and helictites (arrows) in Dechen Cave. (E) Close up of aragonitic stalactites and helictites (arrows) in Huettenblaeser Cave. Note the linear orientation of speleothems along the Variscan fracture (green arrow) acting as a passage for meteoric waters. The red arrow depicts the sample documented here. and the precipitation of various carbonate cements. Like the Windloch Cave, the Huettenblaeser Cave formed in a phreatic meteoric setting by solution-widening of fracture networks. The spatial organisation of the cave points to at least three altitude levels that are most likely related to Quaternary groundwater levels and can be correlated to terraces of the nearby Lenne river. A remarkable feature is the up to 25 m high, solution-widened fractures. Numerous speleothems are found in many cave galleries; these include aragonitic helictites ( Figure 2E). Dated speleothems in the cave ( 230 Th/U time series) have an age of 13.5 ka and younger (Weber et al., 2021) but at least one speleothem dated back to the MIS 7 (ca 216 ka; Lin et al., 2017;Yang et al., 2015). The authors collected the helictite and water samples described in this paper in the Tunnel ('tunnel') Chamber in fall 2021.

| Fieldwork
The collection of helictite and water samples took place in the context of dedicated sampling visits to the cave by the authors and colleagues from local caving societies. Wherever possible, helictite fragments that accumulated gravitationally on the cave floor ( Figure 2C) were sampled to limit the damage to cave decorations. Given that helictites are not related to dripping or flowing water, such as the case for stalagmites and stalactites, water samples were collected from nearby cave pools.

| Optical analyses
Macrophotographic images were taken utilising a Panasonic LUMIX-DC-G9 and a Canon EOS-R 24 camera. Illumination was through a StoneMaster Super Nova LED. Eight separately adjustable segments, a colour temperature of 5700 K and a maximum Illuminance of 2.9 Mlux was used. To reduce vibrations, both cameras are mirrorless system cameras. Given that the depth of field for the different lenses is limited, each image is built by a variable number of single images taken in small steps along the z-axis using a photo stacking unit (StackMaster-Studio, travelling distance 95 μm, resolution 0.1 μm). Hence, each photographic image shown in this paper is a stack consisting of 10 to more than 200 images compiled into the final illustration utilising the Helicon Focus software. The number of photographs required to form an image stack depends on the magnification factor.
To analyse the three-dimensional orientation of the central channel/capillary and its position within the speleothem, micro-computed tomography was applied.
Data were collected in December 2021 at 100 kV with a nanotom m equipped with a nanofocus X-ray tube built by Phoenix (Wunstorf, Germany). A microtomographic data set with a size of 2118 × 1766 × 2396 slices in x-, y-and zdirection and a size of 17.5 was reconstructed. The resulting isotropic voxel size of 17.5 GB is 8.3 μm. The CT-data are archived at the Ruhr-Universität Bochum.
The crystallographic orientation of the aragonitic and calcitic crystals and related mesocrystals building the helictites studied here was determined by electron backscattered diffraction (EBSD) with a symmetry 2 detector (Oxford Instruments) attached to a Scios 2 FIB-SEM (ThermoFisher). Surfaces of thin sections were chemo-mechanically etched using colloidal silica (OP-S) for 5-20 min to remove the uppermost layer of the crystal lattice that was mechanically deformed during thin-section fabrication and coated with a layer of carbon ca 10 nm thick. The data acquisition and analysis were performed using the Aztec and Aztec Crystal software by Oxford Instruments.
To collect the EBSD data from thin sections, the scanning electron microscope (SEM) was operated at a beam energy of 20 kV, an aperture of 60 μm, a working distance of 25 mm and a tilt angle of 70°. Thin sections were mapped in the high-resolution mode using a grid matrix (1913 × 772 points, 1 476 836 total points, 30 h duration) and a step width of 0.45 μm for the aragonitic samples. A step width of 7.88 μm for the longitudinal section and 1.61 μm for the cross-section of the calcitic samples were used. The grid matrix is 470 × 406 points, 190 820 total points and 62 h duration for the longitudinal section and 1947 × 1449 points, 2 821 203 total points, and 64 h duration for the cross-section. The orientations of the crystals in the individual maps were visualised using a rainbowcolour coding, where identical colours indicate identical crystal axis orientations. In addition, orientations of the measured crystallographic axes were plotted as pole figures into the lower hemisphere of a Schmidt net.
Portions of the speleothems were broken into pieces. Subsequently, their surface was examined under a highresolution field emission SEM (HR-FESEM) type Zeiss Merlin with a Gemini II column using secondary electrons in the chamber Everhart-Thornley detector. Due to some samples' complex three-dimensional surface, the specimens were gold coated with a layer of 25 nm. Working distances varied between 5 and 17 mm, acceleration voltage between 5 and 7 kV, and the probe current between 50 and 100 pico-amperes.

| Geochemical analyses of cave waters and speleothems
Element concentration and carbon and oxygen isotope analyses were performed in the laboratories of the Institute of Geology, Mineralogy and Geophysics at the Ruhr-University Bochum or taken from the literature (host rock limestones and hydrothermal dolomites; Niggemann et al., 2018;Pederson et al., 2021aPederson et al., , 2021b. For element concentration measurements (Ca 2+ , Mg 2+ , Na + , Fe 2+ , Sr 2+ , Mn 2+ ), 1 ml deionised water and 1 ml 3 M HNO 3 were added to 1 ml water sample from cave pools using a Dilutor of the type microlab 600 series (Hamilton) resulting in a 3.5% HNO 3 solution. Element concentrations were determined using an ICP-OES (inductively coupled plasma optical emission spectrometer) of the type iCAP 6500 DUO (Thermo Fisher Scientific). Concentrations are given in mg/l, and Mg/Ca ratios are molar ratios. The Multi-element standard solution VIII (VWR Chemicals) and TraceCERT multi-element standard solution 5 (Sigma-Aldrich) have a 1σ-reproducibility for Ca 2+ of 1.2% and 1. 0%, for Mg 2+ of 0.7% for both, for Na + of 1.5% and 1.2%, for Fe 2+ of 0.7% for both, for Sr 2+ of 0.6% for both and for Mn 2+ of 0.9% and 0.7% respectively (n = 6 and n = 10, 02/2022).

| Age dating
Sample preparation and analysis were performed at the Institute for Geosciences, University of Mainz, Germany. The samples were weighed, dissolved in 7N HNO 3 and spiked with a mixed 229 Th-233 U-236 Usolution. Details on the calibration of the spike are given in Gibert et al. (2016). The samples were then evaporated and treated with a mixture of concentrated HNO 3 , HCl and H 2 O 2 to destroy potential organic material. Then, the samples were dried, redissolved in 6 N HCl, and U and Th were separated using ion-exchange columns. Mass spectrometric analyses of U and Th isotopes were performed using a Thermo Fisher Scientific NeptunePlus multi-collector inductively coupled mass spectrometer (MC-ICPMS). All activity ratios and ages were calculated using the decay constants of Cheng et al. (2013) and corrected for detrital Th, assuming a bulk Earth 232 Th/ 238 U weight ratio of 3.8 ± 1.9 for detritus and 230 Th, 234 U and 238 U in secular equilibrium. All errors are given at the 2σ-level and include analytical uncertainties and uncertainties in the spike calibration, the contribution of natural isotopes from the spike and the detrital correction (50%). For further analytical details, the reader is referred to Obert et al. (2016).

| Chronology
The results of 230 Th/U-dating are shown in Table 1. The number of published helictite 230 Th/U-ages is very limited to the authors' knowledge. The ages shown here are not time series data typically reported for stalagmites but represent bulk ages of millimetre-sized subsamples. In the authors' opinion, helicities are actively growing in all three caves. Active and geologically young helictites, both calcitic and aragonitic, are white-translucent to glassy (Figures 2-4). With increasing age, staining with clay minerals and soot is observed ( Figure 4E). In the Windloch Cave, tan-coloured, corroded samples are present ( Figures 2C and 4D). One of these shows an age of 347 ka (Table 1). This suggests that helictite formation in the Windloch Cave goes back at least to the Holstein (Sarnthein et al., 1986) interglacial. The aragonitic helictite from Huettenblaeser Cave shows an age of 0.35 ka, and the calcitic sample is 4 ka. Clearly, these age data represent snapshots, and a more systematic dating campaign will refine the chronology and probably also document helictites from other interglacials.

| Helictite architecture
The helictites described here show a wide range of morphotypes and share many attributes with such described from other caves worldwide (Hill & Forti, 1997;Rowling, 2001). Samples collected in the three caves twist in all directions and protrude out, up and down from the cave floor, wall, ceiling, parent stalactites, stalagmites and soda straws (Figures 2 and 3). The largest and most complex structures are found in the Windloch Cave ( Figure 2A,B). In some cases, helictites, anthodites and soda straw speleothems form a dense meshwork about 1.4 m in diameter (the 'Hydra'; Figure 2A). In one case, downward-growing helictite arrays and an upward growing stalagmite connect to form a complex, tree-like structure about 0.85 m tall ('Tree of happiness'; Figure 2B). In Europe, helictite arrays of comparable size and complexity have previously only been reported from the Asperge Cave (France; Tisato et al., 2015). Regarding the three caves described here, aragonitic helictites display significantly more complex morphotypes than calcitic ones. They are often vermiform, in part filiform, with antler or corkscrew morphologies (see Forti, 1997 andRowling, 2001 for definitions). Helictite lengths range from less than 1 to about 40 cm. On average, helictite diameter is one to several millimetres, but widths of >1 cm are found, too (Figures 2 and 3). Helictites are thickest (vermiform) near their base and thin (filiform) towards their tips. Locally, helictite tubes may converge and coalesce to form a complex composite or corkscrew-like structure ( Figure 3). When reaching critical lengths (often ca 40 cm, but this value depends on the growth orientation), the weight of the helictite overcomes the critical breaking strength, and fragments gravitationally accumulate on the cave floor ( Figure 2C). Calcitic helictites in the Huettenblaeser Cave often display vermiform, antler or corkscrew structures ( Figure 2E).
All samples studied are characterised by a central capillary (or channel). In this paper, the term 'capillary' does not imply that the physics of capillary fluid flow can be applied to all case examples studied, but serves as a descriptive label in the sense of a 'tube that has a very small (hairlike) bore'. Capillaries are, on average, 150-200 μm in diameter and round to irregular in shape. The SEM imaging ( Figure 5) depicts that the inside of individual channels is rugged, with crystallites protruding into the voids ( Figure 5B). Microtomography imaging ( Figure 6) confirms that the geometry of the channels within the aragonitic helictites is consistently irregular, and several orders of channels are observed. The following terminology is applied: first-order, secondorder and third-order channels. The helictite's firstorder channel system is oriented sub-parallel to the main growth direction, may be more or less linear over several millimetres to centimetres, but then abruptly bends and curves sharply, taking up a new growth direction that may deviate by more than 300° from the previous one. Elsewhere, helictites split in two or more tubes (antler morphotype) with comparable diameters. Second-order side-capillaries (referred to as 'canalicules' by Andrieux, 1965) branch away from the main channel and growth axis and may form stubby protrusions when reaching the helictite surface. Less often, second-order capillaries evolve to become first-order features. Conversely, where two or more helictite strands converge, first-order channels may converge into one main channel. Abandoned minor channels that do not reach the helictites' surface and branch off from secondorder channels are referred to as third-order capillaries ( Figure 6). Taking the sub-vertical branch of the aragonitic specimen shown in Figure 6A for reference, the following basic statistics results. The specimen is 35 mm in total length and between 3 and less than 1 mm (tip) in thickness. In this specimen, first-order capillaries amount to a cumulative length of 42 mm, and their diameter ranges between 250 and 200 μm (thinning somewhat towards their end). Second-order capillaries have cumulative lengths of 38 mm. Their diameter ranges between 220 and less than 10 μm (thinning towards the end of the capillary). Two of the secondorder capillaries extend outward and form visible protrusions on the helictite surface (but the capillaries do not reach the surface). Third-order capillaries reach 10 mm in cumulative length and have diameters between 150 and less than 10 μm (at their end). The length ratio of first-order to third-order capillaries is thus 4.2:3.8:1.
In the case of the caves studied here, calcitic helictites are typified by a less complex internal channel geometry than aragonitic case examples. In many cases, the calcite crystals and mesocrystals building the helictite are glassy, revealing the internal structure of the channels ( Figure 4E). Typically, the channel architecture is that of a first-order capillary, commonly between 100 and 200 μm in diameter, with rugged walls resembling the 'pearl-string' morphology described in Kempe and Spaeth (1977).

| Petrography of calcitic and aragonitic helictites fabrics
The petrography of carbonate fabrics building the helictites is documented in Figures 4 and 5. Major differences are found when comparing the seemingly simpler calcitic and the morphologically more complex aragonitic crystallites. Macrophotography images of calcitic helictites from the Dechen Cave display a dense fabric of translucent (glassy) blocky crystals, most of which have more or less idiomorphic diameters of several hundred microns to several millimetres in size (Figures 4E,F and 7A,C). From a mesoscopic perspective, the surface of the calcitic helictites is smooth ( Figure 4E). However, when viewed under the SEM, the surface of individual calcitic helictites is typified by numerous steep rhombohedra indicating a more complex internal fabric ( Figure 5E,F). The terminology of previous workers is followed here with the term 'mesocrystal' used to describe the fibres making up the blocky calcite crystal fabrics. The complex mesocrystal structure of calcitic helictites is documented by EBSD mapping (Figure 7). When mapped in transverse sections ( Figure 7A,B), the complex spatial organisation of the fibrous fabrics becomes obvious. On average, a dislocation of 5-10° in mesocrystal orientation is found. When mapped in a longitudinal direction, divergent arrays of fibrous mesocrystals are observed ( Figure 7C,D). Fibrous mesocrystals fan out from the helictite axis, and a scatter of 5-10 o in their grain dislocation orientation is revealed by EBSD mapping.
In the case of aragonitic helictites, the commonly observed crystal morphologies are acicular (needle), fibrous, platy and lath shaped forms with pseudo-prismatic to flat terminations ( Figure 5A through D). The twinning of some platy and lath-shaped crystals is observed under high SEM magnification. Crystal lengths are on the order of several hundreds of microns. Length:width:thickness (L:W:T) ratios range from ca 100:1:1 (needles) to ca 100:30:5 (platy crystals; Figure 5D). The SEM imaging suggests that the aragonitic helictites are built by a dense, but probably permeable network of crystals ( Figure 5B). The helictite surface is covered by a fur-like array of shingled crystals deviating from the main helictite growth direction at angles of <15° ( Figures 5B,C and 8). At the helictite surface, individual crystal growth directions change abruptly as the helictite twists and bends ( Figure 5A).
The results of EBSD mapping of an aragonitic helictite tip are shown in Figure 8. The orientation of the three axes of the lath-shaped crystallites is quantified. Note, to gain well-constrained evidence on the orientation of the three crystal axes, transmission electron microscopy (TEM) analyses will be required, but these were not compiled in the context of this study. Crystals and mesocrystals with similar orientations are depicted with a similar colour code. The largest crystals grow at the surface, and the smallest crystals are found in the helictite tube's centre. While the central portions of the helictite show a rather uniform orientation of crystal caxes, an increased level of scattering is found towards the helictite surface.

| Geochemistry of water and carbonate samples
Geochemical data of helictites and water samples are shown in Tables 2 and 3 (Supplemental Materials) and Figures 9 and 10. Carbon and oxygen isotope data of cave host rock (Niggemann et al., 2018) and hydrothermal carbonate phases (mainly dolomites; Pederson et al., 2021a) are taken from published work and are shown in Figure 9. Water samples from ponds in all three caves were analysed, as no drip sites are associated with helictites. While pond waters in the three caves arguably represent the bulk water chemistry in F I G U R E 6 (A, B) Details of microtomography imaging of aragonitic helictite from Windloch Cave are seen from two perspectives. The open portions of the central channel (capillary) are shown in red; the aragonitic fabric is translucent. Note at least three orders of channels with numerous aborted third-order channels not reaching the speleothem surface. In the terminology applied here, second-order channels reach the helictite surface and create protrusion, while first-order channels define the main growth direction. these caves, they might not represent the waters from which helictites formed. The Mg/Ca ratios (mol/mol) of pond waters are consistently low and range from 0.09 to 0.21 (Table S3).
Carbon and oxygen isotope data of helictites (Table S2) represent the bulk data of subsamples 1-2 mm in length and do not discriminate for internal variability at a smaller scale. Two distinct geochemical clusters are found. Calcitic helictites from the Dechen Cave plot in the field of other, Holocene calcitic speleothems from the Dechen and the Huettenblaeser caves, albeit in the depleted spectrum (−11‰) of carbon isotope values (Figure 9). In contrast, aragonitic helictites from all sites studied here are consistently less-depleted in 13 C (−0.7 to −6‰) compared to calcitic speleothems from caves in the region and are enriched in 13 C relative to other aragonitic speleothems of Holocene age from Huettenblaeser Cave. In terms of their oxygen isotope data, aragonitic helictites plot in the range of calcitic speleothems (−6.8 to −4‰). One helictite sample from the Windloch Cave clusters near an oxygen isotope value of −3.5‰ and, thus, exhibits higher δ 18 O values than the corresponding speleothems.
Bulk rock isotope values of the Devonian host rock that forms the three caves studied range from +2 to +4‰ for carbon and from −7 to −5‰ for oxygen. These values apply to limestones (>90% calcite) in all three caves studied (Niggemann et al., 2018) except for portions of the host rock affected by Variscan tectonic overprint and hydrothermal dolomitisation. Variscan hydrothermal dolomites in fault and fracture zones range from +2.5 to +4‰ for carbon, that is, are only moderately enriched in 13 C relative to the host limestones, and from −10 to −7‰ for oxygen (Pederson et al., 2021a). The geochemistry of calcitic, and even more so of aragonitic helictites, is not well constrained in the literature. Reinhold (1998) documented petrographic and geochemical data (δ 13 C = +2‰, δ 18 O = −6‰) from fibrous calcites resembling, in the view of the author, filiform helictites. Onuk et al. (2014) published Mg, Sr and Ba concentrations from calcitic helictites from Dragon Belly Cave in Italy.
In the data shown here, the following systematics are observed ( Figure 9): (i) carbon isotope values of calcitic speleothems (all forms) and aragonitic helictites differ F I G U R E 9 Carbon and oxygen isotope values of aragonitic helictites from Windloch and Huettenblaeser caves and calcitic helictites from Dechen Cave. The range of δ 13 C and δ 18 O values of stalagmite carbonate of Holocene age from these caves is shown (Dechen Cave: Niggemann, 2000, N = 50; Huettenblaeser Cave: Weber et al. (2021), aragonite stalagmite n = 300, calcite stalagmites n = 1046). In the absence of speleothem data (other than helictites) from the Windloch Cave, data from the nearby Aggertal Cave are shown for comparison (Jansen, 2006; n = 2). Host rock (limestone) data are from Niggemann et al. (2018), hydrothermal dolomite data are from Pederson et al. (2021a).
The interpretation of these geochemical patterns and the mechanisms involved must remain on the level of a working hypothesis. The main issue is that, in contrast to many studies dealing with stalagmites and flowstones, the water from which these helicitites precipitate cannot be sampled. In order to fully capture the processes involved, a dedicated cave monitoring program extending over several years is required. Judging from the data presently available, the following parameters merit consideration: Devonian host rock mineralogy and geochemistry; the mineralogy and geochemistry of the Variscan (Carboniferous) hydrothermal carbonates (often dolomite) in fault zones acting as fluid conduits; the rainwater uptake of 13 C-depleted soil zone CO 2 in the soil and host rock; prior calcite (and aragonite) precipitation; mineralogy-controlled isotope fractionation; degassing, evaporation and condensation. Moreover, the role of microbially mediated isotope fractionation might be relevant but was not studied in the context of this paper. Note, that none of the helictites studied was covered by a visible biofilm.
Concerning limestone host rock and hydrothermal carbonate isotope geochemistry, no significant differences exist (host rock = +2 to +4‰; hydrothermal dolomite = +2.5 to 4‰), while oxygen isotope values of hydrothermal phases are, as expected, lower (−10 to −7‰) compared to those of the host rock (−7 to −5.5‰; Figure 9). Hydrothermal dolomite precipitation characterises Variscan fault zones in the region (Pederson et al., 2021a), and side strands of these faults crosscut the caves studied. Evidence for a relationship between aragonitic speleothems and structural fluid pathways is that aragonitic speleothems precipitate from faults at cave ceilings ( Figure 2E). Given the rather similar isotope values between Devonian host rock and Variscan dolomites, however, the isotope geochemistry of the hydrothermal phases seems not a significant factor.
The depleted δ 13 C values of stalagmites in all three caves (−11 to −7‰) and the calcitic helictites (≈ −11‰), on the one hand, and the less-depleted δ 13 C values of aragonitic helictites, on the other hand (−5.5 to −0.5‰, Figure 9), are best explained in the context of rainwater percolating through the soil zone where it is in contact with isotopically light CO 2 (−23‰, Cerling et al., 1991).
Typically, speleothem δ 13 C values fall on a mixing line between the soil zone and the host rock carbon isotope end-members. The less-depleted carbon isotope values of aragonitic helictites suggest that isotopically less depleted fluids, percolating through hydrothermal carbonates in fault zones ( Figure 2E), mix with rainfall. Clearly, this model requires verification by direct isotope measurements of hydrothermal fluids from fault zones.
In contrast, the difference in carbon isotope values between aragonitic and calcitic helictites (Figure 9, Δ 13 C ≈ 7‰) cannot, at least not at this magnitude, be explained in the context of mineralogy-related fractionation differences between fluid and calcite and aragonite respectively. Typically, the δ 13 C value of aragonite is 1-2‰ more positive than co-occurring calcite (Romanek et al., 1992). Similarly, the δ 18 O of aragonite is about 1‰ more positive than co-occurring calcite precipitated at the same temperature (Tarutani et al., 1969). Gonzalez and Lohmann (1988) report on aragonitic and calcitic speleothems from Carlsbad Caverns in New Mexico and find that in all cases, the most positive values correspond to aragonite. Lateral calcite-to-aragonite transitions in Moroccan stalagmites are characterised by aragonite being enriched in 13 C by 3‰, and 18 O-enrichment is on the order of 1-1.5‰ relative to calcite (Wassenburg et al., 2012). Similarly, aragonitic and calcitic speleothems in the Grotte de Clamouse, France (Frisia et al., 2002), display a Δ 13 C Ar-Cc of between 3 and 6‰ and a Δ 18 O Ar-Cc of up to 2.5‰ with the aragonitic speleothems consistently being enriched in the heavy isotopes. Concluding, mineralogyrelated fractionation should not be ignored but is perhaps best seen as a second-order factor, riding on top of a dominant fluid-controlled isotope signal.
The fact that localised aragonitic speleothems form in a cave, otherwise decorated by calcitic speleothems suggests that locally, fluid Mg:Ca ratios deviated from the bulk water in the cave. Pond waters in all three caves, representing the bulk cave-water hydrogeochemistry, yield Mg 2+ /Ca 2+ ratios (mol/mol) between 0.09 and 0.21 ( Figure 10; Table S3), that is, a value that is not typical of a fluid from which aragonite is precipitated. Gonzalez and Lohmann (1988) suggest that only when Mg 2+ /Ca 2+ ratios exceed 2.5, and at higher CO 3 2− concentrations, aragonite becomes the dominant speleothem polymorph. Németh et al. (2018) propose that abiotic aragonite precipitation is favoured in waters with a Mg 2+ /Ca 2+ ratio of >1.5, (pH >8.2) and a saturation index of calcite <0.8. Riechelmann et al. (2014) found that aragonite precipitation is favoured when Mg/Ca ratios are ≥0.5, the pH is >8.2 (equivalent to higher CO 3 2− ), and SI calcite <0.8. Two factors might explain locally different Mg:Ca fluid ratios that then result in the precipitation of aragonitic cave carbonates: (i) the dolostone host mineralogy along fault zones and (ii) prior calcite/aragonite precipitation (PCP/ PAP; Wassenburg et al., 2020). The latter process results in CaCO 3 precipitation before reaching the helictite, resulting in increasingly higher Mg 2+ concentrations along the fluid path. Moreover, prior calcite/aragonite precipitation within the karst aquifer and inside the cave leads to further depletion in 12 C of water reaching the cave. In the neighbouring B7 and Bunker caves, drip water δ 13 C-DIC values display a high variability (−5.6 to −12.8‰;Niggemann, 2000;Riechelmann et al., 2011). These values are best explained by the set of parameters discussed here.
Degassing, evaporation and condensation are the remaining processes that require attention. Progressive degassing of CO 2 from the very thin fluid films on speleothems has been discussed, modelled and simulated in laboratory experiments (Dreybrodt & Scholz, 2011;Hansen et al., 2019;Sade et al., 2022;Scholz et al., 2009). In general, in the case of very thin water films, degassing of CO 2 is always fast and much shorter than the time required for equilibration of the dissolved carbon species and precipitation of CaCO 3 (Hansen et al., 2013) and would not affect isotope values in a significant manner. Degassing at the helictite tip is certainly relevant where these features protrude directly from cave walls, and particularly so in the case of the less-permeable calcitic helictites. With reference to helictites growing from the side of open soda straw speleothems, and in the case of permeable aragonitic helictites, much of the fluid degassing does probably take place before reaching the helictite tip.
Concerning the oxygen isotope values reported here, most helictites overlap with calcitic cave carbonates except for one set of samples showing 18 O-enriched values ( Figure 9). It seems possible that these values are best explained by evaporation processes. The rate and magnitude of evaporation depend on the cave studied. The Windloch Cave is characterised by vigorous air circulation (Windloch literally means 'cavern of wind'), and the abundance of gypsum precipitates in the cave point to significant evaporation. The other two caves studied here display weaker air circulation, and evaporation is probably less significant. Condensation may, to some degree, explain the lateral thickening of helictites in caves with weaker air circulation (Klutert and Huettenblaeser caves) and during times characterised by high precipitation amounts and infiltration (winter months).

| Helictite fabrics
A series of previous workers explored the fabrics of various polymorphs of CaCO 3 in stalagmites and flowstones from the viewpoint of carbonate petrology and crystallography (Frisia, 2015;Frisia et al., 2000;Onuk et al., 2014;Reinhold, 1998). Other studies used field (cave) monitoring (Baker et al., 2014;Riechelmann et al., 2014) or laboratory precipitation experiments (Day & Henderson, 2011Hansen et al., 2017;Lü et al., 2019) to explore the relationship between drip-site characteristics, fluid hydrochemistry, the significance of microbial activity and cave carbonate geochemistry and fabrics. Kinetically mediated processes best describe the relationship between fluid chemistry, crystal habit, composition and mineralogy. Gonzalez et al. (1992) suggested that spelean calcite crystal habit is mainly controlled by the supersaturation of the dripping and flowing water. Subsequent workers highlighted that other factors such as the formation of disequilibrium fabrics related to the impact of the dripping water on the stalagmite tip, kinetic processes, inhibitors and foreign ions built into the crystal structure affect crystal morphology (see Riechelmann et al., 2014 for discussion). Most studies dealing with speleothem petrography, however, focussed on calcitic fabrics (Andrieux, 1965;Gonzalez et al., 1992;Onuk et al., 2014). Aragonitic fabrics in speleothems have been less frequently explored (Frisia et al., 2002;Hopley et al., 2009) and refer to, for example, frostwork, that is, aragonite needle clusters radiating away from growth centres and to stalagmite fabrics (Frisia & Borsato, 2010;Wassenburg et al., 2016). Richter and Neuser (2007) studied calcitic helictites from the Breitscheid-Erdbach Cave, situated about 60 km to the south of the Windloch Cave. These authors provide EBSD mapping and described mono-crystalline, but also polycrystalline radial calcite fabrics in helictites and soda straw speleothems. Thin section imaging (under crossed polarisers) of calcitic helictites from Dechen Cave displays extinction patterns indicative of a composite crystal texture (also termed 'aggregate crystal' depending on the author), built by numerous elongated (fibrous) mesocrystals (see Jones (2017) for terminology; Figure 7C,D). The EBSD maps taken perpendicular (transverse section) to the helictite growth axis shown in Richter and Neuser (2007) share important similarities with those measured from samples in the Dechen Cave. The EBSD mapping points to a complex three-dimensional orientation of fibrous mesocrystals deviating away from the main helictite growth axis ( Figure 7D). Small trigonal prismatic calcite crystals are typically observed in transverse sections ( Figure 7B). All of these features are best explained in the context of disequilibrium growth fabrics, display a significant level of complexity and merit future research.
With reference to the helictite aragonitic crystal habits described here, it seems of interest that morphologically very similar (Figure 12), albeit marine aragonite crystals have been reported from (sub)recent concretionary carbonates in the lagoon of Abu Dhabi (Ge et al., 2020). Marine aragonite cement precipitated from saline porewaters (seawater pH 7.7-8.5; salinity 33-100 ppt, Mg/ Ca ratios between 4.7 and 5.2; Pederson et al., 2021b), about 5-10 cm beneath the present-day seafloor (underneath the redox boundary). The precipitation of these features was probably mediated by bacterial sulphate reduction. Along seaward-to-landward transects, acicular aragonites precipitate under the least, and platy (prismatic single crystals) under the most restricted conditions. It is at present unclear to which degree helictite aragonite crystal habits reflect subtle differences in nucleation and precipitation kinetics that could be interpreted in terms of environmental and hydrogeochemical controls similar to their marine counterparts. Therefore, the discussion of these features must presently focus on crystallographic and petrographic aspects.
The aragonite fabrics building helictites in the Windloch and Huettenblaeser caves are intriguing for several reasons. While some of the aragonites are acicular, others are board, platy, rod and lath-shaped and display flat or pseudo-prismatic terminations. The rate of crystal growth at individual crystal surfaces controls the crystal habits. Needle (acicular) aragonites are typified by growth perpendicular to the {001} surface, with {110} surfaces being poorly developed and are typical of high growth rates (Given & Wilkinson, 1985). In the case of needle growth, it is generally assumed that various inhibitors affected {110} surfaces. When needles display increasing degrees of flattening at their tips, fibres with flat (blunt) terminations result. Platy (board-shaped) crystals are best described as prismatic single crystals that lack twinning and display well-developed orthorhombic morphologies ( Figure 5D). Some crystals display swallow-tail-like terminations (Figure 12), taken as evidence for crystal twinning (on {110}), most likely growth-twinning. Others show irregular splitting features, perhaps indicating multiple twinning. The issue of growth-twins in aragonites has been detailed in Németh et al. (2018). In the view of these authors, crystal twinning of orthorhombic aragonite is typical for a monoclinic precursor phase (mAra for 'monoclinic aragonite'). While mAra is crystallographically related to aragonite, it is also associated with hydromagnesite [Mg 5 (CO 3 ) 4 (OH) 2 . 4H 2 O] and magnesite (MgCO 3 ) and later transforms to aragonite (Makovicky, 2012;Németh et al., 2018). If applicable to the case examples studied here, complex nucleation and precipitation pathways typify these speleothems. The possible relationship of the precursor mAra fabrics to Mg carbonates agrees with fluid pathways within hydrothermal fault zone dolomites resulting in high cave water Mg 2+ /Ca 2+ values.
Concluding, the nucleation history, crystallography and the crystal habits of helictites (calcite and aragonite) are complex. Calcitic helictites display a composite crystal fabric built by fibrous calcite mesocrystals. Aragonitic helictites reveal an array of different crystal habits reaching from acicular to well-developed prismatic (board-shaped) morphologies. Evidence for growth-twinning might suggest that monoclinic precursor phases were involved that were initially associated with Mg carbonates but then stabilised to orthorhombic aragonite. The fact that vadosemeteoric aragonite crystals in caves and phreatic-marine aragonite crystals in early diagenetic marine concretions share remarkable morphological similarities points to a first-order control that overrides the differences in precipitation environments and fluid geochemistry.

| Controls on helictite growth and morphology
Reviews of the various formation models of helictites and related fabrics can be found in Hill and Forti (1997), Reinhold (1998) and Onuk et al. (2014). Broadly speaking, these models fall into the following four categories: (i) formation by aerosols, (ii) capillary flow and hydrostatic pressure, (iii) microbially induced precipitation and (iv) others (e.g. changes in piezoelectric pressure potential and osmosis). Speleothems formed from aerosols are not discussed here (Dredge et al., 2013), these features lack a central capillary and are not assigned as helictites (Self & Hill, 2003), and processes such as osmosis cannot explain the features documented here. A conceptual graphic summary of different formation mechanisms as presented in the literature is shown in Figure 13.
By growing artificial, helictite-like features of sodium thiosulfate, Huff (1940) proposed that hydrostatic pressure feeding capillary flow was a relevant process. The theory, particularly the physics and thermodynamics of capillary fluid flow, is reasonably well understood (see Morrow, 1970 for a detailed discussion). Applying these principles to geosciences is complicated by the capillaries' (often) complex geometry and surface properties. Capillary flow can be defined as a fluid movement within the space of a porous material due to adhesion, cohesion and surface tension. Capillary flow allows fluid to move without the assistance of external forces or even against the gravitational pull. If capillary flow through these speleothems takes place, likely supported by hydrostatic pressure when helictites protrude directly from cave walls, then it is likely affected by among other factors (i) the rugged surface of the capillaries' inner walls (Figure 11), (ii) the length of the capillary and its degree of sinuosity (Figures 3 and  6), (iii) the localised narrowing of the capillary diameter ( Figure 6), (iv) spatial differences in the wettability of the carbonate surface, (v) a combination, at least in the case of aragonitic helictites, of channelised flow through the capillary and flow through the crystal meshwork. At present, it seems not clear if helictites, particularly aragonitic ones, are understood from the viewpoint of capillary flow thermodynamics.
These considerations align with the criticism brought forward by Kempe and Spaeth (1977). These authors pointed out that the internal friction resulting from the very irregular capillary walls ( Figure 11) would severely limit capillary pressure. This significantly limits the volume of water forced out at the helictite tip, even when suction due to evaporation at the helictite tip might play a role. Moreover, some helictites grow out of sidewalls of active soda stalactites, and no hydrostatic pressure can build under these conditions. Kempe and Spaeth (1977) suggested that, at least in the case of calcitic helictites, the pearl-string morphology of the central channel ( Figure 4E) reflects a periodic (seasonal) pattern, with smaller ends related to winter and broader ends to springsummer periods. Based on this concept, they calculated growth rates of between 0.02 and 0.04 mm/year for calcitic helictites in Winterberg Cave. If these estimates of precipitation rates hold, then flow rates required for transporting the dissolved species to the site of nucleation and precipitation (several metres per second) through the capillary seem unrealistic.
One important aspect that merits attention is that helictites grow along their long axes and show evidence for thickening, considered evidence for lateral growth (see discussion in Richter & Neuser, 2007). This agrees with Goebel and Reinboth (1972) documenting helictite growth zones. The tip is extended by the deposition of calcium carbonate around the central pore as moisture evaporates and degassing takes place. The rate of evaporation is different in each cave studied but is certainly relevant in the case of the Windloch Cave ('chimney circulation', see discussion of cave ventilation processes in Riechelmann et al., 2019). In the case of aragonitic helictites documented here, the porous meshwork of needles and platy crystals ( Figure 5B) and the presence of second-and thirdorder channels ( Figure 6) might allow for water in the central capillary to also migrate sideward into the helictite and to reach the helictite surface. There, new aragonite crystals are precipitated and induce thickening. Most of the second-order capillaries do not seem to reach the helictite surface but induce a surficial protrusion ( Figure 6). Possible evidence for water percolating through the crystal meshwork is also found in the presence of subrecent (white) aragonite needles growing on the surface of Weichselian (beige) helictites ( Figure 4C).
Concerning calcitic helictites, characterised by a significantly less permeable crystal fabric ( Figure 4E), Richter and Neuser (2007) argued that two basic types are found: one with smooth surfaces and a second type with evidence for crystal growth (rhombohedra) at the helictite surface (the examples described here fall under this category; Figure 5E,F). Richter and Neuser (2007) claim that helictites with smooth surfaces are best explained by precipitation (thickening) from thin water films covering the speleothem surface and perhaps microbially induced carbonate precipitation. In contrast, according to Richter and Neuser (2007), helictites with irregular surfaces point to fluid flow through the capillaries. This model disagrees with that of Andrieux (1965), suggesting that length growth is, in all cases, related to capillary flow and generally thickening is through surface water films.
In the author's view, it seems possible that length growth and thickening are independent, perhaps seasonally controlled processes. During times of high precipitation rates and infiltration into the cave (winter months, Riechelmann et al., 2017), helictite surfaces are possibly covered by thin water films (also condensation), allowing for thickening through precipitation. Similarly, enhanced hydrostatic capillary pressure and flow through the central channel system allows for length growth. During times of reduced precipitation and infiltration, helictite surfaces fall dry and slow growth at the helictite tip takes place from small volumes of water transported through the central channel. This model and, generally, the principles of capillary fluid flow applied to these speleothems require more work.
Another aspect that merits attention is microbially mediated carbonate precipitation and perhaps related isotope fractionation. The issue of microbially induced speleothem formation is extensively discussed in the literature (Banks et al., 2010;Canaveras et al., 2001;Northup & Lavoie, 2001). With regard to helictite fabrics, one of the most specific studies is that of Tisato et al. (2015). This paper deals with calcitic helictite bouquets in the Asperge Cave, France. These features share similarities with the helictites documented from Windloch Cave (Figure 2A) in terms of size and morphological complexity. In the view of these authors, calcitic helictite growth of what is dubbed 'Blue Gallery speleothems' is associated with prokaryotic (gliding) biofilms and microbially induced calcitic helictites coexist with 'common' aragonitic helictites and other speleothems. Microbially influenced mineralisation is triggered by the biofilm that is directed by chemotaxis and which acts as the site of CaCO 3 nucleation. Tisato et al. (2015) argue that the direction of biofilm growthand hence the morphology of the speleothem-is driven by the need for the biofilm to avoid being entombed by the precipitating carbonate. Two basic models are proposed: (i) a passive, microbially influenced mineralisation process, whereby biofilms serve as nucleation media for carbonate precipitation, and biofilm growth is random or driven by chemotaxis; (ii) active control by the microbial consortia directly influencing the growth direction of the helictite in order to obtain ecological advantages. The findings presented in Tisato et al. (2015) are relevant and of general interest, but it remains at present unclear if the 'Blue Gallery speleothems', including spectacular helictite(−like?) features, are directly comparable to the case examples described in this paper. An in-depth study of the microbial consortia that are undoubtedly present in the three caves discussed here must be the focus of future work.
Various authors proposed models explaining the helictites' often erratic growth morphologies, including upward growth (Figure 13). Deposits similar to natural helictites were formed in the laboratory by allowing solutions of different salts to evaporate (Huff, 1940). In the case of these experiments, which may or may not be relevant for natural helicities, the crooked shape results from the chance orientation of crystals. Moore (1954) F I G U R E 1 2 SEM images comparing (A) (sub)recent marine-phreatic aragonite crystals from hardgrounds in the inner lagoon of Abu Dhabi (Ge et al., 2020) with (B) (sub) Recent meteoric-vadose aragonite crystals from Windloch Cave helictites (this paper). The yellow arrows point to a lath-shaped crystal termination displaying evidence for twinning (swallowtail). Lath to board-shaped crystals have flat (green arrow in A and B) to pseudo-prismatic (red arrow in B) terminations.

F I G U R E 1 3
Graphical compilation of discussed helictite formation mechanisms. Calcitic helictites are shown in brown, aragonitic helictites in green. Note, that several of the mechanisms shown here might be independent of the helictite mineralogy. (A and B) surface versus capillary flow. (A) According to Richter and Neuser (2007) calcitic helictites with smooth surfaces (ai) are best explained by precipitation (thickening) from thin water films covering the speleothem surface combined with capillary flow (length growth, see also Andrieux, 1965), and are perhaps affected by microbially induced precipitation. A second type (Aii) is characterised by an irregular surface (rhombohedra) and is controlled by capillary flow. (B) Aragonitic helictites with permeable crystal fabrics. Water is transported through the central capillary (length growth) and seeps through the crystal meshwork (thickening). (C through G) mechanisms controlling growth orientation of helictites. (C) Changes in growth direction by clogging of the central channel, for example, by clay minerals (Onuk et al., 2014). A new channel system forms and the main growth axis of the helictite changes. (D) Localised biofilm (BF) growth on the helictite surface (passive or active control to obtain ecological advantages) might affect helictite growth (Tisato et al., 2015). (E) Airflow direction in a given cave affects helictite growth (discussion and critique in Moore, 1954). (F) The central capillary supplies a locally wetted spot (WS in fi) on the helictite tip and promotes sector-limited crystal precipitation (red in Fii). Small topographic features affect the position of the wetted spot, leading to apparently random changes in the growth orientation (Self & Hill, 2003). (G) Non-deterministic changes of a complex internal system of several orders of channels controlled by impurities, petrographic and crystallographic features (Moore, 1954) results in contorted morphologies. Note, for the aragonitic helictites (Windloch and Huettenblaeser caves) discussed in this paper, the most likely controls on growth morphologies are a combination of parameters shown in F and G. for calcitic helictites (Dechen Cave), the processes discussed in A and C are reasonable explanations. explained helictite curvature by combining the effects of impurities, crystallographic-axis rotation, and stacking of wedge-shaped crystals. The concepts brought forward by Self and Hill (2003) are the ones considered most likely for the case examples studied here. In the view of Self and Hill (2003), the central capillary supplies a locally wetted spot on the helictite's tip, where there is competition for solutes between sectors. Local variations in surface topography, resulting from the extremely irregular morphology of the aragonitic helictite tip at the tens of microns scale ( Figure 4A), affect the position of the wetted spot. These variations promote different growth rates between sectors. Differential sector precipitation rates may result in apparently erratic changes in growth direction for the aggregate as a whole (Self & Hill, 2003). These variations are unique to each helictite as quantitative assessments of growth front azimuths are random (Moore, 1954) and thus cannot be explained by specific airflow directions in a given cave. Precipitation is further affected by water percolating into lower-order capillaries and then into the aragonite crystal meshwork to eventually reach the helictite surface, inducing CaCO 3 precipitation and thickening.
Regarding the morphologically simpler calcitic helictites, Onuk et al. (2014) proposed that clay minerals eventually clogged first-order channels and triggered the formation of a lower-order side channel with a different orientation that evolves into a new, first-order channel inducing a change in growth orientation. No evidence for capillary clogging is found in the specimens studied here, but that does not imply that the model brought forward by Onuk et al. (2014) does not apply to other caves and other helictites. Onuk et al. (2014) also documented that in straight portions of calcitic helictites, crystals are equidimensional. Where the helictites change direction, crystals at the inside of the bend are significantly smaller (ca 100 μm) compared to those at the outside of the bend (ca 500 μm) to compensate for the differences in circumference. With regard to the aragonitic helictites documented here, no statistically relevant differences in crystal sizes are found where these features bend ( Figure 5A through D). Aragonite crystals consistently increase in size from the inner to the outer portions of the helictite (Figures 5B and 8C,D,E).
From the evidence gathered here and data and concepts from published work, it appears that, at present, no single helictite formation model satisfies all boundary conditions. In the case of the aragonitic helictites described here, the authors propose that (i) the sector precipitation model combined with (ii) lateral fluid migration and surficial precipitation controlled by (iii) the complex geometry of the capillary network might explain many of the features observed. With regard to the morphologically more simple and smaller calcitic helictites, (i) hydrostatic capillary combined with (ii) surface flow explain the observations best. The question is raised if the group of speleothems assigned to helictites (but to some degree also anthodites, anemolites, etc.) are closely related in terms of their formation mechanisms and growth mechanics, or rather form a variable group of genetically similar, but in some aspects dissimilar, features. Explanations for the contorted morphologies of helictites often satisfy the observations made in one cave and for one mineralogy but are not necessarily applicable to other caves and morphologies. The authors propose that the search for a unifying helictite model must be abandoned, and emphasis should be placed on recognising a series of processes interacting in a stochastic (but not non-deterministic) manner.

| CONCLUSIONS
Aragonitic and calcitic helictites from three caves in Western Germany show a wide range of morphotypes and share many attributes with those described from other caves worldwide. 230 Th/U-dating points to an age range from recent, actively growing to inactive, tan-coloured specimens as old as 347 ka. Aragonitic helictites from the Windloch Cave site form exceptionally large (1.4 m) and complex structures. When individual helictite branches reach critical lengths (often ca 40 cm), their weight overcomes the critical breaking strength, and fragments gravitationally accumulate on the cave floor.
The novelty of the study is, in particular, to give (i) a detailed and quantitative view of the (micro)structure, crystal orientation and morphology of two different helictite types and (ii) to document the complex internal geometry of the central capillaries (or channel) network. Significant differences between aragonitic and calcitic helictites are present. Calcitic helictites are built by a dense fabric of translucent (glassy) blocky crystals, most of which have more or less idiomorphic diameters of several hundred microns to several millimetres in size. These are best described as composite crystals made of smaller, fibrous mesocrystals. In the case of aragonitic helictites, the crystal morphologies are acicular (needle), fibrous, platy and lath-shaped forms with pseudo-prismatic to flat terminations. The helictite surface is covered by a fur-like array of shingled crystals deviating from the main helictite growth direction at angles of <15°. Remarkably, the meteoricvadose cave aragonite crystals share important morphological similarities with actualistic marine-phreatic aragonite crystals precipitated in the lagoon of Abu Dhabi.
All samples studied are characterised by one to three hierarchical orders of central capillaries. Microtomography imaging confirms that the geometry of the channels within aragonitic helictites is consistently irregular. These capillaries are commonly 150-200 μm in diameter (second-and third-order capillaries narrow to less than 10 μm at their end) and round to irregular in shape. First-order capillaries reach the helictite surface. In calcitic helictites, the channel architecture is that of a first-order capillary, commonly between 100 and 200 μm in diameter, with rugged walls resembling a pearl-string morphology.
The less 13 C-depleted values of all aragonitic helictites contrast their more 13 C-depleted calcitic counterparts. The processes involved are best explained in the context of two different fluids (precipitation infiltration vs. hydrothermal fluids circulating through faults) and their corresponding hydrogeochemistry. The helictite geochemistry is further affected by mineralogy-dependent fractionation, fluid Mg:Ca ratios and prior calcite precipitation. Oxygen isotope values correspond to calcitic and aragonitic speleothems in the region. Higher δ 18 O values are assigned to significant evaporation in the case of the Windloch Cave.
In the case of aragonitic helictites, it seems likely, that water from the central capillary supplies a localised, wetted spot on the helictite's tip, which is then the site of crystal growth. Variations in the position of the wetted spot over time promote different growth rates and directions between different sectors. Water percolates from the capillaries network into the aragonite crystal meshwork, particularly at the end of second-and third-order capillaries. Where water reaches the surface, localised CaCO 3 precipitation takes place, leading to the formation of protrusions and thickening of the helictite. Calcitic helictites display irregular surfaces with crystal rhombohedra, perhaps arguing for seasonal controlled thickening and lengthgrowth patterns related to precipitation amount and infiltration rates. Calcitic helicities are morphologically less complex than aragonitic ones, at least in the case examples studied here.
Helictite mineralogy, fabric and morphology are likely controlled by a parameter set characterised by tipping points that are highly sensitive to even minor environmental and crystallographic changes. Each cave, and even different sites within a single cave, may be characterised by a unique combination of parameters that change over a seasonal (and longer) cycle. A single helictite formation model seems unlikely to satisfy all of these boundary conditions.