Hydrochemical mixing‐zones trigger dolomite formation in an alkaline lake

Dolomite is globally present in past geological records, but rare in modern environments. The mechanisms favouring its precipitation under ambient conditions remain highly debated. This study investigates sediments, containing high concentrations of early diagenetic calcian dolomite, from alkaline Lake Van (Republic of Türkiye, formally Turkey) dating back to 252 ka bp. Powder X‐ray diffraction and scanning electron microscopy evidence suggests that dolomite formation is associated with prior dissolution of aragonite and low‐Mg calcite and a subsequent co‐precipitation with, and/or partial transformation of, high‐Mg calcite into dolomite. The infrequent presence of diatom frustules encapsulated by dolomite suggests, for Lake Van, unusually low pore‐water pH at the time of dolomite formation. Conditions facilitating the preservation of silica, as well as dissolution and subsequent reprecipitation of carbonate phases, could result from periodic reventilations of Lake Van's deep water and an advection of pore fluids with contrasting redox potential and chemical concentration gradients. This continental analogue of the coastal ‘mixing‐zone’ dolomitization model argues that not overcoming a single specific hydrochemical threshold, but highly dynamic and fluctuating conditions trigger dolomite formation in Lake Van.


INTRODUCTION
Many modern dolomite-forming environments are evaporitic in nature (e.g.Brauchli et al., 2016) and consequently the occurrence of dolomite in sedimentary profiles is commonly interpreted as a palaeoenvironmental indicator for a 'negative' water balance.For example, lacustrine dolomite formation is often related to increased evaporation or reduced meteoric water influx, and a falling lake level (e.g.Kelts & Hs€ u, 1978;Degens et al., 1984;Talbot & Kelts, 1986;De Deckker & Last, 1988;Last, 1990;Landmann et al., 1996;Sinha & Smykatz-Kloss, 2003;C ß a gatay et al., 2014;Dean et al., 2015).Evaporitic environments characterized by comparatively high (surface) water temperature, increased alkalinity (and pH), high Mg/ Ca ratios and high salinity may promote the incorporation of Mg into carbonate minerals (Al Disi et al., 2019).Consequently, in many models developed to explain modern and ancient dolomite formation, evaporitic conditions are one of the key environmental components (reviewed by Land, 1985;Tucker & Wright, 1990;Warren, 2000;Machel, 2004).Yet, sedimentological evidence and independent proxies do not always corroborate intuitively suggested evaporative conditions for marine nor lacustrine dolomite formation (e.g.McCormack et al., 2018;Shalev et al., 2019;Reuning et al., 2022).
Dolomite, CaMg(CO 3 ) 2 is a common authigenic mineral in the Phanerozoic rock record but, to date, no recipe or procedure enables precipitating it easily in a laboratory experiment conducted at low temperature.Addressing the 'dolomite problem', research has focused on identifying the key factors that allow for overcoming the kinetic barriers which prevent the incorporation of magnesium ions (Mg 2+ ) into calcium carbonate minerals (CaCO 3 ) to form (proto)dolomite (e.g.Arvidson & Mackenzie, 1999;Petrash et al., 2017;Bontognali, 2019;Kell-Duivestein et al., 2019;Baldermann et al., 2020;Pina et al., 2022).In particular, the presence of microbes (e.g.Vasconcelos et al., 1995;Bontognali et al., 2014;Daye et al., 2019) and their secretions, extracellular polymeric substances (EPS), may play a crucial role for the nucleation of Mg-rich carbonates (Bontognali et al., 2010;Zhang et al., 2012;Roberts et al., 2013;Al Disi et al., 2019).Further, abiotic factors, including the presence of negatively charged clay mineral surfaces (Liu et al., 2019), dissolved silica (Fang & Hu, 2022), manganese (Mn 2+ ) and zinc (Zn 2+ ) ions (Petrash et al., 2015;Daye et al., 2019;Vandeginste et al., 2019), and sulphide (S 2À ) ions (Zhang et al., 2012) catalyse the formation of Mg-rich carbonate phases at low temperature.These phases are possible precursors to ordered dolomite; thus, more than one solution to the 'dolomite problem' appears possible.It is likely that not just one, but a combination of catalytic factors determines whether dolomite or another mineral phase will precipitate in a specific environment (Al Disi et al., 2021).
In addition to the investigation of the abovementioned catalytic factors, research on dolomite also focuses on identifying hydrogeological settings that simultaneously favour dolomite oversaturation and calcite/aragonite undersaturation.A notable example is the mixing-zone dolomitization model, widely popular in the 1980s, which explains the dolomitization of subtidal marine carbonates in shallow burial settings by mixing meteoric freshwater and seawater (Hanshaw et al., 1971;Badiozamani, 1973;Land, 1973;Shatkay & Magaritz, 1987).This model relies on differences in factors like ionic strength, ΣCO 2 , pCO 2 and ionic activities to create conditions predicting the dissolution and dolomitization of preexisting calcium carbonates in the mixing zones of coastal aquifers.Over many years, this model has received strong criticism, both on a theoretical level and also because it was not confirmed through the study of several modern mixing zones (e.g.Hardie, 1987;Land, 1998).Nevertheless, the model was recently re-evaluated integrating new insight on the importance of biogeochemical reactions, associated redox conditions, and mobilization of ions (for example, Mn and Zn) that may play a key role for the formation of dolomite (Petrash et al., 2021).Therefore, mixing zones in coastal marine regions, but potentially also in lacustrine environments, remain interesting sites for investigating the formation of dolomite.
Authigenic dolomite is a minor component in Holocene and Pleistocene sediments of several primarily saline and hypersaline lakes (Last, 1990), but the formation of purely lacustrine dolomite is given less attention compared to its marine counterpart.Within lakes, dolomite can form either within the water column (De Deckker & Last, 1988;Rosen et al., 1988;Coshell et al., 1998;Del Cura et al., 2001;Meister et al., 2011;Murphy et al., 2014;Fussmann et al., 2020) or within the sediment as a primary pore infill or as a secondary precipitate replacing CaCO 3 minerals (Talbot & Kelts, 1986;Rosen & Coshell, 1992).Independent of the mode of precipitation and type of marine or lacustrine setting, dolomite formation may occur stratigraphically discontinuously, i.e. episodically (Meister et al., 2006(Meister et al., , 2008;;McCormack et al., 2018), implying that environmental conditions may change from favouring to hindering dolomite precipitation and vice versa.
The world's largest alkaline lake, Lake Van, is supersaturated with respect to dolomite (SI Dolomite = 3.85 to 4.10; Reimer et al., 2009).Virtually no dolomite precipitates within the water column in the lake today, but high concentrations of authigenic calcian dolomite occur cyclically within finely laminated sediments (with the exception of the Holocene).In Lake Van the timing of these dolomite-rich intervals coincides with periods of rapid Northern Hemisphere temperature increase dubbed Dansgaard-Oeschger cycles (McCormack et al., 2018), and the finely laminated sediments represent humid conditions and a high lake-level (Stockhecke et al., 2014a).These observations clearly contrast previous interpretations of dolomite in Lake Van as an indicator for lake-level lowstands or even lake desiccation (Khoo et al., 1978;Landmann et al., 1996;C ß a gatay et al., 2014) and argue against common models suggesting the necessity of evaporative conditions for dolomite formation.Early diagenetic, deep-water and low-temperature dolomite cyclically occurring in Lake Van sediments testifies to our still deficient understanding of this mineral's formation processes.
Taking advantage of previously published independent Lake Van environmental proxy records, McCormack et al. (2018) demonstrated that dolomite precipitated under cold water conditions during the last 150 ka BP in Lake Van.In addition, and contrary to most dolomitization models, precipitation was shown not to be favoured by an increase in either water salinity, Mg/Ca ratios, pH, alkalinity or sedimentary total organic carbon (TOC).The authors thus proposed that the early diagenetic origin of the dolomite in Lake Van is related to abrupt fluctuations in pore water chemistry and the re-ventilation of the sediment-water interface, typically following a lake-level highstand.It was further hypothesized that these fluctuating conditions at/or close to the sedimentwater interface may have led to microbial stress stimulating EPS production and/or cell mortality and accompanied by an increase in carboxyl groups which effectively catalysed dolomite via complexing and dewatering Mg ions (Bontognali et al., 2010(Bontognali et al., , 2014;;Roberts et al., 2013).
This study investigates, in more detail and with recognition of the environmental and climatic context, how the enigmatic lacustrine dolomite forms during early diagenesis in Lake Van.Building on previous Lake Van dolomite investigations (Landmann et al., 1996;C ß a gatay et al., 2014) and with respect to the dataset already published in McCormack et al. (2018), this record is extended and investigates older sediments now covering three Glacial-Interglacial transitions (ca 250 kyr).A longer timeframe allows for evaluating the impact of Late Pleistocene-Holocene climate change and coincident lake hydrological variability on dolomite formation.Quantitative powder X-ray diffraction (XRD) analyses (Rietveld approach) allows to investigate the role of CaCO 3 phases and high-Mg calcite (HMC) as potential co-precipitate and/or precursor phases in the formation of calcian dolomite.Further, this study explores the mode and timing of dolomite precipitation using XRD, scanning electron microscopy (SEM) imaging and in relation to published independent environmental proxies.Taking advantage of recent carbon and oxygen isotope (d 13 C and d 18 O) studies performed on several other carbonate fractions of Lake Van (McCormack & Kwiecien, 2021) allows to further environmentally contextualize the isotopic composition of dolomite; both published (McCormack et al., 2018) as well as novel dolomite isotope data for the penultimate interglacial presented herein.This study then infers the physicochemical conditions at the sediment-water interface leading to dolomite precipitation in Lake Van.Finally, these results contribute to the wider discussion on newer reinterpretations of the classical 'mixingzone' model (Badiozamani, 1973) with recent reevaluations of, and additions to, the model including a biogeochemical component (e.g.Diloreto et al., 2021;Petrash et al., 2021).These newer contributions to the 'dolomite problem' emphasize the role of dynamic, fluctuating conditions in dolomite precipitation, in the context of marine dolomite formation, which is extended here to also include dolomites precipitating in lacustrine deep water environments.

Material
The sediments from Lake Van (Republic of T€ urkiye, formally Turkey) studied here were recovered in 2010 in the frame of the International Continental Scientific Drilling Program (ICDP) PALEOVAN project (Litt et al., 2012;Litt & Anselmetti, 2014).Ahlat Ridge (AR) composite profile and off-section material were sampled.The samples were collected with a single sample resolution of 2 cm.Event deposits, such as tephra and graded layers, were avoided during sampling.Composite profile depth and age (after Stockhecke et al., 2014b) were assigned to off-section samples by visual correlation based on high-resolution photographs.The lithological description of the composite profile is based on the lithotypes and their genetic interpretation published in Stockhecke et al. (2014a).The data presented here represent the uppermost 67 m of the AR composite profile, corresponding to an age of up to ca 147 ka BP and an interval between 102 m and 119 m, corresponding to an age between 210 and 252 ka BP (Marine Isotope Stages -MIS -7 to 8).

Methods
All samples were wet-sieved with distilled water through a succession of sieves (>250 lm, >125 lm, >63 lm).To avoid significant contributions of large detrital minerals and biogenic carbonates, powder X-ray diffraction analysis (P-XRD) was performed on fine fraction material (<63 lm) which was collected on filter paper, air-dried at room temperature and subsequently ground in an agate mortar prior to analysis.
Powder X-ray diffraction analyses were performed on a Bragg-Brentano diffractometer (PANalytical's Empyrean; Malvern Panalytical, Malvern, UK) equipped with a PIXcel1D detector.Samples were analysed using Cu Ka radiation, applying a tube voltage and current of 45 kV and 40 mA, respectively.The samples were X-rayed in the range from 4 to 65°2h with a step size of 0.0131°and a counting time of 3 s per step.Furthermore, operating conditions included fixed 0.25°divergent and 0.5°antiscatter slits in the incident beam path, incident and diffracted beam 0.04 rad soller slits and a 7.5 mm high antiscatter slit together with a Ni Filter in the diffracted beam optics.
To quantify absolute concentrations of the different carbonate phases in the <63 lm size fraction, 247 P-XRD profiles were analysed using Rietveld refinement.In addition, previously reported P-XRD profiles analysed semi-quantitatively based on mineral Relative Intensity Ratios (RIR) for the last 147 ka BP (McCormack et al., 2018(McCormack et al., , 2019a) ) were reexamined.Therefore, a least squares approach (i.e. the Rietveld approach) was applied to refine a calculated line profile until it matched the measured P-XRD profiles by considering the crystal structure, specimen and instrumental parameters (i.e. unit cell and atomic parameters, mineral crystallinity, particle size, stress and strain, specimen porosity and absorption properties, preferred particle orientation, as well as geometry, configuration and beam geometry of the XRD device).Mineral identification and quantification were carried out using the PANalytical HighScore Plus software package and pdf-4 database at an analytical uncertainty of <3 wt% (Baldermann et al., 2021).
For all samples from MIS 7 to 8 with a high dolomite content and lacking significant interference from other mineral reflections, the degree of dolomite cation ordering was determined from the intensity ratio of the 015 to the 110 peak (I [015]/I [110]) (Goldsmith & Graf, 1958;Hardy & Tucker, 1988; Fig. S1).Further, the equation of Lumsden (1979) was used to semi-quantitatively estimate the CaCO 3 (and MgCO 3 ) contents in dolomite (near-stoichiometric and calcian), low-Mg calcite (LMC) and HMC in mole percentage (mol%) in a suite of samples via the expression: NCaCO 3 = 333.33Ád(104)carbonate -911.99,where NCaCO 3 is the mol% CaCO 3 content in the respective carbonate phase.Quartz was used as an internal standard to correct the position of the dolomite d( 104) reflection.The full width half maximum (FWHM) of the calcian dolomite and HMC minerals was measured using their corresponding d( 104) diffraction peaks for samples with a calcian dolomite content of ≥10 wt%.
Carbon and oxygen isotope analyses were performed in continuous flow mode on a GasBench II coupled to a ThermoFinnigan MAT 253 mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) at the Ruhr-University Bochum following standard procedures (Breitenbach & Bernasconi, 2011).Published dolomite d 18 O and d 13 C values from the last 147 kyr also reported herein were isolated from CaCO 3 phases using 0.27 M disodium ethylenediaminetetraacetic acid (EDTA, McCormack et al., 2018).Because all new samples, from the penultimate interglacial, analysed herein (n = 6) had a dolomite content of >90% relative to aragonite and LMC, EDTA pretreatment was deemed unnecessary with negligible impact of CaCO 3 impurities on the final isotope values.Sample powders of 130 to 240 lg were weighted into borosilicate glass vials and oven-dried at 104°C overnight.Samples were run at 70°C for 2 h, together with international standards NBS19, IAEA603, CO8 and ETH-1 (ISO-A; Meckler et al., 2014).All results are normalized against an in-house prepared dolomite standard (Fra-DOL, grain size <100 lm, d 13 C = 2.74 AE 0.05&, d 18 O = À2.87AE 0.1& Vienna PeeDee Belemnite, V-PDB), which has been normalized to the dolomite standard (Sp€ otl & Vennemann, 2003;M€ uller et al., 2004) prepared by Thorsten Vennemann.The external standard deviation for oxygen and carbon isotope analysis is <0.07&.
Scanning electron microscopy was performed on unsieved, oven dried (50°C) bulk samples, and on the <250 lm sieved fractions.Secondary Electron Images (SEI) were taken from goldsputtered samples using a Gemini 2-Merlin microscope (Carl Zeiss AG, Oberkochen, Germany) operated with an acceleration voltage of 20 kV in high-vacuum mode.
The Principal Component Analysis (PCA) has been conducted using the package FactoMineR in R. The following variables were used: P-XRD mineral content of Ca-dolomite, HMC, nearstoichiometric dolomite, aragonite, LMC, silicate minerals (this study); linearly interpolated XRFdetermined bulk intensities of major elements Si, Al, Ca, K, Ti, Fe, Mn, Cl and element ratios Si/K, K/Ca, Ca/K, Ca/Fe, Si/Al, Mn/Fe, as well as TOC and carbonate content (from Kwiecien et al., 2014, andStockhecke et al., 2014a).The proportion of variances retained by the principal components was determined based on eigenvalues and the PCA was performed using five principal components, which explain 80% of the variance (Fig. S2).The results were visualized using a variable factor map (Fig. S3).

RESULTS
All P-XRD and isotope data is given in Data S1.Five different carbonate phases were identified in the P-XRD patterns of the <63 lm fraction using Rietveld refinement.Three phases show widely varying total concentrations: aragonite (0 to 68.9 wt.%), LMC (0 to 23.5 wt.%) and calcian dolomite (0.3 to 81.5 wt.%).Two other phases stay at background levels: HMC (0 to 5.7 wt.%) and a near-stoichiometric dolomite (0.2 to 6 wt.%).
In terms of stratigraphic occurrence and relative abundance, the calcian dolomite contents reported here are directly comparable to the concentrations reported previously using mineral RIRs for the last 147 ka BP (McCormack et al., 2018).The XRD data from the penultimate interglacial (MIS 7c and MIS 7e) document the highest dolomite concentrations in Lake Van sediments yet (up to 81.5 wt.%), even exceeding the high concentrations of the last Interglacial (up to 62.8 wt.%, Fig. 1, Data S1).
Both d 18 O and d 13 C values of calcian dolomite collected from the penultimate interglacial are within the typical range of values reported previously for the last 147 ka BP (McCormack et al., 2018).Dolomite d 18 O values for MIS 7 range between +6.5 and +7.5& (n = 6, Fig. 1F).Carbon isotope values obtained for MIS 7c (215 ka BP) are between +7.4 and +7.5&, and for MIS 7e range between +3.46 to +4.17& (n = 4, Fig. 1F), respectively.
Scanning electron microscopy imaging documents calcian dolomite crystals appearing as idiomorphic rhombohedra with stepped faces (Fig. 3).Dolomite rhombohedra are typically grown together in aggregates that can reach sizes of >125 lm (present in the 250 to 125 lm sieve fraction).Additionally, SEM imaging captured (to the authors' knowledge) a unique feature in lacustrine sediments, intertwining of dolomite crystals and diatom frustules (or their fragments).While this feature may be undocumented in lacustrine sediments, it bears resemblance to observations on marine deep-sea dolomites (Bernoulli et al., 2004;Meister et al., 2006).

DISCUSSION
The sequence of dolomite precipitation Calcian dolomite, for simplicity from here on primarily referred to as 'dolomite', forms predominantly as pore infill during early diagenesis in Lake Van's sediments (McCormack et al., 2018), without any indication of pseudomorphic replacement of any pre-existing CaCO 3 phases, such as aragonite, LMC, and possibly also HMC (Fig. 3).Still, high concentrations of dolomite typically coincide with low aragonite and LMC concentrations (Fig. 2C and D).Dolomite-rich layers often coincide with a sudden drop in the aragonite to The sudden drop in Ar/(Ar + LMC), coinciding with a high dolomite content suggests that, in Lake Van, a dissolution-reprecipitation process plays a role in the formation of dolomite.As dolomite grows in the sediment, it may take up space previously occupied by other minerals (including CaCO 3 phases) and thereby distort absolute mineral concentrations relative to periods with no dolomite precipitation.However, such a distortion of sedimentary mineral concentrations by growing dolomite would be unlikely to reduce the concentration of aragonite preferably over LMC.Therefore, the sudden drop in Ar/(Ar + LMC) ratios, coinciding with high dolomite concentrations, indicates a preferential dissolution of the thermodynamically less stable (more soluble) aragonite over the more stable (less soluble) LMC (Plummer & Busenberg, 1982).One sample with a total dolomite concentration of 81 wt.% (103.739metres composite depth below lake floor, 215 ka BP) shows neither traces of aragonite nor LMC, suggesting a complete dissolution of both prior to or contemporary with dolomite precipitation within an at least 2 cm (single sample resolution) sedimentary interval.
A positive correlation between HMC and dolomite (Fig. 2A) indicates that HMC formation is linked to that of dolomite (Fig. S3).Both of these phases may precipitate as an early diagenetic  document in alkaline Lake Neusiedl (Austria) a relatively stable ratio of HMC to protodolomite (very high-Mg calcite without dolomite XRD-ordering peaks) within short (<40 cm) core sections.Yet, to the authors' knowledge Lake Van's sediments are the first to document such a relationship between HMC and dolomite throughout a high-resolution (and large-sample size) long sedimentary profile, making them suitable material for further evaluation of early diagenetic Mg carbonate phases and their precipitation processes in a natural setting.The minor to trace concentration of nearstoichiometric dolomite shows only a weak relation to the calcian dolomite content (Fig. 2B) and may, contrary to the latter, be of a detrital origin.A possible source area for the detrital, nearstoichiometric dolomite could be the partly dolomitized Lower Miocene Adilcevaz limestone (Demirstasli & Pisoni, 1965) cropping out in the north of Lake Van (C ß a gatay et al., 2014).Another potential source area could be the dolomitic marble of the Bitlis complex located in the south of Lake Van (Oberh€ ansli et al., 2013).Notably, however, both near-stoichiometric dolomite and calcian dolomite fall into the same PCA group (Fig. S3) indicating a positive relationship.Perhaps at least some of the near-stoichiometric dolomite has already formed from the calcian dolomite within these relatively young sediments, although further research, perhaps involving even older core sections, is required to verify this.

Hydrochemical and environmental context of dolomite formation
The results herein from MIS 7 corroborate previous results, i.e. high concentrations of dolomite commonly occur in Lake Van sediments representing interglacial and interstadial lake level highstands (McCormack et al., 2018).The highest concentrations of dolomite (>81 wt.%) occur in sediments of the penultimate interglacial (MIS 7c, 7e; Fig. 1), sediments that are interpreted to have been deposited under anoxic/suboxic conditions (dubbed Lm and Ll in Stockhecke et al., 2014a).These dolomite-rich layers correspond to a climate optimum evidenced by increased vegetation cover (Pickarski & Litt, 2017) facilitating reduced terrestrial detrital mineral input (Fig. 1).Intermittent presence of diatoms in the sediments supports the notion of a lake level high enough to establish a short-lived outflow.Even a short-lived change from a closed to open hydrological system would lower the alkalinity, salinity and pH, and effectively transform the basin into a freshwater lake (Stockhecke et al., 2014a;North et al., 2017;Tomonaga et al., 2017).Exceptionally high dolomite concentrations in MIS 7 sediments are in line with the concept of dolomite forming along a physicochemical gradient between pore and lake water (McCormack et al., 2018).In this case the hypolimnion became destratified and/or reventilated, once the level of Lake Van fell after the deglaciation highstand.Dolomite is notably absent within Holocene sediments, likely due to the lake level remaining high after an initial rise (Landmann et al., 1996;C ß a gatay et al., 2014;Stockhecke et al., 2014a;Tomonaga et al., 2017) with insufficient Dolomite formation in an alkaline lake 879 physicochemical perturbation of the pore water at the Ahlat Ridge site to induce dolomite precipitation (McCormack et al., 2018).
Dolomite is present predominantly in layers corresponding to a high lake level, but throughout the stratigraphic sequence the physicochemical conditions at the time of its formation were likely variable.Indeed, PCA indicates that dolomite (and HMC) concentration is separate from environmental proxies such as sedimentary redox conditions (XRF-Mn/Fe), organic matter preservation (TOC), detrital mineral input (XRD silicate mineral content and XRF-Ca/K) and biogenic silica production versus detrital input (XRF-Si/K, Fig. S3).Therefore none of these environmental conditions, as interpreted from independent environmental proxies, seem to influence the concentration of dolomite and HMC in the sediments.Dolomite d 18 O values are in a narrow range, indicating that the precipitation took place in the cold, deep water (McCormack et al., 2018).In contrast, the variability of dolomite d 13 C values across the studied interval (À0.2 to +7.6&) argue for different dissolved inorganic carbon (DIC) sources for individual dolomite-rich layers.Other carbonate components, such as epifaunal ostracod valves and early diagenetic aragonite encrustations, forming at the sediment-water interface on decomposing biological remains, have relatively homogenous d 13 C values of À0.3 to +3.9& and +6.7 to +8.4&, respectively (McCormack et al., 2019b;McCormack & Kwiecien, 2021).A notable difference between the d 13 C values of valves from infaunal and epifaunal ostracod species is related to their microhabitat.The infaunal ostracods have lower d 13 C values due to the incorporation of isotopically light carbon sourced from organic matter oxidized in the sediment (von Grafenstein et al., 1999;McCormack et al., 2019b).Oxygenated pore water d 13 C DIC values must be lower in the sediments than those of the average lake water.
Variations in dolomite DIC sources may result from different dolomite formation depths and/or variable contributions of pore and lake water derived DIC.However, variable dolomite d 13 C values between dolomite-rich intervals may, in some cases, also document changes in lake water d 13 C DIC .Samples with mixed surface water precipitates (aragonite and LMC), representing MIS 5e (126 to 129 ka BP, a likely openlake period), have lower d 13 C values, indicating a reduced total lake DIC content and a stronger influence of lower runoff d 13 C DIC (McCormack et al., 2019a).Dolomite from MIS 5e and MIS 7e have similarly low d 13 C values (+4.1 AE 1.64&, n = 8 and +3.9 AE 0.3&, n = 4, respectively).However, dolomite found in the interstadial sediments of the last glacial also has low d 13 C values (reaching À0.24& Fig. 1F).Therefore, dolomite may have precipitated from a mixture of different DIC pools provided from lake and pore water, with the latter being modified by organic matter oxidation; yet, the extent to which each DIC source could have contributed to the dolomite d 13 C value remains to be quantified.
High concentration of dolomite appears to coincide (within a few centimetres of lithological changes and/or adjacent) with occurrences of a qualitatively determined X-ray amorphous phase in the fine-grained sediments (Fig. 1).The occurrence of this amorphous phase (Fig. 1) may be interpreted as the presence of bio-opal, likely siliceous diatom frustules (or fragments thereof), corroborating the presence of diatoms documented by North et al. (2017).Due to Lake Van's high pH value (>9) and Na-CO 3 -Cl-(SO 4 )chemistry (Reimer et al., 2009) diatom frustules are quickly dissolved within the water column and/or at the sediment-water interface (Loucaide et al., 2008;Fritz et al., 2010;North et al., 2017).Consequently, the presence of diatoms in Lake Van sediments suggests a freshening of the lake water (i.e.lowering of salinity and pH) and is an indirect indicator of lake level highstands (Stockhecke et al., 2014a;North et al., 2017).
The occurrence of dolomite enclosing diatom frustules with intricate details preserved (Fig. 3) argues for early cementation of the dolomite (e.g.Bernoulli et al., 2004) as well as against significant vertical movement of 'late' diagenetic, high pH pore fluids though the sedimentary pile, in line with suppressed pore space diffusive transport in Lake Van sediments discussed by Tomonaga et al. (2017).This also implies that the diatom frustules were encased by dolomite precipitating shortly after deposition and from a restricted early diagenetic pore water reservoir.Diatom frustules encapsulated by dolomite show varying stages of preservation, from excellent to largely dissolved (Fig. 3E to H).At least in one case, the SEM image documents dolomite formation likely prior or contemporary with the dissolution of a diatom frustule, because dolomite rhombohedra are filling the frustule cavities (Fig. 3F).This feature argues for a locally highly restricted precipitation of dolomite at a pore space scale allowing for, at the same scale, differences in bio-opal preservation.
All four periods with the highest dolomite concentration (MIS 5c, MIS 5e, MIS 7c and MIS 7e) are either synchronous with or are directly preceded or followed by a higher diatom content (i.e.detectable XRD amorphous phase).Accordingly, the highest dolomite concentrations occur in sediments deposited during periods of particularly high lake levels (Stockhecke et al., 2014a;North et al., 2017;Tomonaga et al., 2017), and precipitate as a result of perturbations in bottom water chemistry (both salinity and pH).Recently, Fang & Xu (2022) demonstrated that high concentrations of dissolved silica (Si) may promote dolomite formation in a high Mg/Ca environment, by promoting Mg incorporation into carbonate minerals and inhibiting aragonite formation.Notably, in Lake Van, high concentrations of dolomite may occur within diatom-rich or diatom-free sediments 1 and 3).The lack of diatom frustules in most of Lake Van's record indicates that their dissolution also takes place in the sediment at times when dolomite formation does not occur, or at least not in significant quantities, as is the case today (North et al., 2017).An increase in Si content alone is unlikely to be responsible for dolomite formation in Lake Van, which must have taken place over a pH and salinity range encompassing both dissolution and preservation of diatom silica.

A lacustrine 'mixing-zone' model for dolomite formation
The results herein do not disclose a clear relation between potential catalytic factors (for example, pH, temperature, salinity, silica concentration and/or high aqueous Mg/Ca ratio) and the formation of dolomite.Instead, they demonstrate ambiguity of overcoming a single specific hydrochemical threshold as a control on dolomite precipitation.This dataset provides strong support to a concept that has gained new momentum in recent years: unstable, fluctuating and highly dynamic hydrochemical conditions trigger the formation of dolomite (Deelman, 1999;McCormack et al., 2018;Fussmann et al., 2020;Diloreto et al., 2021;Petrash et al., 2021).Deelman (1999) suggested that fluctuating pH conditions may result in dolomite formation by allowing for the coprecipitation of metastable and stable phases (for example, aragonite or HMC and dolomite).Conditions opposing the subsequent growth of the metastable phase (such as aragonite undersaturation resulting from periodical mixing of pore waters with different composition that could dissolve aragonite) will favour the continued growth of the stable phase, for instance, dolomite.Deelman (1999) proposed that the required alternations in pH conditions can be found in natural settings (for example, seasonal changes in water chemistry; daily changes caused by the photosynthetic organisms; tidal flooding by water rich in CO 2 ).This hypothesis has not yet been convincingly demonstrated through the study of a modern dolomite-forming environment.However, dolomite has often been found in association with microbial mats, in which daily and seasonal pH fluctuations are caused by a shift between the microbial community structure (e.g.Bontognali et al., 2010;DiLoreto et al., 2019).Microelectrode measurements revealed higher saturation indices with respect to dolomite in association with the photooxic layers, while lower saturation combined with conditions of undersaturation with respect to calcite and aragonite were identified in some anoxic zones of the studied microbial mats (DiLoreto et al., 2019).A similar process may also be at play in the sediment of Lake Van.However, it needs to be considered that geochemical gradients due to a shift between oxic and anoxic conditions most likely do not involve photosynthetic microorganisms in this case, since dolomite formation in Lake Van occurs at depth in the lake, far from the photic zone.
Fluctuation in microbial community composition can influence dolomite formation not only due to metabolism-related shift in pH, but also due to production of EPS with a composition that favours incorporation of Mg into the carbonate minerals.The importance of fluctuating conditions was also included in a recent, revised microbial mediation model for dolomite formation: DiLoreto et al. (2021) monitored seasonal changes in the microbial community composition of mats living in an evaporitic environment (i.e. the Khor al Adaid sabkha in Qatar), a setting that is known to catalyse the formation of dolomite (DiLoreto et al., 2019).Those authors observed that dolomite precipitation occurs when cyclic shifts in microbial community between cyanobacteria and anoxygenic phototrophs create EPS rich in carboxylic functional groups.This particular EPS is not produced when the microbial community is stable, rather, when it is stressed due to an abrupt change in the water's salinity.A process analogue to that described from the microbial mats could also occur in Lake Van's pore waters.As proposed earlier, abrupt climate change driven reventilation of anoxic or suboxic pore waters may destabilize the equilibrium of the microbial community which, before drifting to a composition optimized to the new redox conditions, produce EPS that favour the nucleation of dolomite (McCormack et al., 2018).Interglacial/ glacial and interstadial/stadial changes from wet/warm to dry/cold are ultimately responsible for these dynamic changes in lake and pore water chemistry, and the frequency and intensity of such changes may determine the total dolomite content forming within a distinct stratigraphic layer.Further, the comparison of diatom assemblages between MIS 7e and MIS 5e is interpreted to document less stable bottom water conditions for the former, with more frequent switching between anoxic and oxic conditions (North et al., 2017).Frustules of the diatom Cyclotella meneghiniana, a species common in lakes that fluctuate between freshwater and brackish water states, dominate at an AR composite core depth of 112 m below lake floor (MIS 7e, North et al., 2017), corresponding to the highest dolomite concentration (Fig. 1; Data S1).A higher frequency in bottom water chemistry fluctuations during MIS 7e may have resulted in higher dolomite concentrations compared to MIS 5e (Fig. 1D).
Reventilation and advection of pore fluids with contrasting redox potential and chemical composition may also trigger dolomite formation in the sediments of Lake Van through a process that is more complex than those outlined above.Abiotic and biotic factors may co-occur, creating an ideal window for dolomite formation, which closes again when the gradient between the contrasting fluids runs out.Such a combination of factors may be similar to those described by Petrash et al. (2021) in their recent re-evaluation of the 'mixing-zone model for dolomite formation'.The admixture of compositionally different pore waters does not only produce a gradient of major element concentrations but may also become the locus of enhanced redox sensitive biogeochemical reactions (for example, various microbial respirations, production of a specific type of EPS, mobilization of catalysing ions and dissolved silica in solution).Combined, these biotic and abiotic factors affect Ca-Mg carbonate (dis)equilibrium conditions and may ultimately result in the dissolution of soluble aragonitic and calcitic phases, and the subsequent nucleation of HMC and dolomite.

CONCLUSIONS
The presented data confirm that dolomite formation in Lake Van is ultimately triggered by climatically-induced changes in the lake level and deep water reventilation on the temporal scale of tens to hundreds of years.Two important observations in the ca 250 kyr sedimentary profile include: (i) a pairing of dolomite with HMC, indicating HMC formation either prior to or simultaneously with dolomite, perhaps as a precursor phase; and (ii) episodic presence of diatom frustules encapsulated by dolomite crystals.The latter feature is unique and testifies to a narrow range of the pH of deep water, which facilitated preservation of biogenic silica and simultaneous precipitation of dolomite.To account for these observations, the authors propose a continental analogue of a coastal 'mixing-zone' dolomite formation model.Instead of mixing marine and freshwater endmembers our model involves mixing glacial/stadial, high-alkalinity, oxic and interglacial/interstadial low-alkalinity anoxic deep water.Reventilation of the lake's bottom water and ensuing mixing of lake and pore fluids with contrasting redox potential and chemical composition can establish stratigraphic distinct 'mixing-zones'.These zones are likely to favour biogeochemical processes which dissolve primary, Mg-poor carbonate phases but promote the nucleation and growth of early diagenetic Mg-rich carbonate phases, and maintain only very subtle pH changes.These observations from a continental environment support the notion proposed for the marine realm, stating that transient and dynamic conditions may play a bigger role in dolomite formation than overcoming single kinetic barriers.Finally, the herein documented hydrochemical mixing of pore waters may also trigger episodes of dolomite formation in other alkaline lake settings worldwide.

ACKNOWLEDGEMENTS
The analyses were in part funded from the Deutsche Forschungsgemeinschaft (DFG) grant 268985795.We are grateful to Sebastian Breitenbach, Thomas Reinecke and Rolf Neuser for instrumental assistance.Julian Stromann and Franziska Nitz are thanked for their assistance with the sample sieving.We are thankful for the constructive feedback, criticism and comments of the three anonymous reviewers and Associate  .Variable correlation plot showing the relationships between all proxies.Positively related proxies plot close and form a group, whereas proxies with a negative relation plot on opposite sides of the plot origin (centre), such as the lake productivity proxies and the detrital (runoff) proxies.Note, LMC plots here with the detrital runoff proxies, which is in line with LMC precipitation under close to freshwater conditions within river plume whitings (McCormack et al., 2019a).The length of the arrows indicates the contribution (or representation) of each variable on the factor map. Here, we identify four specific groups: Mg carbonates (Ca-dolomite, HMC and near-stoichiometric dolomite), lake productivity proxies, detrital runoff proxies and summer evaporation proxies.Dolomite and aragonite proxies load most heavily on PC2 with LMC and aragonite showing an inverse correlation to dolomite, further supporting our hypothesis of CaCO 3 dissolution prior to Mg carbonate formation.Environmental proxies (specifically relating to detrital runoff, lake productivity and bottom water oxygenation) and LMC load on PC1 suggesting that environmental conditions developed separately from dolomite formation.XRF-determined bulk intensities of major elements Si, Al, Ca, K, Ti, Fe, Mn, Cl and element ratios Si/K, K/ Ca, Ca/K, Ca/Fe, Si/Al, Mn/Fe, as well as total organic carbon (TOC) and carbonate content are from Kwiecien et al. (2014) and Stockhecke et al. (2014a).

Fig. 3 .
Fig. 3. Scanning electron microscopy (SEM) images of Lake Van calcian dolomite.(A) to (D) Dolomite cement bare of diatom frustules, with details of rhombohedra with stepped crystal surfaces.(E) to (H) Dolomite rhombohedra overgrowing diatom frustules, which in some cases still show intricate details documenting little to no silica dissolution.Dolomite and diatom frustules are exemplarily highlighted by coloured arrows in (E) and (F).(F) Partly dissolved diatom frustule (orange arrow) with dolomite rhombohedra (red arrow) infilling the natural diatom frustule cavities.

Ó
2023 The Authors.Sedimentology published by John Wiley & Sons Ltd on behalf of International Association of Sedimentologists, Sedimentology, 71, 871-886 Dolomite formation in an alkaline lake 881

Ó
2023 The Authors.Sedimentology published by John Wiley & Sons Ltd on behalf of International Association of Sedimentologists, Sedimentology, 71, 871-886Editor Andrea Martin-P erez.Open Access funding enabled and organized by Projekt DEAL.

Figure S3
Figure S3.Variable correlation plot showing the relationships between all proxies.Positively related proxies plot close and form a group, whereas proxies with a negative relation plot on opposite sides of the plot origin (centre), such as the lake productivity proxies and the detrital (runoff) proxies.Note, LMC plots here with the detrital runoff proxies, which is in line with LMC precipitation under close to freshwater conditions within river plume whitings(McCormack et al., 2019a).The length of the arrows indicates the contribution (or representation) of each variable on the factor map. Here, we identify four specific groups: Mg carbonates (Ca-dolomite, HMC and near-stoichiometric dolomite), lake productivity proxies, detrital runoff proxies and summer evaporation proxies.Dolomite and aragonite proxies load most heavily on PC2 with LMC and aragonite showing an inverse correlation to dolomite, further supporting our hypothesis of CaCO 3 dissolution prior to Mg carbonate formation.Environmental proxies (specifically relating to detrital runoff, lake productivity and bottom water oxygenation) and LMC load on PC1 suggesting that environmental conditions developed separately from dolomite formation.XRF-determined bulk intensities of major elements Si, Al, Ca, K, Ti, Fe, Mn, Cl and element ratios Si/K, K/ Ca, Ca/K, Ca/Fe, Si/Al, Mn/Fe, as well as total organic carbon (TOC) and carbonate content are fromKwiecien et al. (2014) andStockhecke et al. (2014a).
Ó 2023 The Authors.Sedimentology published by John Wiley & Sons Ltd on behalf of International Association of Sedimentologists, Sedimentology, 71, 871-886