Is seawater geochem ical composition recorded in marine carbonate?Evidence from iron and manganese contents in Late Devonian carbonate rocks

Zhe W ang·W en Guo,2·Ting Nie·Haoran M a·Tianzheng Huang·Yuanlin Sun·Bing Shen

Abstract Iron and manganese are the important redoxsensitive elements in the ocean.Previous studies have established a series of paleo-depositional redox proxies based on the form and contentof iron in sedimentary rocks(e.g.,degreeofpyritization,FeHR/FeT,Fe/Al).Theseproxies were developed and applied on siliciclastic-rich marine sediments.Althoughmarine carbonate rocks are generally considered to preserve the geochem ical signals of ancient seawater,neither Fe nor Mn content in marine carbonate rocks(Fecarb,Mncarb)has been independently used as a proxy to quantify environmentalcues in paleo-oceans.Both FeandMnare insolubleinoxic conditions(Fe2O3,Fe(OH)3,MnO2),while their reduced forms(Fe2+and Mn2+)are soluble.Therefore,oxic seawater should have low concentrations of dissolved Fe2+and Mn2+,and accordingly carbonate rocks precipitated from oxic seawater should have low Fecarb and Mncarb,and vice versa.To evaluatewhether Fecarb and Mncarb can beused to quantify oxygen fugacity in seawater,wemeasured Fecarb andMncarb ofUpperDevonian marinecarbonate rockscollected from ninesectionsin South China.Fecarbofintraplatformbasinsampleswas signif icantly higher than that of shelf samples,while shelf and basin samples had comparable Mncarb.Themodeling result indicates that the dramatic difference in Fecarb cannot be explained by variation in oxygen fugacity between the shelf and basin seawater.Instead,both Fecarb and Mncarb appear to bemore sensitive to benthic f lux from sediment porewater that is enriched in Fe2+and Mn2+.Porewater Fe2+and Mn2+derive from bacterial iron and manganese reduction;f luxwas controlled by sedimentation rateand the depthof the Fe(Mn)reduction zone in sediments,the latterof which is determ ined by oxygen fugacity at thewater—sedimentinterface.Thus,high Fecarb of thebasin samplesm ight be attributed to low sedimentation rate and/or low oxygen fugacity at the seaf loor.However,invariant Mncarb of the shelf and basin samplesmightbe the consequence of complete reduction ofMn in sediments.Ourstudy indicates that marine carbonate rocksmay notnecessarily record seawater composition,particularly for benthic carbonate rocks.The inf luence of benthic f lux m ight cause carbonate rocks'geochem ical signals to deviate signif icantly from seawater values.Our study suggests that interpretation of geochem ical data from carbonate rocks,including carbonate carbon isotopes,shouldconsider theprocessof carbonate formation.

Keywords Carbonate rocks·Fe content·Mn content·Oxygen fugacity·Benthic f lux

1 Introduction

It isw idely accepted that the seawater geochemical composition can be faithfully recorded in marine carbonate rocks,if diagenetic alteration can be ruled out.As such,reconstruction of seawater geochem ical composition andbiogeochemical cycles in paleo-oceans have been approached w ith various geochem ical proxies extracted from carbonate rocks(Frimmel 2009;Higgins and Schrag 2012;Kampschulte et al.2001;Kampschulte and Strauss 2004;Knolletal.1986;Pogge von Strandmann etal.2014;Webb and Kamber 2000;Zhao et al.2009).For example,carbonate carbon isotopes(δ13Ccarb)have been canonically interpreted as recording the isotopic composition of dissolved inorganic carbon(DIC)in seawater(Kump and Arthur 1999),andthus have beenw idelyusedin chemostratigraphic correlation(Halverson et al.2005;Knoll et al.1986;Zhu et al.2007).In addition,trace amounts of sulfate incorporated into the carbonate crystal lattice(Pingitore et al.1995),collectively known as carbonate associated sulfate(CAS),are believed to record the sulfur isotopic composition of seawater sulfate(Fike etal.2015;Kampschulte et al.2001;Kampschulte and Strauss 2004);such a proxymightbe preserved in early diagenesis and dolom itization(Gill et al.2008;Kah et al.2004).Moreover,it is proposed that seawater Mg isotopic compositions(δ26Mg)can be preserved in marine carbonate,although it is still debated which particular carbonate component is the most reliable archive(Higgins and Schrag 2012;Ma etal.2017;Pogge von Strandmann etal.2014).In addition to variousstable isotope systems,marine carbonate rocks have been used to reconstruct seawater elemental and radiogenic isotopic compositions,including rare earth elements(REEs),and strontium and neodym ium isotopes.Shale-normalized REE data extracted from marine carbonate rocks of differentages demonstrate a sim ilar light REE(LREE)depleted pattern(Nothdurftetal.2004;Webb and Kamber 2000),while Sr and Nd isotopic compositions of paleo-oceans aremostly derived from marine carbonate(Edmonds 1992;Jones and Jenkyns 2001;Richter etal.1992;Shaw and Wasserburg 1985),although bioapatite(e.g.conodonts)has been used as well(e.g.,Ruppeletal.1996;Saltzman etal.2014;Chen etal.2018).

Because the geochem ical composition of carbonate rocks can be altered during diagenesis,evaluationsmustbe carefully assessed before data interpretation.In addition to petrographic observation,which provides the f irst-hand assessment,various geochem ical proxies,including the Mn/Sr ratio(Kah et al.2012),the absolute values of oxygen isotopes,and the correlation of carbon and oxygen isotopes(Bannerand Hanson 1990;Jacobsen and Kaufman 1999;Knauth and Kennedy 2009),have been applied in diagenetic evaluation.For carbonate rocks thathave passed through diagenetic evaluations,geochem ical compositions have been interpreted to record seawater composition.

Using carbonatesasan archive of seawatergeochemical composition assumes thatmarine carbonates precipitated within seawater.However,most carbonates were actually precipitated at seaf loor near the water—sediment interface(WSI)in non-pelagic settings.When precipitating at the seaf loor,marine carbonate isaffected by both seawater and sedimentporewater.Forexample,signif icantbenthic f luxes have been w idely observed along the modern seaf loor.Unlike oxic seawater,benthic f luxes are particularly enriched in Fe2+and Mn2+,generating porewater-seawater concentration gradients and allow ing the diffusion from porewater to seawater(Severmann et al.2010;John et al.2012;Cai et al.2014,2015;Wehrmann et al.2014).In addition,porewater DIC that derives from organic matter degradation can be delivered to the seaf loor(Cai et al.2015),contributing to benthic carbonate precipitation.Under the inf luence of benthic f lux,the geochem ical composition of benthic carbonate can vary from that of seawater.Accordingly,the potential impactof benthic f lux onmarine carbonate rocks needs further exploration.

In this study,we focused on Fe and Mn contents in carbonate rocks.Iron and manganese are both redox-sensitive elements.Ferric Fe(III)is rather insoluble in neutral to basic pH solution,and readily precipitatesas iron oxides(hematite,Fe2O3)or iron oxyhydroxides(Fe(OH)3).Bacterial reduction of ferric Fe in suboxic conditionsgenerates ferrous Fe(II)that is soluble atall pH conditions(Nealson and Myers 1990).Similarly,manganese is dominated by Mn(IV)in the form of MnO2in oxic Earth surface,while Mn(II)mainly derived from anaerobic bacterial reduction ofMn(IV)issoluble(Myersand Nealson 1988).Therefore,both Fe(nM)and Mn(μM)have extremely low concentrations in oxic seawater(Bruland et al.2014).In contrast,both Fe reduction and Mn reduction take place in suboxicanoxic conditions(Canf ield and Thamdrup 2009;Canf ield et al.1993),resulting in an accumulation of Mn2+and Fe2+in high concentrations in anoxic seawater/porewater.An Fe speciationmethod hasbeen developed to reconstruct seawater redox conditions.Two proxies have been used in this methodology:FeHR/FeT(the ratio between highly reactive Fe and total Fe)and Fepy/FeHR(the degree of pyritization or the ratio of pyrite Fe contentw ith respect to highly reactive Fe)(Raiswell and Canf ield 1998;Poulton and Raiswell 2002;Lyons and Severmann 2006;Poulton and Canf ield 2011).The Fe speciationmethod was developed based on siliciclastic sediments;its implication for carbonate rockshasbeen proposed recently(Clarkson etal.2014).

Cathodolum inescence(CL)is an eff icientway to evaluate Mn and Fe contents in carbonate rocks(Barbin et al.1991;Budd et al.2000),and has been used in the evaluation of diagenetic history of carbonates(Pierson 1981).In CL m icroscopy,Mn is the stimulator while Fe is the quencherof lum inescence(Budd etal.2000;Machel1985;Pierson 1981).Carbonate w ith low Mn and Fe concentrations is characterized by non-lum inescence,indicating its deposition in oxic seawater,while carbonate precipitated insuboxic conditions has high Mn but low Fe contents,and displaysbright lum inescence.Incontrast,carbonate formed in anoxic environmentsw ith both high Mn and Fe contents shows dull lum inescence.CL isa powerful tool in determ ining the depositional environment of marine carbonates;however,thismethod can only provide a qualitative estimation of redox conditions or oxygen fugacity during carbonate formation.

Untilnow,neither Fe(II)orMn(II)contents in carbonate(Fecarband Mncarb)hasbeen independently used asa proxy in paleoenvironment studies.As one of the four componentsof reactive Fe in the Fe speciation analysis(Anderson and Raiswell 2004;Poulton and Canf ield 2005),Fecarbalone has not been used in environmental interpretation.Even less is known about themeaning of Mncarb,except that Mn/Sr ratio is used for diagenetic evaluation of carbonate(Jacobsen and Kaufman 1999;Kaufman and Knoll 1995).To explore whether Fecarband Mncarbof marine carbonate can be used to constrain the redox condition of seawater,wemeasured Fecarband Mncarbof Late Devonian marine carbonates from nine sections in South China.Then,a numerical model was developed to evaluate the applicability of Fecarbin paleoenvironment study.

2 Geologic setting

Carbonate samples were collected from nine sections in South China(Fig.1),including the follow ing sections:the Daposhang of Changshun County(Guizhou);Dazhai,Madao,and Changtang of Dushan County(Guizhou);Xiada,Duli-A,and Duli-B of Nandan County(Guangxi),Baisha of Yangshuo County(Guangxi);and Panlong of Wuxuan County(Guangxi).Sampling intervals are bracketed between the Late Devonian(the Frasnian Stage and the Famennian Stage,382.7—358.9Ma)and the early Carboniferous(the Tournaisian Stage,358.9—346.7Ma),during South China was the site of development of numerous offshore carbonate platforms transectedby multiple narrowintraplatformbasins.Such paleogeographic conf iguration m ight be attributed to the N—NE migration of the South China Block since the M iddle Devonian.Trans-tensional tectonic movement resulted in the gradual fragmentation of the southern partof the South China Block and the development of two major sets of trans-tensional riftbasins in GuangxiProvince(Chen etal.2001a).One set of en echlon basins formed as a result of reactivation of antecedent NE-SWsinistral strike-slip faulting along the deep-seated basement zone,while the other set of rhomb-shaped basins m ight be related to movement of a sinistral strike-slip fault(Chen et al.2001a,b,2006).In the Late Devonian to Early Carboniferous,the Dazhai,Madao,and Panlong Sections were locatedoncarbonate platforms,representingshallow marine depositions(Fig.1).The platform succession is characterized by massive algal,oolitic,and shelly wackestone,packstone,and grainstone of the Rongxian Formation(Fig.2a—d).The other six sections were deposited in successional intraplatformbasin environments(Fig.1)represented by the Gubi Formation(from Frasian to lower Famennian)and theWuzhishan Formation(Famennian to lower Tournasian).The Wuzhishan Formation can be correlated w ith the Daihua(Famennian)and Wangyou(Tournasian)Formations in Guizhou Province.The Gubi Formationis composedof thin-tomedium-bedded wackestone and lam inated limestone(Chang et al.2017),while the Wuzhishan Formation is characterized by thinbedded and nodular limemudstone and wackestone,with occasional packstone(Fig.2e,f).Sampling intervals are plotted in Fig.3.The detailed conodont biostratigraphic framework is reported in the Daposhang(Ji1989),Dazhai,Duli-A,and Changtang(Nie et al.2016),and Baisha Sections(Chang et al.2017).A detailed description of studiedsectionsisincludedinthesupplementary information.

3 M ethods

Fresh carbonate samples were split with a rock saw and mirrored thin and thick sectionswere prepared from each split.Sample powders were collected from thick sections using a hand-held m icro-drill(w ith a drill bit 0.2mm in diameter).The sampling was guided by petrographic observation of mirrored thin sections.Three carbonate components were recognized and sampled separately:micrite,calcispar,and biogenic clasts.Although clasts included fossil fragments of brachiopod,echinoderm,gastropod,ostracod,and foraminifera,only the brachiopod shells were large enough for sampling.For brachiopod coquinoid limestone samples(in the Changtang,Duli-A,and DazhaiSections),sample powderswere collected from polished rock slabs using a hand-held m icro-drill.

The dissolutionprocedure followedthe sequential extraction method of carbonate Fe developed by Poulton and Canf ield(2005).A buffering solution consisting of a mixture of acetic acid and ammonium acetate was prepared,and pH 4.5 was precisely achieved by changing the mixing ratio of acetic acid and ammonium acetate.About 50 mg sample powder was carefully weighed in an electronic balance and dissolved in 10m l buffering solution.Reaction was allowed in a shaking table at50°C for 48 h.After centrifugation,0.5m l supernatantwas collected and dried in ahotplate.The insoluble fractionwas re-dissolved in 2%nitric acid,ready forelemental composition analysis.Thisdissolutionmethodcanguaranteecompletedissolution of carbonate minerals(aragonite,calcite,and dolom ite),while oxides(Fe2O3,Fe3O4,and MnO2)and sulf ide(FeS2)remain unaffected(Poulton and Canf ield 2005).

Fig.1 Late Devonian paleogeographicmap of the Yangtze Block,South China.Sample locations aremarked by black dots.1.Daposhang;2.Duli-A;3.Duli-B;4.Xiada;5.Changtang;6.Dazhai;7.Madao;8.Baisha;and 9.Panlong

Theelementalcompositionsweredetermined by Spectra Blue Sop Inductively Coupled Plasma Optical Emission Spectrometry(ICP-OES)at Peking University.A series of gravimetric standard solutions with elemental concentrations ranging from 0.1 to 10 ppm were prepared.The calibration curve of each elementwas f irst established by measuring all the standard solutions before sample analyses.Analytical precision was better than 5%for all elements.In calculations,Ca and Mg were converted to carbonatem inerals,i.e.CaCO3and MgCO3,respectively,and the carbonate fraction was represented by the total mass of CaCO3and MgCO3.The elemental compositions of other elements(Fe,Mn,Sr)were calculated based on carbonate fraction.

4 Results

Fe and Mn contents in the carbonate component(Fecarband Mncarb)are listed in Table S1 and plotted in Fig.4.Before further data analysis,this study considered the effect of carbonate rocks'Mg/Ca on Fecarband Mncarb.The content of Mgcarband Cacarbis related to Mg2+and Ca2+in seawater.Aragonite and high-magnesium calcite(＞4mol%MgCO3)are preferentially precipitated fromaragonite seawater(Mg/Ca＞2mol/mol),while low-magnesium calcite precipitates from a calcite sea(Mg/Ca＜mol/mol)(Hardie 1996;Stanley and Hardie 1998,1999).The carbonate rocks returned abnormally high Fecarbin the case of high-magnesium samples(＞4mol%MgCO3).This phenomenon is presumably due to the substitution of Mg2+in the carbonate lattice for Ca2+;lattice defects created by the smaller ionic radius of Mg2+also cause Fe2+to enter the carbonate lattice(Mazzullo 1992).Therefore,high-magnesium sampleswere excluded from our data analysis.The Fecarband Mncarbresults of the nine sections are summarized in Table 1.

Fig.2 Photomicrographs show ing the texture of Late Devonian carbonate samples:a dolomitic limestone from the bottom part of Panlong Section;b oolitic packstone from the Dazhai Section;c peloid wackestone from the Panlong Section;d shelly packstone from the Dazhai Section,composed of densely packed shells of the rhynchonellid brachiopod Dzieduszyckia;e argillaceous limestone from the Changtang Section,show ing distinguished lime-rich and mud-rich components;and f shelly wackestone from the Changtang Section,consisting of disarticulated shells of the rhynchonellid brachiopod Dzieduszyckia

5 Diagenetic evaluation

The geochem ical composition of carbonate rocks can be altered during diagenesis.Therefore,potential diagenetic alteration should be evaluated before data interpretation.Rock fabric tends to be altered during diagenesis,e.g.elim ination of original fabric during recrystallization as well as non-m im ic replacement in dolom itization(Banner et al.1988).Thus,petrographic observation provides the most straightforward evidence for diagenetic alteration.The texture and fabric of all studied carbonate rocks was recognized.In this study,sample powders were collected by using am icro-m ill(w ith bitsize of 0.5mm),warranting the m illimetric spatial resolution in sampling.Sampling was component-based(i.e.differentiation of m icrite,calcispar,and biogenic clasts),and was guided by petrographic observationin thin section.Inthis method,recrystallization regions and hydrothermal veins,if any,can be avoided during sampling.

CL has been w idely applied in diagenetic evaluation of diagenesis.CL is controlled by the absolute concentrations of Mn and Fe in carbonate(Mncarband Fecarb)and by the Mncarb/Fecarbratio(Pierson 1981).In general,pristine marine carbonate(i.e.precipitated in oxic seawater)that is not signif icantly altered in diagenesis shows non-lum inescence(i.e.low Mn and Fe contents),while diagenetically altered carbonate displays either bright lum inescence(i.e.high Mn but low Fe content,ref lecting early diagenesis in suboxic conditions)or dull lum inescence(i.e.high Mn and high Fe,ref lecting late diagenesis in anoxic conditions).The studied carbonate samples displayed non-lum inescence to dull luminescence(Fig.5).However,we suggest that diagenesis may not be the only interpretation of varying lum inescence(see below).

Strontium and manganese contents in carbonate have been used in diagenesis as well(Kah et al.2012;Gilleaudeau and Kah 2013)because Sr tends to be expelled from the carbonate lattice in early diagenesis,while Mn increases during burial.Based on principle,diagenetic

?Fig.3 Conodont biostratigraphic correlation among the studied sections.The conodont zonation for the Duli-A,Changtang,and Dazhai Sections is adopted from Nie et al.(2016),for the Baisha Section from Chang etal.(2017),and for the Daposhang Section from Ji(1989);the Siphonodella sulcata Zone is amended based on unpublished data.Other sections are based on unpublished data evaluation has been approached w ith Mn/Sr ratios.It is proposed that carbonateswith low Mn/Sr ratios(＜10)can be regarded as least altered,suggesting potential preservation of the stratigraphic trend of geochem ical signals(e.g.carbonate carbonisotopes)and applicability in chemostratigraphic correlation.In contrast,samples with high Mn/Srmay indicate possiblealteration of stratigraphic trends(Kaufman and Knoll 1995).Mn/Sr ratios of the analyzed carbonate samples were low,0.024(2.5%quantile value)to 4.113(97.5%quantile)(Table S1),arguing against diagenetic alteration of our samples.

Although petrographic observation,CL and Mn/Sr ratios indicate that the studied carbonate samples have not been signif icantly altered by diagenesis,some of these canonical interpretationsmay not be valid if carbonate was precipitated nearWSI.

6 Discussion

6.1 Fecarb and M ncarb variation of different carbonate com ponents

Three carbonate componentswere recognized in the studied samples:m icrite,calcispar,and biogenic clasts(Figs.3,S1,S2,S3).There are three possible sources ofm icrite in themodern ocean:disintegration of weakly calcif ied green algae,such as Helimeda and Penicillus(Wefer 1980),breakage of biogenic carbonate grains(micritization)(James and Choquette 1983),and inorganic ormicrobialinduced precipitation fromseawater or within marine porewater(Munnecke and Sam tleben 1996).Among the three possible sources,calcif ied green algae represent the most importantm icrite producer in themodern ocean,but it isunclearwhether the algae thatcan generatem icritehad evolved or become ecologically important by the late Paleozoic(Kaz′mierczaket al.1996;Riding1991;Verbruggen et al.2005).On the other hand,inorganic precipitation of calcispar from normal seawater may not have been favored in the Phanerozoic because of the low carbonate saturation state(Dupraz et al.2009;Mackenzie and Morse 1992).In most cases,precipitation of calcispar occurs during cementation that f ills in the pore spacebetween carbonate and/or siliciclastic grains(James and Choquette 1983).Cementation takes placewhen carbonate saturation is locally elevated,probably due to decomposition of organic matter that elevates pH and alkalinity(Gallagher et al.2014).Calcispar could derive from diagenetic recrystallization,in which micrite and microcrystallinecalcite/aragonitearedissolvedfollowedby reprecipitation of coarser grained carbonate(James and Choquette 1983).Calcispar derived from recrystallization isnormally characterized by a limpid rim and a turbid core.The biogenic carbonate grains in the samples were dom inated by brachiopod shells composed of low-Mg calcite(Ma et al.2017);other skeletons such as echinoderms,foraminifera,and ostracodswere lesscommon(Figs.3,S1,S2,S3).

Fig.4 Fecarb versus Mncarb cross-plots.Open symbols represent partially dolom itized samplesw ith Mg/Ca(mol/mol)＞4%,while solid symbols represent various carbonate components(a,b)or different depositional environment(c,d).a Samples from basin sections;b samples from shelf sections;c micrite samples from different sedimentary facies;and d calcispar samples from different sedimentary facies

Table 1 Summarized measurement of the Fe,Mn contents in carbonate component(Fecarb,Mncarb)from the nine studied sections

Fig.5 Cathodoluminescence photom icrographs of Late Devonian carbonate samples:a non-lum inescence of oolitic packstone from Dazhai Section(DZ-19);b nonlum inescence of peloid wackestone from Panlong Section(PL-3),w ith calcispar show ing dull lum inescence;c non-lum inescence of shelly packstone from coquinoid limestone(upper part of Dazhai Section,DZ-PS-23);and d nonlum inescence of limemudstone from the upper partof Daposhang Section(DPS-4),w ith calcispar show ing dull lum inescence

As compared w ith micrite and calcispar,brachiopod shells returned the lowest Mncarband Fecarbcontents,ranging from0.0(i.e.belowthe detection limit)to 275.4 ppm(mean=32.0 ppm,n=38)and 5.0—2043.7 ppm(mean=243.9 ppm,n=38),respectively.For shelfsamplesw ithlower Fecontents(＜100 ppm),m icrite had higher Mn(mean=82.2 ppm)than calcispar(mean=25.0 ppm),whereas high Fe samples(＞100 ppm)hadoverlappingranges of Mncarb(Figs.6,7).For carbonate samples from basin sections,micrite and calcispar had overlapping ranges of Fecarbbut micrite had higher Mncarbthan calcispar(286.1 ppm vs.93.3 ppm,Figs.6,7).

6.2 Fecarb and M ncarb variations in different depositional environments

Except for two outliers,m icrite of shelf sampleshad lower Fecarbthan basin samples,while micrite of the shelf and basin samples had comparable Mncarb(Figs.6,7).There wasa crude positive correlation between Fecarband Mncarbfor basin samples.Shelf samples w ith low Fe content(＜200 ppm)had abnormally high Mncarb,while Mn content remained low(＜100 ppm)when Fecarbranged from 200 to～1000 ppm.At Fecarbof 1000 ppm,Mncarbreturned a higher level,but seldom exceeded 400 ppm(Fig.5).Samplesw ith high Mn but low Fe contentswere sourced from the Panlong Section(w ith sampling interval transecting the Frasnian-Famennian boundary).The cause of exceptionally high Mn contentof the Panlong samples is unclear.We speculate that Mn enrichment in the Panlong m ight be related to the metallogenesis of the Xialei Mn ores in nearby regions(Zeng and Liu 1999).

6.3 Can Fecarb indicate seawater redox condition?

It is w idely accepted that elemental and isotopic compositions of marine carbonate may record the geochemical signature of seawater if diagenetic alteration can be conf idently excluded(Banner 2004;Pelechaty et al.1996)Thus,carbonate has often been used in paleoenvironment studies.For example,carbonate carbon isotopes(δ13Ccarb)have long been used to indicate isotopic composition of DIC in seawater,and the stratigraphic variation ofδ13Ccarbhas been used for chemostratigraphic correlations at regional and global scales(Kaufman and Knoll 1995).Since Fe is a redox-sensitive element and Fe content in seawater([Fe]sw)is controlled by seawater oxygen fugacity,we explored whether Fecarbcan be used to record the redox condition of paleo-oceans.

In oxic conditions,Fe is dom inated by Fe(III),which is rather insoluble at neutral to basic pH conditions.Fe(III)can be anaerobically reduced to soluble Fe(II)(Nealson and Myers1990).Furthermore,because ferrous iron(Fe2+)has a similar charge but smaller ionic radius compared to Ca2+,Fe2+readily substitutes for Ca2+in the carbonate crystal lattice.The degree of substitution is determ ined by the concentration of Fe2+in solution and the partitioning coeff icient between carbonate and solution,the latter of which is affected by other factors,such as temperature(Morse and Bender 1990).

The amount of Fe2+in seawater is determ ined by oxygen fugacity(f(O2)).The chem icalequation for ferrous iron oxidation by oxygen can be expressed as:

Because Fe2O3and Fe(OH)3are insoluble,the equilibrium for Eq.1 can be expressed as:

where K is the equilibrium constantand[H+],[Fe2+],and[O2]are dissolved proton,ferrous Fe,and O2concentrations in seawater,respectively.Rearranging Eq.2,we get:

Dividing Eq.3 by[Ca2+]on both sides,we get:

i s the Fe/Camolar ratio in seawater.Theamount of Fe in carbonate((Fe/Ca)carb)can be calculated by:

where D is the ratio of partitioning coeff icientsbetween Fe and Ca,i.e.D=DFe/DCa,during carbonate precipitation.

Equation 5 indicates that Fe content in carbonate is inversely correlated w ith the fourth root of O2concentration,suggesting that carbonate would have higher Fecarbwhen precipitating from anoxic seawater(low oxygen fugacity),and vice versa.

Fig.6 Histogramsof Fecarb and Mncarb from different depositional environments:a Fecarb ofm icrite;b Mncarb of micrite;c Fecarb of calcispar;and d Mncarb of calcispar

Fig.7 Box-plots show ing Fecarb and Mncarb of the Late Devonian carbonate samples.Box brackets delineate 25 and 75 percentiles,while the bar in them iddleof the box represents themedian of the values.Whiskers indicate 2.5 and 97.5 percentiles,and hollow dots are outliers.a Fecarb ofmicrite;b Mncarb ofm icrite;c Fecarb of calcispar;and d Mncarb of calcispar from differentsections

Furthermore,if seawater and atmosphere are in equilibrium,the relationship between Fecarband atmospheric p O2level canbe expressedbyusingHenry's law,[O2]=H*p O2:

By using Eq.6,atmospheric p O2can be theoretically reconstructed from Fecarbif seawaterand atmosphereare in equilibrium.However,because the partitioning coeff icients are loosely constrained,atmospheric p O2cannot be precisely calculated.We compared the shelf and basin carbonate samples by assuming,(1)shelf and basin seawater have sim ilar[Ca2+]and pH,and(2)seawater in the shelf environment is in equilibrium w ith atmospheric p O2.The oxygen fugacity in the basin environmentcan be calculated by the follow ing:

where superscripts shelf and basin indicate the shelf and basin samples,respectively,and［O2］basinrefers to dissolved oxygenconcentrationinthe basinenvironment.By assuming the atmospheric p O2level in the Late Devonian was equal to that of present day(0.21 atm)and seawater temperature of 25°C(Sperling et al.2015),Henry's constant for O2is set to 1.3mmol/L/atm.The average values of Fecarbfor the shelf and basin samples were 77.9 and 1026.2 ppm,respectively.The estimated［O2］basinwas 0.009μM or0.286μg/L(Fig.8).Thisvalue is threeorders of magnitudelower thanthethresholdfor anoxic([O2]＜200μg/L)(Libes 2009),suggesting that the basin environment was extremely anoxic and bacteria sulfate reduction occurred.This calculation represents the maximum estimation of［O2］basinby assum ing shelf seawater was in equilibrium w ith atmospheric p O2.In fact,the shelf seawater could be less oxic,if oxygen consumption by organicmatter degradation were considered.

The estimated［O2］basinlevelbased on Fecarbcontradicts paleontological evidence.The presence of benthic animal fossils,such as gastropods,ostracods,and echinoderms,strongly argues against extreme anoxic conditions in the basin sections(Figs.3,S1,S2,S3).Therefore,seawater oxygen fugacity may not be the only control of Fecarb,suggestingthattheapplicationofFecarbin paleoenvironment study is not straightforward.Below,we explore the controlling factors of Fecarb.

6.4 Controls on Fecarb and M ncarb

Equation 6 indicates that Fecarband Mncarbare controlled by the oxygen fugacity of seawater.However,our calculations indicate that the estimated［O2］basinis too low and contradicts paleontological data(Figs.3,S1,S2,S3).Although potential diagenetic alteration cannot be completely ruled outby petrographic observation,CL,and Mn/Sr ratios,diagenesis alone cannot explain systematically higher Fecarbin basin sections.In fact,theaboveestimation assumes that seawater dissolved[Fe2+]is the only Fe source for carbonate,which m ight be true only when all carbonate precipitated from thewater column.In fact,it is well known that Paleozoicmarine carbonatesweremainly produced by benthic carbonate-secreting organisms(Bartley and Kah 2004;Mackenzie et al.2004;Mackenzie and Morse 1992;Tucker and W right 1990),such as brachiopods,echinoderms,mollusks,corals,and bryozoans(Sepkoskiand M iller1985).Althoughmodern foraminifera representone of themost important carbonate producers in the ocean,and exhibitboth benthic and planktonic lifestyle(Gupta 1999),Paleozoic foraminifera were exclusively benthic(falFrerichs 1971;Vachard et al.2010).Benthic carbonate produced at or near SW Iwould be affected by both seawater and sediment porewater.

In themodern open ocean seaf loor w ith relatively high oxygen fugacity,seawater is characterized by low[Fe2+].In contrast,sedimentporewater ismore reduced due to the consumption of dissolved O2by aerobic organic matter degradation.After complete depletion of O2,organic matter undergoes anaerobic degradation by using oxides(e.g.MnO2and Fe2O3)or oxy-anions(e.g.nitrate and sulfate)as electron acceptors(Canf ield and Thamdrup 2009;Canf ield et al.1993).Bacterialmanganese and iron reduction can be expressed as:

These reactions generate Fe2+,Mn2+,and HCO3-that derives from rem ineralization of organic matter,resulting in high concentrations of Fe2+and Mn2+in sediment porewater.As such,Fe2+and Mn2+concentration gradients are generated between porewater and seawater.The upward diffusion of porewater Fe2+and Mn2+into seawater generates benthic f luxes(Caietal.2014,2015;John etal.2012;Severmann etal.2010;Wehrmann etal.2014).When carbonate precipitates at the seaf loor,benthic f luxesof Fe2+and Mn2+can be incorporated into carbonate rocks,resulting in an increase of both Fecarband Mncarb.

Fig.8 Modeling results show ing the relationship between Fecarb and oxygen fugacity of seawater in thebasin environment.Contour lines represent different oxygen fugacity in the shelf facies.100%refers to the equilibrium between seawater and atmosphere;[O2]can be calculated by Henry's Law

The Fe2+and Mn2+f luxes from sedimentporewaterare determ ined by concentration gradients between porewater and seawater,which can be expressed by:

where▽iis the one-dimensional concentration gradientof species i(Mn2+and Fe2+),and liis the depth of the Fe/Mn reduction zone below WSI.Subscripts pw and sw represent porewaterand seawater,respectively.Because both[Mn]swand[Fe]swin oxic seawater are low(μM and nM level,respectively),the concentration gradient is mainly controlled by[Mn]pwand[Fe]pw.High[Mn]pwand[Fe]pwrequire availability of organicmatter and reactive Mn(i.e.MnO2)and Fe(i.e.Fe2O3)in sediments.W ith the consideration of benthic f luxes,Mn and Fe concentration in benthic carbonate can be calculated by:

where KFeis the partitioning coeff icient of Fe between solution and carbonate,DFethe coeff icientof diffusivity of Fe2+,A area,s sedimentation rate,ρthe density of carbonate,and MFethemolecularweight of Fe(56 g/cm3).

By ignoring[Fe]swin oxic seawater and combining Eqs.11 and 12,we arrive at:

Equation 13 indicates that Fecarbis controlled by the follow ing three parameters:[Fe]pw,lFe,and s.Obviously,high Fecarbis favored at low sedimentation rates,small lFe(i.e.shallow depth of iron reduction zone),and higher concentration of[Fe]pw.Ifwe compare the shelf and basin sections investigated in this study,the shelf sections have signif icantly higher sedimentation rate,as indicated by larger stratigraphic thickness(w ithin the same conodont zone)(Fig.2).Low sedimentation rate is accompanied by accumulated organic matter in sediment,enhancing O2consumption and shoaling the iron reduction zone(i.e.lower lFe).Therefore,high Fecarbof the basin samples m ightbe directly attributed to low sedimentation,which in turn leads to high f lux of Fe2+from sediment porewater.

Unlike Fecarbthat showed orders of magnitude difference between shelf and basin samples,the shelf and basin sampleshad an overlapping rangeofMncarb.Thismightbe attributed to(1)Mn reduction before Fe reduction in sediments(Canf ield etal.1993),and(2)Mn content in uppercontinental crust is about 50 times less than Fe(775 vs.39,200 ppm)(Rudnick and Gao 2003).

6.5 Does carbonate record seawater com position?

Ourstudy indicates thatFecarband Mncarbof Late Paleozoic carbonate are strongly affected by benthic f luxes from sedimentporewater,suggesting thatbenthic carbonatedoes not necessarily record seawater geochem ical composition.Therefore,the interpretation of geochem ical data extracted from carbonate samples should consider the processes of carbonate formation.We suggest that only carbonate precipitated w ithin the water column records seawater composition;benthic carbonate precipitated at the seaf loor is suspected to be inf luenced by benthic f lux from sediment porewater.Benthic carbonate is least likely to record seawater composition when the sedimentation rate and seaf loor oxygen fugacity are low,whereas shallow marine carbonate formed in well-ventilated seawater,where benthic f lux is low,is more likely to preserve seawater composition.

Carbonate carbon isotope(δ13Ccarb)is themostw idely used proxy in chemostratigraphic correlation(Kaufman etal.1993;Knoll etal.1986;Zhu etal.2013).Aδ13Ccarbgradient along the shallow-to-deep transect is commonly observed in sedimentary basins.Normally,shallow-water carbonate has higherδ13Ccarbvalues than deep-water carbonate.Traditionally,this gradienthas been interpreted in terms of bathymetric isotopic gradients in seawater(Jiang et al.2007;Lang et al.2016;Shen et al.2011).

A lternatively,our study suggests that theδ13Ccarbgradient could be generated by variation of sedimentation rate between shallow-and deep-water sections.Deposition of12C-enriched carbonate in the deep water environment might be attributed to a larger benthic f lux that delivers12C-enriched DIC from porewater(Cai et al.2014,2015;Emerson etal.2003).Large benthic f lux is favored by low sedimentation rate(Eq.13,Fig.9).In addition,seaf loor m ight have lower oxygen fugacity if the water depth is greater than the thickness of the surface m ixing zone.Therefore,interpretation ofδ13Ccarbdata should also consider the processes of carbonate formation.Sedimentological analyses and petrographic observations are highly recommended to differentiate whether carbonate was precipitated in thewater column orat the seaf loor(i.e.benthic carbonate).In addition,an estimate of sedimentation rate based on high resolution biostratigraphic and chronostratigraphic framework would also provide additional constraints on data interpretation.

W ith the consideration of benthic f lux in carbonate precipitation,there is no longer a clear cutoff between sedimentationanddiagenesis.Sedimentarycarbonate could also record signals of diagenesis,given that benthic f lux delivers porewatermaterial into seawater.Therefore,some traditional approaches or geochemical proxies in diagenetic evaluation warrant reconsideration.For example,bright or dull lum inescence in CL imaging may not necessarily indicate a diagenetic origin of carbonate.Benthic carbonate w ith signif icant input from benthic f lux would display similar luminescence.The Mn/Sr ratio should not be used either,because only Sr loss occurs in sediment,while Mn could be gained during benthic carbonate precipitation.Therefore,diagenetic alterationm ight be overestimated;some data m ight be explained by the inf luence of benthic f lux.

Finally,we recommend using Fecarband Mncarbas proxies to evaluate the potential inf luence of benthic f lux.Samples w ith high Fecarband Mncarbmay indicate strong benthic f lux,and accordingly are less likely to record seawater composition.Only carbonate samples w ith low Fecarband Mncarbmight be suitable for the reconstruction of seawater geochem ical composition.

Fig.9 Schematic diagrams show ing how Fecarb and Mncarb are controlled by benthic f luxes from sediment porewater.Both values are affected by sedimentation rate and the depths of redox boundaries below which iron(manganese)reduction would occur.The thickness of arrows represents the intensity of Fe2+and Mn2+benthic f luxes

7 Conclusions

We measured Fecarband Mncarbof Late Devonian carbonate samples from nine sections in South China.As compared w ith shelf carbonate samples,samples from basin sections had signif icantly higher Fe contents but comparable Mn contents.Modeling indicates that if seawater in the shelf environment was in equilibrium w ith atmospheric p O2,basin seawater would be extremely anoxic w ith dissolved oxygen content＜1μM,which is inconsistentw ith abundant benthic fossils,suggesting that seawater oxygen fugacity may not be the only control on Fecarb.We suggest Fecarband Mncarbmightalso beaffected by benthic f lux,bringing Fe2+and Mn2+from anoxic sediment porewater.High Fecarbin deep-water carbonate m ight have resulted from a relatively low sedimentation rate aswellas the shallow depth of the iron reduction zone in sediments.In contrast,the shelf and basin samples displaying comparable Mncarbm ight be attributed to nearly complete reduction of Mn in sediments,given the low Mn content in the upper continental crust.Our study indicates thatmarine carbonatemay notnecessarily record seawater geochemical composition.Carbonate precipitated on the seaf loor,i.e.benthic carbonate,would also be affected by benthic f lux from sediment porewater.Therefore,interpretation of carbonate geochem ical data should also consider the process of carbonate formation.

AcknowledgementsThis study was supported by National Science Foundation of China(Nos.41172001 and 41772015 to Sun and No.41772359 to Shen).We thank Prof.Linda Kah and one anonymous reviewer for constructive comments and suggestions.