CARBONATE PRECIPITATION, U SERIES DATING,AND U-ISOTOPIC VARIATIONSIN A HOLOCENE TRAVERTINE PLATFORMAT BAD LANGENSALZA - THURINGIA BASIN, GERMANY

�

Norbert FRANK 1,3, Bernd KOBER 2 & Augusto MANGINI 1

ABSTRACT

Travertine and calcareous tufa deposits, are important archives for Quaternary continental climate and archaeology. Here we present a com-prehensive study on Holocene travertine deposition in the Thuringia basin based on a detailed survey of U-series measurements and dating thatdemonstrates rapid accumulation of carbonate during the early Holocene until 8000 years (up to 8 mm yr–1) followed by significant change in tra-vertine accumulation that ended around 5800 years BP at the investigated site. We further demonstrate that systematic changes of the U-isotopiccomposition through time occur in agreement with major changes in travertine accumulation and texture, which are likely to reflect changes ofspring water U-isotopic composition rather than U-series system opening or contamination with non-carbonate particles.

Key words: Travertine, Thuringia, Holocene, U-series, Uranium isotopes.

RÉSUMÉ

PRÉCIPITATION DE CARBONATE, DATATION U/TH ET VARIATIONS ISOTOPIQUES DE L’URANIUM D’UNE PLATEFORME DETRAVERTIN HOLOCÈNE À BAD LANGENSALZA - BASSIN DE THURINGE, ALLEMAGNE

Les travertins et les tufs calcaires constituent des archives importantes pour les études sur le climat continental quaternaire et l’Archéologie.Nous présentons l’étude de la formation d’un travertin holocène du bassin de Thuringe, basée sur un exposé détaillé des données uranium-thorium etdes âges U-Th obtenus, qui démontrent une accumulation rapide du carbonate pendant le début de l’Holocène jusque vers 8000 ans (jusqu’à 8 mmpar an) suivie par un changement significatif dans le rythme d’accumulation du travertin qui s’achève vers 5800 ans BP sur le site étudié. Nous dé-montrons ensuite que les changements systématiques dans la composition isotopique de l’uranium au cours du temps sont en accord avec les varia-tions majeures de texture et d’accumulation du travertin et reflètent les changements de composition isotopique de l’uranium des eaux de sourceplutôt qu’une ouverture du système des séries de l’uranium voire une contamination par des particules non carbonatées.

Mots-clés : Travertin, Thuringe, Holocène, U-Th, isotopes de l’Uranium.

1 - INTRODUCTION

In the past 25 years researchers have shown that con-tinental carbonate deposits, such as travertine can beabsolutely dated to about 350,000 years before presentby means of the 230Th/U disequilibrium method(Blackwell & Schwarcz, 1986; Brunnacker et al.,1983; Frank et al., 2000; Harmon et al., 1980; Mallick& Frank, 2002; Schwarcz et al., 1988; Srdoc et al.,1994; Sturchio et al., 1994). This method is based onthe fact that U-isotopes are easily weathered from anaquifer host rock and are then transported withingroundwater, while the radioactive daughter product230Th is highly particle reactive and thus not abundantin groundwater. Uranium becomes co-precipitatedduring accumulation of secondary carbonate as traver-tine from organic and inorganic calcite or even arago-nite precipitation. Subsequently, 230Th is produced in a

travertine layer from decay of 234U, beginning with theformation of the layer.

Assuming closed system behavior of the travertineand no initial 230Th, one can determine the age of a tra-vertine or calcareous tufa deposit according to the ra-dioactive decay by measuring the present-day activityratios of (230Th/238U) and (234U/238U) (Ivanovich &Harmon, 1992). The (234U/238U) activity ratio ishenceforth given in per mil (‰) as δ234U, whereδ234U = ((234U/238Umeasured)/(234U/238Uequilibrium) – 1) * 1000.234U/238Uequilibrium is equal to 1 as an activity ratio and54.891·10-6 as an atomic ratio (Cheng et al., 2000).

The age uncertainty of this ideal scenario is givensolely by the precision and reproducibility of the mea-surement of both activity ratios and the precision towhich the decay constants are known. However, traver-tine contains aluminosilicate and limestone particles aswell as organic matter, which contribute to the overall

Quaternaire, 17, (4), 2006, p. 333-342

Manuscrit reçu le 06/07/2006, accepté le 13/10/2006

1 Institut für Umweltphysik, Universität Heidelberg, Im Neunehimer Feld 229, D-69120 Heidelberg, Germany2 Institut für Umwelt Geochemie, Im Neuenheimer Feld 234, D-69120 Heidelberg, Germany3 Laboratoire des Sciences du Climat et de L’Environnement-IPSL, Unité mixte CEA-CNRS-UVSQ, Bât.12, Avenue de la Terrasse, 91198 Gif-sur-Yvette, France. E-mail: norbert.frank@Isce.cnrs-gif.fr

amount of U-series nuclides. Moreover initial non-detrial 230Th can be a potential problem and postdepositional infilling with later carbonate generationsin the initial pore space and dead volume, mostly assecondary spar and especially carbonate cement anddetritus is frequent. Thus, any contamination stronglyinfluences the accuracy of an age determination. In ad-dition, even after secondary infilling of the pore spaceand dead volume travertine remains a highly porousrock that is subject to weathering resulting in dissolu-tion and re-precipitation of carbonate, redistribution ofU-series daughters through recoil processes, and U andTh loss or uptake. Classic Karst phenomena such asdissolution are widespread in travertine bodies, as indi-cated by the frequent presence of caves (Adam, 1986;Brunnacker et al., 1982; Frank, 1997; Grewe, 1987;Reiff, 1986; Sommermeier, 1913). In the marine envi-ronment the U-isotopic ratio stays constant over longtime scales (Henderson, 2002) and it can thus be usedto trace U-series system opening. In contrast, in conti-nental carbonates a rapid change of the U-isotopiccomposition of spring water hinders the use of the ini-tial U-isotopic ratio as a quality control instrument forprecise age determination.

To get around these many obstacles, different ap-proaches have been taken. To trace detrital contamina-tion one can use 232Th by either assuming average(238U/232Th) activity ratios of continental crust and ra-dioactive equilibrium (basic correction model) (Franket al., 2000), or by measuring aluminosilicate sedi-ments present at the travertine location (Frank et al.,2000). Studying several coeval samples using so calledisochron techniques is more time consuming but yieldmore sophisticated correction models, as the impact ofinitial 230Th and detritus 230Th can be corrected and ura-nium-series open-system behavior may be identifiedqualitatively if data is not matching a common detrital-carbonate mixing line (Frank et al., 2000; Ludwig &Titterington, 1994; Schwarcz, 1980; Sturchio et al.,1994). Other chronometers such as 231Pa/235U datingmay add information on the quality of ages (Frank,1997), but as the aqueous chemistry of 231Pa is complexand tracer isotopes such as 232Th are missing to quan-tify non-carbonate contamination its application totravertine seems so far quite useless.

Despite complex correction models the initial sam-pling process is a key to obtain precise and accurateages, as major contaminants and weathering featurescan be avoided. For high-resolution studies of “young”travertine the stratigaphic position of the samples isalso crucial in light of a highly irregular precipitationranging from almost horizontal layers and small cas-cades on surfaces having minor slopes, to successivecarbonate cascades and mounds on steeper slopes. Re-cently, Mallick and Frank (Mallick & Frank, 2002)have attempted to obtain ages using a micro-samplingtechnique to guarantee the investigation of primary cal-cite. The results for deposits of 100,000 to 250,000 yrsage are very promising, but beyond this age weatheringbecomes frequently too important to obtain sufficiently

large undisturbed micrite or spar layers (Mallick &Frank, 2002). For young deposits, analytical efficiencyfor 230Th measurements was insufficient to work on mi-cro-samples of less than 100mg size, but micrite domi-nated and associated pore cement may differ by800 years in age (Mallick, 2000) (see also tab. 1). Al-though, the anticipated significant removal of detritalcontamination failed, as micro detritus is present onany scale of a travertine deposit and can be barelyavoided (Mallick & Frank, 2002). However, till nowthe technique to combine detailed petrographic andgeochemical studies prior to U-series dating yield themost convincing results for <250,000 year old depo-sits. Moreover, prior investigations of Frank et al.(Frank et al., 2000) and Mallick and Frank (Mallick &Frank, 2002) have shown that the initial U-isotopiccomposition varies through time. This can be due to (1)U-series open-system behavior assuming constantU-isotopic composition in groundwater, analogous toU-series system opening in corals (Gallup et al., 2002;Gallup et al., 1994; Henderson & Slowey, 2000; Scholzet al., 2004; Thompson et al., 2003; Villemant &Feuillet, 2003) or (2) variable groundwater U-isotopiccomposition during travertine formation, potentiallyindicating changes of groundwater residence time,weathering of host rocks, precipitation, and/or sourceof U-isotopes.

Here we investigate in great detail a travertine de-posit that originated during the Holocene to study thecomplex accumulation pattern in an environmentlargely free of diagenesis. From such a record we caninvestigate the relation of initial carbonate precipita-tion and secondary infilling of initial pore space, andwe can further determine the natural variability of theU-isotopic system in travertine to check whether varia-tions observed in nearby older travertine deposits aredue to weathering and erosion or reflect naturalchanges in groundwater composition and regional en-vironmental conditions.

2 - GEOLOGICAL SETTING

Travertine and calcareous tufa deposits of theThuringia basin originate from meteoric groundwaterrunning through extended limestone and Keuper for-mations of the German Triassic (fig. 1). High ground-water CO2 levels are mainly due to decomposition oforganic matter in the unsaturated soil within thegroundwater formation region, as indicated by averageδ13C values of calcite of –7 ± 1‰ (PDB-scale). BadLangensalza is situated on the slope of the Hainichmountainous region, which has a maximal elevation of495 m a.s.l. Today, the Hainich is the largest connectedwoodland of Germany that has recently become a na-tional park in which the European rain forest can de-velop without significant anthropogenic influence.

Rainfall in the basin is as low as 500 mm per year andthe regional climate has a more continental characterthan comparable sites in Central Europe as the

334

Thuringia basin is completely surrounded by moun-tains. Travertine formation could likely take place du-ring episodes of warm climate, during which the moun-tains were covered with dense woodland as today.During glacial times the region was subject to perma-frost and thus carbonate precipitation was inhibited.The travertine body in Bad Langensalza extends downslope over several square kilometers along the riverSalza from west through east, with the source of car-bonate saturated groundwater near Ufhoven in whichso called soil depression springs along the major faultsystem are still active (springs: small and largeGolcke). The travertine extension and texture has beenextensively investigated for economic reasons as it isstill extracted as a building stone. More than 50drillings have been performed during the past 3 de-cades. The travertine itself has formed on a glacialgravel bed of the river Salza and solely in the less thickeastern part of the deposits near the river Unstrut didtravertine form on alluvial clays. Its thickness is closeto 10 m in the depression at the City centre of BadLangensalza. Its texture was subdivided in 3 commer-cial categories, laminar dense travertine (mostly at thebase of the deposits), porous but laminar travertine(centre of the deposits) and loose alluvial mud inter-sected with travertine and tufa (topmost carbonate

layers and most eastern layers). From the description ofthe drillings those different textures are generally sepa-rated by deposition of reed-encrusted travertine and al-luvial clays likely precipitated in stagnant waters orduring flood events.

At present no carbonate precipitation occurs and thecarbonate rich waters are injected in the river Salza,which has cut its bed into the travertine formation. Thereason for the lack of modern carbonate precipitation isunknown, but is likely due to either deforestation as aconsequence of human activity in this region since pre-roman time, changes in regional climate and precipita-tion, or insufficient hydrological activity to furtherflood the extended travertine body.

Fehler (1998) has found minor cascade structures ofless than 1 m height in the bottom most dense laminardeposits (14 m to about 10 m depth), while the centrallayers and the top layers show little indication of cas-cades probably due to the fact that the initial topographyof the valley, especially the slight depression in the Citycentre has been evened out by carbonate formation.However, it is difficult to reconstruct in great detail thetravertine growth and structure given the complexity ofthe carbonate platform and its extension, as well as thefact that the travertine deposits have been extracted asbuilding stone at least since early medieval times.

335

Fig. 1: Location of travertine sites Bad Langensalza, Burgtonna, Weimar-Ehringsdorf and Bilzingsleben within the Thuringia Basin.Fig. 1 : Localisation des sites de travertin Bad Langensalza, Burgtonna, Weimar-Ehringsdorf et Bilzingsleben dans le bassin de Thuringe.

336L

abC

ode

Dep

th(m

)

238 U

(Fg/

g)δ23

4 Um

(‰)

232 T

h(F

g/g)

(230 T

h/23

8 U)

(230 T

h/23

2 Th)

Age (ka)

δ234 U

i

(‰)

Age

corr

(ka)

HD

-21

572.

40.

384

±0.0

0466

9±3

90.

0118

3±0

.000

150.

0933

±0.0

043

9.3

6.23

±0.3

468

1±4

5.75

0.60

HD

-21

582.

70.

386

±0.0

0171

2±9

0.01

144

±0.0

0022

0.18

23±0

.011

118

.812

.20

±0.7

673

7±9

HD

-21

613.

60.

354

±0.0

0173

1±8

0.02

617

±0.0

0012

0.11

58±0

.001

94.

87.

53±0

.12

746

±86.

300.

56H

D-

2163

4.2

0.27

5±0

.002

837

±21

0.08

921

±0.0

0088

0.15

86±

0.00

561.

59.

80±0

.37

860

±21

HD

-21

644.

50.

347

±0.0

0172

7±1

20.

0060

1±0

.000

100.

0940

±0.0

044

16.6

6.09

±0.2

974

0±1

25.

810.

43H

D-

2166

5.1

0.43

5±0

.001

736

±11

0.06

253

±0.0

0033

0.14

29±0

.002

83.

09.

33±0

.20

756

±11

HD

-21

675.

40.

476

±0.0

0270

7±2

00.

0193

2±0

.000

700.

2272

±0.

0198

17.1

15.4

0±1

.40

739

±20

HD

-L

S75.

40.

320

±0.0

0269

9±1

40.

1132

3±0

.000

220.

1130

±0.0

060

1.0

7.48

±0.4

071

4±1

4H

D-

2168

5.7

0.41

0±0

.001

935

±16

0.00

233

±0.0

0004

0.15

41±0

.004

382

.99.

01±0

.26

959

±16

8.92

0.36

HD

-21

696

0.44

7±0

.001

889

±17

0.00

440

±0.0

0004

0.13

66±0

.006

642

.58.

15±0

.41

910

±18

8.01

0.53

HD

-21

706.

30.

418

±0.0

0186

3±2

20.

0038

0±0

.000

030.

1346

±0.0

040

45.2

8.15

±0.2

688

3±2

28.

010.

39H

D-

2171

6.6

0.40

8±0

.001

900

±13

0.00

188

±0.0

0002

0.15

07±0

.003

010

0.1

8.97

±0.1

992

3±1

38.

900.

27H

D-

2172

6.9

0.42

4±0

.001

873

±17

0.00

355

±0.0

0005

0.15

09±0

.003

655

.09.

12±0

.23

896

±17

8.99

0.35

HD

-21

737.

20.

432

±0.

001

913

±16

0.02

092

±0.

0002

10.

1872

±0.

0055

11.8

11.1

6±

0.35

942

±16

HD

-21

747.

50.

390

±0.0

0191

6±2

30.

0078

3±0

.000

090.

1475

±0.0

058

22.4

8.70

±0.3

793

9±2

28.

400.

56H

D-

2175

7.8

0.42

3±0

.001

938

±23

0.00

368

±0.0

0007

0.15

28±0

.008

453

.69.

00±0

.51

962

±23

8.79

0.66

HD

-21

768.

10.

380

±0.0

0184

3±2

10.

0028

2±0

.000

030.

1359

±0.0

039

56.0

8.32

±0.2

786

3±2

18.

210.

38H

D-

LS6

8.2

0.35

6±0

.001

740

±19

0.01

292

±0.0

0019

0.14

40±0

.007

012

.19.

38±0

.48

760

±19

8.78

0.77

HD

-21

778.

40.

441

±0.0

0292

9±3

60.

0017

1±0

.000

080.

1414

±0.0

094

111.

38.

30±0

.61

951

±35

8.22

0.75

HD

-21

788.

70.

362

±0.0

0290

7±5

00.

0013

7±0

.000

060.

1432

±0.0

179

115.

28.

50±1

.10

930

±50

HD

-21

809.

30.

477

±0.0

0192

5±8

0.00

136

±0.0

0001

0.15

59±0

.002

516

7.0

9.17

±0.1

694

9±8

9.12

0.21

HD

-21

819.

60.

446

±0.0

0395

3±5

10.

0012

3±0

.000

080.

1500

±0.0

101

166.

08.

70±0

.63

980

±51

8.66

0.85

HD

-L

S5b

9.7

0.39

9±0

.002

854

±14

0.00

204

±0.0

0001

0.16

90±0

.012

010

1.1

10.4

0±0

.78

879

±14

10.2

90.

87H

D-

2182

9.9

0.38

2±0

.001

913

±90.

0014

5±0

.000

010.

1620

±0.0

033

130.

29.

60±0

.21

938

±99.

540.

27H

D-

2183

10.2

0.40

8±0

.001

919

±17

0.01

265

±0.0

0012

0.17

10±0

.004

716

.810

.12

±0.2

994

5±1

69.

650.

53H

D-

2184

10.5

0.41

9±0

.001

928

±16

0.00

979

±0.0

0005

0.17

97±0

.002

723

.510

.61

±0.1

895

6±1

610

.25

0.37

HD

-21

8510

.80.

446

±0.0

0191

2±2

80.

0062

4±0

.000

140.

1608

±0.0

081

35.2

10.0

0±0

.48

937

±28

9.33

0.71

HD

-21

8611

.10.

389

±0.0

0193

3±1

60.

0072

0±0

.000

070.

1784

±0.0

053

29.5

10.5

0±0

.31

961

±16

10.2

20.

50H

D-

LS3

t11

.50.

477

±0.0

0493

7±2

10.

0028

6±0

.000

010.

1740

±0.0

080

88.8

10.2

1±0

.48

964

±21

10.1

20.

63H

D-

LS2

c12

.30.

436

±0.0

0591

1±2

50.

0004

8±0

.000

000.

1664

±0.0

030

464.

99.

88±0

.22

937

±25

9.87

0.33

HD

-L

S2t

12.3

0.53

7±0

.005

960

±21

0.00

991

±0.0

0002

0.18

80±0

.005

031

.110

.93

±0.3

299

0±2

110

.65

0.51

HD

-21

9112

.60.

274

±0.0

0091

4±2

00.

0067

0±0

.000

140.

1847

±0.0

072

23.1

11.0

0±0

.48

943

±19

10.6

30.

69H

D-

2192

12.9

0.38

9±0

.001

1241

±34

0.00

379

±0.0

0005

0.22

30±0

.011

570

.011

.30

±0.6

212

81±3

411

.23

0.83

HD

-21

9313

.20.

315

±0.0

0188

7±2

50.

0026

0±0

.000

030.

1621

±0.0

070

60.1

9.74

±0.4

691

2±2

59.

630.

61H

D-

2195

13.8

0.34

0±0

.001

919

±17

0.02

087

±0.0

0011

0.19

84±0

.006

19.

911

.82

±0.4

095

0±1

810

.87

0.79

HD

-21

9614

.10.

334

±0.0

0193

1±1

70.

0414

2±0

.000

270.

2016

±0.0

062

5.0

11.9

4±0

.40

963

±17

All

erro

rsof

conc

entr

atio

ns,i

soto

pic

rati

os,a

ndag

esar

ere

port

edas

2σst

atis

tica

lunc

erta

inti

es.

()

brac

kets

indi

cate

acti

vity

rati

osca

lcul

ated

from

mea

sure

dat

omic

rati

osby

usin

gth

eha

lf-l

ives

give

nby

Che

nget

al.(

2000

)A

ge(k

a)=

Use

ries

age

calc

ulat

edfr

omm

easu

red

acti

vity

rati

osw

itho

utta

king

into

cons

ider

atio

nno

n-ca

rbon

ate

cont

amin

atio

nA

geco

rr(k

a)=

Use

ries

ages

calc

ulat

edco

nsid

erin

ga

detr

ital

(230 T

h/23

2 Th)

acti

vity

rati

oof

0.74

6δ23

4 Um

(‰)

=m

easu

red

Uis

otop

icra

tio

δ234 U

i(‰

)=

init

ialU

isot

opic

rati

oaf

ter

corr

ecti

onof

234 U

deca

y{M

alli

ck,2

002

#565

}H

D-L

Ssa

mpl

esco

rres

pond

tosa

mpl

esta

ken

onth

eou

tcro

pof

the

mai

ntr

aver

tine

body

inB

adL

ange

nsal

zaH

D-N

°=

sam

ples

from

the

dril

lcor

eta

ken

atsi

teM

ilch

gass

ein

dex

c=

cem

ent

inde

xt=

mic

rite

dom

inat

edtr

aver

tine

sam

ple

Tab

.1:U

-ser

ies

data

and

ages

oftr

aver

tine

sam

ples

from

Bad

Lan

gens

alza

,Thu

ring

iaB

asin

,Ger

man

y.Ta

b.1:

Don

nées

uran

ium

-tho

rium

etâg

esU

-Th

obte

nus

sur

letr

aver

tinde

Bad

Lang

ensa

lza,

Bas

sin

deTh

urin

ge,A

llem

agne

.

3 - SAMPLES AND METHODS

For this study we collected travertine and calcareoustufa samples at site Bad Langensalza, in an open quarryin the City centre and from a 14 m drill core recoveredin street Milchgasse about 50 m away from the maintravertine quarry. Both sites are representative for thearea of major carbonate deposition as travertine thick-ness is maximal with almost 10 m mostly banded lami-nar carbonate layers. A detailed cross section of thetravertine is presented in figure 2a and the texture of thedrill core in figure 2b. For our study we collected mate-rial mainly from layers where laminar growth bands aredominant, but a few samples of reed-encrusted tufa andone sample of secondary palisade cement were alsocollected for comparison. Using a titanium-bladed sawwe cut off slabs of some cm3 size, which were pre-cleaned using 0.1 N HNO3 and ultrasound. Sampleswere then carefully rinsed with Millipore water and

dried. Given the young age of the site, we could not ap-ply the micro-sampling technique recently establishedfor older travertine deposits (Mallick & Frank, 2002),thus solely bulk samples have been used for U-seriesdating. However, we prepared thin sections on all slabsfor polarization microscopy to identify the differentcalcite phases. The densest structures are micrite andspar, which build up the growth bands without orienta-tion of the primary minerals. All samples show variableamounts of pore fillings with second or third genera-tion calcite as cement. Also, detritus is present in allsamples and minor traces of residual organic matter.Aliquots of these bulk samples have been used to per-form mass spectrometric U-series dating.

About 300 mg of bulk material was used for U-seriesdating according to the procedures given in detail inFrank et al. (Frank et al., 2000) and Mallick and Frank(Mallick & Frank, 2002). Th and U measurements wereperformed after standard ion exchange chemistry

337

Fig. 2 A and B: Schematic cross section of the travertine body at Bad Langensalza. A) The orange arrow points out the drill location “Mil-chgasse” and the sampling at the travertine outcrop nearby. The size of the cross section corresponds to a few kilometers. B) Summary photo ofthe drill core and description of the travertine texture with depth (m) as obtained from the drill core. The black bar indicates the recovery of tra-vertine material. The observed texture is identical to the one observed in the nearby outcrop.Fig. 2 A et B : Coupe schématique du dépôt de travertin principal à Bad Langensalza. A) La flèche orange indique la position du forage “Milchgasse” etde la carrière de travertin voisine. B) Photos du forage et description de la texture du travertin par rapport à la profondeur obtenue à partir du forage. Labarre noire indique le taux de récupération de matériel. La texture observée est identique à celle de la carrière voisine

(Frank et al., 2000) on a multi-collector mass spec-trometer (Finnigan MAT262). For this study we used anew 233U/236U double spike (FIZ-1), having a 233U/236Uatomic ratio of 0.008439 ± 0.000012 (2σ-mean, N=20)close to the natural 234U/235U ratio of 0.007568 at radio-active equilibrium of (234U/238U). This 233U/236U-spikewas calibrated against reference material NBL-112aand contains 49.00±0.5 ng g-1 [236U]. The amount ofspike was added depending on the sample size to en-sure 234U/233U and 235U/236U atomic ratios closes to 1.Measurements of U reference material NBL-112a yieldon average δ234U values of –38.4 ± 3.3‰ (1σ-STD,N=21) for the course of our experiment, using the de-cay constant of 234U given by Cheng et al. (Cheng et al.,2000). These values agree with recently published va-lues of –37.6 ± 1.0‰ (1σ-STD) and –36.9 ± 1.7% (1σ-STD) by Delanghe et al. (Delanghe et al., 2002), andby Cheng et al. (Cheng et al., 2000). The 229Th-spikeused has a 230Th/229Th atomic ratio of 0.0000050 ±0.0000020 (2σ-mean) and contains 102.94 ± 0.20 ngg–1

[229Th], calibrated against HU-1 reference material byEisenhauer et al. (Eisenhauer et al., 1996). The externalreproducibility of U concentration measurements is0.5‰ (2σ), and that for Th is 0.6‰ (2σ). All measure-ments were performed on the electron multiplier usinga peak jump routine. Thermal fractionation correctionswere carried out for U-isotopes by using the 233U/236Uatomic ratio of the spike. Analytical blanks are veryconstant and amount to 10 ± 5fg 230Th, 0.2 ± 0.1ng232Th, and 0.1 ± 0.03ng 238U. These values are negligi-ble considering the Th and U abundance of our bulksamples.

4 - RESULTS

The travertine of Bad Langensalza has U concentra-tions between 0.27 to 0.54µgg-1 (tab. 1). The bottomlayers between 12.5 m and 14 m depth as well as thetopmost travertine layers between 2 m and 5 m depthhave slightly lower U concentrations of average0.343 ± 0.036 µgg–1, if compared to the central traver-tine body between 5.4 m and 12.5 m depth, which has0.43 ± 0.04 µgg–1 238U (fig. 3a). However, different car-bonate precipitates such as laminar travertine, reed-en-crusted tufa or pore cement do not exhibit significantdifferences in U-concentration.

The measured U-isotopic composition of the traver-tine shows a more pronounced pattern (fig. 3b). Fromthe bottom of the travertine deposit to about 5.4 mdepth the (234U/238U) activity ratio is rather constantwith values of 1.911 ± 0.023, i.e. δ234Uaverage = 911 ±23‰ (N=26). Secular variations are small (< 40‰)except for one sample at 12.9 m depth having a highlyelevated value of δ234U of 1240 ± 34‰ and one sam-ple at 8.2 m depth having a significantly lower valueof δ234U of 740 ± 19‰. Above 5.4 m depth the U-isoto-pic composition systematically drops to mean valuesof δ234Uaverage of 727 ± 50‰ (N=8). A correlation of U-isotopic composition and U-concentration is notevident.

Measured 232Th-concentrations are in general low(< 0.02µg g-1) and (230Th/232Th) activity ratios aremostly > 30 (fig. 3c). Only exceptions are one samplefrom the basal unit of the travertine having 0.04µgg-1

232Th and a few samples from the top having up to

338

Fig. 3A-C: U-series data. a) U concentration, b) U-isotopic composition, c) total 232Th concentration. The scatter and dotted lines corresponds totextural changes in the travertine structure given in the text.Fig. 3A-C : Mesures U/Th. A) Concentration de l’Uranium, B) composition isotopique de l’Uranium et C) concentration en Thorium total (232Th).

0.113 µgg–1 232Th. In the upper unit carbonate precipi-tation is intersected with alluvial clays, which causes alarge amount of detrital particles in the pore space ofthe carbonate fabric. In addition loose fine grain car-bonate sands are frequently intersected with densertravertine precipitates resulting in a significant loss ofmaterial during the drilling of core Milchgasse. Thus,232Th seems mostly derived from leaching of non-car-bonate particles, organic matter and fine grain carbo-nate sands. However, small amounts of 232Th andhenceforth 230Th may result directly from spring watersas Th can be transported within organic complexes oradsorbed to colloids. To test the sensitivity of such non-carbonate contamination on the U-series ages we haveused the measured 232Th concentration to correct mea-sured (230Th/238U) activity ratios solely for detrital230Th-contamination. We applied a very simplistiquecorrection model (Mallick & Frank, 2002) taking intoconsideration a (230Th/232Th) activity ratio of detritus of0.764±0.25, which corresponds to the mean continen-tal crust value (Wedepohl, 1995). Then ages and initialU-isotopic composition (δ234Ui) were calculated by usingthe “ISOPLOT” program provided by K. Ludwig (Lud-wig, 2001) and half-lives provided by Cheng et al.(2000). Age uncertainties are calculated from 2σ statis-tical errors of isotopic ratios and further propagatingthe error of our simple correction model into the agecalculation. The influence of non-carbonate contami-nation on the U isotopic composition is negligible andwas thus not applied here. The applied correctionmodel reduced ages by about 200 years in average andby less than 500 years for 80% of the data. Thus the in-fluence of non-carbonate contamination is rather smallor even insignificant for most of the samples. Further-more, the application of isochron correction models(Kaufman, 1993; Ludwig & Titterington, 1994) as fre-quently applied to contaminated carbonates (Frank etal., 2000; Lin et al., 1996; Sturchio et al., 1994) doesnot provide significant correlations between isotopicratios and does thus not identify a significant additionalsource of Th contamination. However, in 7 cases (seetab. 1) U-series ages have been rejected, due to high an-alytical uncertainties (> 10%) and/or because the car-bonate fabric did reveal dissolution features, organicmatter residues, strong precipitation of Fe-Mn coating,and/or a strong age correction. Dissolution likelycauses U leaching and thus U series open system be-havior, while non carbonate contamination mostly re-sults in strongly elevated Th values. Five of thesesamples are from the topmost travertine sequence ha-ving the most complex and variable texture and elevated232Th concentrations, as mentioned above. Thus thetiming of the topmost travertine sequence has to betaken with precaution due to the lack of precise ages.

Figure 4 summarizes the dating results. Overall, ac-cepted U-series ages range from 11,500 years at the baseof the travertine to ~ 6000 years at the topmost layers.Ages decrease from bottom to top as expected from thestructure of the travertine deposit, but secular variationsof U-series ages are significant throughout the profile.

Within the basal unit between 14.2 to 12 m depth U-series ages oscillate between 10,000 (N=3) and 11,500(N=4) years BP. It is important to note here that wehave measured at 12.3 m depth a travertine sample(LS2t) containing mostly primary micrite and a calcitecement (LS2c) reflecting solely the precipitation ofsecond or third generation calcite as laminated palisadecement. Both carbonate phases differ by 800 years inage with the cement sample being much younger. Thisresult clearly demonstrates the impact of secondaryfilling of pore space that causes secular variations inU-series ages and isotopic composition as we will dis-cuss later. At 12.3 to 12.6 m depth carbonate is precipi-tated at the selected sites in stagnant waters, causingthe deposition of reed-encrusted tufa, which has notbeen dated by U-series dating. Above this layer from12.6 m to roughly 9.5 m depth ages of the laminar tra-vertine are slightly younger compared to the unit below

339

Fig. 4: U-series ages of the Bad Langensalza travertine deposit: CoreMilchgasse and outcrop City Center. Gray and open squares denoteU-series ages that have been corrected for detrital Th-contamina-tion using a 230Th/232Th activity ratio of 0.746b ± 0.3. Triangles re-present uncorrected U-series ages. Five U-series measurementshave been rejected for this age-depth plot, because the travertine fa-bric of these samples clearly indicated strong deposition of Fe-Mnoxides, organic matter, and dissolution. Four of the five rejectedanalyses came from the upper travertine section in which carbonateprecipitation alternates with alluvial clays and in which travertineis highly porous and fine grain carbonate sands are frequent.Fig. 4: Ages U/Th obtenus pour le site Bad Langensalza: forage Milch-gasse et carrière voisine. Les carrés gris indiquent des âges U/Th corri-gés de la contamination en Thorium détritique en utilisant un rapportd’activité de 230Th/232Th-détritique de 0.746 ± 0.03. Les triangles indi-quent les âges U/Th non corrigés. Cinq âges U/Th ont été rejetés dans cegraphique âge – profondeur, car les échantillons ont révélé des tracesde dissolution, de précipitation d’oxyde de fer et manganèse et de ma-tière organique significatives. Quatre de ces cinq échantillons provien-nent de la section supérieure du travertin, très poreuse, dans laquelle laprécipitation de carbonate se fait en alternance avec un dépôt d’argilealluviale et de sable carbonaté.

ranging between 10,500 (N=6) to 9500 (N=2) years BPindicating almost continues accumulation of carbonatewith respect to the laminar travertine underneath it.However, U-series ages seem to oscillate as well be-tween higher and lower ages, but differences are smallwithin error. At 9.5 m loose material was flushed out ofthe drill core so the recovered travertine pieces did notfit together. The layer of loose carbonate sands was alsoobserved at similar depth (± 0.3 m) in the main quarry.At this depth we observe a first, but barely markedchange of U-series ages, as U-series ages above it (9.5to 5.4 m depth) now vary between ~ 8000 years (N=5)and ~ 9500 years (N=6). Thus, the carbonate sand layermay indicate a local interruption of carbonate accumu-lation and erosion some time between 9000 and 10,000years BP. Above this intermediate laminar travertineunit at < 5.4 m depth the texture of the travertinechanges dramatically as laid out in the description ofthe site.

This major change in travertine deposition clearlymarks an important growth interruption of at least 1000years duration between 8500 and 7000 years BP.Within the topmost section (5.4 m to 2 m depth) U-se-ries dating is not evident as carbonate accumulation isfrequently intersected with alluvial clays, carbonatesands, and organic matter rich mud causing significantperturbation of the U-series system. However, the mostreliable ages indicate carbonate accumulation likelybetween 7000 to 5800 years BP. No ages younger than5800 years have been determined probably indicatingthe timing ones carbonate deposition ceased “at the in-vestigated sites”.

5 - DISCUSSION

Travertine precipitation from carbonate rich ground-water is a complex process during which inorganic pre-cipitation through CO2 loss and bacterial inducedcarbonate precipitation interacts, in an environment ofhigh groundwater discharge. Within the Thuringia ba-sin travertine formation is frequent and well known forthe Pleistocene, as vast travertine deposits extendingover several square kilometers and having thicknessesof more than 10 to 20 m have been accumulated. Themost intensely studied sites are at Bilzingsleben,Burgtonna, Weimar, Weimar Ehringsdorf, and BadLangensalza, at which travertine was build up at leastduring the past 300 000 years. In particular at BadLangensalza travertine accumulation occurred duringmost of the Holocene originating from spring wateralong the major fault system of the Thuringia Basin. Incontrast, modern tufa formation is known solely in afew rare places and travertine formation is absent.Travertine accumulated in most of the places citedabove as so called meteogene travertine from springwater enriched in CO2 mostly from soil organic matterdecomposition as little is known about potentialoccurrence of thermal groundwater or other CO2

sources, such as of volcanic origin. Thus, travertine

accumulation must be related to some extend to the lo-cal to regional environmental conditions in theThuringia Basin, such as mean annual temperatures,precipitation, and vegetation cover. The fact that mo-dern travertine deposition is absent or occurs solely ascarbonate tufa must therefore be related to changes ofenvironmental conditions and/or human activity. Ourdetailed survey of U-series ages have revealed thattravertine formation in Bad Langensalza started duringthe early Holocene likely right after the Younger Dryascooling event at 11 000 years BP. Carbonate formationoccurred directly on the glacial gravel bed of the riverSalza, with no indication of a prior warm episode du-ring which alluvial clay could have been deposited.Thus no evidence was found for precipitation of car-bonate during Bølling-Allerød pre-Holocene warm cli-mate. Travertine formation continued throughout theearly Holocene until ~ 8000 years with a mean accumu-lation rate of 3 mm per year. However, U-series agesoscillate between higher and lower values within eachtextural travertine section, without obvious reasonfrom weathering, U-series open system behavior, ordepending on non-carbonate contamination and initial230Th. Consequently, the obtained growth record canonly be interpreted as initial carbonate formation beingfollowed by second and third generation carbonate pre-cipitation from pore fluids (cement or spar) infillingthe pore space and dead volume. It is in particular stun-ning that lower ages observed in a travertine texturalunit correspond to the highest ages of the unit justabove, clearly indicating that continues growth of tra-vertine layers results in successive infilling of the porespace underneath the presently active travertine accu-mulation. From such an at least bimodal carbonate pre-cipitation pattern we estimate the accumulation rate forthe basal units between 14.5 m and 9.5 m depth to be inthe order of 5 ± 1mm per year (11 000 – 10 000 years)and for the unit between 9.5 m and 5.5 m depth to be al-most twice as high at 8 ± 2 mm per year (9500 – 8500years). Then secondary infilling of pore space conti-nued likely for another 500 – 1000 years during whichthe carbonate platform at the two selected sites did notfurther develop or was partly eroded. Then, a growthinterruption of approximately 500 – 1000 years be-comes clearly evident as marked by the important tex-tural change from almost laminar travertine layers tolayers intersected with alluvial clays, organic matterrich mud, and carbonate sands that precipitated be-tween 7000 and 5500 years. However, based on the ob-tained age distribution at two randomly selected sites,which are in addition closely situated we can not infer acarbonate precipitation model for the far more complex3 dimensional structure of this travertine platform ex-tending over several square kilometers. Thus, all inter-pretations of carbonate deposition in light of changingenvironmental conditions seem presently limited bythe lack of constrain on the 3D age structure.

Besides the age distribution we have found that theU-isotopic composition, i.e. δ234U, does dramaticallychange within the travertine sequence (fig. 3b). For

340

most of the travertine profile δ234U values are restrictedto 920 ± 40‰, with two exceptions (see results, tab. 1).Then, at the marked change in texture at 5.4 m depth,values of δ234U drop to 750‰ and remain low for thetopmost section (fig. 3b). Such a systematic change inU-isotopic composition can only be explained throughchanges in U-isotopic composition of the source watersas U-fractionation during diagenesis and recoil pro-cesses can not explain such a significant drop in isoto-pic composition (–200‰) over such small time scales.In addition, changes of groundwater U-isotopic com-position through time are likely to occur as the U-isoto-pic composition depends on the weathering of U fromthe host rocks – here limestone and the overlying soil –which is driven by the groundwater residence time andits geochemical composition. U is weathered from ahost rock causing elevated δ234U in the case were the Usupplying minerals are solely partly dissolved. This is aconsequence of the alpha-recoil process during 238U de-cay within the host rock, which results in a higher mo-bility of the decay product 234U. In contrast, if the hostrock minerals supplying U are completely dissolved nofractionation of 238U and 234U occurs and thus δ234U ~0‰ or negative. In case were the supplying mineralsare partly dissolved the amount of 234U excess dependson the time groundwater stays in contact with the aqui-fer assuming a constant geochemical composition, i.e.constant weathering conditions. Here, reduced resi-dence times results in decreasing 234U excess. However,changes in groundwater flow pattern, i.e. changes inhost rock composition, or variable mixing of waters ofdifferent origin – shallow and deeper aquifers – canalso steer variable U-isotopic composition. Finally,changes in the physico-chemical composition (T, pH,alkalinity, pCO2, etc..) may cause variable weatheringconditions and therefore variable U-isotopic composi-tion for an even constant groundwater flow path anddischarge rate. Thus, from the U-isotopic compositionof travertine alone and having in hand little informationon the groundwater flow and geochemical compositionwe can barely identify whether changes in precipita-tion, vegetation cover, temperature, or alkalinityoccurred as marked by the drop in U-isotopic composi-tion and change in travertine texture. The texture of thecarbonate platform may however give hinds on the en-vironment in which carbonate was precipitated, but forthe moment we are lacking a detailed survey of the top-most travertine layers which could identify whether thecarbonate formed under different water flow condi-tions as compared to the deeper dense laminar traver-tine layers. Geochemical data not discussed here,however indicates little and in particular no systematicchanges in travertine composition. For example,87Sr/86Sr, Sr and Mg concentrations vary little through-out the travertine sequence. Moreover, stable C and Oisotopes have indicated significant variations with res-pect to the travertine texture, but no systematic changesas recorded by the U-isotopic composition have been

found (Frank, unpublished). Thus, to our presentknowledge the composition of travertine precipitatingwaters did not dramatically change with respect to thesignificant change in U-isotopic composition. Thus,one may speculate whether in fact U-isotopes record areduction of the groundwater residence time, whichwould mean that groundwater flow and discharge in-creased during times of declining carbonate formationand finally ended some 5500 years ago. However, theinterpretation of the U isotopic data in light of changesin groundwater flow remains speculative. Despite thelack of understanding how environmental conditionschanged during the final stage of travertine formation itremains stunning that changes in U-isotopic composi-tion are coincident to major changes in travertinetexture.

6 - CONCLUSION

Travertine deposits at Bad Langensalza (Thuringiabasin, Germany) grew with high accumulation rates ofup to 8 mm yr–1 during the early Holocene from 11 500years until about 5800 years. The resulting travertinerecord reflects initial carbonate precipitation followedby continues infilling of the initial pore space and deadvolume with second and third generation calcite mine-rals. This record confirms that northern European tra-vertine formation is restricted to short episodes ofparticular environmental conditions within a climatewarm stage as previously suggested by Frank et al.(Frank et al., 2000) and Mallick and Frank (Mallick &Frank, 2002) for marine isotope stage 5.5 and 5.3. Wefound no evidence of travertine accumulation withinthe late Holocene (< 5800 years) nor during the veryearly Holocene warming (Bølling-Allerød), whichmay reflect missing accessibility or absence of carbo-nate precipitation as a result of unfavorable environ-mental changes.

Significant changes in travertine texture are accom-panied by changes in U isotopic composition of the car-bonate indicating either changes of U supply from thehost rock, i.e. variable weathering, or changes ingroundwater residence time and flow path.

Thus systematic variations of the U isotopic compo-sition in older travertine deposits such as the one ofBurgtonna (MIS 5.5) (Mallick & Frank, 2002) arelikely due to changes in groundwater chemistry ratherthan reflecting U isotope fractionation related to U se-ries system opening through diagenesis and recoil pro-cesses.

In summary further investigations are needed tobetter constrain the complex 3D age structure of thetravertine deposits at Bad Langensalza and to elucidatethe impact of climate and environmental change duringthe Holocene on the geochemical composition of thetravertine layers.

341

ACKNOWLEDGEMENTS

We thank Klaus-Dieter Jäger and Dietrich Mania for having broadour attention to the investigated site Bad Langensalza. We also ackno-wledge Sybille Reuter, Ulrich Hambach, and Ronzon Mallick fortheir help collecting the samples. We are grateful to enterpriseTRACO at Bad Langensalza for their support to perform the drillingat site Milchgasse and the fact that we could easily access the outcropnearby the drill site. We further acknowledge the helpful commentsof P. Deschamps and an anonymous reviewer. This project was fun-ded by the “Deutsche Forschungsgemeinschaft” (DFG), grant: MA821/16-1 and the Heildelberger Akademie der Wissenschaft.

REFERENCES

ADAM K. D., 1986 - Fossilfunde aus dem Cannstatter Sauerwasser-kalken. Fundberichte aus Baden-Wuerrtemberg, 11, 25-61.

BLACKWELL B., & SCHWARCZ H.S., 1986 - U-series analysesof the lower travertine at Ehringsdorf, DDR. Quaternary Re-search, 25, 215-222.

BRUNNACKER K., HENNIG G.J., JUVIGNÉ E., LÖHR H.,URBAN B., & ZEESE R., 1982 - Der Kartstein-Travertin in dernördlichen Westeifel. Decheniana, 135, 179-204.

BRUNNACKER K., JÄGER K.D., HENNIG G.J., PREUSS J., &GRÜN R., 1983 - Radiometrische Untersuchungen zur Datierungmitteleuropäischer Travertinvorkommen. EAZ Ethnographisch-Archäologische Zeitung, 24, 217-266.

CHENG H., EDWARDS R.L., HOFF J., GALLUP C.D.,RICHARDS D.A., & ASMERON Y., 2000 - The half-lives ofuranium-234 and thorium-230. Chemical Geology, 169, 17-33.

DELANGHE D., BARD E., & HAMELIN B., 2002 - New TIMSconstraints on the uranium-238 and uranium-234 in seawatersfrom the main ocean basins and the Mediterranean Sea. MarineChemistry, 80, 79-93.

EISENHAUER A., ZHU Z.R., COLLINS L.B., WYRWOLLK.H., & EICHSTÄDTER R., 1996 - The last interglacial sea le-vel change : new evidence from the Abrolhos islands, West Austra-lia. Geologische Rundschau, 85, 606-614.

FEHLER A., 1998 - Die Travertine von Bad Langensalza Thüringen.Verlag Rockstuhl, Bad Langensalza.

FRANK N., 1997 - Anwendung der Thermionen-Massenspektro-metrie zur Uran-Reihendatierung pleistozäner, mitteleuropäis-cher Travertinvorkommen. Unpublished Dissertation thesis,Univeristy Heidelberg.

FRANK N., BRAUN M., HAMBACH U., MANGINI A., &WAGNER G., 2000 - Warm period growth of travertine during thelast interglaciation in southern Germany. Quaternary Research,54, 38-48.

GALLUP C. D., CHENG H., TAYLOR F.W., & EDWARDS R. L.,2002 - Direct determination of the timing of sea level change du-ring Termination II. Science, 295, 310-313.

GALLUP C.D., EDWARDS L., & JOHNSON R.G., 1994 - The ti-ming of high sea levels over the past 200,000 years. Science, 263,796-800.

GREWE K., 1987 - Hoehlenvermessung am Beispiel der Kartstein-hoehlen. Karst und Hoehle, 1986/1987, 155-163.

HARMON R.S., GLAZEK J., & NOWAK K., 1980 - 230Th/234Udating of travertine from the Bilzingsleben archeological site. Na-ture, 284, 132-135.

HENDERSON G.M., 2002 - Seawater (234U/238U) during the last800 thousand years. Earth and Planetary Science Letters, 199, 97-110.

HENDERSON G.M., & SLOWEY N. C., 2000 - Evidence from U-Th dating against northern hemisphere forcing of the penultimatedeglaciation. Nature, 404, 61-68.

IVANOVICH M., & HARMON R.S., 1992 - Uranium-series dise-quilibrium : Applications to earth, marine, and environmentalsciences. Clarendon Press, Oxford.

KAUFMAN A., 1993 - An evaluation of several methods for determi-ning 230Th/U ages in impure carbonates. Geochimica et Cosmochi-mica Acta, 57, 2303-2317.

LIN J.C., BROECKER W.S., ANDERSON R.F., HEMMING S.,RUBENSTONE J.L., & BONANI G., 1996 - New 230Th/U and14C ages from Lake Lahontan carbonates, Nevada, USA, and a dis-cussion of the origin of initial thorium. Geochimica et Cosmochi-mica Acta, 60, 2817-2832.

LUDWIG K.R., 2001 - ISOPLOT 2.49. Berkeley GeochronologicalCenter special Publication, 1a, 58.

LUDWIG K.R., & TITTERINGTON D.M., 1994 - Calculation of230Th/U isochrons, ages, and errors. Geochimica et CosmochimicaActa, 58, 5031-5042.

MALLICK R., 2000 - Entwicklung einer Mikrobeprobung zur Th/UDatierung und Anwendung an quartären Travertinen aus demThüringer Becken. University Heidelberg.

MALLICK R., & FRANK N., 2002 - A new technique for preciseUranium-series dating of travertine micro-samples. Geochimica etCosmochimica Acta, 66, 4261-4272.

REIFF W., 1986 - Die Sauerwasserkalke von Stuttgart. Fundberichteaus Baden-Wuerrtemberg, 11, 2-24.

SCHOLZ D., MANGINI A., & FELIS T., 2004 - U-series dating ofdiagenetically altered fossil reef corals. Earth and PlanetaryScience Letters, 218, 163-178.

SCHWARCZ H.O., GRÜN R., LATHAM A.G., MANIA D., &BRUNNACKER K., 1988 - The Bilzingsleben archaeologicalsite : new dating evidence. Archaeometry, 30, 5-17.

SCHWARCZ H.P., 1980 - Absolute age determination of archaeolo-gical sotes by uranium series dating of travertines. Archaeometry,22, 3-24.

SOMMERMEIER L., 1913 - Der Kartstein und der Kalktuff vonDreihmuehlen bei Eisfrey in der Eifel. Sonder-Abdruck aus denVerhandlungen des Naturhistorischen Vereins der preuss, Rhein-lande und Westfalens, 70, 302-333.

SRDOC D., OSMOND J.K., HORVATINCIC N., DABOUS A.A.,& OBELIC B., 1994 - Radiocarbon and uranium-series dating ofPlitvice lakes travertines. Radiocarbon, 36, 203-219.

STURCHIO N.C., PIERCE K.L., MURRELL M.T., & SOREYM.L., 1994 - Uranium-series ages of travertines and timing of thelast glaciation in the Northern Yellowstone Area, Wyoming-Mon-tana. Quaternary Research, 41, 265-277.

THOMPSON W.G., SPIEGELMAN M.W., GOLDSTEIN S.L., &SPEED R.C., 2003 - An open-system model for U-series age de-terminations of fossil corals. Earth and Planetary Science Letters,210, 365-381.

VILLEMANT B., & FEUILLET N., 2003 - Dating open systems bythe 238U-234U-230Th method : application to Quaternary reef terra-ces. Earth and Planetary Science Letters, 210, 105-118.

WEDEPOHL K.H., 1995 - The composition of the continental crust.Geochimica et Cosmochimica Acta, 59, 1217-1232.

342