magnesium isotopic composition of the moon

Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

ScienceDirect

Geochimica et Cosmochimica Acta 120 (2013) 1–16

Magnesium isotopic composition of the Moon

Fatemeh Sedaghatpour a,1, Fang-Zhen Teng a,⇑, Yang Liu b,2, Derek W.G. Sears c,3,Lawrence A. Taylor b

a Isotope Laboratory, Department of Geosciences and Arkansas Center for Space and Planetary Sciences, University of Arkansas,

Fayetteville, AR 72701, USAb Planetary Geosciences Institute, Department of Earth and Planetary Sciences, University of Tennessee, Knoxville, TN 37996, USA

c Department of Chemistry and Biochemistry and Arkansas Center for Space and Planetary Sciences, University of Arkansas, Fayetteville,

AR 72701, USA

Received 9 December 2012; accepted in revised form 20 June 2013; available online 8 July 2013

Abstract

The Mg isotopic compositions of 47 well-characterized lunar samples, including mare basalts, highland rocks, regolithbreccias, and mare and highland soils were measured to address the behavior of Mg isotopes during lunar magmatic differ-entiation, constrain the Mg isotopic composition of the Moon, and evaluate the degree of Mg isotopic fractionation betweenplanetary bodies. The d26Mg values range from �0.61 ± 0.03& to 0.02 ± 0.06& in 22 mare basalts, from �0.34 ± 0.04& to�0.18 ± 0.06& in 3 highland rocks, from �0.33 ± 0.05& to �0.14 ± 0.08& in 7 regolith breccias, from �0.23 ± 0.05& to�0.14 ± 0.07& in 6 highland soils, and from �0.41 ± 0.05& to �0.20 ± 0.09& in 9 mare soils. The limited Mg isotopic var-iation among bulk mare and highland soils and regolith breccias indicates negligible Mg isotope fractionation by lunar surfaceprocesses. By contrast, the large Mg isotopic fractionation between low-Ti and high-Ti basalts suggests the source heteroge-neity produced during fractional crystallization of the lunar magma ocean, with ilmenite having lighter Mg isotopic compo-sitions than olivine and pyroxene. Overall, the Moon has a weighted average Mg isotopic composition (d26Mg =�0.26 ± 0.16&) indistinguishable from the Earth (d26Mg = �0.25 ± 0.07&) and chondrites (d26Mg = �0.28 ± 0.06&),suggesting homogeneous Mg isotopic distribution in the solar system and the lack of Mg isotope fractionation during theMoon-forming giant impact.� 2013 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

Isotopic studies of terrestrial and lunar samples can pro-vide insight into the complex processes that govern plane-

0016-7037/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.gca.2013.06.026

⇑ Corresponding author. Present address: Isotope Laboratory,Department of Earth and Space Sciences, University of Washing-ton, Seattle, WA 98195, USA. Tel.: +1 206 543 7615.

E-mail address: [email protected] (F.-Z. Teng).1 Present address: Department of Earth and Planetary Sciences,

Harvard University, 20 Oxford Street, Cambridge, MA 02138.2 Present address: Jet Propulsion Laboratory, California Institute

of Technology, Pasadena, CA 91109, USA.3 Present address: Space Science and Astrobiology Division,

MS245-3, NASA Ames Research Center, Moffett Field, MountainView, CA 94035, USA.

tary accretion and differentiation. For instance, despitethe large variation of oxygen isotopes in the solar system,the Earth and the Moon have identical O isotopic compo-sitions (Clayton et al., 1973; Wiechert et al., 2001; Spicuzzaet al., 2007; Hallis et al., 2010; Liu et al., 2010a), suggestingthat materials in the proto-Moon and proto-Earth werefrom the same sources or well-mixed during the giant im-pact (Wiechert et al., 2001; Pahlevan and Stevenson, 2007).

Magnesium is a moderately refractory element in the so-lar system, with a condensation temperature of �1400 K(Lodders, 2003). In addition to the non-mass-dependentisotope anomalies produced by the decay of short-lived26Al to 26Mg (Lee et al., 1977), Mg isotopes can also frac-tionate mass dependently at high temperatures during con-densation and evaporation processes, as observed fromcalcium–aluminum-rich inclusions (CAIs) (e.g., Clayton


mailto:[email protected]


0

2

4

6

8

10

12

0 5 10 15 20 25

TiO

2 (w

t %)

MgO (wt %)

Mare basaltsHighland rocks

brecciasHighland soils Regolith

Mare soils

Fig. 1. Variation of TiO2 versus MgO contents for lunar samplesstudied here. Data are reported in Table 1.

2 F. Sedaghatpour et al. / Geochimica et Cosmochimica Acta 120 (2013) 1–16

and Mayeda, 1977; Wasserburg et al., 1977; Lee et al., 1979;Niederer and Papanastassiou, 1984; Niederer et al., 1985;Clayton et al., 1988) and chondrules (Galy et al., 2000;Young et al., 2002), and as shown by experimental and the-oretical studies (Davis et al., 1990; Richter et al., 2002,2007). Knowledge of Mg isotopic compositions of terres-trial and extraterrestrial materials, therefore, can poten-tially be used to study the origin, formation, anddifferentiation of planetary objects (Richter et al., 2002;Young et al., 2002; Norman et al., 2006; Teng et al.,2007, 2010; Wiechert and Halliday, 2007). For example,Wiechert and Halliday (2007) found the Earth has heavyMg isotopic compositions relative to chondrites. Theyinterpreted the Mg isotope heterogeneity in the solar systemas reflecting the physical separation and sorting of thechondrules and CAIs in the proto-planetary disk. However,more recent comprehensive studies on terrestrial basalts,komatiites and peridotites indicate that the Earth has aMg isotopic composition indistinguishable from chondriteswithin uncertainties (Teng et al., 2007, 2010; Handler et al.,2009; Yang et al., 2009; Bourdon et al., 2010; Chakrabartiand Jacobsen, 2010; Dauphas et al., 2010; Schiller et al.,2010; Huang et al., 2011; Liu et al., 2011; Pogge vonStrandmann et al., 2011; Xiao et al., 2013).

Studies of the Mg isotopic composition of the Moon canshed light on Mg isotopic distribution in the solar system.Compared to the Earth and chondrites, our understandingof the Mg isotopic composition of the Moon is still a sub-ject of controversy. In-situ analyses reveal homogeneousMg isotopic compositions in lunar mare basalts, glassesand impact melts, similar to the Earth and chondrites at� ± 0.2& amu�1 (2SD) (Warren et al., 2005; Normanet al., 2006). By contrast, Wiechert and Halliday (2007),by using solution method MC-ICPMS with a higher preci-sion (� ± 0.05& amu�1, 2SD), found that Mg isotopiccomposition of the Moon is �0.09& amu�1 heavier thanthat of the Earth, and both the Earth and Moon are heavierthan chondrites. With similar precision, Chakrabarti andJacobsen (2010) reported that the Moon has an identicalMg isotopic composition to chondrites. Nonetheless, thevalues for most terrestrial and chondrite samples reportedby Chakrabarti and Jacobsen (2010) and Wiechert andHalliday (2007) are different from those reported by othergroups (Teng et al., 2007, 2010; Handler et al., 2009; Yanget al., 2009; Bourdon et al., 2010; Dauphas et al., 2010;Schiller et al., 2010; Huang et al., 2011; Liu et al., 2011;Pogge von Strandmann et al., 2011; Xiao et al., 2013).

To better constrain the Mg isotopic composition of theMoon, understand behaviors of Mg isotopes during the lu-nar-forming giant impact and lunar magmatism, and eval-uate the extent of Mg isotopic heterogeneity in the solarsystem, we have studied 47 well-characterized samples fromall major types of lunar rocks returned by the Apollo 11, 12,14, 15, 16 and 17 missions. Our results show limited Mg iso-topic variation in lunar highland rocks, breccias, soil sam-ples and low-Ti basalts, which are similar to terrestrialbasalts and chondrites. By contrast, high-Ti basalts tendto have light Mg isotopic compositions, which may reflectthe source heterogeneity produced during differentiationof the lunar magma ocean (LMO).

2. SAMPLES

The samples investigated in this study are from the samecollection as those studied by Batchelor et al. (1997) andsome of those studied for oxygen and iron isotopes by Spic-uzza et al. (2007) and Liu et al. (2010a). They represent adiverse spectrum of sample types including highland rocks,mare volcanic rocks, regolith breccias, and mare and high-land lunar soils, with a broad range of lunar geological set-tings and geochemical properties (Fig. 1 and Table 1).Petrography, mineralogy, and major- and trace-elementcompositions of these samples are available in the lunarsample compendium (Meyer, 2004–2011).

2.1. Highland rocks

Highland rocks are classified into four major groups:ferroan-anorthosites, Mg-rich suite rocks, gabbros/norites,and alkali-suite rocks (e.g., Lucey et al., 2006). The lunarhighlands, decidedly older than the maria, were subjectedto heavy bombardment, and primary igneous rocks werebroken, mixed, and melted into various types of breccias.Three highland rocks from Apollo 14 and 16 were analyzedfor Mg isotopes in this study (Table 1).

2.2. Mare volcanic rocks

Mare volcanic rocks include various types of basalts thatwere derived from partial melting of cumulates originallyformed by the LMO crystallization. On the basis of thebulk-rock TiO2 content, mare basalts are classified intohigh-Ti basalts (HT) (>6 wt.% TiO2), low-Ti (LT) basalts(1.5–6 wt.% TiO2), and very low-Ti (VLT) basalts(<1.5 wt.% TiO2) (Neal and Taylor, 1992). The basalt sam-ples from the Apollo 12 and 15 missions are mostly low-Tibasalts, whereas those from the Apollo 11 and 17 missionsconsist mainly of high-Ti basalts. Based on mineralogy, andmajor- and trace-element compositions, Apollo 12 basalts

Table 1Chemical compositions of lunar samples investigated in this study.

Sample MgO SiO2 TiO2 FeO CaO

Volcanic rocks:Mare basalts

10020, 222 High-Ti basalt 7.7 40.2 10.6 18.9 12.010049, 89 High-Ti, High-K basalt 7.3 41.0 9.6 18.2 9.912002, 240 Low-Ti basalt 14.8 44.0 2.6 21.8 8.212005, 46 Low-Ti basalt 19.9 41.5 2.7 22.2 6.312051, 48 Low-Ti basalt 7.0 45.7 4.5 20.1 11.212011, 25 Low-Ti basalt 8.7 46.6 3.2 19.4 10.612021, 574 Low-Ti basalt 7.2 46.5 3.5 19.2 11.512052, 330 Low-Ti basalt 8.5 45.8 3.2 20.1 11.015058, 239 Low-Ti basalt 9.5 47.7 1.8 19.6 10.215075, 52 Low-Ti basalt NA NA NA NA NA15475, 187 Low-Ti basalt 8.7 48.3 1.7 19.8 10.315499, 13 Low-Ti basalt 9.1 47.7 1.8 20.2 10.315016, 221 Low-Ti basalt 11.1 44.1 2.3 22.0 9.415555 Low-Ti basalt 11.1 44.15 2.2 22.3 9.470035, 9 High-Ti basalt 9.9 37.8 13.0 18.5 10.071539, 17 High-Ti basalt 5.4 NA 8.6 19.1 12.175015, 1 High-Ti basalt 5.8 41.9 9.2 19.9 11.970215, 322 High-Ti basalt 8.7 37.9 12.9 19.7 10.875075, 171 High-Ti basalt 9.6 38.1 13.3 18.8 10.374255, 185 High-Ti basalt 10.7 38.4 12.5 17.8 10.370017, 524 High-Ti basalt 9.8 38.2 13.3 18.5 10.4

Mare pyroclastic glasses

74220, 758a 14.4 38.5 8.8 22.0 7.7

Highland rocks14310, 631 Impact melt 7.8 47.1 1.2 8.3 12.360315, 30 Coherent poikilitic impact melt 13.2 46.5 1.3 9.3 10.268415, 200 Clast-poor impact melt 4.4 45.5 0.3 4.1 16.3

Regolith breccias10018, 101 8.1 42.4 8.2 16.4 11.910019, 85 6.4 42.4 8.1 16.3 14.010021, 90 8.2 43.2 7.7 16.0 12.110046, 231 9.1 44.0 8.1 16.0 13.010048, 198 7.1 40.4 8.7 16.3 11.010060, 136 8.0 42.1 8.6 17.1 12.610065, 145 8.2 41.2 7.8 16.8 13.2

Highland soils60009, 621 Core, immature-submature 5.9 46.1 0.6 5.2 16.660009, 619 Core, immature-submature 5.9 46.1 0.6 5.2 16.660010, 1257 Core, submature-mature 5.9 46.1 0.6 5.2 16.660013, 319 Core, submature NA 54.8 NA NA NA60014, 197 Core, mature NA 54.8 NA NA NA61501, 120 submature 6.1 44.7 0.5 5.3 15.3

Mare soils12025, 351 Core NA NA NA NA NA12028, 83 Core NA NA NA NA NA70002, 461 Core, submature 10.1 42.3 5.8 15.4 11.270003, 537 Core, submature 10.0 42.6 5.7 15.3 11.070004, 574 Core, submature-mature 10.2 42.2 5.7 16.7 10.570005, 485 Core, mature 9.9 42.1 5.4 16.0 11.870007, 452 Core, submature 10.1 41.7 6.0 15.9 11.370008, 513 Core, immature 9.6 39.5 9.1 18.3 11.270009, 550 Core, submature 10.7 40.4 8.3 17.1 10.9

NA = not available. Major element data (wt.%) of lunar samples are from lunar sample compendium (Meyer, 2004–2011) and mare basaltdatabase (Neal, 2008). The data for 61501 is from Morris (1978) and Morris et al. (1983). Breccia maturity data are from Korotev and Morris(1993), Meyer, 2004–2011 and Morris et al. (1979).

a Orang glass beads.

F. Sedaghatpour et al. / Geochimica et Cosmochimica Acta 120 (2013) 1–16 3

Table 2Magnesium isotopic compositions of reference materials.

Samples d26Mga 2SDb d25Mga 2SDb

IL-lunarc 0.00 0.08 �0.02 0.06�0.07 0.05 �0.04 0.09�0.02 0.08 0.00 0.05�0.03 0.05 �0.03 0.03�0.01 0.07 �0.01 0.05

Averaged (n = 5) �0.03 0.03 �0.02 0.02

IL-granitee �0.06 0.06 �0.02 0.03�0.05 0.06 �0.02 0.04

0.01 0.07 �0.01 0.05�0.05 0.04 �0.05 0.03�0.01 0.05 �0.02 0.05

Average (n = 5) �0.04 0.02 �0.03 0.02

KH-1 olivinef �0.32 0.04 �0.16 0.04�0.29 0.03 �0.12 0.04�0.32 0.06 �0.13 0.03�0.28 0.06 �0.13 0.03�0.32 0.04 �0.16 0.04�0.33 0.05 �0.15 0.05�0.28 0.03 �0.12 0.07�0.28 0.06 �0.17 0.04

Average (n = 8) �0.30 0.01 �0.14 0.01

a dxMg = [(xMg/24Mg)sample/(xMg/24Mg)DSM3 � 1] � 1000,

where x = 25 or 26 and DSM3 is Mg solution made from pure Mgmetal (Galy et al., 2003).

b 2SD = 2 times the standard deviation of the population of n

(n > 20) repeat measurements of the standards during an analyticalsession.

c IL-lunar is a synthetic solution with Mg:Fe:Ca:Ti:Ni =1:4:4:4:0.1.

d Average is weighted average calculated based on inverse-variance weighted model using Isoplot 3.75–4.15.

e IL-granite is a synthetic solution with Mg:Fe:Al:Ca:Na:K:Ti:Ni = 1:5:30:5:10:20:0.1:0.

f KH-1 olivine is an in-house standard that has been analyzedthrough whole-procedural column chemistry and instrumentalanalysis with d26Mg = �0.27 ± 0.07 (2SD) and d25Mg = �0.14 ±0.04 (2SD) (Liu et al., 2010; Teng et al., 2010).


are subdivided into olivine, pigeonite, and ilmenite basalts(Neal and Taylor, 1992; Neal et al., 1994). The mare basaltsfrom Apollo 15 are sub-classified into quartz- and olivine-normative basalts (Chappell and Green, 1973). High-Timare basalts have been classified into ten subgroups basedon their mineralogy and trace-element characteristics (Nealand Taylor, 1992) (Tables 1 and 3).

Mare basalts investigated here include low-Ti basaltsfrom Apollo 12 (n = 6, ranging from olivine, pigeonite toilmenite basalts) and Apollo15 (n = 6, ranging fromquartz-normative to olivine-normative basalts) and high-Ti basalts from Apollo 11 (n = 2, one B3 type olivine basalt(low K, low La) and one A type ilmenite basalt (high K,high La)) and Apollo17 (n = 8, ranging from low K, lowLa to high K and high La basalts) missions (Tables 1 and3). In addition to these mare basalts, orange glass beadsfrom Apollo 17, 74220, were also studied here. Relativeto mare basalts, glass samples are more primitive and hencemay provide more direct information on composition of the

lunar mantle (Delano and Lindsley, 1983; Delano 1986;Shearer and Papike, 1993). Both mare basalts and volcanicglasses are referred as mare basalts in our later discussion.

2.3. Regolith breccias

Lunar breccias, produced by single or multiple impacts,are mixtures of mineral and rock fragments, crystallized im-pact melts, or glassy-impact melts. Polymict breccias areclassified into glassy melt, crystalline melt, fragmental,clast-poor impact melt, granulitic, dimict, and regolithbreccias (Stoffler et al., 1980). The lunar breccias studiedhere are Apollo 11 regolith breccias that consist mostly ofhigh-Ti mare basalt components (e.g., Rhodes and Blan-chard, 1982) (Table 1).

2.4. Lunar soils

Lunar soil is defined to be particles < 1 cm and blanketsthe surface of the Moon. It is formed and altered mainly bymicrometeorite (<1 mm) bombardment, with minor modifi-cations due to solar-wind and cosmic-rays particles(McKay et al., 1991). Composition of lunar soils rangesfrom basaltic to anorthositic with a small meteoritic com-ponent (<2 vol.%) (McKay et al., 1991). The samples stud-ied here include six highland soils from Apollo 16 drillcores, two mare soils from an Apollo 12 drive tube, and se-ven mare soils from Apollo 17 drill cores (Meyer, 2004–2011) (Table 1). The sampling depth for these samples var-ies from 50 cm to 3 m with different degree of maturity andmodal mineralogy, as well as chemical composition (Fryxelland Heiken, 1974; Meyer, 2004–2011).

3. ANALYTICAL METHODS

The Mg isotopic analyses were conducted at the IsotopeLaboratory of the University of Arkansas at Fayettevillefollowing a modified procedure after Yang et al. (2009),Li et al. (2010), and Teng et al. (2007, 2010).

3.1. Sample dissolution

Approximately 1–10 mg aliquots of each sample, basedon the Mg concentration, were dissolved in a mixture ofconcentrated HF–HNO3 (�3:1) (v/v), in Savillex screw-top beakers and heated overnight at a temperature of�160 �C, on a hot plate in a fume hood. The sample solu-tions were then dried at 120 �C. To achieve 100% dissolu-tion, the dried samples were dissolved in a mixture ofHCl–HNO3 (�3:1) (v/v) and heated at a temperature of160 �C for 1–2 days. After evaporation of the solutions intodry substance, the samples were refluxed with concentratedHCl and then evaporated into dryness again. The dried res-idues were finally dissolved in a mixture of 1 N HCl–0.5 NHF, and no residue was observed in the solution.

3.2. Column chemistry and instrumental analysis

The column chemistry involves two steps. The first stepwas performed mainly to separate Ti from Mg by following

Table 3Magnesium isotopic compositions of lunar samples.

Sample d26Mga 2SDb d25Mga 2SDb

Volcanic rocks:Mare basalts

10020, 222 High-Ti basalt �0.56 0.06 �0.28 0.04Replicatec �0.60 0.06 �0.30 0.03Replicate �0.54 0.06 �0.26 0.05Replicate �0.56 0.06 �0.31 0.05Replicate �0.58 0.08 �0.32 0.05Replicate �0.61 0.03 �0.31 0.05Replicate-Tid �0.59 0.04 �0.30 0.06Averagee �0.59 0.02 �0.30 0.0210049, 89 High-Ti, High-K basalt �0.39 0.09 �0.20 0.07Replicate �0.40 0.04 �0.20 0.04

Replicate �0.33 0.05 �0.18 0.03Replicate-Ti �0.35 0.05 �0.14 0.04Average �0.37 0.02 �0.18 0.02

12002, 240 Low-Ti basalt �0.28 0.06 �0.15 0.07Duplicatef �0.30 0.07 �0.13 0.04

Average �0.29 0.05 �0.14 0.0412005, 46 Low-Ti basalt �0.24 0.04 �0.11 0.03

Replicate �0.27 0.06 �0.14 0.07Average �0.25 0.03 �0.12 0.03

12051, 48 �0.27 0.06 �0.13 0.04Replicate �0.30 0.06 �0.17 0.03Average �0.28 0.04 �0.16 0.02

12011, 25 Low-Ti basalt �0.22 0.07 �0.11 0.04Duplicate �0.23 0.08 �0.10 0.05Replicate �0.22 0.06 �0.17 0.10Average �0.22 0.04 �0.11 0.03

12021,574 Low-Ti basalt �0.34 0.04 �0.16 0.04Replicate �0.34 0.06 �0.21 0.05

Replicate �0.28 0.08 �0.16 0.08Average �0.33 0.03 �0.18 0.0312052, 330 Low-Ti basalt �0.20 0.06 �0.10 0.06

Duplicate �0.18 0.08 �0.09 0.04Average �0.19 0.05 �0.09 0.04

15058, 239 Low-Ti basalt �0.32 0.04 �0.16 0.0415075, 52 Low-Ti basalt �0.29 0.05 �0.14 0.05


15475, 187 Low-Ti basalt �0.19 0.04 �0.12 0.03Replicate �0.22 0.08 �0.08 0.08Replicate �0.23 0.06 �0.14 0.04Replicate �0.22 0.03 �0.10 0.04Average �0.22 0.02 �0.12 0.02

15499, 13 Low-Ti basalt �0.21 0.09 �0.12 0.0515016, 221 Low-Ti basalt �0.24 0.06 �0.09 0.03


15555 Low-Ti basalt �0.01 0.05 0.01 0.03Replicate �0.04 0.06 �0.01 0.08

Replicate �0.03 0.08 �0.05 0.08Duplicate 0.02 0.06 0.01 0.04

Average �0.02 0.03 0.00 0.0270035, 9 High-Ti basalt �0.53 0.05 �0.27 0.05

Replicate �0.46 0.04 �0.22 0.03Replicate �0.40 0.09 �0.18 0.09Replicate-Ti �0.43 0.09 �0.24 0.07Average �0.47 0.07 �0.23 0.03


Table 3 (continued)


71539, 17 High-Ti basalt �0.40 0.06 �0.19 0.04Duplicate �0.44 0.06 �0.24 0.04Replicate-Ti �0.45 0.03 �0.24 0.03Average �0.44 0.02 �0.23 0.02

75015, 1 High-Ti basalt �0.53 0.05 �0.26 0.05Replicate �0.47 0.09 �0.25 0.09Replicate-Ti �0.48 0.06 �0.24 0.06Average �0.50 0.04 �0.25 0.04

70215, 322 High-Ti basalt �0.51 0.05 �0.24 0.05Duplicate �0.50 0.06 �0.23 0.04Average �0.51 0.04 �0.24 0.03

75075, 171 High-Ti basalt �0.40 0.06 �0.22 0.04Replicate �0.44 0.05 �0.25 0.05Replicate �0.40 0.09 �0.23 0.09Duplicate �0.42 0.06 �0.17 0.04Replicate-Ti �0.43 0.07 �0.23 0.02Average �0.42 0.03 �0.22 0.01

74255, 185 High-Ti basalt �0.60 0.05 �0.27 0.05Duplicate �0.53 0.06 �0.26 0.04Replicate-Ti �0.50 0.08 �0.24 0.07Replicate �0.52 0.04 �0.25 0.05Average �0.54 0.07 �0.25 0.02

70017, 524 High-Ti basalt �0.57 0.05 �0.27 0.05Replicate �0.51 0.04 �0.24 0.04Average �0.53 0.03 �0.25 0.03

Mare pyroclastic glasses

74220, 758 �0.27 0.05 �0.14 0.05Replicate �0.34 0.04 �0.21 0.03Average �0.31 0.03 �0.19 0.03

Highland rocks14310, 631 Impact melt �0.26 0.08 �0.12 0.04

Duplicate �0.26 0.08 �0.12 0.05Replicate �0.29 0.04 �0.14 0.03Average �0.28 0.03 �0.13 0.03

60315, 30 Plagioclase clasts �0.28 0.09 �0.13 0.05Replicate �0.34 0.04 �0.17 0.04

Average �0.34 0.03 �0.16 0.0368415, 200 Clast-poor impact melt �0.18 0.06 �0.08 0.04

Regolith breccias10018, 101 �0.14 0.08 �0.09 0.06

Replicate �0.22 0.05 �0.11 0.05Average �0.20 0.04 �0.10 0.0410019, 85 �0.29 0.05 �0.17 0.05

Replicate �0.27 0.04 �0.14 0.03Replicate �0.31 0.04 �0.16 0.03Replicate �0.27 0.09 �0.12 0.03Average �0.29 0.02 �0.14 0.02

10021, 90 �0.17 0.07 �0.08 0.0610046, 231 �0.16 0.07 �0.10 0.0610048, 198 �0.17 0.09 �0.08 0.0710060, 136 �0.21 0.07 �0.11 0.0610065, 145 �0.32 0.07 �0.16 0.05

Duplicate �0.33 0.05 �0.16 0.03Replicate �0.28 0.04 �0.11 0.04Replicate �0.29 0.08 �0.13 0.05Average �0.30 0.03 �0.14 0.02

Highland soils60009, 621 Core, immature-submature �0.18 0.06 �0.08 0.0760009, 619 Core, immature-submature �0.17 0.06 �0.08 0.07

(continued on next page)


Table 3 (continued)


Replicate �0.22 0.05 �0.12 0.05Average �0.20 0.04 �0.11 0.0460010, 1257 Core, submature-mature �0.16 0.08 �0.10 0.06

Replicate �0.22 0.05 �0.12 0.05Average �0.20 0.04 �0.11 0.0460013, 319 Core, submature �0.16 0.09 �0.08 0.07


60014, 197 Core, mature �0.14 0.07 �0.07 0.07Duplicate �0.19 0.09 �0.07 0.07Average �0.16 0.06 �0.07 0.05

61501, 120 Highland soil, submature �0.23 0.05 �0.08 0.05Duplicate �0.19 0.07 �0.09 0.05Replicate �0.15 0.06 �0.08 0.03Replicate �0.20 0.04 �0.09 0.01Average �0.19 0.03 �0.09 0.01

Mare soils12025, 351 Core �0.24 0.09 �0.12 0.07

Replicate �0.24 0.05 �0.15 0.05Replicate �0.29 0.08 �0.15 0.05Replicate �0.28 0.06 �0.11 0.03Average �0.26 0.03 �0.12 0.02

12028, 835 Core �0.25 0.07 �0.11 0.05Replicate �0.27 0.05 �0.12 0.05

Average �0.27 0.04 �0.12 0.0370002, 461 Core �0.21 0.08 �0.11 0.0570003, 537 Core �0.20 0.09 �0.11 0.07


70004, 574 Core, submature-mature �0.21 0.08 �0.11 0.07Replicate �0.26 0.05 �0.13 0.05

Average �0.24 0.04 �0.12 0.0470005, 485 Core, mature �0.33 0.07 �0.17 0.05

Replicate �0.30 0.06 �0.14 0.03Average �0.31 0.05 �0.15 0.0270007, 452 Core, submature �0.26 0.08 �0.13 0.0570008, 513 Core, immature �0.36 0.04 �0.16 0.03

Replicate �0.41 0.04 �0.19 0.04Replicate �0.41 0.05 �0.17 0.03Replicate �0.39 0.02 �0.19 0.03Replicate-Ti �0.38 0.05 �0.20 0.06Average �0.39 0.02 �0.18 0.02

70009, 550 Core, submature �0.29 0.08 �0.13 0.07Replicate �0.33 0.06 �0.16 0.04Replicate �0.34 0.05 �0.18 0.05Average �0.33 0.04 �0.16 0.03

a dxMg = [(xMg/24Mg)sample/(xMg/24Mg)DSM3 � 1] � 1000, where x = 25 or 26 and DSM3 is Mg solution made from pure Mg metal (Galy

et al., 2003).b 2SD = 2 times the standard deviation of the population of n (n > 20) repeat measurements of the standards during an analytical sessionc Replicate: Repeat the main-Mg column chemistry from the same stock sample solution.d Replicate-Ti: Repeat Ti column chemistry after the main-Mg column chemistry and isotopic analyses.e Average is weighted average calculated based on inverse-variance weighted model using Isoplot 3.75–4.15.f Duplicate: Repeated measurement of Mg isotopic ratios on the same solution.


the method described in Cook et al. (2006). In brief, asmall-volume (�2 ml) ion-exchange column packed withBio-Rad 200–400 mesh AG1-X8 resin, pre-cleaned (rinsed14 times column volume of 4 N HCl, 1 N HNO3 and18.2 MO Milli-Q� water) was used. The samples wereloaded on the resin in 0.2 mL of 1 N HCl–0.5 N HF. The

first 5 mL of 1 N HCl–0.5 N HF separated Mg from Ti.To assure the accuracy of our data, at least two referencematerials, with comparable Mg:Ti ratios, were included ineach batch of processed samples. To check the efficiencyof Ti column chemistry, the Ti/Mg ratio in purified Mgsample solutions was measured and was <0.05. For most


high-Ti samples, the Ti column chemistry was repeatedafter the main-Mg column chemistry.

In the second step, the main-Mg column chemistry, withBio-Rad 200–400 mesh AG50W-X8 resin in 1 N HNO3

media, was used to separate Mg from other matrix ele-ments, following previously established procedures (Tenget al., 2007, 2010; Yang et al., 2009; Li et al., 2010). Sampleand reference solutions dissolved in 1 N HNO3 (from thefirst step) were loaded on the resin and eluted through thecolumn by 1 N HNO3. Due to Mg isotope fractionationduring ion exchange chromatography (Russell and Pap-anastassiou, 1978; Chang et al., 2003; Teng et al., 2007),the Mg cut was determined to assure 100% Mg yield withno elemental interferences (Teng et al., 2007, 2010; Liet al., 2010). Magnesium yields for all lunar and standardsamples are >99%. The main-Mg column chemistry was re-peated for all lunar samples and reference materials to en-sure that the ratio of the concentration of any interferingcations to that of Mg is <0.05. Magnesium isotopic analy-ses were performed by the standard-bracketing methodusing a Nu Plasma MC-ICPMS. The concentrations ofthe sample solution and standard were matched within10% to avoid the matrix effect caused by concentration mis-match. The purified Mg sample solution (�250–300 ppb Mg in �3% (m/m) or �0.45 N HNO3 solution)was introduced into a “wet” plasma and Mg isotopes wereanalyzed in a low-resolution mode, with 26Mg, 25Mg, and24Mg measured simultaneously, in separate Faraday cups(H5, Ax, and L4).

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

-1.0 -0.8 -0.6 -0.4 -0.2 0.0

δ25 M

g

δ26Mg

Seawater

Earth

setirdnohC

Mare basaltsHighland rocks

Highland soilsMare soils

breccias Regolith

Fig. 2. Magnesium three-isotope plot of lunar samples analyzed inthis study. Compositions of the Earth and chondrites (Teng et al.,2010) as well as seawater (Ling et al., 2011) are also plotted forcomparison. The solid line represents a mass-dependent fraction-ation line with a slop of 0.505. Data are reported in Table 3.

3.3. Precision and accuracy

Long-term external precision and accuracy of Mg isoto-pic analyses at the Isotope Laboratory of the University ofArkansas were evaluated by full-procedural replicate anal-yses of seawater, rocks, minerals, synthetic solutions, andpure Mg Cambridge-1 standard solution (Teng et al.,2010; Li et al., 2010; Ling et al., 2011). The in-run precisionon the 26Mg/24Mg ratio for a single measurement of oneblock of 40 ratios was < ±0.02& (2SD). The long-termexternal precisions for d26Mg and d25Mg are better than±0.07& and ±0.06& (2SD), respectively (Li et al., 2010;Teng et al., 2010; Ling et al., 2011).

In order to evaluate the precision and accuracy of Mgisotope data during the course of this study, full proceduralreplicate analyses were performed on two synthetic solu-tions (IL-granite and IL-lunar) and one in-house standard(Kilbourne Hole (KH-1) olivine) (Table 2). IL-granite andIL-lunar yielded weighted average d26Mg values of�0.04 ± 0.02& (n = 5, Table 2) and �0.03 ± 0.03&

(n = 5, Table 2) respectively, which are identical to the ex-pected value of 0 within uncertainty. The olivine KH-1yielded a weighted average d26Mg value of�0.30 ± 0.01& (n = 8, Table 2), which agrees with the val-ues reported by Teng et al. (2010) and Liu et al. (2010b)(d26Mg = �0.27 ± 0.07&; 2SD, n = 16). The results ofthese synthetic and natural standard solutions confirm theaccuracy and reproducibility of our measurements duringthis course of analyses. In addition, full procedural replicateanalyses of samples and samples processed through

additional Ti-column chemistry yielded identical d26Mg val-ues (Table 3), further assuring the precision and reproduc-ibility of our data.

4. RESULTS

Magnesium isotopic compositions are reported in Ta-ble 2 for reference samples and in Table 3 for lunar samples.All samples analyzed here, in addition to those of the bulkEarth, chondrites, and seawater reported by Teng et al.(2010) and Ling et al. (2011) from the same lab, fall on asingle mass-dependent fractionation line with a best-fitslope of 0.505 (Fig. 2), which is consistent with previousstudies (e.g., Young and Galy, 2004; Li et al., 2010; Tenget al., 2010). d26Mg values range from �0.61 ± 0.03& to0.02 ± 0.06& in mare basalts, from 0.34 ± 0.04&

to �0.18 ± 0.06& in highland rocks, from �0.33 ± 0.05&

to �0.14 ± 0.08& in regolith breccias, from�0.23 ± 0.05& to �0.14 ± 0.07& in highland soil, andfrom �0.41 ± 0.05& to �0.20 ± 0.09& in mare soil sam-ples (Fig. 3). The range of Mg isotopic variation in lunarbasalts (�0.628& for d26Mg) is significantly larger thanthe analytical uncertainty (Table 3 and Fig. 3).

Generally, d26Mg values display a dichotomy betweenlow-Ti and high-Ti basalts. All low-Ti basalts, except forsample 15555, are similar in Mg isotopic compositions, withan average d26Mg value of �0.25 ± 0.10&. By contrast,high-Ti basalts display a larger variation, with an averaged26Mg value of �0.49 ± 0.14&. Basalt sample 15555 hasthe highest d26Mg value among all lunar samples(�0.02 ± 0.03&, 2SD, n = 4). Different aliquots of thissample also display distinct d18O values (5.443–5.769&)(Spicuzza et al., 2007; Liu et al., 2010a). This sample alsodisplays large chemical variations among different chips,which may be attributed to a biased sampling owing tothe large grain size (e.g., Ryder and Schuraytz, 2001).Therefore, the heavy Mg isotopic composition of sample

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

30 35 40 45 50 55 60 65 70 75

Low-Ti basalts

(a)

(b)

High-Ti basalts

Highland rocks

Mare soils

Mg #

δ26M

g

Highland soils

Low Ti High Ti

-0.7

-0.5

-0.3

-0.1

0.1

0 2 4 6 8 10 12 14

δ26M

g

TiO2 (wt %)

breccias Regolith

Fig. 3. Variation of d26Mg versus MgO (a) and TiO2 (b) contentsin lunar samples studied here. The solid line and grey bar representthe average d26Mg of �0.25& and two standard deviation of±0.07& for the Earth (Teng et al., 2010). Data are reported inTables 1 and 3.


15555 may have resulted from the heterogeneous mineraldistribution in this sample. Future studies are needed tobetter understand the mechanisms of producing heavy Mgisotopic composition.

Compared to previous studies of limited lunar samples,our Mg data are systematically lighter than those of Wiec-hert and Halliday (2007), but are heavier than those ofChakrabarti and Jacobsen (2010). The difference betweenlow-Ti and high-Ti basalts analyzed here is similar to thevariation reported by Wiechert and Halliday (2007), butdifferent from Chakrabarti and Jacobsen (2010). Differ-ences among these groups are likely the result of inter-lab-oratory biases, which makes it impractical to conduct cross-comparison among different datasets. However, given thereproducibility for each lab, the range of variation in lunarsamples is a meaningful comparison.

5. DISCUSSION

The Moon is thought to have been partially molten todepths of 400–500 km (Pritchard and Stevenson, 2000;

Canup, 2004) after the giant impact that formed the Moon(Hartmann and Davis, 1975; Cameron and Ward, 1976).Fractional crystallization of this magma ocean (LMO) ledto the formation of a feldspathic crust and a mafic mantle(Warren, 1985). A highly fractionated, late-stage liquidwith significant amounts of ilmenite also crystallized atthe end of the LMO solidification between the deep maficcumulates and the anorthositic crust (Warren, 1985;Shearer and Papike, 1999). This last portion of the highlydifferentiated LMO, termed urKREEP, crystallized beneaththe anorthositic crust (Warren, 1985).

In this section, we first discuss the behavior of Mg iso-topes during lunar soil formation and lunar magmatic dif-ferentiation. Then, we estimate the average Mg isotopiccomposition of the Moon and finally evaluate the extentof Mg isotopic heterogeneity in the early solar system.

5.1. Magnesium isotopic systematics of lunar soils and

regolith breccias

Lunar soils and regolith breccias are mixtures of diverserock types and components, hence may be more representa-tive of the average composition of the Moon than individ-ual crystalline rocks. Previous studies have found largegrain surface-correlated mass-dependent kinetic isotopefractionation effects e.g., for O, S, Si, K, Ca, Mg and Fe(e.g., Epstein and Taylor, 1971; Clayton et al., 1974; Russellet al., 1977; Esat and Taylor, 1992; Humayun and Clayton,1995a; Wiesli et al., 2003). These isotope fractionation ef-fects were sample surface correlated and not representativeof the bulk samples. They were interpreted as results of ionsputtering by the solar wind and galactic cosmic rays, and/or vaporization by micrometeorite impacts followed bygravitational mass selection (loss of the lighter isotopes)and re-deposition of the heavier isotopes (e.g., Switkowskiet al., 1977; Housley, 1979).

The lunar soils and regolith breccias analyzed in thisstudy, except for one sample, have similar Mg isotopic com-positions to the Earth, despite the difference in their bulk-rock compositions (Figs. 2 and 3). This conclusion is similarto previous studies of bulk soils (Esat and Taylor, 1992;Chakrabarti and Jacobsen, 2011), suggesting that irradia-tion or micrometeorite impacts does not affect Mg isotopiccompositions of bulk samples. Hence, these samples can beused to estimate the average Mg isotopic composition ofthe Moon (see details in Section 5.3).

5.2. Magnesium isotopic systematics of lunar basalts

The lunar basalts studied here have variable Mg isotopiccompositions. The low-Ti basalts with <6 wt.% TiO2 havesimilar d26Mg to those of highland rocks, regolith breccias,and mare and highland soils. By contrast, all high-Ti bas-alts (except for one) are light in Mg isotopic compositions(Fig. 3). This trend is similar to the one observed in Wiec-hert and Halliday (2007), despite the inter-laboratory bias.The dichotomy between low- and high-Ti basalts has beenreported in stable isotopes of O, Li, and Fe (Magnaet al., 2006; Seitz et al., 2006; Spicuzza et al., 2007; Crad-dock et al., 2010; Hallis et al., 2010; Liu et al., 2010a). These


differences were suggested to be due to heterogeneous man-tle sources for low- and high-Ti basalts, which are also con-sistent with petrological, mineralogical, geochemical, andradiogenic isotopic studies of these basalts (e.g., Neal andTaylor, 1992; Snyder et al., 1992; Beard et al., 1998; Sheareret al., 2006; Brandon et al., 2009).

Below, we first determine the major mineral phases thatcontrol MgO budget and then discuss the potential factorsthat control Mg isotopic variation in both low-Ti and high-Ti basalts.

5.2.1. Magnesium concentration variations in lunar basalts

For Mg-bearing minerals, clinopyroxene is the mostabundant phase, followed by opaques including ilmenite,chromian ulvospinel and armalcolite in the high-Ti basalts,while olivine and low-Ca pyroxene dominate the low-Tibasalts. Mineral abundances versus MgO plots reveal nosignificant correlation between pyroxene (41–71 vol.%)and MgO (Fig. 4b). However, there is clear positive corre-lation between olivine (0–30 vol.%) and MgO contents inlow-Ti basalts, and a positive correlation between opaqueminerals (15–25 vol.%) and MgO contents in high-Ti bas-alts (Fig. 4a and c). These positive correlations indicate thatolivine controls the MgO variation in low-Ti basaltswhereas opaque minerals play a key role in controllingthe MgO variation in high-Ti basalts, with a buffering roleof pyroxene in both basalt types.

Oliv

ine

(%)

Pyro

xene

(%)

Opa

que

(%)

0

10

20

30High-Ti basalts

Low-Ti basalts

20

40

60

80

0

10

20

30

3 6 9 12 15 18 21MgO (wt%)

-0.7 -0.5 -0.3 -0.1 0.1

δ26Mg

(a)

(b) (e)

(f)(c)

(d)

Fig. 4. Olivine, pyroxene, and opaque modal abundances (%) vs.MgO (wt.%) (a–c) and d26Mg (&) (d–f) of the whole rock samplesof lunar basalts investigated in this study. Mineral abundances forlunar samples are from lunar sample compendia (Meyer, 2004–2011).

Ilmenite [(Fe, Mg)TiO3], armalcolite [(Mg0.5�Fe0.5)Ti2-

O5], and chromian ulvospinel [(Mg, Fe)(Al, Cr)2O4] arethe Fe–Ti oxides containing Mg in lunar basalts. Armalco-lite and chromian ulvospinel can contain significant amountof MgO but they should not significantly affect the MgObudget of high-Ti basalts because of their low modal abun-dance (<0.5 wt.%, Warner et al., 1978). By contrast, themodal abundance of ilmenite in high-Ti basalts can be upto 28% (Papike et al., 1998). This, together with their highMgO content (3–8 wt.% e.g., Papike et al., 1998; Luceyet al., 2006; Van Kan Parker et al., 2011), may effectivelybe sufficient enough to affect the MgO contents of bulk bas-alts (e.g., Papike et al., 1998). Comparison of chemical com-positions of low-Ti and high-Ti basalts supports the aboveconclusion. Both CaO and SiO2 contents in low-Ti andhigh-Ti basalts decrease with MgO for samples investigatedin this study and literature, reflecting crystallization of dif-ferent mineral phases e.g., olivine, pyroxene and plagioclase(Fig. 5a and b). However, at a given MgO value, high-Tibasalts always have lower SiO2 contents than low-Ti basalts(Fig. 5b), indicating the presence of an additional phasethat is enriched in MgO but depleted in SiO2 in high-Ti bas-alts. On the other hand, TiO2 positively correlates withMgO in high-Ti basalts and displays no correlation withMgO in low-Ti basalts (Fig. 5c). If ilmenite contained littleMgO, a negative correlation between TiO2 and MgO wouldbe expected because of the dilution effect. These correla-tions thus further suggest that opaque minerals (i.e., ilmen-ite) not only control TiO2 contents but also play a key rolein MgO variations of high-Ti basalts.

5.2.2. Origins of light Mg isotopic composition of high-Ti

basalts

The mantle sources for low-Ti and high-Ti basalts werecumulates formed in LMO, with the high-Ti basalts fromlate-stage ilmenite-rich cumulates. The O and Fe isotopiccompositions of mare basalts have consistently revealedthat the ilmenite crystallized in the LMO may be the causefor the observed differences (Craddock et al., 2010; Liuet al., 2010a). Our data here also indicate that ilmenitehas a light Mg isotopic composition. d26Mg of lunar basaltsdoes not correlate with Mg# (index of crystallization)(Fig. 3a), or modal abundances of olivine and pyroxene(Fig. 4d and e), suggesting that crystallization of olivineand pyroxene did not significantly fractionate Mg isotopes.Instead, Mg isotopic compositions of lunar basalts decreasewith TiO2 and display a dichotomy between low- and high-Ti basalts (Fig. 3b). In addition, low- and high-Ti basaltsdisplay two clusters with high-Ti basalts enriched in opaqueminerals and light Mg isotopes (Fig. 4f). These suggest thatopaque minerals (i.e., ilmenite) have light Mg isotopic com-positions and shift high-Ti basalts to lower d26Mg values.

The isotopically light ilmenite may result from equilib-rium isotope fractionation in LMO. Studies of natural min-erals and theoretical calculation have found largeequilibrium inter-mineral Mg isotope fractionation (Han-dler et al., 2009; Yang et al., 2009; Young et al., 2009;Liu et al., 2010b, 2011; Schauble, 2011; Li et al., 2011;Wang et al., 2012; Xiao et al., 2013). This fractionation isqualitatively related to the different coordination number


(CN) of Mg bonded to O in minerals, as lower CN results instronger bonds and favors heavy isotopes (e.g., Urey, 1947;Chacko et al., 2001). The CN of Mg in ilmenite is 6 (Ray-mond and Wenk, 1971; Lind and Housley, 1972), identicalto that in olivine and pyroxene. Nonetheless, Schauble(2011) shows that significant Mg isotope fractionation oc-curs among different mineral species from carbonates, sili-cates, to periclase groups, with carbonates lighter andpericlase heavier than silicates, although CN of Mg in allthese mineral groups is 6. Hence, ilmenite as an oxide min-eral could be fractionated from olivine and pyroxene.

0 5 10 15 20 25

MgO (wt%)

40

50

60

SiO

2(w

t%)

Low-Ti basalts

High-Ti basalts

0

6

12

18

TiO

2(w

t%) Ol

(c)

(b)

ArmIlm

30

0

5

10

15

20

CaO

(wt%

)

Ol

An (a)

Fig. 5. CaO, SiO2 and TiO2 (wt.%) vs. MgO (wt.%) in lunar basaltsinvestigated in this study. Grey and light blue diamonds representlow- and high-Ti basalts from mare basalt database (Neal, 2008),respectively. The large pink square on the top left corner on panel(a) represents the anorthosite samples (P90% plagioclase) with thelowest MgO. The rectangular on the lower right corner on thepanel (a) represents the compositions of olivines from low- andhigh-Ti basalts (Papike et al., 1998). Arrows on panel (c) indicatethe accumulation lines for armalcolite (Arm), ilmenite (Ilm), andolivine (Ol) (Papike et al., 1998). Data are from lunar samplecompendia (Meyer, 2004–2011) and mare basalt database (Neal,2008).

Alternatively, the light Mg isotopic composition ofilmenite versus olivine and pyroxene may result from diffu-sion-driven kinetic isotope fractionation in the LMO. Re-cent studies have shown that kinetic Mg and Fe isotopefractionations can occur during chemical and thermal diffu-sion in melts and between melts and minerals (Richter et al.,2008, 2009a,b; Huang et al., 2009; Teng et al., 2011). Cou-pled studies of Mg and Fe isotopes can tell chemical diffu-sion from thermal diffusion as a positive correlationbetween Mg and Fe isotope fractionations is expected forthermal diffusion and a negative correlation is for chemicaldiffusion. Since no d26Mg data are available for ilmenite,comparison of d26Mg and d56Fe data for lunar basalts isused here. Plots of d26Mg and d56Fe with TiO2 displayopposite trends and indicate a negative correlation betweend26Mg and d56Fe in low- and high-Ti basalts (Fig. 6).Although the negative correlation may indicate that chem-ical diffusion can play a role in controlling Mg and Fe iso-topic compositions of lunar basalts, petrological studiesshow that ilmenite in high-Ti basalts is one of the early crys-tallized phases, being in equilibrium with the melt andpyroxene, and it has probably not undergone post-mag-matic re-equilibration (e.g., Warner et al., 1978; Sheareret al., 2006).

Although it is beyond the scope of this study, measure-ments of Mg isotopic compositions of ilmenite and othercoexisting silicates in low- and high-Ti basalts need to be

-0.1

0.0

0.1

0.2

0.3

Liu et al. 2010a

Poitrasson et al., 2004

Weyer et al., 2005

Wiesli et al., 2003

TiO2

δ56Fe

-0.7

-0.5

-0.3

-0.1

0.1

0 2 4 6 8 10 12 14 16

This study

Wiechert and Halliday, 2007

TiO2 (wt %)

26M

g

Fig. 6. d56Fe (Wiesli et al., 2003; Poitrasson et al., 2004; Weyeret al., 2005; Liu et al., 2010a) and d26Mg of lunar basalts (thisstudy, Table 3; Wiechert and Halliday, 2007) versus their wholerock TiO2 contents.


undertaken to solve the issue of whether Mg in lunar ilmen-ite is significantly fractionated from other minerals and theunderlying mechanisms.

5.3. Magnesium isotopic composition of the Moon

Our knowledge on the structure and composition of theMoon comes from samples returned from Apollo and Lunamissions, lunar meteorites, and remote sensing and seismicdata (e.g., Taylor, 1982; Gillis et al., 2004; Weber et al.,2011). Based on these data, the Moon is believed to be dif-ferentiated with a feldspathic crust, a mafic mantle, and apartially molten core. The upper crust is more feldspathicwith more gabbro and anorthositic gabbro and the lowercrust is more mafic with more norite and anorthositic norite(Wieczorek et al., 2006). The upper mantle is enriched inpyroxene while the lower mantle is enriched in Mg-richolivine (e.g., Warren, 1985; Shearer and Papike, 1999).The lunar crustal composition can be estimated by analysisof lunar samples such as ferroan anorthositic, magnesiumand alkali suites, etc., as well as remote sensing data (e.g.,Jolliff et al., 2000; Greenhagen et al., 2010; Chauhanet al., 2012). Unlike crust samples, no sample of lunar man-tle has been discovered to date. The composition of the lu-nar mantle has been mainly studied through the analysis ofpartial melts of the lunar mantle, i.e., lunar basalts and vol-canic glasses (Wieczorek et al., 2006). This method may notgive an accurate estimate of the mantle source compositionbecause of elemental and isotopic fractionation during theinitial crystallization of the LMO. Pyroclastic glasses, withhigher Mg, Ni, and Mg#, compared to the lunar basalts, aresuggested to better constrain the lunar mantle compositionbecause they represent the least differentiated melts (Delanoand Lindsley, 1983; Delano, 1986; Shearer and Papike,1993).

To date, four approaches have been utilized to estimatethe isotopic composition of the bulk Moon: (1) the arithme-tic mean of different types of lunar samples (e.g., Poitrassonet al., 2004); (2) the mean of the most primitive lunar sam-ples, like green picritic glass 15426 (Wiechert and Halliday,2007) or less fractionated material like orange glass 74220and quartz-normative basalts (15058 and 15475) (Magnaet al., 2006); (3) the mean of lunar samples excluding anom-alies (e.g., Weyer et al., 2005); or 4) the mean of differentsample weighted by their volumetric proportions (Liuet al., 2010a). Since different samples contribute to differentvolumetric proportions of the Moon, the arithmetic meanof these samples may not provide a precise estimate ofthe bulk isotopic composition of the Moon. In this study,we first estimate the Mg isotopic compositions of the lunarcrust and mantle separately; then estimate the average com-position of the bulk Moon, based on the weight percentageof the lunar crust and mantle.

Because of the limited Mg isotope fractionation duringtheir formation processes, lunar highland impact-meltrocks and highland soil samples from Apollo 16 are usedto estimate the lunar crust composition. The MgO-weightedaverage d26Mg value (�0.24 ± 0.07&) of these samples(14310, 60315, 68415, 60009, 60010, and 61501) is consid-ered as the average Mg isotopic composition of the lunar

crust. Since MgO contents of samples 60013 and 60014are not available, these samples are not included in the cal-culation. All regolith breccias analyzed in this study arefrom mare regions, which are dominantly basaltic in bulkcomposition. Hence, these breccias, high-Ti basalts, andtwo mare soil samples (70008 and 70009) with >6 wt.%TiO2 are used to estimate the MgO-weighted averaged26Mg value of high-Ti basalts. The other mare soil samplesfrom Apollo 17 with low TiO2 contents are added to low-Tibasalts to estimate their MgO-weighted average d26Mg va-lue. MgO contents of lunar soil 12025 and 12028 are notavailable; hence, these samples are not used to estimatethe MgO-weighted average d26Mg value of low-Ti basalts.These average d26Mg values of low-Ti (excluding 15555)(�0.25 ± 0.05&) and high-Ti (�0.36 ± 0.14&) basalts arethen used to estimate Mg isotopic composition of theirmantle sources. Based on the Clementine remote sensingdata for the lunar surface, low-Ti basalts are 90% of allmare basalts exposed on the lunar surface (Giguere et al.,2000). Using this estimated proportion and the average val-ues of low- and high-Ti mantle sources, the weighted aver-age d26Mg of the lunar mantle is estimated to be�0.26 ± 0.15&. By combining Mg isotopic compositionsof the lunar crust and mantle and their weight percentage(7% and 93%, respectively; Warren, 2005), we estimatethe weighted average d26Mg value of the bulk Moon tobe �0.26 ± 0.16&. This value is considered as the best esti-mate of the average Mg isotopic composition of the Moon.

5.4. Chondritic Mg isotopic composition of the Moon and

Earth

Studies of isotopic compositions of the Moon, Earthand chondrites can help to better understand the solar neb-ula conditions, proto-planetary disc processes, and the ori-gin and evolution of the Earth–Moon system (e.g.,Wiechert et al., 2001; Poitrasson et al., 2004; Wiechertand Halliday, 2007; Paniello et al., 2012). While it is wellestablished now that the Earth has a chondritic Mg isotopiccomposition (Teng et al., 2007, 2010; Handler et al., 2009;Yang et al., 2009; Bourdon et al., 2010; Chakrabarti andJacobsen, 2010; Dauphas et al., 2010; Schiller et al., 2010;Huang et al., 2011; Liu et al., 2011; Pogge von Strandmannet al., 2011; Xiao et al., 2013), there is still uncertainty onthe Mg isotopic composition of the Moon because of thelimited data (Warren et al., 2005; Norman et al., 2006;Wiechert and Halliday, 2007; Chakrabarti and Jacobsen,2010). The wide range of lunar samples measured in thisstudy, and those of terrestrial and chondritic samples ana-lyzed in the same laboratory (Yang et al., 2009; Tenget al., 2010; Liu et al., 2011; Xiao et al., 2013), allows aninternally-consistent comparison of Mg isotopic composi-tions among different planetary materials.

The Mg isotopic composition of the Moon (d26Mg =�0.26 ± 0.16&) estimated here is indistinguishable fromthat of the Earth (d26Mg = �0.25 ± 0.07&; 2SD, n = 139)and chondrites (d26Mg = �0.28 ± 0.06&; 2SD, n = 38)measured in the same laboratory (Teng et al., 2010)(Fig. 7). Hence, our study suggests that the Moon andthe Earth have chondritic Mg isotopic compositions,

5 10 15 20 25

Freq

uenc

y

26Mg

15

30

45

-0.65 -0.45 -0.25 -0.05 0.15

Chondrites x = -0.28 ± 0.06 (Teng et al., 2010)

60 Earth

x = -0.25 ± 0.07 (Teng et al., 2010)

Moon x = -0.26 ± 0.16

(This study)

5

0

10

15

Fig. 7. Histogram of Mg isotopic compositions of the Moon,Earth and chondrites. The vertical dashed line and yellow barrepresent the average d26Mg of �0.28& and two standarddeviation of ±0.06& for chondrites (Teng et al., 2010).


reflecting extensive early disk mixing processes and homo-geneity of Mg isotopes in the solar system. The chondriticMg isotopic compositions of the Moon and the Earth alsoindicate that the Moon-forming giant impact did not frac-tionate Mg isotopes, in contrast from the models proposedfor Fe isotopes (Poitrasson et al., 2004), but consistent withthe lack of Li and K isotope fractionations in theEarth–Moon system (Humayun and Clayton, 1995b;Magna et al., 2006).

6. CONCLUSIONS

This study presents comprehensive Mg isotopic data forlunar samples including most types of available samples todate. The main conclusions are:

(1) d26Mg values range from �0.61 ± 0.03& to0.02 ± 0.06& in mare basalts, from 0.34 ± 0.04&

to �0.18 ± 0.06& in highland rocks, from�0.33 ± 0.05& to �0.14 ± 0.08& in regolith brec-cias, from �0.23 ± 0.05& to �0.14 ± 0.07& in high-land soils, and from �0.41 ± 0.05& to�0.20 ± 0.09& in mare soil samples.

(2) The bulk mare and highland soils and regolith brec-cias have similar Mg isotopic compositions to thoseof the highland rocks, low-Ti basalts and the Earth,which indicates that their formation processes didnot significantly fractionate Mg isotopes.

(3) High-Ti basalts are on average isotopically lighterthan low-Ti basalts, mostly reflecting their heteroge-neous cumulate sources produced during the lunarmagmatic differentiation. The chemical and isotopicstudies suggest that high abundance of ilmenite with

possible lighter Mg isotopic composition than coexis-ting olivine and pyroxene may cause the light Mg iso-topic composition of high-Ti basalts.

(4) Using the average Mg isotopic compositions of thelunar crust and mantle and their volumetric propor-tions, the weighted average Mg isotopic compositionof the Moon is estimated to be �0.26 ± 0.16& ford26Mg and �0.13 ± 0.07& for d25Mg.

(5) The Moon (d26Mg = �0.26 ± 0.16&) has indistin-guishable Mg isotopic composition from the Earth(d26Mg = �0.25 ± 0.07&) and chondrites (d26Mg =�0.28 ± 0.06&), suggesting the homogeneity of Mgisotopes in the solar system and the negligible Mgisotope fractionation during the Moon-forming giantimpact.

ACKNOWLEDGEMENTS

We are grateful to Wang-Ye Li, Shan Ke and Sheng-Ao Liu forhelp in the lab, Kang-Jun Huang for discussion, and WalterGraupner for his help in maintaining the ICP-MS. The constructivecomments from Dimitri Papanastassiou, Paul Tomascak and ananonymous reviewer significantly improved the manuscript. Wealso thank Dimitri Papanastassiou for the discussion and editorialhandling of this manuscript. This work is supported by the Na-tional Science Foundation (EAR-0838227, EAR-1056713 andEAR-1340160) to FZT and NASA Cosmochemistry grant(NNX11AG58G) to LAT. YL participated the revision of this pa-per at Jet Propulsion Laboratory, which is managed by Caltech un-der the contract with NASA.

REFERENCES

Batchelor J. D., Symes S. J. K., Benoit P. H. and Sears D. W. G.(1997) Constraints on the thermal and mixing history of lunarsurface materials and compositions with basaltic meteorites. J.

Geophys. Res. 102, 19321–19334.Beard B. L., Taylor L. A., Scherer E. E., Johnson C. M. and

Snyder G. A. (1998) The source region and melting mineralogyof high-titanium and low-titanium basalts deduced from Lu-Hfisotope data. Geochim. Cosmochim. Acta 62, 525–544.

Bourdon B., Tipper E. T., Fitoussi C. and Stracke A. (2010)Chondritic Mg isotope composition of the Earth. Geochim.

Cosmochim. Acta 74, 5069–5083.Brandon A. D., Lapen T. J., Debaille V., Beard B. L., Rankenburg

K. and Neal C. (2009) Re-evaluating 142Nd/144Nd in lunar marebasalts with implications for the early evolution and bulk Sm/Nd of the Moon. Geochim. Cosmochim. Acta 73, 6421–6445.

Cameron A. G. W. and Ward W. R. (1976) The origin of theMoon. Proc. 7th Lunar Sci. Conf. 120–122.

Canup R. M. (2004) Dynamics of lunar formation. Annu. Rev.

Astron. Astrophys. 42, 441–475.Chacko T., Cole D. R. and Horita J. (2001) Equilibrium oxygen,

hydrogen and carbon isotope fractionation factors applicable togeological systems. Rev. Mineral. Geochem. 43, 1–82.

Chakrabarti R. and Jacobsen S. B. (2010) The isotopic composi-tion of magnesium in the inner solar system. Earth Planet. Sci.

Lett. 293, 349–358.Chakrabarti R. and Jacobsen S. B. (2011) The isotopic composi-

tion of magnesium in bulk lunar soils. Lunar Planet. Sci. XLII.Lunar Planet. Inst., Houston. #2006 (abstr.).

http://refhub.elsevier.com/S0016-7037(13)00360-8/h0005


























Chang V. T.-C., Makishima A., Belshaw N. S. and O’Nions R. K.(2003) Purification of Mg from low-Mg biogenic carbonates forisotope ratio determination using multiple collector ICP-MS. J.

Anal. At. Spectrom. 18, 296–301.Chappell B. W. and Green D. H. (1973) Chemical compositions

and petrogenetic relationships in Apollo 15 mare basalts. Earth

Planet. Sci. Lett. 18, 237–246.Chauhan P., Kaur P., Srivastava N., Bhattacharya S., Ajai, Kiran

Kumar A. S. and Goswami J. N. (2012) Compositional andmorphological analysis of high resolution remote sensing dataover central peak of Tycho crater on the Moon: Implication forunderstanding lunar interior. Curr. Sci. 102, 1041–1046.

Clayton R. N. and Mayeda T. K. (1977) Correlated oxygen andmagnesium isotope anomalies in Allende inclusions: I. Oxygen.Geophys. Res. Lett. 4, 295–298.

Clayton R. N., Hurd J. M. and Mayeda T. K. (1973) Oxygenisotopic compositions of Apollo 15, 16, and 17 samples, andtheir bearing on lunar origin and petrogenesis. Proc. 4th Lunar

Sci. Conf. 1535–1542.Clayton R. N., Mayeda T. K. and Hurd J. M. (1974) Loss of

oxygen, silicon, sulfur, and potassium from the lunar regolith.Proc. 5th Lunar Sci. Conf. 1801–1809.

Clayton R. N., Hinton R. W. and Davis A. M. (1988) Isotopicvariations in the rock forming elements in meteorites. Philos.

Trans. R. Soc. London A 325, 483–501.Cook D. L., Wadhwa M., Janney P. E., Dauphas N., Clayton R.

N. and Davis A. M. (2006) High precision measurements ofnon-mass-dependent effects in nickel isotopes in meteoriticmetal via multicollector ICPMS. Anal. Chem. 78, 8477–8484.

Craddock P. R., Dauphas N. and Clayton R. N. (2010) Mineral-ogical control on iron isotopic fractionation during lunardifferentiation and magmatism. Lunar Planet. Sci. XXXXI.Lunar Planet. Inst., Houston. #1230 (abstr.).

Dauphas N., Teng F.-Z. and Arndt N. T. (2010) Magnesium andiron isotopes in 2.7 Ga Alexo komatiites: Mantle signatures, noevidence for Soret diffusion, and identification of diffusivetransport in zoned olivine. Geochim. Cosmochim. Acta 74,3274–3291.

Davis A. M., Hashimoto A., Clayton R. N. and Mayeda T. K.(1990) Isotope mass fractionation during evaporation ofMg2SiO4. Nature 347, 655–658.

Delano J. W. (1986) Pristine lunar glasses: Criteria, data, andimplications. J. Geophys. Res. 91, D201–D213.

Delano J. and Lindsley D. (1983) Mare glasses from Apollo 17:Constraints on the Moon’s bulk composition. J. Geophys. Rev.

88, B3–B16.Epstein S. and Taylor H. P. (1971) O18/O16, Si30/Si28, D/H and

C13/C12 ratios in lunar samples. Proc. 2nd Lunar Sci. Conf.1421–1441.

Esat T. M. and Taylor S. R. (1992) Magnesium isotope fraction-ation in lunar soils. Geochim. Cosmochim. Acta 56, 1025–1031.

Fryxell R. and Heiken G. H. (1974) Preservation of lunar coresamples: Preparation and interpretation of three-dimensionalstratigraphic section. Proc. 5th Lunar Sci. Conf. 935–966.

Galy A., Young E. D., Ash R. D. and O’Nions R. K. (2000) Theformation of chondrules at high gas pressure in the solarnebula. Science 290, 1751–1753.

Galy A., Yoffe O., Janney P. E., Williams R. W., Cloquet C., AlardO., Halicz L., Wadhwa M., Hutcheon I. D., Ramon E. andCarignan J. (2003) Magnesium isotope heterogeneity of theisotopic standard SRM980 and new reference materials formagnesium-isotope-ratio measurements. J. Anal. At. Spectrom.

18, 1352–1356.Giguere T. A., Taylor G. J., Hawke B. R. and Lucey P. G. (2000)

The titanium contents of lunar mare basalts. Meteorit. Planet.

Sci. 35, 193–200.

Gillis J. J., Jolliff B. L. and Korotev R. L. (2004) Lunar surfacegeochemistry: Global concentrations of Th, K, and FeO asderived from lunar prospector and clementine data. Geochim.

Cosmochim. Acta 68, 3791–3805.Greenhagen B. T., Lucey P. G., Wyatt M. B., Glotch T. D., Allen

C. C., Arnold J. A., Bandfield J. L. and Bowles N. E. (2010)Global silicate mineralogy of the Moon from the Diviner lunarradiometer. Science 329, 1507–1509.

Hallis L. J., Anand M., Greenwood R. C., Miller M. F., Franchi I.A. and Russell S. S. (2010) The oxygen isotope composition,petrology and geochemistry of mare basalts: Evidence for large-scale compositional variation in the lunar mantle. Geochim.

Cosmochim. Acta 74, 6885–6899.Handler M. R., Baker J. A., Schiller M., Bennett V. C. and Yaxely

G. M. (2009) Magnesium stable isotope composition of Earth’supper mantle. Earth Planet. Sci. Lett. 282, 306–313.

Hartmann W. K. and Davis D. R. (1975) Satellite-sized planetes-imals and lunar origin. Icarus 24, 504–515.

Housley R. M. (1979) A model for chemical and isotopicfractionation in the lunar regolith by impact vaporization.Proc. 10th Lunar Planet. Sci. Conf. 1673–1683.

Huang F., Lundstrom C. C., Glessner J., Ianno A., Boudreau A.,Li J., Ferre E. C., Marshak S. and DeFrates J. (2009) Chemicaland isotopic fractionation of wet andesite in a temperaturegradient: Experiments and models suggesting a new mechanismof magma differentiation. Geochim. Cosmochim. Acta 73, 729–749.

Huang F., Zhang Z. F., Lundstrom C. C. and Zhi X. C. (2011) Ironand magnesium isotopic compositions of peridotite xenolithsfrom eastern China. Geochim. Cosmochim. Acta 75, 3318–3334.

Humayun M. and Clayton R. N. (1995a) Precise determination ofthe isotopic composition of potassium: Application to terres-trial rocks and lunar soils. Geochim. Cosmochim. Acta 59, 2115–2130.

Humayun M. and Clayton R. N. (1995b) Potassium isotopecosmochemistry: Genetic implications of volatile elementdepletion. Geochim. Cosmochim. Acta 59, 2131–2148.

Jolliff B. L., Gillis J. J., Haskin L. A., Korotev R. L. and WieczorekM. A. (2000) Major lunar crustal terranes: Surface expressionsand crust mantle origin. J. Geophys. Res. 105, 4197–4216.

Korotev R. L. and Morris R. V. (1993) Composition and maturityof Apollo 16 regolith core 60013/14. Geochim. Cosmochim. Acta

57, 4813–4826.Lee T., Papanastassiou D. A. and Wasserburg G. J. (1977)

Aluminum-26 in the early solar system: Fossil or fuel?Astrophys. J. 211, L107–L110.

Lee T., Russell W. A. and Wasserburg G. J. (1979) Ca isotopicanomalies and the lack of 26Al in an unusual Allende inclusion.Astrophys. J. Lett., L93–L98.

Li W.-Y., Teng F.-Z., Ke S., Rudnick R. L., Gao S., Wu F.-Y. andChappell B. W. (2010) Heterogeneous magnesium isotopiccomposition of the upper continental crust. Geochim. Cosmo-

chim. Acta 74, 6867–6884.Li W.-Y., Teng F.-Z., Xiao Y. and Huang J. (2011) High-

temperature inter-mineral magnesium isotope fractionation ineclogite from the Dabie orogen, China. Earth Planet. Sci. Lett.

304, 224–230.Lind M. D. and Housley R. M. (1972) Crystallization studies of

lunar igneous rocks: Crystal structure of synthetic armalcolite.Science 175, 521–523.

Ling M.-X., Sedaghatpour F., Teng F.-Z., Hays P., Strauss J. andSun W. (2011) Homogeneous magnesium isotopic compositionof seawater: An excellent geostandard for Mg isotope analysis.Rapid Commun. Mass Spectrom. 25, 2828–2836.

Liu Y., Spicuzza M. J., Craddock P. R., Day J. M. D., Valley J.W., Dauphas N. and Taylor L. A. (2010a) Oxygen and iron






















































































































isotope constraints on near-surface fractionation effects and thecomposition of lunar mare basalt source regions. Geochim.

Cosmochim. Acta 74, 6249–6262.Liu S.-A., Teng F.-Z., He Y.-S., Ke S. and Li S.-G. (2010b)

Investigation of magnesium isotope fractionation during gran-ite differentiation: Implication for Mg isotopic composition ofthe continental crust. Earth Planet. Sci. Lett. 297, 646–654.

Liu S.-A., Teng F.-Z., Yang W. and Wu F.-Y. (2011) High-temperature inter-mineral magnesium isotope fractionation inmantle xenoliths from the North China craton. Earth Planet.

Sci. Lett. 308, 131–140.Lucey P., Korotev R. L., Gillis J. J., Taylor A., Lawrence D.,

Campbell B. A., Elphic Feldman B., Hood L. L., Hunten D.,Mendillo M. R., Noble S., Papike J. J., Reedy R. C., LawsonS., Prettyman T., Gasnualt O. and Maurice S. (2006) Under-standing the lunar surface and space-moon interactions. Rev.

Mineral. Geochem. 60, 83–202.Lodders K. (2003) Solar system abundances and condensation

temperatures of the elements. Astrophys. J. 591, 1220–1247.Magna T., Wiechert U. and Halliday A. N. (2006) New constraints

on the lithium isotope compositions of the Moon and terrestrialplanets. Earth Planet. Sci. Lett. 243, 336–353.

McKay D. S., Heiken G., Basu A., Blandford G., Simon S., ReedyR., French B. M. and Papike J. (1991) The lunar regolith. InThe Lunar Source Book Guide to the Moon (eds. G. H. Heiken,D. T. Vaniman and B. M. French). Cambridge Univ. Press, pp.285–356.

Meyer C. (2004–2011) Lunar sample compendium. Available from:<http://curator.isc.nasa.gov/lunar/compendium.cfm> and<http://www.lpi.usra.edu/lunar/samples>.

Morris R. V. (1978) The maturity of lunar soils: Concepts andmore values of Is/FeO. Lunar Planet. Sci. IX. 760–762.

Morris R. V., Lauer Jr. H. V. and Gose W. A. (1979) Character-ization and depositional and evolutionary history of the Apollo17 deep drill core. Proc. 10th Lunar Sci. Conf. 1141–1157.

Morris R. V., Score R., Dardano C. and Heiken G. (1983).Handbook of lunar soils-Part 2: Apollo 16–17. NASA, LyndonB. Johnson Space Center, p. 480.

Neal C. R. (2008) Mare basalt database. Available from: <http://www.nd.edu/~cneal/lunar-L/>.

Neal C. R. and Taylor L. A. (1992) Petrogenesis of mare basalts: Arecord of lunar volcanism. Geochim. Cosmochim. Acta 56,2177–2211.

Neal C. R., Hacker M. D., Snyder G. A., Taylor L. A., Liu Y.-G.and Schmitt R. A. (1994) Basalt generation at the Apollo 12site, Part 1: New data, classification, and re-evaluation.Meteoritics 29, 334–348.

Niederer F. R. and Papanastassiou D. A. (1984) Ca isotopes inrefractory inclusions. Geochim. Cosmochim. Acta 48, 1279–1293.

Niederer F. R., Papanastassiou D. A. and Wasserburg (1985)Absolute isotopic abundances of Ti in meteorites. Geochim.

Cosmochim. Acta 49, 835–851.Norman M. D., Yaxley G. M., Bennett V. C. and Brandon A. D.

(2006) Magnesium isotopic composition of olivine from theEarth, Mars, Moon, and Pallasite parent body. Geophys. Res.

Lett. 33, L15202.Paniello R. C., Day J. M. D. and Moynier F. (2012) Zinc isotopic

evidence for the origin of the Moon. Nature 490, 376–379.Pahlevan K. and Stevenson D. J. (2007) Equilibration in the

aftermath of the lunar-forming giant impact. Earth Planet. Sci.

Lett. 262, 438–449.Papike J. J., Ryder G. and Shearer C. K. (1998) Lunar samples.

Rev. Mineral. Geochem. 36, 1–234.Pogge von Strandmann P. A. E., Elliott T., Marschall H. R., Coath

C., Lai Y.-J., Jeffcoate A. B. and Ionov D. A. (2011) Variation

of Li and Mg isotope ratios in bulk chondrites and mantlexenoliths. Geochim. Cosmochim. Acta 75, 5247–5268.

Poitrasson F., Halliday A. N., Lee D. C., Levasseur S. and TeutschN. (2004) Iron isotope differences between Earth, Moon, Marsand Vesta as possible records of contrasted accretion mecha-nisms. Earth Planet. Sci. Lett. 223, 253–266.

Pritchard M. E. and Stevenson D. J. (2000) Thermal aspects of alunar origin by giant impact. In Origin of the Earth and Moon

(eds. R. Canup and K. Righter). University of Arizona Press,Tucson, pp. 179–196.

Raymond K. N. and Wenk H. R. (1971) Lunar ilmenite (Refine-ment of the crystal structure). Contrib. Mineral. Petrol. 30, 135–140.

Rhodes J. M. and Blanchard D. P. (1982) Apollo 11 breccias andsoils – Aluminous mare basalts or multi-component mixtures.Proc. 12th Lunar Sci. Conf. 607–620.

Richter F. M., Dauphas N. and Teng F.-Z. (2009a) Non-traditional fractionation of non-traditional isotopes: Evapora-tion, chemical diffusion and Soret diffusion. Chem. Geol. 258,92–103.

Richter F. M., Davis A. M., Ebel D. S. and Hashimoto A. (2002)Elemental and isotopic fractionation of Type B calcium-,aluminum-rich inclusions: Experiments, theoretical consider-ations, and constraints on their thermal evolution. Geochim.

Cosmochim. Acta 66, 521–540.Richter F. M., Janney P. E., Mendybaev R. A., Davis A. M. and

Wadhwa M. (2007) Elemental and isotopic fractionation ofType B CAI-like liquids by evaporation. Geochim. Cosmochim.

Acta 71, 5544–5564.Richter F. M., Watson E. B., Mendybaev R., Dauphas N., Georg

B., Watkins J. and Valley J. (2009b) Isotopic fractionation ofthe major elements of molten basalt by chemical and thermaldiffusion. Geochim. Cosmochim. Acta 73, 4250–4263.

Richter F. M., Watson E. B., Mendybaev R. A., Teng F.-Z. andJanney P. E. (2008) Magnesium isotope fractionation in silicatemelts by chemical and thermal diffusion. Geochim. Cosmochim.

Acta 72, 206–220.Russell W. A. and Papanastassiou D. A. (1978) Calcium isotope

fractionation in ion-exchange chromatography. Anal. Chem. 50,1151–1154.

Russell W. A., Papanastassiou D. A., Tombrello T. A., and EpsteinS. (1977) Ca isotope fractionation on the Moon. Proc. 8th

Lunar Sci. Conf. 3791–3805.Ryder G. and Schuraytz B. C. (2001) Chemical variation of the

large Apollo 15 olivine-normative mare basalt rock samples. J.

Geophys. Rev. 106, 1435–1451.Schauble E. A. (2011) First-principles estimates of equilibrium

magnesium isotope fractionation in silicate, oxide, carbonateand hexaaqua magnesium (2+) crystals. Geochim. Cosmochim.

Acta 75, 844–869.Schiller M., Handler M. R. and Baker J. A. (2010) High-precision

Mg isotopic systematics of bulk chondrites. Earth Planet. Sci.

Lett. 297, 165–173.Seitz H. M., Brey G. P., Weyer S., Durali S., Ott U., Munker

C. and Mezger K. (2006) Lithium isotope compositions ofMartian and lunar reservoirs. Earth Planet. Sci. Lett. 245,6–18.

Shearer C. K., Hess P. C., Wieczorek M. A., Pritchard M. E.,Parmentier E. M., Borg L. E., Longhi J., Elkins-Tanton L. T.,Neal C. R., Antonenko I., Canup R. M., Halliday A. N., GroveT. L., Hager B. H., Lee D.-C. and Wiechert U. (2006) Thermaland magmatic evolution of the Moon. Rev. Mineral. Geochem.

60, 365–518.Shearer C. K. and Papike J. J. (1993) Basaltic magmatism on the

Moon: A perspective from volcanic picritic glass beads.Geochim. Cosmochim. Acta 57, 4785–4812.




























http://www.curator.isc.nasa.gov/lunar/compendium.cfm

http://www.lpi.usra.edu/lunar/samples

http://www.nd.edu/~cneal/lunar-L/

http://www.nd.edu/~cneal/lunar-L/
























































































Shearer C. K. and Papike J. J. (1999) Magmatic evolution of theMoon. Am. Mineral. 84, 1469–1494.

Snyder G. A., Taylor L. A. and Neal C. R. (1992) A chemicalmodel for generating the sources of mare basalts: Combinedequilibrium and fractional crystallization of the lunar magma-sphere. Geochim. Cosmochim. Acta 56, 3809–3823.

Spicuzza M. J., Day J. M. D., Taylor L. A. and Valley J. W. (2007)Oxygen isotope constraints on the origin and differentiation ofthe Moon. Earth Planet. Sci. Lett. 253, 254–265.

Stoffler D., Knoll H. D., Marvin U. B., Simonds C. H. and WarrenP. H. (1980) Recommended classification and nomenclature oflunar highland rocks. In: Proc. Conf. Lunar Highland Crust.Pergamon Press, New York and Oxford. pp. 51–70.

Switkowski Z. E., Haff P. K., Tombrello T. A. and Burnett D. S.(1977) Mass fractionation of the lunar surface by solar windsputtering. J. Geophys. Res. 82, 3797–3804.

Taylor S. R. (1982) Planetary Science: A Lunar Perspective. TheLunar and Planetary Inst, Houston.

Teng F.-Z., Dauphas N., Helz R. T., Gao S. and Huang S. (2011)Diffusion-driven magnesium and iron isotope fractionation inHawaiian olivine. Earth Planet. Sci. Lett. 308, 317–324.

Teng F.-Z., Li W.-Y., Ke S., Marty B., Dauphas N., Huang S., WuF. Y. and Pourmand A. (2010) Magnesium isotopic composi-tion of the Earth and chondrites. Geochim. Cosmochim. Acta

74, 4150–4166.Teng F.-Z., Wadhwa M. and Helz R. T. (2007) Investigation of

magnesium isotope fractionation during basalt differentiation:Implications for a chondritic composition of the terrestrialmantle. Earth Planet. Sci. Lett. 261, 84–92.

Urey H. C. (1947) The thermodynamic properties of isotopicsubstances. J. Chem. Soc. (London), 562–581.

Van Kan Parker M., Mason P. R. D. and Van Westrenen W.(2011) Trace element partitioning between ilmenite, armalcoliteand anhydrous silicate melt: Implications for the formation oflunar high-Ti mare basalts. Geochim. Cosmochim. Acta 75,4179–4193.

Wang S.-J., Teng F.-Z., Williams H. and Li S. (2012) Magnesiumisotope variations in cratonic eclogites: Origins and implica-tions. Earth Planet. Sci. Lett. 359–360, 219–226.

Warner R. D., Nehru C. N. and Keil K. (1978) Opaque oxidemineral crystallization in lunar high-titanium mare basalts. Am.

Mineral. 63, 1209–1224.Warren P. H. (1985) The magma ocean concept and lunar

evolution. Annu. Rev. Earth Planet. Sci. 13, 201–240.Warren P. H. (2005) “New” lunar meteorites: Implication for

composition of the global lunar surface, lunar crust and thebulk Moon. Meteorit. Planet. Sci. 40, 477–506.

Warren P. H., Tonui E. and Young E. D. (2005) Magnesiumisotopes in lunar rocks and glasses and implications for the

origin of the Moon. Lunar Planet. Sci. XXXVI. Lunar Planet.Inst., Houston. #2143 (abstr.).

Wasserburg G. J., Lee T. and Papanastassiou D. A. (1977)Correlated O and Mg isotopic anomalies in Allende inclusions,II: Magnesium. Geophys. Res. Lett. 4, 299–302.

Weber R. C., Lin P.-Y., Garnero E. J., Williams Q. and LognonneP. (2011) Seismic detection of the lunar core. Science 331, 309–312.

Weyer S., Anbar A. D., Brey G. P., Munker C., Mezger K. andWoodland A. B. (2005) Iron isotope fractionation duringplanetary differentiation. Earth Planet. Sci. Lett. 240, 251–264.

Wiechert U. and Halliday A. N. (2007) Non-chondritic magnesiumand the origins of the inner terrestrial planets. Earth Planet. Sci.

Lett. 256, 360–371.Wiechert U., Halliday A. N., Lee D.-C., Snyder G. A., Taylor L.

A. and Rumble D. (2001) Oxygen isotopes and the Moon-forming giant impact. Science 294, 345–348.

Wieczorek M. A., Jolliff B. L., Khan A., Pritchard M. E., Weiss B.P., Williams J. G., Hood L. L., Righter K., Neal C. R., ShearerC. K., McCallum I. S., Tompkins S., Hawke B. R., Peterson C.,Gillis J. J. and Bussey B. (2006) The constitution and structureof the lunar interior. Rev. Mineral. Geochem. 60, 221–364.

Wiesli R. A., Beard B. L., Taylor L. A. and Johnson C. M. (2003)Space weathering processes on airless bodies: Fe isotopefractionation in the lunar regolith. Earth Planet. Sci. Lett.

216, 457–465.Xiao Y., Teng F.-Z., Zhang H.-F. and Yang W. (2013) Large

magnesium isotope fractionation in peridotite xenolithis fromeastern North China craton: Product of melt-rock interaction.Geochim. Cosmochim. Acta 115, 241–261.

Yang W., Teng F.-Z. and Zhang H.-F. (2009) Chondritic magne-sium isotopic composition of the terrestrial mantle: A casestudy of peridotite xenoliths from the North China craton.Earth Planet. Sci. Lett. 288, 475–482.

Young E. D., Ash R. D., Galy A. and Belshaw N. S. (2002) Mgisotope heterogeneity in the Allende meteorite measured by UVlaser ablation-MC-ICPMS and comparisons with O isotopes.Geochim. Cosmochim. Acta 66, 683–698.

Young E. D. and Galy A. (2004) The isotope geochemistry andcosmochemistry of magnesium. Rev. Mineral. Geochem. 55,197–230.

Young E. D., Tonui E., Manning C. E., Schauble E. A. and MacrisC. A. (2009) Spinel-olivine magnesium isotope thermometry inthe mantle and implications for the Mg isotopic composition ofEarth. Earth Planet. Sci. Lett. 288, 524–533.

Associate editor: Dimitri A. Papanastassiou


























































































Indirect simultaneous kinetic determination of semicarbazideand hydrazine in micellar media by H-point standard addition

method

A. Safavi a,�, H. Abdollahi b, F. Sedaghatpour a, M.R. Hormozi Nezhad a

a Department of Chemistry, College of Sciences, Shiraz University, Shiraz 71454, Iranb Department of Chemistry, Institute for Advanced Studies in Basic Sciences, Zanjan, Iran

Received 28 May 2002; received in revised form 6 August 2002; accepted 23 August 2002

Abstract

H-point standard addition method (HPSAM) is suggested as a simple and selective method for the determination of

semicarbazide and hydrazine. The reduction of Cu2� to Cu� by semicarbazide and hydrazine in the presence of

neocuproine (Nc) and the subsequent complex formation between Cu� and Nc produced a sensitive spectro-

photometric method for indirect determination of semicarbazide and hydrazine. The difference in the rate of reduction

of Cu2� with semicarbazide and hydrazine in cationic micellar media is the basis of this method. Semicarbazide can be

determined in the range of 0.5�/3.75 mg ml�1 with satisfactory accuracy and precision in the presence of excess

hydrazine. The proposed method was successfully applied to the simultaneous determination of semicarbazide (0.5�/

3.75 mg ml�1) and hydrazine (0.5�/5 mg ml�1) and also to the selective determination of semicarbazide in the presence

of hydrazine in several synthetic mixtures containing different concentration ratios of semicarbazide and hydrazine.

# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Semicarbazide; Hydrazine; Determination; HPSAM

1. Introduction

Hydrazine compounds are an important family

of organic compounds. Their separation and

detection have been of considerable analytical

interest for their wide applications in the chemical

industry and pharmacology [1]. Semicarbazide is a

poison and may be fatal if swallowed. Hydrazine is

a strong reducing agent used as an oxygen

scavenger for corrosion control in boilers and

hot-water heating systems [2]. It is also employed

as a starting material for many derivatives such as

foaming agents for plastics, antioxidants, polymer,

pesticides, plant-growth regulators and pharma-

ceuticals. Moreover, hydrazine, as well as its salts

and methyl and dimethyl derivatives, is used as

rocket fuel, gas generator and explosives. Hydra-

zine is volatile and toxic and is readily absorbed by

oral, dermal or inhalation routes of exposure [3].

Different methods for the determination of hy-

� Corresponding author. Tel.: �/98-711-2284822; fax: �/98-

711-6305881

E-mail address: [email protected] (A. Safavi).

Talanta 59 (2003) 147�/153

www.elsevier.com/locate/talanta

0039-9140/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved.

PII: S 0 0 3 9 - 9 1 4 0 ( 0 2 ) 0 0 4 6 5 - 4

mailto:[email protected]

drazines have been described in the literature.Spectrophotometric methods have been used to

analyze dilute hydrazine solutions using p-di-

methylaminobenzaldehyde [4], vanillin [5], salicy-

laldehyde [6] and silver-gelatin [7]. Optical

chemical sensors for determination of hydrazine

have also been described [8,9]. Flow-injection

method with spectrophotometric [10], fluorimetric

[11] and chemiluminiscence [12] detection havealso been developed for determination of hydra-

zine.

The formation of a charge�/transfer complex

between Cu� and neocuproine (Nc) (2,9-dimethyl-

1,10-phenantheroline) is the basis of the existing

spectrophotometric method for the indirect deter-

mination of reducing agents [13].

Campins-Falco et al. established the fundamen-tals of the H-point standard additions method

(HPSAM) to kinetic data for the simultaneous

determination of binary mixtures or the calcula-

tion of analyte concentration completely free from

bias error [14]. In this method, simultaneous

analysis of two components is possible based on

the difference between the rates of the reaction of

the two components with a common reagent. Twovariants of the method are proposed; one is

applied when the reaction of one component is

faster than that of the other or the latter does not

take place at all [14,15]; the other is used when the

rate constant of two components are time-depen-

dent [14]. Under the assumption that only one of

the species, X , evolves with time, the variables to

be fixed are two times t1 and t2 at which the speciesY does not evolve with time or over the range

between these two times. The absorbance of X at a

given wavelength l and two times t1 and t2 will be

measured, while those of Y under the same

conditions will be measured as equal because Y

is not changing in the time domain of analysis. In a

mixture containing X and Y , standard addition of

X is performed at two times t1 and t2. The intersectof the two straight lines of the standard addition

plots obtained at times t1 and t2 is known as H

point with characteristics of (�/CX , AY), where

(CX ) is the concentration of X and AY is the

analytical signal of species Y . This analytical

signal will enable calculation of concentration of

Y from a calibration curve.

In this work, the first variant of the H-pointstandard addition method is suggested as a simple

and selective method for the determination of

semicarbazide in the presence of hydrazine. This

method is based on the difference in the rate of

reduction of Cu2� with semicarbazide and hydra-

zine and complex formation of Cu� with Nc in

cationic micellar media. The absorbance of the

colored Cu� complex with Nc is then monitoredfor indirect determination of semicarbazide and

hydrazine.

2. Experimental

2.1. Apparatus

The absorbance measurements as a function of

time, at a fixed wavelength were made with a

Shimadzu UV-1601PC spectrophotometer at-

tached to a Pentium 200 MHz computer. Mea-

surements of pH were made with a Metrohm 691

pH-meter using a combined glass electrode.

2.2. Reagents

All chemicals used were of analytical reagent

grade. Triply distilled water was used throughout.

An ethanolic stock solution of Nc, 3�/10�2 M

was prepared by dissolving 0.1629 g of Nc (Fluka)

in ethanol and diluting to 25 ml with the samesolvent. Copper stock solution, 0.1 M, was pre-

pared by dissolving 0.604 g of Cu(NO3)2.3H2O in

water and diluting to 25 ml. Acetic�/acetate buffer

(pH�/5) was prepared by using acetic acid (1 M)

and sodium hydroxide (1 M) solution [16]. A stock

solution of cetyltrimethyl ammonium bromide

(CTAB) (0.1 M) was prepared by dissolving

0.9111 g of CTAB (Merck) in water and dilutingto 25 ml. A stock solution of semicarbazide was

prepared by dissolving 0.0125 g of semicarbazide

(Merck) in water and diluting to 25 ml. Hydrazine

stock solution (500 mg ml�1) was prepared by

dissolving 0.0125 g of NH2NH2.2HCl (Merck) in

water and diluting to 25 ml.

A. Safavi et al. / Talanta 59 (2003) 147�/153148

2.3. Procedure

Stock Nc solution (1 ml), 1 ml of stock Cu2�

solution, 1 ml of buffer solution and 0.6 ml of

stock CTAB solution and appropriate volumes of

semicarbazide and hydrazine containing 2.5�/18.75

mg semicarbazide and 2.5�/25 mg hydrazine were

added to a 5 ml volumetric flask and made up to

the mark with water. For each measurement, :/2ml of the above solution was transferred to the

spectrophotometric cell and the variation of the

absorbance versus time was recorded immediately.

The absorbance was measured at 450 nm with 1 s

time intervals for each sample. Simultaneous

determination of semicarbazide and hydrazine

with HPSAM was performed by measuring the

absorbances at 60 and 220 s after initiation of thereaction.

3. Results and discussion

3.1. Principles of the method

The Cu2��/Nc systems allows the spectrophoto-

metric determination of a reducing agent, Ared,

provided that the redox reaction:

n Cu2��2n Nc�Ared 0 n [Cu(Nc)2]��Aox

is complete with the formation of an equivalent

amount of [Cu(Nc)2]� with respect to the n -

electron reductant, Ared. Copper (II) is a strong

oxidizing agent only when its reduction product,

Cu�, is stabilized by a strong complex-formingligand, e.g. Nc. The standard potential of the

Cu2��/Cu� couple (0.17 V) is shifted to more

positive values by preferential complexation of

Cu�. The stronger the reductant, the more

quantitative will be the reduction of Cu2� with

the subsequent formation of [Cu(Nc)2]� [13].

The reduction of Cu2� to Cu� by semicarba-

zide and hydrazine in the presence of Nc andsubsequent complex formation between resulted

Cu� and Nc produce sensitive spectrophotometric

spectrum for indirect determination of semicarba-

zide and hydrazine.

Difference in kinetic behavior of two analytes

accompanied with mathematical treatment of data

using HPSAM permits simultaneous analysis of

two important hydrazine compounds, such as

semicarbazide and hydrazine. Fig. 1 shows that

the reduction of Cu2� by semicarbazide and

hydrazine has a kinetic behavior and the absor-

bance at 450 nm changes with time during the first

few minutes from initiation of the reaction.

3.2. Effect of operational parameters

Selectivity is one of the most important require-ments of any analytical method. Employing sui-

table experimental conditions and using desirable

mathematical algorithms can improve selectivity.

The kinetics behavior of most chemical species in

some chemical reactions can be controlled by

changing the microenvironment of reaction.

3.2.1. Effect of pH

The sensitivity and rate of reduction and com-

plex formation are dependent on pH of the

medium. Thus, the effect of pH was studied. pH5 was selected as optimum pH. As shown in Fig. 1,

under optimum pH conditions the kinetic behavior

of the system is not completely suitable for

applying the first variant of HPSAM.

3.2.2. Effect of surfactants

Micelles change the effective microenvironment

around dissolved solutes and as a consequence

their physicochemical properties, such as rate

constant, spectral profile etc., can be affected.

Fig. 1. Kinetic curves for 5 mg ml�1 of semicarbazide and 5 mg

ml�1 of hydrazine at pH�/5 in the absence of CTAB.

A. Safavi et al. / Talanta 59 (2003) 147�/153 149

The fact that micelles can accelerate reactions has

been increasingly exploited in the last few years to

improve the features of both catalytic [17�/19] and

non-catalytic [20] kinetics methods [21]. Micelles

can alter reaction kinetics through their rate,

control pathways, change reaction mechanisms

and/or the observed rate constant ratio of two ormore species interacting with a common reagent.

Because micelles can affect the kinetics of reac-

tions; the effects of micelles on the system was

studied under optimum pH. Cetyltrimethyl am-

monium bromide (CTAB) as a cationic surfactant,

sodium dodecyl sulfate (SDS) as an anionic

surfactant and Triton X-100 as a nonionic surfac-

tant were selected. CTAB was found as the bestsurfactant for HPSAM. It was observed that

CTAB increases the rates of both semicarbazide

and hydrazine reactions. This rate enhancement

might be due to the increased reactant concentra-

tions in the micellar pseudo-phase [22]. The

increased rates of reactions in the presence of

CTAB is in such a way that the rate of hydrazine

reaction becomes very fast while that of semicar-bazide is not as fast as the reaction of hydrazine

(Fig. 2). This is a suitable condition for HPSAM

since the reaction of hydrazine is complete in the

time domain of analysis while that of semicarba-

zide is still in progress.

3.3. Analytical performance and figures of merit

Fig. 2 shows the possible applicability of

HPSAM to the simultaneous analysis of semicar-

bazide and hydrazine by the proposed system. As

shown when semicarbazide is selected as the

analyte, it is possible to select several pairs of

time where they present the same absorbance for

hydrazine species. As shown previously by Cam-

pins-Falco et al. [14], the higher the value for slope

increment, the lower the error for analyte concen-

tration. So, only the time pair that gave the

greatest slope increment and at which the change

in absorbance related to hydrazine concentration

was negligible was selected. For this reason, the

time pair of 60�/220 s was employed for obtaining

Fig. 2. Kinetic curves for 5 mg ml�1 of semicarbazide and 5 mg

ml�1 of hydrazine at pH�/5 in the presence of CTAB.

Fig. 3. Plot of H-point standard addition method for simulta-

neous determination of semicarbazide (1.25 mg ml�1) and

hydrazine (1.25 mg ml�1) at different selected time pairs.

Fig. 4. Plot of H-point standard addition method for fixed

hydrazine concentration (1.25 mg ml�1) and different concen-

trations of semicarbazide.


the highest accuracy. Fig. 3 shows several H-point

standard addition calibration lines constructed at

some possible selected time pairs. It is observed

that the 60�/220 s time pair shows the greatest

slope increment.

According to the theory of HPSAM at H-point

(�/CX , AY ), CX (concentration of semicarbazide

at H-point) is independent from the concentration

of hydrazine (Fig. 4) and so Ay (the absorbance

value at H point, in this case the absorbance of

hydrazine) is also independent from the concen-

tration of semicarbazide (Fig. 5). The value of AH

enables calculation of the concentration of hydra-

zine from a calibration curve constructed from

ordinate of several HPSA plots with various

concentrations of hydrazine species.

For the determination of semicarbazide in the

presence of hydrazine, the absorbance increment

as the analytical signal was used. The application

of the HPSAM in the DAt 1�t 2�/Cadded variant

yields the concentration of semicarbazide directly

from the graph [14]. However, in order to ensure

the absence of constant or proportional error from

the calculated concentration, all the possible

DAt 1�t 2�/Cadded lines for semicarbazide should

intersect at the same point, namely that corre-

sponding to the unknown concentration, CH, as

this would indicate that the time evaluation of the

matrix would be a horizontal line. Fig. 6 and Table

1 show the results obtained from employing this

version of HPSAM on a mixture of semicarbazide

and hydrazine. The results obtained by this

procedure were in good agreement with the actual

concentrations.

Under optimum conditions, several synthetic

mixed samples with different concentration ratios

of semicarbazide and hydrazine (1:1, 1:2, 1:10,

1:2.4) were analyzed by using the suggested

method. As can be seen from Table 2, the accuracy

and precision of the results are all satisfactory.

Fig. 5. Plot of H-point standard addition method for fixed

semicarbazide concentration (1.25 mg ml�1) and different

concentrations of hydrazine.

Fig. 6. DA versus added semicarbazide concentration at

different time intervals at 450 nm for a synthetic mixture

containing 1.25 mg ml�1 of semicarbazide and hydrazine.

Table 1

Application of signal increment version of HPSAM in a synthetic mixture containing semicarbazide (1.25 mg ml�1) and hydrazine

(1.25 mg ml�1)

Time interval (s)

55�/512 65�/220 60�/210 80�/210

Found semicarbazide conc. (mg ml�1) 1.27 1.22 1.24 1.26

Recovery (%) 101.6 97.6 99.2 100.8

Relative standard deviation (%) (n�/4) 1.6 1.5 1.4 1.6


Limit of detection was calculated as LOD�/3SCH,

where SCH is the S.D. of several (n�/4) replicatemeasurements of zero concentration of analyte

with HPSAM. The corresponding value obtained

for semicarbazide was 0.03 mg ml�1. In addition,

for showing the applicability of the method,

environmental samples were spiked with semicar-

bazide and hydrazine and the proposed method

was applied to the determination of the analyte.

The results are shown in Table 3. As can be seen,the results are all satisfactory. Semicarbazide and

hydrazine can be determined in the range 0.5�/3.75

and 0.5�/5 mg ml�1 with satisfactory accuracy and

precision.

3.4. Effect of interferences

The effect of hydrazine compounds, phenylhy-

drazine, dinitrophenylhydrazine and isoniazide

was studied. They showed a kinetic behavior the

same as hydrazine and thus interfere with the

determination of hydrazine. Table 4 shows the

results obtained for the analysis of semicarbazide

and hydrazine in the presence of dinitrophenylhy-

drazine. As is obvious, the presence of dinitrophe-

nylhydrazine does not affect the determination of

semicarbazide, but seriously interferes with the

determination of hydrazine.

Table 2

Results of several experiments for the analysis of semicarbazide and hydrazine mixtures

A�/C equation r Present (mg ml�1) Found (mg ml�1)

Semicarbazide Hydrazine Semicarbazide Hydrazine

A60�/0.0755x�/1.315 0.985 0.50 5.00 0.43 5.07

A220�/0.2131x�/1.3746 0.997

A60�/0.0474x�/0.8027 0.993 1.50 3.00 1.26 2.99

A220�/0.1099x�/0.8816 0.999

A60�/0.0723x�/0.4263 0.997 1.25 1.25 1.26 1.36

A220�/0.159x�/0.5357 0.994

A60�/0.0755x�/1.315 0.995 1.50 3.75 1.50 3.55

A220�/0.2131x�/1.3746 0.992

Table 3

Determination of semicarbazide and hydrazine in real sample

Sample Concentration (mg ml�1)

Spiked Found


River Water 2.50 1.50 2.529/0.02 1.469/0.01

River Water 2.00 2.50 2.109/0.07 2.309/0.06

River Water 1.00 1.50 1.029/0.07 1.369/0.03

Table 4

Results of experiment for the analysis of a semicarbazide and

hydrazine mixture in the presence of dinitrophenylhydrazine

(1.25 mg ml�1)

Present (mg ml�1) Found (mg ml�1)


1.25 1.25 1.24 3.10


Acknowledgements

The authors gratefully acknowledge the support

of this work by Shiraz University Research

Council.

References

[1] W.-C. Yang, A.-M. Yu, Y.-Q. Dai, H.-Y. Chen, Anal.

Lett. 33 (15) (2000) 3343.

[2] H.W. Schessl, in: K. Othmer (Ed.), Encyclopedia of

Chemical Technology, vol. 13, 4th ed, Wiley/Interscience,

New York, 1995, p. 560.

[3] J.M. Pingarron, I. Ortiz Hernandez, A. Gonzalez-Cortes,

P. Yanez-Sedeno, Anal. Chim. Acta 439 (2001) 281.

[4] G.W. Watt, J.D. Chrisp, Anal. Chem. 24 (1952) 2006.

[5] S. Amlathe, V.K. Gupta, Analyst 113 (1988) 801.

[6] L.C. Bailey, T. Medwick, Anal. Chim. Acta 35 (1966) 330.

[7] T. Pal, D.S. Maity, A. Ganguly, Analyst 111 (1986) 1413.

[8] C. Gojon, B. Dureault, N. Hovnanian, C. Guizard, J. Sol-

Gel Sci. Technol. 14 (1999) 163.

[9] P. Ortega Barrales, A. Molina Diaz, M.I. Pascual Reguera,

L.F. Capitan Vallvey, Anal. Chim. Acta 353 (1997) 115.

[10] A.A. Ensafi, B. Naderi, Microchem. J. 56 (1997) 269.

[11] A.A. Ensafi, B. Rezaei, Talanta 47 (1998) 645.

[12] A. Safavi, M.R. Baezzat, Anal. Chim. Acta 358 (1988) 121.

[13] E. Tutem, R. Apak, F. Baykut, Analyst 116 (1991) 89.

[14] F. Bosch Reig, P. Campins Falco, A. Sevillano Cabeza, R.

Herreaz Hernandez, C. Molins Legua, Anal. Chem. 63

(1991) 2424.

[15] A. Safavi, H. Abdollahi, F. Sedaghatpour, S. Zeinali,

Anal. Chim. Acta 409 (2000) 275.

[16] J. Lurie, Handbook of Analytical Chemistry (English

translation), Mir, Moscow, 1975.

[17] M.L. Lunar, S. Rubio, D. Perez-Bendito, Anal. Chim.

Acta 237 (1990) 207.

[18] D. Sicilia, S. Rubio, D. Perez-Bendito, Talanta 38 (1991)

1147.

[19] D. Sicilia, S. Rubio, D. Perez-Bendito, Anal. Chem. 64

(1992) 1490.

[20] E. Athanasiou-Malaki, M.A. Koupparis, Anal. Chim.

Acta 219 (1989) 295.

[21] D. Sicilia, S. Rubio, D. Perez-Bendito, Anal. Chim. Acta

284 (1993) 149.

[22] D. Perez-Bendito, S. Rubio, Trends Anal. Chem. 12 (1993)

9.


653 © The Meteoritical Society, 2009. Printed in USA.

Meteoritics & Planetary Science 44, Nr 5, 653–664 (2009)Abstract available online at http://meteoritics.org

Characterization of Antarctic micrometeorites by thermoluminescence

F. SEDAGHATPOUR1 and D. W. G. SEARS1, 2*

1Arkansas Center for Space and Planetary Sciences, University of Arkansas, 202 Old Museum Building, Fayetteville, Arkansas 72701, USA2Department of Chemistry and Biochemistry, University of Arkansas, Fayetteville, Arkansas 72701, USA

*Corresponding author. E-mail: [email protected]

(Received 30 October 2008; revision accepted 30 January 2009)

Abstract–In order to explore the nature and history of micrometeorites, we have measured thethermoluminescence (TL) properties of four micrometeorites, three cosmic spherules, and oneirregular scoriaceous particle, that we found in a survey of 17 micrometeorites. Thesemicrometeorites have TL sensitivities ranging from 0.017 ± 0.002 to 0.087 ± 0.009 (on a scalenormalized to 4 mg of the H3.9 chondrite Dhajala). The four micrometeorites have very similar TLpeak temperatures and TL peak widths, and these distinguish them from CI, most CM, CV, CO, andordinary chondrites. However, the TL properties of these micrometeorites closely resemble those ofthe unusual CM chondrite MacAlpine Hills (MAC) 87300 and terrestrial forsterites. Heatingexperiments on submillimeter chips of a CM chondrite and a H5 chondrite suggest that these TLproperties are have not been significantly affected by atmospheric passage. Thus we suggest that thereis no simple linkage between these micrometeorites and the established meteorite classes, and thatforsterite is an important component of these micrometeorites, as it is in many primitive solar systemmaterials.

INTRODUCTION

Micrometeorites are extraterrestrial particles, <1 mm insize, collected on the surface of the Earth, largely inAntarctica (Maurette et al. 1986, 1991) or in the atmosphere(Brownlee 1985). Their origin could be the asteroids, theMoon, Mars, or comets (Bradley et al. 1988; Love andBrownlee 1993), and they make up the majority of the massaccreted to the Earth from space (Love and Brownlee 1993).Being so small, they are subject to Poynting-Robertson dragand might be more representative of the primitive material ofthe solar system than their larger counterparts, the meteorites(Kortenkamp et al. 2001). Small particles can pass throughthe atmosphere unaltered, but particles in the 50 µm to 1 mmrange may be unaltered, partially altered, or completelymelted depending on entry velocity and angle.

The compositions, mineralogy, and textures ofmicrometeorites (Taylor et al. 2000), and particularlyAntarctic micrometeorites, have been reported by manyauthors (Beckerling et al. 1992; Klöck et al. 1992; Greshakeet al. 1996; Gounelle et al. 1999, 2005; Christophe Michel-Lévy and Bourot-Denise (1992); Noguchi and Nakamura2000). Micrometeorites are mostly chondritic (Kurat et al.1992b; Engrand and Maurette 1998), having mineralogiesand compositions similar to carbonaceous chondrites butlower in Na, Ni, Ca, S, Se, and Zn (Kurat et al. 1992a, 1994;

Maurette 1992). Typically, they consist of iron oxide, ironsulfide, and anhydrous silicate phases (Kurat et al. 1994;Christophe Michel-Lévy and Bourot-Denise 1992; Gengeet al. 2008). The effects of atmospheric heating on themineralogy of micrometeorites have been reported by manyauthors (Rietmejier 1996; Toppani and Libourel 2003;Greshake et al. 1998; Nozaki et al. 2006; Tomioka et al. 2007;Genge et al. 2008). Such measurements suggest that volatileelements, like S, Zn, Ga, and Se will be lost by atmosphericheating, and that temperatures above 800 °C can convertphyllosilicate phases to olivine and pyroxene.

Induced thermoluminescence (TL) properties have beenuniquely successful in evaluating the nature and metamorphichistory of ordinary chondrites, CM chondrites, CVchondrites, and CO chondrites, E chondrites, lunar samples,and basaltic meteorites (Sears et al. 1980, 1982, 1991a,1991b, 1993, 1995b; Sears and Weeks 1983; Guimon et al.1984, 1985, 1995, 1988; Keck and Sears 1987; Keck et al.1987; Ninagawa et al. 1992; Matsunami et al. 1993).Essentially, TL is the light produced by a substance as it isheated and it depends on crystal structure and the presence ofdefects and impurities (e.g., Mn2+ in Ca sites in feldspar, orMn2+ and Cr3+ in forsterite or enstatite, cause luminescence,Fe2+ is a quencher). The strength of induced TLmeasurements (draining the natural TL and applying astandard test dose) is their extraordinary sensitivity, these

654 F. Sedaghatpour and D. W. G. Sears

minerals being detectable in amounts that would be missed bynormal mineralogical and petrographic methods. Thetechnique also allows a certain level of mineral identificationbecause the details of TL production vary with the mineralresponsible. Here we report an induced TL study of fourmicrometeorites collected at Cap Prudhomme, Antarctica(Maurette et al. 1991). Our objective was to explore the valueof thermoluminescence as a means of comparing them withthe known meteorite classes and thereby contribute to anunderstanding of their origin and history. Early phases of thiswork were reported at conferences (Sedaghatpour et al. 2008;Sears et al. 2008).

EXPERIMENTAL

Seventeen micrometeorites with size between 100−160 µm were examined as possible subjects for this study, andfour were selected because of their especially high levels of TLand thus worthy of a detailed study. There were six cosmicspherules (CS) and eleven scoriaceous micrometeorites (SC).For thermoluminescence measurement, the micrometeoriteswere placed in a copper pan in a Daybreak Nuclear andMedical systems TL rig. Heating rates were 7.5 °C/s and the TLproduced between room temperature and 500 °C was recorded.After removal of natural TL by a brief heating to 500 °C in theTL apparatus, the samples were exposed to a 200 mCi 90Sr betasource for 3 minutes and the induced TL measured. Repeatmeasurements were made. We measured TL maximum for anidentifiable peak, the width of that peak, measure as the full-width at half its maximum intensity (FWHM), and thetemperature at which the peak appears. The peak height isnormalized to the TL signal from 4 mg of a non-magneticportion of the Dhajala H3.9 ordinary chondrite; this is referredto as the “TL sensitivity.”

In order to evaluate the effects of atmospheric passage onthe TL properties of micrometeorites, heating experimentswere performed on six submillimeter fragments of a meteoritewith no measurable TL (CM chondrite MAC 88100) and sixfragments of a meteorite with a very high TL sensitivity (theH5 ordinary chondrite Jilin). In this way we could detect boththe formation of TL phosphors, perhaps by the dehydration ofhydrated minerals, and their removal by oxidation or melting.Heating was performed in Lindburg tube furnace with atemperature range of 200 to ∼1200 °C in air at atmosphericpressure. The temperature was controlled with an OmegaOS562 thermal probe with an accuracy of ±1 °C. After theirpreheating TL properties were measured, samples wereplaced in the furnace on a heated quartz rod, removed after30 s, and allowed cool in the atmosphere.

RESULTS

Induced TL was below the detection limit for thirteen ofour particles, but strong signals were detected for four,

designated SC2, CS2, CS3, and CS4 (Fig. 1). The TLsensitivities for our micrometeorites are shown in Table 1.Values range from 0.017 to 0.087 but are generally nearer thelower end of this range, sometimes with a factor of twodifferences between duplicate measurements. The uncertaintyon the TL sensitivity measurements—determined from thereplicate measurements of similar sized Semarkona matrixsamples and the signal-to-noise ratio on the raw data for theseparticles (∼10)—is only about 10%. This suggests that themicrometeorites are heterogeneous and tumbled betweenmeasurements.

The glow curves for the four micrometeorites are shownin Fig. 2. To a first approximation, the overall shapes of theglow curves are similar to those observed for mostchondritic materials, consisting of two major peaks, thelower peak (at 150–200 °C) being the most intense in threeof the micrometeorites, and the higher peak (at ∼400 °C) beingthe most intense for the scoriaceous micrometeorite SC2which has only a weak peak in the 140–240 °C region.

Peak temperatures and peak widths are also listed inTable 1. Peak temperatures range from 165 to 211 °C, thevalues for duplicate measurements on the samemicrometeorite being within about 10%. Three of theparticles have very similar values, the scoriaceousmicrometeorite SC2 perhaps being slightly higher, but therelative weakness of this peak makes this uncertain. TL peakwidths tend to be less uniform because of interference fromthe higher peak, but are within 30–40% for duplicates and inthe range 85 to 133 °C for all the particles.

The results of our heating experiments are shown in Fig. 3.The CM chondrite MAC 88100, which showed nomeasurable induced thermoluminescence before heating, wasunaffected by the heating process (Fig. 3a). The noise on theglow curve (mostly thermal noise) is equivalent to a TLsensitivity detection limit at the 3σ level of about 0.002 underthese experimental conditions, about a factor of 10 lower thanour weaker micrometeorites.

The H5 chondrite Jilin (Fig. 3b) showed no change in TLproperties when heated at 200, 400, 600, and 800 °C, theslight variation from curve to curve reflecting heterogeneityin the six submillimeter Jilin fragments. However, heating attemperatures of 1000 and 1180 °C caused a complete loss ofsignal detectable above the noise.

SAMPLE DESCRIPTIONS

Following measurement of their TL properties, themicrometeorites were imaged by ESEM (Philips ModelXL30) and analyzed with an EDX attachment under highvacuum and running at 10 and 20 keV. Our analyses weremade on whole particles, and are therefore subject to largeuncertainties and are sensitive to weathering products on thesurfaces. Images for those with measurable TL are shown inFig. 1 and will be described below.

Characterization of Antarctic micrometeorites by thermoluminescence 655

Micrometeorite SC2 (Fig. 1a) is a scoriaceous particle thatis an irregular shape ∼150 µm by ∼120 µm. The blocky texturesuggests that it is coarse grained and our EDX data suggests thepresence of forsterite and iron oxides. Thus it is possibly aporphyritic olivine coarse-grained scoriaceous micrometeorite(Genge et al. 2008).

Micrometeorite CS2 (Fig. 1b) is a cryptocrystallinespherule similar to those shown in Fig. 1i-l of Genge et al.(2008). Such objects are usually contain Fe-rich fine-grainedolivine crystals and grains that survive melting and formcrystallization nuclei.

Micrometeorite CS3 (Fig. 1c) is a glassy spherule similar

Table 1. Induced thermoluminescence data for four Antarctic micrometeorites.a

Sample DescriptionSize (µm)

TL sensitivity(Dhajala = 1)

Peak temperature (°C)

FWHM (°C)

SC2 Irregular ~150 × 117 0.036 ± 0.004 211 ± 21 107 ± 16Scoriaceous 0.018 ± 0.002 185 ± 18 133 ± 20

CS2 Ellipsoid ~170 × 140 0.065 ± 0.007 185 ± 18 85 ± 120.087 ± 0.009 165 ± 16 96 ± 14

CS3 Spherule ~155 × 140 0.036 ± 0.004 185 ± 18 125 ± 180.017 ± 0.002 180 ± 18 85 ± 12

CS4 Spherule ~160 × 150 0.019 ± 0.002 175 ± 17 92 ± 140.038 ± 0.004 185 ± 18 89 ± 13

aData for two independent measurements are quoted. Uncertainties were not estimated for individual measurements but experience with similar materialssuggests that 1 uncertainties are ~10% for TL sensitivity, ~10% for peak temperature, and ~15% for peak width. TL sensitivity is the induced TL intensitynormalized to that of 4 mg of bulk sieved Dhajala powder.

Fig. 1. Scanning electron images of the micrometeorites that were found to have high TL levels and are the subject of the present study. a)Irregular scoriaceous micrometeorite SC2. b) Cosmic spherule CS2. c) Cosmic spherule CS3. d) Cosmic spherule CS4. Scale bars for Figs.1a, 1b, 1c, and 1d, are 50 µm, 50 µm, 100 µm, 100 µm, respectively.


to that shown in Fig. 1e of Genge et al. (2008). These objectsfrequently contain FeNi metal grains and vesicles. Our EDXanalysis was consistent with normative Mg-rich pyroxene andolivine.

Micrometeorite CS4 (Fig. 1d) is identical in externalappearance to the S-type spherule shown in Fig. 1i of Gengeet al. (2008), which is also a cryptocrystalline texture. Suchspherules normally contain submicron crystals and significantsubmicron magnetite. Olivine crystal can grow from thesurface inward producing a knobby surface called turtleback,and this appears to be the case with CS4. Again, our EDXanalyses are consistent with this mineralogy.

All our particles contained large amounts of normativehematite (10–60%), presumably due to surface weathering oratmospheric oxidation. However, we also detected normativesilicates olivine and pyroxene in all the particles, in oneinstance a metal grain, and normative feldspar (∼10%).Feldspar has only rarely been observed in micrometeorites(Christophe Michel-Lévy and Bourot-Denise 1992) and we

Fig. 2. Induced TL “glow curves” (plots of light produced as afunction of laboratory heating temperature) for the fourmicrometeorites shown in Fig. 1. As with most extraterrestrialmaterials, the curves consist of a peak in the region 150–200 °C anda broader peak in the region 320–420 which are referred to as the“low” temperature and “high” temperature TL. The curves forthe three cosmic spherules are very similar, but the scoriaceousparticle shows a higher high temperature TL, relative to the lowtemperature TL, compared to the others.

Fig. 3. Glow curves for submillimeter samples of the CM chondriteMAC 88100 (a) and the H5 ordinary chondrite Jilin (b) heated in airfor 30 s at the temperatures indicated. The heat treatment was unableto cause a TL signal to appear in the CM chondrite that initially hadno TL, and the heat treatment was unable to affect the TL of theordinary chondrite with a strong initial TL signal until the heatingreached >1000 °C, when the sample was no longer able to produceddetectable TL. We suggest that in the case of the ordinary chondrite,temperatures this high caused melting of the feldspar responsible forthe TL signal.


assume this component identified by the norm calculation isactually glass.

DISCUSSION

We will first compare the present data with similar datafor CO chondrites, CM chondrites, CV chondrites, and theordinary chondrites. Then we will discuss the mineralogicalimplications of these data and then we will discuss the resultsof our heating experiments and implications for the effect ofatmospheric passage on the TL of our micrometeorites.Finally we will briefly discuss lessons learned from this studyon the measurement of the TL properties of small particles.

Before proceeding, we mention that our present TLmeasurements involved light coming from the surfacebecause, while the test dose of radiation can penetrate thewhole object, these objects are largely opaque. Consistentwith this, on one occasion, we saw considerable variabilityin the level of TL produced by one of the micrometeorites,CS4, as it tumbled in the apparatus between measurements.However, this did not affect the shape of the TL glow curveand any of the conclusions discussed here. The surface is alsousually weathered, and amorphous materials are generallypoor at displaying TL, so we suspect that our TL signal iscoming from essentially unweathered grains that intercept thesurface. We suspect that if we were to break open some of thenonluminescent micrometeorites, we would obtain a signal.However, here we focus on the four micrometeorites thatdisplay a strong signal.

Comparison of Induced TL Properties with Those ofKnown Chondrite Classes

We will not discuss CI chondrites because they do notproduce a measurable TL signal, and we will not discussenstatite chondrites because of their highly characteristic anddistinct TL properties.

Figure 4a shows representative glow curves for COchondrites and the CM chondrites (Keck and Sears 1987).The natural TL survey of Antarctic meteorites included manyCM chondrites that showed no measurable TL (Myers et al.1990). However, MAC 87300, MAC 87301, and MAC 88107(the last not shown in Fig. 4a) are unusual paired CMchondrites that produce TL. Sears et al. (1991) suggested thatthis paired trio of MAC meteorites should be consideredCM3.0, CM3.1, and CM3.1 chondrites, respectively, as manyof their properties resembled those of CO3.0–CO3.2chondrites.

The CO chondrites show a range of fairly well-understood TL properties that reflect the range ofmetamorphism experienced by this class (Keck and Sears1987). The properties they display reflect the formation offeldspar by the crystallization of glass and that the feldspar isin the low-temperature form. The hummocky structure of the

CO3.0 and CO3.2 is not well understood, but probablyrepresents a mixture of minerals including forsterite andthe minerals found in refractory inclusions.Cathodoluminescence (CL) micrography of these meteoritesis consistent with the TL being due to forsterite and refractoryphases (Sears et al. 1991), and the CL of Murchison (whichhas been studied in some depth) suggests that forsterite isabundant and luminescence in these meteorites although in aform not suitable for detection by TL (Sears et al. 1993). Mostprobably, the emission is at longer wavelengths than we candetect in our TL apparatus where heat filters are used tosuppress black body radiation from the sample.

Figure 4b compares peak temperatures for the CO andCM chondrites, and compares them with peak temperaturesfor the micrometeorites. The CO chondrites show a transitionfrom a major peak at ∼250 °C to ∼125 °C as the dominantphosphor changes from small amounts of primary high-temperature feldspar to metamorphism-produced low-temperature feldspar. The micrometeorites plot in a similarrange of peak temperatures to the CM chondrite, suggestingsimilarity to each other and dissimilarity to the COchondrites. The high-temperature peaks in themicrometeorites, seen occasionally in the CO chondrites,especially the least metamorphosed members, resemble thepeaks observed in refractory inclusions from Allende and areprobably due to refractory minerals in these samples.

The CV chondrites have thermoluminescence propertiesthat are similar to those of CO chondrites, reflecting thepresence of refractory phases (with a peak at ∼350 °C in Fig. 5a),high-temperature feldspar (whose peak is at ∼220 °C in Fig. 5a)and low-temperature feldspar (where the peak is at ∼110 °C inFig. 5a). The relative proportions of these and the overall TLsensitivity of the CV chondrites reflect changes caused bymetamorphism (Keck and Sears 1987; Sears et al. 1991). Therefractory inclusions show similar trends in which themetamorphic differences appear to exist from one inclusion toanother. The peak temperatures for the low temperature TLfor the micrometeorites (∼175 °C) are clearly plottingintermediate to the feldspar peaks in the CV chondrites, againsuggesting discrete differences (Fig. 5b). The high-temperature peaks in the micrometeorites also appear to beplotting at higher values than the high temperature peaks forthe CV chondrites (∼400 °C compared with ∼350 °C),suggesting that if refractory minerals are responsible for thesepeaks, they are distinct from those of CV chondrites.

The ordinary chondrites probably have the best-understood TL properties of any meteorite class. In terms of TLsensitivity, the micrometeorites are comparable to 4 mgsamples of petrographic type 3.2–3.4 ordinary chondrites(Table 1). This is surprisingly bright for such small particles.However, the data for their TL peak temperatures and TLwidths (Fig. 6) suggests little, if any, connection with theordinary chondrites with their dominant feldsparthermoluminescence. Ordinary chondrites plot in two discrete


Fig. 4. a) Representative TL glow curves for CO chondrites, of varying metamorphic type, and the paired CM chondrites MAC 87300 andMAC 87301 that are unlike other CM chondrites in displaying a significant TL signal. Induced TL peaks in these samples tend to be at ∼100 °C,∼200 °C, and ∼350 °C (Sears et al. 1991a). Detailed studies of the TL properties of these meteorites and phases separated from them suggestthat these peaks are due to low-temperature (metamorphism-produced) feldspar, high-temperature (primary) feldspar, and refractory phaseslike melilite, respectively (Sears et al. 1991a, 1995). b) A detailed comparison of peak temperatures shows that while the micrometeorites arevery similar to each other, they are distinct from the CO chondrites but resemble the anomalous CM chondrites (including a third pairedmember MAC 88107).


Fig. 5. a) Representative glow curves for CV chondrites showing very similar properties to those of the CO chondrites in Fig. 4 (Sears et al.1995). b) A detailed comparison of peak temperatures shows that in addition to being distinct from CO chondrites, the presentmicrometeorites, at least for the low-temperature TL, are distinct from CV chondrites. However, the micrometeorites display induced TL at∼400 °C which may be analogous to the ∼350 °C peak frequently shown by the CV chondrites and attributed to refractory phases.


regions in Fig. 6, the lower petrographic types (types <3.5) plotin the lower field and the higher types (types 3.5 and above)plot in the higher field. The micrometeorites plot to highertemperatures and smaller widths than the ordinary chondrites.

Forsterite in Micrometeorites

The luminescence properties of solids are governed byenergy transitions in a crystal lattice, the phosphor, whichcontain appropriate luminescence centers, usually transitionmetal ions but sometimes a structural defect (Garlick andGibson 1948; McKeever 1985). Additionally, TL requiressites in the crystal lattice capable of trapping electrons thatcan be released by heating the sample. Removing theelectrons naturally present in the traps and then providing astandard test dose of radiation in the laboratory provides ameans of investigating the number and nature of thephosphors, and to characterize the sample and its history(Sears and Hasan 1986; Sears 1988). Some kinds of materials(for example quartz) have TL peaks whose induced TLproperties can be affected by large radiation doses. This is arare phenomenon only observed to any extent with the 110 °Cpeak of quartz, and it is unlikely that it affected the presentsamples. An important difference between micrometeoritesand the chondrite samples is that the chondrite samples werealmost certainly part of much larger bodies in space and were

shielded from radiation. However, a detailed analysis of thisissue was performed when extraterrestrial materials were firststudied by TL, and we found that their induced TL propertieswere remarkably unresponsive even to the highest radiationdoses (Sears 1980).

The most common phosphor for TL in extraterrestrialmaterials is feldspar, where Mn2+ in the Ca2+ sites causes bothelectron traps and luminescent centers (Geake et al. 1973).Thus the conversion of feldspathic glass to feldspar causes a105-fold increase in TL sensitivity (Sears et al. 1980).Metamorphism also causes the conversion of low-temperature (structurally ordered) feldspar to high-temperature (disordered) and the resulting changes trap depthcauses changes in TL properties reflected in the two ordinarychondrite fields in Fig. 6 (Guimon et al. 1984, 1985).Refractory inclusions from Allende show a movement for thedominant TL peak from ∼100 °C, to ∼200 °C, to ∼400 °C asthe dominant phosphor converts from low-feldspar, to high-feldspar, to melilite during metamorphism (Sears et al.1995b). In the enstatite chondrites, it is the enstatite thatproduces TL with many relatively narrow peaks (Zhang et al.1995).

Forsterite is well known for its cathodoluminescence(CL), which is usually red but occasionally blue (Steele et al.1985; Steele 1986). Trace amounts of any transition metaliron, such as Mn2+ and Cr3+ can cause CL and TL emission byforsterite. Small amounts of Fe2+ in the lattice (greater than∼1 mol%) quenches the TL of most minerals (Geake et al.1973), including olivine, pyroxene, and feldspar, and thisprocess can also have value in understanding the history ofthese samples (Batchelor and Sears 1991). Our EDX analysisand the Genge et al. (2008) descriptions of micrometeoriteswith similar morphologies are consistent with a majorpresence of olivine in these micrometeorites.

Also shown in Fig. 6 are the fields occupied by seventerrestrial forsterites obtained from a variety of volcanic andmetamorphic environments on Earth (Craig and Sears 2009),∼100 µm fragments of Semarkona matrix in which forsteriteis thought to be the TL phosphor (Craig and Sears,Forthcoming) and forsterite-bearing chondrules (Sears et al.1995a). We think it is highly significant that themicrometeorites plot in a field discrete from the ordinarychondrites yet in the same general region as these otherforsterite samples and this suggests that the mineralresponsible for the TL of the micrometeorites is forsterite.

Forsterite appears to be a common component in manyprimitive solar system materials, being observed in lowpetrographic type meteorites, especially in chondrules,(McSween 1977; Olsen and Grossman 1978; Steele et al.1985a; Steele 1986; Jones 1992), in astronomical nebulae(Koike et al. 2002), in comet dust tail (Crovisier et al. 1997),in interplanetary dust particles (Kloeck et al. 1989), deep seaspherules (Steele et al. 1985b), and grains from comet Wild 2returned by the Stardust spacecraft (Zolensky et al. 2006).

Fig. 6. A useful way of comparing the present data with that forordinary chondrites is to consider both the peak temperature and thepeak width. Ordinary chondrites plot in two fields, depending on thestructural state of their feldspar, which is determined by thermalhistory. These fields are 1σ and 2σ variances shown in anunpublished compilation of all available TL data by Jeff Grossman(see Craig and Sears 2009, for more details). Seven terrestrialforsterites (Craig and Sears, Forthcoming), six matrix fragmentsfrom Semarkona (for which the phosphor is also thought to beforsterite, see Craig and Sears 2009), and 11 forsterite-bearingchondrules from Semarkona (Sears et al. 1995a), plot in the fieldsindicated. The presence of the micrometeorites in the forsterite fieldswould suggest that the phosphor responsible for the low-temperatureTL in these micrometeorites is forsterite.


The Thermoluminescence of Micrometeorites: QuenchProducts or Relict Grains?

Micrometeorites CS2, CS3, and CS4 are cosmic spherulesand normally assumed to be completely melted. However,complete melting would result in the absence of a TL signal, sothe presence of a strong TL signal is also evidence for thepresence of unmelted minerals within the spherule.

Relic grains have been reported in micrometeorites,including the types measured here (Genge et al. 2008). Theamounts of relict grains are small, but TL sensitivity is relatedmore to the composition and crystal structure of the grains,rather than their amount. (In chondrule mesostasis TL is beingdetected from 1 part in 105 of the sample, for instance). We donot believe that fine dendrites produced by quench processesproduce significant thermoluminescence because mostcosmic spherules do not have measurable TL. In a recentsurvey, only four out of seventeen micrometeorites—processed in the manner of the present samples—haddetectable induced TL, even though their physical andcompositional properties were similar to those discussed hereand they consisted predominantly of quenched glass.Secondly, while there have been no specific studies of the TLproperties meteorite fusion crusts, measurements made closeto the crust, and therefore include crust, demonstrate that thelevel of induced TL in fusion crust material is essentially zero(Vaz and Sears 1977; Akridge et al. 2000). Meteorite fusioncrusts are also produced by atmospheric melting andsubsequently quenched on the same timescale asmicrometeorites. We are therefore confident that the TL weobserve in the present spherules is due to relict grains.

The Effects of Atmospheric Heating

Our heating experiments enable us to further address theissue of whether the TL we are detecting is due to quenchcrystals or relict grains. During atmospheric passage,minerals can dehydrate and oxidize. Elements can be lost byvolatilization. Amorphous phases could crystallize, althoughthe time scales for alteration are small. Theoretical studiesindicate that small bodies (e.g., <50 µm) can pass through theatmosphere unscathed because they can radiate faster thanthey absorb energy. Larger grains may suffer a degree ofalteration and even partial or complete melting depending onvelocity and entry angle. Our heating experiments wereperformed to provide us with an indication of possible effectsof atmospheric passage on the micrometeorite TL properties.

We first considered the possibility of the heat ofatmospheric passage dehydrating and crystallizing thematerials in the micrometeorite, and thereby creatingluminescent phosphors. Figure 3a shows that a CM chondriteheated for thirty seconds in the laboratory at temperatures of200–1180 °C was not able to cause the formation of TLphosphors. Higher temperatures would cause partial—and

eventually complete—melting, which would destroyphosphors. Thus we think it highly unlikely that the TLproperties of our micrometeorites were caused by theformation of phosphors during atmospheric heating. But whatabout altering the TL properties of phosphors originallypresent in the micrometeorites?

Jilin is an equilibrated ordinary chondrite with very highlevels of induced TL. Heating Jilin for 30 s at temperatures upto ∼800 °C had little or no influence on the induced TLproperties. However, and as discussed above, the phosphor inequilibrated ordinary is feldspar and as the melting point for thefeldspar is approached (∼1110 °C), we see a complete loss ofTL signal. Once melted, the feldspar becomes a glass andglasses do not produce TL. Neither did these glasses crystallizeduring cooling to produce crystals with measurable TL.

To summarize, heating a meteorite with no initial TL doesnot cause a TL signal to appear, but heating a sample with a TLsignal destroys it at ∼1000 °C (where the phosphor is feldspar).Our heating experiments thus suggest that TL properties shouldnot be affected by atmospheric passage unless the phosphorsare completely destroyed by melting. Feldspars are melted atrelatively low temperatures, but if—as we have suggested—themajor phosphor we are seeing in these micrometeorites isforsterite, then significantly higher temperatures are required(∼1600 °C) and the likelihood of their TL properties beingaltered by melting is accordingly reduced.

Measuring the TL Properties of Small Particles

The present measurements were made on equipmentessentially unaltered from the equipment used for many yearson macrosamples (4 mg powders). It is therefore extremelyencouraging that we were able to make the presentmeasurements well above the detection limits of theapparatus. A number of modifications to the apparatus andour procedures would improve our detection limits and enableus to detect TL from smaller or less luminescent particles.First, mechanical changes can be made to the apparatus tobring the sample closer to the detector. Second, focusingoptics can be incorporated into the apparatus. Third, heatingrates can be increased. Fourth, the samples can be brokenopen to reveal interior minerals that have not been exposed toatmospheric alteration or terrestrial weathering.

CONCLUSIONS

Four micrometeorites (three cosmic spherules and onescoriaceous irregular object) have relatively high levels ofinduced TL (comparable with bulk samples of type 3.2−3.4ordinary chondrites), while thirteen objects had no detectablesignal when measured in the same manner. To a firstapproximation these four micrometeorites have similar TLproperties to each other but which distinguish them from CI,most CM, CO, CV, and ordinary chondrites and resemble only


the anomalous MAC 87300 CM-like chondrite. We thussuggest that there is no simple linkage between these objectsand the major chondrite classes but there might be a linkagewith CM-like meteorites. Based on their unusual TLproperties and similarities to terrestrial forsterites, and otherforsterite-rich extraterrestrial materials, we suggest that animportant component of these micrometeorites is forsterite.Heating experiments suggest that atmospheric passage shouldnot have significantly affected the TL properties of thesesamples, although it is possible that the lower melting pointminerals may have melted in the spherules, but not in thescoriaceous particle.

Acknowledgments–We are grateful to Cecile Engrande forsupplying the particles, Jonathan Craig for help with theinitial phases of this work, Walter Graupner for his help inmaintaining the equipment, Matthew Genge and ananonymous reviewer for comments that helped us improvethe paper, and NASA for financial support. We are grateful toHazel Sears for reviewing and proofing this paper prior tosubmission.

Editorial Handling—Dr. Scott Sandford

REFERENCES

Akridge J. M. C., Benoit P. H., and Sears D. W. G. 2000. Terrestrialage measurements using natural thermoluminscence of a drainedzone under the fusion crust of Antarctic ordinary chondrites.Meteoritics & Planetary Science 35:869−874.

Batchelor J. D. and Sears D. W. G. 1991. Thermoluminescenceconstraints on the metamorphic, shock and brecciation history ofbasaltic meteorites. Geochimica et Cosmochimica Acta 55:3831−3844.

Beckerling W., Bischoff A., and Klöck W. 1992. Mineralogy andchemistry of micrometeorites from Greenland and Antarctica.Meteoritics 27:200−201.

Brownlee D. E. 1985. Cosmic dust: Collection and research. AnnualReview of Earth and Planetary Sciences 13:147−173.

Bradley John P., Sandford Scott A., and Walker Robert M. 1988.Interplanetary dust particles. In Meteorites and the early solarsystem. Tucson, AZ: The University of Arizona Press. pp. 861−895.

Christophe Michel-Lévy M. and Bourot-Denise M. 1992. Mineralcomposition in Antarctic and Greenland micrometeorites.Meteoritics 27:73−74.

Craig J. and Sears D. W. G. 2009. The nature and origin of Semarkonafine-grained matrix: A study by induced thermoluminescence.Meteoritics & Planetary Science. This issue.

Craig J. and Sears D. W. G. Forthcoming. The thermoluminescenceproperties of forsterite. American Mineralogist.

Crovisier J., Leech, K., Bockelee-Morvan D., Brooke T. Y., HannerM. S., Altieri B., Keller H. U., and Lellouch E. 1997. Thespectrum of comet Hale-Bopp (C/1995 01) observed with theInfrared Space Observatory at 2.9 AU from the Sun. Science 275:1904−1907.

Engrand C. and Maurette M. 1998. Carbonaceous micrometeoritesfrom Antarctica. Meteoritics & Planetary Science 33:565−580.

Garlick G. F. J. and Gibson A. F. 1948. The electron trap mechanism

of luminescence in sulphide and silicate phosphors. Proceedingsof the Physical Society of London A 60:547–590.

Geake J. E., Walker G., Telfer D. J., Mills A. A., and Garlick G. F. J.1973. Luminescence of lunar, terrestrial, and synthesizedplagioclase caused by Mn2+ and Fe3+. Proceedings, 4th LunarScience Conference. pp. 3181−3189.

Genge M. J., Engrand C., Gounelle M. and Taylor S. 2008. Theclassification of micrometeorites. Meteoritics & PlanetaryScience 43:497−515.

Gounelle M., Maurette M., Engrand C., Brandstätter F., and Kurat M.1999. Minerology of the 1998 astrolabe Antarcticmicrometeorites collection (abstract # 5205). Meteoritics &Planetary Science 34.

Gounelle M., Engrand C., Maurette M., Kurat G., McKeegan K. D.,and Brandstätter F. 2005. Small Antarctic micrometeorites: Amineralogical and in situ oxygen isotope study. Meteoritics &Planetary Science 40:917−932.

Greshake A., Hoppe P., and Bischoff A. 1996. Mineralogy,chemistry, and oxygen isotopes of refractory inclusions fromstratospheric interplanetary dust particles and micrometeorites.Meteoritics & Planetary Science 31:739−748.

Greshake A., Klöck W., Arnet P., Maetz M., Flynn G. J., Bajt S., andBischoff A. 1998. Heating experiment simulating atmosphericentry heating of micrometeorites: Clues to their parent bodysources. Meteoritics & Planetary Science 33:267−290.

Guimon R. K., Weeks K. S., Keck B. D., and Sears D. W. G. 1984.Thermoluminescence as a palaeothermometer. Nature 311:363−365.

Guimon R. K., Keck B. D., and Sears D. W. G. 1985. Chemical andphysical studies of type 3 chondrites—IV: Annealing studies ofa type 3.4 ordinary chondrite and the metamorphic history ofmeteorites. Geochimica et Cosmochimica Acta 19:1515−1524.

Guimon R. K., Lofgren G. E., and Sears D. W. G. 1988. Chemical andphysical studies of type 3 chondrites, IX: Thermoluminescenceand hydrothermal annealing experiments and their relationship tometamorphism and aqueous alteration in type <3.3 ordinarychondrites. Geochimica et Cosmochimica Acta 52:119−127.

Guimon R. K., Symes S. P., Sears D. W. G., and Benoit P. H. 1995.Chemical and physical studies of type 3 chondrites XII: Themetamorphic history of CV chondrites and their components.Meteoritics 30:707−714.

Jones R. H. 1992. On the relationship between isolated and chondruleolivine grains in the carbonaceous chondrite ALHA77307.Geochimica et Cosmochimica Acta 56:3593.

Keck B. D. and Sears D. W. G. 1987. Chemical and physical studiesof type 3 chondrites, VIII: The CO chondrites. Geochimica etCosmochimica Acta 51:3013−3022.

Klöck W., Thomas K. L., McKay D. S., and Zolensky M. E. 1989.Comparison of olivine compositions in IDPs and chondritematrices. Meteoritics 24:286−287.

Klöck W., Beckerling W., Spettel B., Flynn G., and Sutton S. 1992.Bulk composition and mineralogy of Antarctic micrometeorite.23rd Lunar and Planetary Science Conference. p. 697.

Koike C., Chihara H., Koike K., Nakagawa M., Okada M.,Tsuchiyama A., Aok M., Awata T., and Atobe K. 2002.Thermoluminescence of forsterite and fused quartz as acandidate for the extended red emission. Meteoritics & PlanetaryScience 37:1591−1598.

Kortenkamp S. J., Dermott S. F., Floge D., and Grogan K. 2001.Sources and orbital evolution of interplanetary dust accreted bythe Earth. In Accretion of extraterrestrial matter throughoutEarth’s history, edited by Peucker-Eherenbrink B. and SchmitzB. New York: Kluwer/Plenum. pp. 13−30.

Kurat G., Koeberl C., Presper T., Brandstätter F., and Maurettem M.


1992a. Bulk composition of Antarctic micrometeorites: Nebularand terrestrial signatures. Meteoritics 27:246.

Kurat G., Presper T., and Brandstätter F. 1992b. CI-likemicrometeorites from Cap Prudhomme Antarctica. 23rd Lunarand Planetary Science Conference. p. 747.

Kurat G., Koeberl C., Presper T., Brandstätter F., and Maurette M.1994. Petrology and geochemistry of Antarctic micrometeorites.Geochimica et Cosmochimica Acta 58:3879−3904.

Love S. G. and Brownlee D. E. 1993. A direct measurement of theterrestrial accretion rate of cosmic dust. Science 262:550−553.

Matsunami S., Ninagawa K., Nishimura S., Kubon N., Yamamoto I.,Kohata M., Wada T., Yamashita Y., Lu Jie, Sears D. W. G., andNishimura H. 1993. Thermoluminescence and compositionalzoning in the mesostasis of a Semarkona group A1 chondrulesand new insights into the chondrule-forming process.Geochimica et Cosmochimica Acta 57:2101−2110.

Maurette M., Hammer C., Brownlee D. E., Reeh N., and ThomsenH. H. 1986. Placers of cosmic dust in the blue ice lakes ofGreenland. Science 233:869.

Maurette M., and Olinger C. 1991. A collection of diversemicrometeorites recovered from 100 tons of Antarctic blue ice.Nature 351:44−46.

Maurette M. Kurat G., Presper T., Brandstätter F., and Perreau M.1992. Possible causes of depletion and enrichement of minorelements in Antarctic micrometeorites. 22nd Lunar and PlanetaryScience Conference. p. 857.

McSween H. Y. 1977. Chemical and petrographic constraints on theorigin of chondrules and inclusions in carbonaceous chondrites.Geochimica et Cosmochimic Acta 4:1843−1860.

Myers B., Sears H., Hasan F., Score R., and Sears D. 1990. Naturalthermoluminescence data for Antarctic meteorites. AntarcticMeteorite Newsletter 13(2):25−27.

Ninagawa K., Nishimura S., Kubona N., Yamamoto I., Kohata M.,Wada T., Yamashita Y., Lu J., Sears D.W.G., Matsunami S., andNishimura H. 1992. Thermoluminescence of chondrules inprimitive ordinary chondrites, Semarkona and Bishunpur.Proceedings of the NIPR Symposium on Antarctic Meteorites 5:281−289.

Noguchi T. and Nakamura T. 2000. Mineralogy of Antarcticmicrometerorites recoverd from the Dome Fuji Station. AntarcticMeteorite Research 13:285−305.

McKeever S. W. S. 1985. Thermoluminescence of solids.Cambridge; New York: Cambridge University Press.

Nozaki W., Nakamura T., and Noguchi T. 2006. Bulk mineralogicalchanges of hydrous micrometeorites during heating in the upperatmosphere at temperatures below 1000 °C. Meteoritics &Planetary Science 41:1095−1114.

Reitmeijer F. J. M. 1996. The Ultrafine mineralogy of a molteninterplanetary dust particle as an example of the quench regimeof atmospheric entry heating. Meteoritics & Planetary Science31:237−242.

Sears D. W. 1980. Thermoluminescence of meteorites; relationshipswith their K-Ar age and their shock and reheating history. Icarus44:190−206.

Sears D. W. G. 1988. Thermoluminescence of meteorites: Sheddinglight on the cosmos. Nucearl Tracks and RadiationMeasurements/International Journal of Radiation. Applicationsand Instrumentation Part D 14:5−17.

Sears D. W. G. and Hasan F. A. 1986. Thermoluminescence andAntarctic meteorites. Proceeding of the Second Workshop onAntarctic Meteorites, edited by Annexstad J. O., Schultz L., andWänke H.). LPI Technical Report 86−01. Lunar and PlanetaryInstitute, Houston. pp. 83−100.

Sears D. W. G. and Weeks K. S. 1983. Chemical and physical

studies of type 3 chondrites—II: Thermoluminescence ofsixteen type 3 ordinary chondrites and relationships withoxygen isotopes. Proceedings, 14th Lunar and PlanetaryScience Conference, part 1. Journal of Geophyical Research88:B301.

Sears D. W., Grossman J. N., Melcher C. L., Ross L. M., and MillsA. A. 1980. Measuring the metamorphic history ofunequilibrated ordinary chondrites. Nature 287:791−795.

Sears D. W., Grossman J. N., and Melcher C. L. 1982. Chemical andphysical studies of type 3 chondrites—I: Metamorphism-relatedstudies of Antarctic and other type 3 ordinary chondrites.Geochimica et Cosmochimica Acta 46:2471−2481.

Sears D. W. G., Batchelor D. J., Lu Jie, and Keck. D. B. 1991a.Metamorphism of CO and CO-like chondrites and comparisonswith type 3 ordinary chondrites. Proceedings of the NIPRSymposium on Antarctic Meteorites 4:1745−1805.

Sears D. W. G., Hasan F. A., Batchelor J. D., and Lu Jie 1991b.Chemical and physical studies of type 3 chondrites XI:Metamorphism, pairing, and brecciation of ordinary chondrites.Proceedings, 21st Lunar and Planetary Science Conference. pp.493−512.

Sears D. W. G., Benoit P. H., and Lu Jie 1993. Two groups each withdistinctive rims in Murchison recognized bycathodoluminescence. Meteoritics 28:669−675.

Sears D. W. G., Morse A. D., Hutchison R., Guimon R. K., Jie, L.,Alexander C. M. O’D., Benoit P. H., Wright I., Pillinger C.,Xie T., and Lipschutz M. E. 1995a. Metamorphism and aqueousalteration in low petrographic type ordinary chondrites.Meteoritics 30:169–181.

Sears D. W. G., Symes S. P., Guimon R. K., and Benoit P. H. 1995b.Chemical and physical studies of type 3 chondrites XII: Themetamorphic history of CV chondrites and their components.Meteoritics 30:707−714.

Sears D. W. G., Sedaghatpour F., and Craig J. 2008.Thermoluminescence as a technique for determining the natureand history of small particles. Meteoritics & Planetary Science43:A142.

Sedaghatpour F., Craig J. P., and Sears D. W. G. 2008. A study of thethermal history of Antarctic micrometeorites bythermoluminescence: A first look (abstract #1165). 39th Lunarand Planetary Science Conference. CD-ROM.

Steele I. M. 1986. Compositions and textures of relic forsterite incarbonaceous and unequilibrated ordinary chondrites.Geochimica et Cosmochimica Acta 50:1379−1395.

Steele I. M., Smith J. V., and Skirius C. 1985a. Cathodoluminescencezoning and minor elements in forsterites from the Murchison (C2)and Allende (C3V) carbonaceous chondrites. Nature 313:294-297.

Steele I. M., Smith J. V., and Brownlee D. E. 1985b. Minor-elementsignature of relic olivine grains in deep-sea particles—A matchwith forsterites from C2 meteorites. Nature 313:297−299.

Taylor S., Lever J. H., and Harvey R. P. 2000. Number, type andcomposition of an unbiased collection of cosmic spherules.Meteoritics & Planetary Science 35:651−666.

Tomeoka N., Tomeoka K., Nakamaura-Messenger K., and Sekine T.2007. Heating effects of the matrix of experimentally shockedMurchison CM: Comparison with micrometeorites. Meteoritics& Planetary Science 42:4621−4638.

Toppani A. and Libourel G. 2003. Factors controlling compositionsof cosmic spinels: Application to atmospheric entry condition ofmeteoric materials. Geochimica et Cosmochimica Acta 67:1379−1395.

Vaz J. E. and Sears D. W. 1977. Artificially inducedthermoluminescence gradients in stony meteorites. Meteoritics12:47–60.

Zhang Y., Huang S., Benoit P. H., and Sears D. W. G. 1995. Natural


thermoluminscence evidence for major differences in thermalhistory of EH and EL chondrites. 26th Lunar and PlanetaryScience Conference. pp. 1557−1558.

Zolensky M. E., Zega T. J., Yano H., Wirick S., Westphal A. J.,Weisberg M. K., Weber I., Warren J. L., Velbel M. A.,Tsuchiyama A., Tsou P., Toppani A., Tomioka N., Tomeoka K.,Teslich N., Taheri M., Susini J., Stroud R., Stephan T.,Stadermann F. J., Snead C. J., Simon S. B., Simionovici A., SeeT. H., Robert F., Rietmeijer F. J. M., Rao W., Perronnet M. C.,Papanastassiou D. A., Okudaira K., Ohsumi K., Ohnishi I.,Nakamura-Messenger K., Nakamura T., Mostefaoui S.,

Mikouchi T., Meibom A., Matrajt G., Marcus M. A., Leroux H.,Lemelle L., Le L., Lanzirotti A., Langenhorst F., Krot A. N.,Keller L. P., Kearsley A. T., Joswiak D., Jacob D., Ishii H.,Harvey R., Hagiya K., Grossman L., Grossman J. N., Graham G.A., Gounelle M., Gillet Ph., Genge M. J., Flynn G., Ferroir T.,Fallon S., Ebel D. S., Dai Z. R., Cordier P., Clark B., Chi M.,Butterworth A. L., Brownlee D. E., Bridges J. C., Brennan S.,Brearley A., Bradley J. P., Bleuet P., Bland P. A., and Bastien R.2006. Mineralogy and petrology of comet 81P/Wild 2 nucleussamples. Science 314:1735–1739.

Full Paper

Tensammetric Analysis of Nonionic Surfactant Mixtures byArtificial Neural Network

A. Safavi,* F. Sedaghatpour, H. R. Shahbaazi

Department of Chemistry, College of Science, Shiraz University, Shiraz 71454 Iran*e-mail: [email protected]

Received: June 16, 2004Accepted: January 6, 2005

AbstractAn artificial neural network (ANN) model has been developed for tensammetric determination of a series of Brijes(Brij 30, Brij 35, Brij 56, Brij 96) as nonionic surfactants. The tensammetric method is based on the measurement ofthe capacitive current of the mercury electrode after adsorption of surfactants. All Brijes were analyzed in theconcentration range of 1.0 – 100.0 mg mL�1. The proposed method shows good sensitivity and applicability to thesimultaneous determination of mixtures of four Brijes in aqueous solutions.

Keywords: Tensammetric, Nonionic surfactant, Simultaneous determination, Artificial neural network (ANN)

1. Introduction

A large number of synthetic nonionic surfactants can befound in sewage and surface water. Surfactants are the mostimportant synthetic substances with respect to carbonsupply. Their flux to surface water is some orders ofmagnitude higher than fluxes of pollutants like polychlor-obiphenyls (PCB) or polyaromatic hydrocarbons (PAH) [1].Therefore, determination of each of them is important anddifficult with the present state of trace analysis of surfac-tants. On the other hand, such a target is demanding fromthe point of view of routine control of water pollution.Alcohol etoxylates have emerged as the principal nonionicsurfactants in consumer detergent products. Consistentqualities, expanding production capacity of relatively inex-pensive detergents, highly biodegradable alcohols, andincreasing usage in laundry products, particularly heavy-duty liquids, are the principal factors underlying this growth.

Many surface-active substances exhibit tensammetric(adsorption/desorption) peaks [2, 3]. Generally, the analyt-ical results are calculated from the depression of double-layer capacitance or the height of the desorption peak.Various types of surfactants, both ionic and nonionic, as wellas some macromolecular compounds are easily determinedin aqueous solutions by means of their adsorption onsuitable electrodes. The adsorption involves displacementof water molecules by organic molecules which produces ameasurable decrease of the electrode double-layer capacity[4]. The use of the mercury electrode is advantageousbecause of its reproducibly renewable surface. Alternatingcurrent (AC) polarography can be used to measure eitherthe capacity of the electrode double-layer or the capacitivecurrent. This simple methodology has been widely appliedin environmental analysis, polymer analysis, process control,etc. [4]. Direct measurement of surfactants by electro-

chemical methods has proved useful in the determinationand characterization of organic matter in natural andpolluted waters [5, 6]. The height of tensammetric peakscan be used for quantitative analytical purposes. Thedetermination of components of surfactant mixtures is amuch more difficult problem than the individual surfactants.Mixtures of surfactants may undergo complicated mutualinfluences of the components both in the bulk solution [7]and on the electrode surface [4]. It has been shown that themultivariate calibration technique based on singular valuedecomposition and Ho-Kashyp algorithm (SVDHK) can beapplied successfully in cases where interaction between thecomponents of a mixture influences the analytical signal andin cases where this signal is nonlinear with the concentration[8].

The aim of this work is to show how a multivariatecalibration technique based on artificial neural network canbe applied successfully to the simultaneous determination ofseveral nonionic surfactants.

2. Experimental

2.1. Apparatus

Voltammetric measurements were made with a Metrohm694 (Herisau. Switzerland) VA Stand coupled with aMetrohm 693 VA Processor. Voltammetric experimentswere carried out with a three-electrode arrangement with anAg/AgCl, 3 M KCl reference electrode, platinum wirecounter electrode and a multi-mode mercury drop workingelectrode. Dissolved oxygen was removed by purging thesolution under study with argon. pH measurements weremade with a Metrohm 691 pH-meter.

1112

Electroanalysis 2005, 17, No. 12 � 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/elan.200403225

2.2. Reagents

A stock solution (1%) of polyoxyethylene (4) lauryl ether(Brij 30), was prepared by dissolving 1.0000 g Brij 30(Fluka) in water and diluting to 100 mL in a volumetricflask. A stock solution (1%) of polyoxyethylene (23) laurylether (Brij 35), was prepared by dissolving 1.0000 g of Brij35 (Fluka) in water and diluting to 100 mL in a volumetricflask. A stock solution (0.5%) of polyoxyethylene (10) cetylether (Brij 56), was prepared by dissolving 0.5000 g of Brij56 (Fluka) in water and diluting to 100 mL in a volumetricflask. A stock solution (1%) of polyoxyethylene (10) oleylether (Brij 96), was prepared by dissolving 1.0000 g of Brij96 (Fluka) in water and diluting to 100 mL in a standardflask. Britton-Robinson (B-R) buffer was prepared bydissolving boric acid (15g), orthophosphoric acid (8.1 mL)and glacial acetic acid (6.9 mL) in water and diluting to500 mL. Appropriate volumes of this solution were adjust-ed to the required pH value with sodium hydroxide solution(3 M).

2.3. Procedure

All experiments were performed at room temperature. A2 mL aliquot of buffer (B – R) solution (pH 8), and appro-priate volumes of stock solution of each Brij were pipettedinto a 10 mL volumetric flask and diluted to the mark withwater and then transferred to the voltammetric cell.

The solution was purged with argon first for 10 minutesand then for 20 s before each voltammetric step. Subse-quently, the stirrer was switched off for 5 s. Finally, apotential scan was carried out in a negative direction from�500 to �2000 mV in alternating current mode and thecharging current due to adsorption/desorption of thesurfactants was recorded.

Data were processed by using an artificial neural networkmodel in order to find the quantity of each componentpresent in the sample. A back-propagation artificial neuralnetwork having three layers was created using a Visual-Basic software package.

3. Results and Discussion

3.1. Preliminary Experiments

Preliminary experiments were carried out to identify thebest voltammetric technique for quantitative study. Differ-ential pulse (DP) polarography and alternating current(AC) polarography at phase angles 0 and 908 were appliedfor different concentrations of Brij 30 (Fig. 1), Brij 35, Brij56, Brij 96 and tensammetric behavior of each Brij wasstudied. In AC polarography at phase angle 0, shift oftensammetric peak is little. This is an advantageous point formultivariate analysis because changing position of the peakscan affect the results. Also at phase angle 0, the tensam-metric peaks are sharp and the sensitivity is high. Therefore,

AC polarography at phase angle 0 was selected forquantitative determination.

Representative tensammetric responses for the surfac-tants investigated are shown in Figures 2 – 5. Brij 30 (Fig. 2)forms two peaks within the concentration range of 1 – 5.5 mgmL�1. The less negative peak is wide and the other is sharp.The increase in the concentration of surfactant causes theshift of the wide (less negative) peak toward more negative

Fig. 1. a) AC polarogram at phase angle 0; b) AC polarogram atphase angle 90; c) DP polarogram for different concentrations(5 – 3400 mg mL�1) of Brij 30 at pH 9.

1113Tensammetric Analysis of Nonionic Surfactant Mixtures

Electroanalysis 2005, 17, No. 12 � 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

potentials while the other peak stays in position (approx-imately �1770 mV). The wide peak disappears at concen-trations above 5.5 mg mL�1 while the sharp peak increaseswith the increase of concentration of surfactant up to3400 mg mL�1. These findings are very similar to what hasbeen reported previously for other oxyethylated alcohols [9,10].

The tensammetric behavior of Brij 35 is shown in Figure 3.This surfactant shows one broad peak in the concentrationrange of 0.7 – 2.0 mg mL�1. The potential of this peak shiftswith increase of surfactant concentration toward more

negative potentials. The increase in Brij 35 concentrationover the value of 2.0 mg mL�1 causes the appearance ofanother peak. The new peak grows with the increase ofsurfactant concentration up to 1900 mg mL�1 and, simulta-neously, the first peak decreases and finally disappear withincrease in Brij 35 concentration. This trend is also the sameas what has been reported previously for this surfactant [9,10].

Figures 4 and 5 show tensammetric curves for Brij 56 andBrij 96, respectively, which show similar trends as for otherBrij surfactants.

Fig. 2. AC polarogram at phase angle 0 for different concentrations (1 – 3400 mg mL�1) and a) 1 – 5.5 mg mL�1of Brij 30 at pH 9.

Fig. 3. AC polarogram at phase angle 0 for different concentrations (0.7 – 1900 mg mL�1) and a) 0.7 – 26 mg mL�1of Brij 35 at pH 9.

1114 A. Safavi et al.


3.2 Effect of Operational Parameters

3.2.1. Effect of pH

The influence of pH on the tensammetric peak wasstudied in the pH range of 2 to 11 for each Brij (Fig. 6). Inorder, to keep the composition of the buffer constant,when studying the effect of pH, Britton-Robinson bufferswere used. There was no tensammetric peak for any of theBrijes studied in solutions with pH lower than 6. Accord-ing to the results obtained, pH 8 was selected as theoptimum pH for analytical purposes. At this pH value, all

four Brijes have sharp and sensitive tensammetricpeaks.

3.2.2. Effect of Scan Rate

The effect of scan rate over the range of 20 – 120 mV/s wasexamined on the tensammetric peak currents of Brij 30, Brij35, Brij 56 and Brij 96.

The highest tensammetric peak currents were obtainedfor Brij 35, Brij 56 and Brij 96 at a scan rate of 40 mV s�1.Therefore, this scan rate was selected as the optimum scanrate for analytical purposes.

Fig. 4. AC polarogram at phase angle 0 for different concentrations (0.6 – 850 mg mL�1) and a) 0.6 – 5.5 mg mL�1of Brij 56 at pH 9.

Fig. 5. AC polarogram at phase angle 0 for different concentrations (4.5– 3000 mg mL�1) and a) 4.5 – 47.5 mg mL�1of Brij 96 at pH 9.



3.2.3. Effect of Frequency of the Alternating Voltage

The effect of frequency of alternating voltage was studiedover 30 – 240 Hz range for Brij 30, Brij 35, Brij 56 and Brij 96.The height of the tensammetric peak increases withincreasing frequency.

Maximum sensitivity was obtained at frequency of 240Hz. However, this frequency was not selected for quantita-tive determination, because at this frequency, the determi-nation range decreases. Therefore, the frequency of 60 Hzwas selected as the optimum frequency of alternatingvoltage.

3.2.4. Effect of Accumulation Time

At short accumulation times (up to 30 s), peak currentincreased with accumulation time, indicating that beforeadsorptive equilibrium is reached, the longer accumulationtime leads to higher adsorption of the surfactant. The peakcurrent increases upon extending accumulation time to

about 60 s for each Brij and then starts to level off. However,the determination range decreases with increase in accu-mulation time. Therefore, in order to have a larger concen-tration span of the Brijes, the accumulation time of 0 s wasselected for further studies.

3.2.5. Effect of Accumulation Potential

The effect of accumulation potential on the tensammetricpeak current was examined over the potential range of�450to �1800 mV. As shown in Figure 7, the peak current isnearly independent of the accumulation potential except forBrij 30 which decreases with changing potential from�450to �1800 mV. Therefore, an accumulation potential of�500 mV was selected as the optimized potential value.

3.3 Multivariate Calibration with ANN

Figure 8 shows the tensammetric peak for each Brij andtheir mixtures under optimum experimental conditions.There are high overlapping tensammetric peaks for the fourBrijes. Therefore, it seems that the use of a multivariatecalibration technique could be helpful for simultaneousdetermination.

The data obtained from the AC polarogram wereprocessed by ANN, which was trained with the back-propagation of errors learning algorithm. The reduced ACpolarographic data with principal component analysis(PCA) were used as the input for ANN. The first step inthe simultaneous determination of Brijes by ANN method-ology involves constructing the calibration set for themixtures containing Brij 30, Brij 35, Brij 56 and Brij 96. Inthis study, the calibration and prediction sets were preparedrandomly to reduce correlation between concentrations ofBrijes. Thirty mixtures were selected as the calibration setand eleven mixtures were selected as the prediction set foreach Brij. The concentrations for the training set of Brij 30,Brij 35, Brij 56 and Brij 96 were between 1.0 and 100.0 mg

Fig. 6. Effect of pH on tensammetric peak current. Experimen-tal conditions: Eac¼ 500 mV, tac¼ 0 s, 50 mg mL�1 Brij 30, 50 mgmL�1 Brij 35, 50 mg mL�1 Brij 56, 50 mg mL�1 Brij 96.

Fig. 7. Effect of accumulation potential on tensammetric peakcurrent. Experimental conditions: tac¼ 0 s, pH¼ 8, 50 mg mL�1 Brij30, 50 mg mL�1 Brij 35, 50 mg mL�1 Brij 56, 50 mg mL�1 Brij 96.

Fig. 8. AC polarogram for 50 mg mL�1 each Brij; 1) Brij 30, 2)Brij 35, 3) Brij 56, 4) Brij 96, 5) Mixture of Brij 30, Brij 35, Brij 56,and Brij 96 at pH 8.



mL�1 in the calibration matrix. Also, the number of inputswas normalized between �10 and þ10. Calibration set wasused for construction of ANN model, and the independenttest set was used to evaluate the quality of the model.

There are several ways to optimize the parameters of anetwork [11 – 13]. One of them is constructed on theminimum error of prediction of testing set [14, 15]. Theoverall predictive capacity of the model was compared interms of relative standard error, RSE defined as:

RSE %ð Þ ¼ 100�

PN

j¼1Cj � Cj

� �

PN

j¼1Cj

� �2

20

BBB@

1

CCCA

1=2

where Cj is jth component of the desired target (real sample)and Cj is the jth component of output produced by network (predicted sample). Training and testing of the network is aprocess of determination of the network�s topology andoptimization of its adjustable parameters, where we seek theminimum of an error surface in multi-dimensional space.The optimum learning rate and momentum were evaluatedby obtaining those, which yielded a minimum in error ofprediction (of the test set).

To investigate the prediction ability of ANN, neuralnetwork models for individual component, were also madewith respect to output layer considered as a single nodecorresponding to the analyte [16]. The construction of theseANN models is summarized in Table 1. The structure of thenetwork was comprised of three layers, an input, a hiddenand an output layer. The addition of an extra parameter,called bias, to the decision function increases its adaptabilityto the decision problem it is designed to solve. In thisstructure, þ1 was added as a bias.

The most versatile transfer function that can be used tomodel a variety of relationships is the sigmoidal type. In thepresent nonlinear system, sigmoid functions were found tobe optimum for hidden and output layers. The optimumnumber of input obtained was 5 for all Brijes using principal

component analysis (PCA). The number of nodes in thehidden layer was determined by training ANN with differ-ent number of nodes and then comparing the predictionerrors from the test set.

The training was stopped manually when the root meansquare error of the test set remained constant aftersuccessive iteration. The optimum number of iterationcycles or epochs (one pass of all objects, 30 mixtures,through the network) for each component was also ob-tained. Because there are several local minima, the algo-rithm was run from different starting values of initial weightsto find the best optimum. RSEP for Brij 30, Brij 35, Brij 56and Brij 96 were 3.5 %, 4.9%, 2.6%, and 1.7%, respectively.The reasonable relative error for each analyte indicates theaccuracy of the proposed method (Table 2)

For showing the applicability of the model, tap watersamples were spiked with different concentration ratios ofBrijes and these synthetic mixed samples were tested by themodel. The results were all satisfactory (Table 3). This canvalidate the ANN model.

3. 4. Effect of Interferences

The selectivity was assessed by studying the influence offoreign substances on the determination of Brijes. The effect

Table 1. Optimized parameters used for construction of ANNmodels.

Parameter Compounds

Brij 30 Brij 35 Brij 56 Brij 96

Input nodes 5 5 5 5Hidden nodes 5 7 6 4Output nodes 1 1 1 1Learning rate 0.02 0.20 0.10 0.40Momentum 0.125 0.100 0.200 0.200Hidden layer function sigmoid sigmoid sigmoid sigmoidOutput layer function sigmoid sigmoid sigmoid sigmoidNumber of iterations 1920 2000 1250 700

Table 2. Composition of prediction set and relative standard errors for the determination of analytes.

Actual Found

Brij 30 Brij 35 Brij 56 Brij 96 Brij 30 Brij 35 Brij 56 Brij 96

96.0 100.0 50.0 90.0 95.0 97.7 50.5 86.989.0 93.0 17.0 100.0 83.1 99.7 17 99.996.5 100.0 40.0 77.5 98.2 97.7 44.6 76.5

100.0 93 99.0 17.5 96 97.8 98.1 17.315.0 91.0 75.0 99.0 15.8 97.7 73.8 98.6

100.0 90.0 97.5 100.0 99.3 97.7 99.3 99.23.0 99.0 5.5 100.0 3.1 97.7 5.9 99.9

82.0 45.0 9.5 7.5 81.3 46.0 9.6 7.698.0 98.5 90.0 26.0 98.2 97.7 89.9 25.557.0 7.0 50.0 75.0 61.7 6.5 50.5 73.67.5 20.0 17.0 71.0 6.0 19.7 17 73.6

R.S.E. (%) 3.5 4.9 2.6 1.7



of several surfactants at different concentrations wasstudied on tensammetric peak current values of a solutioncontaining 50.0 mg mL�1 of each Brij. Polyethyleneglycoltert-octylphenyl ether (Triton X-100 and Triton X-305) asnonionic surfactant increased the tensammetric peak. Whilepolyoxyethylene sorbitanmonolaurate (Tween 20) and Span80 as nonionic surfactant decreased the tensammetric peak.The effect of N-dodecyl pyridinum chloride (DPC), cetyl-pyridinum chloride (CPC), cetyltrimethyl ammonium bro-mide (CTAB) and N-dodecyltrimethylammonium bromide(DTAB) as cationic surfactants, sodium dodecyl sulfate(SDS) as an anionic surfactant was studied on tensammetricpeak of each Brij. All of them decreased the tensammetricpeak observed for the Brijes. The tolerance limit , defined asthe concentration above which a change of more than threetimes of standard deviation was observed in the signalsobtained for the analytes, was 1.0 mg mL�1 for all the abovesurfactants.

4. Conclusions

The present study demonstrates that AC polarographybased on adsorption and desorption (tensammetry) ofnonionic surfactants can be used to determine several Brijesin aqueous samples. However, interferences from otheradsorbed compounds are the main drawback to the devel-opment of tensammetry as a viable technique for routineanalysis. ANN can be used to determine them simultane-ously and in the presence of known interference. Therefore,the above system can be a potential candidate for sensitiveand selective simultaneous determination of nonionicsurfactants.

5. Acknowledgement

The authors wish to acknowledge the support of this work byShiraz University Research Council.

6. References

[1] P. H. Brunner, S. Capri, A. Marcomini, W. Giger, Water Res.1988, 22, 1465.

[2] B. Breyer, H. H. Bauer, Alternating Current Polarographyand Tensammetry, Wiley, New York 1963.

[3] H. Jehring, J. Electroanal. Chem. 1969, 21, 77.[4] H. Jehring, Elektrosorptioanalyse mit der Wechselstrompo-

larographie, Akademie, Berlin 1974.[5] H. W. Nurnberg, P. Valenta, in The Nature of Seawater (Ed:

E. D. Goldberg), Dahlem Konf., Berlin 1975, pp. 87 – 136.[6] W. Davison, M. Whitfield, J. Electroanal. Chem. 1977, 75,

763.[7] P. Becher, in Nonionic Surfactants (Ed: M. J. Schick),

M.Dekker, New York 1966.[8] M. Bos, Anal. Chim. Acta 1984, 166, 261.[9] A. Szymanski, Z. Lukaszewski, Electroanalysis 1991, 3, 17

[10] A. Szymanski, Z. Lukaszewski, Electroanalysis 1991, 3, 963[11] J. R. M. W. J. Melssen, L. M. Buuydens, G. Kateman, Che-

mom. Intell. Lab. Syst. 1992, 22, 165.[12] J. H. Wikel, E. R. Dow, Bioorg, Medd. Chem. Lett. 1993, 3,

645.[13] M. Bos, A. Bos, W. E. Van der linden, Analyst 1993, 18, 323.[14] N. Maleki, A. Safavi, F. Sedaghatpour, Talanta, 2004, 64, 830.[15] G. Absalan, A. Safavi, S. Maesum, Talanta 2001, 55, 1227.[16] F. Despagne, D. L. Massart, Analyst 1998, 123, 157R.

Table 3. Estimated and actual concentrations of Brijes in synthetic samples.

Actual Found

Brij 30 Brij 35 Brij 56 Brij 96 Brij 30 Brij 35 Brij 56 Brij 96

Sample I 2.5 96.0 86.0 36.0 2.30� 0.06 99.0� 0.06 86.70� 0.01 34.8� 0.15Sample II 89.0 80.0 7.5 5.5 88.50� 0.01 80.0� 1.0 7.80� 0.40 4.05� 0.05



Talanta 64 (2004) 830–835

Single-step calibration, prediction and real samples data acquisitionfor artificial neural network using a CCD camera

N. Maleki∗, A. Safavi∗, F. Sedaghatpour

Department of Chemistry, College of Sciences, Shiraz University, Shiraz 71454, Iran

Received 29 September 2003; received in revised form 23 February 2004 ; accepted 27 February 2004

Available online 28 July 2004

Abstract

An artificial neural network (ANN) model is developed for simultaneous determination of Al(III) and Fe(III) in alloys by using chromeazurol S (CAS) as the chromogenic reagent and CCD camera as the detection system. All calibration, prediction and real samples data wereobtained by taking a single image. Experimental conditions were established to reduce interferences and increase sensitivity and selectivityin the analysis of Al(III) and Fe(III). In this way, an artificial neural network consisting of three layers of nodes was trained by applying aback-propagation learning rule. Sigmoid transfer functions were used in the hidden and output layers to facilitate nonlinear calibration. BothAl(III) and Fe(III) can be determined in the concentration range of 0.25–4�g ml−1 with satisfactory accuracy and precision. The proposedmethod was also applied satisfactorily to the determination of considered metal ions in two synthetic alloys.© 2004 Elsevier B.V. All rights reserved.

Keywords: Artificial neural network; CCD camera; Aluminum; Iron(III)

1. Introduction

There has been a rapid improvement in the technology ofdigital photography, both in terms of hardware and softwareperformance. The strong indications are that this rapid de-velopment will continue over the coming years, fuelled bythe need for cheap, high performance image acquisition forconsumer products and the web[1,2]. The combination ofdigital photography and colorimetric tests potentially offersa route to high throughput qualitative and quantitative ana-lytical measurements.

The use of digital imaging as a detector for colorimetricreactions has great potential for many applications whichinvolve the production of chemical color changes. The spa-tial resolution of the digital images, which can be producedby modern digital cameras facilitates the use of extremelysmall reagent areas, which can save on expensive or unavail-able reagents[3]. Digital camera is based upon the use ofcharge-coupled devices (CCDs). Hence, a digital camera can

∗ Corresponding author. Tel.:+98 711 2284822;fax: +98 711 2286008.

E-mail addresses: [email protected] (N. Maleki),[email protected] (A. Safavi).

act as an analytical detector, and generates a vast amount ofinformation per image. In addition, analysis times may belowered by taking single image, rather than having to scanthe spectrophotometer and change a large number of sam-ples.

The use of artificial neural networks (ANNs) in chemo-metrics has increased during the last decade[4,5]. ANNs areimportant class of pattern recognizer that may have manyuseful chemometric applications[6,7]. The feed-forwardneural networks trained by back-propagation of the errorshave one obvious advantage; there is no need to know theexact form of the analytical function on which the modelshould be built. Therefore, ANNs are suitable for experimen-tal conditions optimization since the relationship betweenevaluating indices and the experimental input parameters iscomplex, nonlinear and often cannot be expressed with acertain mathematical formula. Basic theory and applicationof ANN with back-propagation algorithm to chemical prob-lems can be found in the literature[8–12].

Chemometric methods such as ANN need many samplesfor calibration and prediction set, which is time consumingwith usual techniques such as spectrophotometry becauseof the increase in measuring time due to the need for themeasurement of each sample separately.

0039-9140/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.talanta.2004.02.041

N. Maleki et al. / Talanta 64 (2004) 830–835 831

By using CCD camera as the detection system, all cali-bration, prediction and real samples data can be obtained bytaking a single image and this decreases the analysis time.

Chrome azurol S (CAS) is a sensitive reagent for thedetermination of aluminum. Kashkovskaia and Mutafinused Chrome azurol S to determine aluminum in steels andaluminum bronzes[13]. They measured the absorbanceof the aluminum–chrome azurol S lake at 530 nm, whereBeer’s law is not obeyed. Choosing a wavelength of 567 nmfor absorbance measurements, the calibration curve ofaluminum–chrome azurol S complex was found to obeyBeer’s law from 0 to 1.2�g ml−1 Al [14]. Sensitive spec-trophotometric methods for the determination of Fe(III),based on complexes with a triphenylammonium reagent,chrome azurol S or eriochrome cyanine R (ECR) are de-scribed by Marczenko and Kalowska[15] and Shijo et al.[16]. In most of the reported methods for the determinationof Al(III), iron interferes [14–17]. The same is true for theinterfering effect of Al(III) on Fe(III) determination[15].In the last decade, the use of multivariate techniques suchas partial least square regression has received considerableattention particularly for simultaneous determination ofdifferent species such as aluminum and iron[18,19].

The main purpose of this work is to use CCD camera forsignal recording and application of ANN in simultaneousdeterminations. The proposed method was used for simul-taneous determination of Al(III) and Fe(III) with chromeazurol S.

2. Experimental

2.1. Apparatus

Measurements of pH were made with a Metrohm 691pH-meter using a combined glass electrode. A Reflecta Flec-talux XM 18 was used as a light source. A Canon EOS D30camera was used for taking pictures.

2.2. Reagents

All chemicals used were of analytical-reagent grade.Triply distilled water was used throughout. A stock solu-tion (8 × 10−3 M) of CAS, was prepared by dissolving0.0484 g of CAS (Fluka) in water and diluting to 10 ml ina standard flask. A stock solution (500�g ml−1) of Al(III)was prepared by dissolving 0.6928 g of Al (NO3)3·9H2O(Merck) in water and diluting to 100 ml. A stock solu-tion (500�g ml−1) of Fe(III) was prepared by dissolving0.4305 g of NH4Fe(SO4)2·12H2O (Merck) in water anddiluting to 100 ml. Acetic-acetate buffer (pH 6) was pre-pared by using acetic acid (1 M) and sodium hydroxide(1 M) solution [20]. Stock solution (500�g ml−1) of eachinterfering ion was prepared by dissolving an appropriateamount of the suitable salt.

2.3. Procedure

To a solution containing appropriate amounts of Al(III)and Fe(III) in a 5 ml volumetric flask were added 0.2 ml ofstock CAS solution, 2 ml of buffer solution, and the solutionmixture was made up to the mark with water. For each sam-ple, ca. 2 ml of the above solution was transferred to cells.There were 50 cells as calibration, prediction sets and realsamples sets. All data were obtained by taking an image.A program written in Visual Basic as software was usedto convert colors to red (R), green (G) and blue (B) valuesand ANN, which was trained with the back-propagationof errors learning algorithm, processed these values. Aback-propagation neural network having three layers wascreated with a Visual Basic software package.


3.1. Principles of the method

The spectrophotometric sensitive methods for determina-tion of aluminum and iron are based on systems, includingtriphenylmethane reagents, such as CAS. Chrome azurol Sreacts with Al(III) and Fe(III) ions to form water soluble col-ored complexes. The intensities of the color of CAS–Al andCAS–Fe complexes depend on the concentration of CAS.Since the color of the above complexes is distinct, CCDcamera can be used as a sensitive method for the determi-nation of Al(III) and Fe(III).

3.2. Digital camera and digital color

The analytical data that a digital camera returns are astandard trichromatic response, with 8-bit red, green andblue channels, respectively. Hence a value is returned to theuser ranging from 0 to 255 for each channel. The spectralpower distribution of light source and the spectral reflectancecurve of the object being analyzed both play an importantrole in the value of color obtained through each channel[21]. The colors returned by each of the three channels arereferred to as subtractive colors[21,22]. It should be notedthat the eye sees the converse of the color component that isprimarily absorbed, e.g. if an object appears red, it is becausethis color is being reflected or transmitted by the object andthe blue part of the spectrum is being absorbed[23].

The use of digital images opens the possibility of in-creased usage of processing by computer software withoutthe need for time consuming analysis by conventional spec-trophotometric techniques.

The data were analyzed by a program written in VisualBasic and was then transported to Excel.

In order to obtain RGB values of many samples simul-taneously two methods were implemented. In one method,small flat-bottomed wells were illuminated from the bottomand were photographed. This technique is very similar to

832 N. Maleki et al. / Talanta 64 (2004) 830–835

Fig. 1. Schematic photo of calibration, prediction and real sets used.

the one used by Rose[24]. The problem with this method isthe meniscus of the water surface and its consequent lens-ing effect, which destroys the homogeneity of the image.In addition, reflected light on the surface also poses colorhomogeneity problems. Teflon and other plastics show thesame problems. These problems can be seen from the pho-tos taken by Rose[24].

In order to solve the problems, regular disposable rectan-gular cells were placed in front of a white transparent plateand the plate was back illuminated by photographic lights.The images were taken from the front of the cells.

In order to get as many data as possible from one image,fifty 1-cm plastic disposable cells were cut in half and wereplaced in a vertical rack in five stages as shown inFig. 1.

In order to correct the background, a custom white balancewas set with the plate illuminated with the photographiclight. This procedure is absolutely essential in obtaining highquality data.

3.3. Optimization of experimental variables

3.3.1. Effect of pHThe sensitivity and selectivity are dependent on the pH of

the medium. Thus, the pH of solutions of Al and Fe complexwere varied and R, G and B values were obtained. At pH 6the difference of the RGB values was highest and this pHwas thus used as the optimum pH.

3.3.2. Effect of concentration of indicatorSince CAS is dark yellow at pH 6, its color interferes with

that of the complexes. In order to minimize this effect, same

Fig. 2. Determination range of Al(III) using red, green, and blue bandsobtained from aluminum standard solutions. Determination range usingANN obtained from aluminum standard solutions in the presence of iron.

amounts of Al and Fe were separately added to differentconcentrations of CAS. At 3.2×10−4 M CAS the differencein RGB values was a maximum.

3.4. Multivariate calibration with ANN

The data obtained from the image were processed byANN, which was trained with the back-propagation of er-rors learning algorithm. The first step in the simultaneousdetermination of different metal ions by ANN methodologyinvolves constructing the calibration set for the binary mix-ture Al(III)–Fe(III). In this study, the calibration and pre-diction sets were prepared randomly to reduce correlationbetween concentrations of Al(III) and Fe(III) (Tables 1 and2). Twenty seven binary mixtures were selected as the cal-ibration set and nine binary mixtures were selected as theprediction set for each ion. The training set of Al(III) andFe(III) were between 0.25 and 4�g ml−1, respectively, inthe calibration matrix. Also, the number of inputs was nor-malized between−3 and+3. Calibration set was used forconstruction of ANN model, and the independent test setwas used to evaluate the quality of the model.

The overall predictive capacity of the model was com-pared in terms of relative standard error, R.S.E. which isdefined as follows.

R.S.E. (%) = 100×(∑N

j=1(Cj − Cj)∑Nj=1(Cj)2

)1/2

Fig. 3. Determination range of Fe(III) using red, green, and blue bandsobtained from iron standard solutions. Determination range using ANNobtained from iron standard solutions in presence of aluminum.


Table 1Composition of calibration set and RGB data

Input data Concentration (�g ml−1)

R G B Al(III) Fe(III)

60 27 46 0.75 2.5059 4 124 3.50 1.5038 3 95 2.00 3.2575 4 128 4.00 0.5028 11 63 0.50 4.5079 5 102 3.00 1.0056 3 117 3.00 2.0098 18 53 1.25 1.2572 16 51 1.00 2.00

136 32 57 1.25 0.50103 11 84 2.75 0.5041 9 72 1.25 3.2578 4 115 2.00 0.7525 9 72 1.00 4.5035 25 49 0.00 3.7528 12 59 0.50 4.2541 12 58 1.00 3.0097 40 49 0.50 1.50

172 48 58 1.25 0.00127 105 32 0.00 1.2548 6 88 2.00 2.7575 9 73 2.00 1.5056 11 65 1.50 2.25

158 78 41 0.50 0.5058 6 91 3.50 2.2582 4 124 4.00 0.2576 3 144 4.50 0.50

whereCj is j-th component of the desired target (real sam-ple) andCj is thej-th component of the output produced bythe network (predicted sample).

To investigate the prediction ability of ANN, neural net-work models for individual component were also made withrespect to output layer considered as a single node corre-sponding to the analyte[25]. The construction of these ANNmodels is summarized inTable 3. The structure of networkwas comprised of three layers, an input, a hidden and anoutput layer. The addition of an extra parameter, called bias,to the decision function increases its adaptability to decisionproblem it is designed to solve. In this structure,+1 was ad-

Table 2Composition of prediction set and RGB data

Input data Concentration (�g ml−1)

R G B Al(III) Fe(III)

58 5 75 2.00 2.0036 4 79 1.50 3.5044 5 96 2.00 3.0053 11 59 1.00 2.75

104 19 60 1.50 1.0054 6 80 0.50 2.5079 14 56 1.25 1.7574 2 128 4.00 0.5091 4 145 4.50 0.13

Table 3Optimized parameters used for construction of ANN models

Parameter Compound

Al(III) Fe(III)

Input nodes 3 3Hidden nodes 4 3Output nodes 1 1Learning rate 0.5 0.5Momentum 0.08 0.2Hidden layer function Sigmoid SigmoidOutput layer function Sigmoid SigmoidNumber of iterations 8675 1050

dition as a bias. There are several ways to optimize the pa-rameters of a network[26–28]. One of them is constructedbased on the minimum error of prediction of the testing set[29–32]. Training and testing of the network is a process ofdetermination of the network’s topology and optimizationof its adjustable parameters where we seek the minimum ofan error surface in multi-dimensional space. The optimumlearning rate and momentum were evaluated by obtainingthose, which yielded a minimum in error of prediction (oftest set). The most versatile transfer function that can be usedto model a variety of relationship is the sigmoidal type. Inthe present nonlinear system, sigmoid functions were foundto be optimum for hidden and output layers. The number ofnodes in the hidden layer was determined by training ANNwith different number of nodes and then comparing the pre-diction errors from the test set.

The training was stopped manually when the root meansquare error of the test set remained constant after successiveiteration. The optimum number of iteration cycles or epochs(one pass of all objects, 27 mixtures, through the network)for each component was also obtained.Fig. 4shows a sampleplot of RSEC and RSEP as a function of the number ofiteration for Fe(III) components. Because there are severallocal minima, the algorithm was run from different startingvalues of initial weights to find the best optimum.

RSEP for Al and Fe were 4.8% and 5.6%. The reason-able relative standard error of prediction for each analyteindicates the accuracy of the proposed method (Table 4).

Table 4Composition of prediction set and relative standard errors for analytes

Actual Found

Al(III) Fe(III) Al(III) Fe(III)

2.00 2.00 1.87 2.251.50 3.50 1.44 3.552.00 3.00 1.90 2.961.00 2.75 0.99 2.511.50 1.00 1.33 0.980.50 2.50 0.39 2.401.25 1.75 1.32 1.724.00 0.50 4.08 0.544.50 0.13 4.30 0.19

R.S.E. (%) 4.80 5.61

834 N. Maleki et al. / Talanta 64 (2004) 830–835

Fig. 4. Plots of RSEP (%) and RSEC (%) as a function of the number of iterations for Fe(III).

The concentration of both Al(III) and Fe(III) sampleswere between 0.25 and 4.50�g ml−1. We could use onlyR, G or B value for analysis but using each color band,limits the determination range. In order to broaden thedetermination range and also to simultaneously deter-mine Al(III) and Fe(III), red, green, and blue bands inconjunction with ANN as a nonlinear method was used(Figs. 2 and 3).

3.5. Interferences

The selectivity was assessed by studying the influenceof foreign ions on the determination of Al(III) and Fe(III).The effect of interfering ions at different concentrations wasstudied on RGB values of a solution containing 1.25�g ml−1

of each analyte. The results are summarized inTable 5. Thetolerance limit was defined as the concentration above whicha change of more than three times of standard deviation wasobserved in the signals obtained for the analytes.

Table 5Effect of various ions on the determination of 1.25�g ml−1 Al(III) and1.25�g ml−1 Fe(III)

Ion added Tolerance level (�g ml−1)

Mn(II), Co(II), Ni(II) 200Cd(II) 5Pb(II) 50Ce(II) 100Cu(II) 4Cr(III), Cr(VI) 8Cl−, CH3COO−, Na+, Br− 500a

F−, PO43−, Citrate, Tartarate, Fe(II), Ga(III) 1.25

a Maximum concentration studied.

Table 6Estimated and actual concentration of Al(III) and Fe(III) in syntheticalloys

Sample Al(III) (�g ml−1) Fe(III) (�g ml−1)

Actual Found Actual Found

Ferro-aluminum 1.00 1.15± 0.04 3.00 2.77± 0.04Recovery (%) 115 92.3Ignition pin alloya 0.75 0.75± 0.10 2.00 1.93± 0.01Recovery (%) 100 96.5

a Containing 70.00�g ml−1 Co(II), 16.00�g ml−1 Zn(II),2.00�g ml−1 Fe(III), 0.75�g ml−1 Al(III), 1.50 �g ml−1 Mn(II).

3.6. Real sample analysis

The proposed method was also applied to the determi-nation of Al(III) and Fe(III) in synthetic alloys[33]. Eightcells were related to two real samples and their replication.The results are shown inTable 6. The good agreement be-tween these results and known values indicates the success-ful applicability of the proposed method for simultaneousdetermination of Al(III) and Fe(III) in real sample.

Acknowledgements

The authors gratefully acknowledge the support of thiswork by Shiraz University Research Council.

References

[1] P.M. Epperson, J.V. Sweedler, R.B. Bilhom, G.R. Sims, M.B. Denton,Anal. Chem. 60 (1988) 282A.


[2] Oeean Optics (online), available from:http://www.oeeanoptics.com/homepage.asp(accessed 31 January 2000).

[3] P.M. Epperson, J.V. Sweedler, R.B. Bilhorn, G.R. Sims, M.B. Denton,Anal. Chem. 60 (1988) 327A.

[4] J. Zupan, J. Gasteiger, Anal. Chim. Acta 248 (1991) 1–3.[5] G. Kateman, Cemom. Intell. Lab. Syst. 19 (1993) 135–142.[6] J.R. Long, V.G. Gregoriou, P.J. Gemperline, Anal. Chem. 62 (1990)

1791.[7] J. Schuur, I. Gasteiger, Anal. Chem. 69 (1997) 2398.[8] B.J. Wythoff, Chemom. Intell. Lab. Syst. 18 (1993) 115.[9] P.A. Jansson, Anal. Chem. 63 (1991) 357A.

[10] J. Zupan, J. Gasteiger, Angew. Chem. Int. Ed. Engl. 32 (1993) 503.[11] W. Guo, P. Zhu, H. Brodowsky, Talanta 44 (1997) 1995–2001.[12] Y. Ni, C. Lui, Anal. Chim. Acta 396 (1999) 221–230.[13] E.A. Kashkovaskaia, I.S. Mustafin, Zavodsk. Lab. 24 (1985) 1319.[14] P. Pakains, Anal. Chim. Acta 32 (1965) 57–63.[15] Z. Marczenko, H. Kalowska, Anal. Chim. Acta 123 (1981) 279–287.[16] Y. Shijo, T. Takeuchi, Bunseki Kagaku 20 (1971) 980.[17] R.M. Dagnall, T.S. West, P. Young, Analyst 90 (1965) 13.[18] A.R. Coscione, J. Carlos de Andrade, R.J. Poppi, C. Mello, B.V.

Raij, M.F. de Abreu, Anal. Chim. Acta 423 (2000) 31–40.[19] A.R. Coscione, J. Carlos de Andrade, R.J. Poppi, Analyst 127 (2002)

135.

[20] J. Lurie, Handbook of Analytical Chemistry (English translation),Mir, Moscow, 1975.

[21] E.J. Ciorgianni, T.E. Madden, Digital Color Management: EncodingSolution, Addison Wesley, USA, 1998, p. 9.

[22] http://developer.apple.com/documentation/GraghicsImaging/Conceptual/ManagingColorSync/ColorSync.pdf.

[23] L. Byrne, J. Barker, G. PennarunThomas, D. Diamond, Trends Anal.Chem. 19 (2000) 517.

[24] A.V. Lemmo, J.T. Fisher, H.M. Geysen, D.J. Rose, Anal. Chem. 69(1997) 543.

[25] F. Despagne, D.L. Massart, Analyst 123 (1998) 157R.[26] J.R.M. Smits, W.J. Melssen, L.M. Buuydens, G. Kateman, Chemom.

Intell. Lab. Syst. 22 (1992) 165.[27] J.H. Wikel, E.R. Dow, Bioorg, Med. Chem. Lett. 3 (1993) 645.[28] M. Bos, A. Bos, W.E. Van der linden, Analyst 118 (1993) 323.[29] M. Shamsipur, B. Hemmateenezhad, M. Akhond, Anal. Chim. Acta

461 (2002) 147.[30] G. Absalan, A. Safavi, S. Maesum, Talanta 55 (2001) 1227.[31] M. Baret, D.L. Massart, P. Fabry, F. Conesa, C. Eichner, C. Menardo,

Talanta 51 (2000) 863.[32] Y.B. Zeng, H.P. Xu, H.T. Liu, K.T. Wang, X.G. Chen, Z.D. Hu, B.

Fan, Talanta 54 (2001) 603.[33] http://chemistry.csudh.edu/oliver/chemdata/alloys.htm.

http://www.oeeanoptics.com/homepage.asp

http://www.oeeanoptics.com/homepage.asp

http://developer.apple.com/documentation/GraghicsImaging/Conceptual/ManagingColorSync/ColorSync.pdf

http://developer.apple.com/documentation/GraghicsImaging/Conceptual/ManagingColorSync/ColorSync.pdf

http://chemistry.csudh.edu/oliver/chemdata/alloys.htm

Analytica Chimica Acta 409 (2000) 275–282

Kinetic spectrophotometric determination of V(IV) in the presence ofV(V) by the H-point standard addition method

A. Safavi∗, H. Abdollahi, F. Sedaghatpour, S. ZeinaliDepartment of Chemistry, College of Science, Shiraz University, Shiraz, 71454, Iran

Received 4 June 1999; received in revised form 21 October 1999; accepted 7 November 1999

Abstract

The H-point standard addition method was applied to kinetic data for simultaneous determination of V(IV) and V(V)or selective determination of V(IV) in the presence of V(V). The method is based on the difference in the rate of complexformation between vanadium in two different oxidation states and xylenol orange at pH= 2. Vanadium(IV) can be determinedin the range 0.2–5mg ml−1 with satisfactory accuracy and precision in the presence of excess vanadium(V) and other metalions that rapidly form complexes with xylenol orange under working conditions. The proposed method was successfullyapplied to the simultaneous determination of V(IV) and V(V) and also the selective determination of V(IV) in the presenceof V(V) in several synthetic mixtures with different concentration ratios of V(IV) to V(V). ©2000 Elsevier Science B.V. Allrights reserved.

Keywords:Vanadium; H-point standard addition method; Kinetic method

1. Introduction

Vanadium is an important element in environmentaland biological studies. Vanadium’s role in physiolog-ical systems includes normalization of sugar levels,participation in various enzyme systems as an inhibitorand a co-factor [1] and catalysis of the oxidation ofvarious amines [2]. Although vanadium can exist inoxidation states from II to V in aqueous solution, mostmethods have concentrated on its determination in thestates IV and V, as these are the most common formsencountered in inorganic and biological systems. Alsoin vitro studies showed that the vanadium (IV) and(V) are essential for cell growth atmg dm−3 levels,but can be toxic at higher concentrations. The toxi-

∗ Corresponding author: Fax:+98-71-20027.E-mail address:[email protected] (A. Safavi).

city of vanadium is dependent on its oxidation state[3], with vanadium(V) being more toxic than vana-dium(IV). Recently, there has been renewed interestin the determination of V(IV) in the presence of V(V),since V(IV) is produced by various industrial redoxprocesses (e.g. the Linde Sulpholin process [4]), andis less toxic than V(V) in environmental systems [3].

UV-visible spectrophotometry is the most commontechnique used for vanadium determination, owing tothe sensitivity and selectivity achieved in these reac-tions [5]. With the current choice of reagents, tech-niques and conditions, it is also possible to distinguishbetween the various oxidation states of vanadium. Theanalytical applications of vanadium (IV) have beenreviewed; previously; it was concluded that V(IV) inaqueous solutions forms complexes most easily withreagents containing oxygen or sulfur donor ligands[6–10]. It is thus clear that the presentation of simple,

0003-2670/00/$ – see front matter ©2000 Elsevier Science B.V. All rights reserved.PII: S0003-2670(99)00868-5

276 A. Safavi et al. / Analytica Chimica Acta 409 (2000) 275–282

selective, precise and inexpensive methods for the de-termination of V(IV) in the presence of V(V) is veryimportant.

The H-point standard addition method (HPSAM)is a modification of the standard addition methodthat transforms the incorrigible error resulting fromthe presence of a direct interference in the determi-nation of an analyte into a constant systematic error.This error can then be evaluated and eliminated. Thismethod also permits both proportional and constanterrors produced by the matrix of the sample to becorrected directly. The basis of the method was es-tablished previously [11,12]. Absorbance incrementsas analytical signals were used when only the ana-lyte concentration was required [13]. The method hasbeen applied to eliminate the blank bias error dueto the use of the absorbant blank [14,15], to liquidchromatography [16] and to the analysis of kineticdata [17], with time as an additional variable.

Two variants of HPSAM can be used for the treat-ment of kinetic data [17]. One is applied when thereaction of one component is faster than that of theother or the latter does not take place at all. This vari-ant of the method is based on the assumption that onlyanalyte X evolves with time and the other species Yor interferents do not affect the analytical signal withtime. In this case the variables to be fixed are twotimes t1 and t2 at which the species Y, which doesnot evolve with time or over the range between thesetimes, should have the same absorbance. The othervariant of the method is used when the rate constants ofthe two components are time-dependent. In this case,the two species in a mixture, X and Y, evolve with time,Cx (concentration of analyte) andAy (the absorbanceof interference) can be calculated by plotting the ana-lytical signal1At1−t2 against the added concentrationof X at two wavelengthsλ1 andλ2, provided that theabsorbances of the Y component at these two wave-lengths are the same (AY ) and so are thus the1A t1−t2

values.In this work the first variant of the H-point standard

addition method is suggested as a simple and selectivemethod for the determination of V(IV) in the presenceof V(V). This method is based on the difference in therate of complex formation of V(IV) and V(V) withxylenol orange under acidic conditions.

The well-known metallochromic indicator, xylenolorange (XO) [18,19] has been proposed for the colori-

metric determination of a number of metals [20–23].It is evident that XO shows only slight selectivity,so that direct determination of metal ions with XOusually suffers from serious interferences. AlthoughXO has been used for the determination of vanadiumpreviously [23], to the best of our knowledge, thecomplexation kinetics of V(IV) and XO were notconsidered previously for selective determination ofV(IV) in the presence of V(V).

2. Experimental

2.1. Reagents

All chemicals used were of analytical reagent grade.Triply distilled water was used throughout. Stockvanadium(V) and (IV) solutions (1000mg ml−1), wereprepared from ammonium metavanadate and vanadylsulfate monohydrate (Fluka), respectively, and stan-dardized [24,25]. XO stock solution (0.01 M) wasprepared from its disodium salt (Fluka). The workingsolutions were diluted with water to the appropriateconcentrations. 1000mg ml−1 solutions of the studiedinterfering ions were prepared from appropriate salts.A buffer of pH 2 was prepared by using sodium citrateand hydrochloric acid at appropriate concentrations[26].

2.2. Apparatus

UV-visible absorbance spectra were recorded on aPhilips 8750 spectrophotometer. Measurements of pHwere made with Metrohm 691 pH meter using a com-bined glass electrode. The absorbance measurementsas a function of time, at fixed wavelength, were madewith a Philips PU875 spectrophotometer attached to aPentium 200 MHz computer.

2.3. Procedure

2 ml of buffer solution, 0.2 ml of stock XO solu-tion and appropriate volumes of V(IV) and V(V) wereadded to a 5 ml volumetric flask and made up to themark with water. For each measurement, ca. 2 ml ofthe above solution was transferred to a spectropho-tometer cell and the variation of absorbance vs. time

A. Safavi et al. / Analytica Chimica Acta 409 (2000) 275–282 277

was recorded immediately. The absorbance was mea-sured at 562 nm with 5 s time intervals for each sam-ple. Simultaneous determination of V(IV) and V(V)with HPSAM was performed by measuring the ab-sorbances at just 60 and 240 s for each sample solution.The concentration ranges of V(IV) and V(V) for theconstruction of a HPSA calibration graph were 0.2–5and 0.1–3.0mg ml−1, respectively. The same proce-dure was repeated with 2 ml of samples such as bloodserum, river water, tap water or synthetic samples.


HPSAM as a modified standard addition methodwas proposed to obtain an unbiased analyte concentra-tion when both analyte and interference are present in asample. This method can be applied to kinetic data forthe simultaneous determination of binary mixtures orcalculation of analyte concentrations completely freefrom bias.

The rate of complex formation between XO andV(IV) is slower than V(V) complex formation underacidic conditions. Fig. 1 shows the change in visi-ble absorption spectra of the XO–V(IV) system as afunction of time. The study on the influence of pH onthe rate of V(IV)–XO complex formation showed that

Fig. 1. Variation of visible absorption spectra of XO (3.0× 10−4 M) solution in the presence of 2.0mg ml−1 V(IV), as a function of time.Time interval 1 min.

the maximum change in absorbance in the fixed timeperiod (30–60 s) occurred at ca. pH= 2. The rate ofcomplex formation above pH 3 was so fast that thecomplex formation reaction was almost complete be-fore 30 s and thus the change in absorbance between30 and 60 s was very small (Fig. 2). So, 562 nm andpH 2 were selected for monitoring the kinetics of thesystem. However, under the same condition, the com-plexation of V(V) was very fast.

Fig. 3 shows the possible applicability of HPSAMfor the simultaneous determination of V(IV) and V(V)by the proposed system. The absorbances correspond-ing to V(IV) at 562 nm and at two selected times 60and 240 s arebi andAi , respectively, while those cor-responding to V(V) under the same condition arebandA′. A′ andb are equal because of the fast kineticsof complex formation between XO and V(V). Theyare related, in the simplest assumption, through thefollowing equations [17]:

for V(IV ) Ai = bi + miti 60 ≤ tj ≤ 240

i = 0, 1, ..., n (1)

for V(V) or interferentA′ = b + m tj (m = 0) (2)

where the subscriptsi and j denote the different solu-tions forn additions of V(IV) concentration preparedin order to apply the HPSAM and the time compris-


Fig. 2. Effect of pH on the change in absorbance of the XO–V(IV) system in a fixed time (between 30–60 s).

ing the 60–240 s range, respectively. These two timeswere selected because the change in absorbance re-lated to V(V) concentration was negligible in this timeinterval and also there was an appropriate differencebetween the slopes of the calibration lines for V(IV)at these two times. Thus, the overall absorbances ofthe V(IV)–V(V) mixture at 60 and 240 s are:

at 60 s A60 = bi + b (3)

at 240 s A240 = Ai + A′ (4)

However, applications of HPSAM at the twoaforesaid times are:

Fig. 3. Kinetic curves for (a) 1.0mg ml−1 V(IV) (b) 1.0mg ml−1 V(V) (c) mixture of 1.0mg ml−1 V(V) and 1.0mg ml−1 V(IV).

A60 = b0 + b + M60Ci (5)

A240 = A0 + A′ + M240Ci (6)

which intersect at point H(−CH, AH) ≡ (−CV(IV ),AV(V)) (Fig. 4). M60 and M240 are the slopes of thecalibration lines at 60 and 240 s, respectively.

At the intersection point:

b0 + b + M60CH = A0 + A′ + M240CH (7)

hence

CH = (A′ − b) + (A0 − b0)

(M60 − M240)(8)


Fig. 4. Plot of H-point standard addition method for simultaneous determination of V(IV) (1.05mg ml−1) and V(V) (1.85mg ml−1).

Table 1Results of four replicate experiments for the analysis of V(IV)–V(V) mixtures

A–C equation r Present in the sample (mg ml−1) Found (mg ml−1)

V(IV) V(V) V(IV) V(V)

A240= 0.304Ci + 0.520 0.9998 0.62 1.85 0.63 1.82A60 = 0.192Ci + 0.451 0.9988A240= 0.300Ci + 0.401 0.9997 0.21 1.85 0.23 1.83A60 = 0.190Ci + 0.576 0.9989A240= 0.304Ci + 0.520 0.9998 1.86 0.92 1.76 0.90A60 = 0.188Ci + 0.464 0.9999A240= 0.300Ci + 0.317 0.9997 0.74 0.78 0.74 0.73A60 = 0.188Ci + 0.234 0.9998

As the V(V) complex is assumed not to evolve overthe considered range of time, thenA′ = b and

CH = (A0 − b0)

(M60 − M240)(9)

which is equivalent to the existingCV(IV )(= b0/M60= A0/M240). So, as shown in Fig. 5(a), the value ofCH is independent of the concentration of V(V) in thesample.

Substitution ofCV(IV ) into Eq. (5), yieldsAH = b.The overall equation for the absorbance at the H-pointis simply:

A′ = b = AH = AV(V) (10)

So, again as shown in Fig. 5(b) the value ofAH isindependent of the amount of V(IV) in the sample.

Therefore, the intersection of the straight lines byEqs. (5) and (6) directly yields the unknown V(IV)concentration (CV(IV )) and the analytical signal of the

Table 2Determination of V(IV) in different samples

Sample Concentration of V(IV) (mg ml−1)

Spiked Founda

Blood serum 0.50 0.45± 0.08Blood serum 0.75 0.78± 0.07River water 0.50 0.52± 0.05Tap water 0.25 0.27± 0.04Syntheticb 0.25 0.22± 0.04

a Mean± s.d. (n= 5).b The composition of the synthetic sample was, V(V) 2.5mg

ml−1, Cu(II) 5.1mg ml−1, Cd(II) 40mg ml−1, Pb(II) 12mg ml−1,Ca(II) 6mg ml−1, Zn(II) 2.0mg ml−1, Na(I) 276mg ml−1.


Fig. 5. Plots of H-point standard addition method for (a) fixed V(IV) concentration (1.24mg ml−1) and different concentration of V(V);(b) fixed V(V) concentration (1.85mg ml−1) and different concentration of V(IV).

V(V) species (AV(V)) corresponding tot60 andt240 inthe original samples, as the two times were chosen insuch a way that the latter species had the same ab-sorbance at both times. This analytical signal enablesthe calculation of the concentration of V(V) from acalibration graph. Several synthetic samples with adifferent concentration ratio of V(IV) and V(V) wereanalyzed by HPSAM . As can be seen from Table 1,the accuracy of the results is satisfactory in all cases.In addition, for the validation of the method, environ-

mental and biological complex matrix samples werespiked with V(IV) and the proposed method was ap-plied for determination of the analyte. The results areshown in Table 2; it can be seen that the results aresatisfactory.

An absorbance increment as an analytical signal canbe employed in another version of HPSAM to allowthe analyte concentration,CV(IV ), to be calculated withno systematic, constant or proportional error thanks tothe intrinsic feature of the HPSAM and the nature of


Fig. 6. 1A vs. added V(IV) concentration at different time intervals and 562 nm for synthetic mixtures with (a) 0.77mg ml−1 V(IV) and0.74mg ml−1 V(V) (b) 1.53mg ml−1 V(IV) and 2.50mg ml−1 V(V).

Table 3Application of signal increment version of HPSAM to two synthetic mixtures

Time interval (s)

100–240 150–240 100–150 100–200 80–200 60–240 60–200 90–150 110–170 80–200

V(IV) Found (mg ml−1) 0.73 0.71 0.71 0.76 0.70 1.57 1.57 1.54 1.57 1.56Actual V(IV) (mg ml−1) 0.77 1.53Relative standard deviation (%) 2.6 (n= 5) 0.6 (n= 5)


the method of standard addition. This variant shouldbe of use whenever only the analyte is to be deter-mined or when theA–t curve rather than the compo-sition of the matrix is known. It should also be usefulto apply the single-standard calibration method, whichfeatures greater simplicity and rapidity. This methodcan also be used to diagnose the occurrence of inter-ferences with a given analytical procedure as, in theabsence of errors, the plot of any1At1−t2 against theadded analyte concentration will have a constant point(−CH, 0) [17]. Thus, for the determination of V(IV)in the presence of V(V), the absorbance incrementas an analytical signal was used. The application ofthe HPSAM in the1At1−t2–Caddedvariant yields theconcentration of V(IV) directly from the intercept onthe y-axis. However, in order to ensure the absenceof constant and proportional errors from the calcu-lated concentration, all the possible1At1−t2–Caddedlines for V(IV) should intersect at the same point,namely, that corresponding to the unknown concen-tration,CH, as this would indicate that the time eval-uation of the matrix would be a horizontal line. Fig.6 and Table 3 show the results obtained from em-ploying this version of HPSAM on the synthetic mix-tures of V(IV) and V(V). The results obtained by thisprocedure were in good agreement with those givenabove.

The behavior of the system in the presence of someions was also studied. As (III), Cd2+, Sn (II), Pb2+,Ca2+, Na+ and K+ did not complex with XO un-der working conditions and thus up to 100mg ml−1

of these ions did not interfere with the determinationof 0.50mg ml−1 V(IV). Cu2+, Hg2+, Zn2+, Mg2+,Co2+, Se(IV), Te(IV) and Fe3+ instantly form com-plexes with concentrations up tomg ml−1 level, andbecause complex formation is fast, no interference wasobserved from their presence in the determination ofV(IV). However, only Al3+, Ni2+ and Fe2+ showedan interference, due to slow complex formation at themg ml−1 level.

Acknowledgements

The authors gratefully acknowledge the support ofthis work by the National Research Council of I.R.

Iran (NRCI) as a National Research Project under thegrant number 1081.

References

[1] D.C. Crans, M. Shia Gottlieb, J. Tawara, R.L. Bunch, L.A.Thiesen, Anal. Biochem. 188 (1990) 53.

[2] H.E. Stokinger, in: G.D. Clayton, F.E. Clayton (Eds.), Putty’sIndustrial Hygiene and Toxicology, 3rd Edition, Vol. 24,Wiley-Interscience, New York, 1981, p. 2028.

[3] B. Patel, G.E. Henderson, S.J. Haswell, R. Grzeskowiak,Analyst 115 (1990) 1063.

[4] S. Vermeir, R.De. Haan, Ind. Eng. Chem. Res. 27 (1988)1242.

[5] M.J.C. Taylor, J.F. van Staden, Analyst 119 (1994) 1263.[6] M. Jarosz, Z. Marczenco, Analyst 109 (1984) 35.[7] A. Safavi, A. Massoumi, S. Razi, N. Maleki, B. Haghighi,

Mikrochem. J. 42 (1990) 319.[8] C.K. Verma, S. Kulshrestha, O. Prakash, Chem. Anal. 28

(1983) 669.[9] J. Azanarez, J.C. Vidal, M.A. Navas, C. Diaz, Indian J. Chem.

29A (1990) 1240.[10] M.J.C. Taylor, J.F. van Staden, Anal. Chim Acta 307 (1995)

1.[11] F. Bosch-Reig, P. Campina, P. Campins-Falco, Analyst 113

(1988) 1011.[12] F. Bosch-Reig, P. Campina, P. Campins-Falco, Analyst 115

(1990) 111.[13] P. Campins-Falco, F. Bosch-Reig, J. Vendu Andres, Talanta

39 (1992) 1.[14] P. Campins-Falco, F. Bosch-Reig, J. Vendu Andres, Anal.

Chim. Acta 270 (1992) 253.[15] P. Campins-Falco, F. Bosch-Reig, J. Vendu Andres, C. Molins

Legua, Talanta 41 (1994) 39.[16] P. Campins-Falco, F. Bosch-Reig, R. Heraez Hernandez, A.

Sevillano Cabeza, Anal. Chim. Acta 257 (1992) 89.[17] P. Campins-Falco, F. Bosch-Reig, R. Heraez Hernandez, A.

Sevillano Cabeza, C. Molins Legua, Anal. Chem. 63 (1991)2424.

[18] J. Korbl, R. Pribil, A. Emar, Coll. Czech. Chem. Comm. 22(1957) 961.

[19] J. Korbl, R. Pribil, Chemist–Analyst 46 (1956) 102.[20] K.L. Cheng, Talanta 2 (1959) 61.[21] M. Otomo, Bull. Chem. Soc. Jpn. 36 (1963) 140.[22] K.I. Cheng, Talanta 3 (1959) 147.[23] O. Budevsky, R. Pribil, Talanta 11 (1964) 1313.[24] N.H. Furman, Standard Methods of Chemical Analysis, 6th

Edition, Vol. 1,Van Nostrand, 1962, pp. 1207–1227.[25] I.M. Kolthof, R. Belcher, V.A. Stenger, G. Matsuyama,

Volumetric Analysis, Vol. 3, Interscience, New York, 1957,p. 93.

[26] Lurie, J., Handbook of Analytical Chemistry, (Englishtranslation), Mir, Moscow, 1975.

Full Paper

Ionic Liquids Modify the Performance of Carbon BasedPotentiometric Sensors

Afsaneh Safavi,* Norouz Maleki,* Fatemeh Honarasa, Fariba Tajabadi, Fatemeh Sedaghatpour

Department of Chemistry, College of Sciences, Shiraz University, Shiraz, 71454, Iran*e-mail: [email protected]

Received: October 31, 2006Accepted: December 27, 2006

AbstractA new type of potentiometric sensor based on a recently constructed carbon ionic liquid electrode (CILE) isdescribed. Two kinds of ionic liquids, i.e., N-octylpyridinium hexafluorophosphate (OPFP) and 1-butyl-3-methylimidazoluim hexafluorophosphate (BMFP) were tested as binder for construction of the carbon compositeelectrode. The characteristics of these electrodes as potentiometric sensors were evaluated and compared with thoseof the traditional carbon paste electrode (CPE). The results indicate that potentiometric sensors constructed withionic liquid show an increase in performance in terms of Nernstian slope, selectivity, response time, and responsestability compared to CPE.

Keywords: Carbon ionic liquid electrode, Potentiometric sensor

DOI: 10.1002/elan.200603767

1. Introduction

Ion selective electrodes (ISEs) have become popular inenvironmental and biological fields because of their inter-esting abilities such as simple, quick, low cost, accurate andwide-dynamic-range analysis. The majority of works onISEs are based on the use of polymeric membranes as thesuitablematerial for construction of potentiometric sensors.But finding a suitable internal solution and selection ofcomposition to obtain an elastic and stable membrane areaccounted as a main trouble in these kinds of electrodes.Therefore, in recent years carbon paste electrodes (CPEs)have attracted attention as ion selective electrodes mainlydue to their advantages over membrane electrodes such aschemical inertness, robustness, low cost, renewability, stableresponse, lowohmic resistance, no need for internal solutionand suitability for a variety of sensing and detectionapplications [1 – 9].The CPE based potentiometric sensors which have been

introduced up to now aremainly based on incorporation of aselective agent into carbon paste. The carbon paste usuallyconsists of graphite powder dispersed in a nonconductivemineral oil. Incorporation of mineral oil gives CPE somedisadvantages. Mineral oil is not component-fixed since it isinvolved in various refining of petroleum and processing ofcrude oil, and some unaccounted ingredients may engenderunpredictable influences on detection and analysis [10]. Inaddition they have mechanical problem, their mechanicalstability is something between membrane electrodes andsolid electrodes.Ionic liquids (ILs) make a lot of impact on branches of

analytical chemistry [11, 12]. It has been reported that IL

incorporates several advantages in the membrane electro-des. It was approved that the presence of IL (1-butyl-3-methylimidazolium hexafluorophosphate) as additive andionophore (polyazacycloalkane derivative) at the same timeimproved selectivity of sulfate in PVCmembrane electrode[13]. A successful utilization of imidazolium and phospho-nium based ILs has been reported in various polymermembranes as both plasticizers and ion responsive media[14].Membranes containing IL even promote the character-ization of all-solid-state Ag/AgCl reference microelectrode[15]. Although the above reports have been appearedrecently in the literature on the use of IL in potentiometricsensors, to the best of our knowledge there has been noreport on incorporation of IL in to carbon-based potentio-metric sensors.Recently we introduced and characterized the electro-

chemical behavior of a new type of carbon compositeelectrode called carbon ionic liquid electrode (CILE), inwhich an ionic liquid, N-octylpyridinium hexafluorophos-phate (OPFP), has been used as a binder [16]. Thevoltammetric responses of CILE towards different inorgan-ic, organic and biomolecules have been reported [16 – 18].Room-temperature ionic liquids (ILs) are a good choice asbinder in carbon composite electrodes, because of theirinteresting properties, such as stability, low vapor pressure,low toxicity, low melting temperature, high ionic conduc-tivity and good electrochemical and thermal stability. As wehave demonstrated previously [16], this type of electrodehas lower ohmic resistance than CPE and gives veryreproducible and sensitive voltammetric results. However,the performance of this type of electrode has not yet beentested as a potentiometric sensor.

582

Electroanalysis 19, 2007, No. 5, 582 – 586 C 2007 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim

Besides the lower ohmic resistance of CILE with respectto CPE [16], it is expected that the presence of a hydrophilicbinder in the electrode rather than the hydrophobic, non-component fixed mineral oil used in traditional CPE, cansignificantly alter the response of the electrode as thepotentiometric sensor. The majority of potentiometricsensors use a selective agent in the paste to increaseselectivity of the sensor. However, when the ionophoredoes not form a strong interaction with the ion of interest,the response of the electrodemainly depends on the relativepartition coefficient of the ion of interest in water and in theelectrode.Our main goal in this work is to determine how

incorporation of a hydrophilic binder can change theperformance of a carbon paste electrode. For this purposewe solely investigated the response characteristics ofpotentiometric carbon ionic liquid electrodes withoutincorporation of any selective reagent, and compared thistype of electrode with an unmodified CPE. There are fewreports on the use of unmodified CPEs as potentiometricsensors. It has been reported previously that unmodifiedCPE can be used for the determination of Agþ [19].However, Cu2þ, Fe3þ, Cr3þ, Hg2þ and Al3þ severely inter-fered with this determination. To increase the applicabilityof such a sensor Shamsipur et al. used artificial neuralnetwork for simultaneous potentiometric determination ofAgþ, Hg2þ and Cu2þ [20].In this paper, we report on the successful utilization of

CILE as a potentiometric sensor and will demonstrate thatthe use of ionic liquid increases the analytical performanceof the carbon electrodes as potentiometric sensors. Thecharacteristics of this electrode have been compared withthose of CPE based potentiometric sensors.

2. Experimental

2.1. Reagents and Apparatus

All reagentswereof analytical reagent grade.Triply distilledwater was used throughout. Graphite powder (Fluka),particle size <100 mm and paraffin oil (Merck) were usedfor preparation of CPE. Stock solutions (0.1 M and 0.01 M)of appropriate salts of different ions (all from Merck) wereprepared. Ionic liquids,N-octylpyridinium hexafluorophos-phate (OPFP) and 1-butyl-3-methylimidazoluim hexafluor-ophosphate (BMFP) were synthesized as described else-where [21 – 23]. The electromotive force (EMF) measure-ments were performed at 25� 2 8C with Metrohm 691Digital pH-Meter.

2.2. Procedure

2.2.1. Preparation of Electrodes

The traditional carbon paste electrode (CPE) was preparedby hand-mixing of the graphite powder and paraffin (70%–

30%). The mixture was mixed until a uniform paste wasobtained. The paste was packed into the cavity of a Teflontube. Electrical contact to the carbon paste was establishedvia a stainless steel handle. Fresh surface was obtained byapplying manual pressure to the piston. The resulting freshsurface was polished on a paper until the surface had a shinyappearance. It is important to renew the electrode surfacewhen the solution changed because residual ion will still beadsorbed on the surface of the CPE, which will lead to poorreproducibility.The carbon ionic liquid electrodes (CILEs)wereprepared

similar to CPE. In these electrodes either OPFP or BMFPwere used instead of paraffin with different percentages.When using OPFP which is solid at room temperature, assuggested previously [16], heating the ionic liquid- graphitemixture to a temperature higher than the melting point ofOPFP (m.p. 65 8C) is essential.

2.2.2. Emf Measurements

The CPE or CILE was used as an indicator electrode. Adouble-junction silver/silver chloride electrode (Metrohm)was used as a reference electrode. The potential response ofthe electrode was investigated by measuring the emf valuesof a solution of the desired ion over a concentration range byadding successive aliquots of known concentrations of theion of interest to 25 mL of triply distilled water.

3. Results and Discussion

Preliminary experiments were performed to elucidate theeffect of replacing mineral oil with ionic liquid in the carbonpaste on potentiometric responses. For this purpose thepotentiometric responses of the unmodified CPE andCILEs towards different cations were studied. Selectivity,response time, calibration slope, linear range, and responsestability are important parameters in studies of ion selectiveelectrodes and are going to be discussed herein.

3.1. Electrode Composition

It is known that the sensitivity and linearity for a givenelectrode depend significantly on the electrode composi-tion. For investigation of the electrode composition, sevenelectrodes were prepared. The compositions of theseelectrodes are shown in Table 1. The potentiometric re-sponse of each electrode towards different ions was inves-tigated and the results are compared. The OPFP containingCILE has a higher mechanical stability than BMFP basedCILE and CPEs.

3.2. Potentiometric Responses

Since no modifier was used in the electrodes studied, theselectivity sequence is related to the relative cation partition

583Performance of Carbon Based Potentiometric Sensors

Electroanalysis 19, 2007, No. 5, 582 – 586 www.electroanalysis.wiley-vch.de C 2007 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim

www.electroanalysis.wiley-vch.de

coefficients between water and the electrode. Thus, theresponse of traditional CPE to some cations is incurred byparaffin liquid extraction and by the exchange currentcaused by some cations on the surface of the carbon.Paraffin liquid on the graphite is apt to extract compoundswith nonpolar character. The extent of extraction of lip-ophilic cations is thus expected to be higher in CPE ratherthan CILE which contains ionic liquid with much higherdielectric constant as the binder. Among different cationstested with CPE (electrode no.1), potentiometric responseswere observed for six cations, namely Agþ, Hg2þ, Fe3þ, Cr3þ,Cu2þ and Al3þ. In comparison, the electrodes prepared withionic liquid respond primarily towards cations such as Agþ

Hg2þ and Fe3þ. Tables 2 and 3 show the linear ranges,calibration slopes and correlation coefficients obtained for

potentiometric responses of different electrodes towardsdifferent ions, respectively. These tables also show that theelectrodes containing ionic liquid show better selectivity. Itis also clear from Table 3, that among all the cations tested,the response of all the electrodes were approximatelyNernstian towards Agþ. Potential dependences of electro-des no.1, 3 and 6 on Agþ cation are represented in Figure 1.Thus, in the rest of this study, the responses of differentelectrodes are compared towards Agþ ion. The responsetime of all the electrodes were measured. For this purpose,the average time needed for the electrode to reach 90% ofthe potential response after successive immersion in a seriesof Agþ solutions, each having 10 fold differences inconcentration, are reported in Table 4.

Table 1. Electrodes with different composition.

Composition No.

1 2 3 4 5 6 7

Graphite (%) 70 60 70 80 60 70 80Paraffin (%) 30 – – – – – –N-Octylpyridinium hexafluorophosphate(%) – 40 30 20 – – –1-Butyl-3-methylimidazoluim hexafluorophosphate(%) – – – – 40 30 20

Fig. 1. Potential dependence of Agþ cation for electrodes no.1, 3, and 6.

Table 2. Linear range for different cations using different electrodes.

Electrodenumber

Linear range

Agþ Hg2þ Fe3þ Cr3þ Cu2þ Al3þ

1 9.8� 10�6 – 9.2� 10�3 3.1� 10�5 – 7.4� 10�5 2.0� 10�6 – 4.6� 10�5 2.7� 10�3 – 1.6� 10�2 8.0� 10�6 –3.1� 10�3

2.6� 10�3 –1.6� 10�2

2 1.3� 10�5 – 4.5� 10�3 1.7� 10�5 – 6.4� 10�5 1.1� 10�4 – 6.2� 10�3 2.4� 10�3 – 1.2� 10�2 – –3 6.2� 10�6 – 1.3� 10�2 1.3� 10�5 – 6.7� 10�5 1.4� 10�4 – 1.2� 10�2 4.0� 10�3 – 1.5� 10�2 – –4 2.0� 10�5 – 4.0� 10�4 1.9� 10�5 – 4.9� 10�5 2.1� 10�4 – 7.7� 10�3 1.3� 10�3 – 9.7� 10�3 – –5 3.9� 10�5 – 4.9� 10�3 2.4� 10�5 – 1.8� 10�4 4.0� 10�4 – 7.7� 10�3 – – –6 2.07� 10�5 – 6.33� 10�3 6.5� 10�5 – 1.6� 10�4 1.1� 10�4 – 1.0� 10�3 – – –7 1.1� 10�5 – 6.4� 10�4 3.0� 10�5 – 8.6� 10�5 5.9� 10�4 – 7.2� 10�3 – – –

584 Afsaneh Safavi et al.



As a result of comparing the responses of all electrodestowardsAgþ, electrodeno. 3 containingOPFP, graphitewiththe ratio of 30 :70 displayed the best performance as apotentiometric sensor.

3.1.2. Stability

The equilibrium potentials remained essentially constantfor about 170 min when CILE was used as the sensor, afterwhich only a small divergence was observed. In comparison,the CPE response was constant up to 50 min after which asteady change in potential reading was observed. Thestandard deviations of the potential responses over a period

of 4 h in a 1 mM solution of Agþ ion were 1.5, 2.0 and 4.3(n¼ 25), for electrodes no. 3, 6, and 1, respectively (Fig. 2).This shows that the presence of ionic liquid as the binder,imparts better stability in carbon composite electrodes.

3.1.3. Effect of pH

The effects of pH value have been investigated on CPE andCILEs. The solution was diluted with acetate buffers withpH values in the range of 4 to 6.2 (a limited range due toprecipitation of Agþ at higher pH values) and the responsesfor Agþ were recorded. The obtained results are illustratedin Table 5.Variation of pH values did not cause serious change in

slope. But the best results were obtained at pH 4.1 for thethree electrodes tested. Comparison between the threeelectrodes reveals that CILE with OPFP as the binder showthe best slope at different pH values.

4. Conclusion

This study shows that replacing paraffin with ionic liquid incarbon paste electrodes, modifies the performance of theelectrode as potentiometric sensor. The electrode contain-ing 30%–70% OPFP – graphite offers the best Nerntianslope, response stability, and response time. The potentio-metric responses are very stable and the electrodes can beeasily renewed by removing the surface. This electrodeshows the least variation in response with changing pH from4 to 6 and is thus better suited for work on real samples.

Table 3. Calibration slopes and correlation coefficients for differ-ent electrodes

No. of electrodes Calibration slope (mV/decade)Correlation coefficient (R)

Agþ Hg2þ Fe3þ Cr3þ Cu2þ Al3þ

1 56.2 138.3 121.5 126.9 29.7 183.00.998 0.994 0.995 0.997 0.985 0.994

2 53.0 32.25 81.0 212.4 – –0.993 0.996 0.997 0.992 – –

3 58.8 50.0 103.8 143.6 – –0.997 0.999 0.997 0.997 – –

4 55.9 62.8 92.6 143.6 – –0.984 0.984 0.998 0.980 – –

5 53.2 306.6 104.1 – – –0.998 0.994 0.999 – – –

6 54.7 53.0 153.4 – – –0.997 0.982 0.996 – – –

7 62.3 261.0 92.2 – – –0.998 0.992 0.998 – – –

Table 4. Response times of different electrodes towards Agþ ion.

Type of electrodes 1 2 3 4 5 6 7Response time (s) 15 13 8 9 10 8 10

Fig. 2. Potential stability of electrodes no.1, 3, and 6.

Table 5. Calibration slope for different electrodes towards Agþ inseveral pH range

No. of electrodes pH 4.1 pH 5.0 pH 6.2

1 60.3 63.1 63.13 58.3 60.5 60.36 60.3 63.2 63.7

585Performance of Carbon Based Potentiometric Sensors



5. Acknowledgement

Theauthorswish to acknowledge the support of thiswork byIranian Ministry of Sciences, Research and Technology.

6. References

[1] J. Jezkova, J. Musibva, K. Vytras, Electroanalysis 1997, 9,1433.

[2] M. N. Abbas , J. Pharm. Bio. Anal 2003, 31, 819.[3] M. Shamsipur, S. Ershad, N. Samadi, A. Moghimi, H.

Aghabozorg, J. Solid State Electrochem. 2005, 9, 788.[4] A. Abbaspour, S. M. M. Moosavi, Talanta 2002, 56, 91.[5] H. Ibrahim, Anal. Chim. Acta 2005, 545, 158.[6] M. J.Gismera, D. Hueso, J. R. Procopio, M. T.Sevilla, Anal.

Chim. Acta 2004, 524, 347.[7] K. Vytras, E. Khaled, J. Jezkova, H. N. A. Hassan„B. N.

Barsoum, Fresenius J. Anal Chem. 2000, 367, 203.[8] M. H. Mashhadizadeh, A. Mostafavi, H. Allah-Abadi,

I.Sheikhshoai, Sens. Actuators B 2006, 113, 930.[9] H. Ibrahim, J. Pharm. Biomed. Anal 2005, 38, 624.

[10] H. Liu, P. He, H. Li, C. Sun., L. Shi, Y. Liu, G. Zhu, J. Li,Electrochem. Commun. 2005, 7, 1357.

[11] L. Jing-fu , J. A. Jonsson, G.b. Jiang, TRAC 2005, 24, 20.[12] S. Pandey, Anal. Chim. Acta 2006, 556, 38.[13] C. Coll, R. H. Labrador, R. M. Manez, J. Soto, F. Sancenon,

M. J. SeguO, E.Sanchez, Chem. Commun. 2005, 3033.[14] N. V. Shvedene, D. V. Chernyshov, M. G. Khrenova, A. A.

Formanovsky, V. E. Baulin, I. V. Pletnev, Electroanalysis2006, 18, 1416.

[15] R. Maminska, A. Dybko, W.Wroblewski, Sens. Actuators B2006, 115, 552.

[16] N. Maleki, A. Safavi, F. Tajabadi, Anal. Chem. 2006, 78, 3820.[17] A. Safavi, N. Maleki, O. Moradlou, F. Tajabadi, Anal.

Biochem. 2006, 359, 224.[18] A. Safavi, N. Maleki, F. Tajabadi, Analyst 2007, 132, 54.[19] J. Pei, Q. Yin, J. Zhong, Talanta 1991, 38, 1185.[20] M. Shamsipur, J. Tashkhourian, B. Hemmatinejad, H. Shar-

ghi, Talanta 2004, 64, 590.[21] J. G. Huddleston, A. E. Visser, W. M. Reichert, H. D. Willa-

uer, G. A. Broker, R. D. Rogers, Green Chem. 2001, 3, 156.[22] S. Chun, S. V. Dzyuba, R. A. Bartsch, Anal. Chem. 2001, 73,

3737.[23] J. G. Huddleston, H. D. Willauer, R. P. Swatloski, A. E.

Visser, R. D. Rogers, Chem. Commun. 1998, 1765.

586 Afsaneh Safavi et al.



Research Article

Received: 2 May 2011 Revised: 2 July 2011 Accepted: 2 July 2011 Published online in Wiley Online Library

Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

2828

Homogeneous magnesium isotopic composition of seawater: anexcellent geostandard for Mg isotope analysis

Ming-Xing Ling1,2*, Fatemeh Sedaghatpour2, Fang-Zhen Teng2**, Phillip D. Hays2,Josiah Strauss3 and Weidong Sun4

1State Key Laboratory of Isotope Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences,Guangzhou 510640, P.R. China2Isotope Laboratory, Department of Geosciences and Arkansas Center for Space and Planetary Sciences, University ofArkansas, Fayetteville AR 72701, USA3Water Research Centre, School of Civil and Environmental Engineering, University of New South Wales, Sydney 2052,Australia4Key Laboratory of Mineralogy and Metallogeny, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences,Guangzhou 510640, P.R. China

The magnesium (Mg) isotopic compositions of 40 seawater samples from the Gulf of Mexico and of one seawatersample from the southwest Hawaii area were determined by multi-collector inductively coupled plasma mass spec-trometry (MC-ICP-MS) to investigate the homogeneity of Mg isotopes in seawater. The results indicate that the Mgisotopic composition of seawater from the Gulf of Mexico is homogeneous, both vertically and horizontally, withaverage values for d26Mg=�0.832� 0.068 and d25Mg=�0.432� 0.053 (n = 40, 2SD) – identical to those of seawaterfrom Hawaii (average d26Mg=�0.829� 0.037 and d25Mg=�0.427� 0.033) and to the average literature values of sea-water worldwide (d26Mg=�0.83� 0.11 and d25Mg=�0.43� 0.06, n = 49, 2SD). Collectively, global seawater has ahomogeneous Mg isotopic composition with d26Mg=�0.83� 0.09 and d25Mg=�0.43� 0.06 (2SD, n= 90).The magnesium isotopic composition of seawater is principally controlled by river water input, carbonate precip-

itation and oceanic hydrothermal interactions. The homogeneous Mg isotopic composition of seawater indicates asteady-state budget in terms of Mg isotopes in oceans, consistent with a long Mg residence time (~13 Ma). Consid-ering that seawater is homogeneous, readily available in large amounts, can be easily accessed and processed forisotopic analysis, and has an isotopic composition near the middle of the natural range of variation, it is an excel-lent geostandard for accuracy assessment to rule out analytical artifacts during high-precision Mg isotopic analysis.Copyright © 2011 John Wiley & Sons, Ltd.

(wileyonlinelibrary.com) DOI: 10.1002/rcm.5172

Magnesium is a major element in the Earth’s mantle (MgO=37.8 wt.%)[1] and continental crust (MgO=4.66 wt.%),[2] andit is also abundant in the oceans (Mg%=~0.13 wt.%).[3] Mag-nesium has three stable isotopes, 24Mg, 25Mg and 26Mg, forwhich the representative bulk isotopic compositions are78.99%, 10.00% and 11.01% (atom%), respectively.[4] EarlyMg stable isotopic studies were limited because of the lowprecision (~1%) caused by large instrumental mass fraction-ation effects.[5] The advent of multi-collector inductivelycoupled plasma mass spectrometry (MC-ICP-MS) a decadeago has enabled Mg isotope measurements to be made at aprecision of at least one order of magnitude better than with

* Correspondence to: M.-X. Ling, State Key Laboratory of Iso-tope Geochemistry, Guangzhou Institute of Geochemistry,Chinese Academy of Sciences, Guangzhou 510640, P.R.China.E-mail: [email protected]

** Correspondence to: F.-Z. Teng, Isotope Laboratory, Depart-ment of Geosciences and Arkansas Center for Space andPlanetary Sciences, University of Arkansas, FayettevilleAR 72701, USA.E-mail: [email protected]

Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

other techniques.[5,6] For example, a precision of �0.07%(26Mg/24Mg, 2SD) can be routinely achieved for MC-ICP-MS Mg isotopic analysis,[7] which is sufficient to discriminatemass-dependent Mg isotopic variations in natural samples.[6]

Recent high-precision Mg isotopic studies have significantlyincreased our knowledge of Mg isotope geochemistry by docu-menting the Mg isotopic variations in the mantle and crustalrocks,[5,7–18] and illuminating the processes that may producethese variations, e.g. igneous differentiation, metamorphicdehydration, and continental weathering.[7,9,11,14,15,17,19]

Nevertheless, debate continues regarding: (1) whetherthe earth has a chondritic Mg isotopic composition ornot,[7–9,13,18,20] and (2) the accuracy of Mg isotopic composi-tions reported by different labs for certain geostandards, suchas BCR-1, BCR-2 and San Carlos olivine.[5,9,10,12,13,16,18,20–25]

Both debates may reflect the presence of MC-ICP-MS analyti-cal artifacts in high-precision Mg isotopic analyses con-ducted at different laboratories.[7–9,18,20] Analyses ofgeostandards BCR-1 and BCR-2 in different laboratoriesyielded significantly different Mg isotopic compositions(Fig. 1)[5,9,12,13,20–23,26] (e.g. the d26Mg of BCR-1 varied from�0.58 to �0.09 (Fig. 1(A)) and the d26Mg of BCR-2 variedfrom �0.33 to �0.12 (Fig. 1(B)) – value ranges of seven and

Copyright © 2011 John Wiley & Sons, Ltd.

B

BCR-2

-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0

Liu et al., 2010

Young et al., 2009

Chakrabarti & Jacobsen, 2010

D

SC olivine (Same olivine powder)

-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1

Wiechert & Halliday, 2007

Huang et al., 2009

Teng et al., 2007

Handler et al., 2009

Pearson et al., 2006

SC olivine (Different mineral grains)

C

A

BCR-1

-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4

Young & Galy, 2004

Huang et al., 2009

Bourdon et al., 2010

Wiechert & Halliday, 2007

Teng et al., 2007

Chakrabarti & Jacobsen, 2010

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4

Tipper et al., 2008

Bizzarro et al., 2005

Bourdon et al., 2010

Wombacher et al., 2009

Huang et al., 2009

Teng et al., 2007

Baker et al., 2005

26Mg 26Mg

Figure 1. Magnesium isotopic compositions of geostandards: (A) BCR-1, (B) BCR-2,(C) different grains of San Carlos olivine, and (D) different aliquots of homogeneouspowder made by homogenizing a large amount of San Carlos olivine, distributed byS.B. Jacobsen group from Harvard University.

Homogeneous magnesium isotopic composition of seawater

282

three times, respectively, the typical reported precision –although both standards are expected to have a similar,homogeneous composition. Furthermore, the Mg isotopiccomposition of San Carlos olivine was also widely deter-mined but its value varied greatly in different labora-tories.[9,12,16,20,25] The reported d26Mg values of San Carlosolivine grains vary from �0.68 to �0.06 (Fig. 1(C)). Thisinconsistency may reflect analytical artifacts and/or sampleheterogeneity since different labs worked on different SanCarlos olivine grains; however, analyses of different aliquotsof homogeneous powder made by homogenizing a largeamount of San Carlos olivine also yielded different d26Mgvalues, ranging from �0.55 to �0.19 [10,18,23] (Fig. 1(D)), indi-cating that heterogeneity is not the primary problem.To achieve universal comparability for Mg isotopic data

and to enable assessment of accuracy of data from differentlaboratories, a set of well-accepted standards is necessary.Although the pure Mg standard, Cambridge-1,[27] has beenwidely analyzed in different labs, yielding similar d26Mgvalues,[7,13–15,18,23–25,27–29] its use also does not test for samplepreparation accuracy and reproducibility, since the pure Mgstandard needs no sample preparation as is necessary fornatural samples. Reliance on this standard for inter-laboratorycomparison leaves the Mg isotope science community open tomajor problems in accuracy assessment and comparabilitybecause sample preparation processes, e.g. sample dissolutionand column chemistry have the greatest potential for intro-ducing analytical artifacts and therefore are the analyticalstages most important to monitor.Here we test the possibility of using seawater as a potential

geostandard for Mg isotopic analysis. Seawater matrix

Copyright © 2011Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

element concentrations are much more comparable withthose of natural rocks than a pure Mg standard (Table 1). Inaddition, a seawater sample can be easily accessed, is readilyavailable in large amounts and potentially homogeneous asshown by previous case studies (Fig. 2 and Table 2). In orderto investigate the homogeneity of Mg isotopes in seawaterand the suitability of seawater as a geostandard for futureMg isotopic analysis, we chose 40 seawater samples from dif-ferent locations and depths from the Gulf of Mexico, as wellas a seawater sample from Hawaii to run high-precision Mgisotopic analyses. Our results show that the Mg isotopic com-position of seawater is homogeneous and hence it can be usedas a geostandard for assessment of accuracy.

EXPERIMENTALSamples

Forty seawater samples were collected from 15 sampling sta-tions in the northwestern Gulf of Mexico, including the Texasshelf (20 samples from 5 stations collected in August 2007),Louisiana shelf (10 samples from 5 stations collected in July2008) and Louisiana shelf break (10 samples from 5 stationscollected in August 2008) (Fig. 3). The shelf break is theboundary where the continental shelf meets the continentalslope. At least one surface sample and one bottom samplewith depths ranging from 8 m to 112 m were collected fromeach sampling station (Table 3). One seawater sample fromsouthwestern Hawaii was also analyzed for comparison.The seawater samples were pre-treated by filtration toremove particles and phyto- or zooplankton, and then

wileyonlinelibrary.com/journal/rcmJohn Wiley & Sons, Ltd.

9

Table 1. Comparison of Mg contents and the ratios of other major elements to Mg in major igneous rocks and reservoirs ofthe earth

UCC MCC LCC PM Granite Basalt Peridotite Seawater

Mg (%) 1.50 2.17 4.37 22.80 0.45 4.36 26.90 0.13Ca/Mg 1.72 1.73 1.57 0.11 3.10 1.87 0.01 0.32Na/Mg 1.62 1.16 0.45 0.01 6.69 0.38 0.0006 8.35K/Mg 1.55 0.88 0.12 0.001 8.22 0.10 9.26E-05 0.29

UCC: upper continental crust,[2] MCC: middle continental crust,[2] LCC: lower continental crust,[2] PM: primitive mantle.[1]

Granite: USGS granite standard (G-2), Basalt: USGS basalt standard (BHVO-2), Peridotite: Geological Survey of Japan peri-dotite standard (JP-1), data are from the recommended values of these standards. Seawater data are from Brown et al.[3]

M.-X. Ling et al.

2830

preserved at a temperature below 18 �C prior to columnchemistry (the series of chemical procedures related to theseparation of Mg from the sample solutions).

Analytical methods

The magnesium isotopic analyses were carried out at the Iso-tope Laboratory of the University of Arkansas, Fayetteville,AR, USA. Chemical procedures were conducted in a class10 000 clean laboratory, equipped with a class 100 laminar-flow exhaust hood.[15] Optima grade acids or distilled acids,prepared by sub-boiling distillation from trace-metal gradeacids, and Milli-QW water (Millipore, Billerica, MA, USA)with a resistivity of 18.2 MΩ�cm were used throughout allchemical procedures.[9,15]

Column chemistry and MC-ICP-MS analysis

Volumes of 40 mL of solutions for each of the pre-treated sea-water samples were evaporated to dryness on a hotplate at80 �C. The dried residue was dissolved in 500 mL 1 NHNO3. Magnesium was separated by cation-exchange usingcolumns filled with 200–400 mesh AG50W-X8 resin (Bio-Rad, Hercules, CA, USA), which was pre-cleaned with >20times column volume of 6 N HCl, >5 times column volumeof 1 N HNO3 and Milli-QW water. A 100 mL sample solutionwas loaded onto the columns and Mg was eluted with 1 NHNO3.

[9,15] The procedures for eluting Mg were determinedusing both pure Mg standard solutions and different repre-sentative prepared rock-digestion solutions, in order toachieve 100% recovery of Mg and thus avoid potential frac-tionation during cation exchange.[9] The separation procedure

Average = -0.83 0.11 (2SD, n=49, data from literature)

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

26M

g

Figure 2. Magnesium isotopic composition of seawater sam-ples compiled from literature results. Data are from Table 2.Error bars in this figure represent 2SD uncertainties.

wileyonlinelibrary.com/journal/rcm Copyright © 2011 John Wile

was conducted twice successively, to ensure pure Mg solutionrecovery.

The magnesium isotope ratios were determined by aNu Plasma (Wrexham, UK) MC-ICP-MS instrument inlow-resolution mode, in which the 26Mg, 25Mg and 24Mgisotopes were measured simultaneously in separate Faradaycups.[7] The purified Mg solutions were introduced by aquartz cyclonic spray chamber (Elemental Scientific Inc.,Omaha, NE, USA) to the plasma and analyzed using thesample-standard bracketing method.[5] Ratio measurementsfor all sample solutions were repeated four times withina session.[10] The Mg isotopic results are reported instandard d-notation in per mil relative to DSM3:[27]

dXMg ¼ 103 �XMg=24Mgð ÞsampleXMg=24Mgð ÞDSM3

� 1� �

,

where X refers to 25 or 26.

Precision and accuracy

The precision and accuracy of analyses in the University ofArkansas laboratory were evaluated by repeated, full-procedural analyses of samples and standards with dif-ferent matrices, including a synthetic multi-element standardsolution (IL-Mg-1, concentration ratios of Mg:Fe:Al:Ca:Na:K:Ti = 1:1:1:1:1:1:0.1), rocks, minerals, seawater and Cambridge-1pure Mg standard solution, reported in detail by Tenget al.[7,9] and Li et al.[15] The in-run 26Mg/24Mg ratio precisionfor a single measurement in one block of 40 ratios was lessthan �0.02% (2SD). The internal precision on the measured26Mg/24Mg ratio was<�0.1% (2SD), based on ≥4 repeatruns of the same sample solution during a single analyticalsession. The external precision for d26Mg and d25Mg wasapproximately �0.07% and �0.06% (2SD), respectively, asshown by replicate analyses of synthetic solution, mineral,and rock standards over 2 years.[7] For example, multipleanalyses of the synthetic multi-element standard solution(IL-Mg-1) yielded a d25Mg value of �0.01� 0.06 and ad26Mg value of �0.01� 0.07 (2SD, n= 13),[7] which is con-sistent with the expected value of 0. Furthermore, the re-sults for samples and standards such as Cambridge-1 Mgstandard solution, Kilbourne Hole olivine, SUNY MORBand Allende chondrite, etc., following the full proceduresin this lab, are accurate and consistent with publishedvalues.[7,10,15] During the course of this study, a seawatersample from southwestern Hawaii was also analyzedtogether with other seawater samples. The duplicate

y & Sons, Ltd. Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

Table 2. Magnesium isotopic composition of seawater compiled from literature results

No. Sample name or location d26Mg 2SD* d25Mg 2SD References

1 North Atlantic �0.824 0.040 �0.416 0.080 [41]

2 Mediterranean Sea �0.864 0.122 �0.434 0.105 [5]

3 Atlantic �0.799 0.037 �0.407 0.0074 Mediterranean Sea �0.851 0.244 �0.422 0.121

Seawater �0.84 0.13 �0.43 0.15 [28]

Seawater �0.77 0.10 �0.39 0.04 [42]

Seawater �0.82 0.06 �0.42 0.03 [43]

5 Average# �0.81 0.07 �0.41 0.04IAPSO standard �0.75 0.13 �0.39 0.07 [25]

IAPSO standard �0.83 0.05 �0.40 0.046 Average# �0.79 0.11 �0.40 0.017 IAPSO standard �0.79 0.10 �0.41 0.06 [26]

8 NASS-5 standard �0.84 0.16 �0.43 0.07Hawaii SW �0.83 0.07 �0.42 0.05 [8]

Hawaii SW �0.86 0.02 �0.45 0.02 [15]

Hawaii SW �0.83 0.06 �0.43 0.03 [7]

9 Average# �0.84 0.03 �0.43 0.0310 Bermuda �0.79 0.18 �0.41 0.09 [44]

11 IAPSO standard �0.74 0.07 �0.37 0.04 [45]

12 Northwest Pacific �0.79 0.08 �0.37 0.0513 Northwest Pacific �0.85 0.15 �0.47 0.0914 Northwest Pacific �1.00 0.28 �0.51 0.0115 Northwest Pacific �0.97 0.18 �0.49 0.1216 Northwest Pacific �0.69 0.22 �0.38 0.1317 Northwest Pacific �0.77 0.40 �0.39 0.1618 Northwest Pacific �0.86 0.30 �0.34 0.1219 Northwest Pacific �0.73 0.29 �0.40 0.1420 Seawater �0.83 0.09 �0.44 0.08 [46]

21 IAPSO standard �0.89 0.18 [47]

BCR-403 �0.89 0.06 �0.47 0.11 [24]

BCR-403 �0.96 0.03 �0.46 0.03BCR-403 �0.89 0.14 �0.51 0.08BCR-403 �0.82 0.14 �0.42 0.08BCR-403 �0.87 0.14 �0.47 0.08

22 Average# �0.89 0.10 �0.47 0.0623 IAPSO standard �0.80 0.05 �0.42 0.02 [48]

24 W. Dutch Wadden Sea �0.79 0.03 �0.42 0.0225 IAPSO standard �0.79 0.26 �0.40 0.15 [23]

26 Bermuda �0.82 0.29 �0.41 0.1827 North of Newfoundland �0.84 0.16 �0.44 0.09 [49]

28 North of Newfoundland �0.82 0.06 �0.42 0.0329 North Atlantic �0.87 0.09 �0.44 0.0330 North Atlantic �0.80 0.09 �0.42 0.0931 North Atlantic �0.87 0.03 �0.45 0.0632 North Atlantic, BATS �0.83 0.09 �0.43 0.0633 North Atlantic, BATS �0.86 0.06 �0.44 0.0634 North Atlantic, BATS �0.86 0.06 �0.44 0.0635 North Atlantic, BATS �0.81 0.09 �0.41 0.0636 North Atlantic, BATS �0.87 0.09 �0.46 0.0637 North Atlantic, BATS �0.86 0.09 �0.44 0.0638 Southern Ocean �0.86 0.06 �0.45 0.0639 North Pacific, St. PAPA �0.84 0.09 �0.44 0.0640 North Pacific, St. PAPA �0.79 0.03 �0.40 0.0041 North Pacific, St. PAPA �0.84 0.06 �0.43 0.0642 North Pacific, St. P26 �0.79 0.09 �0.40 0.0643 North Pacific, St. P26 �0.80 0.09 �0.41 0.0644 North Pacific, St. P26 �0.80 0.03 �0.40 0.0645 North Pacific, St. P26 �0.77 0.09 �0.39 0.0346 North Pacific, St. P26 �0.80 0.13 �0.42 0.0347 North Pacific, St. P26 �0.87 0.09 �0.46 0.06

(Continues)


wileyonlinelibrary.com/journal/rcmCopyright © 2011 John Wiley & Sons, Ltd.Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

2831

Table 2. (Continued)

No. Sample name or location d26Mg 2SD* d25Mg 2SD References

48 North Pacific, St. P26 �0.82 0.06 �0.44 0.0049 North Pacific, St. P26 �0.80 0.06 �0.41 0.06

Total average �0.83 0.11 �0.43 0.06

*2SD of total average is calculated using all the Mg isotopic data in this table. 2SD values of some data are calculated from2SE in literature results.#Average of results from the same sample is used to plot in Figs. 2 and 6.

M.-X. Ling et al.

2832

(repeated measurement of different aliquot of the same Mg-cut) and replicate (repeated column chemistry and measure-ment) analyses yielded an average d26Mg=�0.829� 0.037and d25Mg=�0.427� 0.033 (2SD, n= 13) (Table 3), which isconsistent with our previously published data (Table 2).[7,8,15]

RESULTS

The Mg isotopic composition of seawater from the Gulf ofMexico ranges from�0.907 to�0.765 for d26Mg with an averageof �0.832� 0.068 (2SD, n = 40) (Fig. 4 and Table 3), andfrom �0.501 to �0.361 for d25Mg with an average of�0.432� 0.053 (2SD, n = 40) (Table 3). Seawater from south-western Hawaii has d26Mg ranging from �0.872 to �0.813with an average of �0.829� 0.037 (2SD, n= 13), and d25Mgfrom �0.449 to �0.394 with an average of �0.427� 0.033(2SD, n= 13) (Table 3), identical to previously reported valuesof the same seawater sample in this laboratory[7,8,15] (Table 2).In addition, the Mg isotopic composition of seawater from theGulf of Mexico shows no variation as a function of depth, lati-tude, longitude, locations of continental shelf or shelf break(Fig. 5 and Table 3). Overall, the Mg isotopic composition ofseawater analyzed in this study is very homogeneous, and

Houston

Austin Texas

Gulf of

3

412

406&

BR3DBR5A

BR4DBR5C

BR6D

Continentalshelf

Shelf slope

Figure 3. Sketch map of the northwest G(modified from Google Earth). The red diThe locations in Texas shelf and Louisianwhereas those in the Louisiana shelf breaname unavailable) corresponding to Tab


identical to the average value of those compiled from pre-viously published results (d26Mg=�0.83� 0.11, d25Mg=�0.43� 0.06, 2SD, n= 49) (Table 2).

DISCUSSIONS AND CONCLUSION

The residence time for dissolved species in oceans is definedas the average time that a substance remains before removalthrough precipitation or adsorption processes and subse-quent deposition,[30] and it is also called the replacementtime.[31] The residence time is calculated as total componentmass in seawater divided by input or output rate.[30] The resi-dence times of different major seawater constituents vary byover six order of magnitudes.[31] Among the abundant ele-ments in seawater, Mg has a residence time of ~13 Ma, mod-erately close to that of Ca (10 Ma).[31] The mixing time for theearth’ oceans is estimated at 1 kya to 2 kya.[32] The short mix-ing time, long residence time, and balance of input and removalfunctions can result in a homogeneous distribution of Mg inthe oceans with a global average content of ~0.13%.[3] Ourwork, together with previous studies, indicates a homoge-neous distribution of Mg isotopes in the oceans.

New Orleans

Louisiana

Mississippi

Mexico

1D

6C BC12

3B9B

400&401

417&418404&405

&413

407

100km

Baton Rouge

ulf of Mexico and sampling locationsamonds represent sampling locations.a shelf are labeled with station namesk are marked as sample names (stationle 3.


Table 3. Sampling details and Mg isotopic compositions of seawater in this study

Locations Station name* Sample Latitude Longitude Depth(m)

d26Mg 2SD d25Mg 2SD

LA-Shelfx 3B 33 28.5816 �90.7780 0 �0.809 0.066 �0.431 0.073254 28.5816 �90.7780 20 �0.819 0.071 �0.427 0.054

9B 132 28.7214 �90.5516 16 �0.831 0.071 �0.421 0.0547 28.7214 �90.5516 0 �0.841 0.071 �0.427 0.054

6C 125 28.9234 �92.2287 0 �0.785 0.071 �0.424 0.05423 28.9234 �92.2287 25 �0.876 0.071 �0.465 0.054

BC12 150 28.9946 �92.0045 0 �0.840 0.071 �0.454 0.054143 28.9946 �92.0045 20 �0.777 0.061 �0.416 0.063

31D 53 29.3994 �92.5977 0 �0.794 0.061 �0.397 0.063102 29.3994 �92.5977 8 �0.833 0.071 �0.452 0.054

LA-ShelfBreak

406 27.8078 �93.0541 0 �0.832 0.067 �0.450 0.060407 27.8078 �93.0541 69 �0.824 0.071 �0.415 0.054404 27.9544 �92.0146 0 �0.873 0.067 �0.416 0.060405 27.9544 �92.0146 60 �0.852 0.067 �0.438 0.060412 27.8527 �92.9203 0 �0.795 0.061 �0.409 0.063

Duplicate† �0.840 0.084 �0.397 0.063Average �0.818 0.084 �0.403 0.063

413 27.8527 �92.9203 100 �0.803 0.061 �0.405 0.063400 28.3174 �90.5659 0 �0.793 0.061 �0.406 0.063401 28.3174 �90.5659 60 �0.785 0.061 �0.398 0.063417 28.0909 �91.1491 0 �0.841 0.061 �0.401 0.063418 28.0909 �91.1491 112 �0.803 0.061 �0.426 0.063

TX-Shelfx BR4D BR4D-BOT1 28.5588 �95.5835 20 �0.843 0.067 �0.381 0.060Duplicate �0.835 0.068 �0.446 0.050Average �0.839 0.068 �0.413 0.060BR4D-BOT2 28.5588 �95.5835 20 �0.835 0.075 �0.470 0.064BR4D-SFC1 28.5588 �95.5835 0 �0.860 0.080 �0.435 0.069BR4D-SFC2 28.5588 �95.5835 0 �0.839 0.080 �0.473 0.069

BR3D BR3D-BOT1 28.6244 �95.4323 18 �0.876 0.075 �0.478 0.064BR3D-BOT2 28.6244 �95.4323 18 �0.805 0.072 �0.421 0.053BR3D-SFC1 28.6244 �95.4323 0 �0.844 0.072 �0.456 0.053BR3D-SFC2 28.6244 �95.4323 0 �0.872 0.075 �0.484 0.064

BR6D BR6D-BOT1 28.4251 �95.8739 20 �0.861 0.078 �0.456 0.038BR6D-BOT2 28.4251 �95.8739 20 �0.870 0.072 �0.438 0.053BR6D-SFC1 28.4251 �95.8739 0 �0.880 0.075 �0.458 0.064BR6D-SFC2 28.4251 �95.8739 0 �0.774 0.072 �0.398 0.053

BR5A BR5A-BOT1 28.6419 �95.8168 9 �0.853 0.080 �0.430 0.069BR5A-BOT2 28.6419 �95.8168 9 �0.843 0.067 �0.396 0.060BR5A-SFC1 28.6419 �95.8168 0 �0.781 0.072 �0.361 0.053Duplicate �0.832 0.084 �0.501 0.063Duplicate �0.856 0.075 �0.452 0.064Average �0.823 0.084 �0.438 0.064BR5A-SFC2 28.6419 �95.8168 0 �0.809 0.072 �0.390 0.053

BR5C BR5C-BOT1 28.5520 �95.7459 16 �0.765 0.084 �0.409 0.063Duplicate �0.780 0.080 �0.394 0.069Average �0.772 0.084 �0.402 0.069BR5C-BOT2 28.5520 �95.7459 16 �0.888 0.080 �0.458 0.069BR5C-SFC1 28.5520 �95.7459 0 �0.807 0.080 �0.457 0.069BR5C-SFC2 28.5520 �95.7459 0 �0.907 0.058 �0.474 0.060Total average �0.832 0.068 �0.432 0.053

Hawaii Seawater �0.815 0.078 �0.416 0.038Duplicate �0.847 0.062 �0.448 0.055Duplicate �0.813 0.072 �0.417 0.053Duplicate �0.813 0.067 �0.417 0.060Duplicate �0.813 0.071 �0.417 0.054Duplicate �0.813 0.084 �0.417 0.063Duplicate �0.872 0.075 �0.440 0.064Replicate† �0.836 0.068 �0.426 0.050Replicate �0.824 0.068 �0.445 0.050Replicate �0.847 0.058 �0.443 0.060

(Continues)


wileyonlinelibrary.com/journal/rcmCopyright © 2011 John Wiley & Sons, Ltd.Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

2833

Dep

th (

m)

Lat

itu

de

27.5

28.0

28.5

29.0

29.5

120

100

80

60

40

20

0

B

A

Table 3. (Continued)

Locations Station name* Sample Latitude Longitude Depth(m)

d26Mg 2SD d25Mg 2SD

Hawaii Replicate �0.813 0.080 �0.417 0.069Replicate �0.834 0.067 �0.394 0.060Replicate �0.838 0.066 �0.449 0.073Average �0.829 0.037 �0.427 0.033

xLA-Shelf = Louisiana shelf. TX-Shelf = Texas shelf.†Duplicate = repeated measurement of different aliquot of the same Mg-cut. Replicate = repeated column chemistry andmeasurement of the same sample.*The station names of Louisiana Shelf Break are not available.

M.-X. Ling et al.

2834

Controlling factors onMg content and isotopic compositionof seawater

The magnesium content of seawater is controlled by riverwater, groundwater, carbonate precipitation,[31] oceanichydrothermal interactions,[33,34] ion-exchange reactions withclays[33] and dust of continental origin; of these contributoryfactors river water input is the main source of Mg in sea-water.[28,34,35] Magnesium in rivers is mainly derived fromthe weathering of continental rocks and is controlled bywatershed lithology.[36] Under the steady-state seawater Mgconditions thought to have predominated through much ofgeological time,[35] river input was balanced by the importantMg removal – carbonate precipitation[31] and hydrothermalinteractions, e.g. circulation at the mid-ocean ridge (MOR)spreading centers,[34], and references cited therein where Mg isexchanged with Ca,[33] to remove Mg from seawater. Atpresent MOR hydrothermal alteration is thought to be thepredominant Mg removal process; however, syndepositionaldolomite precipitation is thought to have been the predom-inant control on seawater Mg concentrations and on Mg/Caratios over a large extent of the Phanerozoic Eon, out-weighing removal of Mg at the MOR and providing thebalance for the input of riverine Mg.[35]

Consistent with the predominant controls on seawater Mgconcentrations, the Mg isotopic composition of seawaterlargely depends on river-water input, carbonate precipitationand oceanic hydrothermal interactions. The magnesium iso-topic compositions of river waters are highly heteroge-neous.[19,37,38] The flux-weighted d26Mg of global runoff has

-1.00

-0.95

-0.90

-0.85

-0.80

-0.75

-0.70

-0.65

Average = -0.832 0.068 (2SD, n=40, this study)

26M

g

Figure 4. Magnesium isotopic composition of seawatersamples from the Gulf of Mexico.


been estimated at �1.09%, based on sampling of 45 riversacross the globe,[37] which is lower than the average d26Mgof seawater at �0.83%. Carbonate precipitation and clay for-mation may slightly modify the Mg isotopic composition of

Lo

ng

itu

de

-97

-96

-95

-94

-93

-92

-91

-1.00 -0.95 -0.90 -0.85 -0.80 -0.75 -0.70 -0.65

C

26Mg

Figure 5. Lack of correlations between Mg isotopiccomposition and sampling depth, latitude and longitude ofseawater in the Gulf of Mexico: (A) d26Mg vs. depth,(B) d26Mg vs. latitude, and (C) d26Mg vs. longitude.


Average of global seawater = -0.83 0.09 (2SD, n=90)

Average of the mantle = -0.25 0.07

Average of UCC= -0.22 (-0.52 ~ +0.92)

Up to +0.92

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

This study Literature

26M

g

Figure 6. Magnesium isotopic composition of global sea-water samples, including data compiled from literatureresults and in this study. The symbols are the same as inFigs. 2 and 4. UCC: upper continental crust. Data of UCCand mantle are from Li et al.[15] and Teng et al.[7] respectively.


283

seawater at time scales matching riverine input.[5] Further-more, although Mg in calcite has a consistently lower d26Mgvalue than Mg in dolomite by approximately 2%,[5] consider-ing that the Mg content of calcite is significantly lower thanthat of dolomite, dolomite precipitation may be an importantcontrol on seawater Mg isotopic composition. The extent towhich oceanic hydrothermal interactions affect Mg isotopiccomposition of seawater remains poorly understood. Tipperet al.[37] proposed that hydrothermal circulation must prefer-entially scavenge light Mg isotopes from seawater. The factthat the seawater all over the world has a homogeneous Mgisotopic composition implies a steady-state oceanic Mg iso-tope budget.

Seawater as a geostandard for Mg isotopic analysis

All the seawater samples from the Gulf of Mexico and Hawaiianalyzed in this study have homogeneous Mg isotopic com-position, identical to the averages based on literature data(Fig. 6). Furthermore, seawater samples from the Gulf ofMexicowere sampled from awide range of locations (latitude:27.8078 to 29.3994, longitude: –95.8739 to �90.5516), withdepths ranging from 8 m to 112 m. The results indicate thatthe Mg isotopic compositions of seawater are homogeneousboth vertically and horizontally (Fig. 5).Based on the results in this study and literature data (Fig. 6),

global seawater samples have homogeneous Mg isotopiccompositions with d26Mg=�0.83� 0.09 and d25Mg=�0.430.06 (2SD, n = 90). In addition, the ratios of seawater matrixelements to Mg are similar to those of natural samples(Table 1). This, together with the fact that seawater is readilyaccessible in large amounts, can be easily processed for isoto-pic analysis, and has an isotopic composition near the middleof natural range of variation, makes it an excellent geostan-dard for future inter-laboratory accuracy assessment onhigh-precision Mg isotopic analysis.Furthermore, the homogeneous Mg isotopic composition of

seawater may also contribute to further research of seawaterpaleo-temperature, since there are possibly small changes inseawater Mg isotopic composition corresponding to Mg/Cavariation of echinoderms,[34] while the Mg content or Mg/Ca

Copyright © 2011Rapid Commun. Mass Spectrom. 2011, 25, 2828–2836

ratio is largely controlled by temperature.[39,40] Thus, the evo-lution of seawater Mg isotopic composition over time maybe an efficient indicator of seawater temperature in the past.

AcknowledgementsWe thank Yan Xiao for help in the lab and Bing Shen and oneanonymous referee for constructive comments. This workwas supported by NSF (EAR-0838227 and EAR-1056713),Arkansas Space Grant Consortium (SW19002), NaturalScience Foundation of China (NSFC) (Nos. 41103006 and41090370) and the CAS/SAFEA International PartnershipProgram for Creative Research Teams. This is contributionNo. IS-1374 from GIGCAS.

REFERENCES

[1] W. F. McDonough, S. S. Sun. The composition of the Earth.Chem. Geol. 1995, 120, 223.

[2] R. L. Rudnick, S. Gao, in The Crust, (Ed: R. L. Rudnick),Elsevier-Pergamon, Oxford, 2003, pp. 1–64.

[3] J. Brown, A. Colling, D. Park, J. Phillips, D. Rothery,J. Wright, in Seawater: Its Composition, Properties and Beha-viour (1st edn.), Pergamon, Oxford, 1989.

[4] K. J. R. Rosman, P. D. P. Taylor. Isotopic compositions of theelements 1997. Pure Appl. Chem. 1998, 70, 217.

[5] E. D. Young, A. Galy, in Geochemistry of Non-traditional StableIsotopes, (Eds: C. M. Johnson, B. L. Beard, F. Albarede),Mineralogical Society of America, Washington DC, 2004,pp. 197–230.

[6] A. Galy, M. Bar-Matthews, L. Halicz, R. K. O’Nions. Mg iso-topic composition of carbonate: insight from speleothemformation. Earth Planet. Sci. Lett. 2002, 201, 105.

[7] F.-Z. Teng, W.-Y. Li, S. Ke, B. Marty, N. Dauphas, S. Huang,F.-Y. Wu, A. Pourmand. Magnesium isotopic composition ofthe Earth and chondrites. Geochim. Cosmochim. Acta 2010, 74,4150.

[8] W. Yang, F.-Z. Teng, H.-F. Zhang. Chondritic magnesiumisotopic composition of the terrestrial mantle: A case studyof peridotite xenoliths from the North China craton. EarthPlanet. Sci. Lett. 2009, 288, 475.

[9] F.-Z. Teng, M. Wadhwa, R. T. Helz. Investigation of magne-sium isotope fractionation during basalt differentiation:Implications for a chondritic composition of the terrestrialmantle. Earth Planet. Sci. Lett. 2007, 261, 84.

[10] S.-A. Liu, F.-Z. Teng, Y. He, S. Ke, S. Li. Investigation of mag-nesium isotope fractionation during granite differentiation:Implication for Mg isotopic composition of the continentalcrust. Earth Planet. Sci. Lett. 2010, 297, 646.

[11] W.-Y. Li, F.-Z. Teng, Y. Xiao, J. Huang. High-temperatureinter-mineral magnesium isotope fractionation in eclogitefrom the Dabie orogen, China. Earth Planet. Sci. Lett. 2011,304, 224.

[12] F. Huang, J. J. Glessner, A. Ianno, C. C. Lundstrom,Z. Zhang. Magnesium isotopic composition of igneous rockstandards measured by MC-ICP-MS. Chem. Geol. 2009,268, 15.

[13] B. Bourdon, E. T. Tipper, C. Fitoussi, A. Stracke. ChondriticMg isotope composition of the Earth. Geochim. Cosmochim.Acta 2010, 74, 5069.

[14] F.-Z. Teng, W. Li, R. L. Rudnick, L. R. Gardner. Contrastingbehavior of lithium and magnesium isotope fractionationduring continental weathering. Earth Planet. Sci. Lett. 2010,300, 63.

wileyonlinelibrary.com/journal/rcmJohn Wiley & Sons, Ltd.

5

M.-X. Ling et al.

2836

[15] W.-Y. Li, F.-Z. Teng, S. Ke, R. L. Rudnick, S. Gao, F.-Y. Wu,B. W. Chappell. Heterogeneous magnesium isotopic compo-sition of the upper continental crust. Geochim. Cosmochim.Acta 2010, 74, 6867.

[16] M. R. Handler, J. A. Baker, M. Schiller, V. C. Bennett, G. M.Yaxley. Magnesium stable isotope composition of Earth’supper mantle. Earth Planet. Sci. Lett. 2009, 282, 306.

[17] B. Shen, B. Jacobsen, C.-T. A. Lee, Q. Z. Yin, D. M. Morton.The Mg isotopic systematics of granitoids in continental arcsand implications for the role of chemical weathering in crustformation. Proc. Natl. Acad. Sci. USA 2009, 106, 20652.

[18] E. D. Young, E. Tonui, C. E. Manning, E. A. Schauble,C. Macris. Spinel-olivine magnesium isotope thermometryin the mantle and implications for the Mg isotopic composi-tion of Earth. Earth Planet. Sci. Lett. 2009, 288, 524.

[19] E. T. Tipper, M. J. Bickle, A. Galy, A. J. West, C. Pomies,H. J. Chapman. The short term climatic sensitivity of car-bonate and silicate weathering fluxes: Insight from seasonalvariations in river chemistry. Geochim. Cosmochim. Acta 2006,70, 2737.

[20] U. Wiechert, A. N. Halliday. Non-chondritic magnesiumand the origins of the inner terrestrial planets. Earth Planet.Sci. Lett. 2007, 256, 360.

[21] J. A. Baker, M. Bizzarro, N. Wittig, J. Connelly, H. Haack.Early planetesimal melting from an age of 4.5662 Gyr fordifferentiated meteorites. Nature 2005, 436, 1127.

[22] M. Bizzarro, J. A. Baker, H. Haack, K. L. Lundgaard. Rapidtimescales for accretion and melting of differentiated plane-tesimals inferred from 26Al-26Mg chronometry. Astrophys. J.2005, 632, L41.

[23] R. Chakrabarti, S. B. Jacobsen. The isotopic composition ofmagnesium in the Solar System. Earth Planet. Sci. Lett.2010, 293, 349.

[24] E. B. Bolou-Bi, N. Vigier, A. Brenot, A. Poszwa. Magnesiumisotope compositions of natural reference materials. Geo-stand. Geoanal. Res. 2009, 33, 95.

[25] N. J. Pearson, W. L. Griffin, O. Alard, S. Y. O’Reilly. The iso-topic composition of magnesium in mantle olivine: Recordsof depletion and metasomatism. Chem. Geol. 2006, 226, 115.

[26] F. Wombacher, A. Eisenhauer, A. Heuser, S. Weyer. Separa-tion of Mg, Ca and Fe from geological reference materialsfor stable isotope ratio analyses by MC-ICP-MS anddouble-spike TIMS. J. Anal. At. Spectrom. 2009, 24, 627.

[27] A. Galy, O. Yoffe, P. E. Janney, R. W. Williams, C. Cloquet,O. Alard, L. Halicz, M. Wadhwa, I. D. Hutcheon, E. Ramon,J. Carignan. Magnesium isotope heterogeneity of the isoto-pic standard SRM980 and new reference materials for mag-nesium-isotope-ratio measurements. J. Anal. At. Spectrom.2003, 18, 1352.

[28] E. T. Tipper, A. Galy, M. J. Bickle. Riverine evidence for afractionated reservoir of Ca and Mg on the continents:Implications for the oceanic Ca cycle. Earth Planet. Sci. Lett.2006, 247, 267.

[29] J. R. Black, E. Epstein, W. D. Rains, Q.-Z. Yin, W. H. Casey.Magnesium-isotope fractionation during plant growth.Environ. Sci. Technol. 2008, 42, 7831.

[30] J. M. Frank, Chemical Oceanography, (3rd edn.), CRC Press,Boca Raton, 2006.

[31] E. K. Berner, R. A. Berner, in GlobalEnvironment:Water,Air,andGeochemical Cycles, Prentice Hall, Upper Saddle River, 1996.


[32] T. Garrison, in Oceanography: An Invitation to Marine Science(6th edn.), Brook and Cole Publishers, Belmont, CA, 2006.

[33] H. Elderfield, A. Schultz. Mid-ocean ridge hydrothermalfluxes and the chemical composition of the ocean. Annu.Rev. Earth Planet. Sci. 1996, 24, 191.

[34] S. de Villiers, J. A. D. Dickson, R. M. Ellam. The compositionof the continental river weathering flux deduced from sea-water Mg isotopes. Chem. Geol. 2005, 216, 133.

[35] B. H. Wilkinson, T. J. Algeo. Sedimentary carbonate recordof calcium-magnesium cycling. Am. J. Sci. 1989, 289, 1158.

[36] M. Meybeck. Global chemical-weathering of surficial rocksestimated from river dissolved loads. Am. J. Sci. 1987,287, 401.

[37] E. T. Tipper, A. Galy, J. Gaillardet, M. J. Bickle, H. Elderfield,E. A. Carder. The magnesium isotope budget of the modernocean: Constraints from riverine magnesium isotope ratios.Earth Planet. Sci. Lett. 2006, 250, 241.

[38] E. T. Tipper, A. Galy, M. J. Bickle. Calcium and magnesiumisotope systematics in rivers draining the Himalaya-Tibetan-Plateau region: Lithological or fractionation control?Geochim. Cosmochim. Acta 2008, 72, 1057.

[39] J. N. Weber. Temperature-dependence of magnesium inechinoid and asteroid skeletal calcite - reinterpretation ofits significance. J. Geol. 1973, 81, 543.

[40] S. L. Healey, R. C. Thunell, B. H. Corliss. The Mg/Ca-temperature relationship of benthic foraminiferal calcite:New core-top calibrations in the <4 �C temperature range.Earth Planet. Sci. Lett. 2008, 272, 523.

[41] V. T.-C. Chang, A. Makishima, N. S. Belshaw, R. K. O’Nions.Purification of Mg from low-Mg biogenic carbonates for iso-tope ratio determination using multiple collector ICP-MS. J.Anal. At. Spectrom. 2003, 18, 296.

[42] E. T. Tipper, P. Louvat, F. Capmas, A. Galy, J. Gaillardet.Accuracy of stable Mg and Ca isotope data obtained byMC-ICP-MSusing the standard addition method. Chem.Geol. 2008, 257, 65.

[43] E. T. Tipper, J. Gaillardet, P. Louvat, F. Capmas, A. F. White.Mg isotope constraints on soil pore-fluid chemistry: Evi-dence from Santa Cruz, California. Geochim. Cosmochim.Acta 2010, 74, 3883.

[44] J. A. Higgins, D. P. Schrag. Constraining magnesium cyclingin marine sediments using magnesium isotopes. Geochim.Cosmochim. Acta 2010, 74, 5039.

[45] K. Ra, H. Kitagawa. Magnesium isotope analysis of differ-ent chlorophyll forms in marine phytoplankton usingmulti-collector ICP-MS. J. Anal. At. Spectrom. 2007, 22, 817.

[46] P. A. E. Pogge von Strandmann. Precise magnesium isotopemeasurements in core top planktic and benthic foraminif-era. Geochem. Geophys. Geosyst. 2008, 9, Q12015.

[47] P. A. E. Pogge von Strandmann, R. H. James, P. VanCalstern, S. R. Gislason, K. W. Burton. Lithium, magnesiumand uranium isotope behaviour in the estuarine environ-ment of basaltic islands. Earth Planet. Sci. Lett. 2008, 274, 462.

[48] D. Hippler, D. Buhl, R. Witbaard, D. K. Richter,A. Immenhauser. Towards a better understanding ofmagnesium-isotope ratios from marine skeletal carbo-nates. Geochim. Cosmochim. Acta 2009, 73, 6134.

[49] G. L. Foster, P. A. E. Pogge von Strandmann, J. W. B. Rae.Boron and magnesium isotopic composition of seawater.Geochem. Geophys. Geosyst. 2010, 11, Q08015.


magnesium isotopic composition of the moon

Documents