121

Geochemical Journal, Vol. 40, pp. 121 to 133, 2006

*Corresponding author (e-mail: tanimizum@jamstec.go.jp)*Present address: Kochi Institute for Core Sample Research, JapanAgency for Marine-Earth Science and Technology, Monobe-B 200,

Nankoku, Kochi 783-8502, Japan.

Copyright © 2006 by The Geochemical Society of Japan.

Development of rapid and precise Pb isotope analytical techniques usingMC-ICP-MS and new results for GSJ rock reference samples

MASAHARU TANIMIZU1* and TSUYOSHI ISHIKAWA2

1Institute for Research on Earth Evolution (IFREE), Japan Agency for Marine-Earth Science and Technology,2-15 Natsushima, Yokosuka 237-0061, Japan

2Center for Advanced Marine Core Research, Kochi University, Monobe-B 200, Nankoku, Kochi 783-8502, Japan

(Received April 12, 2005; Accepted August 8, 2005)

A rapid and precise routine Pb isotope analytical technique for rock samples was developed using multiple collectorinductively coupled plasma mass spectrometry (MC-ICP-MS). The mass discrimination of the Pb isotope ratios was cor-rected with the 205Tl/203Tl ratio of a Tl isotope reference material (NIST SRM 997) added to samples. The optimization ofthe analytical parameters and the minimization of the memory and matrix effects enabled to acquire the Pb isotope datarapidly (three samples per hour) with a reproducibility of 0.01% by consumption of 150 ng Pb per analysis. Repeatedanalyses of a Pb isotope reference material, NIST SRM 981, gave the average values of 206Pb/204Pb = 16.9308 ± 0.0010,207Pb/204Pb = 15.4839 ± 0.0011, and 208Pb/204b = 36.6743 ± 0.0030 (2σ).

Pb isotope ratios of eleven Geological Survey of Japan rock reference samples were determined using this technique.In anion-exchange chemical separation, complete removal of Hg and Tl from the Pb fraction was confirmed. It was re-vealed that the Tl-normalization minimized the matrix effects caused by remaining impurities and medium, and the repro-ducibility of seven independent analyses of JB-2 was comparable to that of NIST SRM 981. Pb isotope ratios obtained forthe rock reference samples showed a good agreement with those given previously.

Keywords: Pb isotope ratios, GSJ reference rocks, multiple collector ICP-MS, Tl normalization, matrix and memory effects

be resolved by using stable or radioactive double-spikemethod (204Pb-207Pb or 202Pb-205Pb; Woodhead et al.,1995; Todt et al., 1996) or by calculating “zero time in-tercept” of the time-dependent Pb isotope fractionationduring the TIMS measurement (Tuttas and Habfast, 1982),but both methods are not simple and therefore not suit-able for a rapid routine Pb isotope analysis. In recentyears, inductively coupled plasma mass spectrometry(ICP-MS) with multiple collectors (MC) has been intro-duced for the precise isotope analysis. The mass discrimi-nation in MC-ICP-MS is essentially independent of timebut depends on mass number, and the correction of themass discrimination is possible for a given element usingisotopes of other elements in the same mass range. Thisnormalization has been applied to isotope analyses ofmany elements to detect mass-dependent isotopefractionation occurring in natural processes (Johnson etal., 2004). Precise Pb isotope analyses have been alsoperformed with MC-ICP-MS through a normalizationwith 205Tl/203Tl isotope ratios by several authors (Hirata,1996; Belshaw et al., 1998; Rehkämper and Halliday,1998) to avoid the mass-dependent fractionation of TIMS,which is essentially uncorrectable in TIMS without thedouble-spiking. However, subsequent two investigationsof Pb double-spike MC-ICP-MS, which tested the Tl-nor-

INTRODUCTION

Thermal ionization mass spectrometry (TIMS) hasbeen successfully employed for the acquisition of highprecision isotope data for elements with relatively lowfirst ionization potentials, such as Sr and Nd. TIMS isalso applied to the analysis of Pb isotope ratios (206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb), which are essentialfor various geochronological and geochemical studies. Adevelopment of sample loading technique using silica-gel (Cameron et al., 1969) drastically improved the ioni-zation efficiency of Pb in TIMS, which allowed the Pbisotope analysis with significantly smaller sample sizecompared to the electron impact ionization method withgaseous methyl Pb (e.g., Ostic et al., 1967). Pb has onlyone non-radiogenic stable isotope 204Pb, however, and thecorrection of the time-dependent mass fractionation forTIMS usually applied for Sr and Nd isotope analyses us-ing a stable isotope ratio is unavailable. This problem can

122 M. Tanimizu and T. Ishikawa

malization, showed that Tl-normalization is sometimesinaccurate due to inadequate correction of the mass dis-crimination (Thirlwall, 2002; Baker et al., 2004).

In this report, we re-examine the precision and repro-ducibility of Pb isotope ratios using MC-ICP-MS withTl-normalization for rock samples. The ability of MC-ICP-MS and analytical parameter dependence of the pre-cision and reproducibility are discussed. We also reportPb isotope ratios of eleven rock reference samples fromGeological Survey of Japan (GSJ), and compare theirvalues to the previously reported results given by TIMSand MC-ICP-MS.

EXPERIMENTAL PROCEDURES

Mass spectrometryThe instrument used in this study is an MC-ICP-MS

NEPTUNE (Thermo Electron, Germany) in Center forAdvanced Marine Core Research, Kochi University, Ja-pan. This Nier-Johnson-type double-focusing massspectrometer achieves an effective mass dispersion of 17%and produces flat-topped, symmetrical peak shapes. Thisinstrument equips eight movable Faraday collectors andone fixed central channel where an ion beam can beswitched between a Faraday detector and a SEM detec-tor. Further technical information is found in recent pub-lications (Weyer and Schwieters, 2003; Tanimizu, inpress). The Faraday collectors were set to detect the fol-lowing isotopes simultaneously: 202Hg (Low3), 203Tl(Low2), 204Pb (Low1), 205Tl (Center), 206Pb (High1), 207Pb(High2), and 208Pb (High3). The ion current detected by

each collector was converted to voltage through a 1011 Ωresister. The relative differences in the resistors in the nineamplifiers were calibrated against the Center Faraday bypassing a constant electric current. Typical operating con-ditions of the instrument are detailed in Table 1. Onemeasurement run consists of 3 blocks, each of which con-tains 40 cycles (4 sec integration per cycle). Before themeasurement of each block, baseline data were collectedfor 60 sec at m/z = 204.5 (Center Faraday) with the ionbeam being automatically defocused by the software. Eachsingle run takes 15 minutes. Sample solutions were in-troduced with a 50 µL/min flow PFA nebulizer (Elemen-tal Scientific, Inc., USA) in free aspiration mode attachedto dual cyclonic/double Scott glass chamber. Desolvatingnebulizers are useful to enhance sensitivity, but longerwashout time is necessary due to the large memory ef-fect. We did not use the introduction system to enhancesample throughput. On this condition, we routinelyachieved signal intensities larger than 7 × 10–10 A for 1µg/mL of Pb or Tl.

The mass-dependent discrimination in MC-ICP-MSis now understood as an ion repulsion after the skimmingprocess of ions (Gillson et al., 1988), and a simple nor-malization law, “exponential law”, is usually applied forthe correction. In this study, the observed 206Pb/204Pb,207Pb/204Pb, and 208Pb/204Pb ratios were normalized us-ing the 205Tl/203Tl ratio of NIST SRM 997, a Tl isotopereference material, added to the sample prior to the analy-sis. The relationship between the observed and normal-ized Pb and Tl isotope ratios (Robs and Rnorm) is describedby the expression under the exponential law:

Table 1. Typical ICP-MS operating condition. The central argon gas flow rate was adjusted tomaximize the transmission of Pb signal.

ICP Ion Sourcerf frequencies 27.12 MHzrf power 1.2 kW forward, <10 W reflectionFassel-type torch Ar gas flow rates

outer 15 L/minintermediate 0.70 L/mincentral 1.15 L/min

nebulizer PFA Teflon nebulizer (Elemental Scientific, Inc.)spray chamber dual sychronic/double Scott glass chamber (ambient temperature)sample uptake 50 µL/min (free aspiration)

Mass spectrometerion energy 10000 Vextraction 2000 Vtypical transmission

Pb 70 × 10–11 A/ppm at low resolution

Pressures during operationESA 1.0–1.5 × 10–5 Pa (9 × 10–6 Pa during standby)

Analyzer 4-9 × 10–7 Pa (4 × 10–7 Pa during srandby)

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 123

R RA

Anorm obs

f

205 203 205 203205

203

1/ / ,=

( )Tl

R RA

Annorm

nobs n

f

/ / ,204 204204

2=

( )Pb

f fTl Pb= ( ), 3

where f and A are mass discrimination correction factorsand atomic masses of Pb and Tl isotopes, respectively(n = 206, 207, and 208). The certified 205Tl/203Tl ratio of2.3871 for NIST SRM 997 (Dunstan et al., 1980) wasused for the normalization.

A quadrupole ICP-MS ELAN-DRC II (Perkin Elmer,USA) was used for determining blanks, recovery yields,and elution curves of Pb and other elements in the courseof optimization of the chemical procedure.

Isotope standard materials and reagentsA high-purity Pb metal, NIST SRM 981 (Catanzaro et

al., 1968) was used as a Pb isotope standard. About 0.1 gof the Pb metal were weighed and dissolved in 1 M HNO3to prepare a 100 µg/mL stock solution that was subse-quently adjusted to 200 ng/mL with 0.15 M HNO3 forordinary use. A Tl solution of NIST SRM 997 (Dunstanet al., 1980) was also prepared in a similar manner, andwas added to the diluted Pb solution so as to contain 20ng/mL Tl.

All solutions were prepared with 18 MΩ water pro-duced by Milli-Q Element system (MILLIPORE, USA).The HNO3, HBr, and HF reagents used in this study werecommercially supplied high-purity reagents (Cica-Ultrapur (KANTO KAGAKU, Japan), TAMAPURE AA-10 or AA-100 (Tama Chemical, Japan)) or distillates pro-duced with two-bottle Teflon stills using the reagents ofanalytical-grade quality.

Rock reference samplesEleven GSJ “Igneous rock series” rock reference sam-

ples were analyzed in this study. They include basalt (JB-1a, JB-2, and JB-3), andesite (JA-1, JA-2, and JA-3),rhyolite (JR-1 and JR-2), and granite (JG-1a, JG-2, andJG-3). This series shows a wide range in major elementcomposition (SiO2 = 51 to 76 wt.%, MgO = 0.04 to 7.8wt.%, and K2O = 0.4 to 4.7 wt.%; Imai et al., 1995) andPb concentration (5.4 to 32 µg/g; Imai et al., 1995).

Extraction of Pb from rock samplesThe procedure of Pb separation from rock samples

used in this study is similar to that used in a recent TIMSanalysis (Kuritani and Nakamura, 2002). The rock pow-der containing from 200 to 400 ng Pb was weighed in a 7

mL PFA Teflon vial, and decomposed with 0.7 mL 20 MHF and 0.7 mL 8 M HBr at 130°C. After the dryness at140°C, the sample was dissolved with 1 mL 0.5 M HBrand centrifuged. The supernatant was loaded onto the 0.1mL anion exchange resin (Bio-Rad AG 1-X8, 200–400mesh, Br-form) packed in TFE Teflon column. The re-maining precipitate was rinsed with 0.5 mL 0.5 M HBrand subsequently centrifuged, and the supernatant wasthen loaded onto the same column. The Pb fraction wascollected with 1 mL H2O after the elution of other ele-ments using 2.5 mL of a mixed acid composed of 0.25 MHBr and 0.5 M HNO3. The use of the mixed acid ratherthan HBr alone enables to efficiently separate Pb from Tland other coexisting elements with similar affinities tothe resin (Strelow, 1978). The Pb fraction was evaporatedto dryness at 120°C, and the dried sample was dissolvedwith 0.15 M HNO3. Appropriate volume of the NIST SRM997 Tl solution (10 µg/mL) was added to obtain 200 ng/mL Pb and 20 ng/mL Tl in the final sample solution. Tominimize the Pb contamination from experimental envi-ronment, all chemical procedures were carried out in aclass-1000 clean room and class-100 clean benches. To-tal procedural blank of Pb was from 22 to 40 pg, whichwas negligible in our routine analysis.

RESULTS AND DISCUSSION

Evaluation and minimization of memory effect and ma-trix effect

In ICP-MS, samples can be introduced into the instru-ment in the form of a solution, which contrasts with TIMSthat uses a salt of the analyte evaporated onto the metalfilament. This sample introduction system of ICP-MSleads to a rapid and easy sample handling, while skilledsample loading techniques onto the filament and the sub-sequent re-evacuation of the ion-source chamber are nec-essary in TIMS. However, we must pay attention to thememory effect of previous sample solutions caused byadsorption or sticking of elements onto the surfaces ofthe nebulizer, spray chamber, torch, and interface. Vola-tile elements like Pb tend to show relatively large memoryeffects, and the complete elimination of the previous sam-ple signal is generally not easy. The memory effect of Pbin our introducing system was tested using the 200 ng/mL Pb solution of NIST SRM 981. A pair of the solutionswas prepared, one of which was the usual 0.15 M HNO3medium while the other contained only trace amount ofHNO3 (0.002 M). After introducing each Pb solution forthe half an hour, the introducing system was washed withpure 0.5 M HNO3. The washout profiles of 208Pb areshown in Fig. 1. It is evident that the residual Pb signalderived from the 0.15 M HNO3 solution is more quicklyremovable than that from the 0.002 M HNO3 solution,which indicates a stronger adsorption of Pb from the

124 M. Tanimizu and T. Ishikawa

quasi-H2O solution onto the surfaces of the sample intro-duction system. A similar experiment was carried out forZn (Tanimizu, in press), and its quicker washout profilethan Pb reveals the significantly large adsorbability ofPb. As the remaining 208Pb signal becomes almost con-stant at around 4 × 10–15 A after 5 minute washout (Fig.1), we adopted the washout time of 5 minutes with 0.5 MHNO3 in our routine Pb isotope analysis. The level of theresidual Pb memory is less than 0.01% of the signal in-tensity in our Pb isotope analysis with 200 ng/mL Pb(208Pb 6 × 10–11 A), and low enough to be negligible inour analysis (see later discussions). 15 minutes for dataacquisition plus 5 minutes for washout yield total ana-lytical time of 20 minutes, capable of measuring threesamples per an hour.

Another factor that must be taken into considerationin MC-ICP-MS is matrix effect. The matrix effect in ICP-MS can be classified into spectral interferences andnonspectral interferences. The spectral interferences, inwhich overlapping mass peaks are added to the analytesignal, originate from the elements in gas, air, water, andmedium. It is previously reported that the signal of Hgisotopes derived from plasma support gas may be ob-served (Rehkämper and Halliday, 1998). Since 204Hg over-laps with 204Pb, 202Hg signal was monitored for the cor-rection of this interference. In our analytical condition,the 202Hg signal was about 2 × 10–16 A, and the estimated204Hg/204Pb ratio was as low as less than 2 × 10–4. Otherinterferences caused by many polyatomic ions are pecu-liar to ICP-MS in addition to isobaric interferences, astypically exemplified by the ArO and ArN interferenceson Fe isotopes. Fortunately the Pb mass range is high

Fig. 1. Washout profile of NIST SRM 981 Pb solutions. Opencircles and crosses are data points from 200 ng/mL Pb solu-tions prepared with 0.15 M HNO3 and 0.002 M HNO3, respec-tively.

Fig. 2. Mass spectrums around mass number (a) 209 and (b)204 observed using a SEM detector with an RPQ energy filter.The influences of hydride species, 208PbH and 203TlH were es-timated from 31 × 10–11 A 208Pb and 35 × 10–11 A 203Tl (in Fara-day detector) using the pure NIST Pb and Tl solutions, respec-tively. The signals on 209 and 204, which may include traceimpurity signals of 209Bi, 204Pb, and 204Hg, are all less than 1 ×10–15 A, and the maximum PbH/Pb and TlH/Tl ratios are cal-culated to be less than 4 × 10–6. The tailings to the both peaksare less than 2 × 10–16 A, which are comparable to the speci-fied abundance sensitivity of 0.3 ppm with the RPQ filter.

enough to escape from the Ar-originated polyatomic in-terferences, and no significant background signal wasobserved. A possible polyatomic interference in this massrange is caused by hydride species of analytes, PbH andTlH: 206PbH, 207PbH, and 203TlH may affect the signals

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 125

of 207Pb, 208Pb, and 204Pb, respectively. The maximuminfluences of these polyatomic ions were estimated bymonitoring the signals at m/z = 209 for 208PbH and 204for 203TlH using the pure NIST Pb and Tl solutions, re-spectively. The obtained PbH/Pb and TlH/Tl ratios wereboth less than 4 × 10–6 (Fig. 2), and the maximum influ-ences on Pb isotopes are calculated to be 2 × 10–5 in ourPb (200 ng/mL) with Tl (20 ng/mL) solutions, negligiblein our Pb isotope analysis.

Nonspectral interference refers to matrix-inducedchange in signal intensity that is unrelated to the pres-

ence or absence of spectral components. The detail of thistype of interferences is summarized in a literature (Evanand Giglio, 1993). The matrix concerned here is predomi-nantly HNO3. Therefore the matrix-induced change wasexamined with various HNO3 concentrations. As shownin Fig. 3, gradual signal changes were observed both forPb and Tl against HNO3 concentration. We adapted 0.15M HNO3 in our routine Pb isotope analysis to maximizethe signal intensities of Pb and Tl.

Fig. 3. Matrix effect on Pb and Tl signal intensities caused byvarious HNO3 contents in medium.

Fig. 4. Results of repeated analyses of a 200 ng/mL NIST SRM981 Pb solution. Average values of the Tl-normalized Pb iso-tope ratios are: 206Pb/204Pb = 16.9308 ± 0.0010, 207Pb/204Pb =15.4839 ± 0.0011, and 208Pb/204Pb = 36.6743 ± 0.0030 (2σ).

Fig. 5. Pb isotope composition of NIST SRM 981 reported inliteratures compared with our data. The relative differences ob-tained by (a) TIMS and (b) MC-ICP-MS are shown withoutuncertainties. Their methods are listed in Table 2. Our value isconsistent with most of the previous MC-ICP-MS data with Tl-normalization within ±0.005%/amu, while large deviations(0.03%/amu) are observed from the accepted values (Thirlwall,2000, 2002; Baker et al., 2004). Note that the TIMS data seemto be deviated from the expected straight linear mass-depend-ent fractionation trends due to anomalous behavior of Pb aspointed out previously (Thirlwall, 2000).

126 M. Tanimizu and T. Ishikawa

Accuracy and reproducibility in analyses of NIST SRM981

The Pb isotope ratios of the 200 ng/mL NIST Pb solu-tion with 20 ng/mL Tl were repeatedly measured in tendays over a period of three weeks. The results are shownin Fig. 4. The MC-ICP-MS was optimized to maximizePb signal intensity in each day one hour after the plasmaignition, and no replacement or cleaning of the sampleintroduction system was carried out during this three weeksession. The resultant average values were 206Pb/204Pb =16.9308 ± 0.0010, 207Pb/204Pb = 15.4839 ± 0.0011, and208Pb/204Pb = 36.6743 ± 0.0030 (uncertainties in 2σ, n =51). The relative uncertainties are from ±0.0059 to±0.0078%, which shows an excellent reproducibility ofour analysis. This reproducibility is superior to the re-sults obtained by TIMS or MC-ICP-MS with Tl-normali-zation reported so far, and comparable to those of the lat-est MC-ICP-MS results with 207Pb-204Pb double-spike(Table 2).

There is a slight discrepancy that Pb isotope ratiosobtained in this study are systematically lower than therecently accepted values with 207Pb-204Pb double-spike(Thirlwall, 2000, 2002; Baker et al., 2004; see Table 2).The relative deviation makes a linear trend against massnumber with an average slope of about 0.03%/amu(Figs. 5a and 5b). On the other hand, our values are con-sistent with most of the previous MC-ICP-MS data withTl-normalization (Hirata, 1996; Belshaw et al., 1998;Rehkämper and Halliday, 1998) within ±0.005%/amu(Fig. 5b) except White et al., (2000). As pointed out pre-viously (Rehkämper and Halliday, 1998), the slight de-viation of the MC-ICP-MS data from the accepted datamay be due to a slight difference between fPb and fTl thatare assumed to be identical in Eq. (3). Recently anotherkind of correction is proposed empirically (Albarède etal., 2004), but in the present stage, mechanism of the massdiscrimination in ICP-MS is not sufficiently understood,and its better understanding will lead a more adequatetheoretical correction law in future studies.

Confirmation of matrix- and memory-independence of thePb isotope data

The signal intensity of Pb in MC-ICP-MS changeswith matrix composition, which we showed in the caseof HNO3 concentration (Fig. 3), but the matrix also af-fects the observed Pb isotope ratios. As shown in Fig. 6,the Pb isotope ratios obtained without Tl-normalization(open circles) vary with HNO3 concentration up to 0.1%/amu. However, the ratios after the Tl-normalization(closed circles) are independent of HNO3 concentration,which reveals that the matrix effect is efficiently elimi-nated by the Tl-normalization.

Figure 4 demonstrates the excellent performance ofMC-ICP-MS in reproducibility for the repeated analyses

Tabl

e 2.

P

b is

otop

e ra

tios

of

NIS

T S

RM

981

det

erm

ined

by

seve

ral

TIM

S an

d M

C-I

CP

-MS

met

hods

. U

ncer

tain

ties

are

2 s

tand

ard

devi

a-ti

ons

of t

heir

rep

eate

d an

alys

es. U

ncer

tain

ties

sho

wn

in i

tali

cs a

re t

he v

alue

s re

-cal

cula

ted

here

fro

m t

he d

iffe

rent

Pb

isot

ope

rati

os i

n th

esa

me

lite

ratu

res.

Ref

eren

ceM

etho

d20

6 Pb/

204 P

b2σ

207 P

b/20

4 Pb

2 σ20

8 Pb/

204 P

b2σ

TIM

SC

atan

zaro

et

al. (

1968

)ab

solu

te is

otop

ic d

eter

min

atio

n16

.937

0.01

115

.491

0.01

236

.721

0.02

6W

oodh

ead

et a

l. (1

995)

207 P

b-20

4 Pb

doub

le-s

pike

16.9

370.

004

15.4

920.

005

36.7

080.

013

Tod

t et a

l. (1

996)

202 P

b-20

5 Pb

doub

le-s

pike

16.9

356

0.00

2215

.489

10.

0028

36.7

006

0.01

08N

ohda

(19

99)

sam

ple-

stan

dard

bra

kett

ing

16.8

908

0.00

5815

.428

50.

0056

36.4

982

0.01

47T

hirl

wal

l (2

000)

207 P

b-20

4 Pb

doub

le-s

pike

16.9

408

0.00

2215

.495

50.

0026

36.7

225

0.00

80K

urit

ani

and

Nak

amur

a (2

002)

zero

-tim

e co

r rec

tion

16.8

846

0.00

1715

.421

50.

0024

36.4

779

0.00

78

MC

-IC

P-M

SH

ira t

a (1

996)

ma s

s-c o

rre c

ted

pow

er l

aw,

205 T

l/20

3 Tl

= 2

.387

116

.931

10.

0090

15.4

856

0.00

836

.680

0.02

10H

alli

day

(199

8)ex

pone

ntia

l la

w, 20

5 Tl/

203 T

l =

2.3

871

16.9

295

0.00

5515

.481

80.

0051

36.6

671

0.01

28B

elsh

aw e

t al.

(199

8)ex

pone

ntia

l la

w, 20

5 Tl/

203 T

l =

2.3

875

16.9

320.

0068

15.4

870.

006

36.6

830.

015

Whi

te e

t al.

(200

0)ex

pone

ntia

l la

w, 20

5 Tl/

203 T

l =

2.3

871

16.9

467

0.00

7615

.489

90.

0039

36.6

825

0.00

78T

hirl

wal

l (2

002)

207 P

b-20

4 Pb

doub

le-s

pike

16.9

417

0.00

2915

.499

60.

0031

36.7

240.

009

Bak

er e

t al

. (20

04)

207 P

b-20

4 Pb

doub

le-s

pike

16.9

416

0.00

1115

.499

90.

0011

36.7

258

0.00

31

Thi

s st

udy

expo

nent

ial

law

, 20

5 Tl/

203 T

l =

2.3

871

16.9

308

0.00

1015

.483

90.

0011

36.6

743

0.00

30

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 127

of the same sample. However, in the routine analysis ofnatural samples, the Pb isotope ratios to be measured aredifferent one by one, and inevitably the Pb isotope com-position of the remaining Pb signal memory after 5 minutewashout should vary significantly. If an analyzed samplehas an extreme Pb isotopic composition, the memory mayaffect the measured value of the next sample like the caseof zinc in double-spike analysis (Tanimizu et al., 2002),although we had shown that the level of the Pb memoryis less than 0.01% of the sample signal in usual analysis(Fig. 1). To evaluate the possible influence of the memoryeffect, we prepared a Pb solution with 206Pb/204Pb = 21,207Pb/204Pb = 16, and 208Pb/204Pb = 42 extracted from anarcheological bronze sample (unpublished results). ThisPb isotope composition resembles that of HIMU-typeocean island basalt, which have one of the highest Pb iso-tope ratios in the earth (206Pb/204Pb ≅ 21.0, 207Pb/204Pb ≅15.8, and 208Pb/204Pb ≅ 40.2; Zindler and Hart, 1986). Asthe Pb isotope composition of NIST SRM 981, on theother hand, is similar to that of igneous rocks with thelowest Pb isotope ratios, alternate analyses of these twocontrasting samples provide an extreme case study forevaluating the memory effect. In this experiment(Fig. 7), the NIST SRM 981 data analyzed just after theHIMU-like bronze sample (closed circles) were all indis-tinguishable from the other data (open circles). Thus it isevident that the memory effect is negligible for the analy-sis of almost all terrestrial rock samples in our methodwith the washout time of 5 minutes.

Pb isotope compositions of GSJ rock reference samplesPrior to the Pb isotope analysis of GSJ rock reference

Fig. 6. Matrix effect on observed 206Pb/204Pb ratios (open cir-cles) and Tl-normalized 206Pb/204Pb ratios (closed circles)caused by medium (HNO3). Note that the vertical scales arealmost the same, and their uncertainties are far smaller thanthe symbol sizes. It is evident that the matrix effect is efficientlyeliminated by the Tl-normalization.

Fig. 7. Data of NIST SRM 981 alternately analyzed with a so-lution which has a HIMU-like extreme Pb isotope composition.Closed and open circles indicate the NIST 981 data taken justafter the analysis of the HIMU-like sample and the data of theothers, respectively. These two kinds of data are indistinguish-able each other, indicating that the memory effect is negligi-ble.

samples, the Pb fraction chemically separated from JB-2(Pb = 5.36 µg/g, Hg = 4.78 ng/g, Tl = 42 ng/g, and Zn =108 µg/g; Imai et al., 1995) was examined by using thequadrupole ICP-MS to confirm the adequacy of the Pbseparation. As reported previously (Strelow, 1978), theuse of the 0.25 M HBr-0.5 M HNO3 mixed acid as aneluant was quite effective to remove Hg, Tl, and otherelements with strong affinity to the anion-exchange resinin Br- form, for example, Zn. When 0.5 M HBr alone wasused as an eluant instead of the mixed acid, the observed202Hg/208Pb, 205Tl/208Pb, and 66Zn/208Pb count ratios forthe obtained Pb fraction were 1 × 10–6, 2 × 10–4 and 1,respectively, which shows incomplete removal of theseelements. On the other hand, the 66Zn/208Pb ratio observedin our usual procedure was as low as 7 × 10–4, and the202Hg and 205Tl signals were not detected. The recoveryyield of Pb was >99%.

The Pb isotope ratios of eleven GSJ rock referencesamples determined in this study are summarized in Ta-ble 3. These data were obtained in the same period of theNIST 981 analysis. To evaluate the reproducibility in themeasurement of rock samples, seven independent analy-ses were made for JB-2. The values obtained were 206Pb/204Pb = 18.3315 ± 0.0025, 207Pb/204Pb = 15.5460 ± 0.0021,and 208Pb/204Pb = 38.2240 ± 0.0055 (uncertainties in 2σ,n = 7). Although the observed relative uncertainties of±0.014% are slightly larger than those for NIST SRM 981(from ±0.0059 to ±0.0078%), this reproducibility is com-parable to the latest double-spike data (Thirlwall, 2000,2002; Baker et al., 2004). Furthermore, it is worth noting

128 M. Tanimizu and T. Ishikawa

that in the Pb solution of JB-2 #5, Al and Mg were in-tently added to give the concentration levels similar toPb. The 206Pb/204Pb ratio of the run #5 without Tl-nor-malization deviates from those of the previous data by0.1%, and this matrix effect remains in the next run #6(Fig. 8a). However, after Tl-normalization, the Pb iso-tope ratios of the runs #5 and #6 are consistent with theresults of the other runs (Fig. 8b). This clearly demon-strates that the matrix effect caused by small amount ofimpurity elements due to incomplete column separationis efficiently eliminated by Tl-normalization as in the caseof the matrix effect associated with medium (Fig. 6).

The Pb isotope ratios of GSJ rock reference samplesreported by several authors are summarized in Appendix,in which all the data including ours are normalized basedon the deviation of the reported NIST SRM 981 values

from the latest value with double-spike (Baker et al.,2004). The JB-2 data are extracted and compared withprevious results (Koide and Nakamura, 1990; Matsumotoet al., 1993; Ishikawa and Nakamura, 1994; Taylor andNesbitt, 1998; Nohda, 1999; Ishizuka et al., 2003; Bakeret al., 2004) in Fig. 9. All JB-2 data are almost consistentwithin ±0.1% relative deviations; 206Pb/204Pb =18.33~18.35, 207Pb/204Pb = 15.53~15.57, and 208Pb/204Pb= 38.20~38.30.

Our Pb isotope data of the GSJ rocks are comprehen-sively plotted against the previous TIMS data (Koide andNakamura, 1990; Matsumoto et al., 1993; Nohda, 1999)in Fig. 10. Most of the data in Fig. 10 gather around theideal 1:1 lines within ±0.3% relative deviations, showingthe consistency between our data and the TIMS data. Al-though the reasons of the three times larger deviation of

weight/mg 206Pb/204Pb 2SE 207Pb/204Pb 2SE 208Pb/204Pb 2SE

JB-1a (6.76) 64.0 18.3713 0.0009 15.5504 0.0009 38.6601 0.0030

JB-2 (5.36)#1 76.7 18.3325 0.0007 15.5472 0.0007 38.2269 0.0020#2 69.0 18.3319 0.0010 15.5464 0.0011 38.2246 0.0034#3 126.6 18.3321 0.0008 15.5464 0.0007 38.2247 0.0020#4 65.6 18.3324 0.0009 15.5467 0.0009 38.2261 0.0026#5 100.3 18.3300 0.0010 15.5449 0.0010 38.2204 0.0032#6 60.3 18.3311 0.0009 15.5456 0.0010 38.2222 0.0030#7 39.5 18.3303 0.0010 15.5451 0.0009 38.2232 0.0025average 18.3315 15.5460 38.22402SD 0.0025 0.0021 0.0055

JB-3 (5.58) 96.7 18.2839 0.0011 15.5228 0.0011 38.1996 0.0034

JA-1 (6.55)#1 70.1 18.3019 0.0009 15.5343 0.0010 38.2381 0.0034#2 31.8 18.3021 0.0011 15.5334 0.0009 38.2366 0.0025average 18.3020 15.5338 38.2374

JA-2 (19.2) 30.8 18.3898 0.0009 15.5927 0.0009 38.6202 0.0030

JA-3 (7.7) 66.4 18.3171 0.0008 15.5539 0.0009 38.3736 0.0028

JR-1 (19.3) 21.3 18.3493 0.0009 15.5490 0.0009 38.3662 0.0028

JR-2 (21.5) 22.7 18.3454 0.0008 15.5482 0.0009 38.3605 0.0028

JG-1a (26.4) 14.6 18.6057 0.0008 15.6102 0.0008 38.6874 0.0028

JG-2 (31.5) 17.9 18.6050 0.0007 15.6345 0.0008 38.9829 0.0024

JG-3 (11.7)#1 34.1 18.3531 0.0009 15.5641 0.0010 38.4480 0.0030#2 37.0 18.3529 0.0009 15.5640 0.0010 38.4457 0.0030average 18.3530 15.5640 38.4469

Table 3. Pb isotope ratios of eleven GSJ rock reference samples obtained in this study. Pbconcentrations (Imai et al., 1995) are described in parentheses after the sample names inµg/g.

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 129

Fig. 8. 206Pb/204Pb ratios of JB-2 (a) without Tl-normalizationand (b) with Tl-normalization. Uncertainties are statistical twostandard error of one analysis. Arrows on the right side indi-cate ±0.01% relative deviation. Although the runs #5 and #6suffer from the matrix effect caused by impurity, their Tl-nor-malized ratios are consistent with the other data by the effec-tive elimination of the effect. The run #7 data were obtainedafter the half a year from the other runs.

Fig. 9. Comparison of Pb isotope ratios of JB-2 obtained inthe previous studies and this study. Open circles, half-filled cir-cles, and a closed circle are obtained by TIMS, double-spikeMC-ICP-MS, and this study (Tl-normalization), respectively.Error bars indicate uncertainty in each analysis expressed in2σ.

the Pb isotope ratios compared to JB-2 alone and a largerdeviation of JA-3 are not clear, one possible explanationmay be given by isotopic heterogeneities of acidic rocksamples in mg sample size. Alternatively, this deviationmay be due to the matrix effect caused by impurities suchas Cd, Zn, and Tl on the filament as pointed out in a pre-vious TIMS analysis (Kuritani and Nakamura, 2002).

The effectiveness of Tl-normalization to actual rocksamples and comparable precision to the double-spikeanalyses achieved here were proven by Fig. 9 in contrast

130 M. Tanimizu and T. Ishikawa

to the previous opinions that Tl-normalization is not suf-ficient to correct the mass discrimination (Thirlwall, 2002;Baker et al., 2004). Theoretically Pb double-spike tech-nique is superior to Tl-normalization in excluding thedifference in f values (Eq. (3)) as discussed above, thoughisotopically enriched spikes and separate calculation ofPb isotope ratios are necessary. In this respect, Tl-nor-malization does not need care of the Pb spikes and therelated isotopically variable memory effect, and data re-duction can be treated on time. These are advantages ofTl-normalization compared to the Pb double-spikemethod. The cause of apparent inaccuracy of the Tl-nor-malization method relative to the Pb double-spike methodargued previously (Baker et al., 2004) is still unknown,and replicate analyses of a well-characterized standardrock sample in addition to NIST SRM 981 are recom-mended in current situation.

Comments on applications of MC-ICP-MS for precise andaccurate isotope analysis

Isotope analysis using MC-ICP-MS always needs thegreatest care to the memory and matrix effects and thepolyatomic interferences. In addition, the elements whichcause isobaric interferences to analyte must be strictlyremoved from the sample. Unlike TIMS, ICP ionizes al-most all elements efficiently, and the isotope ratios of theionized elements are significantly fractionated from theoriginal values (e.g., IUPAC values; Rosman and Taylor,1998) through the interface due to more than one orderof magnitude larger mass discrimination effect comparedto TIMS. This means that only small amount of impurityremained in samples may cause serious isobaric interfer-ence that is difficult to correct accurately. Furthermore,if the correction for the mass discrimination with anotherelement in the same mass range is applied to MC-ICP-MS, it is essential that the standardization element is oncecompletely removed from the sample before the additionof the isotope reference material. This is because the re-sidual standardization element might show extremelyfractionated isotope ratios due to mass-dependentfractionation during natural processes or chemical sepa-ration especially that with ion-exchange resin (Maréchaland Albarède, 2002). Of course the analyte must be freefrom this mass-dependent fractionation during chemicalpurification, and complete recovery (>95%) of the analyteis required from the sample. Finally, a high-purity iso-tope reference material whose isotopic abundances arewell known must be selected as the standardization ma-terial. The uncertainty of the isotopic abundances and theimpurity of the material directly affect the accuracy ofthe corrected isotope ratios of the analyte. From thesereasons, we paid a great deal of attention to achieve com-plete recovery of Pb and complete removal of Hg and Tlin the chemical procedure as well as to minimize the

Fig. 10. Comparison of Pb isotope ratios of GSJ rock refer-ence samples obtained by previous TIMS analyses and our MC-ICP-MS analysis. Typical uncertainty in the TIMS analyses isalmost identical to the symbol sizes. Most data gather aroundthe ideal 1:1 lines (solid lines) within ±0.3% relative devia-tions (broken lines).

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 131

matrix effect and memory effects, and chose NIST SRM997 for the standardization material in our Pb isotopeanalysis. As the conditions of background spectrum andpolyatomic interferences are dependent of each instru-ment and analytical setting, their inspection in each ana-lytical method is required. We stress that the satisfactionof the above factors is essential for the precise and accu-rate isotope analysis using MC-ICP-MS.

In this paper, we showed considerable influences ofthe matrix effects on the Pb isotope ratios obtained with-out Tl-normalization. This type of matrix effect directlyaffects the reproducibility in isotope analyses of severalelements such as Li, Ca, and Fe, which do not have ad-equate standardization elements, as pointed out by sev-eral authors (Bryant et al., 2003; Fietzke et al., 2004;Malinovsky et al., 2003). Sample-standard bracketingcorrection has been applied for these elements, but theanalysis must be carried out under the strict matrix con-trol. We must keep in mind the fact that the isotope analy-ses of these elements with MC-ICP-MS suffer from thesame problem as in the TIMS analyses on this point.

Acknowledgments—We are grateful to Y. Hirao, Beppu Univ.for providing Pb solutions of archeological bronzes and forhelpful discussion. The paper benefited from helpful and con-structive reviews by J. Baker and an anonymous reviewer. Wewould also like to thank N. Ohkouchi and H. Kitazato for theirencouragement and support regarding this work.

REFERENCES

Albarède, F., Telouk, P., Blichert-Toft, J., Boyet, M., Agranier,A. and Nelson, B. (2004) Precise and accurate isotopicmeasurements using multiple-collector ICPMS. Geochim.Cosmochim. Acta 68, 2725–2744.

Baker, J., Peate, D., Waight, T. and Meyzen, C. (2004) Pb iso-topic analysis of standards and samples using a 207Pb-204Pbdouble spike and thallium to correct for mass bias with adouble-focusing MC-ICP-MS. Chem. Geol. 211, 275–303.

Belshaw, N. S., Freedman, P. A., O’Nions, R. K., Frank, M.and Guo, Y. (1998) A new variable dispersion double-fo-cusing plasma mass spectrometer with performance illus-trated for Pb isotopes. Int. J. Mass Spectrom. 181, 51–58.

Bryant, C. J., McCulloch, M. T. and Bennett, V. C. (2003) Im-pact of matrix effects on the accurate measurement of Liisotope ratios by inductively coupled plasma massspectrometry (MC-ICP-MS) under “cold” plasma condi-tions. J. Anal. At. Spectrom. 18, 734–737.

Cameron, A. E., Smith, D. H. and Walker, R. L. (1969) Massspectrometry of nanogram-size samples of lead. Anal. Chem.41, 525–526.

Catanzaro, E. J., Murphy, T. J., Shields, W. R. and Garner, E.L. (1968) Absolute isotopic abundance ratios of common,equal-atom, and radiogenic lead isotopic standards. J. Res.Nat. Bur. Stand. 72A, 261–267.

Dunstan, L. P., Gramlich, J. W. and Barnes, I. L. (1980) Abso-lute isotopic abundance and the atomic weight of a refer-

ence sample of thallium. J. Res. Nat. Bur. Stand. 85, 1–10.Evan, E. H. and Giglio, J. J. (1993) Interferences in inductively

coupled plasma mass spectrometry. J. Anal. At. Spectrom.8, 1–18.

Fietzke, J., Eisenhauer, A., Gussone, N., Bock, B., Liebetrau,V., Nägler, Th. F., Spero, H. J., Bijima, J. and Dullo, C.(2004) Direct measurement of 44Ca/40Ca ratios by MC-ICP-MS using the cool plasma technique. Chem. Geol. 206, 11–20.

Gillson, G. R., Douglas, D. J., Fluford, J. E., Halligan, K. W.and Tanner, S. D. (1988) Nonspectroscopic interelementinterferences in inductively coupled plasma massspectrometry. Anal. Chem. 60, 1472–1474.

Hirata, T. (1996) Lead isotopic analyses of NIST standard ref-erence materials using multiple collector inductively cou-pled plasma mass spectrometry coupled with a modifiedexternal correction method for mass discrimination effect.Analyst 121, 1407–1411.

Imai, N., Terashima, S., Itoh, S. and Ando, A. (1995) 1994 com-pilation of analytical data for minor and trace elements inseventeen GSJ geochemical reference samples, “Igneousrock series”, Geostand. Newslett., 19, 135–213.

Ishikawa, T. and Nakamura, E. (1994) Origin of the slab com-ponent in arc lavas from across-arc variation of B and Pbisotopes. Nature 370, 205–208.

Ishizuka, O., Taylor, R. N., Milton, J. A. and Nesbitt, R. W.(2003) Fluid-mantle interaction in an intra-oceanic arc: con-straints from high-precision Pb isotopes. Earth. Planet. Sci.Lett. 211, 221–236.

Johnson, C. M., Beard, B. L. and Albarède, F. (eds.) (2004)Geochemistry of non-traditional stable isotopes. Reviewsin Mineralogy & Geochemistry Vol. 55, The MineralogicalSociety of America.

Koide, Y. and Nakamura, E. (1990) Lead isotope analyses ofstandard rock samples. Mass Spectrom. 38, 241–252.

Kuritani, T. and Nakamura, E. (2002) Precise isotope analysisof nanogram-level Pb for natural rock samples without useof double spikes. Chem. Geol. 186, 31–43.

Malinovsky, D., Stenberg, A., Rodushkin, I., Andren, H., Ingri,J. Öhlander, B. and Baxter, D. C. (2003) Performance ofhigh resolution MC-ICP-MS for Fe isotope ratio measure-ments in sedimentary geologic materials. J. Anal. At.Spectrom. 18, 687–695.

Maréchal, C. and Albarède, F. (2002) Ion-exchange fractionationof copper and zinc isotopes. Geochim. Cosmochim. Acta66, 1499–1509.

Matsumoto, A, Hirao, Y. and Togashi, S. (1993) Determinationof lead isotope ratios of GSJ rock reference samples. Bull.Geol. Surv. Japan 44, 649–657.

Nohda, S. (1999) Precise isotopic measurements of lead byfused silica-gel. Geochem. J. 33, 133–139.

Ostic, R. G., Russell, R. D. and Stanton, R. L. (1967) Addi-tional measurements of the isotopic composition of leadfrom stratiform deposits. Can. J. Earth. Sci. 4, 245–269.

Rehkämper, M. and Halliday, A. N. (1998) Accuracy and long-term reproducibility of lead isotopic measurements by mul-t iple-collector inductively coupled plasma massspectrometry using an external method for correction ofmass discrimination. Int. J. Mass Spectrom. 181, 123–133.

132 M. Tanimizu and T. Ishikawa

Rosman, K. J. R. and Taylor, P. D. P. (1998) Isotopic composi-tions of the elements 1997. Pure Appl. Chem. 70, 217–235.

Strelow, F. W. E. (1978) Distribution coefficients and anionexchange behavior of some elements in hydrobromic-nitricacid mixtures. Anal. Chem. 50, 1359–1361.

Tanimizu, M. (2006) An evaluation of a new multiple collectorinductively coupled plasma mass spectrometer and its po-tential performance. Frontier Research on Earth Evolution2003–2004 (in press).

Tanimizu, M., Asada, Y. and Hirata, T. (2002) Absolute iso-topic composition and atomic weight of commercial zincusing inductively coupled plasma mass spectrometry. Anal.Chem. 74, 5814–5819.

Taylor, R. N. and Nesbitt, R. W. (1998) Isotopic characteristicsof subduction fluids in an intra-oceanic setting, Izu-BoninArc, Japan. Earth. Planet. Sci. Lett. 164, 79–98.

Thirlwall, M. F. (2000) Inter-laboratory and other errors in Pbisotope analyses investigated using 1 207Pb-204Pb doublespike. Chem. Geol. 163, 299–322.

Thirlwall, M. F. (2002) Multicollector ICP-MS analysis of Pbisotopes using a 2 207Pb-204Pb double spike demonstratesup to 400 ppm/amu systematic errors in Tl-normalization.

Appendix. Pb isotope ratios of eleven GSJ rock reference samples with JB-1and JG-1 reported so far. All data are normalized to the latest double-spikevalues of NIST SRM 981 (Baker et al., 2004) by adjusting their reported val-ues of NIST SRM 981 to the reference values as relative fractionation factors.

Chem. Geol. 184, 255–279.Todt, W., Cliff, R. A., Hanser, A. and Hofmann, A. W. (1996)

Evaluation of 202Pb-205Pb double spike for high-precisionlead isotope analysis. Earth Processes: Reading the IsotopicCode, Geophysical Monograph 95.

Tuttas, D. and Habfast, K. (1982) High precision lead isotoperatio measurements. Finnigan MAT Appl. Rep. 51.

Weyer, S. and Schwieters, J. B. (2003) High precision Fe iso-tope measurements with high mass resolution MC-ICPMS.Int. J. Mass Spectrom. 226, 355–368.

White, W. M., Albarede, F. and Telouk, P. (2000) High-preci-sion analysis of Pb isotope ratios by multi-collector ICP-MS. Chem. Geol. 167, 257–270.

Woodhead, J. D., Volker, F. and McCulloch, M. T. (1995) Rou-tine lead isotope determination using a lead-207-lead-204Double spike: a long-term assessment of analytical preci-sion and accuracy. Analyst 120, 35–39.

Zindler, A. and Hart, S. (1986) Chemical geodynamics. Ann.Rev. Earth Planet. Sci. 14, 493–571.

APPENDIX

206Pb/204Pb 207Pb/204Pb 208Pb/204Pb Reference

JB-1 18.3594 15.5775 38.6723 Nohda (1999)

JB-1a 18.3620 15.5397 38.6252 Matsumoto et al. (1993)18.3832 15.5666 38.7151 This study

JB-2 18.3401 15.5567 38.2204 Koide and Nakamura (1990)18.3331 15.5557 38.2644 Koide and Nakamura (1990)18.3220 15.5367 38.1981 Matsumoto et al. (1993)18.3180 15.5367 38.2311 Matsumoto et al. (1993)18.3391 15.5597 38.2504 Ishikawa and Nakamura (1994)18.4154 15.5525 38.3297 Taylor and Nesbitt (1998)18.3394 15.5713 38.2932 Nohda (1999)18.3366 15.5617 38.2725 Ishizuka et al. (2003)18.3435 15.5619 38.2784 Baker et al. (2004)18.3436 15.5624 38.2786 This study

JB-3 18.2931 15.5387 38.2224 Koide and Nakamura (1990)18.2911 15.5317 38.2044 Koide and Nakamura (1990)18.2650 15.5207 38.1641 Matsumoto et al. (1993)18.3014 15.5409 38.2682 Nohda (1999)18.2958 15.5389 38.2540 This study

Rapid and precise Pb isotope analysis of GSJ reference rocks by MC-ICP-MS 133

Appendix. (continued)

206Pb/204Pb 207Pb/204Pb 208Pb/204Pb Reference

JA-1 18.3001 15.5387 38.2364 Koide and Nakamura (1990)18.3131 15.5567 38.2614 Koide and Nakamura (1990)18.3121 15.5437 38.2444 Koide and Nakamura (1990)18.3110 15.5507 38.2912 Matsumoto et al. (1993)18.3090 15.5437 38.3442 Matsumoto et al. (1993)18.3187 15.5604 38.3295 Nohda (1999)18.3138 15.5504 38.2925 This study

JA-2 18.4301 15.6457 38.7625 Koide and Nakamura (1990)18.4161 15.6527 38.7825 Koide and Nakamura (1990)18.4040 15.6137 38.6782 Matsumoto et al. (1993)18.4020 15.6117 38.6582 Matsumoto et al. (1993)18.4017 15.6089 38.6751 This study

JA-3 18.2801 15.5247 38.2624 Koide and Nakamura (1990)18.2801 15.4896 38.1764 Koide and Nakamura (1990)18.2860 15.5137 38.2451 Matsumoto et al. (1993)18.2890 15.5207 38.1981 Matsumoto et al. (1993)18.3290 15.5701 38.4282 This study

JR-1 18.3771 15.5837 38.4614 Koide and Nakamura (1990)18.3740 15.5807 38.4682 Matsumoto et al. (1993)18.3620 15.5667 38.4292 Matsumoto et al. (1993)18.3612 15.5652 38.4208 This study

JR-2 18.3541 15.5617 38.3854 Koide and Nakamura (1990)18.3581 15.5757 38.4164 Koide and Nakamura (1990)18.3520 15.5547 38.3882 Matsumoto et al. (1993)18.3600 15.5647 38.4142 Matsumoto et al. (1993)18.3573 15.5644 38.4151 This study

JG-1 18.5662 15.5787 38.5364 Koide and Nakamura (1990)18.5371 15.5947 38.5894 Koide and Nakamura (1990)

JG-1a 18.5671 15.5957 38.6542 Matsumoto et al. (1993)18.5771 15.6047 38.6852 Matsumoto et al. (1993)18.5871 15.6147 38.7192 Matsumoto et al. (1993)18.6177 15.6265 38.7425 This study

JG-2 18.6232 15.6617 39.0645 Koide and Nakamura (1990)18.6162 15.6527 39.0315 Koide and Nakamura (1990)18.5891 15.6377 39.0023 Matsumoto et al. (1993)18.6031 15.6277 38.9803 Matsumoto et al. (1993)18.6171 15.6508 39.0384 This study

JG-3 18.3931 15.6097 38.5974 Koide and Nakamura (1990)18.3470 15.5557 38.4182 Matsumoto et al. (1993)18.3490 15.5647 38.4402 Matsumoto et al. (1993)18.3649 15.5802 38.5016 This study